Modelling Faecal Coliform Dynamics in Streams and ...

Modelling Faecal Coliform Dynamics in Streams and Rivers by Jeremy Wilkinson, B.Sc. (hons).

This thesis is submitted in fulfilment of the requirements for the degree of Doctor of Philosophy at the University of Wales, Aberystwyth, September 2000.

Modelling Faecal Coliform Dynamics

Declaration This work has not previously been accepted in substance for any degree and is not being concurrently submitted in candidature for any degree.

Signed

(Richard Jeremy Wilkinson)

Date

Statement 1 This thesis is the result of my own investigations, except where otherwise stated. Other sources are acknowledged by footnotes giving explicit references. A bibliography is appended. Signed


Date

Statement 2 I hereby give consent for my thesis, if accepted, to be available for photocopying and for inter-library loan, and for the title and summary to be made available to outside organisations. Signed


Date

ii


Abstract This thesis reports work undertaken to improve modelling of faecal coliform dynamics in river systems. A broad literature review introduces the many environmental processes that influence faecal coliform concentrations in rivers and highlights approaches to modelling. Channel flushing and resuspension of organisms from the riverbed can significantly lower the quality of the overlying water. A series of experiments investigated the impact of reservoir releases on the entrainment of organisms from storage within the channel. The results suggest three phases of transport: wavefront transport by near bed steep-fronted waves, wavefront entrainment with transport at the slower mean velocity, and burst-sweep cycle entrainment maintaining elevated concentrations during continued high flow. A model is presented and demonstrated to reproduce the timing and concentration of the entrainment episodes to a high degree of accuracy. The model has the potential for application to the transient movement of other particle-associated contaminants. Sunlight is the major influence on faecal coliform die-off. Direct measurements of solar irradiance are commonly not available a method is presented for estimating irradiance from daily sunshine hours. The estimated irradiance is incorporated into water quality models that effectively reproduce the variations in faecal coliform concentration over a seven-year period. Measurements of irradiance or sunlight hours should be one of the prime considerations for any monitoring programme concerned with bacterial concentrations in environmental waters.

iii


Acknowledgements The project that formed the basis of this work was conducted while employed by the Institute of Hydrology, whose support is very gratefully acknowledged. The work was funded by the Department of the Environment (PECD Reference No. 7/7/385) with a contribution from the National Rivers Authority (NRA). The NRA (later the Environment Agency) made valuable contributions by provision of water quality and hydrometeorological data. Welsh Water plc. and North West Water provided water quality data for their monitored raw water intakes. Simon Durbin at Cwm Rheidol Power Station kindly agreed to carry-out our reservoir release plans, as did Peter Joyce at Yorkshire Waters Blubberhouses Depot on the Washburn and Tim Harrison of Severn Trent Region Environment Agency who operate the Clywedog Dam. Peter Cunnington at Ness Gardens provided paper records of meteorology data for hand copying, and Helen Stafford at Centre for Coastal and Marine Sciences, Bidston provided raw data from the Bidston meteorological site. Thanks to the staff of Acer Environmental who undertook the microbiological enumerations for the field experiments.

Thanks to David Kay for his kindness, understanding and belief in my work over the years. Thanks also to Mark Wyer for his sense of fun, it was a great pleasure working with him in the field. Thanks also to my colleagues at the Institute of Hydrology; Danny Butterfield, whose support in the field was excellent. Alan Jenkins was also very supportive. Thanks also to Paula Hopkins and other members of the CREH team. Finally I received a subtle hint from Keith Beven, which was repeated by Pam Naden. This was quite helpful, once I realised what they were getting-at, thank you both.

Thanks to those of you who didn't believe I'd ever write-up, your doubt helped to spur me on (perhaps as intended). Special thanks to friends and family whose support, encouragement and editorial comments have been invaluable.

iv


Contents Declaration Abstract Acknowledgement Contents List of Tables List of Figures

ii iii iv v viii x

Chapter 1 - Introduction 1 Chapter 2 - Literature Review 2 2.1 Faecal coliform die-off dynamics 2 2.1.1 Quantifying faecal coliform die-off 3 Quantifying faecal coliform die-off 4 Total coliform, faecal coliform and E.coli counts 6 Comparison of data 7 2.1.2 Physical influences on faecal coliform die-off 8 Seasonal patterns 8 The effect of sunlight 10 Temperature 12 Soil moisture and survival on the catchment surface 14 2.1.3 Physicochemical and nutrient effects 16 pH and faecal coliform die-off 16 Dissolved oxygen content 18 Particulates, settlement and other effects 19 Attachment to particles and surfaces 20 Particles, nutrients and FC relationships in natural waters 21 Settlement and other survival enhancing effects 24 Survival on the river-bed 25 2.1.4 Microbial Interactions 28 Microbial predators and competitors of faecal coliforms 28 Removal experiments 29 The influence of predation on faecal coliform die-off 30 2.1.5 Summary of die-off effects 34 2.2 Sources and Transport of Faecal Coliforms 36 2.2.1 Sources of faecal contamination 36 Typical sources of faecal contamination and their concentrations 37 2.2.2 Faecal contamination from diffuse sources 38 2.2.3 Transport into the river system 39 Mechanisms of transport through the soil system and run-off processes 40 Field observations of transport processes 42 2.2.4 Fluvial dynamics 45 Faecal coliform resuspension 46 2.3 Modelling faecal coliform die-off and transport 50 2.3.1 Bacterial die-off models 50 2.3.2 River models for faecal coliform dynamics 53 The advection dispersion equation (ADE) 53 A simple river application 55 Two-dimensional steady-state sediment transport model for rivers 56 Estuarine transport modelling with the ADE 57 Water Quality Modelling using the Active Mixing Volume (AMV) theory 60 v

Modelling Faecal Coliform Dynamics The QUASAR model A process-based model of bacterial dynamics in upland streams

62 64

Chapter 3 - Methodology and Study Rationale 3.1 The modelling problem 3.2 The provision of data 3.3 The field experiments 3.3.1 The field sites 3.3.2 Sampling and analysis techniques 3.3.3 Experimental releases 3.4 The River Dee: Huntington Intake 3.5 Methodologies for model development 3.4.1 Development of the fluvial dynamics model 3.4.2 Die-off and longer term modelling

69 69 70 73 73 76 77 79 82 82 83

Chapter 4 - Hydrology 4.1 Open channel flow and wave propagation 4.1.1 Classification of flows 4.1.2 Kinematic and dynamic (gravity) waves 4.1.3 Steep fronted waves 4.2 Afon Rheidol 4.2.1 Discharge 4.2.2 Flow routing 4.3 Afon Clywedog discharge 4.4 River Dee, depths, travel-times and local runoff 4.4.1 Water depth Huntington Intake 4.4.2 Travel times, velocity and mixing parameters 4.4.3 A rainfall runoff model for catchment delivery

84 84 85 86 87 89 89 92 96 99 100 104 105

Chapter 5 - Faecal Coliform Fluvial Dynamics 5.1 Results of the reservoir release experiments 5.1.1 River Washburn 5.1.2 Afon Clywedog 5.1.3 Afon Rheidol I 5.1.4 Afon Rheidol II 5.1.5 Summary of responses 5.1.6 Proposed storage / entrainment process 5.1.7 Supporting observations from related fields 5.2 The bacterial entrainment model 5.2.1 The new model 5.3 Modelling the reservoir releases 5.3.1 River Washburn and Afon Clywedog 5.3.2 Model application to the Afon Rheidol 5.3.2.1 Example modelling results 5.3.2.2 The effect of successive hydrographs 5.4 Summary

110 110 110 114 119 123 127 128 130 135 136 142 143 149 150 152 157

Chapter 6 - Modelling results 6.1 Die-off relationships 6.1.1 The die-off equations 6.1.2 A comparison of the die-off inducing influences

158 158 158 165 vi

Modelling Faecal Coliform Dynamics 6.1.3 Summary of die-off effects 6.2 Seasonal and diurnal variations in die-off 6.2.1 Extending the environmental data record 6.2.1.1 Irradiance Clear sky irradiance Cloud influenced irradiance 6.2.1.2 Extending the water temperature record 6.2.1.3 Interpolating the turbidity record 6.2.2 Variation in daily die-off over seasonal cycles 6.2.2.1 Depth averaged irradiance 6.2.2.2 Settlement: the apparent die-off 6.2.2.3 The seasonally variable die-off function 6.2.3 Diurnal variations in die-off 6.2.3.1 Converting to an hourly time interval 6.2.3.2 Estimates of hourly die-off 6.3 Modelling daily and seasonal changes 6.3.1 An alternative measure of die-off, X90 6.3.2 Water quality model applications 6.3.2.1 The full four input model 6.3.2.2 A reduced model with constant input 6.3.2.3 A two input model for daily data 6.3.2.4 A three input model without mixing zones 6.3.3 Summary

166 167 167 168 168 171 172 174 175 176 177 177 182 182 183 188 189 191 191 195 196 199 200

Chapter 7 - Conclusions and recommendations 7.1 The entrainment model 7.1.1 Model application 7.1.2 Enhancements to the field experiments 7.1.2.1 Precision of stage records 7.1.2.2 Need for direct discharge measurements 7.1.2.3 Continuous conductivity, turbidity, temperature and suspended sediment samples 7.1.2.4 Repeat experiments 7.1.3 Bed-store-re-charge and die-off 7.2 Catchment delivery 7.3 Real-time die-off 7.4 Model use 7.5 Towards a bacterial dynamics “toolbox” 7.6 Simple guidelines for recreational water users

202 202 204 204 204 205 205

References

211

Appendix I: Appendix II: Appendix III: Appendix IV:

Freshwater die-off rates Die-off in the soil system Transport in the sub-surface zone Data sets collated during this study

206 206 207 208 209 209 210

225 228 232 233

vii


List of Tables Table 2.1 Summary of die-off and transport related factors resulting in high winter faecal coliform concentrations in lowland streams and rivers (reverse each factor to give summer effects). Table 2.2 pH range over which faecal coliform die-off in water is at a minimum (pHkmin). Table 2.3 E.coli die-off rate at different nutrient concentrations. The tests used filter sterilised water from the River Coquet dosed with different proportions of sterile sewage effluent (after Evison, 1989). Table 2.4 Faecal coliform die-off in different sediments (recalculated from data of Burton et al., 1987; *Sherer et al., 1992). Table 2.5 Hostile organisms that contribute to faecal coliform die-off. Table 2.6 Organisms removed from natural water samples by successive filtration with decreasing filter pore sizes. Table 2.7 Die-off rate k (d-1) (and percentage reduction in die-off rate with each removal) in clean and sewage contaminated river water following serial removal of natural biota by filtration, temperature = 15ºC, (source data, Flint, 1987). Table 2.8 Summary of the main causative effects of environmental variables and factors on the survival of faecal coliforms. Table 2.9 Sources of faecal contamination, with typical faecal coliform concentrations. Table 4.1 A comparison of the most important features of gravity waves and kinematic waves in natural channels (after Dingman (1984)). Table 4.2 Rating of discharge Qi in cumecs at the Aberffrwd Dam, from power generated, P (kW), by the Cwm Rheidol Power Station Generator Set No.4. Table 4.3 Observed and estimated flow characteristics for the Afon Rheidol artificial hydrograph and bacterial peaks. Table 4.4 Rating parameters for the compound Crump weir at Cribynau, the second sampling site on Afon Clywedog. Table 4.5 Width, depth and channel cross-section area just upstream of Farndon on the River Dee (based on data presented in Gurnell (1997)). Table 4.6 Discharge, travel time and mean flow velocity in the River Dee. Table 4.7 Estimates of water depth, channel cross section area, flow velocities, reach volume and mixing parameters for the River Dee. Table 4.8 Values of IHACRES parameters for modelled catchments presented in Littlewood and Jakeman (1993).Time constants are in days unless otherwise specified. Table 4.9 Summary of parameters of the rainfall runoff model for 20 minute and daily data. Table 4.10 Summary of rainfall and runoff values for 1991, for 20 minute and daily data (r and r* are the rainfall and effective rainfall annual averages over the period of record, 1986-92).

viii


Table 5.1 Geometric mean, maximum and minimum faecal coliform concentrations at two sampling sites on the River Washburn, during a step change in discharge, 26 May 1993. Statistics are given for the full data period and the three distinct phases of the event. Table 5.2 Times of travel and calculated velocities for various features of the reservoir release response of the River Washburn 26 May 1993. Table 5.3 Summary statistics for the observed faecal coliform concentrations at the upstream and downstream sampling sites on the Afon Clywedog. Table 5.4 Summary of wave speeds and bacterial peak travel-velocities in m.sec-1 for the reservoir release experiment on the Afon Clywedog. Table 5.5 Geometric mean (GM) faecal coliform concentrations in Afon Rheidol artificial hydrograph event response water, 17 February 1993. Table 5.6 Summary statistics of faecal coliform concentrations in the Afon Rheidol, during a combined storm-runoff and artificial flow event, 7 April 1993. Table 6.1 Conversion factors for comparison of light intensity values. Table 6.2 Summary of power law model die-off rates, ki, over ranges of irradiance investigated in original studies. Table 6.3 Summary of model die-off values at different temperatures as presented in Figure 6.2. Table 6.4 Summary of variables and parameters in the die-off component of the faecal coliform model. Table 6.5 Calculated faecal coliform die-off rate for different daily irradiance values. Water depth z=1m, annual mean suspended solids concentration SS=9.64 and the annual mean darkness die-off rate kd=0.335 (I0, kd and SS were taken from data for the River Dee near Chester in 1989). Table 6.6 Calculated faecal coliform die-off rate and percentage difference from mean, for different suspended solids concentrations. Water depth z=1m, annual mean daily irradiance I0=257.37 and the annual mean darkness die-off rate kd=0.335 (I0, kd and SS were taken from data for the River Dee near Chester in 1989). Table 6.7 The change in die-off from that at mean pH over the range of pH that might be expected in upland and lowland rivers in the UK. Table 6.8 Summary table indicating the main effects, typical die-off values and relative influence of the main variables of interest. Table 6.9 Ratio of observed irradiance minima at Krawaree, NSW, Australia (=35.81S) (after Kirk, 1983) and calculated maximum potential irradiance (units are W.m-2). Table 6.10 Summary statistics for variables and time variable parameters for estimating depth averaged irradiance for the River Dee at Huntington from 1 January 1986 to 10 November 1993. Table 6.11 Summary of parameters used to derive daily die-off estimates for the River Dee.

ix

Modelling Faecal Coliform Dynamics Table 6.12 Summary statistics for variables and time variable parameters for estimating overall dieoff in the River Dee at Huntington from 1 January 1986 to 10 November 1993.

Table 6.13 Summary statistics for variables and time variable parameters for estimating depth averaged irradiance for the River Dee at Huntington from 1 November 1991 to 18 November 1992. Table 6.14 Summary statistics for variables and time variable parameters for estimating hourly dieoff in the River Dee at Huntington from 1 November 1991 to 18 November 1992. Table 6.15 Summary statistics for estimated hourly overall die-off (per day), ktot, and various derivatives for the River Dee at Huntington from 1 November 1991 to 10 November 1993. Table 6.16 Parameters and constants for the model result presented in Figure 6.21.

List of Figures Figure 2.1 Graphical derivation of die-off rate k and 90% reduction time T90 assuming simple first order decay dynamics. Figure 2.2 Seasonal variation in faecal coliform concentration at the Pynes Raw Water intake on the River Exe in Devon, UK, between 1989 to 1992. Figure 2.3 Plot of light attenuation factor, , per metre depth against suspended matter concentration (after Pommepuy et al. 1992). Figure 2.4 Scatter plot of log10 median E.coli count (per 5g) against log10 soil moisture content (% of dry weight) for a catchment in the Yorkshire Dales (after Hunter and McDonald, 1991a). Figure 2.5 Plot of observed (after McFeters and Stuart, 1972) and modelled (solid line) of E.coli dieoff rate per day against pH. Figure 2.6 Survival of faecal coliform in raw river water at 20ºC showing a rapid decline in numbers in the aerated flask, little change in numbers in the static flask and N2 flushed flasks (after Zerfas, 1970). Figure 2.7 (a). Representation of the charge distribution between an E.coli cell and a solid surface. (b). Graph showing the net attraction energy between surfaces resulting from attractive Van der Waals and repulsive electrostatic forces as a function of separation distance in high and low conductivity (salt) solutions (after Marshall 1979). Figure 2.8 The influence of successive removal of sections of natural microbial communities on E.coli die-off in (a) "clean" river water, (b) sewage contaminated river water (after Flint 1987) and (c) estuarine water (after Enzinger and Cooper 1976). Figure 2.9 Seasonal variation in die-off of E.coli in (a). sewage contaminated and (b). clean water bodies, showing the effect of successive removals of groups of natural biota. Note that raw samples exhibit the most rapid die-off and that die-off is at its greatest during the summer months (after Verstraete and Voets, 1972). Figure 2.10 Scatter plots of log10 faecal coliform load against stage height for inputs from a. overland flow and b. non-matrix throughflow (after Hunter et al., 1992).

x

Modelling Faecal Coliform Dynamics Figure 2.11 The relative importance of bacterial input rates from the catchment land store and the channel sediment store to stream bacterial dynamics, with distance downstream (after Hunter et al., 1992).

Figure 2.12 Propagation of artificially generated hydrograph with response of total coliforms (TC) and E.coli concentration at locations (a) 400m, and (b) 2500m downstream of the hydrograph source (after McDonald et al., 1982). Figure 2.13 The response of total coliforms (TC) and E.coli concentration to a step change in stage height (after McDonald et al., 1982). Figure 2.14 Bacterial die-off curves as predicted by models in the literature (after Crane and Moore, 1986). Figure 2.15 A graphical representation of the functioning of the advection dispersion equation (ADE) as applied to model the concentration of a conservative contaminant when there is (a),dispersion without advection, (b), advection but no dispersion, and (c), when the two affects are combined. Figure 2.16 Profile of simulated and measured coliform bacteria concentrations for the Tigris river in Iraq (after, Al-Layla and Al-Rizzo, 1989). Figure 2.17 Plot of washout limitation function, for Sat=1000, Equation 2.3.34. Figure 4.1 Release discharge from Aberffrwd Dam, Qi, smoothed Qi (Qi ave), and downstream stage at the sampling site on the Afon Rheidol. Figure 4.2 Wave speed and discharge, and stage-discharge relationship for the sampling site on the Afon Rheidol. Figure 4.3 Time-series of discharge in 13 successive model reaches of 684.6 m length between Aberffrwd Dam to Blaengeufordd on the Afon Rheidol. Figure 4.4 Time-series of release discharge, Qi, at Aberffrwd and routed discharge, Qo', at Blaengeufordd. Figure 4.5 Release discharge from Aberffrwd Dam on Afon Rheidol, 7 April 1993. Figure 4.6 Routed discharge, Qo', and discharge from stage, Qo'', at Blaengeufordd, and Q2' is Qo'' routed to Site2. Figure 4.7 Stage discharge relationship for Bryntail Weir. Figure 4.8 Programmed stage, hi prog, and observed stage, hi obs, at Bryntail Weir on Afon Clywedog. Figure 4.9 Three part rating curve for the compound Crump weir at Cribynau on Afon Clywedog. The legend shows the rating for the component parts of the rating curve (see also Table 4.9). Figure 4.10 Discharge in cumecs at Bryntail, Qi, and Cribynau, Qo. Stage at Cribynau, ho, is also shown to demonstrate the stage distorting effect of the Crump weir. Figure 4.11 Observed discharge and water surface levels on the River Dee for 1995. Figure 4.12 Water surface levels with the tidal component removed.

xi

Modelling Faecal Coliform Dynamics Figure 4.13 Chester Weir level against Ironbridge level (tidal component removed). Figure 4.14 Difference in water surface level between Chester Weir and Ironbridge plotted against discharge.

Figure 4.15 Estimated water depth at Huntington against discharge at Manley Hall. Figure 4.16 Estimated inflow Qi and slow-flow component Qs, with rainfall r*. Figure 4.17 Estimated discharge Qik, with discharge at Manley Hall on the River Dee and adjusted rainfall r*. Figure 5.1 Faecal coliform responses observed at the upstream and downstream ends of a 2 km reach of the River Washburn to a step increase in stage. Figure 5.2 Faecal coliform responses observed at upstream and downstream ends of a 3.6 km reach of the Afon Clywedog downstream of the Clywedog Dam. Figure 5.3 Summary of wave speed and flow velocities for the monoclinal flow rises and resulting bacterial entrainment peaks on the Afon Clywedog (see Table5.3). Figure 5.4 Velocity of observed bacterial peaks plotted against observed wave speeds including arrival times of bacterial peaks (values in m/s). Figure 5.5 Variations in faecal coliform load through the period of the first artificial flow event on the Afon Rheidol. Qi is release flow and Qo is discharge at the sampling site. Figure 5.6 Variations in faecal coliform concentration through the period of the first artificial flow event on the Afon Rheidol. Figure 5.7 Time-shifted faecal coliform concentration and discharge plotted against time relative the start of the release and the first arrival of the event water downstream. Figure 5.8 Time of arrival of bacterial peaks and flow wave relative to the beginning of the release. Figure 5.9 Time series plots of estimated discharge and observed faecal coliform concentration at Blaengeufordd (top) and Site 2, 7 April 1993. Figure 5.10 The main faecal coliform responses on the Afon Clywedog (top) and Afon Rheidol (7 April 1993, middle; 17 February 1993, bottom), displaying the similarity in feacal coliform response to artificial hydrographs. Figure 5.11 Observed and modelled faecal coliform concentrations in response to a step change in flow on the River Washburn. Figure 5.12 Model fit with fully dynamic AMV and fully time varying parameters. Figure 5.13 Model fit to the Clywedog experimental data using the latest version of the model with single channel store. The entrainment time-lag is calculated from the difference between wavefront arrivals and the arrival of the bacterial peaks. Figure 5.14 Model fit to the Clywedog data with the latest version of the model with distributed storage and falling wave entrainment. Figure 5.15 Variation in die-off rate (T90 = 1/ktot) and model variables associated with rapid release channel store bacterial numbers in response to the experimental release on the Clywedog.

xii

Modelling Faecal Coliform Dynamics Figure 5.16 The time varying parameters associated with model result in Figure 5.12. "Mean" is the average of the summed parameters over the duration of the model run.

Figure 5.17 Model fit with fixed parameters, using mean parameter presented in Figure 5.16 above and the mean value of bo/Df for the upstream input term. The model fit was adjusted using the input variables Nsk=20.Q2 cfu.sec-1, Nik=1000 cfu.sec-1. Figure 5.18 Changes in flow on the Afon Clywedog represented by various normalised parameters, demonstrating alternative functions for the distribution of bacterial accumulations in the channel. Figure 5.19 An example of model output using wetted perimeter as the entrainment threshold variable. As can be seen this results in the majority of stored organisms being entrained at lower discharge. The time shift parameter =0. Figure 5.20 Model fit to the Afon Rheidol event of 17 February 1993, entrainment is driven by discharge from stage. Figure 5.21 Model result using the single store version, driven by the input discharge (smoothed and delayed to coincide with observed stage). This version of the model included falling wave entrainment. Figure 5.22 Discharge, Qi, from Aberffrwd dam and routed discharge, Qo, at the field site at Blaengeufordd on the Afon Rheidol for the three days immediately preceding and including the experimental release. Figure 5.23 Model result for successive flow events. Figure 5.24 Model channel store charging-up and undergoing entrainment during events presented in Figure 5.23. Figure 5.25 Change in storage of 25 model channel bacterial sub-stores. Figure 5.26 Detail of model response for the 17 February event. Figure 5.27 Model output with zero travel-time delay between discharge and bacterial response. The time-shifted smoothed release discharge is used as the driving variable. Figure 5.28 As Figure 5.27, but entrainment is driven by discharge from observed stage. Figure 6.1 Plots of power law models fit to faecal coliform die-off rate per day for a range of light intensities and for various studies. Figure 6.2 Plots of modelled die-off, kT, against water temperature as fit to data from various studies. Figure 6.3 Plots of calculated die-off (ki = Iz,avg) over the seasonal range of daily irradiance, I0, that might be expected for a UK river, for a. different suspended solids (SS) concentrations (z=1m) and b. various water depths (SS=9.64). Figure 6.4 Ideallised curves of clear sky irradiance, Imax, observed irradiance, Io, at Bidston Observatory, and modelled irradiance, Io', based on sunshine hours data for Ness Gardens. Figure 6.5 Mean daily cloud influenced irradiance,I0 ,as observed at Bidston Observatory, and modelled,Io', from sunshine hours at Ness Gardens and calculated clear sky irradiance, Imax .

xiii

Modelling Faecal Coliform Dynamics Figure 6.6 Plots of air temperature at two meteorological sites and water temperature in the River Dee at Huntington at 9am from November 1991 to November 1992. Figure 6.7 Plots of modelled (Tw') and observed water temperature in the River Dee at Huntington for the period of record of faecal coliform data;1986 to 1993. Figure 6.8 A sample of the time-series for turbidity, suspended solids and discharge for the River Dee at Huntington, showing modelled suspended solids concentration. Figure 6.9 a. Estimated mean daily surface irradiance (top), I0 , and depth averaged irradiance, Iz,avg,(i.e. reduced by water depth and attenuation by particulates). b. Estimated die-off, ktot(s), and the component die-off time-series, ki, kT, kpH and ks. Figure 6.10 Estimated die-off, ktot and kT (die-off due to temperature alone) and log10 faecal coliform concentration in the River Dee between August 1988 and May 1990. Figure 6.13 Estimated hourly T90 (1/ktot) with maximum and minimum die-off curves for the River Dee from November 1991 to November 1992. Figure 6.14 The relationship between minimum and maximum die-off rate, ktot, (top) and the relationships between die-off rate and T90 for peak die-off (centre) and minimum die-off (bottom). Figure 6.15 Two five day samples of diurnal cycles in estimated die-off at mid-summer (top) and mid-winter (bottom). Figure 6.16 Ninety-percent die-off travel-time, X90, and faecal coliform concentration at Huntington Intake. Figure 6.17 Percentage reduction in bacterial concentration in the 6 km reach from Ironbridge to Huntington Intake. Figure 6.18 Model inputs and output for the River Dee, November 1991- November 1992. The input WwTW, is the "point-source" input referred to in the text. All units are in CFU per 100ml. Figure 6.19 Model output with constant single input value of 4000 cfu per 100ml. Figure 6.20 Model parameters for the single constant input model of faecal coliform concentrations in the River Dee. Figure 6.21 Model output for the daily ADZ with quick flow input and constant upstream input. Figure 6.22 The quickflow run-off input component for the simulation of the faecal coliform behaviour in the River Dee. Figure 6.23 Variations in model output and observed faecal coliform concentrations for the River Dee at Huntington Intake, plotted on logarithmic and natural scales. Map 1 Afon Rheidol site map. Map 2 River Washburn site map. Map 3 Afon Clywedog site map. Map 4 River Dee site map.

xiv


1. Introduction

1 Introduction ________________________________________________________________________ "Determining the concentration of sanitary indicator bacteria (in this case faecal coliforms) is an important element in assessing the risk to public health from water bourne diseases in water intended for public supply and/or used for recreational purposes" Hunter (1999). The purpose of this study was to examine the key processes influencing faecal coliform concentrations in streams and rivers and to produce a model that embodies those processes which could be tested against event based and longer term data.

Faecal coliform bacteria are found in the intestines of all warm-blooded animals and their presence in natural waters is indicative of faecal contamination and the potential presence of pathogenic organisms. Faecal coliforms enter the natural environment when excreted in faeces and reach the riverine environment in surface runoff from rainfall, in water percolating from land adjacent to the channel, in effluent from waste water treatment works or when directly defecated into the stream. Of the vast numbers defecated only a small proportion is likely to reach the stream channel from the catchment surface. The vast majority are killed on the surface by sunlight or through dehydration, or are filtered out in the soil system. On entering the stream the coliforms have to compete for nutrients and are grazed upon by native microbes. Sunlight is the major die-off influence in natural waters, surface effects offer extended survival to the contaminant organisms through the potential to cluster with particles and settle into darker waters or simply by coating them with an armour of clay. By attachment they can effectively shelter from sunlight and hostile predators. By 1


1. Introduction

settling the organisms can enter an environment where the sun may not reach and where nutrient availability is improved. Where the organisms deposit they may form pockets of storage capable of resuspension by storm flows or the passage of a boat and contribute to the reduction of the overlying water quality. Ultimately the organisms will either die-off or be washed out to sea, where they contribute to bathing water contamination.

Given this complexity of processes influencing the faecal coliform concentrations in the riverine environment the challenge was to provide improved methods for modelling this behaviour. New models have been developed and are tested against reservoir release response data and long-term monitoring data for a drinking water abstraction intake. The models prove highly effective and can reproduce the observed data to a high degree of accuracy.

In general, existing models for faecal coliform dynamics make certain simplifying assumptions. Firstly, that bacterial die-off occurs at a constant rate and secondly that organisms that have been deposited on the river bed can no longer affect the concentration in the water. The new models are not limited by such assumptions and include terms for the effects of sunlight and water temperature on die-off, and the reduction of light intensity by suspended matter. The work includes a technique for estimating daily mean solar radiation from sunshine hours. The bacterial transport related components of the models include supply limited catchment run-off, slow-flow inputs, point source inputs and deposition to and entrainment from storage on the channel bed.

The full bacterial transport and die-off model has the potential for applications to other microbial contaminants and indicators. The fluvial dynamics model has the potential for applications in non-microbial fields for modelling particulate associated contaminants such as heavy metals, radio-nuclides and persistent organics. A key recommendation of this study is for the widespread use of instrumentation for solar radiation or sunshine hours for longterm monitoring of microbial contamination, measurements of turbidity are of vital importance if relative estimates of die-off are to be meaningful. In addition to this baseline data, the development of a probabilistic framework for the prediction of periods of poor water quality should form the basis of a follow-on study. A simple calculation based on flow velocity and time variable die-off rate is presented and could form the basis of such an index. 2


2. Literature Review

2 Literature Review ________________________________________________________________________ This literature review concentrates on the processes affecting faecal coliform numbers in streams and rivers and previous attempts to model faecal coliform behaviour. The chapter has three main sub-sections that introduce, in turn, the causes of faecal coliform die-off, their transport into and within the stream system, and finally mathematical models that have been used in attempts to simulate these processes. The chapter commences with die-off, the primary process affecting faecal coliform numbers; once released from their initial growth medium, i.e. faecal matter, the rate at which the organisms die will increase as the conditions become less like those within the parent material.

2.1 Faecal Coliform Die-off Dynamics Faecal coliforms are present in the intestines of all warm-blooded animals. Once excreted from this optimal growth environment their numbers start to decline. On the catchment surface moist dark conditions prolong their survival, whereas sunlight and drying conditions accelerate die-off. Organisms that survive long enough to be transported into the river system have one key survival need met, that of moisture. In the water the factors which most accelerate die-off are sunlight, acidity, temperature and microbial effects. Particles aid survival by shading the contaminant organisms from harmful sunlight, shielding them from predators and assisting in their settlement to the riverbed where the conditions favour extended survival. 2



Die-off is the primary process affecting faecal coliforms, continually reducing their numbers irrespective of other processes. If the organisms are dispersed in a river, they continue to die-off. If they are diluted by rainfall and washed into the soil water system the number of organisms falls as they die. Were the organisms to remain stationary, unaffected by dispersing or diluting effects, their numbers would also fall as they died. The important consideration about die-off is how quickly it occurs, i.e. the rate of die-off. An appreciation of die-off and the relative affects of different die-off influences is, therefore, an essential prerequisite to any study of faecal coliform dynamics.

Basic background information about the study of faecal coliforms is presented before examining specific die-off effects. The section is sub-divided according to the nature of the factors causing die-off. For example, sunlight and temperature are considered to be primary physical environmental factors influencing the rate of die-off, these are in the first group of processes. Soil moisture and water chemistry are dependent state variables of the soil and river systems, respectively, and are in the second group of physico-chemical factors. Where equations that relate a particular environmental variable to faecal coliform die-off are available in the literature, these are presented. For certain variables there is sufficient data to derive equations for die-off due to the effect of that variable. Chapter 6 develops equations for these variables and illustrates the results using data for the River Dee in North Wales. Finally, microbial interactions are examined and the broad causative effects of the different influences are summarised in Section 2.1.5.

2.1.1 Quantifying faecal coliform die-off The study of coliform survival dynamics spans some ninety years e.g. Chick (1908) and a consequence of this long history is the variety of terminology, different approaches to quantifying die-off, and the enumeration of different coliform groupings. The corollary of this is that care should be taken when attempting to compare the results of different studies. A detailed examination of their methods is essential as well as the ability, given certain assumptions, to convert these disparate results into one common system. For this reason information is presented about how die-off is quantified, the differences between coliform counts, and problems associated with the reporting of experimental conditions.

3



Quantifying faecal coliform die-off In order to compare the results from different studies of faecal coliform survival and assess the relative importance of the different influences on survival, die-off must be quantified in some consistent way. The two mathematical indices most commonly used to describe faecal coliform die-off are introduced below and some of the problems encountered when attempting to quantify a behaviour affected by many different factors are highlighted. The two commonly used indices are the die-off rate coefficient, k, and T90, the time taken for a population to fall to 90% of the initial value. The die-off rate coefficient k is derived from first order decay dynamics as described by Chick's Law (Chick, 1908), which describes a simple exponential decline from an initial population:

N t = N 0 10 -kt where, N0 is the initial population, and Nt is the population at time t > 0. The die-off rate coefficient, k, is the amount the log10 population falls per unit time and is given by

k =

log10 N 0 - log10 N t t

where t is the time taken for N to change from N0 to Nt.

For

k = 0 there is no die-off i.e., Nt = N0 k < 0 indicates growth and k > 0 indicates die-off.

The numerical value of k can be found graphically by plotting the log10 bacterial calculation against time and finding the difference between two values of bacterial numbers, N2 from N1 (Figure 2.1), for unit time (e.g. hour/day) on the x-axis. Similarly, T90 is the time taken for the population to fall by one log10 cycle from the initial count and

k =

1 T 90

or, k =

2 T 99

or, k =

3 T 99.9

etc,

where T99 is time taken for a 99% reduction in population. T99.9 would be the time for a 3log cycle reduction in population, i.e. 99.9% reduction in population. McFeters and Stuart (1972) used T50 as their measure of die-off, this being the time taken for a 50% reduction in the initial bacterial population. Assuming that simple first order decay is appropriate T50 values can be converted to die-off rate k by the relationship

k =

log10 0.5 T 50 4



In reality a time delay before the onset of die-off in the bacterial population is often observed (Crane and Moore, 1986; Flint, 1987; Enzinger and Cooper, 1976). In such situations the expression of k is more problematic because it must be decided whether k is calculated for the period including that delay, or the period from the onset of die-off alone. It is not clear from the literature which approach is commonly used. It might be argued that exclusion of the time delay would yield a more realistic value of k given that no die-off occurs during the initial period. Alternatively, this exclusion might be interpreted as giving a die-off rate that is too rapid. T90 includes time delay and its' use in such cases would be unambiguous in that it is simply a measure of the time taken for a ninety percent reduction in population; how the population changes to reach that point is not incorporated in the value.

Figure 2.1 Graphical derivation of die-off rate k and 90% reduction time T90 assuming simple first order decay dynamics.

Care should thus be exercised when making direct comparisons of die-off rates taken from different studies. Further potential causes of inconsistency in die-off results concern the microbial composition of the sample being counted, the type of count being made, and variations in the experimental or environmental conditions used. For example differences in light and/or temperature conditions can result in very different faecal coliform die-off rates (see Section 2.1.2).

Total coliform, faecal coliform and E.coli counts Mack (1977) gives an excellent history of the development of the total coliform count as an index of bacterial water quality. In 1880 Eberth discovered that Salmonella typhi produced 5



typhoid fever in man and Gaffky isolated the typhoid organism in 1884. In 1885, Escherisch, examining a stool sample, discovered many organisms, similar to that which produced typhoid, one of which was named Bacillus coli, later becoming Escherichia coli (Mack, 1977), and identified as being characteristic of the faeces of warm-blooded animals (Dutka, 1973). The easy and unambiguous methodology used to identify coliforms meant that this group became a key means of indicating the presence of faecal contamination in aquatic environments (Dufour, 1977). The coliform group, which gives the total coliform count, however, comprises a variety of organisms including Escherichia, Klebsiella, Enterobacter and Citrobacter types. Of these Escherichia is the only ‘undoubted inhabitant of the gastrointestinal tract’, Klebsiella, Enterobacter and Citrobacter are only found in relatively small numbers in faeces and can easily be found on vegetation and in soils (Dufour, 1977). A total coliform count, therefore, does not indicate contamination that is specifically faecal in origin.

Eijkman (1904) sought an assay that would detect coliforms that were more specifically faecal in origin. This lead to the development of the faecal coliform, sometimes referred to as the thermotolerant coliform, index which involves incubation at a higher temperature, 44ºC rather than the lower 37ºC used for the total coliform count. Fewer organisms can survive at the higher temperature, hence the term thermotolerant. The specificity of the faecal coliform count was a great improvement, but there were still occasions when industrial effluent samples with high concentrations of carbohydrate materials and no evidence of warm-blooded faecal contamination gave rise to a positive test. The positive result was usually due to Klebsiella species (Dufour, 1977). In human and animal faeces, however, 90-95 % of the coliforms are E.coli (Dufour, 1977) and evidence from microbiological laboratories analysing environmental samples suggests that between 80 and 85% of faecal coliforms in freshwater are E.coli (Godfree, 1994). Until 1983, in the UK, results were expressed as E.coli but the test in fact gave a faecal coliform count. Further confirmatory procedures are required to give an E.coli count; the production of gas from lactose and indole, in the indole test, carried-out at 44ºC (HMSO, 1969).

Clearly slight differences in faecal coliform enumeration are possible between studies. Given the assumption that the majority of faecal coliforms are E.coli, and that results gained by virtually identical methods may be expressed as faecal coliform or E.coli counts, for the purposes of this study the two indices have been treated as being synonymous. A further 6



justification of this statement is that the aim of this study is to examine processes and cause and effect relationships in order to develop an improved model of faecal coliform dynamics in streams and rivers. The absolute numbers of organisms are not important for this purpose. Furthermore, it is not an examination of the value of specific indicator organisms or the accuracy of specific bacterial enumerations. Other issues relating to experimental conditions used in die-off studies may be of greater significance to the results summarised in the current chapter.

Comparison of data The matter of reporting experimental methods and conditions is of great importance when attempting to compare studies undertaken by a number of different authors. If the authors fail to list all experimental conditions used in their experiments, the best information that can be obtained from comparative assessments of the work is only qualitative. Crane and Moore (1986), for example, examined a wide range of data from studies of bacterial die-off in the environment. Die-off rates were found to be highly variable, spanning several orders of magnitude. Environmental factors, such as the non-linearity of pH and temperature effects, were assumed to be the main cause of the differences and attempts to relate die-off rates to these using multiple regression were unsuccessful.

Improvements in the reporting of experimental results and the use of graphical analysis techniques in this study has assisted in the comparison of results and these data are used to develop relationships for faecal coliform die-off with environmental variables (see Chapter 6).

7



2.1.2 Physical influences on faecal coliform die-off The primary physical controls on the rate of faecal coliform die-off in the environment are sunlight, temperature and the availability of moisture. Moisture availability is important where the organisms are exposed to the open air; faecal coliform delivery to streams from moist soil areas can be much greater than that from dry areas (Hunter and McDonald 1991a). Sunlight accelerates the rate of faecal coliform die-off on both the catchment surface and in river water, indeed UV light is an important method of disinfecting wastewaters and water for potable supply (Wyer et al., 1994). Faecal coliform die-off increases logarithmically with increases in temperature and by nearly an order of magnitude over the range of temperatures observed in UK rivers. Thus sunlight and temperature changes will have a large effect on the persistence of faecal coliform in natural waters.

In this section equations relating faecal coliform mortality to sunlight intensity are presented and data from a variety of studies of temperature effects on die-off are plotted together for comparison. Chapter 6 makes use of these data to derive equations for die-off and temperature to be used in the proposed model of faecal coliform dynamics.

Seasonal patterns Seasonal cycles in faecal coliform concentration in natural waters are driven by annual changes in temperature, sunlight intensity, river flow (e.g. Cohen and Shuval, 1973; Davenport et al., 1976; Gordon, 1972), the supply of organisms from the catchment (see Hunter and McDonald 1991a), and the intensity of hostile activity by native biota (Verstraete and Voets, 1972). In upland areas winter concentrations tend to be lower than are those in summer. Faust and Goff (1977) found that pasture and forest were similar seasonally, producing peak coliform discharges in summer. The lower winter concentrations result from dilution by high flows and the exhaustion of diffuse sources of organisms from the land (Hunter and McDonald, 1991a). The lack of dilution afforded by low summer flows results in more concentrated run-off and hence higher stream concentrations and may also reflect seasonal livestock practices. The seasonal patterns in faecal coliform concentrations are controlled by the supply of organisms to the stream. In lowland streams and rivers the highest concentrations tend to be observed during the winter months (e.g. Barbe and Francis, 1995; Hirn et al., 1979), summer concentrations are much lower (Figure 2.1, see also Chapter 6). The high winter concentrations result from a combination transport and die-

8



off processes (Table 2.1). Faust and Goff, (1977) observed the highest faecal coliform runoff from cultivated land in the winter, probably as a result of manure application. 10000

CFU per 100ml

1000

100

Observed FC Moving average 10 Jan-89

Jul-89

Jan-90

Jul-90

Jan-91

Jul-91

Figure 2.2 Seasonal variation in faecal coliform concentration at the Pynes Raw Water intake on the River Exe in Devon, UK, between 1989 to 1992.

An exception to this general case, for lowland rivers, might be expected where a sewage effluent contributes large numbers of organisms relative to the concentrations in the receiving water. This would be similar to the upland stream case, where higher winter flows giving adequate dilution, resulting in lower winter concentrations. During the summer low flows the dilution afforded to the effluent would be less resulting in higher concentrations. However, despite the winter dilution, the greater flow velocities that might be expected, and the reduction in die-off inducing forces relative to summer mean that the overall flux of organisms is likely to be greatest during the winter.

The following sections examine influences on faecal coliform die-off responsible for these general seasonal patterns in more detail and, where available, introduce equations relating die-off to environmental variables.

9



Table 2.1 Summary of die-off and transport related factors resulting in high winter faecal coliform concentrations in lowland streams and rivers (reverse each factor to give summer effects). Die-off related factors

Transport related factors

Fewer hours of sunlight

Washoff from the catchment surface

Reduced light intensity

Reduced residence in river reaches

Lower temperature

Stirring of sediment

Protection by particles Reduction in light penetration Wetter soil conditions Slower biological purification processes More favourable pH The effect of sunlight This section examines the processes and mechanisms by which sunlight causes faecal coliform die-off, gives examples of contrasting die-off rates under brightly lit and dark conditions. Finally simple equations are introduced relating faecal coliform die-off in water to solar radiation.

Sunlight affects the die-off rate of faecal coliforms directly through cell damage and indirectly by altering other characteristics of the physical environment the organisms encounter. This latter component includes such effects as raising water temperature, and enhancing evaporation and moisture loss, which enhance die-off. Sunlight also drives photosynthesis thus increasing the oxygen content of water; this also accelerates die-off (Zerfas, 1970). A further factor driven by sunlight is algal populations that grow in sunny conditions; it is possible that algae produce substances toxic to coliforms (e.g. Mezrioui et al., 1994; Verstraete and Voets, 1972). Sunlight is, thus, an important variable in driving faecal coliform die-off.

The effect of sunlight on faecal coliform die-off can be dramatic. Water samples exposed to bright sunlight have been observed to undergo 90% die-off in only a few hours, as opposed to a number of days in darkness (e.g. Fujioka and Siwak, 1985; Gannon et al., 1983; Bellair 10



et al., 1977; Kunkle and Meiman, 1968). Similarly, an exposed lawn site yielded a T90 of around 5 days; die-off took three times longer on a shaded hill-slope (Van Donsel et al., 1967). There are insufficient data to adequately quantify the effect of sunlight on the catchment surface. Although, the factors that are likely to control die-off on the ground surface are those that influence moisture conditions and the strength of sunlight falling on the surface. These include vegetation, slope, aspect and factors relating to the convergence of streamlines / flow-paths. Thus, the characterisation of die-off on the catchment surface requires attention.

The ability of sunlight to purify bacterially contaminated waters is influenced by the longterm water quality of the water body in question. This is due, not only to direct physical mechanisms but also to changes in biological processes. For example, a study of two lakes, one sewage contaminated, the other relatively clean, showed that the die-off due to sunlight was greater in the clean lake water (Verstraete and Voets, 1972). In clean water the depth of light penetration will be greater (see Pommepuy et al., 1992). This enhances the direct lethal effect sunlight has on faecal coliforms and can raise oxygen which benefits the competing biota further accelerating coliform die-off (Verstraete and Voets, 1972; Zerfas, 1970). In poorer quality waters where light attenuation is significant, the benefits of sunlight will be reduced. The suspended matter, which reduces the depth of light penetration, also shields the organisms which may be further protected by attaching to the particles (see Section 2.1.3). Pommepuy et al. (1992) found 73% of observed light attenuation in fresh river water to be due to suspended and dissolved organic matter. The data presented in Figure 2.3 yielded the following relationship between suspended matter and light attenuation,

 = 0.22 C sm

0.78

where Csm is the concentration of suspended matter in mg.l-1 (Pommepuy et al., 1992). The use of such a relationship is examined in Chapter 6.

11



Figure 2.3 Plot of light attenuation factor, , per metre depth against suspended matter concentration (after Pommepuy et al. 1992).

Sunlight, in the form of solar radiation, has a major impact on faecal coliform survival. In darkness the organisms may persist for many days, in bright sunlight, however, populations can be destroyed in a few hours. The relationship between solar radiation and faecal coliform die-off is linear and the above equations may readily be incorporated into models of faecal coliform dynamics (see Chapter 6). Temperature is driven by solar radiation and varies in a similar manner over short and long time-scales. A number of authors have studied the effects of temperature on faecal coliform die-off and the main findings of these studies are outlined in the next section.

Temperature In this section the effect of temperature on faecal coliform die-off in water is examined. Dieoff in natural waters is shown to increase exponentially with temperature (e.g. McFeters and Stuart, 1972; Flint, 1987; see Chapter 6). The affect of temperature on die-off is diminished in sewage contaminated waters, indicating the need to account for broader water quality when attempting to model faecal coliform dynamics.

As demonstrated above a change from bright sunlight to overcast conditions can reduce the rate of die-off by a factor of ten; such a change might occur rapidly with the passage of a cloud. Large changes in sunlight intensity, occurring in a matter of minutes, might be 12



expected to contribute to the inherent noisiness of bacterial concentrations in rivers. The physical controls on temperature mean that large variations are unlikely to occur at timescales shorter than a few hours. There may be a strong diurnal pattern in air temperature, but the diurnal effect on water temperature is likely to be damped out relative to longer-term, i.e. seasonal influences. Hence, temperature is likely to be important in determining the long-term die-off conditions in a river.

Many studies have shown that faecal coliform die-off is more rapid at high, rather than lower, temperatures (e.g. Kunkle and Meiman, 1968; McFeters and Stuart, 1972; Flint, 1987; Evison, 1989). Early field studies highlighted the difficulty of separating out the effects of temperature from those of sunlight (Kunkle and Meiman, 1968). McFeters and Stuart (1972) overcame these problems by the use of dialysis chambers under laboratory conditions. The chambers allowed the passage of water and nutrients while maintaining the test organisms in a fixed volume under carefully controlled laboratory conditions. The dialysis chambers were inoculated with naturally occurring E.coli and suspended in filtered stream water in a recirculating flume in order to simulate the conditions experienced in a natural river system. The results of this were very similar to the other results for fresh river water (e.g. Flint, 1987; Evison, 1989)(see Chapter 6).

Two other studies of clean river water produced both higher (Flint, 1987) and lower (Evison, 1989) die-off rates than that of McFeters and Stuart. This might be due to the chemistry of the river waters; McFeters and Stuart found that die-off was more rapid in low conductivity water. An alternative explanation might be found in the different treatment of the experimental waters. The unfiltered river water resulted in the greatest die-off rates (Flint, 1987), whereas filter sterilised (at 0.2um) water (Evison, 1989) resulted in the slowest die-off. A number of studies (in particular, Verstraete and Voets, 1972; Enzinger and Cooper, 1976; Flint, 1987) have shown that successive removal of parts of the natural stream biota reduces the self-purification ability of a water (see Section 2.1.4). The dialysis membranes used by McFeters and Stuart effectively filtered the water at 0.45um pore size, resulting in a microbial composition somewhere between the other studies and, perhaps, the die-off results they observed.

13



A simple equation relating faecal coliform die-off and temperature has been derived from the literature data and is presented in Chapter 6. For a list of example die-off rates for freshwater see Appendix I.

In summary, the higher the water temperature the greater the increase in faecal coliform dieoff rate. In sewage impacted waters the effect of changes in temperature on the rate of faecal coliform die-off is less than in clean river water. Out of the water, different factors effect faecal coliform die-off. The following section investigates these factors.

Soil moisture and survival on the catchment surface A brief consideration of faecal coliform survival in soils and on the catchment surface is given here. This is pertinent to the broader understanding of the delivery of faecal coliforms to the stream channel and of seasonal variations in stream concentration where the delivery of organisms from the catchment contributes significantly to the numbers observed in the stream.

Figure 2.4 Scatter plot of log10 median E.coli count (per 5g) against log10 soil moisture content (% of dry weight) for a catchment in the Yorkshire Dales (after Hunter and McDonald, 1991a).

Faecal coliform survival is enhanced in moist soils and in locations that favour the continuity of cool moist conditions, i.e. shaded, well vegetated areas in soils with good moisture retention. Moisture is perhaps the most important factor in determining bacterial 14



survival (Van Donsel et al., 1967). Moist land areas have been found to contribute significantly greater numbers of faecal bacteria to surface waters than dryer land areas (e.g. Hunter and McDonald, 1991a: Bagdasaryan, 1964), and survival is greatly extended during periods of water-logging (Chandler et al., 1981), see Figure 2.4.

Moisture content and retention properties of the soil are in turn influenced by other factors that may indirectly affect bacterial survival, for instance, soil type and vegetation cover (e.g. Beard, 1940; Hunter and McDonald, 1991a). Vegetation type, as an indicator of long-term soil moisture regime, could be used as a marker for zonation of a catchment into active and passive areas (Hunter and McDonald, 1991a). Active zones would comprise boggy/moist areas where bacterial survival is enhanced and run-off is actively generated. Reduction in bacterial input to surface waters could be achieved by preventing the access of livestock to such zones (Hunter and McDonald, 1991a). Vegetation type and cover also influences soil moisture conditions. Dense grasses afford protection from adverse environmental conditions, limiting the effect of sunlight, wind and high temperatures (Chandler et al., 1981; Zyman and Sorber, 1988). Where soil moisture levels are maintained artificially by the application of liquid organic wastes long survival times can be expected; T90 values of 9 days for pastures and 11 days in top soil have been observed (Chandler et al., 1981). Liquid animal wastes have high moisture content and good moisture retention properties (Crane and Moore, 1986). Chandler et al. (1981) found that the faecal coliform concentration of the waste applied to various test plots was more important in determining soil concentrations than the rate of application. The high moisture levels also meant that the soil type of the plots had no significant effect on the rate of die-off.

These results suggest that the use of soil moisture maps in determining bacterial source areas and for avoiding dense stocking of the kind of the “active zones” discussed by Hunter and McDonald (1991a) would be valuable to both water quality modellers and catchment management planners. Appendix II lists die-off rates for a variety of soil treatments.

15



2.1.3 Physicochemical and nutrient effects Section 2.1.2 introduced the effects of the primary physical factors, sunlight, temperature and moisture, on faecal coliform die-off. This section examines secondary physicochemical influences and effects that relate to the direct environment surrounding the contaminant organisms. The impact of extremes of pH and dissolved oxygen, both readily measurable variables, are introduced, and the complicated survival enhancing processes produced by the presence of particles are discussed.

These relationships and effects are grouped together as secondary factors in that they are internal state variables of the soil or water system under consideration. Temperature is also an internal state variable driven by insolation. The sun, however, not only drives temperature but acts directly on the contaminant organisms. During cloud cover the direct solar effect is absent, the temperature, however, still varies both diurnally and seasonally and should be treated as an independent variable for the purposes of die-off. Dissolved oxygen content on the other hand is itself determined by both temperature and sunlight. As temperature increases the amount of oxygen that can remain dissolved in water is reduced. Sunlight drives photosynthesis in algae and macro-phytes thereby consuming carbon dioxide and returning oxygen into the water. In the absence of light the respiration of dead organic matter and consumption of oxygen may exceed the rate at which oxygen dissolves into the water from the atmosphere, thus oxygen levels fall. Under extreme conditions of anoxia perhaps resulting from eutrophication the microbes that breakdown the organic materials take oxygen from other sources such as dissolved oxides; these reduction reactions cause an increase in hydrogen ion concentrations thus increasing acidity.

These variables are thus driven by inter-dependant processes and have been accounted for in existing models (Whitehead and Hornberger, 1984) which could be used to drive die-off relationships in a model of faecal coliform dynamics.

pH and faecal coliform die-off In the same way that exposure to strong sunlight causes rapid die-off of faecal coliforms, both high and low pH cause a dramatic increase in die-off of faecal coliforms, compared to neutral conditions (Figure 2.5). This effect has been observed in soils (Cuthbert et al., 1955, McCalla, 1964) and stream-waters (Cohen, 1922; McFeters and Stuart, 1972). For soils this may be important when considering the major source areas of organisms in upland areas. 16



For example, in acid peat soils, pH=2.9 to 4.5, die-off within a few days has been observed, as opposed to several weeks in limestone soil where pH is around 8 (Cuthbert et al., 1955). Cuthbert et al.(1955), suggested that acidic conditions not only increased the rate of faecal coliform die-off but may also have affected nutrient availability and the action of antimicrobial agents. Similarly in-situ tests with mine drainage water, pH=3, caused a reduction in faecal coliform 106 to 100 FC.100ml-1 in only two hours (Carlson-Gunnoe et al., 1983). The reduction in neutral water over the same period was negligible. Die-off appears to be at a minimum at neutral pH (Table 2.2).

Figure 2.5 Plot of observed (after McFeters and Stuart, 1972) and modelled (solid line) of E.coli die-off rate per day against pH.

At non-neutral pH the normal cell functioning of a faecal coliform bacterium is upset by the inhibition of metabolite transfer through the cell walls and direct cell damage (McCalla, 1964; Singleton and Sainsbury, 1981). Although similar in that they inhibit metabolite transfer, the exact mechanism of this disturbance under acid (low pH) and alkaline (high pH) conditions is different. Under acid conditions hydrogen ions are abundant and become absorbed to the cell walls inhibiting the replacement of essential cations, sodium, potassium and calcium, within the organism. The leakage of potassium and other compounds essential to cell functioning may occur. Under alkaline conditions the binding of heavy metals to cell membranes may cause death or injury by inhibiting the transfer of metabolites.

Table 2.2 pH range over which faecal coliform die-off in water is at a minimum (pHkmin). 17

Modelling Faecal Coliform Dynamics pHkmin 5.0 < pH > 6.4 5.5 < pH > 7.5

2. Literature Review Reference Cohen (1922) McFeters and Stuart (1972)

The relationship between pH and faecal coliform die-off indicated by the data of McFeters and Stuart (1972), where die-off is a minimum at neutral pH increases under acid and alkali pH, can be represented by a hyperbolic cosine. This relationship is derived in Chapter 6, which also provides an indication of the significance of pH on faecal coliform die-off in a UK river.

Dissolved oxygen content The dissolved oxygen (DO) status of an aquatic system is a function of water quality, cycles in DO are associated with solar energy and photosynthesis by primary producers on both diurnal and annual time-scales. Zerfas (1970) showed that faecal coliform die-off was more rapid in well aerated water compared to water flushed with nitrogen gas; in order to deplete the water of oxygen (see Figure 2.6). Verstraete and Voets (1972) found die-off to increase with dissolved oxygen content in sewage contaminated lake waters. Other authors have proposed that extended faecal coliform survival in ice-covered waters results from oxygen depletion (Hirn et al., 1979; Gordon, 1972; Davenport et al., 1976).

Figure 2.6 Survival of faecal coliform in raw river water at 20ºC showing a rapid decline in numbers in the aerated flask, little change in numbers in the static flask and N2 flushed flasks (after Zerfas, 1970).

18



These observations are in agreement with the comments about the effects of solar radiation, that the cleaner a water body is initially, the greater the ability to self-purify. As DO is both readily measurable and long-term data on DO levels are available for many rivers in the UK. Thus if a river is known to have low DO status, it would be appropriate to qualitatively assign a lower die-off rate to that stretch of river for the purposes of modelling. Unfortunately there is currently insufficient data to quantify such a relationship.

Particulates, settlement and other effects The following section considers how faecal coliform organisms become attached to particles, how the particles shield the organisms from sunlight and predators, and enhance settlement to the riverbed into pockets of stored organisms. These processes not only influence faecal coliform survival but are important in determining faecal coliform dynamics in fluvial systems (Jenkins, 1984); the pockets of stored bacteria can be resuspended into the flowing water above (see Section 2.2.2). Certain subtle and complex processes relating to nutrients are also indicated.

The impact of particulates and the supply of nutrients on faecal coliform survival are complex (Milne et al., 1991). Suspended particles reduce light penetration into the water column (Pommepuy et al., 1992). Fine particles coat the organisms shielding them from light and predators (Roper and Mitchell, 1978). Attachment to particles aids settlement (Milne et al., 1986; Gannon et al., 1983), where conditions on the river bed favour prolonged survival relative to those in the water column (see Hendricks and Morrison, 1967; Verstraete and Voets, 1972). Channel-bed storage represents a major source of organisms (Van Donsel and Geldreich, 1971) capable of lowering the quality of the overlying water when suitably disturbed (Gary and Adams, 1985; Jenkins 1984; Matson et al., 1978; Grimes, 1975). Attempts to quantify this survival have produced wide variations in die-off rate, a possible cause of this uncertainty is the inability to create realistic conditions in the laboratory (e.g. Sherer et al., 1992; Burton et al., 1987).

19



Attachment to particles and surfaces Faecal coliforms can become attached to any solid surface such as sediment particles, rocks, plant and animal surfaces, organic matter, and to each other (Marshall, 1979). They may also become coated by clay particles (Roper and Mitchell, 1978). A description of the mechanisms underlying this attachment is given below.

Solid objects generally have net positive surface charge that attracts negative ions (cations) in solution. These cancel-out the positive charge sites forming a layer of strongly adsorbed cations a few nanometres thick known as the Stern-layer. Beyond this layer the electrostatic forces decay almost exponentially, as does the difference in numbers of cations to anions (Figure 2.7), this zone is the diffuse layer (Marshall, 1979; White, 1979). Solid objects also exhibit non-electrostatic surface forces, these are Van der Waals forces which have a weak attractive effect. Because of the high surface density of cations the charge of surfaces in water is negative. As like charges repel, the net electrostatic force acting between two surfaces is one of repulsion; they should push each other apart. However, in most natural waters the Van der Waals attraction force exceeds the electrostatic repulsion and the surfaces stick together.

Figure 2.7 (a). Representation of the charge distribution between an E.coli cell and a solid surface. (b). Graph showing the net attraction energy between surfaces resulting from attractive Van der Waals and repulsive electrostatic forces as a function of separation distance in high and low conductivity (salt) solutions (after Marshall 1979).

20



In waters of very low conductivity, the scarcity of ions in solution means that the diffuse layer extends beyond the reach of the Van der Waals forces, thus nearby surfaces will repel (Figure 2.7b). In most fresh-waters the conductivity is sufficiently high that the diffuse layer does not extend beyond the influence of the Van der Waals attraction force. The particles or surfaces come to rest at a distance where the attraction and repulsion forces are in equilibrium. By this mechanism faecal coliforms, and other organisms, become loosely attached to each other, to stationary surfaces in the stream and to suspended particles. The organisms may be released from their attachment by rising flow associated with storm events (Kunkle, 1970), or other disturbances such as a person paddling (Sherer et al., 1988), the passage of a vessel (Pettibone et al., 1996) or dredging operations (Grimes, 1974).

Adsorption capacity varies depending on the surface area to volume ratio of a material, the greater the surface area to volume ratio the greater the number of charge sites, weight for weight. Clay minerals have excellent adsorptive properties, silts, fine sands and coarser sands have successively lower adsorptive capacity (Marshall, 1979). In sandy sediments the actual particle size distribution is important. Unlike clays, which have internal charge surfaces available for adsorption, only the outer surface of a sand grain is available for adsorption. Thus, the smaller the mean particle size of sandy sediments the greater the availability of adsorption sites in a given volume of sediment. The degree of adsorption also varies between species of bacteria (White, 1979) and the degree of saturation of adsorption sites. The likelihood of organisms becoming attached increases in waters of pH above 8.0, with the addition of calcium, magnesium, and, in soils, with decreasing soil moisture (Bitton, 1980).

Particles, nutrients and faecal coliform relationships in natural waters Faecal coliforms can enter a river system either as freely suspended organisms or attached to particles. If sufficient particles are present in the water, around a fifth of the freely suspended organisms are likely to attach rapidly (Milne et al., 1986). Attachment to particles will have a major impact on the survival of the organisms. The size of the particles will determine how readily they settle to the riverbed. Of the organisms that become attached to particles the majority are likely to be associated with particles of less than 30 m diameter. Jenkins (1984) found that 60% of attached organisms were associated with solids of settling velocities equivalent to mineral grains of diameter 30 m or less. Other studies, however, found that a greater proportion (90%) of the organisms were associated with even 21



smaller particles, of the order of 0.45-10m diameter (Gannon et al., 1983; Auer and Niehaus, 1993). This difference between the respective studies may be due to the nature of the particles, i.e. the relative distributions of particle sizes. Finer particles, in particular clays, have a greater surface area to volume ratio than larger particles such as sand grains, and have a greater capacity to attach organisms. Thus, the greater the proportion of finer particles available, the greater the proportion of organisms that will attach to that fraction.

Small particles have been shown, by electron microscopy, to form a coating around E.coli cells capable of protecting the organisms from the parasitic attack of bdellovibrio (Roper and Mitchell, 1978). Particles also protect faecal coliforms from the suns harmful rays (Section 2.1.2). This may be achieved through a combination of the particle coating effect, mentioned above, and the reduction of light penetration into the water column (Pommepuy et al., 1992).

The relative sizes of particles to organisms may be significant in determining both transport and survival dynamics. Very small particles such as clays form a coating around larger organisms (Roper and Mitchell, 1978) affording protection and increasing the likelihood of settlement. If evenly coated with similar size particles it is possible that the coating will not be easily detached from the organism. Where particles are of the same order of size as the organism, the protection afforded by the particle may be minimal and the likelihood of detachment higher. If the particles are larger than the organism, the particle might develop a coating of organisms and hence the organisms adopt the settlement characteristics of that particle. In this case the protection afforded by the particle may be less and the organisms may be easily detached in turbulent flows.

Particles, therefore, play an important role in the survival of faecal coliforms in the water column. In the absence of sediment particles E.coli in sea water have been found to aggregate, or cluster; the organisms at the centre of the cluster being protected from the osmotic stress of the sea water (Findlay et al., 1990). Findlay et al. (1990) described this clustering as a "self-protection" mechanism. The clustering may, however, be a side effect of being in sea water; high ionic strength solutions, such as sea water, favour the binding of particles (Marshall, 1978). Hence organisms that drift together in sea water may not be able to avoid sticking together. Thus, the clustering effect is a fortunate phenomenon for faecal coliforms in sea water, and the term "self-protection" is perhaps a misnomer in this case. 22



Some fresh-waters do have high conductivity (Robson et al., 1996) and it is possible that some degree of protection by aggregation occurs in such waters. This mechanism may be limited to relatively quiescent waters where suspended particles are absent and the organisms are not "pulled-apart" by the turbulent shear forces encountered in faster flowing water.

The relationship between faecal coliform survival and the supply of nutrients is similarly complicated. In the absence natural biota, the addition of nutrients to samples has been shown to prolong E.coli survival (Hendricks and Morrison, 1967; Evison, 1989). Brettar and Höfle (1992) also observed extended E.coli survival in lake "meso-cosms" dosed with growth media. The meso-cosms were in the form of large membrane bags containing natural lake water dosed with E.coli. In this case the extended survival was due to the protection afforded by algal particles, the growth of which was stimulated by the nutrient additions, unlike the direct metabolic effect observed by Hendricks and Morrison (1967) and Evison (1989).

Table 2.3 E.coli die-off rate at different nutrient concentrations. The tests used filter sterilised water from the River Coquet dosed with different proportions of sterile sewage effluent (after Evison, 1989). Nutrient supply

0.025

0.25

2.5

25

0.136

0.316

0.279

0.043

(% Sterile sewage concentration)

Die-off rate (d-1)

An abundant supply of nutrients may, therefore, enhance faecal coliform survival. Table 2.3 reproduces results that show extended E.coli survival at both high and low nutrient concentrations compared to intermediate nutrient levels (see Evison, 1989), i.e. faecal coliforms can survive for extended periods under conditions of nutrient starvation as well as excess. Indeed, E.coli cells are able to enter a state of dormancy under conditions of nutrient starvation and although not detectable by standard culture enumeration methods, the organisms remain viable and capable of returning to a culturable state when conditions are appropriate (Brettar and Höfle, 1992; Roszak and Colwell, 1987).

Faecal coliform survival under different conditions of nutrient availability is complicated by the presence of particles. Milne et al. (1991) found that E.coli survival in sea water was enhanced by either the provision of adequate nutrients or suspended solids, but not both. 23



The same initial nutrient concentration was used in these experiments and the suspended sediment concentration varied. The sea water was sterile so no indication is available about whether the observed effects would be repeated in the presence of the full natural biota. E.coli die-off was slowest at low suspended sediment (SS) concentrations. This was thought to be due to the ability of the particles to tightly adsorb nutrients. Hence at low SS concentrations the nutrients are available in the water to be metabolised, reducing the rate of die-off. As the SS concentration rises the nutrients become bound to the particles and as a result unavailable to the organisms which starve and die-off more rapidly. With further additions of SS, however, there are sufficient particles to protect the faecal coliforms which now die-off less rapidly. Parallel tests carried out at 20ºC and 5ºC showed that at low suspended sediment concentrations and low temperatures faecal coliform die-off was actually accelerated. This specific mechanism produces a result that is contrary to the general observation that suspended solids reduce coliform die-off (e.g. Pommepuy et al. 1992; Roper and Mitchell, 1978). At 20ºC maximum die-off was observed with 12.5 mg SS per litre, at 5ºC peak die-off occurred with 5 mg SS per litre. This result may have been caused by reduced bacterial metabolism at the lower temperature, i.e. the organisms were less able to utilise the available nutrients and hence died more rapidly than at the higher temperature (Milne et al., 1991). While interesting results in themselves and insightful into the smaller detail of the interactions between faecal coliforms and particles, it is likely that these small effects would not contribute significantly to the overall pattern of die-off observed in the field. The main factors would be the simple mechanisms such as the shielding from sunlight penetrating the water afforded by the presence of particles (Section 2.1.2).

Settlement and other survival enhancing effects. The presence and attachment of faecal coliforms to particulates has the potential to significantly accelerate the settlement of the organisms to the riverbed. In the absence of particulates or settlable particles E.coli cells will remain in suspension and are unlikely to settle to the riverbed and as a result be more susceptible to die-off effects. Milne et al. (1986) found that freely floating E.coli cells attach rapidly to available particles and other studies have shown that the majority are likely to be associated with particles of less than 30 m diameter (Gannon et al., 1983; Jenkins, 1984; Auer and Niehaus, 1993). Milne et al. (1986) found that in the presence of mud and silts in estuary water E.coli deposition was a function of time and the deposition rate was directly proportional to suspended solids (SS) 24



concentration. Jenkins (1984) showed that up to 75% of E.coli settled out within the first few minutes of experiments in still water using natural sediment from the River Washburn in Yorkshire. Auer and Niehaus (1993) found settling rates in Onondaga Lake, New York of around 1.2 m.day-1 for faecal coliforms associated with particles of between 0.45-10 m in diameter. In streams and rivers such low settlement rates might only be expected at quiescent margins or in regulated stretches where a high water level is maintained at low flow for the purposes of navigation.

Biskie et al. (1988) carried out slurry injection into Bear Creek a small slowly flowing stream in Central Oregon. The slurry was composed of 5% fresh cattle faeces mixed with stream water. After injection into the stream a portion of the organisms were immediately washed free from the manure particles and travelled downstream with the flow as a solute. The bulk of the organisms (95%) were found to have settled within 50 m of the injection point, at the stream bottom they either died or were released gradually over a period of hours following the injection. The bulk injection of slurry exceeded the carrying capacity of the small stream demonstrating the rapid settlement of the contaminant organisms and associated particles.

In the absence of particles Milne et al. (1986) found that there was no detectable faecal coliform settlement from sewage final effluent, even after 3 hours. With mixtures of effluent and estuarine water deposition did occur but was not a function of time and the deposition rate exhibited no correlation with SS concentration. The E.coli cells were found to be equally likely to attach to non-settlable effluent particles or estuarine silts and mud and the difference in deposition was attributed to the different settling characteristics of the particles.

Survival on the river-bed On the riverbed, faecal coliforms accumulate in a thin layer a few millimetres thick (Jenkins, 1984). Stephenson and Rychert (1982) found stream bed concentrations of up to 760 times the numbers in the overlying water, in other studies this figure was 2500 (Matson et al., 1978). Conditions favour extended survival as a result of improved nutrient availability (Hendricks and Morrison, 1967), the reduction of light penetration and stabletemperature effects (Verstraete and Voets, 1972). Verstraete and Voets (1972) showed that die-off was less variable at the mud-water interface than in the overlying water. During the 25



summer months die-off at the mud-water interface was half that observed in the overlying water, suggesting that sunlight was not accelerating die-off at depth, the long-term variation at the mud-water interface followed the temperature of the overlying water.

Investigations of faecal coliform survival in sediments of differing mineral composition show no consistent pattern, highlighting the difficulties in attempting to assess faecal coliform die-off in sediments (e.g. Sherer et al., 1992; Burton et al., 1987; Allen, 1953) (Table 2.4).

The main limitation of studies examining faecal coliform survival in sediments has been the difficulty of reproducing representative conditions. Hendricks and Morrison (1967) observed the growth of enteric bacterial numbers in sterile nutrient extracts from stream bottom sediments at temperatures above 10ºC, this led them to suggest that extensive growth may occur in bottom sediments. The data presented in Section 2.1.4 show that in die-off experiments with raw natural samples growth does not occur, in fact die-off is rapid, growth is only observed in tests where naturally occurring predators are absent.

Table 2.4 Faecal coliform die-off in different sediments (recalculated from data of Burton et al., 1987; *Sherer et al., 1992). Clay/silt/sand ratio

% Organic matter content

Total Kjeldahl Nitrogen (ppm)

Total Phosphorous (ppm)

E.coli die-off rate k (d-1) 1

75 : 25 : 0

14.8

3.18

9.98

0.126

28 : 55 : 18

6.2

16.32

6.90

0.228

25 : 51 : 24

5.2

13.97

14.30

0.236

12 : 76 : 11

9.0

24.10

7.70

0.150

2 : 0 : 98

0.7

-

-

0.207

34 : 34 : 32

-

-

-

*0.0159

14 : 12 : 74

-

-

-

*0.0189

1

Value as quoted in source text (suspected to be in error). Other values recalculated from graphical data show values in source text to be widely in error. Units of values in source text also quoted incorrectly.

The die-off rates in the two studies presented in Table 2.4 are a factor of ten different. This highlights the difficulty of representing field conditions in the laboratory. Sherer et al. (1992) mixed cow manure with their sediment samples, thus introducing the liquor and organic matter from the manure to the sediment, as well as distributing the faecal organisms 26



evenly throughout the test samples. Biskie et al. (1988) showed that 95% of organisms injected, with manure slurry, into a creek settled within 50 m downstream of the injection point. It is likely, however, that the components would separate, the liquid would be diluted into the flow and the solids would disperse and perhaps settle to the surface of existing deposits. The experiments were undertaken in darkness at 8ºC, and given the nutrient rich environment provided by the manure and absence indigenous benthic organisms, the die-off rates observed were low (Table 2.4) (Biskie et al., 1988).

The results presented by Burton et al. (1987) are problematic in that there are inconsistencies between the tabulated results, graphical representations of raw data and units used for bacterial survival. The study also attempted to mimic the physical and chemical conditions in different freshwater sediments using a continuous flow laboratory microcosm. The requirements for simulating field conditions adequately were, however, only partly met. The water used to circulate over the sediment samples, although reconstituted to the major ion concentrations observed in the field, was sterile. The sediment samples were stored at 4ºC for up to two weeks before use, the test organisms were initially distributed evenly throughout the sediment samples and no reference was made to the lighting conditions used during the tests. The quoted rates of die-off were far in excess of any values presented in this literature review for die-off in water. Die-off rates recalculated from data presented graphically gave more realistic results indicating that the values quoted in the source material were quoted per hour and should have been per day. Once this was established the quoted die-off slopes were superimposed on the graphs presented and found not to coincide with the graphical data. Burton et al. (1987) assert that there were no statistically significant differences between replicate tests, if this is true then the recalculated values presented in Table 2.4 are believed to give a better representation of their test results. Even when recalculated, however, the die-off rates were a factor of ten faster than those observed by Sherer et al. (1992) and still within the range that might be expected in water.

Given that these two sets of results represent two extreme cases it would seem appropriate to chose a die-off value somewhere between the two as a starting point when setting parameters for modelling faecal coliform stream dynamics.

2.1.4 Microbial interactions 27



A wide range of processes affecting the survival of faecal coliforms have been examined these include primary environmental variables and characteristics of water and soil quality. This section introduces studies that have investigated microbial interactions that influence faecal coliform numbers in natural waters. Natural raw water samples were used in these studies; the aim being to highlight die-off behaviour that might be expected under natural microbial conditions as opposed to studies which were interested in E.coli behaviour in isolation, i.e. under sterile conditions. The following section presents the types of organism known to prey on, or compete with, faecal coliform. The methods used to isolate the effects of the different groups of predating organisms are described. Finally the findings of various studies are presented, these indicate that the organism causing the greatest proportion of faecal coliform die-off differs according to the type of aquatic system being studied, i.e. clean freshwaters, sewage contaminated waters and estuarine waters. The results also highlight the rapid rates of die-off in raw samples compared to those that have had the naturally occurring microbes removed.

Microbial predators and competitors of faecal coliforms In natural waters faecal coliforms find themselves in a hostile environment where, without fresh inputs from external sources, their numbers will decline rapidly. This decline is accelerated by the presence of other organisms that compete for nutrients and adsorption sites, graze directly upon and are parasitic towards faecal coliforms (Verstraete and Voets, 1972; Roper and Marshall, 1978). The main groups of these organisms are summarised in Table 2.5.

The grazers are at the larger end of the size spectrum of hostile organisms and include multicelled invertebrates such as protozoa and rotifera. Rotifera have thread like limbs (cillia) which beat rhythmically to produce motion, meaning that they can move freely in search for food, rather than waiting for it to come to them. Rotifera might be expected in the upper layers of lake waters (Singleton and Sainsbury, 1981). However, this does not mean that they would not be present in, for example, the still waters of a regulated lowland river; an environment not dissimilar to that of a shallow lake. Certain protozoa are capable of motion,

Table 2.5 Hostile organisms that contribute to faecal coliform die-off. Type of organism

Hostile effect

Protozoans and rotifera

Grazing 28



Native bacteria

Competition for nutrients

Bdellovibrios, bacteriophage

Parasitism

they have either cillia or a flagellum (a long tail), and actively graze on bacteria (Singleton and Sainsbury, 1981; Brettar and Höfle, 1992). Competitor organisms include naturally occurring bacteria, and any other organisms competing for the same nutrient supply as faecal coliforms. Parasites of faecal coliforms include bdellovibrio (parasitic bacteria) and bacteriophage (parasitic viruses) referred to as coliphage when specific to coliforms, these organisms infest and live within the host until they multiply and hatch-out rupturing the hosts cell walls and thus killing it in the process (Singleton and Sainsbury, 1981). The following sub-section introduces the types of experiments researchers have used to investigate the proportion of faecal coliform die-off caused by the different hostile organisms.

Removal experiments In sterile water samples, faecal coliform populations remain unchanged for periods far in excess of those for samples containing natural microbes (Flint, 1987; Enzinger and Cooper, 1973; Verstraete and Voets, 1972). In order to investigate the proportion of the observed die-off caused by the different groups of hostile organisms, microbiologists undertake “removal” experiments. By removing groups of organisms from samples and comparing the rates of die-off between treatments the impact of the different groups can be examined. Methods of removal include filtration, dosing with antibiotics and autoclaving. Autoclaving is a very effective method of high temperature sterilisation used by microbiologists. Removal by filtration involves samples of raw water undergoing a series of filtrations at successively smaller pore sizes. At each filtration the range of organisms in the filtrate is reduced (Table 2.5). The filtrates are dosed with a predetermined number of laboratory cultured E.coli cells and held under fixed temperature and light conditions, sub-samples are extracted at regular intervals for enumeration to examine the rates of decline in E.coli numbers.

To investigate the influence of algae and protozoa in the absence of naturally occurring bacteria, samples are dosed with an anti-biotic. Anti-biotic resistant E.coli strains are used as

29



the contaminant organism. Die-off in autoclaved samples is assessed in order to provide the sterile sample case against which the other results can be compared.

The data presented below are from studies that have investigated the relative impact of different micro-organism groups on faecal coliform survival.

Table 2.6 Organisms removed from natural water samples by successive filtration with decreasing filter pore sizes. Size of filter pore (m)

Type of organism removed

5

Algae and protozoa

3

Protozoans

1.2

Micro-cilliates and micro-flagellates

0.8

Bacteria

0.45

Bacteria/phage

0.22

Bdellovibrio (coliphage may remain)

The influence of predation on faecal coliform die-off The studies presented below demonstrate the differing impact that the groups of microorganisms present in natural waters have on faecal coliform die-off, but more importantly they highlight the significance of microbial effects in the overall die-off of faecal coliforms in natural systems.

Verstraete and Voets (1972) and Flint (1987) found that the removal of indigenous bacteria produced the greatest reduction in die-off in relatively unpolluted lake and river waters, respectively (Figures 2.8, 2.9 and Table 2.7). Figure 2.8 shows the initial E.coli concentration of each experimental treatment, the rapid die-off in raw samples and prolonged survival with subsequent removals. The sterile samples show no discernible reduction in E.coli numbers over the approximately 15-day period for which data is presented.

30



Figure 2.8 The influence of successive removal of sections of natural microbial communities on E.coli die-off in (a) "clean" river water, (b) sewage contaminated river water (after Flint 1987) and (c) estuarine water (after Enzinger and Cooper 1976). Table 2.7 Die-off rate k (d-1) (and percentage reduction in die-off rate with each removal) in clean and sewage contaminated river water following serial removal of natural biota by filtration, temperature = 15ºC, (source data, Flint, 1987). River water sample

raw (unfiltered) water

particles, protozoa, algae removed

bacteria removed

autoclaved (sterile)

Clean river water Sewage impacted

0.47 (0%)

0.44 (7%)

0.0286 (87%)

0.0 (6%)

0.25 (0%)

0.2 (20%)

0.105 (38%)

0.0 (42%)

In clean lake and river water the removal of protozoa and particulates (including algae) had little effect on die-off rates. In sewage contaminated water, however, the removal of protozoa and algal particles had a greater affect, accounting for around 15-20% of die-off. 31



Bacteria made no major contribution to die-off. In sewage contaminated water bacteriophage and a bactericidal agent caused the most die-off. The bactericidal agent was of uncertain origin and referred to as a highly motile 'ultramicro-bacteria' (Flint, 1987) and a 'thermo-labile toxin' a toxin destroyed by heating (Verstraete and Voets, 1972). Conversely, in estuarine waters Enzinger and Cooper (1976) found the key die-off-promoting organism to be protozoa. In a more recent study Brettar and Höfle (1992) examined the development of successive groups of predator organisms that developed in response to an injection of faecal coliforms into eutrophic lake water. Infection of E.coli by bacteriophage was found to be of no relevance to overall die-off. Predator populations grew and declined in response to the size of 'food' available to graze upon. In the first few days of the experiments the majority of E.coli cells were free in suspension (i.e. not attached to particles) and grazing by flagellates was the dominant cause of die-off. Further reductions in free-floating E.coli resulted from their attachment to particles. In response, the numbers of flagellates declined and larger multi-cellular rotifers, K.cochlearis and Conchillus Unicornis, became the dominant grazer population. These organisms are capable of grazing on particles ranging from the size of faecal coliform cells (0.8 to 1.2 m) up to 20 m (Brettar and Höfle, 1992). After ten days the remaining E.coli (now only existing where attached to particles) had been grazed to undetectable levels by the macro-zooplankton Daphnia. Daphnia are able to feed upon particles of up to 35 m. This not only demonstrates the succession of predators that develops in response to an input of contaminant organisms, it indicates the changing nature of the contaminant population with time. The E.coli cells were initially floating freely within the water mass but with time and predation the only surviving cells were found attached to particles. This attachment increases the possibility that such organisms will survive sufficiently long to settle to the channel-bed where conditions are favourable to prolonged survival (see Section 2.1.3) and, perhaps, eventual disturbance and resuspension by the rising flow caused by a storm event.

Verstraete and Voets (1972) demonstrated the seasonal variation in microbial activity (Figure 2.9). The fastest die-off rates were found in samples collected during the summer when microbial activity is at its greatest. Die-off was slowest in samples collected during winter due to reduced microbial metabolism as a result of lower temperatures and reduced sunlight (and hence primary productivity). There is less energy available and this slows 32



down the whole food web, reducing the numbers of predatory organisms and the rate at which they can feed on the introduced faecal coliforms.

Figure 2.9 Seasonal variation in die-off of E.coli in (a). sewage contaminated and (b). clean water bodies, showing the effect of successive removals of groups of natural biota. Note that raw samples exhibit the most rapid die-off and that die-off is at its greatest during the summer months (after Verstraete and Voets, 1972).

In summary, naturally occurring organisms accelerate faecal coliform die-off. Die-off in the presence of these organisms is much more rapid than when they are absent. The implication of this is twofold. Firstly, it highlights the importance of using raw waters to gain a realistic indication of die-off under natural conditions. Secondly it raises questions about the nature of the contamination, i.e. whether it is continuous in the form of an effluent discharge, or transient, for example, a spill of farm animal slurry. In the case of a continuous discharge into a stream or river, it might be expected that the microbial system would be in some kind of quasi-equilibrium. That is to say that the size of the natural predator population would be appropriate to the numbers of contaminant organisms received, and hence rapid die-off could be expected. In the event of a major transient pollution incident the numbers of contaminant organisms would swamp the natural system, there would be insufficient predators to make an impact on the contaminant population. This would be analogous to the sterile sample case and die-off would be less rapid. This would only be the case initially as natural mixing processes would tend to dilute and disperse the contaminants and hence reduce their numbers with respect to the numbers of predators, the predator population 33



would also be expected to grow in response to the abundance of food. Clearly this alone is a complex area worthy of further consideration, especially in terms of the trophic status of a particular stream or river; nutrient poor (oligotrophic) systems would be expected to suffer far more severely than less impoverished systems with an already well developed microfauna. Conversely, in eutrophic systems or systems already heavily contaminated prolonged survival would be expected.

2.1.5 Summary of die-off impacts Table 2.8 summarises the main causative effects of changes in environmental variables and water quality conditions on the survival of faecal coliforms, as well as, the time-scales over which the largest variations in these variables occur.

Section 2.3 introduces models for describing faecal coliform die-off. The theme of faecal coliform die-off modelling is then developed in Chapter 6 with comparisons of the relative significance of the different die-off inducing factors and the development of equations and example applications of the equations to a UK river. The following section highlights sources of faecal contamination, the pathways by which coliforms enter a stream system from the catchment surface and the transport contaminant organisms once in the river system.

Table 2.8 Summary of the main causative effects of environmental variables and factors on the survival of faecal coliforms. Variable

Main effect(s)

Time scale(s)

Solar radiation

Die-off proportional to solar radiation

Seasonal, diurnal, hourly

Temperature

Die-off increases exponentially with

Seasonal, diurnal

temperature Soil moisture

Drying conditions accelerate die-off

Seasonal with daily effects

pH

Die-off accelerated either side of neutral

Fluvially and photosynthetically driven

pH

effects

Dissolved oxygen

Die-off faster in well oxygenated waters

Diurnal, seasonal

Particulates

Reduces die-off due to sunlight

Episodic; fluvially driven

Predators

Compete with, prey-on and parasitise

Short term continuous internal process

contaminant organisms

34



2.2 Sources and transport of faecal coliforms There are a wide variety of sources of faecal coliforms in catchments. Wherever there are accumulations of warm-blooded animals there will be large numbers of faecal coliforms released into the environment, either directly by defecation or by the disposal of faecal waste to land or to rivers. If the organisms present in this waste survive they will ultimately be transported into nature's drainage system.

In this section typical sources of faecal coliforms in catchments are highlighted with particular attention to contamination from agricultural land. This leads into a discussion of the hydrological processes that facilitate the transport of the organisms from the land surface into the river system. Current understanding of fluvial transport processes is introduced.

The summary of faecal coliform sources has been included in order to give an indication of the variety of sources and typical concentrations of organisms that might be expected either directly or by run-off. The focus is then changed to highlight faecal contamination from agricultural land-uses. The aim being, to concentrate on poorly defined (Jenkins, 1984) natural hydrologically driven diffuse processes as opposed to direct piped discharges to rivers.

2.2.1 Sources of faecal contamination The sources of faecal contamination of surface waters are many and varied and the concentration of coliforms in any given stretch of water may comprise organisms from a variety of sources. There are "point" sources, such as discharges of domestic and farm effluents, combined sewer overflows. These inputs can be quantified by measuring the flow and concentration over a range of conditions. Diffuse, or "non-point" sources are the other inputs that cannot be defined as coming from a specific point. For the purposes of this study, diffuse sources are run-off from the catchment surface. Diffuse inputs are more difficult to quantify than point sources because of difficulties in sampling multiple transport pathways and other factors that influence the passage of the organisms into channels. The simplest way to characterise coliform delivery from diffuse sources is to monitor the catchment stream outlet of a given area, this provides an integrated signal from that area but little 36



additional information with which specific transport pathways and mechanisms may be examined.

Typical sources of faecal contamination and their concentrations Table 2.9 summarises typical sources of faecal coliforms giving concentrations of organisms per gram of waste or per 100 ml of liquid waste. The main sources of contamination are

Table 2.9 Sources of faecal contamination, with typical faecal coliform concentrations. Source of Faecal contamination

Nature of Input

1. Non-point sources,

Diffuse inputs to the catchment surface, transported to surface waters by hydrological processes.

Natural Inputs

Faeces of indigenous fauna e.g. rodents (FC per gram) [Geldreich and Kenner, 1969]. Inputs from nesting or roosting bird colonies (FC per gram faeces) [Standridge, 1979].

2.105

Agricultural; use of farm animal wastes as fertilisers on arable and pastureland, dirty water reuse schemes for irrigation/disposal, intensive stock grazing.

109 per 100ml [Evans and Owens, 1972]

Sewage sludge disposal; application to arable land and forestry as fertiliser. Recreational; directly from contact water sports, contamination of rivers draining catchments supporting enhanced recreational activity [Geldreich and Kenner, 1969].

107 per g human faeces

Run-off from agricultural land with artificial drainage [Evans and Owens, 1972].

10-105 per 100ml

Avian Enhanced

Typical values, all in FC

3.107-4.108

2. Point sources

Localised, easily quantifiable sources, such as discharges piped directly into a receiving-water.

Sewage treatment

Piped discharges of treated domestic (examples, FC/100ml) and industrial effluent [Cohen and Shuval, 1973; Menon, 1985].

Urban runoff, storm drainage

Combined sewer overflows(1), storm drains(2), contamination from urban runoff(3), street gutters(4). Storm induced inputs may result in extremely high "first-flush" concentrations due to scour of material in pipes. Storm sewerage may operate before sufficient dilution by receiving water can be achieved. (All values per 100ml) [(1&2)Burm and Vaughn, 1966 (in (3&4)Geldreich et al.,1968)].

Industrial effluents

Discharges from food processing industries; poultry/meat/potato plants, fruit/vegetable canneries etc [Menon, 1985].

102-107 per 100ml

Landfill sites

Leachates contain faecal coliforms, organisms can survive within sites for long periods [Blannon and Peterson, 1974].

106 per 100ml (In waste 105-107 per g)

10-106 final 108 raw (1)

107 106 (3) 3 10 -5.104 (4) 2 10 -104

(2)

37



sewage effluents and waste from agriculture. Other less obvious sources of faecal contamination are food processing waste waters; bacterial levels in vegetable processing wastes are attributed to the soils in which they were grown, the use of animal fertilisers, farm animals and poor quality irrigation water. In meat processing wastes, straw, blood, flesh, fat, offal and manure from the intestines of the slaughtered animals result in similar bacterial concentrations to the faeces of farm animals (Geldreich and Kenner, 1969). Standridge et al. (1979) found that outbreaks of various diseases on a lakeside recreational beach resulted from faecal contamination by a permanent water fowl population on the shore. Similarly, gull populations have the potential to lower the quality of night-roost water bodies as well as streams receiving run-off from nesting sites. Gulls may feed on agricultural land, at landfills and sewage works and breed on moorland, often using reservoirs for safe night-roosts. One herring gull may excrete around 1.8 billion faecal coliforms in a day; 1.9 billion is typical for a human (Gould and Fletcher, 1978). Landfill leachates contain high concentrations of indicator bacteria (Donnelly et al., 1981). Niemi (1985) examined the potential of fish farms as a source of faecal contamination but the majority of faecal coliforms were traced to run-off from bird droppings on surrounding land. Fish are cold-blooded, they do not have a permanent faecal coliform flora in their intestines and in remote regions their faeces rarely contain faecal coliforms (Geldreich 1970).

2.2.2 Faecal contamination from diffuse sources On the land-surface, faecal coliforms may remain locked-up in parent faecal material for extended periods. Evans and Owens (1972) found that organisms continued to be released for up to 4 months after being deposited on the catchment. An examination of run-off from "standard cow pies" has shown faecal coliform release declines exponentially, once dry on the surface, rainfall intensity becomes an important determinant of the run-off quality, with lower rainfall rates producing greater concentrations of organisms (Kress and Gifford, 1984). Once released they can become adsorbed to soil and organic matter and survive for long enough to provide a semi-permanent reservoir capable of contaminating the surrounding aquatic environment following transport by hydrological processes (Hunter and McDonald, 1991a). The supply of organisms from diffuse sources is determined by landuse, applications of waste, and livestock densities (Kay and Stoner, 1989; Owens and Evans, 1972; Faust, 1977). Run-off from grazed areas generally has higher bacterial 38



concentrations than un-grazed areas (e.g. Van Donsel et al., 1967; Meiman and Kunkle, 1967; Wyer et al., 1994), in some cases up to five or ten times greater (Doran and Linn, 1979).

Faust and Goff (1977) found faecal coliform concentrations to correlate with flow, stream length (r2=0.799) and basin size (r2=0.9), 68% of faecal coliforms discharged were from pasture, 17% from forest and 15% from cultivated land. The annual rates of faecal coliform discharge from the different land-uses were, for pasture 89.107 FC/Ha/yr, forest 22.107 FC/Ha/yr, and cultivated land 19.8.107 FC/Ha/yr. The faecal coliform discharge per basin unit area was found to be similar on a site to site basis and no significant difference was found on a season by season basis. The daily rate of discharge from a population density of 0.6 animals per hectare was between 7.5.106 FC/Ha/day and 669.106 FC/Ha/day (Faust, 1976). Robbins, Howells and Kriz (1972) estimated that a beef population of 5.7 animals per hectare produced 1450.106 FC/Ha/day compared to 725. 106 FC/Ha/day from land devoid of farm animals. Faust and Goff (1977) showed that of the total numbers of organisms discharged per hectare per month, 68% came from pasture, which comprised only 18% of the total study area. However, on average less than 1% of the organisms produced by grazing livestock were found to be washed from the catchment surface, travelling no further than 6.7 km in the streams.

These studies that have attempted to quantify faecal coliform run-off from diffuse sources indicate that agricultural land can deliver large numbers of organisms to the stream network, the following section investigates the soil and hydrological processes that contribute to this run-off.

2.2.3 Transport into the river system An examination of the soil and run-off mechanisms that determine the quality of run-off from diffuse sources and is necessary to understand how run-off contributes to stream coliform dynamics. The combination and magnitude of inputs, outputs and die-off effects will determine the observed behaviour in any given river reach. For example, in lowland areas, transport by water infiltrating the soil mass may account for the majority of organisms 39



reaching surface waters from non-point sources. Such inputs, at any bank-side location, may be small compared with inputs from single point sources, but the total contribution to the stream coliform loading may be significant. In upland areas the majority of organisms are transported by surface runoff and non-matrix soil through-flow during wet weather. Under base-flow conditions inputs are provided by near channel return flows, as well as soil matrix through-flow. The relative importance of these pathways depends on the nature of the catchment, the antecedent soil moisture status and the occurrence of rainfall, its duration and intensity.

Mechanisms of transport through the soil system and run-off processes The main pathways of faecal coliform transport from diffuse sources to the stream network are through the soil system and in surface run-off. In the soil system the pathways for faecal coliform transport can be separated into two routes. The first is via the inter-granular pore spaces of soil blocks (matrix through-flow) and the second is via macropores. These are, as the name suggests, large voids within the soil system, cracks and root-pipes; they bypass the soil-matrix. The flow through these voids is variously referred to as bypass flow, non-matrix through-flow and macropore-flow (Germann et al., 1987; Hunter et al., 1992). Appendix III provides example travel distances of bacteria in porous media.

Matrix through-flow input occurs at or below the stream surface and accounts for a large proportion of the water input to a channel, but as a result of filtering by the soil matrix may only contribute relatively small numbers of coliforms (Hunter et al., 1992, Matthess et al., 1988). Processes, such as filtration, which retain bacteria within the soil matrix are known as "deposition mechanisms" (Hornberger et al., 1992; Corapcioglu and Haridas, 1985). The main deposition mechanisms are straining or filtration in the contact zones of adjacent pores, sedimentation in the pore spaces, and adsorption (Matthess et al., 1988; Corapcioglu and Haridas, 1985). The presence of organic material can also limit the extent of bacterial transport in soils (Yates and Yates, 1988).

Macropores tend to be concentrated near to the soil surface and can form extensive connected systems providing rapid drainage to the stream channel (e.g. Hunter et al., 1992). Land-drains are man-made macropore systems and also provide a rapid transport pathway through the soil for coliforms (Evans and Owens, 1972). Despite providing a rapid drainage 40



pathway, macropore flow is likely to be infrequent. The voids only become active when infiltration excess or saturation excess occurs. The condition of infiltration excess is when the rainfall rate or surface accumulation of water exceeds the ability of the soil matrix to soak it up. Saturation excess occurs when the soil mass is saturated. Both conditions result in the excess water spilling into the adjacent cracks and pipes and draining rapidly through the macropore system (Germann et al., 1987). Frequent or short duration rainfall events are unlikely to cause macropore flow and the sub-soil transport of organisms will be dominated by matrix through-flow (Germann et al., 1987).

Small-scale surface run-off features that have been observed to contribute to surface run-off include seeps, springs, return flows and protostreams (Hunter et al., 1992). Protostreams are transient surface drainage channels that develop during rainfall where subsurface and surface flows combine to produce a defined and recognisable channel (Hunter et al., 1992). Except for protostreams, these run-off features result from saturation of the soil system through either rainfall excess or occur where the hill slope intercepts the local water table. Return flows from macropores may contain many organisms, however, seeps from the soil matrix will have undergone some degree of filtration and thus contain relatively fewer organisms (Hunter et al., 1992; Harvey, 1989).

Thus faecal coliforms on the land surface have a number of pathways by which they can enter the stream. They might travel through the soil system via the fine pore spaces of the soil, or in the large voids caused by cracking or roots within the upper soil layers. Alternatively, if conditions suit they may travel in surface flow, or by some combination of these pathways. The next section presents the findings of process studies in the UK and elsewhere which have investigated faecal coliform transport via these pathways. The significance of different flow pathways is highlighted in terms of location within the catchment and also the temporal dynamics with respect to storm events.

Field observations of transport processes Kunkle (1970) studied upland catchments with permeable soils in the US, finding that bacterial contributions from areas away from stream margins were small compared to those derived near channel and grazing had minimal impact when carried-out away from the stream margins. This coincided with hydrological observations that upland areas contributed 41



little or no overland flow during storms; most storm runoff originated on saturated areas which built-up along channel edges (Kunkle, 1970). Hunter and McDonald (1991a) found that moist areas, where overland flow was preferentially generated, contributed significantly higher numbers of faecal coliforms to surface waters than drier areas.

Hunter et al. (1992) examined the relative contributions of inputs from overland flow, matrix through-flow, and macropore flow to stream bacterial numbers in a small research plot adjacent to the River Skell in upland Yorkshire. At 11 sites along the stream the faecal coliform load contributed by overland flow was 2 orders of magnitude higher than macropore flow and 5 orders of magnitude higher than matrix through flow. The relatively small bacterial loading from matrix through-flow was due to the combination of low concentrations and low flows (Hunter et al., 1992).

Overland flow was a major contributor to stream bacterial load during both base and stormflow conditions. Seeps, springs, return flows and protostreams all contributed to overland flow (Hunter et al., 1992). Considerable variation was found in the median faecal coliform load values at overland flow sites. These were derived from either bacteriologically pure return flows or highly concentrated flow in protostreams, or some combination of these (Hunter et al., 1992). Hunter et al. (1992) suggested that protostreams were important in the transport of coliforms from the catchment surface to the streambed store between storm events. This latter observation is valuable in that it provides a possible explanation of stream-bed recharge between storms, despite the fact that large numbers of organisms are washed into the stream during a storm (see Section 2.2.4).

Germann et al. (1987) suggested that macropore flow was unlikely in response to frequently occurring or short duration rainfall events. This appears to be the case at the macropore flow site (P14) along the Skell (Hunter et al., 1992) where there were no samples at stream stage heights of less than 30 cm (Figure 2.10,b). Field observations showed that water and entrained organisms reaching non-matrix through flow sampling sites came via root systems and non-biological voids draining from the catchment surface on the land close to the stream (Hunter et al., 1992).

42



Figure 2.10 Scatter plots of log10 faecal coliform load against stage height for inputs from a. overland flow and b. non-matrix throughflow (after Hunter et al., 1992).

Faecal coliform input load and concentration from both overland and macropore flow sites exhibited a strong positive correlation with stream stage height suggesting that the bacterial input rate was largely determined by rainfall conditions (Figure 2.10) (Hunter et al., 1992). The greater the volume and duration of rainfall the greater the load of organisms delivered to the stream.

At a quarter of the overland flow sites, however, faecal coliform concentration was found to fall as flow rate in the stream increased. Hunter et al. (1992) suggested that this reduction in numbers resulted from localised depletion of the supply of organisms on the land surface. Faecal coliform load at these sites, however, still increased with flow thus there must have been sufficient organisms from non depleted areas to offset the diluting effect of the rising flow. It was suggested that areas prone to depletion may have been those subject to bacterial removal by continual water movement, i.e. they flow during both dry and wet weather conditions (Hunter et al., 1992).

There were few samples of zero bacterial concentration at macropore flow outlets and Hunter et al. (1992) suggested that such sites were the end-points of extended drainage systems. This is a reasonable proposition, in that organisms would be supplied from a large surface area; the probability of a large area supplying no organisms is less than the probability that a small area would supply no organisms. Smith et al. (1985) also found that 43



transport via macropores could result in significant contamination many tens of metres from the source of bacteria. Thus when rainfall is sufficient, macropore systems provide a rapid bypass of slower soil matrix transport and can contribute large numbers of organisms to the stream network.

In areas where the soil is cultivated macropore flow is unlikely to be a major route for faecal coliform transport to the river network. Smith et al.(1985) found that in soil columns where the natural soil structure was destroyed, i.e. macropore features were absent, there was no significant bacterial transport beyond a few metres. Where the soil structure was intact transport via macropores was rapid. Thus the process of ploughing, which regularly destroys the network of macropores, effectively eliminates one route by which organisms may reach the stream network. It is possible that soil compaction by grazing animals may also disrupt macro-pore flow; this may also cause enhanced surface run-off. Indeed, Hunter et al. (1992) suggested that inputs from the catchment surface become less significant with distance travelled downstream and hence into lowland areas where cultivation is more common. At the same time greater human density results in increased inputs from effluent discharges. Inputs from the channel-bed store will also be greater in lowland areas, because reduced channel slope and the presence of navigable reaches result in greater settlement and storage of organisms between storm flows compared to upland reaches (Figure 2.11).

Figure 2.11 The relative importance of bacterial input rates from the catchment land store and the channel sediment store to stream bacterial dynamics, with distance downstream (after Hunter et al., 1992). 44



In general the area adjacent to the stream will be the most important source area for organisms reaching a watercourse. Overland flow and macropore flow, contribute the greatest numbers of organisms in upland areas. Overland flow may also be important between storms, replenishing the channel bed-store. Extended macropore systems provide rapid drainage from large areas but only “operate” at high flows. Matrix throughflow is the main drainage pathway for organisms in lowland areas and contributes water to the river network but not major numbers of organisms due to filtering within the soil. In lowland areas inputs from point sources and from bed-store resuspension are likely to be more important than run-off. Thus with knowledge of current land-use and topographical information a simple assessment can be made of areas important in delivering organisms to the stream network. This is valuable not only to the modeller but also to catchment planners who can use the information to target areas for improving run-off quality.

2.2.4 Fluvial dynamics The settlement, storage and subsequent resuspension of organisms are important mechanisms in the transport of faecal coliforms in river systems. These processes drive the background to the fieldwork and model development presented in Chapters 3, 4, 5 and 6. Section 2.1.3 shows how faecal coliforms become attached to particles with which they may settle from the water column into concentrated pockets of storage on the river-bed. The organisms stored in these pockets, when suitably disturbed, can be resuspended into the flowing water above. This section demonstrates that these pockets can return significant numbers of organisms to the flowing water when stream flow increases in response to a rainfall, or artificial flow, event.

Faecal coliform resuspension Studies in the United States (Morrison and Fair, 1966; Kunkle and Meiman, 1967; McSwain and Swank, 1977) demonstrated enhanced coliform concentrations during high or rising flows and suggested the possibility of a link between increased coliform concentrations, suspended sediment concentration and the stream-bed area. Elder (1978) showed that the supply of organisms available at high flows was finite and became exhausted by successive flow events. Meiman and Kunkle (1967) observed sharp increases in concentration during 45



storm flow declining to near zero on the receding limb of the hydrograph. A seasonal pattern of bacterial accumulation in the stream channel had been observed long ago (Streeter 1934); periods of low flow favouring sedimentation and high flow causing depletion. However, no data was available for UK streams at that time.

Kay and McDonald (1980a) undertook a more thorough examination of the sources and behaviour of faecal coliforms in two upland UK streams; Capelshaw Beck and the River Washburn in Yorkshire. Sampling for total coliform during periods of constant flow over three 24-hour periods in summer and winter established that background concentrations were highly variable (between 2 and 600 counts per 100ml). Storm flow events were also sampled and significant increases in concentration were observed during the rising limb of each event.

The existence of reservoirs in the Washburn system made it possible for stream flow to be varied artificially. This facility was used to increase the flow without bacterial inputs from the catchment, in order to investigate the change in coliform concentration and distinguish between the sources of organisms contributing to previously observed peak concentrations. If the main source of organisms was run-off from the catchment, the artificially produced hydrograph would not increase the stream coliform concentration, the low bacterial concentration reservoir water would actually dilute the numbers in the stream (McDonald et al., 1982). The first release showed this not to be case. A marked bacterial peak, of similar magnitude to the natural flow events, coincided with the peak stage. This showed that the bacterial peak was caused by the entrainment of organisms stored within the channel, but it was not clear whether the organisms came from the channel bed or banks, or both. This was tested with a second release designed to cause streambed disturbance with minimal stage rise. A six-fold increase in velocity was achieved which again produced an increase in coliform concentration of similar magnitude to the natural hydrograph, suggesting that direct entrainment of organisms from the channel bed was the main source of organisms (McDonald et al., 1982). Further releases on the Washburn demonstrated the increase in peak bacterial concentration with propagation of the hydrograph downstream, suggesting that this resulted from the continued accumulation of organisms resuspended from storage within the channel (Figure 2.12). The finite nature of the channel supply of organisms was also demonstrated during an extended artificial high flow. A bacterial peak, coincided with 46



the initial increase in flow and the concentration subsequently fell to pre-release values (Figure 2.13), this observation was repeated on the Afon Tyweryn in North Wales (Kay and McDonald, 1982).

Figure 2.12 Propagation of artificially generated hydrograph with response of total coliforms (TC) and E.coli concentration at locations (a) 400m, and (b) 2500m downstream of the hydrograph source (after McDonald et al., 1982).

Figure 2.13 The response of total coliforms (TC) and E.coli concentration to a step change in stage height (after McDonald et al., 1982).

A final confirmatory test, involving a further reservoir release on the Washburn, immediately after a major natural flow event, was made. If the natural hydrograph had not flushed the channel bed of organisms then the artificial release would. The release resulted in a dilution of the stream concentration (McDonald et al., 1982) indicating that natural run-

47



off events wash organisms into the channel and flush out the organisms stored on the riverbed.

Moore et al. (1988) sampling Bear Creek in Oregon made an intensive study of faecal coliform concentrations during a flow event, observing elevated concentrations at two locations on the stream. The greatest increases at the downstream site were thought to be due to run-off from the adjacent cattle feed lot. Analysis of suspended solids concentrations showed a similar pattern of increase. A previous reservoir release made on Bear Creek doubled the flow from 0.074 cumecs to 0.150 cumecs for a period of 8 hours (Sherer et al., 1988). This produced a nineteen-fold increase in faecal coliform concentration from nonevent conditions (Moore et al., 1988). This was less than the increase in concentration induced by the rainfall-event. That yielded increases in concentration that were 0.5 and 2 times greater than the reservoir release. The largest increase was downstream of the feedlot, indicating a significant supply of organisms from this source. Thus although, Kay and McDonald (1980a) found that the resuspension of organisms from storage on the stream-bed was the main source of observed storm event bacterial peaks, this is not always the case; the magnitude of the different sources of organisms will vary from site to site. Indeed, Hunter et al.(1992) suggested that the relative significance of bacterial inputs from channel bed and the catchment surface is likely to depend on the sampling location within the river system. Inputs from the bed would be expected to increase in significance with distance downstream where lower flow velocities predominate and greater settlement occurs.

In summary, faecal coliforms that have become concentrated in pockets of storage on the river bed during dry weather periods, are capable of returning significant numbers of organisms to the flowing water above when flow conditions are sufficiently energetic to disturb and resuspend them. The numbers will accumulate in the water as the flood pulse passes downstream resuspending more organisms with its passage. The channel supply of organisms is finite and in the aftermath of a flow event will be depleted and not capable of supplying more organisms until replenished by organisms that have settling-out. Chapter 5 builds on the processes introduced above, presenting a new model for faecal coliform stream dynamics.

48



2.3 Modelling faecal coliform concentrations in rivers This section introduces models that have been used to describe faecal coliform die-off and stream dynamics. These are wide ranging and include a variety of modelling approaches. For example, multivariate statistical approaches relate bacterial concentrations to a number of driving variables using simple statistical relationships (Hunter and McDonald, 1991b; Kay and McDonald, 1983; Mahloch, 1974). This purely "black-box" approach pays little attention to the physical nature of the system or its internal processes. Simple deterministic first order decay functions have been used for describing the exponential die-off of some initial bacterial population (Crane and Moore, 1986; Mahloch, 1974; Streeter, 1934). This approach is only appropriate for closed systems, for example a beaker in a laboratory subject to constant environmental conditions. Minor variants to this type of model are demonstrated. Crane and Moore (1986) provide an excellent review of these simple models. Models of more complex systems such as lakes, estuaries and rivers need to incorporate terms for hydrodynamic and diffusion effects (Canale et al., 1993; Crowther, 1991; Johnson and Ford; 1987). Once fluid motion and diffusion has been approximated adequately, then terms to describe the non-conservative behaviour of faecal coliforms can be added; such as the first order decay functions mentioned above.

Since the aim of this study has been to develop an improved conceptual or process-based model for faecal coliform stream dynamics, this section concentrates on the physical and conceptual type modelling applications found in the literature. An overview is provided of the many die-off models as well as a more detailed examination of selected river model applications.

2.3.1 Bacterial die-off models Crane and Moore (1986) undertook a detailed review of bacterial die-off modelling and the following section summarises this work.

Simple first-order die-off models applied to faecal coliform are generally based on Chick's Law (Chick, 1908):

Nt = 10- k.t N0

(2.1)

50



The equation describes logarithmic die-off of a population N over a time period 0 to t, with constant die-off rate k per unit time. Figure 2.14 shows form curves for bacterial die-off from models based on Chicks Law; Equation 2.1 gives curve 1. Curve 2 might be observed if there was a period of growth prior to the onset of die-off at t = t1 (Equation 2.2).

N t = -k (t -t 1 ) 10 N0

(2.2)

More frequently data presented in the literature exhibit a time-delay to the onset of die-off during which the faecal coliform population remains unchanged (for example, Figure 2.8). In this situation it would be appropriate to introduce a case whereby for t  t1, Nt = N0, and for t > t1, Equation 2.1 applies.

A changing die-off rate may be observed over time as susceptible sub-groups of a population die-off more rapidly leaving longer lived sub-groups such that the overall die-off rate takes the shape of curve 3 in Figure 2.14. Frost and Streeter (1924) proposed a model for the die-off of a population composed of two sub-groups with different tolerances to environmental stresses. This can be expanded to give a series with any number of subgroups, a to i are the proportions of the initial total population with the respective die-off rates ka to ki (Equation 2.3).

Nt = a 10 -k a t + b 10- k b t  ...i 10 -k i t N0

(2.3)

A number of other models in the literature produce a function similar in shape to curve 3, a good example is Equation 2.4 (see Fair et al.,1971).

Nt = 10 -kt(1/n) N0

(2.4)

Here, n is a non-uniformity coefficient and n = 1gives a type 1 curve, n > 1 gives a type 3 curve and n < 1 produces a type 4 curve for the case where the die-off rate increases with time. This occurs during the chlorination process (Crane and Moore, 1986).

Equation 2.1 is the basic equation used for modelling faecal coliform die-off, in applications to real situations the die-off rate, k, has to be adjusted to account for the affects discussed in Section 2.1, i.e. for the influence of temperature and solar radiation etc. Equations for the influence of solar radiation have already been presented in Section 2.1. Equations for the effect of temperature on die-off rate coefficient include simple linear relationships (Qin et al., 1981; Canale et al., 1973), 51



k T = a  bT

(2.5)

and, the Arrhenius or Van't Hoff equations used to relate temperature and rate reaction coefficients (Auer and Niehaus, 1993),

k T = k 20 (T  20)

(2.6)

Figure 2.14 Bacterial die-off curves as predicted by models in the literature (after Crane and Moore, 1986).

In Equation 2.5, a is the y axis intercept, i.e. the die-off rate when T = 0, and b is the slope. The data presented in Figure 6.2, however, indicate that the relationship between temperature and faecal coliform die-off rate is not linear. Equation 2.6 can be used to fit the data presented in Figure 6.2, , sets the slope of the curve, for  = 1 the curve is flat, i.e. dieoff does not vary with temperature. For  < 1, die-off rate decreases with increasing temperature and for  > 1, increases with temperature.  =1.1, gives an approximate fit to the data of Evison (1989) presented in Figure 6.2 with a k20 of 0.3 per day. This model can be adjusted to fit all of the data presented in Figure 6.2, but despite this apparent versatility, it assumes some a priori knowledge about the die-off rate under a particular set of conditions in the system being studied. While this may be acceptable for a controlled laboratory experiment where the conditions are repeatable, in an application of a water quality model where so many factors influence the overall die-off rate this equation would be too rigid. It is unlikely that the die-off rate at a given temperature would be the same from one occasion to the next and changes in k20, as well as, , change the slope of the die52



off curve. Thus the equation not only fixes the die-off curve in one plane, it has two parameters to be assigned during model calibration. In Chapter 6 an alternative equation, derived from the literature data, is presented. This alternative equation merely assigns a slope to the curve and allows the overall die-off rate to float within a multi-parameter space under the influence of sunlight, temperature and pH, and only an initial overall die-off estimate is required to run the model.

As indicated previously (Section 2.1 and 2.2), die-off is just one of the influences on faecal coliform concentration at any point within a river system. The observed concentration is effectively a mass-balance of inputs and outputs; upstream and lateral inflows, resuspension from the stream-bed, and deposition, die-off, dispersion and outflow. The next section gives a brief overview of models that have been developed for, adapted or applied to faecal coliform modelling.

2.3.2 River models for faecal coliform dynamics There are few water quality models that have been adapted to model faecal coliform concentration and fewer still developed specifically for that purpose (e.g. Jenkins, 1984). Water quality models for rivers require a number of components in order to work successfully. Firstly a hydrodynamic component is necessary to model changes in water depth and velocity as the flow rate rises or falls with the passage of a hydrograph event. Secondly, the model must be able simulate the mixing processes within the channel. To these basic components terms may be added for faecal coliform die-off and the transfer of organisms to and from storage on the riverbed.

This section gives a brief overview of the component parts of water quality models and provides examples of attempts to model faecal coliform numbers.

The advection-dispersion equation (ADE) The ADE was derived from work on turbulent diffusion in pipes (Taylor, 1954) and has been in common use in water quality modelling since the early 1970s. It has remained popular in practical applications (Henderson-Sellers et al., 1988), despite the inability to adequately reproduce observed tracer profiles (Elder, 1959; Fischer, 1967; Nordin and Sabol, 1974; Day, 1975). The ADE has two terms, one for advection; longitudinal movement with the flow, the other describing dispersion; the effect of mixing on the 53



Figure 2.15 A graphical representation of the functioning of the advection dispersion equation (ADE) as applied to model the concentration of a conservative contaminant when there is (a),dispersion without advection, (b), advection but no dispersion, and (c), when the two affects are combined.

pollutant concentration. The model can be developed for one, two and three-dimensional transport depending on the nature of the system to be modelled. In applications to rivers a one-dimensional formulation is generally used, based on the false assumption that the flow is well mixed throughout the channel cross-section: 2 C C  C + D 2 = -U t x t

(2.7) 54

Modelling Faecal Coliform Dynamics where, x

= longitudinal distance

C

= solute concentration

U

= mean flow velocity

D

= dispersion coefficient.


This is the simplest form of the ADE, without additional terms for inflows or sinks, as it would be applied for a conservative solute, i.e. one that is not transformed by physical, chemical or biological reactions. Figure 2.15 provides a graphical representation of the functioning of the ADE. It shows how a pulse of a solute of initial concentration C0 entering a water body at point x0 at time t0 changes through time, and includes hypothetical cases (Figure 2.15a and b) to demonstrate the components of the model. In Figure 2.15a the flow velocity U = 0, i.e. there is no lateral transport, as might be expected in a still water body; only dispersion acts on the solute concentration which falls with time as it disperses into the surrounding water. If there were advection only, i.e. D = 0, the pulse would be carried downstream unaltered by mixing or dispersion (Figure 2.15b). These two effects combine to give Figure 2.15c showing transport downstream and attenuation of the solute concentration. The ADE is utilised in a wide variety of models describing, for example, salinity variation, sediment transport and bacterial transport (Crowther, 1991), it may also be applied, to flood-wave propagation (Whitehead et al., 1979). A number of examples of studies using the ADE follow.

A simple river application Al-Layla and Al-Rizzo (1989) made a simple application of the ADE in Iraq. A onedimensional version of the ADE was used to simulate a steady-state decline in total coliform count along 75km of the Tigris River downstream of the "Saddam" Dam. In a steady-state application the river flow is not varied, this greatly simplifies the modelling procedure, since the velocity, U, and coefficient of dispersion, D, are constant for each reach. Microbial dieoff was modelled using a simple first order decay coefficient (see Section 2.3.1). Figure 2.16 shows the profile of measured and simulated coliform concentrations with distance downstream of the dam.

Whilst the general trend of coliform decline is effectively modelled, there is some inadequacy in either the selection of die-off rate or the identification of contaminant inputs 55



which may account for the deviations of the modelled coliform concentration from the observed data.

The above example represents a very simple application of the ADE used to simulate the decline in coliform numbers with distance downstream of a dam. The presence of the dam means that flow is easily controlled; in this case held steady. Under these conditions a steady-state application of the ADE is adequate. In systems where flow variations are driven by natural processes, such as, rainfall-runoff or tidal cycles a more involved approach is necessary to model the changes in solute concentrations and faecal coliform numbers.

Figure 2.16 Profile of simulated and measured coliform bacteria concentrations for the Tigris river in Iraq (after, Al-Layla and Al-Rizzo, 1989).

Two dimensional steady-state sediment transport model for rivers Uchrin and Weber (1979, 1980, 1981) highlighted the importance of the solid phase, i.e. particulates, in determining the behaviour and fate of pollutants, berating the fact that, at the time, virtually all water quality models did not incorporate this important link. They reported the use of a two-dimensional steady-state formulation of the ADE (Equation 2.8) to describe the transport and distribution of suspended solids and associated pollutants, including faecal coliforms, in a steady turbulent river channel: 2 2 C C C  C  C = -U + D 2 +W +D 2 t x x z z

where, C

(2.8)

= concentration of suspended solids, 56


and,

x

= horizontal channel distance,

z

= vertical distance above bed,

W

= particle settling velocity,

U

= mean velocity of ambient fluid and particles,

D

= turbulent diffusion coefficient for particles.


A generalised bottom boundary condition was formulated to allow for complete settling or re-entrainment of particles such that:

D

C = - WC z

(2.9)

and, substituting Equation 2.9 into Equation 2.8 and rearranging gives: 2 C C C  C U +D 2 = W (1   ) x z t x

(2.10)

where,  is the settlement / entrainment coefficient with a value between zero and one;  = 0 for complete settling, and  = 1, for complete re-entrainment (Uchrin and Weber, 1979). Particle settling velocity, W, is treated as a randomly distributed function, generated from quiescent column settling data, which can be calibrated to account for turbulence. Given these boundary conditions and definitions, the two-dimensional partial differential equations (PDE) can be solved numerically. The resultant solutions can then be used to generate a rate law for specific water quality parameters of the form;

dC = k Cn dt

(2.11)

where t is the time of travel and k and n are constants relating to particle and hydrodynamic conditions. For uniform particle fall velocity, n has been found to approach unity.

This suspended solids model has been incorporated into a lake system model, which utilises a finite section approach, and applied to a recreational lake in South East Michigan in the United States. In the application of this model bacterial removal from the water column, other than by sedimentation, i.e. predation, die-off, etc., was lumped into a first order function parameterised from field data.

Estuarine transport modelling with the ADE Crowther (1991) presents a one-dimensional application of the ADE to the transport of sediment and the dispersal and inactivation of E.coli in the Clyde Estuary in Scotland.

57



Continuity and momentum equations (Equations 2.12 and 2.13, respectively) were used to model water movement in the estuary:

(W + W s )

 h Q + = LL t x 

Q  Q h gAR  + [ ] + gAS f + gA + =0 t x A x 2  x

(2.12)

2

where Q

and

(2.13)

= discharge (cumecs)

h

= stage height (m)

t

= time (sec)

x

= distance downstream (m)

W

= width of estuary at surface (m)

Ws

= average width of storage areas (m)

L

= lateral inflows per unit length (m2.sec-1)



= density (kg.m-3)

A

= cross sectional area of flow (m2)

R

= hydraulic radius (m)

Sf

= friction slope.

The model assumes that storage areas such as docks and basins do not affect the momentum balance in the estuary and that lateral inflows add no momentum to the flow (Crowther, 1991). The effect on solute concentrations of storage areas such as docks was also ignored; the docks are long and narrow with small entrances, and tend to be poorly mixed. The ADE was used to model salinity, which is treated as a conservative solute: 2 Q S S  S L =+ D 2 + SL A x A t x

(2.14)

and extended to account for sediment, 2 E - Dp  Cs Q  Cs  Cs L =+D + C sL + Cs 2 t A x A A x

(2.15)

bacterial behaviour, 2  Cb Q  Cb  Cb L =+D + C bL + k C b 2 t A x A x

where S

(2.16)

= salinity

Cs, Cb = concentration of sediment and bacteria respectively, D

= dispersion coefficient,

E

= amount of sediment eroded from channel bed, 58


and,


Dp

= deposition rate,

A

= cross-sectional area of flow,

L

= lateral inflow,

k

= bacterial die-off rate, f(sunlight, salinity, temperature and suspended solids).

For die-off Crowther (1990) presents a scheme similar to that of Auer and Niehaus (1993), Qin et al. (1991) and Sarikaya and Saatchi (1987):

k = k d  I z ,avg

(2.17)

Unlike Auer and Niehaus (1993), the equation for kd uses a formula derived from observations made in laboratory experiments examining the relationship between die-off and salinity, temperature and suspended solids concentration (Milne et al., 1991) and described in Section 2.1.2:

k d  a(T ,Cs ) Se b( T ,Cs ) .S

(2.18)

Values for a(T,Cs) and b(T,Cs) are assigned from tables in the data files that support the computer program of the model (Crowther, 1990). The value of  is also provided in the same tables.

The model does not link E.coli dynamics to sediment behaviour. The bacteria are transported by the flow without transfer to or from storage with sediment on the estuary floor. The reason for this apparent omission may be the nature of the supply of organisms, since Milne et al. (1986) found that there was no significant settlement of organisms from mixtures of sewage effluent and sea water. It is also likely that the extreme turbulence experienced in an estuary would be sufficient to keep the majority of organisms in suspension (see Section 2.1.3).

Sediment (Equation 2.19) is modelled using the Krone formula, relating entrainment, E and deposition, Dp to channel bed shear velocity (u*) and critical velocities of deposition and entrainment, vd and ve. Equations 2.19 and 2.20 show the functions used (Crowther, 1991).

For deposition u* < vd,

D p = (1 -

u*

2

vd

2

)W b vs

(2.19)

and for erosion u* > ve, 59


E=(

u*

2

ve

2


- 1) W b 

(2.20)

Where,  is the erosion rate, Wb the bed width and vs the settling velocity. Shear velocity, u*, is given by;

u* = |

Q | A

k loge { 30.2

H

(2.21)

}

ks

where, ks is the Nikuradse sand roughness coefficient, H is depth, and k, the von Karman constant for flow with sediment, is 0.174.

This application of the ADE highlights the potential complexities of modelling estuary systems. The ebb and flood of the tide complicates the relationship between flow and water level as well as producing high flow velocities that cause the sediment to be moved cyclically up and down the channel. The application requires detailed hydrodynamic, sediment and bacterial dynamics models with many equations, coefficients and parameters. Unfortunately, Crowther (1991) never completed the application of this model to the E.coli concentrations, it was therefore not possible to assess the validity of the model.

Water Quality Modelling using the Active Mixing Volume (AMV) theory A more easily applied and physically plausible approach to river mixing is provided by the Active Mixing Volume theory (see Price et al., 2000). The AMV theory was born out of efforts by Valentine and Wood (1977) to improve the performance of the ADE when modelling tracer dispersion. The ADE alone is not capable of reproducing the time profiles of a solute pulse passing a point in a natural channel (e.g. Henderson-Sellers et al., 1988). These profiles have a rapid rise in concentration followed by an extended tailing-off in concentration after the peak of the tracer pulse has passed the point of observation. Valentine and Wood (1977) added another term to the ADE to simulate the effect of a proportion of the channel flow being diverted into zones of storage, referred to as "deadzones", where the water would circulate slowly, gradually releasing the diverted water back to the main flow. When the main channel concentration is falling, water of higher concentration, from dead-zones, returns to and extends the tailing-off of the solute pulse. Beer and Young (1983) went one step further, showing that the main ADE terms for

60



dispersion and advection could be discarded entirely, successfully using the dead-zone component alone to model dispersion.

The dead-zones, that dominate dispersion within a channel, comprise many pockets of recirculatory flow over a wide range of size scales, ranging from areas of storage marginal to the main flow, such as reverse flows on bends or in pools, and smaller features such as turbulent eddies and wakes associated with roughness elements in the channel. These features create a spectrum of areas of slow and fast transport.

A critical assumption made in applying this model is that the sum of the dispersive effect of the dead-zones within a reach can be modelled as if they are a single dead-zone and hence the model has, until recently, referred to as the aggregated dead-zone model, or ADZ. This is something of a misnomer and Wallis (1993) refers to the model as an active mixing zone model (AMZ), since the dispersion elements are dynamic zones within the flow. The model itself merely represents a mixing volume within the flow and not actual zones of mixing, and is now referred to as an "active mixing volume" model, or AMV, for which there are a variety of applications besides solute transport in rivers (see Young, 1992). HendersonSellers et al. (1988) discuss the relative merits of the ADE and AMV approaches to dispersion modelling. Not only is the AMV approach felt to offer a better conceptualisation of mixing processes in natural channels and reproduce time-series of tracer concentrations derived from field experiments, it does not suffer from the numerical dispersion encountered when approximating the ADE in computer applications (see Young and Wallis, 1993).

For a conservative solute, or contaminant, under conditions of constant discharge the AMV model in its simplest is given by,

dVe xt = Q u t - - Q xt dt where, Q

(2.22)

= discharge

Ve

= active mixing volume

ut

= upstream input concentration

xt

= downstream outflow concentration



= pure time delay (for advection)

The concentration of a conservative solute in Ve is governed by changes in the inflow concentration ut. If ut is greater than the concentration in the mixing volume, xt, that concentration rises. If ut is less than xt, xt falls. Advection is accounted for by a pure time61



delay , and, for a pulse of solute passing down a reach, is the time elapsed between the solute entering the top of the reach and the first increase in concentration as it arrives at the end of the reach. Ve/Q=T, is the AMV residence time, and determines the model dynamics. A large residence time results in a slow model dynamic and vice versa. For example, in a fast flowing reach where solute transport is dominated by advection the AMV is small with respect to the rate of flow through the reach. In lowland reaches, where artificial structures maintain unnaturally high water levels at low flow velocities, as for example in the lower reaches of the River Dee (see Chapter 6), the residence time of the reach will be much greater since the mixing volume is high with respect to the flow.

The AMV is used for river mixing and dispersion processes in the model presented in Chapters 5 and 6 of this thesis where the application is described in greater detail. The following sub-section introduces one of the few, if not the only, model to combine faecal coliform dynamics and an AMV type approach to contaminant transport in rivers.

The QUASAR Model The QUASAR (QUAlity Simulation Along Rivers) model embodies AMV theory and incorporates additional terms for a variety of water quality parameters including E.coli. The model was developed from work by Whitehead and Young (1975) that investigated the impact of the new town of Milton Keynes on the water quality of the Bedford-Ouse in the UK. The model is applied on a reach by reach basis down a river system, performing a mass balance of flow and water quality sequentially downstream. The reach boundaries are determined by points in the river where there is a change in the water quality or flow due to the confluence with a tributary, the location of a sewage treatment final effluent discharge, an abstraction, or the location of weirs.

The model uses a "kinematic" approach to flow routing based-on the assumption that the propagation of a flood pulse, or wave, is analogous to the passage of a pulse of a conservative pollutant (Himmelblau and Yates, 1968). The model uses an equation of identical form to Equation 2.22, but where xt and ut are replaced by Qj-1 and Qj

d V j Q j dt

= Q j (Q j 1 - Q j )

(2.23)

62



and V j and Q j are the mean or reference reach volume and flow about which small changes in flow occur. V j / Q j   j , the time constant. Written in continuous time transfer function form Equation 2.23 becomes,

Q j ( s ) 1 = Q j 1 ( s) (1   j s ) n

(2.24)

where, s is the Laplace operator and n is the number flow routing components, or "dynamic lags", in series (Whitehead et al., 1979). The model is fitted to observed data by adjusting the number of lags and the time constant. For very long reaches a pure time delay can be added to account for the first arrival of a flood-wave. The approach has similarities to, and advantages over, the Muskingum-Cunge flow routing method, which gives a good approximation to the ADE

Q Q 2 Q = -u + D 2 t x x where, Q is discharge, D

(2.25) is the dispersion coefficient, and u is the kinematic wave

velocity, which is faster than the mean velocity, U (Equation 2.7).

The advantages of the "dynamic lag" model is that it does not suffer from "non-minimum phase behaviour" of which can result in reductions in reach output flow in response to sharp increases in flow to the reach input, it is also less complicated (see also Beven and Wood, 1993).

The model is further simplified by the fact that it may be sufficient to use the actual reach volume, V, and flow, Qj, instead of the reference volume and flow, V j and Q j , and Qj-1 and Qj, instead of the small variations in flow, Qj-1 and Qj.

Reach volume may then be determined from V=Al, where l is the reach length and A is the cross-sectional area given by

A = hb 

h2 s

(2.26)

for a trapezoidal channel, where b is the base width and s, the bank slope and h the reach depth. This approach is applied in Chapter 4.

63



The only parameter left undefined is n, the number of lags. The choice of n was aided by the assumption that flood-wave and tracer propagation is analogous, such that flow and water quality may be modelled with the same equations and parameters (Whitehead et al., 1979). The number of lags is determined from the model fit of the dispersion of a conservative tracer (Whitehead et al., 1979).

The model formulation for E.coli includes a temperature dependant decay rate, k, such that

Qj dC j Q j = C j 1 - C j + I j - k C j dt V V

(2.27)

where, Cj is the concentration of E.coli at the outlet of the jth component and Cj-1 is the output of the preceding component. Ij is sources and sinks due to effluents, abstractions etc. The model has been applied to UK rivers, such as the River Tawe to assess heavy metal pollution and the River Thames to assess the movement and distribution of nitrates and algae (Whitehead and Williams, 1982, Whitehead and Hornberger, 1984). The model can be applied to a long list of water quality determinands (E.coli, nitrate, dissolved oxygen (DO), biochemical oxygen demand (BOD), ammonia, ammonium ion, temperature, ortho-phosphate, pH, and a "conservative" water quality parameter).

The AMV / QUASAR approach has been used as the basis for the modelling developments presented in Chapters 5 and 6. The next section describes one of the few models developed specifically for the purpose of modelling faecal coliform transport in streams and rivers.

A "process-based" model of bacterial dynamics in upland streams This model was derived from the work of the University of Leeds School of Geography study group (Jenkins, 1984; Kay and McDonald, 1982; McDonald et al., 1982; and Kay and McDonald, 1980a) who investigated faecal coliform fluvial dynamics in Yorkshire, the results are summarised in Section 2.1 and 2.2. These studies highlighted the importance of river-bed storage in determining faecal coliform fluvial dynamics by demonstrating the impact of organisms entrained into a rising flow without catchment inputs. The resultant model employs a mass balance formulation for the bacterial concentration in the water

C = I+ W - S - D

(2.28)

and numbers stored in the river-bed

N = S - W - D

(2.29) 64



where, I represents inputs from land-surface, W is entrainment from the river bed (Jenkins (1984) uses the term "washout", the term entrainment is used in the current text for consistency), S is settlement into bed storage and D is net die-off as a result of environmental stresses.

The discussion of this model starts with the delivery of organisms from the catchment into the stream channel. Non-point source inputs are derived from both "quick" and "base" flow runoff components. The base-flow component comprises, soil throughflow and groundwater flow; responsible for maintaining inputs during dry periods. The quick-flow component is the rainfall response component resulting in surface run-off and non-matrix throughflow. It was assumed that rainfall induced processes increase in intensity through the storm and that inputs to the channel increase linearly with discharge. Base-flow, or background, and discharge related input were assumed to be diluted by the volume of flow. The input of bacteria to the stream from the surrounding catchment (I) is, therefore, given by

IQQ IB I = IB + = + IQ Q Q Q

(2.30)

where, IB is background input, IQ is discharge related input and Q is discharge.

Bacterial inputs from the catchment surface enter the water only, i.e. the organisms do not enter riverbed storage directly from the catchment. Inputs to riverbed storage are by settlement from the flowing water alone as an internal function of the reach. Increased inputs raise the riverbed store bacterial numbers indirectly by providing larger numbers of bacteria for settlement. Jenkins (1984) suggested that settlement was unlikely to remain linear as discharge increased, on the basis that turbulence would reduce the rate of settlement, but not the actual numbers of organisms settling since the discharge is greater. Hence, the numbers of organisms settling to riverbed storage,

S = P set

1 Q C = pQC 1 + Q/ PQ

(2.31)

where, Pset is the rate of settlement, or the proportion of the total load settling per unit time, PQ is the discharge at which Pset halves. Jenkins (1984) went on to calculate the increase in numbers of bacteria per unit stream bed area due to settlement,

dN S = dt vw

(2.32)

65



where, N is the number of bacteria in the riverbed store per unit bed area, v is velocity and w is the channel width. Jenkins (1984) then replaced the terms for velocity and channel width with discharge on the assumption that vw = Q/h, where h is the mean flow depth and h  QF, where F  0.5, such that

vw =

Q Q

(2.33)

0.5

and,

dN C = P set Q0.5 dt 1 + Q / PQ

(2.34)

which defined the change in numbers of organisms stored in the riverbed due to settlement in terms of flow alone.

Entrainment is modelled in a similar manner to settlement, as a function of flow, and is assumed to occur at some threshold discharge below which there is no entrainment. Jenkins recognised that this was a simplification of a more complicated fluvial process involving behaviour such as sediment release following cobble movement and the adherence characteristics of the bacteria to the riverbed substrate and sediment particles. It was assumed that the entrainment of bacteria from the riverbed store would follow a suspended sediment rating curve such that

W = a Qb N

(2.35)

and b was assigned a value of 2, and as entrainment was assumed to commence at some threshold discharge, QT, 2

W = a( Q2 - QT )N

(2.36)

and,

a=

T H2 2 QT + 1

(2.37)

where, TH2 is the proportion of the bacterial store entrained when Q = QT + 1 m3.s-1. This allows a to be conveniently formulated in terms of perceived system behaviour or field observations.

A further refinement of the entrainment term was to limit the number of organisms that can be washed-out of storage at each time step. N is replaced by 66


f(N) =


N. Sat N + Sat

(2.38)

where Sat is the maximum value of f(N), which only approaches Sat for values of N much greater than Sat (Figure 2.17). This was seen as representing the entrainment capacity of the flow (Jenkins, 1984), i.e. for a very large bed store the flow could only entrain from a small proportion of the organisms available at each time step. It is also a convenient way of preventing all the available organisms from being washed-out too rapidly.

The full differential equations for the change in concentration in the water and the change in bacterial numbers in the riverbed store are

a( Q 2 - Qt2 )N I B dC = - K c C - pC + + + IQ dt Q Q and

dN = - K n N + pQC - a( Q 2 - Qt2 )N dt where, Kc and Kn are bacterial die-off coefficients in water and the bed store, p is the settlement coefficient (Equation 2.31).

Figure 2.17 Plot of washout limitation function, for Sat=1000, Equation 2.38.

Jenkins et al. (1984) presented model results for applications to event data presented in Kay and McDonald (1980a) and in the example presented achieved a good reproduction of the 67



timing and magnitude of the observed bacterial peaks, but was unable to reproduce the behaviour observed in this study. The need for further research to clarify the processes upon which the model was based was suggested (Jenkins et al., 1984), in particular the investigation of catchment supply; the focus of work by Hunter et al. (1992, 1999), and development in the downstream direction, i.e. investigations away from the headwater areas.

This model provides the key step towards improving the modelling of faecal coliform concentrations, giving an excellent basis for the current study. In the current study the main principles of this model are adopted and a new scheme proposed that provides a simple conceptual formulation with fewer parameters and in the entrainment module (Chapter 5) and including die-off terms and other factors relating to bacterial delivery from the catchment in addressing issues of longer-term variations in concentration in a large lowland reach (Chapter 6).

68


3. Study Rationale and Methods

3 Study Rationale and Site Information This chapter describes the direction of the work in this thesis, outlines the approaches used, details the field studies and analytical methods used. Site descriptions are provided, as well as, background information about the sites where this is of relevance to the investigations conducted in the later chapters.

3.1 The Modelling Problem The aim of this study was to produce an improved process-based model for predicting faecal coliform concentrations in stream and river systems. The literature review has highlighted the breadth of work on faecal coliform die-off dynamics and studies of riverine transport and delivery to the stream system from catchment sources. Existing models, with the exception of Jenkins (1984), only incorporate a simple die-off function into a general river transport and dispersion model. Some of these are steady-state only and others have a hydrodynamic component for fully-dynamic applications. For steady-state applications, i.e. when the flow is constant or near constant, a simple formulation with die-off alone may meet the needs of the modeller or water quality planner, for example to examine the decay of an effluent input to a river with distance downstream. This is not to suggest that such an approach is satisfactory, since the calibrated die-off rate reflects both die-off and settlement and a proportion of the organisms are not dead, but may remain on the riverbed for extended periods until, resuspended, they may lower the quality of the overlying water. Thus a fully dynamic approach, i.e. one that accounts for natural flow variations, is required to account 69



for the behaviour of faecal coliforms in river channels. Such a model should also account, in some way, for the storage and resuspension of organisms on the riverbed. Clearly, the model of Jenkins (1984), satisfies the requirement for transfer to and from a riverbed store, but is not formulated to account explicitly for flow routing or the mixing characteristics of the channel.

So far, the discussion centred on the shorter time-scale modelling of faecal coliform fluvial dynamics, this study is, however, not merely concerned with fluvial effects. At longer timescales, i.e. annually, as well as, diurnally changes in temperature and solar radiation have a marked impact on faecal coliform concentrations (see Chapter 6); faecal coliform die-off under summer conditions can be an order of magnitude more rapid than in winter (Section 6.2). This degree of variation highlights the importance of these influences and demands that their effect on die-off is accounted for in a dynamic model of faecal coliform behaviour.

With this background to faecal coliform dynamics in mind the following objectives were established: i.

the need to provide data of high quality for the purposes of examining faecal coliform dynamics,

ii.

to use this data to further test / validate existing models of faecal coliform dynamics, and

iii.

to re-formulate, or develop new, model structures on the basis of the observed behaviour.

Strategies, initially proposed, to address the above objectives and the choice of the eventual approach taken are presented in the following sub-section.

3.2 The Provision of Data The first stage for providing data for this study was to discover what existing data sets were available in the UK, to examine the length of record, frequency of sampling and the presence of other supporting variables, such as discharge, water temperature, irradiance and turbidity. On the basis of these findings, the need, or otherwise, for experiments, or regular monitoring, was assessed.

70



The majority of faecal coliform data in the UK is held by the private water supply companies and the Environment Agency and is associated with the routine monitoring of river locations with abstraction for drinking water supplies, or where summertime surveys of recreational water quality have been carried out. See Appendix IV for a listing of water quality data-sets including bacteriological determinands collated during this study.

For a data set to be of use in this study it had to meet one or more of the following criteria. The data should: i.

Span a continuous period of at least one year. This was in order to provide useful information on annual variability.

ii.

Be sampled at a high frequency over several flow events in order to indicate the timing and magnitude of storm responses.

iii.

Be sampled at least daily to provide adequate information about long-term dynamic variations in concentration (see Whitehead, 1978).

iv.

Have multiple sites along one river for the purposes of reach based input - output modelling.

v.

Have supporting physico-chemical information and be located near a flow gauging site.

vi.

Be derived from samples collected and microbiologically enumerated to recognised national standards, in order to assure the quality and validity of the data.

This last point was assessed by means of a detailed questionnaire sent to the laboratories of the contributing agencies, and showed that they all used the membrane filtration (MF) technique to the standard set out by Report No. 71 "The Bacteriological Examination of Water Supplies", HMSO, 1983. In other respects the data were very variable (see Appendix IV), with only a small number of data sets meeting the above criteria. The best of these was for the River Dee at the Huntington Works drinking water abstraction point near Chester. Daily weekday samples were collected at this site and data for the six years from 1 January 1986 until 18 November 1991 were made available. The data set included temperature, turbidity and pH, and was considered to be of great value for the purposes of investigating annual cycles of faecal coliform behaviour. Of the other sites the River Exe in Devon had two sampling locations which could be treated as input / output data for a series of reaches (work on this data is presented in Wilkinson et al., 1995b). Work on the River Dee data set forms the basis of Chapter 6. There were no existing data sets sampled sufficiently 71



frequently to assist investigations into faecal coliform fluvial dynamics, although data from the experiments of Sherer et al. (1988), Jenkins (1984), Kay and McDonald (1982), McDonald et al. (1982), and Kay and McDonald (1980a) were available for model testing, if required. Literature data, presented in Chapter 6, on die-off related to environmental factors was used to develop improved die-off equations (see Section 3.4.1 and Section 6.1).

Having established the availability of data suitable for model testing over longer timescales, strategies for providing event-based data, for examining faecal coliform fluvial dynamics, were devised. Three approaches were considered: a) Monitoring the bacterial response of a natural channel to artificially induced changes in flow. b) Natural Hydrograph Events. c) Controlled experiments in an artificial channel. The first option, monitoring the bacterial response of a natural channel to artificially induced changes in flow, would involve making controlled releases from an impoundment of some kind and sampling during the passage of the generated flood-wave at locations downstream. Stream-flow manipulation has certain advantages over normal hydrological events. From a purely logistical point of view, the timing and amplitude of the hydrograph can be planned, this avoids the expense and wasted time of periods spent in the field with a monitoring team and equipment awaiting the on-set of a suitable event. More importantly, such experiments would provide data to quantify the interaction between the flow and riverbed bacterial numbers in the absence of inputs from the catchment surface, combined sewer overflows, or urban runoff. This was the approach chosen, but a consideration of the other approaches helps to explain why the first option was the best choice.

While, the main aim of the study was to examine channel response without catchment inputs. The possibility of switching sampling from the Afon Rheidol to the adjacent unregulated channel of the Afon Ystwyth, if a suitable run-off event occurred was considered. This would have given the opportunity to sample during the passage of a natural flow event. It would also have overcome the expense of keeping sampling staff and a fully equipped and manned mobile laboratory on standby when an adequate natural event could not be guaranteed. No suitable natural event occurred and the reservoir release experiments were carried-out as programmed. 72



The third option of controlled experiments in an artificial channel would enable the examination of entrainment and settlement of organisms, but would have certain drawbacks. It would require the transport of bed substrate and organisms, for suspension, into the laboratory environment with all the attendant health risks associated with handling potentially pathogenic material. A large-scale flume, such as would be required for this study, is a closed-loop system; the channel water has to be recirculated, i.e. the output of the reach becomes the input. This may have implications for the survival conditions encountered by the bacterial load, as well as, being unrealistic in terms of the passage of a flow event. For example, in response to an increase in flow, the initial flood pulse of entrained organisms would be followed by the successively dispersed shadow as it passed repeatedly through the system. For these reasons it was decided to work in natural river channels.

3.3 The Field Experiments Having established the need to provide detailed event based faecal coliform data for the purpose of examining the behaviour of the organisms in response to flow events, and that this could best be achieved by artificial flow manipulation, a search for suitable sites was carried-out.

3.3.1 The Field Sites Potential methods for forcing artificial changes in flow include releases from upland impoundments and reservoirs or from hydroelectric schemes and even the outlets of interbasin transfer schemes. A large number of sites were identified, however, the eventual choice of sites was influenced by a combination of factors: i.

The nature of the channel morphology.

ii.

Permission of access and opportunities for artificial flow releases.

iii.

Access to laboratory facilities.

iv.

Time and funding available, particularly for microbiological enumeration which was one of the single highest costs during the study.

v.

Personal risk and the ease of collecting mid-stream samples.

73



The three sites that were eventually sampled are the Afon Rheidol in Mid-Wales, which flows through Aberystwyth into the Irish Sea, the Afon Clywedog in the headwaters of the River Severn and the River Washburn, a tributary of the River Wharfe north of Leeds.

ABERYSTWYTH

Afon Melindwr

Site 1: Blaengeuffordd

Site2

N

AFON RHEIDOL

0

Aberffrwd Dam

1 km

Map 1: Afon Rheidol site map.

In the Rheidol catchment, controlled releases were provided from the Rheidol Hydroelectric Scheme, operated by Powergen plc. The river was sampled on a straight reach at Blaengeuffordd (NGR, SN64008053, Site 1; Map 1) 8.9 km downstream of the Cwm Rheidol reservoir in the catchment flood plain. The reach is characterised by partially confined irregularly meandering pool riffle sequence with a bed slope of approximately 0.0029. The soil is clay loam. Faecal inputs are derived from the grazing of sheep and cattle, there is also a domestic effluent input from a small displacement activated works serving a population of approximately 200 people in the dormitory village of Capel Bangor. A daily programme of releases is made in the Rheidol to provide peak and off-peak power for export to the national grid. A consequence of this regime is that the system is well flushed and sediment movement through the system is limited by the various impoundments that comprise the system. Any organisms flushed by the experimental hydrograph might, therefore, be expected to represent inputs to the study reach in the period between releases.

On the River Washburn, releases are made between a series of four impoundments operated by Yorkshire Water plc. as part of their water resource management activities. Sampling for faecal coliforms was carried out during one of a number of regular releases, catering for white-water canoeists. The canoeing releases are made between Thruscross and Fewston Reservoirs (see Map 2). The channel in this reach is a pool riffle sequence with a slope of around 0.011 confined within a narrow valley. The channel banks are stabilised by trees, 74



boulders and in places by the underlying bedrock. The valley floor is covered by clay loam soil and is grazed by sheep and cattle. Either end of a central 1.9 km reach was sampled, from NGR SE15705695 (Site 1; Map 2) at the upstream site to SE16605540 (Site 2; Map2) just upstream of Fewston.

Site 1: Upstream

Direction of flow

N Site 2: Downstream 0

2km

RIVER WASHBURN

Map 2: River Washburn site map.

The Afon Clywedog in the Upper Severn was sampled at either end of a 3.6 km reach (NGR SN91408675 to SN94398553, Sites 1 and 2; Map 3) immediately downstream of the Llyn Clywedog reservoir, operated by the Severn-Trent Region of the Environment Agency. The reach is topographically confined and initially of step-pool configuration, rapidly changing to a pool riffle sequence of bed-slope 0.0079. The bank-side and valley slopes are grazed by sheep, inputs of domestic waste water are minimal. The soil is clay loam.

75



Site 1: Bryntail (upstream)

N

Site 2: Cribynau (downstream)

AFON CLYWEDOG 0

2k

Map 3: Afon Clywedog site map.

3.3.2 Sampling and Analysis Techniques Sampling was commenced prior to the releases in order to establish initial concentrations against which to compare the event response results. Release waters were also sampled at or just downstream of the reservoirs in order to characterise the quality of the water entering the study reaches. Stage, temperature and conductivity were recorded at each sampling interval. Samples were taken by wading, as near as was possible, to the centre of the channel flow and from approximately 0.6 of the flow depth. Samples of 400ml were collected in presterilised containers held near the base and plunged to the appropriate depth with the neck pointing upstream to avoid contamination by hand. Samples were stored in the dark prior to transportation to the laboratory for analysis. Standard UK methods were used (H.M.S.O., 1983: Page 46). In the laboratory samples were first diluted in 99 ml and 90 ml of sterile Ringers solution (Oxoid BR52) to provide serial dilutions. Generally 10 ml and 100 ml of original sample was filtered through 47 mm diameter 0.45 m membrane filters (Gelman GN6). The filters were placed on a Lauryl Sulphate Broth (Oxoid MM615) on absorbent pads in 55 mm petri dishes prior to pre-incubation four hours at 30C. The plates were then incubated for a further 14 hours incubation at 44C (± 0.25C). Thermotolerant coliform enumeration followed H.M.S.O. (1983: Section 7.9.4.2, Page 46). The count at 18 hours is technically a faecal coliform organism or thermotolerant coliform count (see H.M.S.O., 1983, Section 7.9.2, Page 45). No confirmatory procedure was adopted to define the 76



numbers of Escherichia coli within the overall thermotolerant coliform group enumerated. This is standard practice for raw waters (H.M.S.O., 1983: Page 46). From each sample six enumerations were made; three lots of 100ml and 10ml, respectively. The use of two filtering volumes ensures ease of counting both high and low numbers of organisms. Triplicate analysis improves the confidence interval about an estimate by a factor related to the square root of the number of replicates (Fleisher and McFadden, 1980; Fleisher et al., 1993), in this case, a 1.73 fold improvement in accuracy. The results are expressed as colony forming units (cfu) per 100ml. Duplicate samples were taken at intervals of 10 samples and blank control filtrations of Ringers solution were run for quality control purposes.

Since data were required for model identification and construction, a constant sampling interval was adopted for each field experiment. This is unlike previous studies (e.g. McDonald et al., 1982) where objectives were to highlight peak bacterial numbers during stage rise. The choice for a fixed or constant maximum time interval in this study was because a uniform sampling interval is required for the application of finite difference approximation techniques, which have been used in computerising model differential equations for development and testing purposes.

3.3.3 Experimental releases The experimental reservoir releases undertaken in this study were devised intentionally to follow-on from, and supplement, the work of McDonald et al. (1982) and Kay and McDonald (1982, 1980). Given the existence of these data it was decided that the resources available for fieldwork would best be utilised by generating a small number of new and more detailed event data sets, rather than a greater number of less intensively sampled events. This was done in order to capture more information about the very dynamic fluvial behaviour of faecal coliforms but at the same time recognising the cost of this strategy in terms of greater uncertainty surrounding a smaller sample of experiments.

The details of the four release experiments carried-out are as follows:

i. Afon Rheidol, 17 February 1993. The release was designed to simulate a natural hydrograph of the order that might be expected on the adjacent Afon Ystwyth which is 77



unregulated and of similar catchment area and bedrock geology (Milne et al., 1981). The operational constraints of the power station meant that the hydrograph had a maximum flow of 14.1 m3.sec-1, equivalent to 1200 kW, the maximum output from the generator set at the Aberffrwd Dam. This rose from 1.72 m3.sec-1, 100 kW, the around the minimum compensation flow permitted on the river. For ease of implementation the smooth design hydrograph was converted to a hydrograph with half hourly steps and values expressed in kW generated. This was a much simpler schedule for the generator operators to implement. This first sampling run on the Afon Rheidol produced 75 samples. Stage, temperature, conductivity were read at the same time as the water samples were taken, at 10 minute intervals prior to the rising limb of the hydrograph, when the interval was reduced to every 5 minutes, and reverted to 10 minute intervals on the falling limb. Sampling commenced at 09:20 HRS and continued until 18:30 HRS, after the hydrograph recession.

ii. Afon Rheidol, 7 April 1993. For this release a more complex discharge profile was designed, in order to examine specific effects. During the night preceding this second release there was a significant period of rainfall; 19 mm at Fron Goch, NGR SN 60508250, 3 km WNW of the sampled reach, and the highest daily rainfall in April. This raised the river level and turbidity, due to runoff from the adjacent catchment and sub-catchments. Since the laboratory was in place with microbiologists sampling went ahead as planned. With hindsight it might have been better to abandon this run and return after a period of dry weather. Sampling was carried-out at the previous location plus a second site 2 km further downstream (NGR SN62218089, Site 2; Map 1), and additional samples were taken just downstream of the Aberffrwd Dam (NGR SN69387944) to assess the bacteriological quality of the water entering the study reach. A total of 102 samples were collected, resulting in a total of 594 enumerations and 99 values; three sample pots were lost during transit.

iii. River Washburn, 26 May 1993. The third experiment coincided with a canoeing event for which the Thruscross Reservoir scour valves were opened-up to provide white water for the canoeists. This event provided what was in effect a step-change in flow, raising the stage by 30 cm upstream and 80 cm downstream in less than 30 seconds. The increase in discharge was from 5000 m3/day to 100,000 m3/day (0.058 to 1.57 m3/sec). Prior to the release samples were taken at 5 minute intervals to provide a background level against which to compare the effect of the flood-wave. On arrival of the flood-wave samples were collected at 2.5 minute intervals for the next 1.5 hours. This extended period of sampling at 78



high frequency was necessary to assess whether thorough flushing of the stream channel had occurred and that no unexpected increases in concentration were missed after the main response. A total of 71 samples were collected at the two sites.

iv. Afon Clywedog, 28 May 1993. This was the last of the experiments to be undertaken and involved a rapid re-deployment of the microbiologists and field sampling team to MidWales. This timing was necessitated by the operational requirements of the Severn-Trent Region of the Environment Agency who manage the Llyn Clywedog reservoir operations. The allowance for releases of water additional to compensation flow was 500,000 m3, and a fifth of this was allocated for this experiment with the stipulation that the reservoir must be at 100 % capacity (or spilling) before the additional release could be made. Thus the amplitude and duration of the event was designed to take-in this limitation, with the total design release equalling 90,000 m3, a margin of 10 % less than the allocation, in order to give leeway for operational errors. The hydrograph was built-up from around 1.47 m3.sec-1 in a series of discrete steps, each held for a period of half an hour, to a peak discharge of 12.10 m3.sec-1 after 1.5 hours, receding to compensation flow over the next 5 hours. Sampling commenced at 10:10 HRS and continued until 16:30 HRS. Samples were taken every 5 minutes on the rising limb of the hydrograph and then every 10 minutes on the recession. One hundred and seven samples were collected from the Afon Clywedog during this last experiment.

The release flow data and experimental results are presented in Chapter 4 and Chapter 5.

3.4 The River Dee, Huntington Intake The River Dee is a major source of water for domestic and industrial consumption, Rushbrooke and Beaumont (1984) noted that the Dee provides as much water each day as all of the Lake District reservoirs. The mean annual rainfall in the Dee headwaters near Bala is around 2500 mm and only 600 mm near Chester (Changxing et al., 1999), the catchment area is 1816 km2 (Weston, 1979) producing a mean annual flow of approximately 37 m3.sec1

(Lambert, 1988). The Huntington Water Treatment Works (see Map 4) provides an

average of 365 000 m3.d-1 of potable water (Taylor et al., 1993) and is protected by an intake protection system (Rushbrooke and Beaumont, 1984). The situation at the Huntington raw water intake on the River Dee is relatively complicated. The river is very highly regulated, Chester Weir, which is approximately 5 km downstream of the intake, results in 79



the river backing-up as far as 20 km upstream with low slope of 0.000005 (Weston 1979). The river also suffers from tidal reversals which range from minor "blips" to backwater effects extending as far as 30 km back upstream of Chester Weir. A spring tide flow reversal during low flow can result in a peak reverse flow approaching 60 m3.sec-1 but such reversals typically last only 1.25 hours. The water driven into the channel only drains out slowly and does not empty before the next tide arrives. From Chester Weir downstream the Dee estuary is 40 km long and regulation flows help to guard against the build-up of salinity in the estuary, which was more common prior to 1965 when Llyn Celyn was completed. The residual flow at Chester Weir is only 4.2 m3.sec-1 and peak chloride levels in excess of 2000 mg.l-1 have been reported (Weston, 1979; Sumner, 1982) indicating that sea water occasionally reaches this location.

Map 4: River Dee site map.

Sumner (1982) suggests that around 29% of high tides top Chester Weir resulting in rapid increases in turbidity, ammonia and bacteriological counts, and periodic cessation of pumping at the Dee-side abstraction; Heronbridge, which is downstream of Huntington is only affected to a minor extent. The water quality of the Dee is typical of water that drains both upland and lowland catchments (Taylor et al., 1993), with variations in quality resulting from both season and weather patterns.

80



Ammonia monitoring on the Dee has been used to indicate possible pollution from combined sewer overflows (CSOs), waste water treatment works (WwTWs), farm slurry, agricultural run-off and other sources for concentrations in excess of 0.3 mg.l-1 (Rushbrooke and Beaumont, 1984). Ammonia levels are higher during the winter, summer nitrification may result in lower concentrations and the occasional peaks result from agricultural run-off (Taylor et al. 1993). pH is depressed during the winter months in response to high flows as might be expected (e.g. Wilkinson et al. 1997; Neal et al. 1998).

Sumner (1982), Nuttall (1982) and Nuttall and Parry (1987) report on E.coli concentrations in the River Dee. Sumner (1982) reports concentrations of between 0 and 55 000 cfu per 100 ml with a median of 925 for the Sutton Hall raw water abstracted at the Heronbridge and Dee-side intakes between 1970 and 1976; tidally affected samples were included in these statistics. Nuttall (1982) took samples at three sites including Iron Bridge, and two sites further upstream, from early 1976 to 1978. The most upstream site was just below Bala and the second site at Overton upstream of the Dee meanders. At Iron Bridge the river is very slow flowing and receives 0.47 m3.sec-1 of sewage effluent; the effluent included waste water from 20 major sources within the catchment (Nuttall, 1982; Nuttall and Parry, 1987). In addition Nuttall (1982) mentions samples taken over a 24 hour period from the Huntington Works intake, unfortunately these data were not reported in that 1982 publication, and these have been lost (Nuttall pers. comm., 2000). Nuttall (1982) reported a significant seasonal pattern in E.coli concentrations at Iron Bridge with fluctuations in count of up to 1000-fold, with reduced numbers through the summer months increasing in the Autumn and remaining elevated throughout the winter months. Summer-time turbidity of 5 NTU or less is conducive to efficient self-purification of the river by settlement or die-off in the slower flowing reaches (see Nuttall, 1982; see also the bacterial die-off estimates presented in Section 6.2). Nuttall also found that each of the three sites monitored responded to faecal contamination on the same dates, the largest peaks were found to coincide with increased inflows at the largest of the sewage works in the catchment. Other important factors contributing to the bacterial load in the Dee listed by Nuttall and Parry (1987) include the scour of sewage effluent, storm-water discharges, leisure activities and nonpoint discharges of agricultural effluent.

Changxing et al. (1999), Gurnell et al. (1998) and Gurnell (1997) have been studying the geomorphology of the Dee meanders between Bangor-is-y-Coed and Farndon. The main 81



focus of the study (Gurnell, 1997) is the interaction of tides, non-synchronous mainstream and tributary flows and hydraulic structures such as weirs with the transmission of water and sediment during floods and their impact on bank sedimentation. Gurnell (1997) showed that there was a zone of preferential in-channel sediment storage, after a large flood, near the head of the tidally influenced zone of the Dee. Other than these studies there is no other recent work on the Dee reported in the literature. The data presented in Gurnell (1997) was used to calculate water depth at Huntington Intake (Chapter 4) this was used in Chapter 6 to estimate bacterial die-off rates. Gurnell (1997) noted that the combination of tidal weirs and reservoir regulation is common on heavily managed European rivers and represents an important area for research.

Chapter 6 describes the process of estimating bacterial die-off in the River Dee and the process of attempting to model the bacterial concentrations.

3.5 Methodologies for Model Development A two-point approach was taken to model development making use of the two types of data available. The first stage was to examine the fluvial dynamics model component. Secondly, die-off equations were developed using literature data, these were used with the long-term monitoring data to estimate die-off over seasonal and diurnal cycles. A brief description of these stages is given below. Chapter 6 develops these procedures and demonstrates their results.

3.4.1 Development of the fluvial dynamics model During the development of the entrainment model a variety of approaches were investigated before a successful or satisfactory result was achieved. These included applying the Jenkins model to the event data, modification of the AMV dispersion model to incorporate the Jenkins type, and alternative, riverbed store functions, as well as, recursive time-series analysis to identify a "black-box" model from the event data. It soon became apparent that these approaches would not be capable of reproducing the behaviour observed during the field experiments. The eventual model structure was devised to simulate the observed behaviour in response to the artificial hydrograph events. Having identified a model structure that satisfied the functional criteria indicated by the experimental data, a simple formulation of the model was applied in the finite difference backwards approximation (see 82



Young and Wallis, 1993). A simple parameterisation was carried-out in order to investigate whether the model performed as expected. The results of this initial investigation are reported by Wilkinson et al. (1995a). The model was found to function satisfactorily but required further attention to the parameterisation. This was in order to characterise the experimental reach dispersion properties, apply a die-off rate consistent with the literature, and to undertake flow modelling (Chapter 4). In addition, a variety of methods for initialising the model and relating discharge (Q) to entrainment were investigated in single and multiple reach versions of the model (Chapter 5).

3.4.2 Die-off and longer term modelling Literature data for die-off related to different environmental variables was used to develop equations for the effect of temperature and pH, adequate equations for sunlight and die-off were already available. The equations were devised in such a way that once an initial value of the total die-off rate was estimated, the die-off rate could then adjust automatically according to the three driving variables, sunlight, temperature and pH. Seasonal data from existing water quality monitoring data was used to investigate long-term die-off and was incorporated into the overall model structure. Additional environmental data on rainfall, air temperature, sunshine hours and solar radiation was collated and used to reconstruct the water temperature record and the solar radiation (irradiance) record (see Section 6.2). The reconstructed data and raw values were used in investigations of seasonal and diurnal dieoff processes. The die-off data was used in combination with a fully dynamic mixing zone model of the River Dee between Iron Bridge and the Huntington Intake. The rainfall data was used in a runoff model designed to estimate the runoff from the 800 km2 unregulated part of the catchment downstream of the Manley Hall gauge. This and the other hydrological work is described in Chapter 4.

83


4. Hydrology

4 Hydrology This chapter describes the estimation of discharges, wave speeds, flow velocities, water depths and rainfall-runoff required for the interpretation of the reservoir release responses (Chapter 5) the estimation of light attenuation for calculations of die-off in the River Dee (Chapter 6) and the development of a catchment delivery model.

4.1 Open channel flow and wave propagation The relationship between wave speed and mean flow velocity is of particular relevance to this study. The step-changes in discharge used in the reservoir release experiments resulted in steep fronted waves that propagated downstream disturbing and entraining faecal coliforms (see Chapter 5). The nature of the waves depended on the magnitude of the increases (and decreases) in discharge relative to the preceding flow condition. The literature describing these hydraulic phenomenon spans many decades (e.g. Rouse, 1946 and Martin and McCutcheon, 1998), and the terms used to classify different types of wave has evolved accordingly. This section attempts to summarise this literature and highlight the features of greatest relevance to the interpretation of the reservoir release experiments.

84


4. Hydrology

4.1.1 Classification of flows Dingman (1984) indicated the scale of forces acting in open-channel flows spanning six orders of magnitude, i.e. from a trickle solid surface to the open ocean. For flows in the range encountered in this thesis the dominant forces are gravity balanced by turbulence. Classifications of the flow can be useful since they provide a quick measure of certain welldefined characteristics. In this study the focus is on boundary layer flows, where there is a fixed boundary, i.e. the channel, and the flow velocity at the boundary is zero.

Open channel flows are classified according to the relative magnitude of gravity, viscous and turbulent forces, the rate of change of mean velocity or depth with time, and the rate of change of the mean velocity or depth in the downstream direction. The ratio Vy/v, where V is velocity, y is depth and v viscosity is known as the Reynolds Number (Re); for open channel flows y is replaced by R, the hydraulic radius A/P, where A is the flow cross section area and P is the wetted perimeter. For Re < 500 viscous force dominates and the flow is laminar; the flow viscosity damps out disturbances to the flow. For Re > 500 the turbulent force becomes significant, the flow may appear laminar, but disturbances may grow into full turbulence or subside depending on the frequency, amplitude and persistence of the disturbance. For Re > 2000 turbulent force dominates and the flow will be fully turbulent. The flow is said to be transitional for 500 < Re < 2000. Most natural flows are well within the turbulent state (Dingman, 1984).

The second classification of open channel flows is the ratio of turbulent to gravitational forces: V / gy, this ratio is known as the Froude number, F, and is the ratio of the wave speed to the velocity of the flowing water. For F < 1, the wave speed exceeds the flow velocity and waves caused by disturbances can propagate upstream. Such flows are said to be sub-critical. In supercritical flows, F > 1, the flow velocity exceeds the wave speed and waves cannot propagate upstream. For F = 1 the flow is said to be critical. Sub-critical flows are described as tranquil and super critical flows are described as rapid or shooting. For flows in channels of irregular cross section the mean depth, or hydraulic depth (y) is used, where y  A / w and w is the top width of the flow.

85


4. Hydrology

4.1.2 Kinematic and gravity waves The only types of wave of significance to hydrological studies are the gravity wave and the kinematic wave (Dingman, 1984). Table 4.1 gives a summary of the main features of the two types of wave. A gravity wave is one type of dynamic wave, other dynamic waves include those caused by elastic forces or surface tension (Dingman, 1984). A good example of a dynamic wave is the ripples caused by a stone dropped onto the surface of a pond (Malley and Ward-Smith, 1998). Surges, such as a tidal bore or a hydraulic jump, are examples of gravity waves (Dingman, 1984). These phenomenon persist as long as the forcing conditions remain and only occur in response to large sudden changes in discharge (Henderson, 1966; Chow, 1959) such as might result from hydro-power operations, canal gate activity or other similar rapid changes in flow. The step features of the hydrographs observed in the Washburn and the Clywedog experiments had characteristics of surge type gravity waves (Figures 5.1 and 5.2), although there is a kinematic phenomenon that exhibits similar characteristics. Section 4.1.3 discusses the effects of these very rapid changes in discharge in more depth.

Table 4.1 A comparison of the most important features of gravity waves and kinematic waves in natural channels (after Dingman (1984)). Gravity Wave

Kinematic Wave

Cause

Any disturbance that causes a local change in elevation of the water surface

Local accumulation of water, for example resulting from an influx of runoff

Celerity

Vg = v   g y and cg =  g y

Direction

Upstream and downstream

Dissipation

Frictional resistance causes rapid dissipation. Hydraulic jumps persist as long the forcing conditions persist

ck = (5/3) v Downstream only Gradual dissipation

Chow (1959) refers to kinematic wave flow as uniformly progressive flow, the main characteristic of which is that it does not change shape as it propagates downstream. Dingman (1984) describes the kinematic flood wave as a "translatory wave of stable form... that travels down the channel at constant velocity". In theory, the front of the kinematic flood wave will tend to steepen, since the deeper components of the wave travel faster than the shallower ones (Henderson, 1966). The wave ultimately takes on a stable form, since dispersion and attenuation effects tend to halt further steepening of the wave front (Henderson, 1966). Another important feature of the kinematic wave is that the wave height 86


4. Hydrology

tends to be conserved as the wave propagates downstream (Martin and McCutcheon, 1998). No forces or accelerations are involved in kinematic wave translation. The net forces are zero and kinematics is in fact the study of motion in the absence of forces (Martin and McCutcheon, 1998). The overall form of the hydrographs on the Rheidol and the Clywedog, i.e. the broad variation in discharge over the period of the whole of each experiment, exhibited kinematic behaviour (Figures 4.4, 4.6 and 4.10).

For a kinematic wave, the speed of the wave is proportional to the rate of change of discharge with depth, this can be evaluated by using the Manning equation (see Dingman, 1984). In a uniform rectangular channel, if the slope and roughness of the channel are constant, ck = 5/3 v, where ck is the kinematic wave velocity measured relative to the channel bank (Bedient and Huber, 1988). A natural flood wave is a kinematic wave that has dynamic waves moving ahead and behind (Henderson, 1966). The kinematic wave speed will be less than the leading dynamic wave for F < 2. The kinematic wave assumptions are only violated at very flat slopes or for very large increases in flow, and for normal floods the dynamic wave fronts attenuate very rapidly (see Henderson, 1966). In natural channels the relationship between mean flow velocity and kinematic wave speed varies according to channel geometry (Dingman, 1984).

Special cases of gravity and kinematic waves are relevant to the reservoir release experiments carried-out in this study. The following section introduces these special cases.

4.1.3 Steep fronted waves The step increases in discharge in the Washburn and the Clywedog resulted in steep fronted flow waves propagating down those channels. The two types of wave already discussed above, the gravity wave and the kinematic wave can have a steep front. Gravity wave types of steep fronted wave include surges, bores and moving hydraulic jumps (Malley and WardSmith, 1998). Kinematic type steep fronted waves are referred to as the monoclinal wave (Chow, 1959) or the Kinematic Shock (Henderson, 1966). This text will use the term "kinematic shock" first used by Lighthill and Whitman (1955). Chow (1959) suggested that such wavefronts can be assumed to be made up of a series of very small waves placed one on top of the other. The top most waves have greater depth and hence velocity, and will

87


4. Hydrology

overtake the lower waves in a forward direction. The waves tend to combine and form a single large wavefront which is steep and stable.

Dingman (1984) relates gravity and kinematic waves using Froude number. In shallow water gravity waves have celerity (or wave speed measured relative to the moving liquid), cg = gy. Tidal bores are an example of an abrupt gravity wave (Bedient and Huber, 1988). The downstream velocity of a gravity wave is Vg = v + cg. A kinematic wave in a rectangular channel travels at ck = 5/3 v, thus when cg = 2/3 v, Vg = ck, the two waves have the same downstream velocity relative to the stream bank. From the definition of the Froude number, Fr = v / cg, when Fr = 3/2, Vg = ck. For Fr > 3/2, Vg < ck and for Fr < 3/2, Vg > ck . When Fr > 3/2, flows become unstable and roll waves form spontaneously (Dingman, 1984). This may have been the case in the Washburn, unfortunately no mean velocity value is available for that experiment so it is difficult assess based on wave speed and velocity. Martin and McCutcheon (1998) suggest that the dynamic wave, as forerunner to the main kinematic wave, may push material ahead of it along the wave front in a "first flush" effect. Rouse (1946) used the relative depths of the initial flow, y1, and the wave height, y = y2 - y1 to give an indication of the type of wavefront that is likely to form. If y / y1 < 1 an undular wave will form. This has a wavefront asymptotic to the water surface ahead of the wave and smooth curvilinear waves following the front. If y / y1 > 1, the wavefront will break resulting in a sharp discontinuity in the water surface.

Similarly, the overrun critical depth can be used to assess the likely nature of a flow wave (Henderson, 1966; Chow, 1959). A wavefront propagating along a channel has the same discharge ahead of and immediately behind it, because the water with the increased discharge is travelling more slowly than the wave. This is termed the overrun discharge (Chow, 1959), Qr = ( c - v1 ) A1 at the front and behind the wavefront Qr = ( c - v2 ) A2 (Henderson, 1966). Calculation of the overrun-critical-depth yc = 3 Qr2 / g gives an indication of the likely behaviour of the wave. If yc > y1 (the downstream depth), then the wavefront will have a near-vertical shock front (Henderson, 1966; Chow, 1959).

The measure of critical depth or relative wave height for the Washburn (Section 5.1.1) indicates that a steep fronted kinematic shock would be expected. For the Clywedog stepped hydrograph the discharge increments were sufficient to result in shock fronted kinematic 88


4. Hydrology

waves (see Figure 4.10). On the Rheidol the release profile did not have such distinct steps, the study reach was very long, and the flood-plain reach had a shallow slope. The result of this was that by the time the wave front reached the sampling location, any shock features had dissipated leaving a smooth hydrograph (Figure 4.1).

4.2 Afon Rheidol 4.2.1 Discharge Discharge, Qi (m3.sec-1, or cumecs), into the study reach of the Afon Rheidol was estimated from a rating curve for the No. 4 Generator Set operated by Powergen plc. This generator provides all of the flow past the Aberffrwd Dam and the rating curve gives a relationship between power generated, P (kW), and flow in cubic feet per second Qi (cusecs). To calculate the discharge at Aberffrwd it was necessary to digitise this rating curve, fit a polynomial relationship to the curve and convert flows from cusecs to cumecs (Table 4.2).

Table 4.2 Rating of discharge Qi in cumecs at the Aberffrwd Dam, from power generated, P (kW), by the Cwm Rheidol Power Station Generator Set No.4. 1 digitised mm, mm to Qi cusecs Fit, Qi cusecs from (mm x 4) R2=0.9999 rating curve 0 8.6 34.3 33.3 100 15.2 60.8 60.8 200 22.4 89.7 91.0 300 30.1 120.5 123.7 400 39.7 158.6 158.6 500 48.7 194.8 195.5 600 58.6 234.2 234.4 700 67.9 271.5 275.0 800 78.7 314.6 317.1 900 89.4 357.8 360.6 1000 100.6 402.5 405.3 1100 111.5 446.2 450.9 1200 124.7 498.9 497.4 1 -7 3 -3 2 Fit, Qi (cusecs) = -0.3.10 P + 0.14.10 P + 0.262P + 33.279

P (kW generated)

Qi cumecs (Qi cusecs/35.316) 0.94 1.72 2.58 3.50 4.49 5.54 6.64 7.79 8.98 10.21 11.48 12.77 14.09

Discharge at the sampling location was equally straightforward to calculate, the channel section approximated to a shallow sided trapezoidal channel over the range of observed discharges (Table 4.3). This resulted in an approximately linear stage discharge relationship (Figure 4.2). The release discharge was found to propagate down the channel with minimal 89


4. Hydrology

alteration in shape other than the averaging out of the step increments (Figure 4.1). This was the desired outcome since a shape as close to a natural hydrograph as possible was originally intended. The step increments were necessary for ease of implementation by the dam operators. In order to estimate the downstream discharge at the sampling site the release discharge was transformed using a simple symmetrical moving average scheme taking the 16 59

14

Stage, cm

10 8 39 6 4

29 h 19 7:00

Discharge, cumecs

12 49

Qi

Qi ave

2 0

9:00

11:00

13:00

15:00

17:00

19:00

Figure 4.1 Release discharge from Aberffrwd Dam, Qi, smoothed Qi (Qi ave[rage]), and downstream stage at the sampling site on the Afon Rheidol.

mean of two bands of four values either side of the central point. Plotting this smoothed input discharge with the stage data so that the two traces have the same y-axis amplitude, it can be seen that they are almost identical, other than at the very beginning of the hydrograph. The apparent loss of the initial rise in the stage trace may have been the result of a number of factors: the faster travel times of the deeper waves catching-up with the initial increases, the dissipation of these small initial increases with transport downstream, and the lack of sensitivity of the stage measuring technique for detecting very small variations.

Table 4.3 Observed and estimated flow characteristics for the Afon Rheidol artificial hydrograph and bacterial peaks. Stage, h (cm)

Discharge, Time lag Wave Qi between, speed, (cumecs) flow wave c = 1.175

Velocity of Ratio, bacterial v/c peaks,

Crosssection area,

A = by + y2/s (trapezoidal

90


24.5 25.5 29.5 35 39.5 44.5 50 54 55.4

and bacterial peaks (minutes) 25.6 26.2 27.7 28.9 29.5 29.9 30.4 30.6 30.7

2.02 2.40 3.94 6.02 7.47 8.93 11.08 12.44 13.26

4. Hydrology

Q0.0911 (m/s)

v = 0.4544 c + 0.461 (m/s)

1.25 1.27 1.33 1.38 1.41 1.43 1.46 1.48 1.49

1.03 1.04 1.07 1.09 1.10 1.11 1.13 1.13 1.14

channel), b = 4.5 m s = 0.025 (m2) 2.1 2.3 3.1 4.5 5.8 7.4 9.5 11.1 11.7

A = Q/v (m2)

0.822 0.817 0.801 0.788 0.781 0.776 0.770 0.766 0.764

2.0 2.3 3.7 5.5 6.8 8.0 9.8 11.0 11.7

Downstream discharge was calculated from the stage discharge relationship, producing a discharge time-series retaining the features of the observed stage. The error between the

1.6

14

1.5

12 Discharge, cumecs

Wave speed, c

discharge volume upstream and downstream was +2.5%.

1.4 1.3 1.2 1.1

10 8 6 4

1.0

2

0.9

0 0

5

10

15

y = 0.3546x - 6.5936 2 R = 0.9987

20

30

Discharge, Qo

40

50

60

Stage, cm

Figure 4.2 Wave speed and discharge, and stage-discharge relationship for the sampling site on the Afon Rheidol.

Water depth, y, in Table 4.3 was estimated from the stage discharge relationship by subtracting the zero discharge stage from the stage values and converting to SI units. The depth, y, was then used to estimated flow cross-section area, A, for a trapezoidal channel (Equation 2.26), the bank slope, s, and base width, b, were adjusted to give a fit to the crosssection area values calculated from A = Q/v. The discharge, travel-time and water depth values were used in the modelling work presented in Chapter 5.

91


4. Hydrology

4.2.2 Flow routing This section introduces a simple approach to flow routing that provides discharge estimates for multiple reaches along the Afon Rheidol experimental reach. It was also found that the dispersion term, b0, could be used directly in the mixing component of the fluvial dynamics model (Chapter 5). The mixing parameter provides the dispersion that spreads the entrained bacterial peaks in the wake of the flow wave.

The approach used was that applied in the Bedford River Ouse study reported by Whitehead et al. (1979) and is equivalent to the kinematic wave approximation of the St. Venant equations (see also Beven and Wood, 1993; and Young, 1986 for comments on this method). The method was introduced in Chapter 2, Section 2.3. The technique treats the propagation of a flood-wave as being analogous to the dispersion of a conservative tracer such that in the simplest form

dQ Q = (Q - Q) dt V i Which in the finite difference backwards approximation becomes

Qk  a1Qk 1  b0 Qi ,k  where,

a1  1  b0 , and b0  t

Qk 1 Vk 1

and k is the temporal index and t is the discretisation interval, such that if t=t at k then t=t+t at k+1, thus if Qk is the discharge at the current time increment then Qk-1 is the discharge at the pervious time interval, t-t.

Since, bo, is defined in terms of discharge and volume and as no reach volume data was available, the value was derived from the relationship between wave speed and discharge based upon the times travel (Table 4.3). V  Q.tt , where tt is travel time, or time of first arrival of the new discharge, and tt  L / c , where L is reach length, and c is wave-speed, such that

1 c V  Q.L , and b0  t c L

92


4. Hydrology

In application of the flow routing algorithm to the Afon Rheidol it was both necessary and convenient to divide the 8.9 km reach into sub-reaches that were multiples of the sampling interval. This was done for two reasons. Firstly, over the entire reach length there would be greater differences in discharge from the top of the reach to the bottom of the reach than for a smaller length interval. Hence the use of Q at the downstream end of the reach would not be representative of the conditions at other locations. Secondly, the choice of a distance equivalent to a multiple of sample intervals merely simplifies the transfer of data from one model node, or reach outlet, to the input of the next. 16 14

Discharge (cumecs)

12 10 8 6 4 2 0 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00

Figure 4.3 Time-series of discharge in 13 successive model reaches of 684.6 m length between Aberffrwd Dam to Blaengeuffordd on the Afon Rheidol.

In the Afon Rheidol the initial travel-time between the up- and downstream ends of the study reach was 26 t. The smallest integer subdivision of this was 13 t so the reach was split into 13 sub-reaches with equal length (684.6 m) and travel time of 2 t at the minimum flow. Since there is little variation in travel time through the study reach the assumption of a fixed time delay of two time intervals has little impact on the flow routing result. Figure 4.3 shows the reach-by-reach routed flows in the Rheidol for the period of the experimental hydrograph. The discharge from stage at the sampling location calculated in Section 4.1.2 is shown to give a comparison of the two methods. Since no gauging was undertaken at the sampling site it has not been possible to verify the accuracy of the routed flow estimates. The absolute accuracy of the flow estimates is not crucial to this study and the values should 93


4. Hydrology

be taken as indicative of the likely discharge at the sampling site. The timing of the routed discharge was the most important consideration because of the relevance it has to the corresponding bacterial behaviour (see Chapter 5).

An alternative means of parameterising this simple flow routing algorithm which allowed volume, V, and discharge, Q, in the routing parameter, bo, to be varied "independently" was to use the water depth at the sampling location to derive a reach volume, V=AL (see Chapter 2, Section 2.3). L is reach length and A is channel cross-section area which was approximately trapezoidal at the sampling location, (Equation 2.26). Now the routing parameter is

b0  t

Qk 1 . Ak 1 L

16 14

Discharge, cumecs

12

Qi Qo' Qo

10 8 6 4 2 0 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00

Figure 4.4 Time-series of release discharge, Qi, at Aberffrwd and routed discharge, Qo', at Blaengeuffordd. Figure 4.4 demonstrates the routing result that best coincides with the timing and form of the discharge from stage at presented in Section 4.2.1 (Qo in Figure 4.4). This second method resulted in a reduced maximum flow compared to the first method and greater attenuation of the central peak flow of the stepped release hydrograph. The main assumption using this method is that the volume calculated from the stage at the sampling site, represents the volume in the entire reach. It was necessary to take a fraction of the reach volume to fit the model. The cross-section area constants are given in Table 4.3 above. 94


4. Hydrology

For the second experiment on the Rheidol (7 April 1993) discharge at Blaengeuffordd (Site 1), Qo, had to be calculated afresh. This was necessitated by the rainfall-runoff component, and the fact that at the peak of the artificial event the stage at Blaengeuffordd was much higher, for a given discharge, than during the previous experiment. This was probably due to the greater overall volume of water in the channel, and reduced water surface slope, because the second and larger flow peak was preceded by a smaller one (Figure 4.5). This might have a "backing-up" effect that required a greater stage, and hence head of water, to maintain a given flow.

D ischarge , Q i cume cs

14 12

Qi

10 8 6 4 2 0 6 :43

9 :07

11 :31

13:55

Tim e HRS :M IN

Figure 4.5 Release discharge from Aberffrwd Dam on Afon Rheidol, 7 April 1993.

In order to estimate Qo, the routing algorithm used for the first event was applied to the second set of release data. A power relationship was calculated for paired values of Qo' and ho, the observed stage at Blaengeuffordd, and used to calculate Qo'' from ho, thus providing estimates of discharge in a time-series retaining the original features of the stage record. Comparison of the stage derived discharge and the routed flow shows that the differences in discharge decline with time, the stage derived discharge being greater than the routed value, initially. The rainfall-runoff would have added to the release discharge during the early phase of this second experiment (Figure 4.6). The stage-derived flow, Qo'', was substituted for the routed flow and routed further, to the second sampling site. This Site 2 routed flow, Q2', was not rated to the observed stage, h2.

95


4. Hydrology

12

Discharge, cumecs

11 10 9

Qo' Qo'' Q2'

8 7 6 5 4 9:07

10:19

11:31

12:43

13:55

15:07

16:19

Time HRS:MIN

Figure 4.6 Routed discharge, Qo', and discharge from stage, Qo'', at Blaengeuffordd, and Q2' is Qo'' routed to Site2.

4.3 Afon Clywedog Discharge The sampling locations in the Afon Clywedog were chosen because they both had rated control structures giving estimates of discharge in the study reach. The upstream site, Bryntail, is the location where the input flow to the study reach, Qi, was controlled by the reservoir operators according to the programmed release devised prior to the experiments. The stage discharge relationship presented in Figure 4.7 was derived from the paired values of stage, hi, and discharge, Qi, in megalitres per day, representing the programmed steps in the release hydrograph. Discharge values were converted to cumecs prior to calculating the rating curve. Readings of stage from the weir stage board were taken at the same time as samples, with additional stage readings in between samples to give stage values every 2.5 minutes in order to improve capture of the changes in flow. This proved to be a worthwhile exercise since the timing of the different stages of the actual release, hi obs, did not coincide with the programmed release, hi prog., (see Figure 4.8). The first programmed rise in stage was to have been at 10:30 am. When sampling commenced at 10:05 am the stage was already above the background level of 0.3 metres. Since the bacterial concentrations of the upstream site compared to Cribynau, the second sampling site downstream, were significantly lower, the loss of this initial data did not affect the findings of the study.

96


4. Hydrology

14

Discharge (cumecs)

12

Qi = 26.072hi2 - 0.403hi - 0.7563 R2 = 0.9995

10 8 6 4 2 0 0.25

0.45

0.65

Stage, hi (m)

Figure 4.7 Stage discharge relationship for Bryntail Weir. 0.75

hi prog Stage, h metres

0.65

hi obs

0.55

0.45

0.35

0.25 9:30

10:30

11:30

12:30

13:30

14:30

15:30

16:30

17:30

Time HRS:MIN

Figure 4.8 Programmed stage, hi prog, and observed stage, hi obs, at Bryntail Weir on Afon Clywedog. The compound Crump weir at Cribynau, the downstream sampling site, had not been calibrated since the late 1970's, the Environment Agency (pers. comm.) indicated that no gauging had been undertaken after 1979 and that the rating curve should be used with caution. The weir has a three part rating (Table 4.4, Figure 4.9) given by the standard rating curve Q = C ( h  a ) n where C and n are constants and a is the stage at zero flow relative to the different components of the weir.

97


4. Hydrology

Table 4.4 Rating parameters for the compound Crump weir at Cribynau, the second sampling site on Afon Clywedog. a, stage at zero flow

C, constant multiplier

n, constant power

-0.0783 0.0308 -0.0696

3.223763 2.714501 3.090559

1.08831 1.27612 3.81467

Stage, ho, at transition to next weir component 0.374 1.067 -

Stage, ho metres

10

1

Q=3.22(ho-0.0783)^1.088 Q=2.71(ho+0.0308)^1.276 Q=3.09(ho-0.0696)^3.815

0.1 0.1

1

10

100

Discharge cumecs

Figure 4.9 Three part rating curve for the compound Crump weir at Cribynau on Afon Clywedog. The legend shows the rating for the component parts of the curve (see also Table 4.4).

The discharge values calculated from this rating were, in general, consistent with those at Bryntail upstream (Figure 4.10). The base flow values were greater, as would be expected, but there was some attenuation of the higher discharges, above 9 cumecs. Causes of this discrepancy may include changes in bed form that have affected the rating and errors associated with the inaccuracy of reading a staff gauge rather than having a stage recorder and stilling well. The nature of the Crump weir is such that relatively large changes in discharge at higher flow result in only small changes in stage as can be seen in Figure 4.10. These errors are once again not critical to the findings of the study, but highlight the importance of adequate stage recording to eliminate one source of uncertainty in the results. 98


4. Hydrology

14 Qi

12

Qo

2.6

ho

8 2.1 6 4

1.6

2 0

Stage , ho (m)

Discharge cume cs

10

1.1

-2 -4 9:36

0.6 10:48

12:00

13:12

14:24

15:36

16:48

18:00

T ime HR S:M IN

Figure 4.10 Discharge in cumecs at Bryntail, Qi, and Cribynau, Qo. Stage at Cribynau, ho, is also shown to demonstrate the stage distorting effect of the Crump weir.

4.4 River Dee, depths, travel-times, local runoff This section describes work to derive a depth discharge relationship, examine times of travel at different discharges and to estimate a local rainfall runoff component for faecal coliform delivery from the immediate catchment draining to Huntington Intake (see Map 4). A depth discharge relationship was necessary for the calculation of die-off estimates, since water depth is an important factor influencing light penetration of the water column (see Chapter 6). Times of travel were used to investigate the delivery of organisms to the study reach; the 90% die-off travel distance, X90, gives an informative measure of travel time combined with die-off (Chapter 6). Travel time data provided flow velocities for the input to the Iron Bridge end of the study reach and discharge divided by channel cross section area was used for velocity in the Huntington reach.

4.4.1 Water Depth at Huntington Intake Water level data for 1995 have been used to estimate water depth at Huntington Intake and to generate a depth discharge relationship. Figure 4.11 presents raw level data for Chester Weir and Iron Bridge and discharge at Manley Hall; the influence of the tides is clearly 99


4. Hydrology

visible, level values are measured relative to a local Ordnance Survey Datum. The first stage of the operation was to remove the tidal component from the level records. This was achieved by taking a 10 hour centred minimum. This effectively clipped-off the tidal blips (Figure 4.12). Figure 4.13 demonstrates a consistent relationship between the water level at Chester Weir and that at Iron Bridge. The differences in level between the two sites were plotted against discharge and a polynomial equation used to produce a line of best fit (Figure 4.14). The lines of points deviating from the main bulk of values are the result of backwater effects at high flows with high tides. The relationship between H (the stage difference from Iron Bridge downstream to Chester Weir) and discharge (Q) is,

H  a1Q 3  a 2 Q 2  a3 Q  C , where a1 = -5.5.10-7, a2 = 1.2.10-4, a3 = 5.49.10-3 and C = -2.214.10-2, for discharge Q > 170 m3.sec-1, H = 1.68.

Chester Weir is approximately 6 km downstream of the Huntington Intake, and the reach between Iron Bridge and Chester Weir is 12 km long. For a low flow of Q = 12 m3.sec-1 at Manley Hall, the level difference between Iron Bridge and Chester Weir is 6.2 cm, a slope of 1:194 000, which is consistent with the values presented by Weston (1979). The water surface elevation between the two sites was assumed to vary linearly with distance downstream. The Huntington Intake is approximately half way between Iron Bridge and Chester Weir. The water level at Huntington, therefore, at Huntington would be the level at Chester Weir plus half of the elevation difference over the entire 12 km reach. Then by subtracting the bed elevation at Huntington, the water depth could be estimated. Just upstream of Farndon, the channel-bed elevation is approximately 1.6 m AOD (Gurnell, 1997), with a very shallow slope steepening gradually in an upstream direction. If the bedslope is assumed to be equal to the water-surface slope at low flow, projecting the channel bed elevation forward 14 km to Huntington would give a value of approximately 1.53 m AOD.

100

4. Hydrology

350

7

300

6

250

5

200 Discharge, Manley Hall

150

Ironbridge, level 100

4 3

Level (mAOD)

Discharge, Q (cumecs)


Chester Weir, level

50

2

0 Dec-94 Feb-95 Mar-95 May-95 Jul-95 Aug-95 Oct-95 Dec-95

1

Figure 4.11 Observed discharge and water surface levels on the River Dee for 1995.

7 Ironbridge level

Level (mAOD)

6.5

Chester Weir level

6 5.5 5 4.5 4 Dec-94

Mar-95

Jul-95

Oct-95

Figure 4.12 Water surface levels with the tidal component removed. The small drops in stage at Chester Weir are due to upstream abstractions.

101


4. Hydrology

Chester Weir level (mAOD)

5.1 5 4.9 4.8 4.7 4.6 4.5 4.4 4.3 4.2 4

4.5

5

5.5

6

6.5

7

Ironbridge level (mAOD) Figure 4.13 Chester Weir level against Iron Bridge level (tidal component removed).

2.5

Level difference (m)

2

1.5

1

0.5

0 0

50

100

150

200

250


Figure 4.14 Difference in water surface level between Chester Weir and Iron Bridge plotted against rated discharge at Manley Hall.

Having estimated water depth over a wide range of flows it was possible to derive a depth discharge relationship for use in estimating die-off in the Huntington reach. Figure 4.15 102


4. Hydrology

shows estimated depth, z (m), at Huntington plotted against discharge, the depth discharge relationship is z  0.000083Q 2  0.001477Q 1.75  0.004856Q 1.5  0.001655Q  2.73 . 5

Depth, z (m)

4.5

4

3.5

3

2.5 0

50

100

150

200

250


Figure 4.15 Estimated water depth at Huntington against discharge at Manley Hall.

Table 4.5 Width, depth and channel cross-section area just upstream of Farndon on the River Dee (based on data presented in Gurnell (1997)). Width

Mean depth

Cross-section 2

Associated

(m)

(m)

area (m )

discharge (m3/s)

Near bank-full

35

4.4-4.6

150-170

114-341

Intermediate

30-32

4

100-130

40-82

Low level

24-30

3-3.4

80-100

n/a

Gurnell (1997) reported survey data collected in 1973 by the then Dee and Clwyd River Authority for multiple channel cross-sections along the meandering section of the river in the flood-plain above Farndon. Reach characteristics were surveyed relative to three levels in the river; near bank-full, an intermediate level and a lower level. The downstream end of the reach was within the zone of tidal influence and backwater effects from Chester Weir. The approximate channel widths, mean depths and cross-sectional area are presented in Table 4.5 and are consistent with the depths presented in Figure 4.15.

103


4. Hydrology

4.4.2 Travel times, velocity and mixing parameters Travel time values for a range of discharges were digitised from plots presented by Lambert (1983). The estimates were based on tracer experiments carried-out in June 1982 by Welsh Water and Lancaster University (Rushbrooke and Beaumont, 1984) and adjusted with data from pollution incidents (see Lambert, 1983, for a full description of this study). The discharge, travel time and velocity data are presented in Table 4.6, the channel length between Manley Hall and Farndon is approximately 35 km, thus the velocity is given by travel time divided by 35 km. Fitting the data with a standard power law v = aQ b renders the coefficients a = 0.0474 and b = 0.7717.

Table 4.6 Discharge, travel time and mean flow velocity in the River Dee. Discharge (m3/sec) at

Travel Time of event peaks,

Mean velocity

Manley Hall

Manley Hall to Farndon (hours)

(m/sec)

12

29

0.335

14

25

0.390

16.5

20

0.486

20

17.5

0.556

40

12.4

0.748

80

8

1.215

100

6

1.620

Table 4.7 summarises the depth, cross section area and velocity values estimated for a range of discharges.

The velocities presented in Table 4.6 represent the behaviour in the natural channel upstream of the reach between Iron Bridge and Chester Weir. The ponded Iron Bridge to Chester Weir reach is much deeper than upstream and consequently the velocities in the lower reach cannot be expected to be as high as from Manley Hall to Farndon. A more representative velocity was calculated from Q / A (Table 4.6). The values of water depth calculated for Huntington Intake were used to estimate channel cross section area on the assumption of a trapezoidal channel with base width 20 m and bank slope of 0.3 (see Section 4.2). The channel cross section area was multiplied by the reach length to give reach volume and the mixing volume parameter estimated from Q / V.Df. 104


4. Hydrology

Table 4.7 Estimates of water depth, channel cross section area, flow velocities, reach volume and mixing parameters for the River Dee. Discharge, Depth, z Q (m) (m3/s) 8 16 32 64 128

2.80 2.90 3.11 3.52 4.14

Huntington Cross section area, velocity, v=Q/A A (m2) (m/s) 82.2 0.097 86.0 0.186 94.5 0.339 111.8 0.572 139.9 0.915

Farndon velocity, v = aQb (m) 0.235 0.401 0.684 1.166 1.988

Reach volume, V = A.L (m3) 493316 516072 566905 671052 839448

Mixing parameter, Q / V.Df (1/s) 0.000054 0.000103 0.000188 0.000318 0.000508

4.4.3 A Rainfall Run-off model for catchment delivery The River Dee has a total catchment area of 1816 km2 (Weston, 1979), the catchment area to the flow gauge at Manley Hall is only 1019.3 km2 (Weston, 1979). Inflows from the Rivers Clywedog and Alyn as well as Aldford and Pulford Brooks drain a large part of the additional area of approximately 800 km2 downstream of Manley Hall. This section describes the adaptation of a rainfall runoff model which was used to estimate the discharge in the remaining 800 km2 catchment.

The River Dee drains Snowdonia National Park where annual rainfall is around 2500 mm. The headwaters are augmented by discharges from three man-made lakes (see Chapter 3). The highly regulated discharge passes through the Vale of Llangollen and meanders across the Cheshire Plain. At Manley Hall the discharge record clearly reflects the regulation with periods of very flat constant flow at low flows (Figure 4.16). The subsequent 800 km2 part of the catchment is unregulated in the upper reaches and receives around 760 mm rainfall annually (Weston, 1979). This un-gauged lower area must add to the overall discharge arriving at Huntington Intake and due to spatial variability in rainfall there might be peaks that would not be observed at Manley Hall. In order to provide time-series of discharge at Huntington Intake, rainfall from a nearby site was used to generate a local run-off component. This component was used in the catchment run-off bacterial input term to the river model presented in Chapter 6. The rainfall used was that for Ness Gardens, 17 km to the north-west. The mean annual rainfall at Ness Gardens for the six year period from 1 January 1986 to 31 December 1992 was 767 mm, coinciding with the value quoted by Weston (1979).

105


4. Hydrology

The rainfall runoff model used was based-on the principle of IHACRES (e.g. Jakeman et al., 1990). The model is a simple black-box transfer function model with three main components; a loss module for evapotranspiration, and a high-pass and low pass filter for quick and slow flow run-off components. The loss module was based-on that presented by Littlewood and Jakeman (1994). The modified version presented here includes solar radiation, since evaporation is a function solar energy received (see Penman, 1948), such that

  I T rk* = 1  0.5 0   I max Tmax 

 rk 

where rk is the measured rainfall, r*k is rainfall minus evapotranspiration, T and I0 are air temperature and mean daily irradiance (see Section 6.2.1.1), respectively. The soil moisture status is represented by

  *  w1 r s k =  1 1  k  1  (1   w ) z  where w is the time constant of the soil moisture store. The value for w was taken from the mean of values presented in Littlewood and Jakeman (1993) for the River Exe in Devon and the River Teifi in Wales, 10 days and 15 days, respectively. The model is presented in the finite difference backwards approximation (see Young and Wallis, 1993). In the discrete time form, zn is the "z-operator" and is a mathematical convenience to aid notation and to help simplify equations. For example, if the rainfall at time t, rt = rk and the sampling interval is given by t, then the rainfall at t + t is rk+1, which is z1rk, similarly z-1rk = rk-1, i.e. the previous value of rainfall. Finally, the effective runoff volume, uk, is the product of effective rainfall (m) and catchment area (m2);

uk = A

sk * rk s max

where smax is used to scale soil moisture storage between 1 and 0. In the IHACRES model where the output runoff volume is known, a scaling constant is used to adjust the effective rainfall volume to the observed discharge volume (Littlewood and Jakeman, 1994). In this application the run-off volume is unknown; the requirement was for a run-off series representative of local rainfall effects.

The effective run-off volume, is routed through the two-box transfer function model to produce a resultant flow with a quick and slow flow component given by 106


Qik =

4. Hydrology

bq bs  uk , 1 1  as z 1  a q z 1

Equation 4.4.3.1

the model differs only slightly from the IHACRES model in that the user decides the separation between fast and slow flow and the time constant, T. The parameter relationships are, a q = e

 t . q

,

bq = (1 - aq) and a s = e

 t . s

bs = (1 - as)(1 - ), where,  = 1 / T, , is the quick-flow /

slow-flow ratio for the separation of the effective discharge volume, uk, and has a value of between 1 and 0. 0

60

5 10 15

40

20 30

25 30

20

Rainfall, mm

Discharge, cumecs

50

35 40

10

45 Qi, local runoff 0 Jan-87

Qs, slowflow

r* 50

Apr-87

Jul-87

Oct-87

Jan-88

Apr-88

Jul-88

Oct-88

Figure 4.16 Estimated inflow Qi and slow-flow component Qs, with rainfall r*.

Ideally for an un-gauged catchment, quick and slow-flow time constants would be based on values for a nearby gauged catchment of similar size, topography, soil, bedrock and landuse etc. The mean of the time constants for the Teifi and Exe were used in this application (Table 4.8). Figure 4.16 shows the estimated additional discharge, Qi, for the 800 km2 continuation of the catchment, the slow flow component Qs and the effective rainfall, r*. Figure 4.17 shows the total discharge for Huntington Intake with the lower discharge at Manley Hall and effective rainfall, r*.

107


4. Hydrology

Table 4.8 Values of IHACRES parameters for modelled catchments presented in Littlewood and Jakeman (1993).Time constants are in days unless otherwise specified. Catchment

Area, km2

Vq

Vs

Tq

Ts

Iq

Is

Tw

River Exe

601

0.67

0.33

3.1

53

0.97

0.03

10

Afon Teifi

894

0.67

0.33

2.7

51

0.97

0.03

15

Nant y Gronwen

0.7

0.4

0.6

4.4 hr

90 hr

0.92

0.08

86 hr

Current study

800

0.6

0.4

2.7

51

0.97

0.03

12.5*

200.0

0

180.0

5

160.0

10

140.0

15

Q, Manley Hall

120.0

20

Q, Huntington

100.0

25

r* 80.0

30

60.0

35

40.0

40

20.0

45

0.0 Jan-87

Rainfall, mm

Discharge, cumecs

*Tw is the mean of the values for the Exe and Teifi.

50 Apr-87

Jul-87

Oct-87

Jan-88

Apr-88

Jul-88

Oct-88

Figure 4.17 Estimated discharge Qik, with discharge at Manley Hall on the River Dee and adjusted rainfall r*.

Table 4.9 Summary of parameters of the rainfall runoff model for 20 minute and daily data.

 = 0.6

 -6

Quick-flow

4.287.10

Slow-flow

2.27.10-7

T

Daily data

20 minute data

days

a

b

a

2.7

0.6905

0.1857

0.9949

0.00308

51

0.9806

0.00776

0.9997

0.00011

b

Table 4.10 Summary of rainfall and runoff values for 1991, for 20 minute and daily data (r and r* are the rainfall and effective rainfall annual averages over the period of record, 1986-92). Note that Weston (1979) quoted values of r=760 mm and mean discharge Q = 37 m3.sec-1. 108


4. Hydrology

r = 767

Daily data

20 minute data

Q

Q

r*=560

Qq

Qs

Qi=Qq+Qs

Qq

Qs

Qi=Qq+Qs

Manley

Huntington

Average

7.38

4.22

11.6

7.58

4.55

12.09

27.7

39.77

5 percentile

0.169

2.81

3.83

0.166

3.24

3.99

10.07

14.79

99 percentile

31.87

7.95

39.13

34.96

7.95

45.59

104.5

129.51

Tables 4.9 and 4.10 summarise the model parameters and results for the two different time scales. The mean discharge estimated for Huntington Intake is consistent with the value of 37 m3.sec-1 quoted by Weston (1979).

The quick and slow-flow components of the estimated additional discharge were used directly at the two different time-scales in attempts to model the bacterial concentration at Huntington Intake (Chapter 6). The slow-flow, Qs, was used to provide a continuous baseflow input of organisms and Qq, the quick-flow used to drive a supply-limited delivery model in line with the observations of Hunter et al. (1992, 1999) (see also Chapter 2).

109


5. FC Fluvial Dynamics

5 Faecal Coliform Fluvial Dynamics This chapter describes the results of the artificial hydrograph experiments outlined in Chapter 3. A description of the results of each individual experiment is given, followed by a discussion of the results as a whole, and their interpretation, in the context of the possible entrainment processes occurring. The development of the entrainment model is presented and finally model performance is discussed with examples from modelling the faecal coliform behaviour observed in the field experiments.

5.1 Results of the reservoir release experiments The results of each of the field experiments carried-out are presented below and the potential sources of faecal coliforms within the stream channels is discussed, including a consideration of possible entrainment mechanisms. An examination of observations from related fields; fine sediment transport and the study of algae in rivers, provides useful insights for inferences about faecal coliform behaviour in stream channels. Individual site descriptions are given in Chapter 3 and flow routing is outlined in Chapter 4.

5.1.1 River Washburn This event provided what was in effect a "step-input" response resulting from the opening of the control valves on Thruscross Dam. Figure 5.1 demonstrates the steepness of the wavefront. The flow wave was heard some time before it arrived at the downstream site and was described as being like a tidal bore. It was necessary to retreat to the bank prior to the 110



arrival of the wave; the stage rose from 10 cm to 80 cm in less than 30 seconds. The water of the wavefront was visibly very turbid, but the high turbidity declined rapidly with the passage of the front. 2.0

1200 Downstream response

1.8 1.6 1.4

800

1.2 1.0

600

0.8 400

FC2

Stage, m

FC (cfu/100ml)

1000

0.6

h2

0.4

200

0.2 0

0.0

400

1.2 Upstream response

FC1

h1

0.9

200 0.6 0 16:00

16:30

17:00

0.3 17:30

Figure 5.1 Faecal coliform responses observed at the upstream and downstream ends of a 2 km reach of the River Washburn to a step increase in stage.

Table 5.1 gives a brief statistical summary of the data for the two sites. The data is separated into three periods; an initial period of low concentration prior to the arrival of the flow wavefront, a steep peak in concentration coinciding with the wavefront, followed by a rapid decline to a steady but elevated concentration. Similar characteristics were reported by McDonald et al., (1982) and Kay and McDonald (1980a, 1982). Figure 5.1 is plotted with identical y-axis scales to highlight the downstream accumulation of entrained organisms propagating with the wavefront. The true bacterial peak downstream may not have been captured during sampling as suggested by the truncated appearance. Kay and McDonald (1982) sampled the wavefront on the Tryweryn very intensively to capture the bacterial peak. In the current study a 10 to 15 second sampling interval might have been necessary to capture the peak. A simple interpolation, the intersect of the slopes of the points on either side of the peak, would suggest that the maximum concentration might have been in excess 111



of 1600 cfu per 100ml. Following the passage of the wavefront, the faecal coliform concentration declined rapidly within 5 minutes to the near steady post-peak levels (Table 5.1).

Table 5.1 Geometric mean, maximum and minimum faecal coliform concentrations at two sampling sites on the River Washburn, during a step change in discharge, 26 May 1993. Statistics are given for the full data period and the three distinct phases of the event. All values

Pre-arrival

Wavefront

Post-peak

Site 1 FC1

GM (n)

112 (41)

37 (4)

161 (3)

124 (34)

Upstream

Max, min

195, 28

46, 28

195, 131

160, 98

Log10 Std. Dev.

0.176

0.108

0.086

0.059

Site 2 FC2

GM (n)

179 (30)

61 (4)

895 (3)

175 (23)

Downstream

Max, min

1093, 51

72, 51

1093, 630

281, 133

Log10 Std. Dev.

0.297

0.082

0.132

0.085

Table 5.2 Times of travel and calculated velocities for various features of the reservoir release response of the River Washburn 26 May 1993. A mean velocity was inferred from the arrival of the second minor bacterial peak. This coincided with the expected timing based-on kinematic theory but was not confirmed by direct measurement. Reach length = 2000 m

Travel time,

Velocity, m/s

minutes Wavefront rise mid-point 16:37:24.5 hrs (ck)

22.4

1.488

Bacterial peak 16:40:00 hrs

23.75

1.404

Minor bacterial peak 16:52:30 hrs

37.5

0.889

Mean velocity from kinematic wave speed, v = 3/5 ck

37.3

0.893

The fact the bulk of the entrained organisms were concentrated at the flow wavefront might suggest that the water flow velocity was the same as the wave speed. This is not likely given the discussion in Section 4.1. Assuming the wave speed were greater than the flow velocity, and if it was behaving as a kinematic shock as suggested in Section 4.1.3, the bacterial peak would be expected to arrive some 15 minutes after the wave front (Table 5.2). A small secondary bulge in the response occurred at around 16:52, this coincides with the estimated mean flow velocity (Table 5.2). The responses in the Tryweryn (Kay and McDonald, 1982) were also to step increases in discharge (stage rises of around 0.3 m), the faecal coliform 112



concentration peaked simultaneously with the stage rise. This was also the case for a previous release in the Washburn (McDonald et. al., 1982). The Tryweryn has a slope of 0.0108 similar to that of the Washburn (slope = 0.0111).

The significance of considering the kinematic shock in terms of a moving hydraulic jump is that the flow appears to pass from super-critical ahead of the wave to sub-critical behind the wave; the jump is a moving point of criticality (Abbott, 1979). To recap from Section 4.1.3, examining the propagation of the wave as a hydraulic jump in a lagrangian frame of reference, i.e. the observer travels with the wave, the fluid would appear to travel from the smaller depth y1 with discharge, Q1, into the greater depth, y2. However, since the wavefront travels ahead of the mean water movement there is a zone on the deeper (upstream) side of the wavefront that has depth y2 but not discharge Q2, because this is lagging behind the wavefront. This zone still has the initial smaller discharge of Q1, this is the overrun with discharge Qr = Q1, cross-section area, A2 and velocity, v2. The effect of this is to produce a zone of suction, causing lift and turbulence. Water is drawn from ahead into the overrun as the wave passes downstream. Rouse (1946, pages 144 to 146) demonstrates this turbulence to great effect. The wave can be imagined to be suction front, the turbulence associated with the front being so extreme (and in this case very close to the channel bed, y2 >> y1) that it violently disturbs the bed surface detaching the available organisms and particles which are then locked into the wavefront. A proportion of the entrained material will escape to flow at v2 and arrive at the sampling location after the wavefront (Table 5.2). In situations where the wave height is small with respect to initial depth, y1, the potential for the turbulence caused by the wave to pull material directly into the wavefront is likely to be diminished.

Following the passage of the wavefront, a large proportion of the available organisms might have been entrained and flushed out of the channel. Previous studies have suggested this (Kay and McDonald, 1982) and McDonald et al. (1982) suggested that the channel supply of organisms was finite for any specific discharge. This behaviour is exhibited in data presented in Kay and McDonald (1980a), where successive hydrographs and multiple rises in stage resulted in successive bacterial peaks. Jeje et al. (1991) observed this kind of effect for multiple storm-flow peaks and the entrainment of sediment where peaks of concentration occurred with each flow peak. This is an important observation and the Clywedog reservoir release highlights this phenomenon.

113



Finally, the continued elevation of the FC concentration at both sampling sites above the pre-release concentrations following the passage of the wavefront suggests an additional source of organisms (Figure 5.1, Table 5.2) or perhaps just continued entrainment under the fast flowing conditions established for canoeing. For instance turbulent bursting has been shown to cause a kind of sporadic stochastic entrainment (Garcia et al., 1996).

5.1.2 Afon Clywedog The artificial hydrograph on the Afon Clywedog consisted of a series of steep step increases in discharge followed by a similar number of steps back down to the initial flow rate (Figure 5.2). The corresponding faecal coliform concentrations demonstrate the characteristics of initial low concentrations in the pre-release phase followed by a number of bacterial peaks on the hydrograph rise and into the recession (Figure 5.2). The concentrations downstream were a factor of ten greater than upstream (Table 5.3).

The bacterial peaks in the downstream faecal coliform response appear to be associated with the step changes in discharge, on both increases and decreases in stage. The bacterial peaks coincided with the flow wavefronts but also appeared after the passage of the wavefronts. There were two initial small peaks at 11:15 and 11:20 occurring on the first rising discharges (up to 5.8 cumecs). Note that this first major increase in discharge comprised two waves, the second faster wave had caught-up with the first. It is also important to stress at this point that the significance of these smaller bacterial peaks is not certain since they are defined by only one sample. Faecal coliform enumerations are known to produce noisy data, although with triplicate analysis the accuracy of the results is increased by a factor of 1.73 (see Fleisher and McFadden, 1980, Fleisher et al., 1993).

114


11:37:30

5. FC Fluvial Dynamics 14

11:55

2000

Downstream

12:10 12:50

12

11:20

FC2 Qo 8

1000

6

13:05 11:15

Flow (cumecs)

CFU per 100ml

10 1500

4 500 2 0

0 15

500 Upstream 0 10:00

11:00

FC1

12:00

13:00

14:00

15:00

Qi

16:00

0 17:00

Figure 5.2 Faecal coliform responses observed at upstream (Bryntail) and downstream (Cribynau) ends of a 3.6 km reach of the Afon Clywedog downstream of the Clywedog Dam (see Section 3.3.1, Map 3). Hydrograph rise samples were taken every 5 minutes and every 10 minutes on the recession. Note also that the first two discharge increments had merged by the time they reached Cribynau.

Table 5.3 Summary statistics for the observed faecal coliform concentrations at the upstream and downstream sampling sites on the Afon Clywedog. All values

Non-rise

Rise

Upstream

GM (n)

30 (52)

14 (24)

59 (28)

(Bryntail)

Log10 Std. Dev.

0.357

0.087

0.218

Max, min

258, 9

21, 9

258, 28

Downstream

GM (n)

355 (55)

147 (27)

832 (28)

(Cribynau)

Log10 Std. Dev.

0.444

0.122

0.301

Max, min

2033, 84

259, 84

2033, 193

Following a slight drop in concentration came the first major bacterial peak during the quasi-steady flow of 5.8 cumecs with a sampled peak at 11:37:30. The concentration then 115



declined slightly prior to a third peak at the rise in discharge from 5.8 to 8.4 cumecs. This was the last of the peaks that appeared to coincide with an increment in discharge; from midday up to the hydrograph recession the bacterial peaks occurred only during periods of quasi-steady flow. Following the onset of the hydrograph recession, at around 13:00, there was a further small peak in concentration (13:05, Figure 5.2).

2.5 c1 (pre-rise)

Velocity (v), wave speed (c) m/s

Wave speeds, c c2 (first rise)

2.0 c3 (mid-rise)

c = fit [mid-rise]

1.5 c = fit [c1,2,3]

v (FC peaks)

1.0 v' = 3/5 c(fit[mid-rise])

Flow velocities, v v = fit [FC peaks]

0.5 3

5

7

9

11

Discharge, Qo cumecs

Figure 5.3 Summary of wave speed and flow velocities for the monoclinal flow rises and resulting bacterial entrainment peaks on the Afon Clywedog (see Table 5.4).

The relevance of the timing of the peaks relative to the step increments in discharge only became apparent when modelling commenced and the distributed channel storage model proposed in Section 5.3 was tested. The initial modelling work (presented in Wilkinson et al. (1995a and b)) demonstrated the need for a fixed time delay between the arrival of the stage increments and the corresponding model bacterial peaks (see Figure 5.12). This delay placed the model peaks within a few minutes of the observed peaks (those at 11:37:30, 12:10 and 12:50 marked in bold in Figure 5.2). This initial modelling work fed-back to an improved interpretation of the data and the realisation of the importance of the flow routing theory presented in Chapter 4. An examination of the wave speeds and velocities associated with the travel times of the wave fronts and bacterial peaks confirmed a clear relationship between the two (Table 5.4, Figure 5.3). Thus, data interpretation and modelling proceeded

116



together as an iterative process. Modelling provided insights into data interpretation. This improved interpretation was used to enhance the model.

Wilkinson (1945) found that the mid-points of rise or fall stages were best for determining the velocity of an observed wave, these timings were used in the current study. An examination of kinematic wave theory (e.g. Henderson, 1966), shows that for a wide rectangular channel, the mean velocity of the moving water should be approximately 0.6 vk, the kinematic wave velocity. The corresponding values for a triangular channel and a wide parabolic channel are 0.752 vk and 0.694 vk (Chow, 1959). In this case, the ratio of wave speed to the velocity of the bacterial peaks varies from 0.7 to 0.65. This strongly suggests that the bulk of the entrained organisms were moving with the mean water velocity. There are a number of factors that might affect the ratio of the wave speed to flow velocity. Firstly the variation in form and roughness in a natural channel is likely to affect the flow velocity relative to the wave speed. The peak bacterial concentration might appear to travel faster than the mean velocity if the organisms are concentrated towards the sampling location, since the overall distribution will have a shorter distance to travel. The errors associated with a coarse sampling interval might mean that the estimated velocities are incorrect; this seems less likely given the consistent relationship shown in Figure 5.3. The number of points is also small but, again, the systematic relationship seems to discount this as a source of major error. Table 5.4 Summary of wave speeds and bacterial peak travel-velocities in m.sec-1 for the reservoir release experiment on the Afon Clywedog. Discharge Qo (m3/s)

c1 Pre-rise

c2 First rise

c3 Mid rise

Wave speed, Velocity, v, v' = 0.694 c Ratio, v/c c, from fit [mid-rise] bacterial peak

3.67 5.82 8.67 11.47

1.20

1.20 1.60 1.85 2.40

1.20 1.50 1.78 2.12

1.22 1.47 1.80 2.12

2.00 2.40

0.84 1.00 1.17 1.39

0.85 1.02 1.25 1.47

0.70 0.67 0.66 0.65

Figure 5.4 shows an approximately straight-line relationship between wave speed and peak bacterial travel velocities for the Clywedog. The linear relationship between wave speed and flow velocity was useful since it allowed the timing of the bacterial peaks with respect to the flow waves to be estimated accurately (Section 5.3).

117



1.5 1.4 12:50 Velocity of FC peaks

1.3 1.2 12:10

1.1 1.0 11:37:30 0.9 0.8

11:15

0.7 1.1

1.3

1.5

1.7

1.9

2.1

2.3

Wave speed, c

Figure 5.4 Velocity of observed bacterial peaks plotted against observed wave speeds including arrival times of bacterial peaks (values in m/s).

These results demonstrate behaviour unlike that observed in the Washburn. The Washburn release was so extreme that the majority of entrained organisms got locked into the travelling flow wavefront. In the Clywedog there were small peaks associated with the first three discharge increments (at 11:15, 11:20 and 11:55), but the majority of the organisms travelled at the mean flow velocity. The last rising wave was not accompanied by a bacterial peak. At the higher stages it is possible that the water depth is such that the wave transition disturbs the bed without drawing the organisms directly into the wavefront, leaving the material to be transported at the mean flow velocity. At the lower discharges, the waves disturb the organisms and the disturbance lifts a proportion of the organisms sufficiently to enter the wavefront. In the case of the Washburn, the initial stage was so low that the disturbance was only a small distance the bed, hence the organisms getting sucked into the wavefront. There was also a bacterial peak on the first falling stage. This wave was equivalent to a negative surge, these waves tend to dissipate and spread-out since the deeper wavelets travel faster than the shallower ones, so that the upper forward part of the wave travels away from the back of the wave. The wavefront presents a transition that is less severe than for a rising wave and the associated lift and turbulence is likely to be less. Hence the erosional capacity is reduced, add to this the contracting entrainment field of the flow, 118



the potential for entrainment on the Clywedog recession was greatly reduced. The last bacterial peak of 13:05, most likely reflects this process, the last contact, as the flow recedes, with a region that only encountered entrainment from the earlier transition from 8.69 to 12.1 cumecs. The remaining channel area within the field of entrainment underwent a total of four flow transitions and might be expected to be depleted to the extent that further disturbance would not entrain sufficient organisms to impact on the overlying water quality.

A similar effect is suggested in the data for the first Rheidol experiment, the results are presented below.

5.1.3 Afon Rheidol I Table 5.5 shows the overall geometric mean (GM) concentration of the data set plus the GM concentration during the hydrograph rise and either side of that period. The faecal coliform response can be split into two main phases as for the Clywedog; a period of enhanced concentration coinciding with the rising limb and peak of the hydrograph and lower concentrations following the peak flow.

Table 5.5 Geometric mean (GM) faecal coliform concentrations in Afon Rheidol artificial hydrograph event response water, 17 February 1993. All values

Non-rise

Hydrograph rise

GM (n)

76 (75)

57 (41)

108 (34)

Log10 Std. Dev.

0.21

0.14

0.18

Max, min

277, 33

192, 33

277, 38

The rising-limb bacterial response is characterised by a sequence of periods of enhanced concentration punctuated by sharp declines to near background levels. The peak concentration was 277 cfu/100ml the minimum was 38 cfu/100ml. The water entering the system from the Aberffrwd Dam had faecal coliform concentration of less than 30 cfu/100ml. Figures 5.5 and 5.6 show the faecal coliform load and concentration variations with time and discharge.

119



16

3.E+07 Qo

14

3.E+07

Qi FC load

2.E+07

10 8

2.E+07

6

FC load, cfu

Discharge, cumecs

12

1.E+07

4 5.E+06

2 0 07:30

0.E+00 09:30

11:30

13:30

15:30

17:30

Figure 5.5 Variations in faecal coliform load through the period of the first artificial flow event on the Afon Rheidol. Qi is release flow and Qo is discharge at the sampling site. 16

300 Qo

14

Qi FC

200

10 8

150

6

100

FC (cfu/100ml)

Discharge, cumecs

12

250

4 50

2 0 07:30

0 09:30

11:30

13:30

15:30

17:30

Figure 5.6 Variations in faecal coliform concentration through the period of the first artificial flow event on the Afon Rheidol.

Figure 5.5 demonstrates that the greatest bacterial load discharged occurs towards the hydrograph peak, suggesting that the greatest numbers of organisms are entrained during that period. The peak coliform concentration occurred before the first rise in stage was 120



detected. This suggests that the situation was not as straightforward as that observed in the Clywedog and Washburn (Figure 5.6). The concentrations that followed this first peak were all less.

Once again, the changes in faecal coliform concentration on the Rheidol are best explained relative to the timing of the flow wave propagation from the upstream to downstream sites. Section 4.2.1 presented the results of estimating discharge and travel times and highlighted the time lag between the flow and the slower water containing the entrained bacterial peaks. Other than "smearing-out" (or dispersion) the flow transitions the hydrograph did not significantly change in shape. This meant that it was possible to use simple graphical techniques to examine the bacterial response with respect to the changes in flow (Figure 5.7 and Figure 5.8). 15 380

FC Qi smoothed

330

10

Qi FC (cfu/100ml)

5 230 Peak 4, 12:25 Peak 2, 11:00

180

0

Peak 5, 12:55

Discharge, cumecs

Peak 1, 10:00

280

130 -5 80 Peak 3, 11:30

480

450

420

390

360

330

300

270

240

210

180

150

120

90

60

30

0

-10 -30

30

Figure 5.7 Time-shifted faecal coliform concentration and discharge plotted against time relative the start of the release and the first arrival of the event water downstream.

By modifying the stepped input hydrograph with a symmetrical moving average function a smoothed discharge trace with the same shape and form as the downstream stage record was generated (Section 4.2.1). Time-shifting the faecal coliform concentration data with the stage to coincide with the smoothed upstream discharge, demonstrated that peaks in the bacterial concentration follow the flow increments in a similar way to that seen in the Clywedog (Section 5.1.2). Unlike the Clywedog, however, the long reach length means that 121



the for discharges above 5 cumecs, the bacterial peaks arrived at the sampling site after the wave from the next increase in flow (Figure 5.8). Or to put this another way, the flow waves catch-up with and overtake the bacterial disturbances caused by the preceding flow increment. Since individual waves are undetectable at the sampling site it seems unlikely that the later waves would have sufficient energy, as they catch-up, to cause further entrainment into the pulse they catch-up with.

The variability in concentration on the hydrograph recession may reflect small disturbances by the step reductions in flow, but also random disturbances of the river bed caused by the continued high and turbulent flow, in a similar manner to that suggested for the Washburn. The cause of the peak at 15:30 (Figure 5.6) is uncertain, it may have been a sudden disturbance of material caused by cattle crossing and perhaps defecating directly the channel. Other potential sources of contamination include a small waste water treatment works that serves the hamlet of Capel Bangor, the works discharges on a pulse basis. This source may also explain the initial Peak 1 (Figure 5.7) that precedes the first detectable rise in stage (Figure 5.6). The routed discharge (Figure 4.3) suggests a rise in discharge coinciding with Peak 1 (Figure 5.6). Given that the stage values were taken from manual readings of a staff gauge without stilling, it is possible that some minor increase in stage may have passed unnoticed. It seems unlikely, however, that an imperceptible initial disturbance of the flow would be capable of producing a faecal coliform response of the magnitude of Peak 1. This likelihood is further negated by the suggestion of the other experimental results, that a major increase in discharge is required to cause significant entrainment of organisms stored in the channel-bed.

Compared to the Clywedog or Washburn, the concentrations observed in the Rheidol were much lower. This may have been due to the previous days high flow of 11 cumecs (see Figure 5.22), and perhaps also a low rate of bacterial input to the channel. The lack of a discrete bacterial peak on the next flow rise after Peak 1, may have been due to the size of Peak 1; it may simply have swamped the response to the next step rise in flow. Fewer samples were taken in this initial period so an individual response may have been missed.

122



350

Time, minutes

300 250 200 Wave arrivals

150

FC peak arrivals

100 0

5

10

15

Discharge, cumecs

Figure 5.8 Time of arrival of bacterial peaks and flow wave relative to the beginning of the release.

The gap in response between Peaks 3 and 4 to the discharge rise from values of 5 and 9 cumecs may have been a result of localised depletion of the channel supply of organisms, perhaps resulting from the previous days high discharge. If this was the case it would have implications for the mechanism of settlement. For example, if there are zones of preferential settlement in the channel, for example large dead-zone features that are flushed between specific flow ranges, then a response less regular than that observed in the Clywedog might be expected (see Carling et. al., 1994). Organisms do accumulate at different rates under different flow conditions as has been demonstrated in the Washburn (Jenkins, 1984). The question that remains to be answered is whether this heterogeneity of zones of accumulation can result in pulses of bacterial entrainment that are of sufficient magnitude not to be integrated out by mixing and longitudinal transport in the channel. The results presented in Section 5.3.2.1 suggest that this is not the case, i.e. that heterogeneity of storage zones is not important.

5.1.4 Afon Rheidol II A series of step and ramp features comprised the second experimental release on the Rheidol (Figure 4.5). During the second phase of the experiment, from 12:00 Hrs, a response of similar order to that of 17 February was observed (Figures 5.6 and 5.10). The channel bacterial response during the initial phase was overwhelmed by inputs from a storm 123



flow event resulting from a significant period of rainfall lasting up to a few hours before sampling commenced.

Table 5.6 Summary statistics of faecal coliform concentrations in the Afon Rheidol, during a combined storm-runoff and artificial flow event, 7 April 1993. Phase 1 includes storm runoff, Phase 2 is dominated by the artificial release discharge.

Site 1, FC1 Blaengeuffordd

Site 2, FC2

GM (n)

All values

Phase 1

Phase 2

455 (50)

1183 (21)

228 (29)

[9:30 to 11:50 Hrs]

[12:00 to 16:00 Hrs]

Log10 Std. Dev.

0.392

0.132

0.186

Max, min

1700, 119

1700, 567

441, 119

GM (n)

672 (49)

1770 (25)

245 (24)

[9:30 to 12:25 Hrs]

[12:30 to 16:30 Hrs]

New site,

Log10 Std. Dev.

0.456

0.188

0.064

downstream

Max, min

3233, 189

3233, 666

346, 189

Figure 5.9 shows a time-series of discharge and faecal coliform concentration for the two sampled locations. The bacterial response at both sites is characterised by two distinct phases; a period of enhanced concentrations followed by a period of smaller and less variable concentrations. Table 5.6 summarises the geometric mean (GM) concentrations for the two phases and the overall period of data for the two sites, as well as, the maximum and minimum enumerated values at both sites. Considering the response in terms of the two phases, the effect of the rainfall event, during the first phase, is clearly seen at the second sampling site (Figure 5.9). The initial FC concentrations were almost twice those at Blaengeufordd, the likely source of the elevated levels in this section of the river was a small tributary draining agricultural land north of the main channel. This tributary discharges into an old ox-bow, blind at the upstream end and connected to the main channel at the downstream end. The run-off into the ox-bow was highly turbid, consequently, the ox-bow water entering the main channel was also very turbid.

124



3500

2500

11

Qo''

10 9

2000

8 1500

7

1000

6

500

5

0

4

3500

12 11

FC2

3000

FC (cfu/100ml)

FC1

10

Q2'

2500

9 8

2000

7 1500

6

1000

5 4

500 0 9:20

Discharge, cumecs

FC (cfu/100ml)

3000

Discharge, cumecs

12

3 2 10:20

11:20

12:20

13:20

14:20

15:20

16:20

Time HRS:MIN

Figure 5.9 Time series plots of estimated discharge and observed faecal coliform concentration at Blaengeufordd (top) and Site 2, 7 April 1993.

The faecal coliform concentrations declined with increasing discharge as the reservoir release water diluted the inputs to the main river. Four samples taken at regular intervals during the release from the reservoir outlet, were found to be relatively pure. The FC concentrations of these samples were between 10 and 28 FC per 100ml, around one hundredth of the concentrations found downstream. The elevated levels found at the downstream sites during the storm response demonstrate the degree of bacterial contamination of the run-off entering the channel.

During the second phase of the response, which followed the decline in storm flow from the tributaries (from around midday onwards), FC concentrations returned to levels closer to those of the first experiment of 17 February (Figure 5.10). The geometric mean FC concentration of this second phase (Table 5.6, 228 cfu/100ml) was still approximately three times greater than that of the first experiment (Table 5.5, 76 cfu/100ml). The final peak in 125



concentration at Site 1, followed the last steep flow increment of this experiment (refer to Figure 4.5 where this step is more clearly apparent) in a similar way to the responses in the other experiments.

10

Qo

1500

8

1000

6 4

500

2

0 10:50

0 11:20

11:50

12:20

12:50

13:20

13:50

14:20

14:50

500

FC (cfu/100ml)

12

12

400

FC1

11

Qo''

10 9

300

8 7

200

6 100 12:00

Discharge, cumecs

FC (cfu/100ml)

FC2

2000

Discharge, cumecs

14

2500

5 12:30

13:00

13:30

14:00

14:30

15:00

15:30

16:00

Figure 5.10 The main faecal coliform responses on the Afon Clywedog (top) and Afon Rheidol (7 April 1993, bottom), displaying the similarity in faecal coliform response to artificial hydrographs.

At no point during the hydrograph rises was there a dilution by the cleaner upstream water to concentrations below the non-rise levels (Figure 5.10). This lack of dilution suggests a sufficient supply of organisms to mix into the flowing water and maintain the faecal coliform concentration above that of the incoming water from upstream. The troughs in concentration demonstrate temporary diminutions or depletions in supply prior to a further source becoming active and producing a further peak in concentration. The observed peaks and troughs in the first Rheidol experiment might suggest discontinuities in the supply of organisms from the channel over the range of observed discharges. These may reflect spatial heterogeneities in storage within the channel and the variety of sources different in nature both physically and/or biologically (Section 5.1.7 possible within channel sources of organisms). Alternatively, the response in the Rheidol may merely have been a result of the 126



already depleted channel supply of organisms, as indicated by the generally low concentrations. A final point worthy of consideration is that the Rheidol study reach is a lowland channel in the flood-plain with shallower bed-slope than the other two sites, thus direct comparison of the results should be undertaken with care.

In future data collection exercises for the purpose of further model testing, continuous stage, conductivity and turbidity measurements would be helpful to determine the exact timing of events.

5.1.5 Summary of responses On the Washburn a solitary steep shock-fronted kinematic wave resulted from a release for slalom canoeing. The wavefront was sampled and sampling was continued after it passed. The wave passed the downstream sampling site in under 30 seconds. The wave propagation is analogous to a moving hydraulic jump. The wave overrun, the zone between the wavefront and the slower bulk water flow, is a zone of extreme lift and turbulence. On the Washburn the shallow initial depth placed the overrun disturbance very close to the channel bed and the majority of the disturbed material was effectively sucked and locked into the "rolling" wavefront. A small proportion of the entrained material escaped to arrive later at mean flow velocity. Entrainment episodes associated with turbulent bursting is a likely cause of the elevated concentrations during the continued high flow.

On the Clywedog the artificial hydrograph was a series of shock fronted rising and then falling waves. Some limited wavefront entrapment of entrained material may have occurred in response to the first flow increments, but was not apparent at greater water depths. The majority of faecal coliforms lifted by the turbulent wavefronts was left in suspension to travel with the bulk flow at the mean flow velocity. The falling waves on the recession, almost identical to decaying negative surges, resulted in further entrainment but the effect was less pronounced than that observed on the rising waves. The reduced entrainment on falling waves was possibly due to depletion of the channel supply of organisms, but may also have been due to the spreading-out of the wavefront. The deeper wave components travel faster than the shallower ones, this causes the wave to steepen on a rising front and spread on a falling wave, as a result, the falling waves have a less dramatic impact on the channel as they propagate downstream. The responses on the Rheidol demonstrated similar 127



behaviour to that of the Clywedog and Washburn, steep rises in discharge caused episodes of entrainment with the entrainment peaks lagging behind the wavefronts.

5.1.6 Proposed storage / entrainment process The various experiments reported in the literature and the experiments described above have demonstrated consistent patterns in faecal coliform responses to artificial flow events and during natural hydrograph events.

A general outline process for the settlement, storage and entrainment of faecal coliforms (and particulate phases with parallel behaviour) is as follows: i.

Settlement: Attachment and or flocculation of organisms enhances their settlement; they settle throughout the channel but only deposit where flow conditions or a cohesive substrate prevents their immediate resuspension.

ii.

Storage: Organisms are stored everywhere within the channel wetted-perimeter on plant surfaces, cobbles and rocks, amongst sediment accumulations, as particles within macrophyte beds etc. (see Section 5.1.7).

iii.

Entrainment: Organisms stored within the channel are detached or lifted back in to the water column by changes in the nature of the flowing water above. Three modes of entrainment were suggested by the behaviours observed in the reservoir release experiments: a. Wave-front (overrun) entrapment (wave-front entrainment); where a steepfronted wave, with wave height much greater than the preceding water depth, effectively sucks and holds disturbed organisms in the turbulent wake of the wave. The disturbed material travels at the wave speed. This mechanism may also include the wave-front pushing suggested by Martin and McCutcheon (1998). b. Wave-front disturbance ("rapid" entrainment). The organisms lifted are not held in the overrun. The wave front may be less steep and the wave height small or not greater than the initial water depth. These flow transitions may merely lift the organisms into the water column to be transported at the mean flow velocity. Note that this wave-front mechanism is also relevant on falling waves, i.e. step reductions in discharge. c. Non wave-front turbulent disturbance ("slow" entrainment) of bed and bank sources. This third mode of entrainment results from continued turbulence 128



during elevated discharges. This tends to cause many small and irregular disturbances of bed and bank. The sum of these small disturbances is sufficient to maintain elevated faecal coliform concentrations above those encountered at lower rates of flow.

The initial interpretation of the reservoir release responses presented in Wilkinson et al. (1995a and b), involved a simple conceptualisation the spatial heterogeneities of channel form and bacterial sources within a channel. The source zones were considered as two symmetrical bands on either side of the central body of flow. At any steady flow there would be a central region depleted of organisms (coinciding with the thalweg). The bands on either side of this depleted core can accumulate settling organisms from the water column. These zones of accumulation become supply areas at higher discharges. Further entrainment can only occur if there is another increase in discharge. In this way the channel was assumed to maintain a flow memory which is gradually erased by settlement at low flows. This conceptualisation is replaced by entrainment type "b" above. This early approach produced satisfactory modelling results (see Section 5.3.1), but it is accepted that this was a naive representation of the observed faecal coliform responses.

In the case of the proposed processes above and in reference to steep fronted waves: Overrun entrapment (or wave-front pushing) may only be significant where the wavefront passes sufficiently close to the bed. Bank side regions may be the zones where this effect occurs to the greatest extent in natural channels, since this is where the waves come closest to the bed. Towards the centre of the channel where the water is deeper, lift and mean velocity transport is more likely to be the main mode of transport. On successive waves the bank side might be relatively depleted and wavefront entrainment peaks less likely.

Where a more gradual increase in discharge occurs, i.e. associated with a natural rainfallrunoff event, the occurrence of a steep-fronted kinematic shock is unlikely (Henderson, 1966). The mode of bed resuspension might be analogous to the idealised symmetrical model proposed by Wilkinson et al.. The turbulent field of influence of the flow spreading outwards and areas where conditions were previously sufficiently quiescent that no entrainment occurred encounter sufficient turbulence for material to be disturbed and reentrained. In natural events, however, the supply of organisms from catchment runoff and storm sewer overflows may swamp the response in the channel. In modelling the River Dee, 129



the channel entrainment component of the model was not required (see Section 6.3.2.2). In fact it is likely that each modelling application for faecal coliform dynamics will require a different combination of "plug-in" components to suit the situation at the site of interest. This theme of a modular approach to model application is expanded in Section 6.3 and Chapter 7.

5.1.7 Supporting observations from related fields Various evidence exists in the literature to give qualitative support to the process of storage and entrainment proposed above, this is presented below and, where relevant, related to the observations of the field experiments and those presented in the literature.

Gannon et al. (1983) and Auer and Niehaus (1993) studying settlement in North American Lakes found that 90% of settled organisms were associated with particles of clay or silt size (0.45 to 10 m diameter) and settlement velocity in the order of 1.2 m d-1. The Americans consider clay size particles to be non-settlable (APHA, 1981). Graham (1990), however, studying the siltation of clay-size particles suggested that particles "must deposit everywhere in a stream under all conditions of water velocity and turbulence". Reynolds (1979) proposed a model for the deposition of particles from turbulent flow based simply upon water depth and stokian (still-water) settling velocity of the particles. Particles, or in this case faecal coliforms, if we infer an analogous process, will only accumulate where the flow conditions, or a "sticky" substrate, precludes their immediate resuspension (McCave, 1984).

With regard to storage areas within a river channel, Jenkins (1984) investigated faecal coliform concentrations in the channel of the Washburn finding that organisms were "stored" on and in a variety of substrates and locations in the channel. A text of relevance for the consideration of the nature of a river channel for the storage of faecal coliforms is Reynolds (1992), referring to habitats for algae in a river. Reynolds suggests that since rivers are able to maintain a spatially and temporally diverse array of microhabitats they collectively offer an almost infinitely "patchy" environment. McCave (1984) refers to zones of accumulation with sticky surfaces in pool-riffle reaches, these include the mucilagenous surfaces of diatom stalks and the structural matrix of the periphyton into which particles can infiltrate and become trapped (Hoagland et. al., 1982). Epilithic algal growths on hard surfaces in the boundary layer use sticky, gelatinous secretions or alternatively "hold-fast" 130



stalks to attach firmly and maintain their positions even when main channel velocities are approaching 2 m.s-1 (Reynolds, 1992). It is likely that moving objects such and pebbles and stones would abrade and break-up epilithon releasing the material gradually. Reynolds (1992) also mentions episammic and epipelic algal groups, these attach to sand grains and fine sediments respectively in zones where the flow conditions permit the formation of temporary poorly consolidated gravel deposits. The species of algae concerned form loose mats or films of gliding or aggregating filaments, nets or clusters of unicells. These can represent a significant proportion of the fluvial microflora but are susceptible to increases in discharge and the associated rise in shear forces and rates of flushing. These groups are likely to swept into suspension rapidly and might represent a source for associated faecal coliforms that would be entrained by passing flow waves.

Wohl and Cenderelli (2000) and Beschta et al. (1981) demonstrate the differential disturbance of fine sediments and bed-load in response to reservoir releases, fine suspended sediment concentrations were observed to peak on the sharp rise of the hydrographs and decline very rapidly following the passage of the wavefront. Particles may also infiltrate the matrix of coarser sediment in a channel-bed (Milhous, 1982). Milhous (1982) proposed a simple conceptual model for the release of fines from the matrix of gravel-bed rivers, this also results in a wavefront response of the type observed in the Washburn. Wohl and Cenderelli (2000) and Beschta et al. (1981) present data that suggests bed-load motion was a more continuous and somewhat erratic process than the rapid entrainment of fine sediment. Such an erratic process might in part contribute to the variations in the elevated bacterial concentrations during continued periods of high flow along with the burst-sweep cycle process observed by Garcia et al., (1996). Wass and Leeks (1999) present continuous turbidity data for a natural hydrograph event in the River Don, UK. This event shows certain similarities with the reservoir release characteristics with a sharp increase in turbidity during the steepest part of the hydrograph, this peaks as the hydrograph rise slows, then declines to near constant elevated turbidity until the hydrograph recession when it declines exponentially to pre-storm values. This might suggest an initial flush of bed and nearchannel catchment material followed by a period of sustained erosion from channel and catchment sources. Unfortunately, Wass and Leeks do not present the data in the context of sediment fluvial dynamics through the course of the event. It does however present an interesting snapshot of a very large data-set collected as part of the LOIS programme (see

131



Wilkinson et al., 1997) and more importantly demonstrates similar behaviour in terms of concentration to the bacterial dynamics discussed here.

Neal et al., (1996) refer to a study by Jeje et al. (1991) and the work of Walling and Webb (1982) in their examination of acid available (AAP) trace metal concentrations in rivers entering the Humber Estuary. They link sediment exhaustion and hysteresis effects to the changes in AAP metals through hydrograph events, suggesting that during closely spaced storms suspended sediment concentrations during the second event may be much lower than on the first event due to exhaustion effects.

Plants in a river can act as source and sink areas for fine sediments. The blades and stems of macrophytes, if sufficiently dense, can provide a microenvironment of reduced velocity and turbulence suitable for the accumulation of fine particles (HRL, 1985; Dawson, 1985). Jenkins (1984) investigated the behaviour of E.coli within the bed of the River Washburn finding that the organisms were heterogeneously distributed. In areas of sediment accumulation concentrations of organisms increased during low flow and decreased during higher discharges. At bare-bed sites the behaviour was similar but less pronounced. Jenkins found that weed covered sites in slow flowing water accumulated organisms during low discharges and became depleted at higher discharge. In fast flowing water, however, this behaviour was reversed, the accumulation occurred at higher discharges when organisms entrained elsewhere were effectively combed-out of the flow by the weeds. Macro-phytic algae become firmly attached to stones and rocks and other solid surfaces, although the upper limits of their tolerance to turbulence and velocity are not well quantified (Reynolds, 1992).

Carling et al. (1994) and Tipping et al. (1993) highlight the importance of dead-zone features as preferential storage areas for fine particles and associated contaminants. Deadzones may include well defined pools, regions of retardation associated with channel banks, water within macrophyte beds and quiescent water behind cobbles and boulders. The potential for the replacement of low flow dead-zone features by volume equivalent new dead-zone features at higher flows (Carling et al., 1994) might be one way that zones of preferential accumulation and supply are maintained over a range of flows. Entrained organisms from one area may settle-out in other quiescent areas to be resuspended once more on a higher discharge. 132



Reynolds (1992) suggested that the distribution of algal types in suspension to some extent represent the various components of the benthic algal assemblage relative to their susceptibilities to flushing-out by the flow. In upland streams Reynolds suggests that the principal means of disturbing algae is through enhanced stream-flow resulting in the movement of stones, and the compression of the boundary layer which exposes more prominent algae to mechanical stress. In lowland reaches where flow velocities are slower, Reynolds suggests that enhanced discharge may have sufficient effect to resuspend fine materials. Reynolds concludes that these effects will be generically similar from river to river but the specific behaviour will be unique to each system.

Possible alternative channel derived sources of organisms include erosion of the channel banks and of cattle faecal deposits on the channel margins and bars. The latter may certainly be a source along the Rheidol, where cattle gather to drink from gravel bars, here faecal material can accumulate and be washed away when inundated during rising flows.

Whether bank erosion is likely to be a significant source of organisms in uncertain. Kay and McDonald, (1980) produced a hydrograph with minimal stage rise, intended to disturb the bed of the Washburn without causing bank-wash, this produced a faecal coliform response of similar magnitude to previous events which had included both bed and bank disturbance. They suggested that the channel-bed was the dominant source of organisms.

Soils can contain in excess of a few hundred faecal coliforms per gram (e.g. Hunter and McDonald, 1991a) this material may contribute to the overall channel loading at high flows, but the process of entrainment might be of the gradual episodic / erratic type associated with turbulent bursting and cobble movement.

Lawler (1995), gives a good overview of bank erosion processes. The erosional processes are split into three main groups; sub-aerial preparation processes, direct fluid entrainment and mass failure processes. Freeze-thaw action and desiccation are both sub-aerial processes that either cause transfer of material to the flow or prepare it for fluvial removal. Freezethaw action tends to decline with altitude and hence distance downstream, in conditions of inversion this pattern can be reversed. Desiccation tends to be related to summer air

133



temperature and rates of evapotranspiration. These processes act externally to the river and are related to microclimatic variability.

Direct fluid entrainment of bank material is related to boundary shear stresses; for which Lawler uses stream power as a loose approximation. Stream power reaches a maximum in the mid-basin zone with minima in the flood-plain and headwaters. The dominance of fluid entrainment processes mid-basin is reinforced by the progressive reduction in bank material grain size with distance downstream. Headwater sediment is commonly coarse and flood plain sediments often cohesive clay materials, whereas middle reaches may have fine and coarse sand which are more readily eroded (Lawler, 1995).

The third group of bank erosion effects is mass failure (see Lawler, 1995 and Ashbridge, 1995); the collapse of soil blocks from the banks. Catastrophic failure can lead to sudden slugs of fine sediment entering the flow, or fallen blocks may erode gradually "leaking" material over long periods. The study of stability for bank failure is well advanced and critical bank heights above which failure is a likelihood can be defined. Denser materials have a smaller critical height, this is offset in increasingly cohesive materials which have higher critical heights. Mass failure is dominantly a lowland reach process where meandering channels cut through fine alluvial material with high banks. Since none of the field sites studied in the reservoir release experiments were in reaches with high banks, the mass-failure source of material seems unlikely to have been a major contributor to the faecal coliform behaviour observed. The banks of the Washburn showed little evidence of bank erosion, the banks were protected by mature tree trunks and boughs and were otherwise well vegetated or cobble armoured, the situation was similar for the Clywedog, although small patches of exposed alluvium were present. The Rheidol differed, in that it was a flood plain location with a gravel-bed with overlying alluvium. Ashbridge (1995), gives an overview of bank erosion processes with particular reference to the River Culm in Devon where the main processes of bank erosion are through corrasion, collapse and flushing away of the collapsed material. Corrasion is the undercutting of lower, cohesionless materials such as gravels, and tends to give a time delayed response associated with extended scouring at high flows. Frequently bank material will collapse on the falling limb of a hydrograph once the support of the adjacent water is lost. Material may be washed away rapidly or remain at the base of the bank to be washed away by a subsequent high flow. Weathering of the banks, desiccation cracking and frost action can weaken bank material making it prone to more 134



rapid erosion. Seasonal soil moisture variation is also important, at high pore water pressures the soils loose their cohesion. It is possible that the gradual corrasion of submerged banks contributed to the general supply of organisms that maintained concentrations above those of the release waters in the Rheidol and perhaps also in the Clywedog. It is also possible that the post hydrograph recession periods of extended enhanced concentration in the Rheidol (7 April 1993) and the peaks in bacterial concentration in the Clywedog were the result of organisms released from eroded bank materials.

Walling et al. (1999) used a fingerprinting technique to assess the relative proportions of sediment from different sources in the River Wharfe and Yorkshire Ouse in the UK, where 37% and 23% of the load weighted mean concentrations, for the period 1994-1997, were from bank sources the remainder was from topsoil sources. During high flows, bank sources comprised a greater proportion of the overall load than during low flows. The technique, in very simple terms, works on the basis of examining the chemical and mineralogical characteristics of material from typical source sites and examining suspended sediment samples for their relative proportions of the different components using a multivariate statistical approach.

5.2 The bacterial entrainment model The results presented above in Section 5.1 suggest entrainment mechanisms resulting from the disturbance of organisms, attached to or stored within the wetted perimeter of a stream or river channel, by the passage of a propagating flow-wave and expansion of the field of the main turbulent flow body. This section gives some background to the steps taken in model development, describes a new model devised and developed with the event data to reproduce the observed behaviour, and presents the modelling results.

The early development of the model for faecal coliform fluvial dynamics followed attempts to apply the model of Jenkins (1984), as well as, simple threshold entrainment models. The model of Jenkins (1984) (Chapter 2) used equations for entrainment and settlement from a single channel store of organisms with non-linear functions and threshold values to set "knee-points" in response curves to limit the rate of entrainment. Since the Jenkins model

135



could not reproduce the behaviour observed in the field experiments and threshold approaches were over-simplistic an alternative model was devised.

5.2.1 The new model The basic form of the new model follows the mass-balance plan used by Jenkins (1984). The river reach is treated as a two-box system; an upper box for the changes in concentration in the water,

X =I+E-S -D and a lower box for the organisms stored in the associated part of the channel

N = S - E - D where, X is the mass balance of inputs, I, entrainment, E, settlement, S, and die-off, D. The number of organisms in channel storage, N, is the sum of settlement and losses due to entrainment and die-off.

The key process of the model for this application is entrainment. The model has two components for entrainment, a slow entrainment mechanism associated with the expansion of the turbulent body of the flow with associated processes of bursting and of cobble disturbance. The second part of the entrainment system relates to rapid entrainment associated with the wavefront processes proposed in Section 5.1.6.

An essential part of the proposed model is some form of mixing process. Active-mixingvolume theory was used as the basis for this (see Chapter 2). Investigation of active mixing volume modelling in rivers has shown that only a proportion of the water volume in the river reach is effectively involved in mixing and dispersing a tracer or contaminant pulse (e.g. Wallis et al., 1989); hence the term active mixing volume. The proportion of the river reach volume, V, termed the "effective volume", Ve, is said to be the "dispersive fraction" (Df) of V, and has been shown by experimentation at different discharge rates in a number of UK streams and rivers to be weakly dependent on discharge, decreasing slightly at high flows (e.g. Wallis et al., 1989; Beven and Carling, 1992). As Df 

Q b Q Ve and b0  t k , 0  t k , Vk Df V Vek

and since the variation in Df is small with respect to discharge, it has been assumed to be constant for the purposes of this application. The mixing volume residence time, T = Ve/Q, 136



and gives a measure of the time a conservative tracer would reside in the study reach (Price et al., 2000). On the assumption that hydrograph propagation is analogous to the dispersion of a conservative pollutant (Whitehead and Hornberger, 1981) the routing parameter, b0, was used in the water quality model to determine mixing characteristics through the artificial hydrograph events (see Chapter 4). In the one dimensional approach used here, an important simplifying assumption made is that the water is completely mixed across the channel cross-section (see Henderson-Sellers et al. 1990).

Price et al., also present the full AMV equation for time varying Q and V written in the mass-balance arrangement, which for a conservative solute is

Q dMot Qt 1 dQt 1 dVet  Mit    t    Mot . dt Vet Vet Qt dt Vet dt  where Mi is the mass entering the reach and Mo is the mass leaving the reach outlet. Since

dX t dVet dMo dVet X t   Vet  Xt dt dt dt dt Substituting and rearranging for the change in concentration of solute with time gives

Q dX t Qt 1 dQt 2 dVet   U t    t   X t . dt Vet Vet Qt dt Vet dt  This is the basic AMV for a conservative solute. For modelling faecal coliforms the additional terms for die-off, settlement, rapid and slow entrainment, and lateral inflows have to be added to the mass-balance, which becomes;

Q  dX t Qt Ne  Ni 1 dQt 2 dVet  U t    t    k s  ktot t  X t  t dt Vet Vt  Vet Qt dt Vet dt  where V is the volume of water in the reach and ks is the settling rate, ks = vs / zt, where vs is the stokian settling velocity and zt is the water depth, ktott is the die-off rate. The inputs are Net, which is the sum of the entrainment inputs, and Ni, catchment derived inputs.

Taking the finite difference backwards approximation of the one-dimensional differential equation and rearranging for Xk gives

X k = t

   Q  2(Vek  Vek 1 ) Qk  Qk 1  Qk  X k 1  .... U k  1  1   t  k  k s  k tot     Vek Ve Q    Vek k k   .....

t ( Nek  Ni ) . Vk 137



In this fully dynamic formulation the possibility of instability becomes an issue. In situations where there is a smooth transition from low to high flow and vice versa it is easy to choose a temporal discretisation interval, t, that keeps the changes in Q and Ve small with respect to the absolute values of Q and Ve from one time step to the next. For greater

t, where there changes in Q and Ve may be large with respect to Q and Ve, there is the potential that the model will be unstable, i.e.

Q  2(Vek  Vek 1 ) Qk  Qk 1 t  k  k s  k tot    1 Vek Qk  Vek  which results in instability. This was a concern with the experiments carried-out in the current study, where rapid changes in discharge and volume resulted from the stepped reservoir releases. Instability was also an issue when modelling the River Dee, the high values of summer die-off requiring a short time-step (Section 6.3). Figure 5.16 shows the relative magnitude of the parameter components for the model of the Afon Clywedog event.

In the more common steady-state applications for solute dispersion (e.g. Green et al., 1994) both Q and V are constant and the additional terms are not necessary. The effect of the additional terms are as follows. On rising flows, discharge and volume increase and hence the influence of Xk-1 on the current value of X is reduced, this is sensible given that more water of different concentration is flowing into the system. On a hydrograph recession the effect is reversed and the influence of Xk-1 on the next value is enhanced. Put another way, the influence of the input concentration U is reduced, which is logical since less water with concentration U is entering the hypothetical river reach.

The bed-store component of the model was tested in two formulations, one used a single channel-store the other used a distributed channel store, the two alternatives produced very similar results. It was also found that the model could be greatly simplified by reducing the mixing volume part of the model to a single constant parameter, and ignoring die-off which was very slow compared to the other processes operating, such that

X k = bU k  1  1  b  tk s X k 1 

t ( Nek  Ni ) Vk

The model was tested on the Afon Clywedog data with and without the additional terms and found to produce very similar results in the different cases (Figures 5.12 and 5.17). 138



The mass-balance equation for the single bed-store component was as follows

dNTt = k sVt X t - k b NTt  Net dt and

 d 2 Qt  dQt e Net =   w  r   s Qt2  NTt  Newt  Nert  Nest 2 dt dt   where NTt is the total number of organisms, w is the wave-front entrapment constant, r is the wave-front disturbance constant, for simplicity this will be referred to as "rapid" entrainment. s is the non wave-front entrainment constant, this will be referred to as "slow" entrainment. The second derivative of discharge is used for wave-front entrapment, where the entrained organisms enter the wave and travel with it in the overrun. It was necessary to make this conditional so that it only worked on rising discharge. The rapid entrainment is time lagged by e. The organisms disturbed by the wave are lifted into suspension but are not trapped in the wave overrun and the resultant bacterial peaks propagate at the mean flow velocity. On the Clywedog since the main phase of entrainment was associated with postwave mean velocity transport the time-delay was calculated from the relationships between wave speed and discharge and flow velocity and discharge (Table 4.5).

The earlier alternative version of the model uses a distributed channel store with multiple sub-stores for the rapid entrainment component (see Wilkinson et al., 1995a and b). This version was found to produce almost identical results to the preceding single store version, but with the added utility of accounting for the discrete changes in storage though-out an event. In effect it maintains a flow memory of the history of entrainment in the channel. This allows for the hypothetical possibility that part of the channel can become depleted of organisms, but storage remains in regions flushed at higher flows.

The channel-store is constructed by dividing the range of observed discharges for the study period, Qmin to Qmax, into a number of sub-stores, n, this gives what is in effect a "bed-store access interval",

Q =

Q max - Q min n

The number of sub-stores undergoing entrainment is

ne 

Qt - Q min Q 139



where Qt is the discharge at time t. At constant discharge ne sub-stores are empty and further entrainment can only occur if the discharge increases. The initial rapid-supply number of organisms for the study reach, NR0, is then divided into n-ne sub-stores resulting in Nt,j organisms in each store;

N t, j =

NR0 . n  ne

The stores assigned between Qt and Qm at any time t, are free from entrainment and able to accumulate organisms settled from the flowing water. This allows for re-accumulation of entrained organisms into other areas where accumulation is still possible. In the short term the areas of storage in the channel may maintain a "flow-memory". This would depend on the time elapsed between events, the water depth, settling velocity of the particles and the concentration of organisms in the water. A first flow event might deplete an area of storage and further entrainment would only be possible with a second higher discharge which would flush organisms from previously undisturbed areas. If the organisms can be redistributed on rising flows, then they can also re-accumulate in previously depleted areas once the conditions no longer cause entrainment.

The net change in rapid channel supply numbers of organisms NRt is

dNRt = k sVt X t -  r dt

ne

N

t

 k sVt X t  Nert

n=0

where all variables and constants are as previously defined. The reduction in numbers of organisms in each of ne sub-stores undergoing entrainment is the sum of settlement minus entrainment and die-off,

dN t, j k sV = X t - ( r  k b) N t, j dt n where, kb is the net bacterial die-off rate on the river bed. The change in storage in the stores not subject to entrainment is settlement minus die-off,

dN t, j k sV X t - k b N t, j . = n dt Settlement is treated as occurring in all parts of the channel not just in quiescent zones; accumulation will only occur if settlement exceeds entrainment and die-off (see Graham, 1990).

140



The slow entrainment component of the channel supply is identical to the slow component of the single store formulation, such that

dNS t = k sVt X t -  s Qt2 NS t dt where, NSt is the size of the slow release channel store. The slow entrainment component relates to the sum of entrainment due to random burst/sweep action and cobble disturbance and perhaps the erosion of bank material during extended high discharges. Given this, it would be almost impossible to account for this behaviour with a distributed formulation in a meaningful way.

The results presented in the next section are a series of examples taken from the overall period of model development. Consistent results were achieved for the River Washburn and Afon Clywedog with a number of model variants. A great deal of work was carried-out with the Afon Rheidol data and a cross-section of these results are presented. It became apparent that further experiments would be needed in the Rheidol to confirm the exact nature of the behaviour occurring.

The model version used for the majority of the results presented below differs only in minor detail to the latest versions of the models presented above. The model uses the distributed channel store sub-system described above and the discrete time equation for the concentration of faecal coliforms in the water column is

X k = t

 2(Vek  Vek 1 ) Qk  Qk 1  Qk v    Q  X k 1  ....  U k  1  1   t  k  s  k tot    Vek Vek Qk    Vek z k   .....

t (Nerk  Qk2 Ns  Ni ) Vk

where Ns is a constant slow release entrainment input (equivalent to the slow entrainment variable presented above), Ni is the sum of inflows from the catchment, and Nerk is as defined above. The slow entrainment acts almost identically to that in the latest version of the model, but in this case the exhaustion of the supply of organisms for slow entrainment is assumed not be significant over the time span of the experiments.

141



5.3 Modelling the reservoir releases The model development procedure is an on-going or evolutionary process with new data and modelling results influencing decisions about validity of model structures or mechanisms which may be modified or updated to incorporate new ideas or alternative approaches. The model results presented below should be considered in this light; they represent the output or testing of different ideas from on-going development. The results are therefore preliminary and are not considered in the same sense as modelling results from, for example, hydrological modelling where calibration and validation runs can easily be carried-out for different sites as well as offering the potential for identifying numerous calibration optima within a multi-parameter hyper-space. In the current study the model parameters are chosen to reflect values presented in the literature in order to support the perceived physical behaviour represented by the model, rather than choosing any value however unrealistic simply to achieve a model fit. Having chosen parameters based-on observation, adjustments to model fit, where possible, were made by altering the numbers of organisms available for entrainment or from catchment inputs. In this way it is assumed that the well founded parts of the model can be trusted and the highly variable and unknown inputs to the system are treated as the main influence on model fit.

5.3.1 River Washburn and Afon Clywedog The River Washburn was the most readily modelled (Figure 5.11), in such an extreme case, a major step change in flow, effects such as die-off could be ignored, and settlement was irrelevant given that this was a full scale entrainment episode (see Section 5.1.3). Parameter values were chosen manually to give the best fit as measured with the Rt2 statistic. Rt2 is a statistic of goodness of fit used in time-series modelling (see Young and Benner, 1991, Price

et al., 2000);

Rt 2 = 1 

 ( y k  yˆ k ) 2  ( yk ) 2

where,  ( y k  yˆ k ) 2 is the variance of the squared model errors and  ( y k ) 2 is the variance of the squared observed values. When the model errors are small Rt2 tends to 1.

142


5. FC Fluvial Dynamics 1.200

2000 Rt2=0.915

1800

CFU per 100ml

1400

0.800 FC 1

1200

Model FC 2

1000

0.600

Obs. FC 2 800

Stage

0.400

600 400

Stage (metres)

1.000

1600

0.200

200 0 16:12

0.000 16:40

17:09

17:38

18:07

Figure 5.11 Observed and modelled faecal coliform concentrations in response to a step

change in flow on the River Washburn.(Parameter values were as follows: Df=0.35, b0min=0.457, b0max=0.748, NT0=1.8.107 cfu, ktot=0.6d-1, vs=1.2m.d-1, n=10, =0.2min-1, Nsk=10 cfu.sec-1, Nik=25 cfu.sec-1).

The initial conditions to be set for the fluvial dynamics model are the size of the channel store and the distribution of the organisms within the store. For the Washburn and Clywedog the organisms could be assumed to be evenly distributed across the channel substores. For the Rheidol two alternative approaches were investigated for setting the numbers of organisms in the store as described in Section 5.3.2. Figure 5.12 shows the model fit to the Clywedog data using the earlier version of the model. The good fit to the observed data (Rt2 = 0.923) suggests that the Afon Clywedog was just as readily modelled as the Washburn. The fit was achieved by setting e to 10 minutes. Without the time delay the modelled bacterial peaks occur earlier than those observed (dotted line). In the latest version of the model a second derivative term was introduced for wave overrun transport, where the entrained organisms have been sucked into the wavefront and travel at the wave-speed rather than the slower mean water velocity (see Figure 5.13).

143



2500 Model FC (time shifted)

Rt2=0.923

2000

12

Observed FC

1500

8

Flow

6

1000

Flow (cumecs)

CFU per 100ml

10 Model FC (no time shift)

4 500 2 0 09:21

0 10:33

11:45

12:57 14:09 Time HRS:MIN

15:21

16:33

Figure 5.12 Model fit with fully dynamic AMV and fully time varying parameters. (Df=0.3,

b0min=0.194, b0max=0.506, NT0=1.7.108 cfu, kwmin=0.10 hr-1, kwmax=0.27 hr-1, vs=1.2m.d-1, n=10, =0.1min-1, Nsk=100.Q2 cfu.sec-1, Nik=50 cfu.sec-1 and =10 minutes).

This latest simplified version of the model has only one main channel store, as described in Section 5.2. The derivative terms for wave overrun and rapid entrainment were constrained so that responses were only possible on rising flow waves. Alternatively, by using the absolute value of dQ/dt, i.e. +ve and -ve = +ve, it was possible to allow entrainment on falling, as well as, rising waves. Figure 5.14 used the most recent distributed store version

144



Modelled FC 2000

12

Observed FC

CFU per 100ml

1500

Rt2 = 0.990

10

8

6

1000

Flow (cumecs)

Flow

4 500 2

0 10:30

11:30

12:30

13:30

14:30

15:30

0 16:30

Figure 5.13 Model fit to the Clywedog experimental data using the latest version of the

model with single channel store. The entrainment time-lag is calculated from the difference between wavefront arrivals and the arrival of the bacterial peaks.

Modelled FC 2000

12

Observed FC

CFU per 100ml

1500

Rt2 = 0.943

10

8

6

1000

Flow (cumecs)

Flow

4 500 2

0 10:30

11:30

12:30

13:30

14:30

15:30

0 16:30

Figure 5.14 Model fit to the Clywedog data with the latest version of the model with

distributed storage and falling wave entrainment. 145



of the model with entrainment on both rising and falling flow waves, the two additional small peaks on the hydrograph recession are due to falling wave entrainment. This version of the model did not have wave overrun entrainment, hence the small wavefront peaks in concentration were not modelled. With the distributed store version of the model, the depletion of parts of the channel flushed at lower discharges means that there is not much remaining storage to be entrained by the falling waves, as would be expected. In the single store version of the model, relatively large numbers of organisms remain in the store after the hydrograph rise and bacterial peaks were produced on each falling wave. Since there was minimal falling wave entrainment this feature was disabled. 12

1.8E+08 1.6E+08

10 1.4E+08 1.2E+08 1.0E+08 6

4

2

0 09:21

ktot

8.0E+07

Nrk

6.0E+07

NTk

4.0E+07

Sum Nrk

2.0E+07

cfu

T90 Hours

8

0.0E+00 10:33

11:45


15:21

16:33

Figure 5.15 Variation in die-off rate (T90 = 1/ktot) and model variables associated with

rapid release channel store bacterial numbers in response to the experimental release on the Clywedog.

Figures 5.15 and 5.16 demonstrate the changes in parameter values, channel storage and the episodes of entrainment through the course of the model run presented in Figure 5.12. The use of fixed parameter values in the mixing component of the model was found to produce better results (Figure 5.17) than with time varying parameters. With fixed parameter values based on the mean of the values presented in Figure 5.15, the model reduces to:

X k = 0.3348U k  1  0.6360 X k 1 

t (0.1Nrk e  Qk2 Ns  Ni ) . Vk 146



1

0.8

0.6

Sum of parameters

bo/Df

2(V(k)-V(k-1))/V(k)

(Q(k)-Q(k-1))/Q(k)

delta t.ks

delta t.ktot

Mean

0.4

0.2

0

-0.2 9:21

10:33

11:45

12:57

14:09

15:21

16:33

Figure 5.16 The time varying parameters associated with model result in Figure 5.12

above. "Mean" is the average of the summed parameters over the duration of the model run.

2500


Rt2=0.954

2000

12

CFU per 100ml

10 Flow 1500

8 6

1000

4

Discharge (cumecs)

Observed FC

500 2 0 09:21

0 10:33

11:45

12:57

14:09

15:21

16:33

Figure 5.17 Model fit with fixed parameters, using mean parameter presented in Figure

5.16 above and the mean value of bo/Df for the upstream input term. The model fit was adjusted using the input variables Nsk=20.Q2 cfu.sec-1, Nik=1000 cfu.sec-1.

147



1.2 Wetted Perimeter Discharge

Normallised parameter

1.0

0.8

Flow cross sectional area Log10(Q) Velocity

0.6

0.4

0.2

0.0 09:21

10:33

11:45

12:57

14:09

15:21

16:33

Figure 5.18 Changes in flow on the Afon Clywedog represented by various normalised

parameters, demonstrating alternative functions for the distribution of bacterial accumulations in the channel. 14

2500 Model output

12 2000 CFU per 100ml

10 Flow

1500

8 6

1000

4

Discharge (cumecs)

Observed FC

500 2 0 09:21

0 10:33

11:45

12:57

14:09

15:21

16:33

Figure 5.19 An example of model output using wetted perimeter as the entrainment

threshold variable. As can be seen this results in the majority of stored organisms being entrained at lower discharge. The time shift parameter =0. By using different entrainment driving variables different responses can be achieved. Figure 5.18 presents a variety of flow derived variables that can be substituted as the entrainment 148



driving variable. The variables have been normalised for ease of comparison. Each of the alternatives will result in greater entrainment at lower discharge (Figure 5.19).

These examples demonstrate that the faecal coliform responses in the Washburn and Clywedog could be readily modelled and that the features of the model offer a great deal of versatility in the nature of the response that can be achieved. In particular the timing of the response to the disturbing influence is easily varied and the nature of the response, i.e. whether it is focussed at higher or lower flows, can be controlled. The model can accommodate depletion of the channel supply of organisms and also include entrainment by falling waves.

5.3.2 Model application to the Afon Rheidol The bacterial response of the Rheidol was more difficult to model than the Washburn and Clywedog responses. The low concentrations in the first experiment were due to a large generating discharge made from Aberffrwd Dam the previous evening, this meant that the channel supply did not have time to recover prior to the experiment. The steep spike at the beginning of the response could not be modelled, and this raised a question about the first arrival of the event release downstream. It is unlikely that the small initial increase in discharge would be capable of causing such a spike of concentration, although, when examined as a contribution to the bacterial load, the spike was relatively small (see Section 5.1.3). An alternative source for this spike and that at 15:30 Hrs was the waste-water treatment works that serves the Hamlet of Capel Bangor. The works discharges on a pulse basis and may have been responsible for the spikes observed. The second experiment was swamped by rainfall-runoff which delivered large numbers of organisms from the catchment, only in the latter part of the experiment could a possible channel response be discerned.

The following section presents modelling results for the first of the two experiments, and demonstrates a simple approach to bacterial accumulation in the channel and the effect of successive flow peaks. A multiple reach version of the model was run but this did not provide additional data of value in determining the nature of the Rheidol response, and this work has not been included.

5.3.2.1 Example modelling results 149



Figure 5.20 gives an example of modelling the faecal coliform response of the Rheidol in the first of the experiments (17.2.1993). The earlier model version is used and is the same in all respects as that used in 5.3.1 above, except that the entrainment driving variable is Q2 rather than Q, such that 2 2 2 Qmax - Qmin Q 2 - Qmin and ne  t . n Q 2

300

16 Observed FC Modelled FC Qo

CFU per 100ml

250

200

14 12 10

150

8 6

100

Discharge, cumecs

Q 2 =

4 50

0 08:24

2

09:36

10:48

12:00

13:12

14:24

15:36

16:48

0 18:00

Figure 5.20 Model fit to the Afon Rheidol event of 17 February 1993, entrainment is driven

by discharge from stage. (Parameter values were as follows: Df=0.35, b0min=0.243, b0max=0.824, NT0=1.7.107 cfu, kw=0.15 d-1, vs=1.2m.d-1, n=100, =0.18min-1, Nsk=8.Q2 cfu.sec-1, Nik=150 cfu.sec-1 and =25 minutes).

Since the discharge time-series only retains minor vestiges of the features of the original stepped release discharge from Aberffrwd Dam, the bacterial response generated by the model lacks distinct peaks. By smoothing and time-shifting the release discharge, which propagated approximately conservatively in overall form (Figure 4.1), and using this to drive entrainment, it was possible to produce a pulsed response (Figure 5.21; the new single store version of the entrainment model used for Figure 5.13 was used here). For the result presented in Figure 5.21 Q, rather than Q2, was used as the entrainment driver. This version also includes falling wave entrainment, which accounts for the improved hydrograph recession fit to the observed variations in concentration.

150



300 Modelled FC

14 Observed FC

250

12

Flow

10 8

150

6

Flow (cumecs)

CFU per 100ml

200

100 4 50 2 0 9:00

10:00

11:00

12:00

13:00

14:00

15:00

16:00

17:00

0 18:00

Figure 5.21 Model result using the single store version, driven by the input discharge

(smoothed and delayed to coincide with observed stage). This version of the model included falling wave entrainment.

Although the release discharge wavefronts were dissipated before they could reach the sampling site, they must have been conserved for some proportion of their travel downstream. If this was the case, it would not be unreasonable to expect the faecal coliform response to show some residual evidence of the effects of flow waves disturbing the bed and causing pulses of entrainment. Figure 5.7 demonstrates that the bacterial peaks did relate to the wavefronts of the release discharge. It was suggested in Section 5.1.3 that the gap in the faecal coliform response was due to the previous evenings high flow (see Figure 5.22 below). The model bacterial peaks, in Figure 5.21, coincide with the release discharge increments. In the actual response there was no peak on the flow rise from 5 to 9 cumecs. This is where the largest model bacterial peak is produced. If this peak were removed from the model response, the timing, if not the magnitude, of the other peaks coincide well with the observed peaks. This may add strength to the argument that part of the channel had been depleted, especially given that the later peaks were greater, which may reflect the possibility that areas of channel storage had remained relatively intact and were thus capable of giving a greater response than the areas that had been depleted. If this is true it diminishes the argument that specific channel accumulations might have a dominating impact on the bacterial response. Over the 8.9 km reach it seems more likely that spatial 151



heterogeneities would be integrated out. If, however, downstream of Aberffrwd Dam the organisms accumulate in channel storage with distance downstream. When the wave fronts are strongest, the channel supply will be minimal, and downstream where the waves have all but dissipated away the channel supply of organisms might be much greater.

By running the model over the preceding period including the previous high discharge it was hoped to confirm the proposition that partial channel depletion was responsible for the response in the first Rheidol experiment. At the very least it was an opportunity to demonstrate model operation over successive flows.

5.3.2.2 The effect of successive hydrographs 16 14 Discharge, cumecs

12

Qi Qo

10 8 6 4 2 0 2/18/93 0:00

2/17/93 12:00

2/17/93 0:00

2/16/93 12:00

2/16/93 0:00

2/15/93 12:00

2/15/93 0:00

Figure 5.22 Discharge, Qi, from Aberffrwd dam and routed discharge, Qo, at the field site

at Blaengeufordd on the Afon Rheidol for the three days immediately preceding and including the experimental release. In order to investigate the effect of previous flow (Figure 5.22) on the supply of organisms for subsequent entrainment, the model was run for the three days leading up to the 17 February 1993 experiment. The channel-store was recharged by deposition from the water column on the assumption of steady low-level bacterial input to the channel over the period of low flow.

152



350 Model FC

14

300

Discharge

CFU per 100ml

10 200 8 150 6 100

4

50

0 2/16/93 12:00

Discharge, cumecs

12

250

2

2/17/93 0:00

2/17/93 12:00

0 2/18/93 0:00

Figure 5.23 Model result for successive flow events.

Figures 5.23 shows the model response to the successive events. The faecal coliform response to the larger second event is diminished compared to the first. Figure 5.24 shows the change in over-all bed-storage. Figure 5.25 shows the variation in each of the 25 stores used on this application. The majority of stores are flushed by the first flow event, the remainder are unaffected, recharge continues between the two flow events, the second event flushes the available storage (up to the maximum discharge encountered during the period of interest). Figure 5.26 gives a close-up of the 17 February model response which is initially diminished with a large peak towards the end of the response.

153



2.5E+07

3.0E+06 Bed-store total, NTk

Entrainment, Nek

2.0E+06

CFU

1.5E+07 1.5E+06 1.0E+07 1.0E+06 5.0E+06

CFU per time step

2.5E+06

2.0E+07

5.0E+05

0.0E+00 2/16/93 0:00

2/16/93 12:00

2/17/93 0:00

2/17/93 12:00

0.0E+00 2/18/93 0:00

Figure 5.24 Model channel store charging-up and undergoing entrainment during events

presented in Figure 5.23.

16

1.8E+06 Stores not flushed by first event

1.6E+06

14

1.4E+06 1.2E+06

CFU

10 1.0E+06 8.0E+05 6.0E+05 4.0E+05

Stores depleted by first event

8

Recharge by deposition between events

6 4 2

2.0E+05 0.0E+00 2/16/93 12:00

Discharge, cumecs

12

2/17/93 0:00

2/17/93 12:00

0 2/18/93 0:00

Figure 5.25 Change in storage of 25 model channel bacterial sub-stores.

154



300

16 Observed FC 14

250

Modelled FC

CFU per 100ml

200

10

150

8 6

100

Discharge, cumecs

12

Qo

4 50

0 9:00

2 0 11:00

13:00

15:00

17:00

Figure 5.26 Detail of model response for the 17 February event. Figures 5.27 and 5.28 are included to show the model response using zero time delay (e) between the entrainment driving discharge and the arrival of the bacterial response. In the previous examples the time-delay values are derived from the data presented in Table 4.2. Setting the entrainment time delay to zero improves the placement of the peaks and troughs in concentration, but the response ends too late. With the time delay the overall timing of the model response is better (Figure 5.26). The use of the smoothed release discharge as the driving variable (Figure 5.27) results in a peak in the location of the large peak at the beginning of the response. If a minimal time delay were correct this would imply that the main source of organisms stored in the channel of the Rheidol were located near to the sampling location; perhaps downstream of the Capel Bangor waste water treatment works. The model seems to work satisfactorily for the successive flow events, but on the basis of this one result it is difficult to make any strong conclusions about the bacterial response of the Rheidol without further field investigations. Having made this statement it is, however, important to point-out that the response does follow the behaviour exhibited in the other experiments. The results from the Washburn and Clywedog experiments and the associated modelling results give a much stronger indication of the entrainment behaviour of faecal coliforms in upland channels. 155



300

16 Observed FC 14

250

Modelled FC

CFU per 100ml

200

10

150

8 6

100

Discharge, cumecs

12

Qo

4 50

0 9:00

2 0 11:00

13:00

15:00

17:00

Figure 5.27 Model output with zero travel-time delay between discharge and bacterial

response. The time-shifted smoothed release discharge is used as the driving variable. 300

16 Observed FC 14

250

Modelled FC

CFU per 100ml

200

10

150

8 6

100

Discharge, cumecs

12

Qo

4 50

0 9:00

2 0 11:00

13:00

15:00

17:00

Figure 5.28 As Figure 5.27, but entrainment is driven by discharge from observed stage.

156



5.4 Summary The results of a series of experiments to investigate the entrainment of faecal coliforms in response to reservoir releases have been presented. The results demonstrated the importance of steep fronted waves in flushing organisms from storage within the channel. The results suggested three modes of entrainment and transport in response to steep fronted waves: i.

Wave-front entrapment resulting in organisms being caught-up in the propagating wavefront and travelling at the wave speed.

ii.

Bed disturbance by the wave overrun, resulting in organisms being lifted into the water column and moving downstream at the mean flow velocity.

iii.

Non wave-front entrainment associated with increased discharge and associated turbulence, possibly resulting from burst-sweep cycles and cobble motion.

A new model was presented that incorporates the three modes of entrainment listed above. The new model has a component for channel storage, depth and concentration dependent settlement and mixing within the channel. The model has a small number of parameters and requires a minimal amount of field data. The model is capable of reproducing the observed bacterial entrainment peaks to a very high degree of accuracy.

Chapter 6 examines faecal coliform die-off in response to irradiance in a ponded reach of the River Dee near Chester. The principle of the entrainment model is used for catchment delivery in response to local rainfall run-off.

157



that the well founded parts of the model can be trusted and the highly variable and unknown inputs to the system are treated as the main influence on model fit.

5.3.1 River Washburn and Afon Clywedog The River Washburn was the most readily modelled (Figure 5.11), in such an extreme case, a major step change in flow, effects such as die-off could be ignored, and settlement was irrelevant given that this was a full scale entrainment episode (see Section 5.1.3). Parameter values were chosen manually to give the best fit as measured with the Rt2 statistic. Rt2 is a statistic of goodness of fit used in time-series modelling (see Young and Benner, 1991, Price et al., 2000);

Rt 2 = 1 

 ( y k  yˆ k ) 2  ( yk ) 2

where,  ( y k  yˆ k ) 2 is the variance of the squared model errors and  ( y k ) 2 is the variance of the squared observed values. When the model errors are small Rt2 tends to 1. 2000

1.200 Rt2=0.915

1800

CFU per 100ml

1400

0.800 FC 1

1200

Model FC 2

1000

0.600

Obs. FC 2 800

Stage

0.400

600 400

Stage (metres)

1.000

1600

0.200

200 0 16:12

0.000 16:40

17:09

17:38

18:07

Figure 5.11 Observed and modelled faecal coliform concentrations in response to a step change in flow on the River Washburn.(Parameter values were as follows: Df=0.35, b0min=0.457, b0max=0.748, NT0=1.8.107 cfu, ktot=0.6d-1, vs=1.2m.d-1, n=10, =0.2min-1, Nsk=10 cfu.sec-1, Nik=25 cfu.sec-1).

The initial conditions to be set for the fluvial dynamics model are the size of the channel store and the distribution of the organisms within the store. For the Washburn and 143



Clywedog the organisms could be assumed to be evenly distributed across the channel substores. For the Rheidol two alternative approaches were investigated for setting the numbers of organisms in the store as described in Section 5.3.2. Figure 5.12 shows the model fit to the Clywedog data using the earlier version of the model. The good fit to the observed data (Rt2 = 0.923) suggests that the Afon Clywedog was just as readily modelled as the Washburn. The fit was achieved by setting e to 10 minutes. Without the time delay the modelled bacterial peaks occur earlier than those observed (dotted line). In the latest version of the model a second derivative term was introduced for wave overrun transport, where the entrained organisms have been sucked into the wavefront and travel at the wave-speed rather than the slower mean water velocity (see Figure 5.13). 2500


Rt2=0.923

2000

12

Observed FC

1500

8

Flow

6

1000

Flow (cumecs)

CFU per 100ml

10 Model FC (no time shift)

4 500 2 0 09:21

0 10:33

11:45


15:21

16:33

Figure 5.12 Model fit with fully dynamic AMV and fully time varying parameters. (Df=0.3, b0min=0.194, b0max=0.506, NT0=1.7.108 cfu, kwmin=0.10 hr-1, kwmax=0.27 hr-1, vs=1.2m.d-1, n=10, =0.1min-1, Nsk=100.Q2 cfu.sec-1, Nik=50 cfu.sec-1 and =10 minutes).

This latest simplified version of the model has only one main channel store, as described in Section 5.2. The derivative terms for wave overrun and rapid entrainment were constrained so that responses were only possible on rising flow waves. Alternatively, by using the absolute value of dQ/dt, i.e. +ve and -ve = +ve, it was possible to allow entrainment on falling, as well as, rising waves. Figure 5.14 used the most recent distributed store version

144



Modelled FC 2000

12

Observed FC

CFU per 100ml

1500

Rt2 = 0.990

10

8

6

1000

Flow (cumecs)

Flow

4 500 2

0 10:30

11:30

12:30

13:30

14:30

15:30

0 16:30

Figure 5.13 Model fit to the Clywedog experimental data using the latest version of the model with single channel store. The entrainment time-lag is calculated from the difference between wavefront arrivals and the arrival of the bacterial peaks.

Modelled FC 2000

12

Observed FC

CFU per 100ml

1500

Rt2 = 0.943

10

8

6

1000

Flow (cumecs)

Flow

4 500 2

0 10:30

11:30

12:30

13:30

14:30

15:30

0 16:30

Figure 5.14 Model fit to the Clywedog data with the latest version of the model with distributed storage and falling wave entrainment.

145



of the model with entrainment on both rising and falling flow waves, the two additional small peaks on the hydrograph recession are due to falling wave entrainment. This version of the model did not have wave overrun entrainment, hence the small wavefront peaks in concentration were not modelled. With the distributed store version of the model, the depletion of parts of the channel flushed at lower discharges means that there is not much remaining storage to be entrained by the falling waves, as would be expected. In the single store version of the model, relatively large numbers of organisms remain in the store after the hydrograph rise and bacterial peaks were produced on each falling wave. Since there was minimal falling wave entrainment this feature was disabled. 12

1.8E+08 1.6E+08

10 1.4E+08 1.2E+08 1.0E+08 6

4

2

0 09:21

ktot

8.0E+07

Nrk

6.0E+07

NTk

4.0E+07

Sum Nrk

2.0E+07

cfu

T90 Hours

8

0.0E+00 10:33

11:45


15:21

16:33

Figure 5.15 Variation in die-off rate (T90 = 1/ktot) and model variables associated with rapid release channel store bacterial numbers in response to the experimental release on the Clywedog.

Figures 5.15 and 5.16 demonstrate the changes in parameter values, channel storage and the episodes of entrainment through the course of the model run presented in Figure 5.12. The use of fixed parameter values in the mixing component of the model was found to produce better results (Figure 5.17) than with time varying parameters. With fixed parameter values based on the mean of the values presented in Figure 5.15, the model reduces to:

X k = 0.3348U k  1  0.6360 X k 1 

t (0.1Nrk e  Qk2 Ns  Ni ) . Vk

146



1

0.8

0.6

Sum of parameters

bo/Df

2(V(k)-V(k-1))/V(k)

(Q(k)-Q(k-1))/Q(k)

delta t.ks

delta t.ktot

Mean

0.4

0.2

0

-0.2 9:21

10:33

11:45

12:57

14:09

15:21

16:33

Figure 5.16 The time varying parameters associated with model result in Figure 5.12 above. "Mean" is the average of the summed parameters over the duration of the model run.

2500


Rt2=0.954

2000

12

CFU per 100ml

10 Flow 1500

8 6

1000

4

Discharge (cumecs)

Observed FC

500 2 0 09:21

0 10:33

11:45

12:57

14:09

15:21

16:33

Figure 5.17 Model fit with fixed parameters, using mean parameter presented in Figure 5.16 above and the mean value of bo/Df for the upstream input term. The model fit was adjusted using the input variables Nsk=20.Q2 cfu.sec-1, Nik=1000 cfu.sec-1.

147



1.2 Wetted Perimeter Discharge

Normallised parameter

1.0

0.8

Flow cross sectional area Log10(Q) Velocity

0.6

0.4

0.2

0.0 09:21

10:33

11:45

12:57

14:09

15:21

16:33

Figure 5.18 Changes in flow on the Afon Clywedog represented by various normalised parameters, demonstrating alternative functions for the distribution of bacterial accumulations in the channel. 14

2500 Model output

12 2000 CFU per 100ml

10 Flow

1500

8 6

1000

4

Discharge (cumecs)

Observed FC

500 2 0 09:21

0 10:33

11:45

12:57

14:09

15:21

16:33

Figure 5.19 An example of model output using wetted perimeter as the entrainment threshold variable. As can be seen this results in the majority of stored organisms being entrained at lower discharge. The time shift parameter =0. By using different entrainment driving variables different responses can be achieved. Figure 5.18 presents a variety of flow derived variables that can be substituted as the entrainment 148



driving variable. The variables have been normalised for ease of comparison. Each of the alternatives will result in greater entrainment at lower discharge (Figure 5.19).

These examples demonstrate that the faecal coliform responses in the Washburn and Clywedog could be readily modelled and that the features of the model offer a great deal of versatility in the nature of the response that can be achieved. In particular the timing of the response to the disturbing influence is easily varied and the nature of the response, i.e. whether it is focussed at higher or lower flows, can be controlled. The model can accommodate depletion of the channel supply of organisms and also include entrainment by falling waves.

5.3.2 Model application to the Afon Rheidol The bacterial response of the Rheidol was more difficult to model than the Washburn and Clywedog responses. The low concentrations in the first experiment were due to a large generating discharge made from Aberffrwd Dam the previous evening, this meant that the channel supply did not have time to recover prior to the experiment. The steep spike at the beginning of the response could not be modelled, and this raised a question about the first arrival of the event release downstream. It is unlikely that the small initial increase in discharge would be capable of causing such a spike of concentration, although, when examined as a contribution to the bacterial load, the spike was relatively small (see Section 5.1.3). An alternative source for this spike and that at 15:30 Hrs was the waste-water treatment works that serves the Hamlet of Capel Bangor. The works discharges on a pulse basis and may have been responsible for the spikes observed. The second experiment was swamped by rainfall-runoff which delivered large numbers of organisms from the catchment, only in the latter part of the experiment could a possible channel response be discerned.

The following section presents modelling results for the first of the two experiments, and demonstrates a simple approach to bacterial accumulation in the channel and the effect of successive flow peaks. A multiple reach version of the model was run but this did not provide additional data of value in determining the nature of the Rheidol response, and this work has not been included.

5.3.2.1 Example modelling results 149



Figure 5.20 gives an example of modelling the faecal coliform response of the Rheidol in the first of the experiments (17.2.1993). The earlier model version is used and is the same in all respects as that used in 5.3.1 above, except that the entrainment driving variable is Q2 rather than Q, such that 2 2 2 Qmax - Qmin Q 2 - Qmin and ne  t . n Q 2

300

16 Observed FC Modelled FC Qo

CFU per 100ml

250

200

14 12 10

150

8 6

100

Discharge, cumecs

Q 2 =

4 50

0 08:24

2

09:36

10:48

12:00

13:12

14:24

15:36

16:48

0 18:00

Figure 5.20 Model fit to the Afon Rheidol event of 17 February 1993, entrainment is driven

by discharge from stage. (Parameter values were as follows: Df=0.35, b0min=0.243, b0max=0.824, NT0=1.7.107 cfu, kw=0.15 d-1, vs=1.2m.d-1, n=100, =0.18min-1, Nsk=8.Q2 cfu.sec-1, Nik=150 cfu.sec-1 and =25 minutes).

Since the discharge time-series only retains minor vestiges of the features of the original stepped release discharge from Aberffrwd Dam, the bacterial response generated by the model lacks distinct peaks. By smoothing and time-shifting the release discharge, which propagated approximately conservatively in overall form (Figure 4.1), and using this to drive entrainment, it was possible to produce a pulsed response (Figure 5.21; the new single store version of the entrainment model used for Figure 5.13 was used here). For the result presented in Figure 5.21 Q, rather than Q2, was used as the entrainment driver. This version also includes falling wave entrainment, which accounts for the improved hydrograph recession fit to the observed variations in concentration.

150



300 Modelled FC

14 Observed FC

250

12

Flow

10 8

150

6

Flow (cumecs)

CFU per 100ml

200

100 4 50 2 0 9:00

10:00

11:00

12:00

13:00

14:00

15:00

16:00

17:00

0 18:00

Figure 5.21 Model result using the single store version, driven by the input discharge

(smoothed and delayed to coincide with observed stage). This version of the model included falling wave entrainment.

Although the release discharge wavefronts were dissipated before they could reach the sampling site, they must have been conserved for some proportion of their travel downstream. If this was the case, it would not be unreasonable to expect the faecal coliform response to show some residual evidence of the effects of flow waves disturbing the bed and causing pulses of entrainment. Figure 5.7 demonstrates that the bacterial peaks did relate to the wavefronts of the release discharge. It was suggested in Section 5.1.3 that the gap in the faecal coliform response was due to the previous evenings high flow (see Figure 5.22 below). The model bacterial peaks, in Figure 5.21, coincide with the release discharge increments. In the actual response there was no peak on the flow rise from 5 to 9 cumecs. This is where the largest model bacterial peak is produced. If this peak were removed from the model response, the timing, if not the magnitude, of the other peaks coincide well with the observed peaks. This may add strength to the argument that part of the channel had been depleted, especially given that the later peaks were greater, which may reflect the possibility that areas of channel storage had remained relatively intact and were thus capable of giving a greater response than the areas that had been depleted. If this is true it diminishes the argument that specific channel accumulations might have a dominating impact on the bacterial response. Over the 8.9 km reach it seems more likely that spatial 151



heterogeneities would be integrated out. If, however, downstream of Aberffrwd Dam the organisms accumulate in channel storage with distance downstream. When the wave fronts are strongest, the channel supply will be minimal, and downstream where the waves have all but dissipated away the channel supply of organisms might be much greater.

By running the model over the preceding period including the previous high discharge it was hoped to confirm the proposition that partial channel depletion was responsible for the response in the first Rheidol experiment. At the very least it was an opportunity to demonstrate model operation over successive flows.

5.3.2.2 The effect of successive hydrographs 16 14 Discharge, cumecs

12

Qi Qo

10 8 6 4 2 0 2/18/93 0:00

2/17/93 12:00

2/17/93 0:00

2/16/93 12:00

2/16/93 0:00

2/15/93 12:00

2/15/93 0:00

Figure 5.22 Discharge, Qi, from Aberffrwd dam and routed discharge, Qo, at the field site

at Blaengeufordd on the Afon Rheidol for the three days immediately preceding and including the experimental release. In order to investigate the effect of previous flow (Figure 5.22) on the supply of organisms for subsequent entrainment, the model was run for the three days leading up to the 17 February 1993 experiment. The channel-store was recharged by deposition from the water column on the assumption of steady low-level bacterial input to the channel over the period of low flow.

152



350

16 Model FC 14

300

Discharge

CFU per 100ml

10 200 8 150 6 100

4

50

0 2/16/93 12:00

Discharge, cumecs

12

250

2

2/17/93 0:00

2/17/93 12:00

0 2/18/93 0:00

Figure 5.23 Model result for successive flow events.

Figures 5.23 shows the model response to the successive events. The faecal coliform response to the larger second event is diminished compared to the first. Figure 5.24 shows the change in over-all bed-storage. Figure 5.25 shows the variation in each of the 25 stores used on this application. The majority of stores are flushed by the first flow event, the remainder are unaffected, recharge continues between the two flow events, the second event flushes the available storage (up to the maximum discharge encountered during the period of interest). Figure 5.26 gives a close-up of the 17 February model response which is initially diminished with a large peak towards the end of the response.

153


5. FC Fluvial Dynamics 3.0E+06

2.5E+07 Bed-store total, NTk

Entrainment, Nek

2.0E+06

CFU

1.5E+07 1.5E+06 1.0E+07 1.0E+06 5.0E+06

CFU per time step

2.5E+06

2.0E+07

5.0E+05

0.0E+00 2/16/93 0:00

2/16/93 12:00

2/17/93 0:00

2/17/93 12:00

0.0E+00 2/18/93 0:00

Figure 5.24 Model channel store charging-up and undergoing entrainment during events

presented in Figure 5.23.

1.8E+06

16 Stores not flushed by first event

1.6E+06

14

1.4E+06 1.2E+06

CFU

10 1.0E+06 8.0E+05 6.0E+05 4.0E+05

Stores depleted by first event

8

Recharge by deposition between events

6 4 2

2.0E+05 0.0E+00 2/16/93 12:00

Discharge, cumecs

12

2/17/93 0:00

2/17/93 12:00

0 2/18/93 0:00

Figure 5.25 Change in storage of 25 model channel bacterial sub-stores.

154



300 Observed FC

14 250

Modelled FC

CFU per 100ml

200

10

150

8 6

100

Discharge, cumecs

12

Qo

4 50

0 9:00

2 0 11:00

13:00

15:00

17:00

Figure 5.26 Detail of model response for the 17 February event. Figures 5.27 and 5.28 are included to show the model response using zero time delay (e) between the entrainment driving discharge and the arrival of the bacterial response. In the previous examples the time-delay values are derived from the data presented in Table 4.2. Setting the entrainment time delay to zero improves the placement of the peaks and troughs in concentration, but the response ends too late. With the time delay the overall timing of the model response is better (Figure 5.26). The use of the smoothed release discharge as the driving variable (Figure 5.27) results in a peak in the location of the large peak at the beginning of the response. If a minimal time delay were correct this would imply that the main source of organisms stored in the channel of the Rheidol were located near to the sampling location; perhaps downstream of the Capel Bangor waste water treatment works. The model seems to work satisfactorily for the successive flow events, but on the basis of this one result it is difficult to make any strong conclusions about the bacterial response of the Rheidol without further field investigations. Having made this statement it is, however, important to point-out that the response does follow the behaviour exhibited in the other experiments. The results from the Washburn and Clywedog experiments and the associated modelling results give a much stronger indication of the entrainment behaviour of faecal coliforms in upland channels. 155



300 Observed FC

14 250

Modelled FC

CFU per 100ml

200

10

150

8 6

100

Discharge, cumecs

12

Qo

4 50

0 9:00

2 0 11:00

13:00

15:00

17:00

Figure 5.27 Model output with zero travel-time delay between discharge and bacterial

response. The time-shifted smoothed release discharge is used as the driving variable. 300

16 Observed FC 14

250

Modelled FC

CFU per 100ml

200

10

150

8 6

100

Discharge, cumecs

12

Qo

4 50

0 9:00

2 0 11:00

13:00

15:00

17:00

Figure 5.28 As Figure 5.27, but entrainment is driven by discharge from observed stage.

5.4 Summary 156



The results of a series of experiments to investigate the entrainment of faecal coliforms in response to reservoir releases have been presented. The results demonstrated the importance of steep fronted waves in flushing organisms from storage within the channel. The results suggested three modes of entrainment and transport in response to steep fronted waves: i.

Wave-front entrapment resulting in organisms being caught-up in the propagating wavefront and travelling at the wave speed.

ii.

Bed disturbance by the wave overrun, resulting in organisms being lifted into the water column and moving downstream at the mean flow velocity.

iii.

Non wave-front entrainment associated with increased discharge and associated turbulence, possibly resulting from burst-sweep cycles and cobble motion.

A new model was presented that incorporates the three modes of entrainment listed above. The new model has a component for channel storage, depth and concentration dependent settlement and mixing within the channel. The model has a small number of parameters and requires a minimal amount of field data. The model is capable of reproducing the observed bacterial entrainment peaks to a very high degree of accuracy.

Chapter 6 examines faecal coliform die-off in response to irradiance in a ponded reach of the River Dee near Chester. The principle of the entrainment model is used for catchment delivery in response to local rainfall run-off.

157


6. Long-term data: die-off and modelling

6 Long-term data: die-off and modelling In this chapter on bacterial die-off and long term modelling, equations relating faecal coliform die-off to environmental variables are presented and the relative influence of faecal coliform die-off demonstrated. The remainder of the chapter is dedicated to the derivation of long-term daily and hourly die-off variables driven by environmental variables. These are then used to model the variations in faecal coliform concentration observed in an eight year period of weekday data from the North West Water plc. abstraction at Huntington on the River Dee near Chester, England.

6.1 Die-off relationships 6.1.1 The die-off equations Three sets of equations are presented below relating faecal coliform die-off to solar irradiance, water temperature and pH. The equations relating faecal coliform die-off to irradiance and suspended solids, are taken from the literature (e.g. Auer and Niehaus, 1993; Pommepuy et al., 1992). The equations for temperature and pH are derived from common observations in the literature. These are the only environmental variables used in the model for the purposes of calculating die-off rate. Chapter 2 highlighted the importance of microbial interactions in the self-purification of natural waters, but these interactions are many and complicated and it would be difficult to measure their effects readily for the purposes of driving a water quality model. A model incorporating the microbial functions 158



would have many additional parameters that would be difficult to assign a meaningful value. The model would have more parameters than variables and would be considered over-parameterised. In the current model the effects of microbial interactions on faecal coliform die-off are considered to be an internal process of the aquatic system. Their effects are assumed to be implicit within the functions driven by the readily measured, or derived, physical and chemical variables, i.e. irradiance and suspended particles, temperature and pH. This assumption is supported by the use of parameter assignments associated with observations of faecal coliform die-off behaviour in natural river or lake-water in the presence of the natural micro-flora and fauna. Thus the functions incorporate not only the direct influence of the driving variable on the coliform organisms but also the hidden effect on the natural microbes and the changes, or otherwise, in their influence on the coliforms. The different equations are presented below.

Faecal coliform die-off is related to the strength of sunlight (solar irradiance) by the power law  ki = I

where,  = die-off, ki, at I=1 and  = the slope of log10(ki) against log10(I). Figure 6.1 and Table 6.2 present power law models fit to the results of five studies examining the relationship between light intensity and faecal coliform die-off in fresh and salt-waters and in waste treatment lagoons, under field and laboratory conditions. Light intensity values have been converted where necessary to W.m-2 (see Table 6.1, for conversion factors) and die-off rates are expressed as per day. The data of Auer and Niehaus (1993), Evison (1989) and Sarikaya and Saatchi (1987) produce a value of  of around 1, and similar values for  (Table 6.2) and have been used in later calculations of die-off.

Table 6.1 Conversion factors for comparison of light intensity values. Alternative Unit of Light Intensity

Equivalent Light Intensity in W.m-2

1 calorie.cm-2.day-1

0.4845

1 Einstein.m-2.hr-1

6.3131.10-5

159



Table 6.2 Summary of power law model die-off rates, ki, over ranges of irradiance investigated in original studies. Pommepuy et al. (1992), sea water.

I W/m2 0.25 0.5 1 2 4 8 16 32 64 128 256 512 1024 2048

0.207 0.305 0.450 0.663 0.977 1.441 2.124 3.132 4.617 6.807 10.035 14.795

= =

0.45 0.56

Auer and Niehaus (1993), hypereutrophic lake.

Evison (1989), clean freshwater, 15 °C

Evison (1989), clean sea water, 15 °C

(0.0171) (0.00002) 0.068 0.137 0.274 0.547 1.094 2.189

0.0171 1

(0.0091)

0.166 0.603 2.189 7.947 28.847

1.641 3.448 7.242 15.211 31.951

0.00002 1.86

0.0091 1.07

Bellair et al., (1977), sea water 18.5-26 °C

Sarikaya and Saatchi (1987), waste ponds, 2530 °C

(0.6643)

(0.0018)

2.129 2.848 3.810 5.098 6.821 9.126 12.209

0.059 0.117 0.234 0.468 0.937 1.874

0.6643 0.42

0.0018 1

Sarikaya et al. (1987), waste ponds, 2631 °C

Sarikaya et al. (1987), batch study, 24-27 °C

(0.0077)

(0.0109)

0.123 0.246 0.493 0.986 1.971

0.174 0.349 0.698

0.0077 1

0.0109 1

100 Pommepuy et al. (1992), seawater. Auer and Niehaus (1993), hypereutrophic lake.

ki (per day)

10

Evison (1989), clean freshwater, 15 degC. Evison (1989), clean seawater, 15 degC.

1

Bellair et al., (1977), seawater 18.526 degC Sarikaya and Saatchi (1987), waste ponds, 25-30 degC.

0.1

Sarikaya et al. (1987), waste ponds, 26-31 degC. Sarikaya et al. (1987), batch study, 24-27 degC.

0.01 0.1

1

10 100 1000 Irradiance, Io (W/m2)

10000

Figure 6.1 Plots of power law models fit to faecal coliform die-off rate per day for a range of light intensities and for various studies.

160



For sunlight penetration in water irradiance is usually denoted Iz,avg, the depth averaged irradiance (see Auer and Niehaus, 1993; Qin et al., 1991 and Sarikaya and Saatchi, 1987);

I z,avg =

I0 { 1 - e - z } z

where, I0 is the irradiance at the water surface,  is the light attenuation coefficient and z is average water depth (Auer and Niehaus, 1993; Qin et al., 1991; Sarikaya and Saatchi, 1987). Pommepuy et al. (1992) found 73% of observed light attenuation in fresh river water to be due to suspended and dissolved organic matter, such that

 = 0.22 C sm

0.78

where Csm was the concentration of suspended matter in mg.l-1. Gameson and Gould (1975) reported that the effect of sunlight on coliform die-off was additive and independent to that of temperature such that

k = k T +  I z,avg  . Die-off in natural waters increases exponentially with temperature such that at higher temperatures, a small change in temperature will cause a greater change in die-off rate than at lower temperatures. This relationship can be represented by the simple equation T

k T2 = k T1 10

where T is the change in temperature (C) and, kT1 and kT2, are the previous an current dieoff rate, respectively.  is the slope of log10 kT against temperature (Figure 6.2);

=

log10 k T 2 - log10 k T 1 T .

An alternative formulation for kT is

k T = 10

K 0 T

.

This gives the die-off for any temperature, relative to the reference die-off at T = 0 C, k0,

= log10 k 0 . and K 0 Table 6.3 summarises the values of  and k0 of models fit to the data of various authors. The mean values of  and k0 from Table 6.3 are 0.0285 and 0.1067, respectively. These have been used in Section 6.2 to illustrate the effect of the various die-off functions with environmental data.

161



Table 6.3 Summary of model die-off values at different temperatures as presented in Figure 6.2. Temperature

Mitchell and Qin et al., Starzyk, 1975. 1991. Waste River water. pond.

0 3 5 10 15 20 25 30

0.167 0.182 0.192 0.220 0.253 0.290



0.012 0.1671

k0

Evison, 1989. McFeters and Flint, 1987. River water. Stuart, 1972. River water Lake water. d/s WwTW.

Flint, 1987. River water, u/s WwTW

0.077 0.086 0.113 0.149 0.197 0.260 0.342

0.037 0.045 0.076 0.128 0.214 0.360

0.062 0.110 0.196 0.349 0.621

0.204 0.240 0.282 0.331 0.389 0.457

0.232 0.313 0.423 0.570 0.769 1.038

0.024 0.0652

0.045 0.0270

0.05 0.0349

0.014 0.1739

0.026 0.1722

Auer and Niehaus, 1993. Hypereutrophic lake water.

0.610 0.610

0.000 0.610

kT (per day)

1.00

0.10 Mitchell and Starzyk (1975), river water Qin et al. (1991), waste pond Evison (1989), river water McFeters and Stuart (1972), lake microcosm Flint (1987), river d/s WwTW Flint (1987), d/s WwTW Model mean

0.01 0

10 20 Temperature, degrees Celsius

30

Figure 6.2 Plots of modelled die-off, kT, against water temperature as fit to data from various studies.

162



For the third physical variable it has been possible to derive a relationship with pH. Observations of the effect of different pH conditions on faecal coliform die-off have shown that die-off is increased at non-neutral pH, i.e. die-off is accelerated, as conditions become increasingly acid or alkaline. This approximates a hyperbolic cosine function,

k pH = k min cosh a pH k min  pH  where, kpH is die-off due to pH, pHkmin is pH at minimum die-off (kmin), and a is a constant of proportionality. Fitting this equation to the data of McFeters and Stuart (1972) (see Figure 2.5),

k pH  0.135 cosh 0.4456.5  pH  . The parameters used to fit the data of McFeters and Stuart (1972) have been used in the illustrations below. The application of the equation is such that the value kmin is replaced by the prevailing die-off value at pH 6.5.

The three die-off components are combined to give a value of total, or overall, die-off per day,





 (T T )  k tot  k d 10 2 1  I z , avg cosh a ( pH min  pH )  .

Equation 6.1

The die-off effects are treated in the following progression

k tot  k d f (T )  f ( I 0 , SS )  f ( pH ) i.e. the effect of changing temperature on the darkness die-off rate is calculated, to this are added the effect of irradiance and particulates (SS). The result is multiplied by the pH effect. Since pH is an internal quality of the water or aquatic system this seems reasonable. For example at higher temperatures or irradiance rates, when the contaminant microbial population is already heavily stressed, if the pH deviates from around neutral, the increase in die-off would be far more dramatic than during dull or cool conditions. If pH were treated as an additive effect then the increase in die-off would be the same under high or low stress conditions.

The die-off component of the model has a maximum requirement of five variables and six parameters. The assignment of parameters is considerably simplified by the use of rate constants derived from literature data. Table 6.4 gives a summary of the variables and parameters used in the die-off component of the faecal coliform model, highlighting whether they are assigned from available data, or have to be optimised. Only one parameter has to be assigned, this is the slope of the temperature response, , which can be chosen according to the degree of sewage contamination of the water being studied. The other

163



parameters are assumed to be valid for a wide variety of applications, while accepting that further investigations of die-off are necessary.

Table 6.4 Summary of variables and parameters in the die-off component of the faecal coliform model. Variable

Parameters

Assignment

data

Log10 slope, 

Choice of 2 means

data or derived

Slope, 

Literature, fixed

data

Slope and exponent, b, c

Literature, fixed

Water depth, z m8

data or derived

Dependent on method

Typical function

Acidity, pH

data

Multiplier, minimum, a,

Literature, fixed

Temperature, TC -2

Irradiance, I0 W.m

Suspended solids, SS mg.l

-1

pHk,min

Water depths vary significantly in the direction of flow and across a channel with the variety of channel forms in any river reach (see Bathurst, 1993). The depth value in the model is used as a general indication and the effects of spatial variations in depth on die-off rate are assumed to be averaged-out by mixing within the channel.

The die-off model in initialised with the assignment of a value of darkness die-off rate. This may be chosen from Figure 6.2. Once this initial value has been chosen the model is run and die-off is automatically adjusted by the driving variables. Finally, in cases where pH does not vary significantly the effect might be ignored. In other cases such as in rivers affected by acid mine drainage or in upland streams affected by acidic episodes and generally low pH (e.g. Neal et al., 1997) the incorporation of the pH function would be appropriate (see Table 6.7, below).

The following section takes typical or mean values from the literature for the various components of the die-off model and gives a comparison of their relative effects for changes in the driving variables observed in natural streams and rivers.

164



6.1.2 A comparison of the die-off inducing influences Tables 6.5, 6.6 and Figure 6.3 provide an illustration of values for faecal coliform die-off that might be expected in a UK river under various solar radiation, water depth and suspended solids conditions. The values presented in Table 6.5 were calculated using the equations presented in Section 6.1.1.

Table 6.5 Calculated faecal coliform die-off rate for different daily irradiance values. Water depth z=1m, annual mean suspended solids concentration SS=9.64 and the annual mean darkness die-off rate kd=0.335 (I0, kd and SS were taken from data for the River Dee near Chester in 1989). I0 (Watts/m2)

kd + Iz,avg

Mean

257.36

0.595

5 %ile

29.23

0.364

-61.2

95 %ile

663.61

1.006

59.1

% difference to mean

7

7

z=0.2

SS=0.5

6

6

SS=1.0

z=0.28 z=0.39

SS=2.0

4

SS=8.0

3

SS=32

SS=16 SS=64

z=0.77 z=1.08

3

z=1.51 z=2.11

2

1

1

0

0

500 Io

z=0.55

4

2

0 a.

5

SS=4.0

ki (per day)

ki (per day)

5

0

1000 b.

500 Io

1000

Figure 6.3 Plots of calculated die-off (ki = Iz,avg) over the seasonal range of daily irradiance, I0, that might be expected for a UK river, for a. different suspended solids (SS) concentrations (z=1m) and b. various water depths (SS=9.64). The plots show the increase in die-off with irradiance and the reduction in die-off caused by greater suspended solids concentrations and water depths.

165



Table 6.6 Calculated faecal coliform die-off rate and percentage difference from mean, for different suspended solids concentrations. Water depth z=1m, annual mean daily irradiance I0=257.37 and the annual mean darkness die-off rate kd=0.335 (I0, kd and SS were taken from data for the River Dee near Chester in 1989). SS (mg/l)

kd + Iz,avg

Mean

9.46

0.595

5 %ile

0.5

0.769

76.8

95 %ile

37.08

0.457

-77.3

% difference to mean

Table 6.7 The change in die-off from that at mean pH over the range of pH that might be expected in upland and lowland rivers in the UK.

1 2

pH

cosh(0.445(6.5-pH))

kpH

Upland river1

mean 5 percentile 95 percentile

5.0 4.3 7.6

1.23 1.52 1.12

0% +23% -9%

Lowland river2

mean 5 percentile 95 percentile

7.8 7.1 8.7

1.17 1.03 1.52

0% -12% +30%

Data for the Afon Hafren at Plynlimon (Neal et al., 1996b). Data for lowland rivers taken from Robson and Neal (1996).

Table 6.7 indicates the effect of pH over the range of pH values that might be observed in upland and lowland rivers in the UK. The greatest effect in an upland river would be observed at high flows when pH is at a minimum resulting in a 23% increase in die-off. In lowland rivers die-off would increase at lower flows when pH is highest; a 30% increase compared to that at mean pH.

6.1.3 Summary of die-off effects Table 6.8 has been produced in order to give a rapid indication of the broad causative effects of the different die-off factors and for the purposes of assisting in the designation of model parameters. The effects on die-off, the time scales at which those effects act, and, where data is available, ranges of die-off values with the proportional change in die-off over a typical range of that environmental variable are included in the table.

166



Table 6.8 Summary table indicating the main effects, typical die-off values and relative influence of the main variables of interest. Values calculated from equations, tables and values presented in Section 6.1.1 or taken from tables in Section 6.1.2. Variable

Main effect(s)

Time scale(s)

Die-off values (rate, per day)

Typical changes in die-off

Solar irradiance

Power law relationship with solar radiation

Seasonal, diurnal, hourly

0.05 - 10 per day

10 fold increase in die-off from 10 to 100 W/m2

Suspended solids

Die-off decreases exponentially with [SS]

Seasonal, event based

Temperature

Exponential increase with temperature

Seasonal, diurnal

pH

Die-off accelerated either side of neutral pH

Fluvially driven (also influenced by photosynthetic effects)

60 % reduction in die-off from 1 to 10 mg/l, z-=1m 0.1 - 1.00 per day

+ 103 % (20 C) - 48 % (2 C) + 30 % (pH 7.1) - 12 % (pH 8.7)

6.2 Seasonal and diurnal variations in die-off Having demonstrated the relative effects of the main die-off influencing variables in Section 6.1, this section looks at variations in die-off in relation to real environmental data at daily and hourly time-steps. The investigation is based-on an eight year period of data for the River Dee near Chester on the England - North Wales border (see Chapter 3 for a site description). The environmental data was derived from a variety of sources and the often short periods of overlap within and between data-sets required that the records be extended.

6.2.1 Extending the environmental data record As highlighted in Section 6.1 there are a number of important variables that influence faecal coliform die-off in the water column. In order to investigate these effects for the River Dee data-set it was necessary to improve their overlap. Of the four main variables, observations of irradiance were only available for a the last two years of record. Water temperature was available for the first four and a half years. Suspended solids and turbidity had a period of overlap of four years. A variety of methods were used to extend the data over the full eight years of faecal coliform data (1986-93), these include the use of supplementary air temperature data, sunshine hours and river discharge and are described below.

167



6.2.1.1 Irradiance Irradiance data for these investigations were provided by the Centre for Coastal and Marine Sciences at Proudman Oceanographic Laboratory, Bidston Observatory, Birkenhead. The data available only went back as far as November 1991, therefore an alternative source of data going back to 1986 was sought. Daily sunlight hours at Ness Gardens, on the Wirral, was retrieved from paper records along with cloud cover, air temperature and precipitation for the full eight year period. These data were combined with calculated clear-sky irradiance values to produce a modelled mean daily irradiance, this was validated against the two year period for Bidston Observatory. A further investigation examined an alternative approach to deriving solar radiation for sites where no suitable meteorological observations are available and used water temperature to derive solar radiation. This latter method although achieving a high correlation coefficient (Rt2 =0.974) was less successful than the sunshine hours method and is not reported here.

Clear-sky irradiance It was necessary to calculate ground-surface incident irradiance under clear sky conditions in order to derive an upper limit for estimates of cloud influenced irradiance. In this case irradiance arriving at the water's surface of the River Dee approaching the Huntington Works intake. The first stage in this process was to programme-in equations to calculate irradiance from standard equations for clear sky irradiance received by a horizontal unit surface area. This irradiance programme could then be used to provide solar radiation values through daily cycles or mean daily values for any latitude.

Solar radiation arrives at the upper atmosphere of Earth at an approximately constant rate known as the Solar Constant, the mean value of which is 1366.3 +/- 0.7 W.m-2 (Pap et al., 1999). The irradiance received by a horizontal surface at the base of the atmosphere depends on the angle of that surface to the solar beam and the atmospheric clarity (MacIllveen, 1986). The angle of the receiving surface to the solar beam depends on latitude, , time of year and the time of day (Kirk, 1983); the solar declination, , and the solar elevation, , respectively. The solar declination is the effect of the 23.5° angle of the earth's axis to the vertical which gives rise to the seasons as the earth orbits the sun. The solar elevation results from the daily rotation of the planet. Although standard tables for solar declination exist, Spencer (1971) derived the following relationship

 = 0.39637  22.9133 cos  4.02543 sin   0.3872 cos 2  0.052 sin 2 168



where  is the day angle;  = 2d / 365 and d is day number, 0 on the first of January and

364 on 31 December. The solar elevation sin   c1  c 2 cos  where  is the time angle

 = 2t / 24 and t is the time in decimal hours. The constants c1 and c2 are functions of latitude, , and day of the year (solar declination, ); c1  sin  sin  and c 2  cos  cos  . Having computed the solar elevation, sin, for the required period and time interval the potential irradiance per unit horizontal area, I0, can be calculated; I 0  I n sin  where In is the irradiance at the base of the atmosphere normal to the solar beam. The intensity of light reaching the base of the atmosphere is reduced by the mass of optically active material in the atmosphere and the solar elevation affects the path length of the light through the atmosphere (MacIllveen, 1986). At midday the path length is a minimum and increases towards the beginning and end of the day. The fraction of the solar flux which penetrates to the surface in the direct solar beam is known as the Zenith Transmissivity, 90, and has a value of approximately 0.8 (MacIllveen, 1986). The transmissivity at any solar elevation

    90 1 / sin  and the horizontal irradiance is then calculated from I 0  F0  sin  where F0 is the solar constant.

The value of irradiance arrived at is the potential, or maximum, that might be expected under clear sky conditions for the given date, time and latitude. Minimum values of irradiance for an overcast sky during summer, winter and at the equinoxes in New South Wales, Australia (Kirk 1983) are around a tenth of the clear sky value calculated from the above equations (Table 6.9). Table 6.9 Ratio of observed irradiance minima at Krawaree, NSW, Australia (=35.81S)

(after Kirk, 1983) and calculated maximum potential irradiance (units are W.m-2). Winter

Summer

Equinox

Observed minima

48.4 (14/6/76, 11:45)

78.5 (23/12/76, 14:30)

92.0 (4/3/77, 10:30)

Potential irradiance

454.7

852.3

849.8

Ratio min/potential

0.106

0.092

0.108

Since the River Dee data has a daily time step, it was necessary to calculate the irradiance averaged over a 24-hour period, by approximate integration using Simpsons Rule (see, Stroud, 1970) and dividing the result (total daily solar energy received) by 86400 seconds. Figure 6.4 shows a half-year, midwinter to midsummer, daily mean irradiance. 169



400 Imax Averaged daily irradiance, W/m2

350 300

Io Bidston Io', f(sunshine hours) Imax (cosine)

250 200 150 100 50 0 Jan-92

Feb-92

Mar-92

Apr-92

May-92

Jun-92

Figure 6.4 Ideallised curves of clear sky irradiance, Imax, observed irradiance, Io, at

Bidston Observatory, and modelled irradiance, Io', based on sunshine hours data for Ness Gardens.

An approximation to the full equations for daily mean irradiance over an annual cycle is to use a cosine function I 0,t = I 0, w int er  I 0 1  cos  t  / 2 . Using the winter minimum and summer maximum to scale the cosine to the limits of the calculated irradiance it was found that the simple cosine gave a maximum +30% error around winter. This is because the trough of the irradiance curve calculated with the full equations is broader than that of the simpler cosine function (the Summer peak is also narrower, see Figure 6.4). An approximation to the full equations for solar elevation for a diurnal sequence of irradiance uses a half sinusoid (MacIllveen, 1983); I 0,t = I 0, max sin  N  where I0,max is the maximum irradiance at noon and N = 0.133 cos 1  tan  tan   is the day length, t is time elapsed since sunrise and s is the sunrise time given by cos s = tan  tan  (Kirk, 1983). This method results in an over approximation of irradiance of greater than 100% at times less

170



than 2 hours after sunrise and greater then 2 hours before sunset on mid-summers day; the overall error in the mean solar radiation for the 24-hour period is 19% (MacIllveen, 1983). Given this error and the fact that the midday irradiance needs to be calculated in any case, the full solar radiation equations have been used in the following investigations.

Cloud influenced irradiance from sunshine hours Cloud influenced irradiance was estimated using two methods; primarily from sunshine hours and when unavailable from cloud cover data for Ness Gardens. By calculating the ratio of sunshine hours to the number of daylight hours each day it was possible to scale the peak incoming radiation to reproduce the values for Bidston (Figure 6.4 and 6.5). For example, if the sun shines for 50% of the daytime period, the model assumes that 50% of the daily mean radiation has been delivered to the point of interest. This value is adjusted for albedo effects. MacIllveen (1983) suggests albedo values for different cloud types ranging from approximately 20% for cirrus to 70% for nimbostratus. For the Bidston observatory data, when compared to the calculated maximum irradiance curve, the upper and lower five percentile albedo values were approximately 80% and 20%. The model has been optimised to give the same upper and lower five percentiles of albedo for the estimated irradiance, I0'.

 t I ' 0 ( sh) = I max  0.18  0.03  0.85 d t sh 

  

where, I0'(sh) is irradiance from sunshine hours, tsh, and td is the number of daylight hours for the particular day and is given by t d = 24(1  cos 1 (tan  tan  ) /  ) . , is a serially uncorrellated noise variable with zero mean, used to increase the variability of I0,sh during the winter period.

Where there were missing values in the sunshine record, data on cloud cover was used as an alternative; Cc   I 0,Cc = I 0  0.95  0.937 , 10   Cc is the degree of cloud cover; Cc=0 for clear skies and 8 for total cloud cover. The functions were optimised manually and an Rt2 value of 0.993 was achieved (Figure 6.4 and 6.5).

171



450

Averaged daily irradiance, W/m2

400

Clear sky irradiance, Imax Observed Irradiance (Bidston)

Irradiance from Sunshine Hours

350 300 250 200 150 100 50 0 Jan-91

Jul-91

Dec-91

Jun-92

Dec-92

Jun-93

Figure 6.5 Mean daily cloud influenced irradiance,I0 ,as observed at Bidston Observatory,

and modelled,Io', from sunshine hours at Ness Gardens and calculated clear sky irradiance, Imax .

The actual variable used combines the observed an estimated values; where values of observed irradiance were available these were used in place of the estimated values.

6.2.1.2 Extending the Water Temperature Record The water temperature record for the River Dee was discontinuous from August 1990. This was the latter period of the microbiological data record. By making use of the relationship between air and water temperature estimates of water temperature could be made to cover this period. Comparison of water temperature in the River Dee and air temperature for two nearby meteorological sites (Ness Gardens and Bidston Observatory) showed that water temperature follows air temperature very closely (Figure 6.6). This relationship has been found for other rivers (Smith, 1981; Calandro, 1973; Johnson 1971).

172



25

Temperature at 9am, deg.C

20

15

10

5

0 Nov-91

Jan-92

Mar-92

May-92

Jul-92

Sep-92

Nov-92

-5 Twater River Dee

Tair Bidston Laboratory

Tair Ness Gardens

-10

Figure 6.6 Plots of air temperature at two meteorological sites and water temperature in

the River Dee at Huntington at 9am from November 1991 to November 1992. 30 Tw Tw'

Temperature, deg.C

25

20

15

10

5

0 Jan-86

Jan-87

Jan-88

Jan-89

Jan-90

Jan-91

Jan-92

Jan-93

Figure 6.7 Plots of modelled (Tw') and observed water temperature in the River Dee at

Huntington for the period of record of faecal coliform data;1986 to 1993.

173



In Figure 6.7, Tw, is the observed water temperature in the River Dee at Huntington and Tw' is the modelled temperature. Tw' was arrived at through the following sequence of operations on the original Tair: 1. Set all negative values to zero. 2. Calculate a seven day mean, T'air. 3. Then calculate;

Tw' k = aTw' k 1 

T ' air C z

where, k denotes the current time

interval so Twk-1 is the previous value of water temperature. Water depth is given by z, thus at greater values of water depth the air temperature has a smaller affect on water temperature. The constants a and C were used to fit the equation. The method achieved an Rt2 of 0.980. The values of a and C were, 0.7 and 0.4, respectively. 4. The values of Tw' were then substituted into the time-series Tw where there were missing values to form a composite series of observed and modelled values.

6.2.1.3 Interpolating the Turbidity Record The turbidity and suspended solids records for the River Dee at Huntington from 1986 to the end of 1993 are patchy and discontinuous. The turbidity record covers the latter part of the eight year period and the suspended solids data the former. There is a period of overlap during the central period and the two variables show a close correlation as is commonly found (e.g. Gippel 1989; Wass and Leeks, 1999). For the purposes of die-off modelling, turbidity plays an important role in influencing light penetration into the water column, as well as, providing survival niches or even a protective coating for contaminative organisms (see Chapter 2). In order to examine die-off over the full record of data it was necessary to generate from the available data a variable capable of providing a consistent indication of the suspended solids concentration in the river.

The method for deriving the suspended solids values presented in Figure 6.8 and used later to determine time varying die-off rates was as follows: 1. Correlate suspended solids (SS) and turbidity (turb); SS = 1.21 turb - 0.737, R2=0.58. 2. Calculate missing turbidity values from suspended solids using 1 above. 3. Correlate turbidity with river discharge; turb = 0.2476 Q - 0.8075. 4. Calculate turbidity, where no suspended solids values are available, from discharge. The resultant time-series comprises original turbidity values, plus most of 1987 from suspended solids correlation and a small proportion of values from river discharge. 174



Turbidity

Discharge, Turbidity, Suspended Solids

200

Suspended solids Discharge 150

Suspended solids (model)

100

50

0 Nov-86

Jan-87

Mar-87

May-87

Jul-87

Sep-87

Nov-87

Figure 6.8 A sample of the time-series for turbidity (units, NTU), suspended solids(mg/l)

and discharge (m3/s) for the River Dee at Huntington, showing modelled suspended solids concentration.

Having generated overlapping time-series of the die-off controlling variables it was possible to investigate the variation in estimated die-off in the River Dee over the eight year period of record. The following section presents the results of some of these investigations.

6.2.2 Variation in daily die-off over seasonal cycles Section 6.1 demonstrated the marked effects of sunlight, particulates and to a lesser extent temperature on faecal coliform die-off. The following section demonstrates the variation in die-off values that might be expected for a real river situation, where die-off is rapid during the summer months and slow in the winter time. A time-variable die-off rate is essential for any water quality model likely to be used over seasonal cycles. The following examples demonstrate the variation in die-off based on daily mean or spot sampled data. Section 6.2.3 demonstrates significant diurnal variation in die-off rate in addition to the seasonal and day to day variations presented below.

6.2.2.1 Depth averaged irradiance 175



Water depth and suspended matter affects net die-off in a river reach by influencing the penetration of light. The deeper the water and the greater the turbidity, the slower the resulting die-off will be (Section 6.1.2). This sub-section demonstrates this affect for a timeseries of data for the River Dee, showing the reduction in light intensity averaged over the water depth over two hydrological years (Figure 6.9a in Section 6.2.2.3). Discharge is plotted on a reversed scale, for clarity, to show the periods of greatest effect on depth averaged irradiance. Depth averaged irradiance should more correctly be referred to as received irradiance averaged through the depth of water. Depth averaged irradiance was calculated from the equations presented in Section 6.1.1 using raw data and the data derived in Section 6.2.1. The relationship between water depth and discharge is presented in Section 6.3.1. Over the eight year period of record the mean depth averaged irradiance is 38.2 W.m-2 compared to the surface value of 88.3 W.m-2, the minimum reduction in irradiance was approximately 1% and the maximum greater than 99% (Table 6.10). Note that the greatest reductions in irradiance coincide with high flow periods, when flow velocities are greatest, hence the highest bacterial loads would be expected during winter high flows.

Table 6.10 Summary statistics for variables and time variable parameters for estimating

depth averaged irradiance for the River Dee at Huntington from 1 January 1986 to 10 November 1993. z

SS

-1

m

-1

Max

280.4

4.41

223.85

Min

8.1

2.80

Mean

30.2

5%ile

Q 3

m .sec

mg.l

nu

Io

Iz,avg

-2

-2

Iz,avg/Io

W.m

W.m

25.73

343.5

223.16

0.999

0.05

0.02

3.7

0.05

0.010

3.07

8.67

1.37

88.3

38.23

0.377

9.7

2.82

1.69

0.35

5.2

0.81

0.063

95%ile

90.5

3.82

29.00

4.26

253.5

130.22

0.644

Median

18.3

2.93

4.10

0.76

66.7

21.76

0.400

SD

29.2

0.33

15.09

1.91

79.1

42.57

0.187

Section 6.2.2.3 uses the depth averaged irradiance, water temperature and pH to calculate the various components of the time-variable die-off function for literature parameters, demonstrating the relative effects of the different components for the same period of data. The period of data was chosen arbitrarily to highlight the fluctuations in die-off to the environmental variables; the full period of data is too long to display to good effect. 176



6.2.2.2 Settlement: the apparent die-off Faecal coliform settlement is considered in Chapters 2 and 5 and is of importance to the examination of die-off in the River Dee with the ponded lower reaches from Farndon to Chester weir (Chapter 3). During summer low flows settlement may be an important contributor to the over-all loss of organisms from the water column. Reynolds (1979) proposed a model for the deposition of particles from turbulent flow based simply upon mean water depth, z, and stokian (still-water) settling velocity, vs, of the particles. The settlement rate, ks = vs / z. This represents a gross settlement within the water column and does not account for the actual deposition of particles. The net rate of deposition will decrease with increasing flow because increased turbulence will maintain particles and associated organisms in suspension to a greater extent than at lower flow. In the massbalance model presented in Chapter 5 this reduction in deposition is accounted for automatically within the model. For the purpose of the illustration presented below it was necessary to use a simple function k s = v s z (1  Q Q95%ile ) to reduce loss by settlement with increasing discharge. The settlement velocity used in the estimates presented below was 1.2 m.d-1 after Auer and Niehaus (1993) and is consistent with the findings of Milne et al., (1986) studying coliform settlement with estuarine mud and silts.

6.2.2.3 The seasonally variable die-off function The overall die-off rate, ktot, was calculated using the equations presented in Section 6.1 with parameters as summarised in Table 6.11. The temperature slope parameter, , used was the mean of the values presented in Table 6.3, excluding the high values of Evison (1989) and McFeters and Stuart (1972). For die-off due to sunlight the value of  used was the mean of values presented in Table 6.2, the seawater values were excluded since they were much higher than those for freshwater. The value of  for freshwater found by Evison (1989) was very low and was also excluded. An alternative set of parameters for the light attenuation coefficient, , presented in Reynolds (1992) were tested but gave significantly less attenuation. These parameter values were derived from the study of particulates in lakes and lake feeder streams. The parameter values used were derived from a study of the Morlaix Estuary (Pommepuy et al., 1992), where conditions are closer to those observed in the tidally influenced lower Dee. The affect of tidal incursions on water depth and turbidity has not been assessed. Tidal reversals along the Huntington Intake reach last for around 1.25 hours (Weston, 1979). Sumner (1982) noted that the tidal reversals on the Dee enhance bacterial concentrations and turbidity. The modelling results in Section 6.3 show that the 177



variations in faecal coliform concentration on the Dee are dominated by die-off and fluvial processes. Since the tidal incursions are of such short duration an investigation of their impact would have required additional sampling to provide data of greater temporal resolution and was beyond the scope of this study.

Table 6.11 Summary of parameters used to derive daily die-off estimates for the River Dee.

Variable

Parameters

Assignment

Temperature, TC

, k0

0.0285, 0.1067 (mean Table 6.3)

, ( = 1)

0.00932 (mean Table 6.2)

b, c

0.22, 0.78 (after Pommepuy et al., 1992)

Water depth, z m

c, f

0.905, 0.2281

Acidity, pH

a, pHk,min

-0.445, 6.5

Irradiance, I0 W.m-2 Suspended solids, SS mg.l

-1

The resulting die-off time-series is presented in Figure 6.9 with the individual die-off components: temperature, sunlight and pH (where, kpH = ktot -( kT + ki )), as well as loss due to settlement, ks (ktot(s) = ktot + ks). The plots demonstrate die-off patterns over two contrasting years, 1988 with frequent summer storms and reduced die-off over the year, 1989 commenced with a series of compounded storms followed by an extended period of base-flow. The general seasonal pattern is of minimum die-off in winter and the greatest dieoff during the summer months. Irradiance dominates spring, summer and autumn die-off and is in excess of die-off due to temperature during those months. During the winter months and periods of enhanced flow the irradiance is much lower (Figure 6.9a) and is reduced by the attenuation of the turbid water. During 1988 the flow in the Dee rarely fell to compensation flow for more than a few days, it was a cloudy year (I0 was reduced relative to 1989), and die-off was reduced relative to 1989. As a consequence of these conditions, higher flows and slower die-off, more organisms would be expected to get transported greater distances along the river and hence delivery to the reach approaching the Huntington intake would be greater.

At the end of the period of high flows, towards the end of April 1989, die-off increases sharply and remains elevated until the next high flows in November 1989. Note also in Figure 6.9a that there are periods when discharge is low, but irradiance is still depressed, due to cloud cover, for example in June 1989.

178

350

0

300

50

250

100

200 150

Io Iz,avg

150

Q

200

100

250

50

300

0 Dec-87

a.

Mar-88

Jun-88

Sep-88

Jan-89

Apr-89

Jul-89

Oct-89

3.5

0

Die-off rate (per day)

3

0.5 ktot(s)

2.5

1

ki

2 1.5 1

kT

1.5

kpH

2

ks

2.5

0.5

3

0 b.

Dec-87

Discharge, Q


Settlement (per day)

Mean Daily Irradiance


3.5 Mar-88

Jun-88

Sep-88

Jan-89

Apr-89

Jul-89

Oct-89

Figure 6.9 a. Estimated mean daily surface irradiance (top), I0 , and depth averaged

irradiance, Iz,avg,(i.e. reduced by water depth and attenuation by particulates). Discharge is plotted to demonstrate the timing of the maximum reduction in light penetration.(Irradiance values are in W.m-2, discharge is in m3.sec-1). b. Estimated die-off, ktot(s), and the component die-off time-series, ki, kT, kpH and ks.


overall die-off in the River Dee at Huntington from 1 January 1986 to 10 November 1993. T °C

kT

Iz,avg

ki

per day

W.m-2

per day

pH

kpH

ks

ktot(s)

per day

per day

per day

Max

24.5

0.534

223.16

2.0799

8.31

0.470

0.425

3.26

Min

1.1

0.115

0.05

0.0005

6.19

0.000

0.000

0.14

Mean

10.9

0.231

38.23

0.3594

7.47

0.066

0.338

0.99

5%ile

3.5

0.134

0.81

0.0074

7.09

0.008

0.000

0.23

95%ile

19.1

0.374

130.22

1.2323

7.84

0.205

0.420

2.15 179



Median

10.6

0.213

21.76

0.1983

7.50

0.041

0.393

0.85

SD

5.1

0.079

42.57

0.4056

0.24

0.069

0.118

0.58

0

4.5

0.2

3 0.3 2.5 0.4

2 log10 FC

0.5

kT 0.6

1

-0.5

Log10 CFU per 100ml

4.5

0

4

0.5 3.5

1 1.5

3

2

2.5 2

2.5

log10 FC

3

ktot(s)

3.5

1.5 b.

Die-off per day

3.5

1.5 a.

0.1

Dec-87

Die-off per day

log10 CFU per 100ml

4

Mar-88

Jun-88

Sep-88

Jan-89

Apr-89

Jul-89

Oct-89

Figure 6.10 Estimated die-off, ktot and kT (die-off due to temperature alone) and log10 faecal

coliform concentration in the River Dee between August 1988 and May 1990. The overall die-off rate, ktot , appears to be a better descriptor of FC concentration, during the summer, than die-off due to temperature alone. During the winter months irradiance becomes a minor component of the die-off function. This is reflected in the mean values of the die-off components for the eight year span of estimated values; ki is only moderately greater than kT (Table 6.12).

When compared to the actual faecal coliform concentrations the die-off function is found to describe the major fluctuations in concentration quite effectively. Figure 6.10 contrasts the relative dynamics of the ktot and kT to the observed faecal coliform (FC) concentration of samples from the Huntington raw water intake. The FC concentration has been log10 transformed and plotted on a reverse scale in order to aid comparison. Die-off due to temperature alone kT lags the overall die-off, ktot. Using irradiance to derive a die-off function brings the peak die-off forward and better reflects the observed periods of low FC 180



concentration. During the winter period or periods of high flow, die-off ceases to be the major control on FC concentrations in the Dee, when catchment run-off and fluvial dynamics are likely to dominate (Figures 6.9 and 6.10, see also Chapter 2). In the River Dee the system effectively switches between die-off and settlement processes at low flow, and fluvial processes at high flow. The lower section of the river is ponded against Chester Weir and acts like a still water body at low flow.

The use of irradiance to drive faecal coliform die-off has other advantages over the use of temperature alone, in that the effect of particulates is automatically incorporated in the equation for depth averaged irradiance; hence the sharp increase in die-off after the end of April 1989 high flows. The very rapid variation in mean irradiance that is associated with daily weather patterns and changes in cloud cover also introduces a large degree of stochasticity into the die-off estimate. Water temperature is an inherently damped state variable relying on the slow energy balance response of the water body due to the large specific heat capacity of water. Faecal coliform concentrations in the River Dee also vary very rapidly from day to day, so it is more natural to adjust the die-off rate with irradiance rather than temperature.

Section 6.3 combines the die-off variable with catchment run-off and fluvial information to model the River Dee faecal coliform concentration. Section 6.2.3 estimates die-off over diurnal cycles and demonstrates results for contrasting seasons and weather conditions.

181



6.2.3 Diurnal variations in die-off The previous section demonstrated the variation in overall die-off using daily mean values of irradiance. This section demonstrates the effect of the much greater effect on die-off of the large diurnal variations in irradiance, i.e. night time and day time die-off.

The development of an hourly model was originally planned to facilitate the investigation of diurnal die-off behaviour, to overcome the daily time interval being both slower than the dynamic (temporal) characteristics of the observed bacterial concentrations, and too long to run a water quality model. This latter point was mainly an issue of model stability since the summer-time clear sky die-off rate was well in excess of 1 (see Section 6.2.2.2).

The main microbiological and water quality data-set for the Huntington intake spans 7 years with an approximately daily sample interval. The investigation of diurnal effects was limited by the availability of sufficiently frequently sampled data. At the Bidston Observatory, measurements of irradiance only commenced in November 1991 and the Huntington intake microbiological enumerations of River Dee water ceased in November 1992. Therefore the examination was restricted to the 1 year of overlap.

6.2.3.1 Converting to an Hourly Time Interval In order to examine diurnal die-off it was necessary to prepare the available data into a consistent form. There were only four variables with a sufficient initial temporal resolution for conversion to an hourly time step. These were: i.

Air temperature, which was converted to water temperature using a similar model to that presented in Section 6.2.1.2. The source data was from the Bidston Observatory Meteorological Station, recorded every ten minutes, which was found to correlate very closely to that for Ness Gardens (see Figure 6.6). An hourly centred mean was calculated for each new time interval.

ii.

Ten minute solar radiation data for Bidston was averaged so that each new hourly value was composed from the current value and the values over the previous hour. In the case of solar radiation it is the radiation history leading up to the current time which will have influenced the aggregated rate of die-off, i.e. the current bacterial concentration is a reflection of what has happened in the previous time interval (in this case an hour). 182



iii.

Rainfall values were summed for the previous hour.

iv.

Discharge values every 15 minutes were provided by the Environment Agency. These data were simply decimated, i.e. each hourly value was taken.

A value of turbidity or suspended solids was essential for determining the degree of light attenuation. Since there is no suspended solids data for the latter period of the water quality data set, turbidity was used. Hourly values were linearly interpolated from the daily observed values. Table 6.13 summarises the basic statistics for the variables used to calculate diurnal die-off, the values are consistent with those presented in Table 6.10. The most obvious difference is in irradiance at the midday maximum.


depth averaged irradiance for the River Dee at Huntington from 1 November 1991 to 18 November 1992. z

SS

-1

m

-1

Max

178.9

4.377

69.00

Min

8.4

2.807

Mean

28.4

5%ile

Q 3

m .sec

mg.l

nu

Io

Iz,avg

-2

-2

Iz,avg/Io

W.m

W.m

5.98

931.3

697.5

1.00

0.05

0.02

0.0

0.0

0.00

3.053

5.16

0.73

97.5

45.8

0.48

10.0

2.824

0.79

0.18

0.3

0.1

0.14

95%ile

73.9

3.640

17.00

2.01

531.6

245.3

0.78

Median

20.0

2.952

3.15

0.54

6.6

3.1

0.51

SD

23.3

0.279

6.20

0.62

172.6

86.7

0.19

6.2.3.2 Estimates of hourly die-off Table 6.14 summarises the hourly time interval die-off data, the values are consistent with the daily time interval data presented in Table 6.12, excepting the limits of the data which are higher and lower, since the data represent a broad diurnal spread rather than a mean. The maximum die-off due to irradiance is particularly high, greater than 6 per day or equivalent to 3.8 hours if expressed as T90 for the chosen parameter values (Table 6.11). The mean over-all die-off rate is 0.65 per day, a T90 of 1.5 days, slow, but still relatively rapid compared to die-off due to temperature alone, kT, which has a mean T90 of 4.2 days. Note

183



that pH is absent from Table 6.14; since kpH was a minor part of the daily die-off value (Table 6.12) pH was not included in the calculations.

One of the major issues associated with examining this data which has a strong diurnal pattern was finding a means of visualisation that was immediately transparent. There was also the need to consider scale issues because with the data presented here, both diurnal and seasonal variations were of interest. Sub-sets of the data presented below demonstrate typical diurnal patterns (Figure 6.15). The data were also modified to present the longerterm variations more effectively. Detailed statistical tables of data grouped by hour or month or by hydrological behaviour, i.e. high flow or low flow, would provide numbers to characterise the general patterns within the data, but purvey little of the dynamic behaviour. Tables are provided in appropriate sections but these were used to give a summary of results and as a cross-check of the numbers shown in the graphs. The use of scatter-plots to demonstrate relationships was also investigated, but very often the noise in the data exceeded any systematic relationship, or alternatively the relationships observed might be time variable meaning that scatter-plot or correlation analysis was of minimal value. The graphical approach adopted here offers an effective way of presenting the relationships and patterns in the data.


hourly die-off in the River Dee at Huntington from 1 November 1991 to 18 November 1992. T °C

kT

Iz,avg

ki

ktot

per day

-2

per day

per day

W.m

Max

28.7

0.724

697.5

6.500

6.813

Min

1.6

0.119

0.0

0.000

0.119

Mean

11.5

0.241

45.8

0.427

0.668

5%ile

4.0

0.139

0.1

0.001

0.158

95%ile

19.3

0.386

245.3

2.286

2.599

Median

10.9

0.221

3.1

0.029

0.305

SD

4.7

0.079

86.7

0.808

0.838

Finally, the examination of diurnal data offers the potential to test whether an hourly timescale is relevant when investigating changes in bacterial concentrations in a slow flowing lowland river. For example, at low flow if the river is visualised as a conduit flowing slowly with fixed inputs at various locations, the environmental stresses, in this case dominantly 184



irradiance, will act at all points along the system simultaneously; the die-off dynamics observed at one point will be reflected at any other location. At low flow, the system is quasi-stationary as the mixing processes have little relevance to the overall bacterial dynamics. This theme is pursued in Section 6.3, where, the intention of applying the mixing-volume approach (described in Chapter 2) was abandoned and examples of various simpler alternatives are presented. 9

10000 Hourly T90 Maximum die-off Minimum die-off

7 Die-off, T90 (days)

Observed FC

6

9000 8000 7000 6000

5 5000 4 4000 3

CFU per 100ml

8

3000

2

2000

1

1000

0 Oct-91

0 Dec-91

Feb-92

Apr-92

Jun-92

Aug-92

Oct-92

Figure 6.13 Estimated hourly T90 (1/ktot) with maximum and minimum die-off curves for the

River Dee from November 1991 to November 1992. The observed faecal coliform concentration data are plotted to demonstrate the coincidence with the maximum die-off curve.

Figure 6.13 presents estimates of diurnal variations in die-off over a two year period. Dieoff is presented as T90 with running maximum and minimum die-off used to highlight the relative changes in the night-time troughs and solar peaks in estimated die-off. The observed faecal coliform concentration data are included in the plot and demonstrate the coincidence with peak die-off. Figure 6.14 presents the die-off maxima and minima as the

185



rate, ktot, and as T90. Peak die-off follows peak irradiance and minimum die-off follows water temperature. In the case of peak die-off, the die-off rate, ktot, demonstrates the summer-time dynamics to better effect than T90 which is more effective for displaying the winter dynamics. Figure 6.14 further highlights the difference between peak daytime and night-time die-off T90, which during the summer months varies between 10.5 hours and around 4 days, respectively. Table 6.15 gives a more detailed summary of the hourly die-off estimates. Section 6.3 combines T90 with flow velocity to produce a ninety percent travel distance, or X90, which reflects both die-off and fluvial behaviour and gives a good coincidence with the observed bacterial concentrations.

Table 6.15 Summary statistics for estimated hourly overall die-off (per day), ktot, and

various derivatives for the River Dee at Huntington from 1 November 1991 to 10 November 1993. hourly

running 24

running 24

running 24

ratio

value

hour mean

hour max.

hour min.

mean:max

Max

6.481

2.324

6.481

0.522

0.943

Min

0.119

0.143

0.199

0.119

0.157

Mean

0.650

0.650

1.818

0.237

0.400

5%ile

0.163

0.213

0.331

0.141

0.266

95%ile

2.404

1.451

4.239

0.355

0.658

Median

0.317

0.527

1.549

0.222

0.371

SD

0.775

0.412

1.226

0.068

0.121

Statistics for hourly ktot

Figure 6.15 shows the difference between diurnal cycles in die-off during mid-summer and mid-winter. The mid-summer peak daily die-off due to irradiance, ki, is in excess of ten times greater than that in mid-winter, the day length is also longer and consequently the running average overall die-off is approximately ten times greater than in winter. An examination of the corresponding sunshine hours values for each day shows that 26 June 1993 was an overcast day with only 0.1 hours of sunshine, the following two days both had 12 hours of sunshine or more. The 26 June ki value is equivalent to the mid-winter value with a clear sky.

186



6

ktot (max)

5

ktot (min)

4

24 hour mean

3 2 1

Dec-91

Feb-92

Apr-92

Jun-92

Aug-92

Oct-92


6 4

8

2

6

0

ktot (max)

-2

4

T90 (ktot [max])

2

-4 -6 Oct-91


10

0 Dec-91

Feb-92

Apr-92

Jun-92

Aug-92

Oct-92

0.5

16

0.4

14

0.3

12

0.2

10

0.1

ktot (min)

0

8

T90 (ktot [min])

-0.1

6

-0.2

4

-0.3 Oct-91

Die-off T90 (days)

0 Oct-91

Die-off T90 (days)


7

2 Dec-91

Feb-92

Apr-92

Jun-92

Aug-92

Oct-92

Figure 6.14 The relationship between minimum and maximum die-off rate, ktot, (top) and the

relationships between die-off rate and T90 for peak die-off (centre) and minimum die-off (bottom).

187




6 kT ki ktot (mean)

5 4 3 2 1 0 25-Jun-93

26-Jun-93

27-Jun-93

28-Jun-93

29-Jun-93

30-Jun-93

23-Dec-92

24-Dec-92

25-Dec-92


0.5 kT ki ktot (mean)

0.4 0.3 0.2 0.1 0.0 20-Dec-92

21-Dec-92

22-Dec-92

Figure 6.15 Two five day samples of diurnal cycles in estimated die-off at mid-summer (top) and mid-winter (bottom).

This section has demonstrated the highly variable nature of bacterial die-off over both seasonal and diurnal cycles, and how environmental data can be used to estimate die-off rates on the basis of data or relationships in the literature. The next section combines the themes highlighted in the literature review, Chapter 5 on fluvial dynamics and the work on die-off presented so far.

6.3 Modelling daily and seasonal changes This section presents work that gives an improved means of modelling or simulating the faecal coliform behaviour in the River Dee. The themes of river transport and bacterial dieoff driven by environmental variables are combined with a mixing-zone approach to modelling dispersion. Different approaches to bacterial input to the model are examined and

188



alternative formulations of the model components are considered to provide solutions with the minimum number of parameters that provide an acceptable result. The nature of the data available for the River Dee meant that the modelling could be carried-out at two different time-scales. There was high resolution environmental data for one year only and daily data for a seven year period from 1986 to 1992. In addition to the modelling work, while examining data relationships it was found that combining flow velocity and die-off rate produced a time-series with many of the features present in the coliform data.

6.3.1 An alternative measure of die-off, X90 The delivery of organisms from a point upstream to a downstream location in a river, ignoring resuspension effects, is a trade-off between die-off and rate of transport within the flowing body of water. The faster the organisms travel the greater the surviving number, or concentration, will be at a given downstream location. The combination of flow velocity and die-off produces a new term, X90, the ninety percent die-off travel distance. Like T90, the time taken for a ninety percent reduction in bacterial concentration, X90, is the distance that the organisms would travel in the T90 time. X90 is simply the product of reach averaged flow velocity and T90, or since T90 = 1 / ktot, X90 = v / ktot. Alternative estimates of reach averaged flow velocity, v, were derived from time of travel data and water depth assuming a trapezoidal channel (see Chapter 4). The faster of the two velocity estimates, from time of travel data for the reach between Manley Hall and Farndon was found to produce a time series that better reflected the changes in faecal coliform concentration at Huntington Intake (Figure 6.16). The use of the slower Iron Bridge to Huntington velocity caused the variations in high flow estimates of X90 to be exaggerated. The Manley Hall / Farndon velocity is less representative of the velocity in the Huntington reach but more representative of the velocity of delivery from the higher non-ponded part of the river system. The Iron Bridge / Huntington velocity data have much lower values than those for Farndon, with 99% of values less than 1, thus the high values are relatively exaggerated and the low flow, high die-off, values diminished.

An important property of the X90 variable to the die-off work in general, is that it offers the potential to estimate the reduction in bacterial concentration that might be expected between different locations along the river (Figure 6.17). The percentage reduction in concentration is calculated from (1  10  ktot . x v )  100% (see Chapter 2), where x is the reach length and v is mean flow velocity. An obvious extension of this work would be to produce a probability of 189



poor water quality based-on flow velocity and die-off estimates. This would provide a simple operational indicator of times of expected poor water quality without the complexity of a full mathematical model.

100000

10000

10000

1000

1000

100

100

10

FC cfu per 100ml

100000

X90, km

1000000

X90 FC

10 Jul-87

1 Jul-88

Jul-89

Jul-90

Jul-91

Jul-92

Figure 6.16 Ninety-percent die-off travel-time, X90, and faecal coliform concentration at Huntington Intake.

Reduction in concentration, %

120 100 80 60 40 20 0 Jul-87

Jul-88

Jul-89

Jul-90

Jul-91

Jul-92

Figure 6.17 Percentage reduction in bacterial concentration in the 6 km reach from Iron Bridge to Huntington Intake. 190



6.3.2 Water quality model applications This next sub-section presents applications of the water quality model presented in Chapter 5 in various arrangements of decreasing complexity with daily data for 1986 to 1993 and 20 minute data for 1992.

6.3.2.1 The full four input model The full water quality model is presented in this application with dynamic mixing and transfer to and from channel storage as presented in Chapter 5. In order to run the fully dynamic model it was necessary to produce time series of the driving variables with a 20 minute time interval. In the daily time-step model it was necessary to use effective parameters, i.e. values that have been adjusted to compensate for instability problems. In this case it was possible to use more empirically determined parameter values based-on the relationships presented elsewhere in the thesis. The basic structure of the model was similar to that presented in Chapter 5, but because of the specific circumstances relating to the characteristics of the Lower Dee reach approaching the Huntington Intake, special consideration had to be made to the connection of the various inputs to the model. The main issue was regarding whether the component inputs should be treated as an upstream input, which would route them through the mixing model of the reach, or whether they should be treated as direct inputs at the sampling location. The inputs to the model were: i.

A point-source input influenced by die-off and diluted into the flow at the upstream end of the reach.

ii.

A "slow" entrainment component (Q2 Ns) as presented in Chapter 5, but additionally influenced by die-off and then diluted into the flow upstream.

iii.

Rainfall-runoff driven by the quickflow component of the estimated response of the remaining 800 km2 of unregulated catchment downstream of the Manley Hall gauge (see Chapter 4).

iv.

The rapid entrainment component of the fluvial dynamics model, simplified for improved computational efficiency.

A range of variations in the inputs and their interactions were investigated and found to give minor changes to the response of the model. An example of this involved splitting the rainfall run-off and rapid entrainment inputs between the upstream and downstream ends of the model reach. This arrangement acknowledges that there will be rainfall-runoff and entrainment throughout the channel network up to the modelled reach input, and also inputs 191



to the modelled reach itself. In Chapter 5 the reservoir release experiments and modelling exercise suggested the possibility that entrainment in the immediate vicinity of the sampling location was important in determining the dynamic behaviour observed, i.e. the peaks and troughs of concentration would be more pronounced since the organisms had not undergone mixing and had not been dispersed and integrated along the study reach.

As shown in Section 6.2 die-off determines both the long-term and shorter term variations in concentration, so these experiments became an exercise in attempting to apply a more complex model to see whether the results could be improved-upon. The main aim was to attempt to model the peaks of concentration perceived to be the result of hydrometeorological stimuli. One important aspect that was not built into the model or investigated in depth was the effect of tidal reversals, this should be built into any further investigations on the Lower Dee.

The details of the model application and the equations used are described below. To recap from Chapter 5 the discrete time version of the model used to simulate the behaviour observed in the reservoir release models was

X k = t

   Q  2(Vek  Vek 1 ) Qk  Qk 1  Qk v  X k 1  .... U k  1  1   t  k  s  k tot     Vek Vek Qk    Vek z k   .....

t (Nrk e  Qk2 Ns  Ni ) , Vk

where, since the upstream inputs were negligible the entrainment inputs were added at the reach outlet. In the application to the Dee, although a variety of arrangements were investigated, the following arrangement will be described since it was deemed to be the most physically reasonable and produced justifiable results:

X k = t

 2(Vek  Vek 1 ) Qk v    Q  .... (Ues k  Us k  Urq k )  1   t  k  s  k tot   Vek Vek    Vek z k 

.....

Qk  Qk 1   X k 1  Xek , Qk 

where, the input concentration from slow entrainment, Ues k 

(1  k surf )Qk2 Nes Vk

, and Nes

is a constant.

192



The point-source input

Us k 

(1  k surf )(Us k 1Vk  Ns ) Vk

, Ns, the number of point-source organisms per time step,

is also a constant.

The quickflow runoff input component

Urq k 

 r (Qq k  Qq k 1 ) Nl k Vk

, where Nlk is the size of the land-store and

Nl k  (1  ( r (Qq k  Qq k 1 )  k surf )) Nl k 1  Nil , where r is the run-off entrainment rate and Nil, the number of organisms input to the land store at each time step, is a constant. The quickflow runoff component is conditional on the quickflow discharge, Qqk. For (Qqk - Qqk1)

< 0, (Qqk - Qqk-1) is constrained to zero and hence Urqk = 0, i.e. if the discharge is not

increasing there is no entrainment from the catchment surface. The die-off rate used is ksurf = kT + I0. The surface irradiance, I0, is used rather than the depth averaged irradiance, Iz,avg, since the organisms are derived at or near the surface of the water, or the surface of the land.

The rapid entrainment input Xe k 

 (v k  v k 1 ) Nrk Vk

, where Nrk is the size of the channel

store and is the sum of additions by settlement and losses due to entrainment and die-off;

Nrk  (1  ( (v k  v k 1 )  k z )) Nrk 1  v s z X k Vk , and vk is the flow velocity. Reach averaged flow velocity is used to drive the entrainment function in this application since it biases the majority of entrainment towards changes in flow at the lower end of the discharge range (see Chapter 5). The rapid entrainment component is conditional on increases in velocity in the same way that land-store run-off was conditional on quickflow discharge. For (vk - vk-1) < 0, (vk - vk-1) is constrained to zero and hence Xek = 0. In addition the land store bacterial number is prevented from falling below zero. The die-off rate used for the channel store was kz = kT + Iz, where Iz = I0 e-z, the light intensity at the base of the water column.

In the mixing component of the model the term Q/Ve is used instead of v/L.Df, as in Chapter 5. For the Dee, the only estimates of v available are based on times of travel for a nonponded reach between Manley Hall and Farndon, these were found to be far higher than 193



values estimated from Q/A for the Huntington reach. The estimates of channel cross-section area, A, were calculated from water depth at Huntington Intake on the assumption of a trapezoidal channel with base width 20m and bank-slope of 0.3 and produced values in agreement with Gurnell (1997)(see Chapter 4).

100000 10000 1000 100 10 Obs. FC 1 Oct-91

Dec-91

Feb-92

Model output Apr-92

Jun-92

9am value Aug-92

Oct-92

1000000 Channel flush

100000

Runoff

10000 1000 100 10 1 Oct-91

Dec-91

Feb-92

Apr-92

Jun-92

Aug-92

Oct-92

100000 10000 1000 100 10 Ww TW 1 Oct-91

Dec-91

Feb-92

Apr-92

Jun-92

Slow entrainment Aug-92

Oct-92

Figure 6.18 Model inputs and output for the River Dee, November 1991- November 1992. The input WwTW, is the "point-source" input referred to in the text. All units are in CFU per 100ml.

194



Figure 6.18 shows the four model inputs, and output including the 9am value that coincides with the daily sample bacterial concentration. The baseflow non-point source input and the point-source input provide the baseline input concentration and the run-off and channel entrainment inputs provide the pulses of microbial input. There is insufficient detail in the faecal coliform data to assess the validity of this degree of complexity, especially given that the baseline concentration can be simulated using the constant single input model with time varying parameters below.

6.3.2.2 A reduced model with constant input In this version of the model there is only one input which is a constant value, the time variable parameters of the model produce the dynamics of the resulting output concentration (Figure 6.19). The model fit was improved by increasing the reach length to reduce the b0 parameter to the same order of magnitude as the die-off parameter (Figure 6.20). The model equation was

X k = t

   Q   Qk U   1   t  k  k tot   X k 1 . Vek    Vek  

100000 Constant input

Obs. FC

Model output

9am value

10000

1000

100 Oct-91

Dec-91

Feb-92

Apr-92

Jun-92

Aug-92

Oct-92

Figure 6.19 Model output with constant single input value of 4000 cfu per 100ml.

The results with daily data were similar to those presented above, although for the longer period it was necessary to introduce a trend multiplier to fit the data over the seven year record (see below). 195



0.1

0.01

ktot

bo+ktot

bo

0.001 Oct-91

Dec-91

Feb-92

Apr-92

Jun-92

Aug-92

Oct-92

Figure 6.20 Model parameters for the single constant input model of faecal coliform concentrations in the River Dee.

6.3.2.3 A two input model with daily data Output from this version of the model is presented in Figure 6.21 below (Table 6.16 details the parameters used). This version of the model uses the mixing volume formulation with the rainfall run-off delivery model. The slow entrainment / baseflow input component was set to zero and the settling velocity was also set to zero, effectively switching-off the channel entrainment function. Point source input was set to be constant, and treated as the upstream input to the mixing model undergoing die-off within the mixing volume component of the model. The quickflow runoff input was added at the model outlet. This meant that the spikiness of the rainfall-runoff response was retained since it had not been smoothed-out by mixing in the reach. The model equation was:

 Q    Q X k =  t k Us k  1  t  k  k tot  X k 1  Xrq k .trend  Vek     Vek   Where Xrqk is Urqk as described in Section 6.3.2.1 above. The point-source input

Us k  Ns Vk , Ns, is a constant number of point-source organisms per time step. The

196



100000

CFU per 100ml

10000

1000

100

10

Model FC Observed FC

1 Jun-87

Jun-88

Jun-89

Jun-90

Jun-91

Jun-92

Figure 6.21 Model output for the daily ADZ with quick flow input and constant upstream input.

"trend" term was used to adjust the result of the output, the trend indicates a 65% reduction in the long-term faecal coliform concentration over the seven years of observation. The cause of the declining concentrations was not investigated but could be the result of improvements in waste water treatment or catchment management; the ammoniacal nitrogen concentration shows a similar decline. Alternatively it may have been due to the increase in temperature observed over the period of record. This effect was not detected in the die-off rate, which was probably due to the temperature component forming a small part of the dieoff rate. Neither was any account taken of variation in solar flux when estimating irradiance. A subsequent examination of a 20 year record of satellite measurements of the solar flux (Pap et al., 1999) shows an 11 year cycle that has a 0.1% amplitude, consequently the influence of this variation on ki is insignificant. The changes in air temperature from 1986 to 1992 do increase and decrease with the variation in solar flux and this apparent relationship might be worthy of further investigation in relation to bacterial die-off. Further examination of solar flux data might however be of value since different wavelengths of light are known to have a stronger die-off effect than others (e.g. Gameson and Saxon, 1967), and some spectral components of the solar flux vary to a greater extent than the overall variation in the overall flux (e.g. Donnelly, 1991).

197



Table 6.16 Parameters and constants for the model result presented in Figure 6.21. Parameter, constant

Value (units)

Dispersive fraction, Df

0.25

Reach length, L

6000 (m)

Upstream input, Us

3.3.107 (cfu per 100 ml)

Quickflow input, Niq

1.8.109 (cfu)

Figure 6.22 shows the quickflow runoff component; an interpretation of what it does in the model is that it represents rapid runoff from the catchment surface via non-matrix throughflow and direct run-off. The supply of organisms was assumed to be limited as indicated by Hunter et. al. (1992). The model die-off and mixing terms accounted for the majority of the variability in concentration at the time-scale represented. The die-off term accounted for the summer reductions in concentration and the mixing term accounted for the way increased discharge transported organisms closer to the sampling point and reduced the time in transit and potential for die-off.

CFU per 100 ml

1.E+05 1.E+05

Xq

8.E+04 6.E+04 4.E+04 2.E+04 0.E+00 Dec-85

Apr-87

Sep-88

Jan-90

Jun-91

Oct-92

Figure 6.22 The quickflow run-off input component for the simulation of the faecal coliform behaviour in the River Dee.

This version of the model reproduced the longer-term changes in faecal coliform concentration and also some of the shorter-term flashiness resulting from rainfall effects. A similar result was acheived without the mixing-zone component, by simply summing the various inputs to the system (see below).

198



6.3.2.4 A three input model without mixing-zones While investigating building the full mass-balance model it was found that the variations in faecal coliform concentration at Huntington on the River Dee could be reproduced to a reasonable degree using a very simple scheme (Figure 6.23). This model worked on the 100000

2.5E+04 Xk

Obs. FC

Uk

Xrqk

Xsk

2.0E+04

1000

1.5E+04

100

1.0E+04

10

5.0E+03

1 Jun-87

CFU per 100ml

CFU per 100ml

10000

0.0E+00 Jun-88

Jun-89

Jun-90

Jun-91

Jun-92

Observed FC

14000

Simple Model FC 12000

CFU per 100 ml

10000 8000 6000 4000 2000 0 Jun-87

Jun-88

Jun-89

Jun-90

Jun-91

Jun-92

Figure 6.23 Variations in model output and observed faecal coliform concentrations for the River Dee at Huntington Intake, plotted on logarithmic and natural scales.

199



assumption that there were three typical inputs contributing to the FC concentration at the point of interest on the river. These were: an upstream input, i.e. the organisms travelling with the water flowing down the main channel of the river: a local rainfall-runoff input from the catchment immediately feeding the reach approaching the sampling location: a pointsource input from piped discharges to the system. The model equations were as follows:

X k = (U k  Xs k  Xrq k ).trend where Xrqk was Urqk as described in Section 6.3.2.1 above. The upstream input concentration,

Uk 

(1  k tot )Qk Nu , Vk

and Nu is a constant input per time step. The point-source input, Xs k  (1  k tot ) Ns Vk and Ns is the number of point-source organisms per time step, which was also a constant.

The only adjustments used to fit the model were made to the input constants (numbers of organisms) of each of the driving variables. This method was repeated with the enhanced frequency data and demonstrated that although the longer-term variations in concentration could be recreated, there was still insufficient information to reproduce the short-term variations in concentration.

6.3.3 Summary The study has demonstrated the importance of die-off in determining the variations in faecal coliform concentration in the River Dee over time-scales of days, weeks and months. The whole system is affected by the same die-off dynamics. The die-off of organisms from different locations or sources might be different due to local conditions, e.g. shading from light or soil moisture status, but the overall peaks and troughs would be expected to follow the same pattern.

The results show that the major variations in bacterial concentration over a long period could be reproduced with estimates of die-off, the transport behaviour in a river reach and a few assumptions about the inputs to the system. The model incorporates a rapid rainfall driven input component local to the sampling site (as distance from the sampling site increases the peaks of run-off become dispersed to a greater extent). The results suggest that what arrives at the sampling site is determined by a trade-off between die-off and transport. 200



This trade-off between die-off and transport potential was described in a simple sense by the die-off distance variable, X90. It was demonstrated that the solar component of die-off is still the most important factor effecting faecal coliform concentrations over diurnal cycles, and Chapter 5 demonstrated a model capable of reproducing short term discharge related surges in regulated channels. There still remains a need to provide data suitable to test the combined die-off and entrainment behaviour at, say, an hourly time-step over a series of diurnal cycles during winter and summer conditions and at high and low flows.

201


7. Conclusions and recommendations

7 Conclusions and Recommendations "Determining the concentration of sanitary indicator bacteria (in this case faecal coliforms) is an important element in assessing the risk to public health from waterbourne diseases in water intended for public supply and/or used for recreational purposes" Hunter (1999). The purpose of this study was to examine the key processes influencing faecal coliform concentrations in streams and rivers and to produce a model that embodied those processes which could be tested against event based and longer term data. The work does more than this, it sets out the basis for developing a bacterial dynamics modelling toolbox with broad potential for application to a wide range of modelling issues throughout a catchment. The study presents model components that no commercially available water quality modelling package currently provides. These existing packages are not capable of modelling the complex behaviours presented in this thesis. This final chapter summarises the work and sets out tasks required to bring this work forward to provide easily usable set of software.

7.1 The Entrainment Model The results of the reservoir release experiments carried-out in this study and the data presented in the literature demonstrate that, in certain circumstances, settlement, storage in the channel and re-entrainment of faecal coliforms can have significant water quality impacts. The processes involved in this cycle can be explained by the following:

202

Modelling Faecal Coliform Dynamics i.


Settlement: Attachment and or flocculation of organisms enhances their settlement; they settle throughout the channel but only deposit where flow conditions or a cohesive substrate prevents their immediate resuspension.

ii.

Storage: Organisms are stored everywhere within the channel wetted-perimeter on plant surfaces, cobbles and rocks, amongst sediment accumulations, as particles within macrophyte beds etc. (see Section 5.1.7).

iii.

Entrainment: Organisms stored within the channel are detached or lifted back in to the water column by changes in the nature of the flowing water above. Three modes of entrainment were suggested by the behaviours observed in the reservoir release experiments: a. Wave-front (overrun) entrapment (wave-front entrainment); where a steep-fronted wave, with wave height much greater than the preceding water depth, effectively sucks and holds disturbed organisms in the turbulent wake of the wave. The disturbed material travels at the wave speed. This mechanism may also include the wave-front pushing suggested by Martin and McCutcheon (1998). b. Wave-front disturbance ("rapid" entrainment). The organisms lifted are not held in the overrun. The wave front may be less steep and the wave height small or not greater than the initial water depth. These flow transitions may merely lift the organisms into the water column to be transported at the mean flow velocity. Note that this wave-front mechanism is also relevant on falling waves, i.e. step reductions in discharge. c. Non wave-front turbulent disturbance ("slow" entrainment) of bed and bank sources. This third mode of entrainment results from continued turbulence during elevated discharges. This tends to cause many small and irregular disturbances of bed and bank. The sum of these small disturbances is sufficient to maintain elevated faecal coliform concentrations above those encountered at lower rates of flow.

These component processes are incorporated into the entrainment model and the observed faecal coliform responses can be reproduced to a high degree of accuracy. The model has a small number of parameters and requires a minimal amount of field data.

7.1.1 Model Applications 203



The entrainment model has the potential to be applied to other fields such as sediment transport and the transport of sediment or particulate associated contaminants, such as, radio-nuclides (e.g. Rowan, 1995), heavy metals (e.g. Neal et al., 1999; Berndtsson, 1990), pesticides and any other contaminant with a significant particulate phase (see Foster et. al., 1995). Application to natural phenomenon such as flushing following rainfall runoff events as well as artificially induced changes in flow are also possible. In the case of natural events the wave-front associated entrainment might not be required. For artificially induced changes in discharge, such as hydro-power operations, reservoir releases for white-water canoeing events or in man-made systems such as canals, where steep-fronted waves might result from certain operations, the full model would be required.

It was also shown in Chapter 6 that the principles of the entrainment model were readily transferred to model the wash-off of organisms from the catchment surface. The issue of determining the bacterial numbers available from the land is discussed in Section 7.4.

7.1.2 Enhancements to the field experiments There are a number of important points which would have reduced the uncertainties surrounding the results and findings presented from the reservoir release experiments.

7.1.2.1 Precision of stage records The first of these concern the measurement and monitoring of flows in the experiments. In all of the experiments stage was measured by manual observations of water surface elevation by staff gauge alone. Observations were made, at best, on each sample and in between samples, and, at worst, only with each sample. The disadvantages of this approach were only too obvious when the issue of wave-front timing and bacterial arrivals had to be addressed. Not only were the values too infrequent to capture the onset of changes in water level, but they were also not sufficiently precise to detect small changes in stage. There was no stilling around the gauges which meant the fluctuations in the turbulent water surface prevented accurate readings from being made. This was compounded by observer discrepancies. This problem could have been eliminated by the installation of continuous stage recording equipment with appropriate stilling wells. These would have provided exact records of stage fluctuations both during and before the experiments commenced to give a record of the flow history leading up to the experiments. Extra stage recorders at

204



intermediate sites between the release point and the downstream sampling sites would have allowed further examination of the dissipation of the flow waves as the propagated.

7.1.2.2 Need for direct discharge measurements In addition to precise stage records for the sites, actual measurements of discharge by the velocity area method in the Rheidol and Washburn would have been valuable. In the Rheidol these estimates could have been used to confirm how much the release hydrograph changed overall shape as it propagated. In the Washburn a measure of discharge would have been useful for estimating the total number of organisms entrained and also in examining the nature of the flow wave (see Section 5.1.1).

7.1.2.3 Continuous conductivity, turbidity, temperature and suspended sediment samples Continuous time-series of turbidity, conductivity and temperature in parallel with a continuous stage record would have greatly simplified the explanation and interpretation of the release experiments. Both changes in temperature and conductivity would have indicated changing proportions of water from different sources in the channel, i.e. the reservoir water might differ in temperature from the water already in the channels. Given that the flow waves precede the moving water, the arrival of new water (or a change in concentration) behind the wave-front would help to confirm the processes determining the timing and magnitude of the bacterial entrainment episodes.

Continuous turbidity measurements would have highlighted the peak entrainment episodes in relation to the wave-front positions and arrival of mean velocity water. It might also have confirmed or otherwise the likelihood of other minor bacterial peaks that were not captured during sampling. Additional samples of suspended material could have been examined to determine the nature of the material, i.e. the mineral and organic composition. This data would assist with assessing the different supply components within the channel and the timing of their activity with respect to the phases of the experiments.

7.1.2.4 Repeat experiments

205



In an ideal world, repeat experiments incorporating the above recommendations would be valuable to provide information about the issues of timing of bacterial peaks and to determine bed-store recharge.

7.1.3 Bed-store re-charge and die-off One major component of the entrainment model that was not addressed was how to determine the numbers of organisms in the channel store. Determining the initial bed-store bacterial numbers is an important element in setting-up the model. The model fits achieved in Chapter 5 were arrived at in part by adjusting the initial size of the bed store. This part of the model is also linked to catchment delivery and settlement processes. If the rate of supply from the catchment is better defined and the rate of settlement is appropriate, then the accumulation of organisms in the channel will occur as an internal process of the model. 7E+05

16

6E+05

14

CFU

10 4E+05 8 3E+05 6 2E+05

Discharge, cumecs

12

5E+05

4

1E+05

2

0E+00 2/16/93 16:48

0 2/16/93 21:36

2/17/93 2:24

2/17/93 7:12

2/17/93 12:00

2/17/93 16:48

Figure 7.1 Reach by reach modelled channel store bacterial numbers in the Afon Rheidol. The dashed line is discharge. The solid lines are bed store faecal coliform numbers, the smallest numbers are for the upstream reach and the greatest numbers downstream.

Die-off in the channel store is linked to die-off in the water column. This helps to balance the accumulation from settlement. Irradiance reaching the bed store would be calculated as the irradiance penetrating to the mean depth to the channel bed, rather than the depth averaged irradiance. In this way the die-off due to sunlight would automatically be adjusted to suit the incident light condition, the turbidity of the overlying water and the water depth. 206



Additional experiments to provide information about the accumulation of organisms in the channel were devised but the cost of the exercise meant that these were not carried-out. A multiple reach version of the model was also investigated. This version of the model used a fixed per unit length input term for catchment supply with settlement and transport at a steady discharge this resulted in the increasing accumulation of organisms in the channel store with distance downstream (Figure 7.1).

7.2 Catchment Delivery As shown above, catchment delivery is of direct significance to the accumulation of organisms in channel bed store. This issue is of greater significance to all studies of microbial contamination of surface waters. In Chapter 6 a simple rainfall runoff model is linked to a bacterial delivery model based on the entrainment model presented in Chapter 5. This runoff delivery module was used to generate a realistic time-series of input bacterial numbers for the River Dee model (Section 6.3.2). Model calibration through-out this study was not carried-out in the conventional manner of using a numerical optimisation routine. In the case of the River Dee model, examination of the water quality data and environmental variables, i.e. sunlight and discharge, showed that these two variables had the greatest influence on the observed bacterial variations. The model was calibrated by choosing robust parameter estimates taken from the literature for the time-variable die-off component and known relationships between discharge and travel time in the study reach. The fit of the model was then achieved by merely adjusting the supply of organisms to the various model nodes, i.e. the upstream input and the run-off and point source delivery components. The working assumption was that the model structure and parameters are well defined, but the inputs are not well known. The challenge for the future is to better define the ranges of bacterial concentrations that might be expected from land areas under differing uses. Wilkinson et al. (1995b) present data demonstrating strong positive correlations between geometric mean faecal coliform concentrations and land classifications and stocking information for a range upland Welsh catchments and larger catchments elsewhere in the UK. The strongest positive correlations were with ADAS Grade IV land (this is land in valley bottoms used for silage production and lamb fattening and some dairy/beef production), total animals per km2 and ITE Land-cover percentage cultivated land. More recent studies have provided many data sets for sites through-out the UK (see Stapleton et al., 2000a, 2000b, 1999; Wyer et al., 1999) providing bacterial numbers at both high and 207



low flow. These will be used to produce improved relationships between landuse and bacterial delivery under different flow conditions (Wyer, pers. comm.). The ultimate aim should be to provide a probabilistic framework that gives a range of possible bacterial concentrations for a given land type, this could be linked with a rainfall runoff model, GIS based land-use information and the runoff entrainment model of Chapter 6 to produce input to the river network from diffuse sources.

7.3 Real-time die-off Chapter 6 presents the development of a time-variable die-off rate driven by environmental variables for use in water quality models. The proposition is that sunlight being a global variable impacts on bacterial numbers throughout a catchment to a greater or lesser extent. Irradiance is a key variable that should be incorporated into all investigations of contaminant microbial dynamics, as well as, models that hope to determine bacterial die-off realistically. Time-variable die-off can be applied to catchment delivery from point and diffuse sources as well as the main river channel. The ability to use die-off in a real-time framework is aided by the availability of instrumentation to give continuous measurements of solar radiation, turbidity, temperature, water depth and discharge. These variables can be fed directly into the full river transport model as applied to the River Dee (Section 6.3.2), or more simply to produce X90 the 90% reduction travel distance (Section 6.3.1), or the percentage reduction in concentration for a given reach length (Figure 6.17). These simple measures link die-off and travel time in a river reach and have potential for protecting recreational water users in reaches downstream of waste water treatment discharge and also for abstraction intake protection. A further possibility is to use X90 as a measure of whether a discharge from a known point source in a river system is likely to have a significant impact at a downstream site of interest. This could be extended in larger catchment management assessments to highlight key discharges for monitoring and/or investment for improvement.

7.4 Model use Section 7.3 has already suggested some uses for a real-time variable die-off rate. These included the protection of recreational water users and drinking water abstraction intakes, by predicting periods of poor and good water quality. Real-time die-off is transferable to other water environments, for example coastal bathing waters and lakes and impoundments with 208



contact water sports uses, for example Lake Windermere (see Stapleton et al., 2000a). The die-off work presented in this thesis has also been used in investigations of bacterial water quality in Cardiff Bay, the Irvine Coast in the West of Scotland, and in the Bristol Channel. The inclusion of real-time die-off in the model for the Irvine Coast resulted in a very accurate reproduction of the observed concentrations (Wyer pers. comm.).

7.5 Towards a bacterial dynamics "toolbox" The faecal coliform dynamics model includes a variety of components and the discussion above has mentioned uses specific to particular component parts, i.e. real-time die-off applications and in Section 7.1.1 and 7.2 applications for the entrainment component in other water quality parameters or contaminants, as well as, delivery from the catchment. The direction this modelling framework appears to be taking, is towards the development of a "bacterial contamination modelling toolbox" where the variety of model components can be connected in a configuration to suit the application. The components would be considered as a series of plug-ins. The model components already used in this study include: 

A bacterial entrainment model for transfer to and from river bed storage



A river mixing module based-on mixing zone theory



A real-time die-off model with utilities for generating irradiance from cloud cover



A combined rainfall runoff and catchment entrainment model



A point source model for waste water inputs

The other components that could be derived from aspects of the other models include: 

A flow routing module



A storm sewer over flow component. This would incorporate rainfall runoff and entrainment sub models to simulate storm response pulse input



Specific land use related catchment delivery sub-models



An incident simulation module, to simulate for example some catastrophic failure of farm infrastructure injecting large volumes a waste into a river. The user selects a location, time and number of organisms, the module feeds into the system and the user can see the impact at the site of interest



A link to epidemiological studies that can highlight the probable health risks associated with consumption or other contact with water of a given faecal coliform concentration

209

Modelling Faecal Coliform Dynamics 


A help and appendix system that includes the die-off charts and equations presented in Section 6.1 as well as other essential information to assist in the use of the system.

The development of this suite of modules into a user-friendly system presents a significant programming challenge. A Windows based user interface with a flow chart visualisation of a main channel to which selected components could be added is required. By clicking on each component the user would bring-up a box with the key input rates or coefficients. The system could be run in real-time mode, feeding in data from telemetred sensors, or alternatively scenarios could be run with example data. The user could then click on nodes within the model and look at the most recent period of data, or the data for the scenario chosen.

If a new package is not a realistic option for these tools it should be possible to modify an existing package in order to achieve a greater use of the principles presented in this work and thereby assist in reducing the risks to health posed by faecal contamination of water.

7.6 Simple guidelines for recreational water users Finally at a more basic scale it should, on the basis of some of the processes highlighted in this thesis, be possible to produce a simple chart of dos and don’ts for people at risk from water that might have faecal contamination. For example for hill-walkers, where not to go to the toilet and where to avoid taking water from and times when drinking un-boiled surface water would be unwise, i.e. during or immediately after rainfall. This kind of basic information should be available to all at risk recreators on interpretation boards at all access points to recreational areas. With this kind of advice it is possible the incidence of some avoidable, unpleasant and in cases fatal, illness could be reduced.

210


References

References: Abbot. M. B. 1979. Computational Hydraulics. Pitman, pp 12, 13. Al-Layla M. A. and H. M. Al-Rizzo 1989: A Water Quality Model for the Tigris River downstream of the Saddam Dam, Iraq. Hydrological Sciences Journal 34:6, 687-704. Allen L. A., J. Grindley and E. Brooks. 1953. Some chemical and bacterial characteristics of bottom deposits from lakes and estuaries. J. Hyg. 51, 185-194. American Public Health Association. 1981. Standard methods for the analysis of water and wastewater, 15th Ed., American Public Health Association inc., Washington D.C. Ashbridge D. 1995. Processes of river bank erosion and their contribution to the suspended sediment load of the river Culm, Devon. Chapter 13 in Foster I. D. L., Gurnell A. M. and Webb B. W. (Eds) Sediment and Water Quality in River Catchments. John Wiley and Sons Ltd. Auer M. T. and Niehaus S. L. 1993. Modeling Faecal Coliform Bacteria - 1. Field and Laboratory Determination of Loss Kinetics. Wat. Res. 27, 693-701. Bagdasaryan G. A. 1964. Survival of Viruses of the Enterovirus Group in Soil and on Vegatation. J. Hyg. Epidemiol. Microbiol. Immunol. 8, 497-505. Barbe D. E. and J. C. Francis. 1995. An analysis of seasonal fecal coliform levels in the Tchefuncte River. Wat. Res. Bull. 31:1. Bathhurt J. C. 1993. Flow resistance through the channel network. Chapter 4 in Beven K. and M. J. Kirkby (Eds.), Channel Network Hydrology. Wiley. pp. 67-98. Beard P. J. 1940. Longevity of Eberthella tyhposus in Various Soils. Am. J. Publ. Health 30, 10771082. Bedient P. B. and W. C. Huber. 1988. Hydrology and Floodplain Analysis. Addison-Wesley. 650pp. Beer T. and P. C. Young. 1983 Longitudinal dispersion in natural streams. J. Environ. Engin. Div ASCE 109, 1049-1067. Bellair J. T., G. A. Parr-Smith, I. G. Wallis. 1977. Significance of diurnal variations in fecal coliform die-off rates in the design of ocean outfalls. J. Wat. Poll. Con. Fed. 2022-2030. Berndtsson R. 1990. Transport and sedimentation of pollutants in a river reach: a chemical mass balance approach. W. R. R. 26:7, 1549-1558. Beschta R. L., W. L. Jackson and K. D. Knoop. 1981. Sediment transport during a controlled reservoir release. Wat. Res. Bull. 17:4, 635-641. Beven K. and E. F. Wood. 1993. Flow routing and the Hydrological Response of Channel Networks. Chapter 5 in Beven K. and M. J. Kirkby (Eds.), Channel Network Hydrology. Wiley. pp. 99128. Biskie H. A., B. M. Sherer, J. A. Moore, J. R. Miner and J. M. Buckhouse. 1988. Fate of manure point loading into a rangeland stream. ASAE Paper 88-2081. ASAE, St. Joseph, Michigan, U.S. Bitton G. 1980. Introduction to Environmental Virology. Wiley Interscience, New York, 1980.


References

Blannon J. C. and M. L. Peterson, 1974. Survival of Faecal Coliforms and Faecal Streptococci in a Sanitary Landfill. USEPA. News of environmental research in Cincinatti. 12 April. Brettar I. and M. G. Höfle. 1992. Influence of ecosystematic factors on survival of Escherischia coli after large-scale release into lake water mesocosms. App. Env. Microbiol. 58:7, 2201-2210. Burm R. J. and R. D. Vaugn. 1966. Bacteriological Comparison Between Combined and Separate Sewer Discharges in Southeastern Michigan. J. Wat. Poll. Control. Fedn. 38:3, 400. Burn D. H. and E. A. McBean. 1986. Non-linear Optimisation Modelling of Coliform Bacteria. Water, Air and Soil Pollution 32, 183-200. Burton G. A., D. Gunnison and G. R. Lanza. 1987. Survival of Pathogenic Bacteria in Various Freshwater Sediments. App. Environ. Microbiol. 53:4, 633-638. Cabelli V. J. 1978. New Standards For Enteric Bacteria. Chapter 2 in Mitchell R. (Ed), Water Pollution Microbiology. pp233-271. John Wiley USA. Calandro A. J. 1973. Analysis of Stream Temperatures in Louisiana. Louisiana Dept. of Public Works, Technical Report 6. Canale R. P., R. L. Patterson, J. J. Gannon and W. F. Powers. 1973. Water quality models for total coliform. J. Wat. Pollut. Control Fed. 45, 325-336 Canale R. P., M. T. Auer, E. M. Owens, T. M. Heidtke and S. W. Effler. 1993. Modelling fecal coliform bacteria-II. Model development and application. Wat. Res. 27, 703-714. Carling P. A., H. G. Orr and M. S. Glaister. 1994. Preliminary observations and significance of deadzone flow structure for solute and fine particle dynamics. Chapter 7 in Beven K., P. Chatwin and J. Millbank (Eds.), Mixing and Transport in the Environment. John Wiley and Sons. Chadwick A. J. and J. C. Morphett. 1993. Hydraulics in Civil and Environmental Engineering. E.&F.N. Spon. 577pp. Carlson-Gunnoe N. E., C. B. Law, G. K. Bissonnette. 1983. In Situ Persistence of Indicator Bacteria in Streams Polluted With Acid Mine Drainage. Wat.Res. 17:9, 1119-1124. Chandler D. S., I. Farran and J. A. Craven. 1981. Persistence and Distribution of Pollution Indicator Bacteria on Land Used for Disposal of Piggery Effluent. App. Environ. Microbiol. 42:3, 453-460. Changxing S., G. Petts and A. Gurnell. 1999. Bench development along the regulated, Lower River Dee, UK. Earth surface processes and landforms 24:2, 135 Chick H. 1908. An investigation of the laws of disinfection. J. Hyg. 8, 92-158 Chow V. T. 1959. Open channel hydraulics. McGraw-Hill. 680pp. Cohen B. 1922. Disinfection studies. J. Bacteriol. 7, 183. Cohen J. and H. I. Shuval. 1972. Coliforms, Faecal Coliforms and Faecal Streptococci as Indicators of Water Pollution. Water, Air and Soil Polln. 2, 85-95. Corapcioglu M. Y. and A. Haridas. 1985. Microbial Transport in Soils and Groundwater: A Numerical Model. Adv. Wat. Res. 8, 188-200. Crane S. R. and J. A. Moore. 1986. Modelling Enteric Bacterial Die-off: A Review. Water, Air and Soil Polln. 27, 411-439.


References

Crowther J. M. 1991. Microbiological Pollution Indicators in Marine Waters. Final Report of the DoE Project. Cuthbert W. A., J. J. Panes and E. C. Hill. 1955. App. Environ. Microbiol. 12, 63. Davenport C. V., E. B. Sparrow and R. C. Gordon. 1976. Faecal Indicator Bacteria Persistence Under Natural Conditions in an Ice Covered River. Appl. Environ. Microbiol. 32, 527-536. Davis E. M. and C. S. Valentino. 1985. Die-off Characteristics of Enterococci and Other Indicator Bacteria Groups in Lake Houston. Proc. Wat. Qual. Technol. Conf. Houston. Tex. Paper No.P1-20, 509-524. Dawson F. H. 1989. Ecology and management of freshwater plants in lowland streams. Report, Freshwater Biological Association 57, 43-60. Day T. J. 1975. Longitudinal dispersion in streams. W. R. R. 11, 909-918. Dingman S. L. 1984. Fluvial Hydrology. W. H. Freeman. 383pp. Donnelly J. A., P. V. Scarpino and D. Brunner. 1982. The survival of Indicators in Lysimeters and Landfills. Wat. Sci. Tech. 14, 1558-1559. Donnelly R. F. 1991. Solar UV spectral irradiance variations. J. Geomagnet. Geoelect. 43, 835-842. Doran J. W. and D. M. Linn. 1979. Bacteriological quality of runoff water from pastureland. App. Environ. Microbiol. 37. Dufour A. P. 1977. Escherischia coli; the Faecal Coliform. In Hoadley A. W. and B. J. Dutka (Eds). Bacterial Indicators/Health Hazards Associated with Water. Am.Soc. for Testing and Materials. 48-58. Dutka B. J. 1973. Coliforms Are an Inadequate Indicator of Water Quality. J. Env. Health 36, 39-46. Elder J. F. 1987. Indicator Bacteria Concentrations as Affected by Hydrologic Variables in the Apalachicola River, Florida. Water, Air and Soil Polln. 32, 407-416. Elder J. W. 1959. The dispersion of marked fluid in turbulent shear flow. J. Fluid Mech. 5, 544-560. Enzinger R. M. and R. C. Cooper. 1976. Role of Protozoa in the Removal of Escherischia coli from Estuarine Waters. App. Environ. Microbiol. 31:5, 758-763. Evans M. R. and J. D. Owens. 1972. Factors Affecting the Concentration of Faecal Bacteria in Landdrainage Water. J. Gen. Microbiol. 71, 477-485. Evans M. R. and J. D. Owens. 1973. Soil Bacteria in Land-drainage Water. Wat. Res. 7, 1295-1300. Evison L. M. 1989. Comparative Studies on the Survival of Indicator Organisms and Pathogens in Fresh and Sea Water. Wat. Sci. Tech. 20:11/12, 309-315. Fair G. M. and J. C. Geyer. 1954. Water Supply and Waste Disposal. John Wiley and Sons Inc. N.Y. Fair G. M., J. C. Geyer and D. A. Okun. 1971. Elements of Water Supply and Waste Disposal. John Wiley and Sons Inc. N.Y. Faust M. A. 1976. Coliform bacteria from diffuse sources as a factor in estuarine pollution. Wat. Res. 10, 619-627.


References

Faust M. A. and N. M. Goff. 1977. Basin size, water flow and land-use efffects on fecal coliform pollution from a rual watershed. In Correll. D. (Ed.), Watershed research in Eastern North America. Vol. II. Chesapeake Bay Center for Environmental Studies Smithsonian Institution, Edgewater, Maryland 21037. Findlay J. S., J. C. Curran, D. P. Milne, J. M. Crowther, S. G. Wallis, B. J. B. Wood. 1990. The Selfprotection of Escherischia coli in Seawater. J. I. W. E. M. 4, 451-456. Fischer, H. B. 1967. The mechanics of dispersion in natural streams. J. Hydr. Div. Proc. ASCE 93, 187-216. Flint K. P. 1987. The Long-term Survival of Escherischia coli in River Water. J. App. Bacteriol. 63, 261-270. Foster I. D. L., A. M. Gurnell and B. W. Webb (Eds). 1995. Sediment and water quality in river catchments. John Wiley and Sons. Frost W. H. and H. W. Streeter. 1924. In Streeter H. W. 1934. A Formulation of Bacterial Changes Occurring in Polluted Water. Sewage Wks. Journal 6:2, 208-235. Fujioka R. S. and E. B. Siwak. 1985. The Cidal Effect of Sunlight on Alternative Microbial Indicators of Water Quality. Proc. Wat. Qual. Tech. Conf. Houston.Texas. Gameson A. L. H. and D. J. Gould. 1975. Effects of Solar Radiation on the Mortality of some Terrestrial Bacteria in Sea Water. In A. L. H. Gameson (Ed.). Proceedings of the International Symposium on Discharge of Sewage from Sea Outfalls. Pergamon Press, London. Gameson A. L. H. and J. R. Saxon. 1967. Field studies on the effect of daylight on mortality of coliform bacteria. Wat. Res. 1, 279-295. Gannon J. J., Busse M. K. and Schillinger J. E. 1983. Fecal coliform disappearance in a river impoundment. Wat. Res. 17, 1595-1601. Garcia M., Y. Nino and F. Lopez. 1996. Laboratory Observations of Particle Entrainment into Suspension by Turbulent Bursting. Chapter 3 in Ashworth P. J., S. J. Bennet, J. L. Best and S. J. McLelland. (Eds.) Coherent Flow Structures in Open Channels. John Wiley and sons. Gary H. L., and J. C. Adams. 1985. Indicator bacteria in water and stream sediments near the Snowy Range in southern Wyoming. Water, Air and Soil Pollution 25, 133-144. Geldreich E. E. 1970. Applying Bacteriological Parameters to Recreational Water Quality. J. Am. Wat. Wks. Assoc. 62, 113-120. Geldreich E. E. and B. A. Kenner. 1969. Concepts of Faecal Streptococci in Stream Pollution. J. Wat. Poll. Cont. Fed. 41:8, 336-351. Geldreich E. E., L. C. Best, B. A. Kenner, D. J. Van Donsel. 1968. The Bacteriological Aspects of Storm-water Pollution. Wat. Polln. Cont. Fed. 40:11, 1861-1872. Germann P. F., M. S. Smith, and G. W. Thomas. 1987. Kinematic Wave Approximation to the Transport of Escherischia Coli in the Vadose Zone. W. R. R. 23:7, 1281-1287. Gippel C. J. 1989. The Use of Turbidity Instruments to Measure Stream Water Suspended Sediment Concentration. Monograph Series No. 4, Department of Geography and Oceanography, University College, Australian Defense Force Academy, Canberra. Godfree A. F. 1994. Personal communication.


References

Gordon R. C. 1972. Winter Survival of Faecal Bacterial in a Sub-Arctic Arctic River. USEPA. EPAR2-72-013. Gould D. J. and M. R. Fletcher. 1978. Gull Droppings and Their Effects on Water Quality. Wat. Res. 12:9, 665-672. Graf. W. H. 1971. Hydraulics of Sediment Transport. McGraw-Hill. Graham A. A. 1990. Siltation of stone-surface periphyton in rivers by clay-sized particles from low concentrations in suspension. Hydrobiologia 199, 107-115. Green H. M., K. J. Beven, K. Buckley and P. C. Young. 1994. Pollution prediction with uncertainty. Chapter 6 in Beven K., P. Chatwin and J. Millbank (Eds.), Mixing and Transport in the Environment. John Wiley and Sons. Grimes D. J. 1974. Release of Sediment-Bound Coliforms by Dredging. App. Microbiol. 29:1, 109111. Gurnell A. M. 1997. Adjustments in river channel geometry associated with hydraulic discontinuities across the fluvial-tidal transition of a regulated river. Earth surface processes and landforms 22:10, 967-985. Gurnell A. M., M. Bickerton, P. Angold, D. Bell, I. Morrisey, G. E. Petts and J. Sadler. 1998. Morphological and ecological change on a meander bend: the role of hydrological processes and the application of GIS. Hydrological Processes 12, 981-993. HMSO. 1983. The bacteriological examination of water supplies 1982. Reports on Public Health and Medical Subjects No. 71. Methods for the examination of waters and associated materials. HMSO, London. Hagedorn C., E. L. McCoy and T. M. Rahe. 1981. The Potential for Groundwater Contamination from Septic Effluents. J. Environ. Qual. 10:1. Harvey R. W. 1989. Transport of bacteria in a contaminated aquifer. USGS Toxic substances hydrology program. Proceedings of of the technical meeting, Pheonix, Arizona. 183-188. Henderson F. M. 1966. Open Channel Flow. MacMillan. Henderson-Sellers B., P. C. Young and J. Ribiero da Costa. 1988. Water quality models: Rivers and reservoirs. In: Proceedings of the international symposium on water quality modelling of agricultural non-point sources, Part 1. Utah State University, Logan, Utah. ARS-81. Hendricks C. W. and S. M. Morrison. 1967. Multiplication and Growth of Selected Enteric Bacteria in Clear Mountain Stream Water. Wat. Res. 1, 567-576. Hendry G. S. and A. Toth. 1981. Some Effects of Land Use on Bacteriological Water Quality in a Recreational Lake. Wat. Res. 16, 105-133. Himmelblau D. M. and R. V. Yates. 1968. A new method of flood routing. Wat. Res. 4, 1193-1199. Hirn J., H. Viljamaa and M. Raevouri. 1979. The Effect of Physiochemical, Phytoplankton and Seasonal Factors on Faecal Indicator Bacteria in Northern Brakish Water. Wat. Res. 14, 279285. Hjulström F. 1935. The morphological activity of rivers as illustrated by the Rivers Fyris. Bull. Geol. Inst. Uppsala 25:III.


References

Hoagland K. D., S. C. Roemer and J. R. Rosowski. 1982. Colonisation and community structure of two periphyton assemblages, with emphasis on the diatoms (Bacillariphyceae). Am J. Bot. 69, 188-213. Hornberger G. M., L. M. Mills, and J. S. Herman. 1992. Bacterial Transport in Porous Media: Evaluation of a Model using Laboratory Observations. W. R. R. 28:3, 915-938. Hornung M., S. Le-grice., N. Brown and D. Norrise. 1990. The role of geology and soils in controlling surface water acidity in Wales. pp 55-66 In: Edwards R. W., A. S. Gee and J. H. Stoner (Eds.), Acid Waters in Wales. Kluwer academic publishers, Dordrecht. Hydraulic Research Laboratory. 1985. The Hydraulic Roughness of vegetation in open channels. Report IT 281. Hydraulics Research Ltd., Wallingford. Hunter C. and A. T. McDonald. 1991a. Seasonal changes in the sanitary bacterial quality of water draining a small upland catchment in the Yorkshire Dales. Wat. Res. 25:4, 447-453. Hunter C. and A. T. McDonald. 1991b. The Occurence of Coliform Bacteria in the Surface Soils of Two Catchment Areas in the Yorkshire Dales. J. I. W. E. M. 5:5, 534-538. Hunter C., A. T. McDonald, K. J. Beven. 1992. Input of Faecal Coliform Bacteria to an Upland Stream Channel in the Yorkshire Dales. W. R. R. 28:7, 1869-1876. Hunter C., J. Perkins, J. Tranter and J. Gunn. 1999. Agricultural Land-use Effects on the Indicator Bacterial Quality of an Upland Stream in the Derbyshire Peak District in the UK. Wat. Res. 33:17, 3577-3586. Jakeman A. J., I. J. Littlewood and P. G. Whitehead. 1990. Computation of the Instantaneous Unit Hydrograph and Identifiable Component Flows with Application to Two Small Upland Catchments. J. Hydrol. 117, 275-300. Jeje L. K., O. Ogunkoya, J. M. Oluwatimilehin. 1991. Variation in suspended sediment concentration during storm discharges in three small streams in the upper Osun Basin, Central Western Nigeria. Hydrol. Processes 5, 361-369. Jenkins A. 1984. The Effect of Fluvial Processes on the Dynamics of Sanitary Bacteria in Upland Streams. Ph.D. Thesis, University of Leeds. Jenkins A., M. J. Kirkby, A. T. McDonald, P. Naden and D. Kay. 1984. A Process Based Model of Faecal Bacterial Levels in Upland Catchments. Wat. Sci. Tech. 16, 453-462. Johnson F. A. 1971. Stream temperatures in an alpine area. J. Hydrol. 14, 322-336. Johnson M. C. and D. E. Ford. 1987. Two-Dimensional Modelling of Storm Inflow Processes at DeGray Lake Arkansas. Proc. DeGray Lake Symposium Tech. Rep. E-78-4. Kay D. and A. T. McDonald. 1980a. Towards an understanding of the effects of catchment management on water quality. Technical Paper. Lancaster IBG. Kay D. and A. T. McDonald. 1980b. Reduction of coliform bacteria in two upland reservoirs: the significance of distance decay relationships. Wat. Res. 14:5, 305-318. Kay D. and A. T. McDonald. 1982. Enteric bacterial entrainment in a recreational channel following reservoir releases for competitive canoeing. Cambria 9:2, 61-68. Kay D. and A. T. McDonald. 1983. Predicting coliform concentrations in upland impoundments: Design and calibration of a multivariate model. Appl. Environ. Microbiol. 46:3, 611-618.


References

Kay D. and J. Stoner. 1988. The effects of catchment land use on stream water quality in an acidsusceptible region of West Wales: The implications for compliance with EC drinking water and fishery directives. Applied Geography 8, 191-205. Kay D., M. Wyer, A. T. McDonald and N. Woods. 1990. The Application of Water Quality Standards to UK Bathing Waters. J. I. W. E. M. 4, 436-441. Kirk J. T. O. 1983. Light and photosynthesis in aquatic ecosystems. Cambridge University Press. Kress M. and G. F. Gifford 1984. Fecal Coliform Release from Cattle Fecal Deposits. Wat. Res. Bull. 20:1, 61-66. Kott Y. 1977. Current Concepts of Indicator Bacteria. In Hoadley A. W. and B. J. Dutka (Eds). Bacterial Indicators/Health Hazards Associated with Water. Am. Soc. for Testing and Materials. 3-13. Kunkle S. H. 1970. Sources and Transport of Bacterial Indicators in Rural Streams. Proc. of Symposium on Interdisciplinary Aspects of Watershed Management. 3-6 August. Bozemann, Montana. Kunkle S. H. and J. R. Meiman. 1968. Sampling Bacteria in a Mountain Stream. Colorado State University Hydrology Paper No.28. Lambert A. O. 1983. Time of travel of pollution on the lower Dee. Welsh Water Internal Report No. 77:013A. Lawler D. M. 1995. The impact of scale on the processes of channel-side sediment supply: a conceptual model. In. Effects of scale on interpretation and management of sediment and water quality (Proceedings of a Boulder Symposium, July 1995). IAHS Publ. no. 226. Lighthill M. J. and G. B. Whitman. 1955. On Kinematic Waves, I, Flood Movement in Long Rivers. Proc. Roy. Soc. London 229, 281-316. Littlewood I. G. and A. J. Jakeman. 1993. Characterisation of quick and slow stream flow components by unit hydrographs for single- and multi- basin studies. Chapter 12 in Robinson M. (Ed.) Methods of hydrological basin comparison. Institute of Hydrology report 120, 99-111. Littlewood I. G. and A. J. Jakeman. 1994. A new method of rainfall-runoff modelling and its applications in catchment hydrology. Chapter 6 in Zannetti P. (Ed) Environmental Modelling, Vol. II: 143-171. Computational Mechanics Publications, Southampton, UK. MacIllveen J. F. R. 1986. Basic Meteorology. Van Nostrand Reinhold (UK). pp. 457. Mack W. N. 1977. Total Coliform Bacteria. In A.W. Hoadley and B.J. Dutka (Eds.) Bacterial Indicators / Health Hazards Associated with Water. American Society for Testing and Materials. STP 635. pp55-64. Mahloch J. L. 1973. Comparative Analysis of Modelling techniques for Coliform Organisms in Streams. App. Microbiol. 27:2, 340-345. Marshall K. C. 1979. Reactions of Microorganisms, Ions and Macromolecules at Interfaces. In Ellwood (Ed). Interfaces in Microbial Ecology. Harvard U.P, Cambridge. Martin J. L. and S. C. McCutcheon. 1998. Hydrodynamics and Transport for Water Quality Modelling. Lewis. 794pp.


References

Matson E. A., S. G. Hornor, J. D. Buck. 1978. Pollution Indicators and other Microorganisms in River Sediment. J. Wat. Polln. Control Fed. January, 13-19. Matthess G., A. Pekdeger and J. Schroeter. 1988. Persistence and transport of bacteria and viruses in groundwater - a conceptual evaluation. J. Contam. Hydrol. 2, 171-188. McCalla T. M. 1964. Why Does H+ Ion Become Toxic to Soil Bacteria? Soil Sci. Soc. Am. Proc. 6, 165-167. McCave I. N. 1984. Erosion, transport and deposition of fine grained suspended sediments. pp. 35-70 in Stow D. A. and D. J. W. Piper (Eds), Fine grained sediments: deep water processes and facies. Blackwell Scientific Publications, Oxford. McDonald A. T., D. Kay and A. Jenkins. 1982. Generation of Faecal and Total Coliform Surges by Stream Flow Manipulation in the Absence of Normal Hydrometeorological Stimuli. App. Environ. Microbiol. 44, 292-300. McFeters G. A. and D. G. Stuart. 1972. Survival of Coliform Bacteria in Natural Waters: Field and Laboratory Studies with Membrane Filter Chambers. App. Microbiol. 24:5, 805-811. McSwain M. R. and W. T. Swank. 1977. Fluctuations in Naturally Occuring Populations of Enteric Bacteria in Oligotrophic Streams of Western North Carolina. USDA Forest Service Paper SE-158. Menon A. S. 1985. Salmonellae and Pollution Indicator Bacteria in Municipal and Food Processing Effluents in the Cornwallis River. Can. J. Microbiol. 31, 598-603. Mezrioui N., B. Oudra, K. Oufdou, L. Hassani, M. Loudiki and J. Darley. 1994. Effect of microalgae growing on wastewater batch culture on Escherischia Coli and Vibrio Cholerae survival. Wat. Sci. Tech. 30:8, 295-302. Milhous R. T. 1982. Effect of Sediment Transport and Flow Regulation on the Ecology of Gravelbed Rivers. Chapter 29 in Hey R. D., J. C. Bathurst and C. R. Thorn (Eds), Gravel Bed Rivers. John Wiley and Sons Ltd. Milne D. P., J. C. Curran and L. Wilson. 1986. Effects of Sedimentation on the Removal of Faecal Coliform in Estuarine Water. Wat. Res. 20:12, 1493-1496.

Milne D. P., J. C. Curran, J. S. Findlay, J. M. Crowther and B. J. B. Wood. 1991. The Effect of Dissolved Nutrients and Inorganic Suspended Solids on the Survival of E.coli in Seawater. Wat. Sci. Tech. 24:2, 133-136. Milne D. P., J. C. Curran, J. S. Findlay, J. M. Crowther and S. G. Wallis. 1989. The Effect of Estuary Type Suspended Solids on Survival of E.coli in Saline Waters. Wat. Sci. Tech. 21:3, 61-65. Milne R. A., P. V. Varallo, P. G.Hilder, S. Boneham, R. Rees. 1981. Fish populations in the rivers Ystwyth and Rheidol in relation to toxic zinc, lead and cadmium concentrations. Draft Final Report. Welsh Water Authority. Mitchell D. O. and M. J. Starzyk. 1975. Survival of Salmonella and other indicator microoganisms. Can. J. Microbiol. 21, 1420. Mitchell R. and C. E. Chamberlain. 1978. Survival of Indicator Organisms. Chapter 2 in Berg G. (Ed.). Indicators of Viruses in Water and Food. Ann-Arbour Science. pp15-37.


References

Moore J. A., J. Smith, S. Baker, and J. R. Miner. Evaluating coliform concentrations in runoff from various animal waste management systems. Special Report 817. Agric. Exp. Station. Oregon State University, Corvallis. Morrison S. M. and J. F. Fair. 1966. Influence of Environment on Stream Microbial Dynamics. Hydrology Paper No.13. Colorado State University. N.R.A. 1991. Proposals for Statutory Water Quality Objectives. Neal C., C. J. Smith, H. A. Jeffery, H. P. Jarvie and A. J. Robson. 1995. Trace element concentrations on the major rivers entering the Humber Estuary, NE England. J. Hydrol. 182, 37-64. Neal C., H. P. Jarvie and T. Oguchi. 1999. Acid-available particulate trace metals associated with suspended sediment in the Humber rivers: a regional assessment. Hydrol. Processes 13, 1117-1136. Niemi M. 1985. Faecal Indicator Bacteria at Freshwater Rainbow Trout (Salmo gairdneri) Farms. National Board of Waters, Helsinki. Publication of the Water Research Institute Report No.64:49. Nordin, C. F. and G. V. Sabol. 1979. Empirical data on longitudinal dispersion in rivers. USGS, Lakewood, 20-74. Nuttall D. The populations, characterisation and activity of suspended bacteria in the Welsh River Dee. J. App. Bacteriology 53, 49-59. Nuttall D. and O. T. Parry. 1987. The effect of environmental factors on coliphages in the Welsh River Dee. Letters in Applied Microbiology 4, 117-120. Oragui J. I. and D. D. Mara. 1981. Occurrence of Rhodococcus Coprophilus and Associated Actinomycetes in Faeces, Sewage, and Freshwater. App. Environ. Microbiol. 42:6, 10371042. Pap J. M., C. Frohlich. 1999. Total solar irradiance variations. J. Atmos. Solar-Terrestrial Physics. 61, 15-24. Peixoto J. P and A. H. Oort. 1992. Physics of Climate. American Institute of Physics, New York, pp. 520. Penman H. L. 1948. Natural evaporation from open water, bare soil and grass. Proc. Roy. Soc. Lond. 193, 120-145. Pettibone G. W., K. N. Irvine, K. M. Monahan. 1996. Impact of a ship passage on bacteria levels and suspended sediment characteristics in the Buffalo River, New York. Wat. Res. 30:10, 25172521. Phelps E. B. 1944. Stream Sanitation. Chapter 6 in, Public Health Aspects of Bacterial Pollution. John Wiley and Sons Inc. N.Y. Piper B. S., C. A. L. Ishemo and P. G. Whitehead. 1991. River Water Quality Models for Environmental Planning and Management. Unpublished Paper IH Wallingford. Price L. E., P. Goodwill, P. C. Young and J. S. Rowan. 2000. A data-based mechanistic modelling (DBM) approach to understanding dynamic sediment transmission through Wyresdale Park Reservoir, Lancashire, UK. Hydrol. Processes 14, 63-78. Pommepuy M., J. F. Guillaud, E. Dupray, A. Derrien, F. Le Guyader & M. Cormier. 1992. Enteric bacteria survival factors. Wat. Sci. Tech. 25:12, 93-103.


References

Qin D., J. P. Bliss, D. Barnes and P. A. FitzGerald. 1991. Bacterial (total coliform) die-off in maturation ponds. Wat. Sci. Tech. 23, 1525-1534. Reynolds C. S. 1979. Seston sedimentation: experiments with lycopodium spores in a closed system. Freshwat. Biol. 9, 55-76. Reynolds C. S. 1992. Algae. Chapter 9 in Calow P. and G. E. Petts (Eds.) The Rivers Handbook: Hydrological and Ecological Principles. Blackwell Scientific Publications. Robbins J. W. D., D. H. Howells and G. J. Kriz. 1972. Stream pollution from animal production units. J. Wat. Polln. Control. Fed. 44, 1563-1544. Robson A.J., C. Neal, J.C. Currie, W.A. Virtue, A. Ringrose. 1996. The water quality of the River Tweed and its tributaries. Report No. 128. Institute of Hydrology, Wallingford. Rouse H. 1946. Elementary Mechanics of Fluids. John Wiley. Roper M. M. and K. C. Marshall. 1978. Effects of clay mineral on microbial predation and parasitism of Escherischia coli. Microbial Ecology 4, 279-289. Rowan J. S. 1995. The erosional transport of radiocaesium in catchment systems: a case study of the Exe Basin, Devon. Chapter 18 in Foster I. D. L., A. M. Gurnell and B. W. Webb (Eds), Sediment and water quality in river catchments. John Wiley and Sons. Roszak D. B. and R. R. Colwell. 1987. Survival strategies of bacteria in the natural environment. Microbiol. Rev. 51:3, 365-379. Rushbrooke J. N. and F. Beaumont. 1986. River Dee Scheme and intake protection systems. J. Wat. Engineers and Scientists 40:2, 173-192. Sarikaya H. Z., A. M. Asce and A. M. Saatci. 1987. Bacterial Die-off in Waste Stabilization Ponds. J. Env. Eng. 113:2, 366-382. Sarikaya H. Z., A. M. Asce, A. M. Saatci and A. F. Abdulfattah. 1987. effect of pond depth on bacterial die-off. J. Env. Eng. 113:6, 1350-1362. Sherer B. M., J. R. Miner, J. A. Moore and J. C. Buckhouse. 1988. Resuspending organisms from a rangeland stream bottom. ASAE. 31:4, 1217-1222. Sherer B. M., J. R. Miner, J. A. Moore and J. C. Buckhouse. 1992. Indicator bacteria survival in stream sediments. J. Env. Qual. 21:4, 591-595. Singleton P. and D. Sainsbury. 1981. Dictionary of Microbiology. John Wiley and Sons Ltd. Smith K. 1981. The prediction of river water temperatures. Hydrol. Sci. Bull. 26, 19-32. Smith M. S., G. W. Thomas, R. E. White, and D. Ritonga. 1985. Transport of Escherischia coli Through Intact and Disturbed Soil Columns. J. Env. Qual. 14:1, 87-91. Spencer J. W. 1971. Fourier series representation of the position of the sun. Search. 2, 172. Sorensen D. L., S. G. Eberl and R. A. Dicksa. 1989. Clostridium perfringens as a Point Source Indicator in Non-Point Polluted Streams. Wat. Res. 23:2, 191-197. Stainer R. Y., E. A. Adelberg and J. L. Ingraham (Eds.). 1984. General Microbiology. 4th.Edn. MacMillan Press Ltd.


References

Standridge J. H., J. J. Delfino, L. B. Kleppe and R. Butler. 1979. Effect of Water Fowl (Anas platyrhynchos) on Indicator Bacteria Populations in a Recreational Lake in Madison, Arkansas. App. Environ. Microbiol. 38:3, 547-550. Stapleton C. M., J. Wilkinson, D. Kay, L. Fewtrell and M. D. Wyer. 2000a. Lake Windermere bathing waters bacterial impact study. Report to the Environment Agency (N.W. Region). 156pp plus Figures and Appendices. CREH, University of Wales, Aberystwyth. Stapleton C. M., J. Wilkinson, D. Kay, L. Fewtrell and M. D. Wyer. 2000b. Morecambe Bay bathing waters bacterial impact study. Report to the Environment Agency (N.W. Region). 199pp plus Figures and Appendices. CREH, University of Wales, Aberystwyth. Stapleton C. M., J. Wilkinson, M. D. Wyer, L. Fewtrell, and D. Kay. (1999) Seascale and St Bees bathing waters bacterial impact study. A confidential report to Environment Agency (N.W. Region). 55pp (plus Figures). CREH, University of Wales, Aberystwyth. Stephenson G. R. and R. C. Rychert. 1982. Bottom Sediment: a Reservoir of Escherischia coli in Rangeland Streams. J. Range. Man. 35:1, 119-123. Streeter H. W. 1934. A Formulation of Bacterial Changes Occurring in Polluted Water. Sewage Wks. Journal 6:2, 208-235. Sumner A. M. 1982. Some experiences in the treatment of River Dee water for public supply. J. Inst. Water. Engineers and Scientists 33:5, 455-466. Taylor G. I. 1954. The dispersion of matter in turbulent flow through a pipe. Proc. Roy. Soc. London Ser. A223, 446-468. Taylor G. S., P. Hillis and I. Walker. 1993. Pilot-plant trials on the River Dee at Huntington. J. I. W. E. M. 7:4. Tipping E., C. Woof and K. Clarke. 1993. Deposition and resuspension of fine particles in a riverine 'dead zone'. Hydrol. Processes 7, 263-277. Uchrin C. G. and W. J. Weber. 1979. Modelling of Transport Processes for Suspended Solids and Associated Bacteria in River-Harbour-Lake Systems. Proc. symposium on Processes Involving Contaminants and Sediments Modelling Section. Honolulu, Hawaii, 1-6 April. Uchrin C. G. and W. J. Weber. 1980. Modelling Suspended Solids and Associated Bacteria in Natural Water Systems. ASCE Env. Eng. 8-10, 160-166. Uchrin C. G. and W. J. Weber. 1981. Modelling Stormwater Runoff Induced Bacterial Contamination of Belleville Lake (MI). Proc. ASCE Water Forum '81. 944-952. Valentine E. M. and I. R. Wood. 1977. Lonigtudinal dispersion with Dead Zones. J. Hydr. Div. ASCE 103, 975-990. Van Donsel D. J. and E. E. Geldreich. 1971. Relationships of Salmonellae to Faecal Coliforms in Bottom Sediments. Wat. Res. 5, 1079-1087. Van Donsel D. J., E. E. Geldreich and N. A. Clarke. 1967. Seasonal Variations in Survival of Indicator Bacteria in Soil and Their Contribution to Storm-water Pollution. App. Microbiol. 15:6, 1362-1370. Verstraete W. and J. P. Voets. 1972. Comparative Study of E.coli in Two Aquatic Ecosystems. Wat. Res. 10, 129-136.


References

Walling D. E. and B. W. Webb. 1992. Water Quality I. Physical Characteristics. In, Calow P. and G. E. Petts (Eds.), The Rivers Handbook: Hydrological and Ecological Principles. Blackwell Scientific Publications. Walling D. E. and B. W. Webb. 1982. Sediment availability and prediction of storm period sediment yields. IAHS Publ. 137, 327-337. Walling D. E., Owens P. N. and Leeks G. L. J. 1999. Fingerprinting suspended sediment sources in the catchment of the River Ouse, Yorkshire, UK. Hydrol. Processes. 13, 955-975. Wallis S. G. 1993 Aggregated mixing zone modelling of solute transport in rivers. Proc. B.H.S. 4th. National Hydrology Symposium. Cardiff. Wallis S. G. and P.C. Young. 1986. The aggregated dead zone (ADZ) model for dispersion in rivers. Proc. BHRA Conference on water quality modelling in the inland natural environment. BHRA. Plymouth. pp. 421-433. Wallis S. G., J. M. Crowther, J. C. Curran, D. P. Milne and J. S. Findlay. 1989a. Consideration of One-dimensional Transport Model of The Upper Clyde Estuary. Chapter 3 in Palmer H. A. (Ed.), Advances in Water Modelling and Measurement. BHRA. Wallis S. G., P. C. Young. and K. J. Beven. 1989b. Experimental investigation of the aggregated dead zone model for longitudinal solute transport in stream channels. Proc. Inst. Civ. Engrs. 87:2, 1-22. Wass P. D. and G. L. J. Leeks. 1999. Suspended sediment fluxes in the Humber catchment, UK. Hydrol. Processes 13, 935-953. Weiss C. M. 1951. Adsorption of E.coli on River and Estuarine Silts. Sewage and Ind. Wastes 23:2, 227-237. Weston A. E. 1979. The measurement of interactive freshwater and tidal flows in the River Dee, North Wales. J. Inst. Water Engineers and Scientists 33, 69-79. White R. E. 1979. Introduction to the Principles and Practice of Soil Science. Blackwell Scientific Publications. Whitehead P.G. 1978. WATERCRES Study of Water Quality in the Lower Molonglo, Murrumbidgee Burrinjuck Region of the ACT and NSW. First Progress Report. CRES Report No. AS/R19 1978. ANU, Canberra, Australia. Whitehead P.G. and P. C. Young. 1975. A dynamic-stochastic model for water quality in part of the Bedford Ouse River System. In G. C. Vansteenkiste (Ed.), Computer simulation of water resources systems. North Holland / American Elsevier, Amsterdam / New York. Whitehead P.G. and G. Hornberger. 1984. Modelling algae behaviour in the River Thames. Wat. Res. 18, 945-953. Whitehead P.G., P. C. Young and G. Hornberger. 1979. A Systems Model of Stream Flow and Water Quality in the Bedford-Ouse River - I. Stream Flow Modelling. Wat. Res. 13, 1155-1169. Whitehead P.G. and R. J. Williams. 1982. A Dynamic Nitrate Balance Model for River Basins. IAHS Exeter Conference Proceedings, IAHS publication No 139. Wilkinson J. H. 1945. Translatory waves in natural channels. Trans. Am. Soc. Civ. Eng. 110, 12031225.


References

Wilkinson R. J., A. Jenkins, M. Wyer and D. Kay. 1995a. Modelling faecal coliform dynamics in streams and rivers. Wat. Res. 29:3, 847-855. Wilkinson R. J., A. Jenkins, M. Wyer and D. Kay. 1995b. Modelling faecal coliform concentrations in streams. IH Report No 127, IH Wallingford, UK. Wilkinson W. B., G. L. J. Leeks, A. Morris and D. E. Walling. 1997. Rivers and coastal research in the Land Ocean Interaction Study. Sci. Tot. Environ. 194/195, 5-14. Wohl E. E. and D. A. Cenderelli. 2000. Sediment deposition and transport patterns following a reservoir sediment release. W.R. R. 36:1, 319-333. Wyer M. D., G. Jackson, D. Kay, J. Yeo and H. Dawson. 1994. An assessment of the impact of inland surface water input to the bacteriological quality of coastal waters. J. I. W. E. M. 8, 459-467. Wyer M. D., J. Crowther, D. Kay, C. M. Stapleton and J. Wilkinson. 1999. Faecal indicator organism sources and budgets for the Tendring Peninsula, Essex. Report to Anglian Water. Centre for Research into Environment and Health. University of Wales, Aberystwyth. Yates M. V. and S. R. Yates. 1988. Modelling Microbial Fate in the Subsurface Environment. CRC Critical Reviews in Environmental Science. 17:4, 307-344. Young C. C. and H. Greenfield. 1923. Observations on the Variability of the Bacterium coli Group Under Natural and Artificial Conditions. Am. J. Public Health. 13, 270. Young P. C. 1986. Time series methods and recursive estimation in hydrological systems analysis. In Kraijenhoff D. A. and J. R. Moll (Eds.), River Flow Modelling and Forecasting. Dordrecht: Reidel, pp.129-80. Young P. C. 1992. Parallel processes in hydrology and water quality - a unified time-series approach. J. I. W. E. M. 6:5, 598-612. Young P. C. and S. Benner. 1991. microCAPTAIN handbook Version 2.0. Centre for Research on Environmental Systems, Intstitute of Environmental and biological Sciences, Lancaster University. Young P. C. and S. G. Wallis. 1986. The Aggregated Dead Zone Model for Dispersion in Rivers. Proc. International Conference on Water Quality Modelling in the Inland Natural Environment. Bournemouth, England. Young P.C. and S. G. Wallis. 1993. Solute Transport and Dispersion in Channels. Chapter 6 in Beven K. and M. J. Kirkby (Eds.), Channel Network Hydrology. Wiley. pp.99-128. Zanoni A. E., W. J. Katz, H. H. Carter and R. C. Whaley. 1978. An In-Situ Determination of the Disappearance of Coliforms in Lake Michigan. Wat.Poll.Cont.Fed. 78, 321-329. Zerfas J. W. 1970. Survival of Faecal Coliforms and Faecal Streptococci in River Water. Unpublished M.Sc Thesis. South Dakota State University. Zyman, J. and C. A. Sorber. 1988. Influence of Simulated Rainfall on the Transport and Survival of Selected Indicator Organisms in Sludge Amended Soils. J. Wat. Poll. Con. Fed. 60, 20152110.


APPENDIX I: Die-off in fresh-water

APPENDIX I: Freshwater die-off rates of coliform bacteria measured in situ (after Mitchell and Chamberlain, 1978). System

Temperature Indication

T90 (hr)

k (day-1)

Ohio River

Summer (20C) Winter (5C)

47 51

1.18 1.08

Upper Illinois River

June-September October and May December-March April and November

27 22 95 53

2.04 2.52 0.58 1.03

Lower Illunois River

June-September October and May December-March April and November

27 63 90 80

2.04 0.89 0.62 0.70

3.64

15.12

"Shallow turbulent stream" Missouri River

Winter

115

0.48

Tennessee River (Knoxville)

Summer

53

1.03

Tennessee River (Chattanooga)

Summer

42

1.32

Sacramento River

Summer

32

1.73

Cumberland River

Summer

10

5.52

2.1

26.4

100

0.50

135

0.41

Glatt River Groundwater stream

10C

Leaf River (Mississippi) Wastewater lagoon

7.9-25.5C

276-79

0.20-0.70

Maturation ponds

19C

28 33

1.99 1.68

Oxidation ponds

20C

21.3

2.59

225


APPENDIX I: Die-off in fresh-water

Die-off rate, k (days-1)

Length of study

Season or temperature (C)

pH

Organism type

Aquatic system description

Bacterial die-off in the freshwater environment (after Crane and Moore, 1986).

Well water inoculated with pure cultures (field, membrane filter) (McFeters et al., 1974)

Coliforms Entercocci Coliform Streptococci Streptococcus equinus Streptococcus bovis Shigella dysenteriae Streptococcus sonnei Streptococcus flexneri Salmonella paratyphi A Streptococcus paratyphi D Streptococcus typhimurium Streptococcus typhi Vibrio cholerae Streptococcus paratyphi B

7.48

10.12

4 days

0.285 0.221 0.277 0.249 0.485 0.128 0.217 0.198 0.181 0.303 0.253 0.303 0.809 0.673 2.022

Stream water (membrane filter) McFeters and Stuart, 1972)

field study

Escherichia coli

4.6

5 days

lab study

E. coli

8.37 8.10 8.10

1.970 3.140 0.151 0.231 0.495 0.990 1.386 6.930 0.630 0.433 0.330 0.347 0.770

20 days

0.192 0.144 0.256 0.288 0.256 0.288 0.383 0.461 0.177 0.144 0.288 0.329

Modelling Faecal Coliform Dynamics in Streams and ...

Modelling Faecal Coliform Dynamics in Streams and ...

Suggest Documents

Spatial and temporal faecal coliform variations in surface ... - CiteSeerX

Coliform Bacteria in Sierra Nevada Wilderness Lakes and Streams

Risk Assessment of Total and Faecal Coliform Bacteria From Drinking ...

a comparison between faecal sterols and coliform ... - Revistas USP

Level of Faecal Coliform Contamination of Drinking Water Sources ...

Development of a Protocol for Sampling Faecal Coliform Bacteria ...

Comparison of methods to analyse imprecise faecal coliform count ...

Generic Modelling of Faecal Indicator Organism ... - MDPI

Seston dynamics in southern Appalachian streams - CiteSeerX

Modelling with Streams in Daisy and The SchemEngine Project

modelling valve dynamics and flow in reciprocating

Modelling Lobbying Behaviour and Interdisciplinarity Dynamics in

Mathematical Modelling in Population Dynamics

Phosphorus and Nitrogen Dynamics in Streams Associated with Wildfire

Phosphorus dynamics in selected wetlands and streams of the lake ...

Trout Production Dynamics and Water Quality in Minnesota Streams

Faecal coliforms, faecal enterococci, Salmonella Typhi and ...

Dynamics of copepod faecal pellets in relation to a Phaeocystis ...

DISCUSSION: PERSISTENT COLIFORM CONTAMINATION IN

Coliform Bacteria and Drinking Water

The effect of algal biomass on the removal of faecal coliform from ...

Modelling flux tube dynamics

Dynamics & Modelling - Mechatronics Academy

Dynamics & Modelling - Mechatronics Academy