the effect of sampling frequency on chemical ... - Science Direct

PII:

S0269-749

I (96)00028-0

Environmental Pollution, Vol. 93, No. 2, pp. 147 157, 1996 Copyright © 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0269-7491/96 $15.00 + 0.00

ELSEVIER

THE EFFECT OF SAMPLING FREQUENCY ON CHEMICAL PARAMETERS IN ACID-SENSITIVE STREAMS P. A. Brewin, a B. Reynolds, b P. A. Stevens, b A. S. Gee C & S. J. O r m e r o d a aCatchment Research Group, School of Pure and Applied Biology, University of Wales Cardiff, PO Box 915, Cardiff CF1 3TL, UK ~lnstitute of Terrestrial Ecology, Bangor Research Unit, University of Wales Bangor, Deiniol Road, Bangor, Gwynedd LL57 2UP, UK CEnvironment Agency, Rivers House, St Mellons Business Park, St Mellons, Cardiff CF3 0LT, UK (Received 22 August 1995; accepted 26 February 1996)

Abstract We assessed the effects of different simulated sampling regimes (weekly, fortnightly, monthly and bimonthly) on parameters describing the water chemistry of 72 streams in acid-sensitive areas of Wales. For pH, sulphate, total hardness and aluminium, reduced sampling frequency had no discernible or systematic effect on the apparent annual mean chemistry relative to the values derived from weekly data. Standard deviations and coefficients of variation were either unaffected, or were reduced. However, sampling frequency had a moderate effect on mean p H when the data were separated into seasons: winter mean pH increased on average by 0.13 units and summer means decreased on average by 0.12 units, when using bimonthly data relative to weekly. Extreme values were detected less effectively at lower sampling frequencies, significantly altering the intercept and~or the slope of the strong relationships between the means and minimum p H or maximum Al. These effects almost certainly reflect the exclusion of extreme events (summer drought and winter floods) from low sample frequencies and reveal limitations in the use of mean values from periodic sampling programmes for summarising some aspects of site chemistry. Nevertheless, previously established relationships between mean stream chemistry, land use and stream biology were still strong at all sampling frequencies. Clear recommendations about the needs to fully parameterise episodic fluctuations depend on unanswered questions about: (i) whether biota respond to mean or episodic chemical conditions and (ii) whethe.r baseflow chemistry, episodic fluctuations, or some combination of these, will best reflect trends in acidification. Copyright © 1996 Elsevier Science Ltd

Environmental Pollution, 1992), there are still few studies which consider sampling problems posed by episodic forms of pollution (e.g. Adams et al., 1992). Episodic fluctuations are particularly apparent in the case of acidification and probably have important consequences for aquatic biota (Weatherley & Ormerod, 1991; Davies et al., 1992; Ormerod & Jenkins, 1994). However, resource and logistical constraints often limit sampling regimes below the continuous frequency necessary to detect all episodic changes. Instead, spot sample data are often used to parameterise stream chemistry, for example as mean values, which are assumed to integrate chronic and episodic conditions (e.g. Weatherley & Ormerod, 1991). This approach has been used in assessing relationships between catchment features, stream chemistry and biology (Ormerod et al., 1989; Wade et al., 1989; Rutt et al., 1990) and is also important in modelling studies (Cosby et al., 1994), in the assessment of temporal trends (Nisbet et al., 1995) and critical loads (Kreiser et al., 1993). Any effects on such simple parameters by sampling frequency would clearly have ramifications in all these cases. Here we use weekly chemical data from 72 streams in the acid-sensitive areas of Wales to investigate the effects of four different simulated sampling frequencies, ranging from weekly to bimonthly, on the apparent chemistry of four key determinands. We also assess whether different simulated sampling frequencies would have affected our ability to detect relationships between chemistry, land use and stream biota. Although these data are now over 10 years old, they represent one of the few sets on which these assessments could be made.

Keywords: Acidification, sampling frequency, episodic pollution, Wales, water chemistry.

METHODS

INTRODUCTION Although there has been some discussion of the sampling frequency required to effectively eharacterise ambient water quality, or to monitor and detect effects on rivers by effluents (e.g. Sanders & Adrian, 1978; Casey et al., 19813; Hulme et al., 1985; Royal Commission on

Between October 1983 and September 1984, biological and weekly chemical data were collected by the Welsh Water Authority from 104 stream sites located in the major acid-sensitive regions of upland Wales (see Ormerod et al., 1989 for methods). At the same sites, macroinvertebrates were collected from riffles during April and July 1984 (Wade et al., 1989) and catchment land use was assessed (Ormerod et al., 1989). Thirty-two

148

P . A . Brewin et al.

sites had incomplete runs of chemical data over the sampling year and were therefore excluded from this analysis. The weekly pH, sulphate, total hardness and aluminium data from the remaining 72 sites were used to assess the effect of sampling frequency on apparent stream chemistry. These determinands were chosen because SO42- is currently the major acidifying anion in Welsh streams and inorganic AI, H + and Ca 2 ÷ are primary factors determining the biological responses to acidification (Ormerod et al., 1987; Brakke et al., 1994). For each stream the following parameters were calculated for all determinands: mean concentration, standard deviation (SD), coefficient of variation (CV, i.e. SD/mean, as a percentage), and the maximum and minimum values, for the following four sampling frequencies: 1. 2. 3. 4.

All 52 samples per site (i.e. weekly samples) 26 samples per site (i.e. fortnightly sampling) 13 samples per site (i.e. monthly sampling) 6 samples per site (i.e. bimonthly sampling).

For the fortnightly, monthly and bimonthly sampling intervals, data points were removed at regular calendar intervals from the sampling year to represent, as closely as possible, real sampling intervals beginning from a common starting date. Sites were separated according to their annual mean pH into six groups, each covering a pH range of 0.5 units, so that any systematic effects of different sampling frequency could be assessed across the range of acidity (Table 1). Within each group, means and SDs were calculated for each parameter at the different sampling frequencies. To detect seasonal patterns the data were disaggregated into summer (April September inclusive) and winter (October-March inclusive) periods and the analysis repeated. Significant differences between groups and between sampling frequencies were assessed by analysis of variance. To determine whether any effects were influenced by the basis of site division the analysis was repeated with sites ranked according to their annual mean pH and divided into six groups of 12 sites. Regression analysis and analysis of covariance (ANCOVA) were used to assess the effects of sampling frequency on the relationships between chemical parameters, biology and catchment land use. The total number of macroinvertebrate species recorded at each site and the percentage of each catchment covered by Table 1. The pH range of groups used in analysis of effects by various sampling frequencies and the number of sites in each group

Group

pH interval

Number of sites per group

1

< 5.0

6

2 3 4 5 6

5.0-5.5 5.5-6.0 6.0-6.5 6.5-7.0 >7.0

10 15 18 17 6

conifer forestry over 20 years old were used for this analysis, as these features have previously revealed important relationships with stream chemistry (e.g. Ormerod et al., 1989; Ormerod & Wade, 1990).

RESULTS For all four determinands varying the sampling frequency had no discernible or systematic effect on mean chemistry relative to the weekly means, or on the standard deviations of means around the overall mean (Fig. 1). Instead, the most pronounced differences occurred between some acidity groups, with standard deviations around the overall means greatest for pH, aluminium and sulphate at the most acid sites and greatest for total hardness at the more circumneutral sites (Fig. 1). Separating the data into winter and summer seasons left this effect qualitatively unchanged, except for a modest but systematic decrease in apparent mean pH across groups during the summer of between 0.08 and 0.16 units (mean 0.12), and an increase in apparent mean pH at all but the most circumneutral sites during the winter of between 0.04 and 0.21 units (mean 0.13), with decreasing sampling frequency (Fig. 2). These effects were unchanged when sites were divided evenly with 12 sites in each group. The standard deviations of weekly samples were around 6-12% of the mean for pH, 20-40% for hardness, 25 50% for sulphate and 60-105% for aluminium (Fig. 3). For pH, these coefficients of variation (CV) were affected more by differences between groups than by sampling frequency, with significantly higher CV at sites with a mean pH of between 5.5~.0 than the more acidic or circumneutral site groups (ANOVA; weekly CV, F¢sm)=4.14; p 6 than at the most acid sites between weekly and monthly sampling frequencies (F(5,71)=2.45, p=0.044, Table 2). This effect significantly increased the slope of the relationship between mean and minimum pH at monthly and bimonthly frequencies in comparison to weekly values (ANCOVA; F=4.8, p=0.031; Table 2). A similar effect was evident for maximum aluminium, which decreased more at acid sites, so that both the

Sampling frequency and stream, chemistry

149

CO

I I

~0 I

I

I

I

I I

I I

oo

~o

cO

LO

'

I

,'~

' c02~

oO

];

04

O" C CO

04

E

E CO

'/

oO

~0 ;-~

CO

I

I

I

I

O4

C4

~,

I CO

\

I I I

!

,

!

|

i

i

!

~5 ( e O O e 0 t_16w) s s e u p J e q l e l O | u e e l A I

o

l

i

d

• ~. 0

O

d

~5

~5

(~_16tu) w n ! u ! w n l e CO

I I I

UeelAl

I I

' I

O.

0

I

~O

&

I I

I

o3

oO

I

'

I

cO

E

@4

cq

I

oo

'II

~4

l.O

'

I

I I

.~

E o3

0O

I

I

04

' I '

I

I

I

I

t-

oO

~0

@4

E CO CO

I

04

I I

'

I

I

o3

'

o4

o

I I

CO

I

I

-H

I

iiI

I

I |

i

g

I~.

r~

~5

~5 H d Ueel/'i

l

v

O 04

! Od

! 0

(~.16w) e l e q d l n s

i

I

00

!

!

~.

,.4

£

..~

UeelAI CO

6

,

,

1

1248

,

~ L ~

.

,

.

.

2

1248

,

,

,

,

,

°

,

°

3

4

1248 1248 Sampling lntewal (weeks)

,

°

,

,

5

1248

,

,

,

,

6

1248

,

t-

-r( R .

4

4.5

5.5

6

6.5

7.5

,

,

,

1

1248

,

,

,

,

,

.

2

1248

,

,

°

,

,

,

,

,

.

,

,

,

,

3

4

1248 1248 Sampling Intewal (weeks)

,

Summer

,

,

,

5

1248

,

,

,

6

1248

Fig. 2. Mean ( ± S D ) pH for 72 streams, in six pH groups (Table 1), at a sampling frequency of 1, 2, 4 and 8 weeks during summer and winter periods.

Site group

4

4.5

5

~; 5.5

¢..

-1-

6.5

7

7.5

Winter

=__.

;E

b~

q.h

Sampling frequency and stream chemistry

I

:

I

,

(

~I" oJ ~-

I I

I

I

I

;

:

151

. , ..

/

¢.D t~

', ,,f" , I

t,O

I

I

I

"-

I

E

I

' j/

I

e-,,

c~

I

,,,¢

oJ

I

I

t~ v

I

I

'i

I

~o

I I

~'-

I

,.Q

~E

I

I

I

~r

I

e~

I

co

|

i

I

0

0

0

0

,

i

o

0

0 ~'

i

i

0 ~

0 0

sseup~eq lelOl Io A O ueeIAl

e~

"r"

I I

/

(

I

I

I

/

I

I

I ,

0..I

I

cM

I

I

( I

I

,

)

I

I

/ \

0

¢"

I

('

I

co

I

I

,

co

CO

I

I

)

I

I

I I

|

I

I

I

I

I

"~

e-,

I

I

,

i

i

0 OO

0 ~ID

0 ~"

0 O,J

o

I.

wn!u!wnle peAIOSS!p Io A O ueel/~

E

I ¢.D

Lc~

I

I I

I

I

CO

I

I

'

1 O~I

,

I

I

I

.CO

I

,O,I

/

'

I

I

I

I

I

I

e-,

.CO ,,,~. ,C'J

e..

I e-,

I I

I

I

I

I

I I

I

I

I

.OD

I I I

,O,J v

I

I I

'I

I I

I

I

z~

I .~

I

I

.00

E t~ o3

I I

" I

I I

-H

I

I

= 0

I

'E

" ( ' I I ,

i

i

I

CO

-e'.-

{M

I

I

e

i

|

cM

O0

I I

I !

I

0

w 0

.r-

I J 0

i O

i O

| 0

v O

i O

._,9.

|

0

0

Q. 0

Hd io AO u~el~l

e|eqdlns peAIoss!p jo A O uReIAI z~

6

3.5

4.5

5

1

1248

o

,

,

2

1248

,

o

,

•

•

o

,

,

3

4

1248 1248 Sampling Intewal (wee~)

,

.

.

5

1248

o

.

.

o

,

6

1248

,

0.8

~ 0.2

e-

•~ 0.4

E

t~

._= E 0.6

C

,

•

1

1248

,

i

•

•

,

,

2

1248

,

t •

,

,

,

.

.

.

.

.

3

4

1248 1248 Sampling Interval (weeks)

.

.

5

1248

,

,

,

6

1248

,

,

(0.55) (0.81) (0.24) (0.15) (0.04)

-0.31 (0.44) 0.64a(0.07) 0.13 (0.24) 2.05 (0.16) 1.73 (0.04)

a (±SD)

0.63 0.90 0.06 2.08 1.72

a ( + SD)

+ + + +

+ + + +

0.92a(0.07) 0.87 (0.04) 0.20 (0.03) 0.38 (0.09) 0.006 (0.001)

b (±SD)

0.69 (0.09) 0.80 (0.05) 0.21 (0.04) 0.39 (0.08) 0.006 (0.001)

b ( -4-SD)

68.7 89.0 31.8 24.9 30.4

r2

45.4 78.1 33.1 28.2 32.0

r2

~Values are significantly different from weekly values--according to analysis of covariance p < 0.05.

Mean pH vs minimum pH Log mean AI #gl-~ vs log maximum Ai #glMean pH vs log number of species (Fig. 5) Log mean AI #gl -I vs log number of species (Fig. 6) Log mean AI #gl- l vs % conifer cover (Fig. 7)

Regression parameters

Mean pH vs minimum pH Log mean AI #gl -~ vs log maximum A1 #gl -I Mean pH vs log number of species (Fig. 5) Log mean AI #gl -I vs log number of species (Fig. 6) Log mean AI #gl-i vs % conifer cover (Fig. 7)

Regression parameters

1

0.12 0.75 0.09 2.06 1.74

(0.46) (0.08) (0.25) (0.16) (0.04)

a ( + SD)

153.5 563.6 25.6 18.2 28.8

F

-0.48 (0.46) 0.46a(0.06) 0.05 (0.25) 2.08 (0.17) 1.76 (0.04)

a (±SD)

Sampling interval (weeks)

58.l 249.9 27.1 21.7 30.9

F

Sampling interval (weeks)

+ + + +

+ + + +

0.95"(0.08) 0.93a(0.03) 0.21 (0.04) 0.39 (0.09) 0.006 (0.001)

b (±SD)

0.80 (0.08) 0.85 (0.04) 0.21 (0.04) 0.38 (0.09) 0.006 (0.001)

b ( + SD)

2

68.1 92.3 32.5 24.9 30.2

r2

61.4 86.3 31.7 25.3 28.9

r2

152.6 845.0 26.5 18.2 28.5

F

111.3 440.1 25.5 18.6 26.9

F

Table 2. The effects of different sampling intervals on the relationships between chemical variables, the total number of invertebrate species and the percentage cover by coniferous plantations greater than 20 years old for 72 stream sites in Wales (all F values are significant at p < 0.001)

Fig. 4. Mean (+ SD) minimum pH and maximum aluminium for 72 streams, in six pH groups (Table 1), at a sampling frequency of 1, 2, 4 and 8 weeks.

Site group

e-

E

E .~ 5.5

•r

6.5

t O

Sampling frequency and stream chemistry slope and intercept of the relationship between mean and maximum aluminium were significantly different for bimonthly compared to weekly sampling frequencies (ANCOVA; slope, F = 4.4, p = 0.038; intercept, F = 14.6, p20 years old)

Fig. 7. Relationship between percentage of catchment covered by conifer forest greater than 20 years old and mean aluminium concentration at a sampling frequency for aluminium of 1, 2, 4 and 8 weeks (see Table 2).

156

P . A . Brewin et al.

Table 3. Mean 95% confidence intervals around mean pH and mean aluminium concentration for 72 stream sites in Wales at different sampling intervals Frequency

Weekly Fortnightly Monthly Bimonthly

Number of sites 72 72 72 72

Mean confidence intervals pH

Aiuminium mg litre -1

0.32 0.46 0.66 1.16

0.04 0.06 0.09 0.15

bedrock. One implication would be that streams of intermediate pH would need more intense sampling in programmes aiming for means which had high precision. By contrast, the coefficients of variation of aluminium at circumneutral sites and of total hardness across all sites declined with longer sampling intervals. Since the overall means were mostly unaffected, a low sampling frequency for these determinands would give apparently greater precision in the detection of mean conditions. These effects probably reflect the decreased likelihood of sampling extreme flow events at lower sampling frequencies, with proportionately more samples being drawn from baseflow conditions, and hence more uniform chemistry. Thus, the greater precision given by a low sampling frequency on the one hand will be offset by a decrease in representativeness on the other, especially when there is still a risk that samples drawn by chance from extreme conditions will have a disproportionately large effect due to the smaller sample size. Such effects might suggest the possibility of an unconventional strategy of purposely sampling to avoid variable flow conditions. This strategy would have value where the detection of extremes was not considered important and in instances where baseflow reflected important chemical trends--perhaps even those due to acid deposition. However, sampling frequency may influence the detection of spatial or temporal trends by affecting the sensitivity of some statistical tests aimed at assessing differences between sites or years. Analysis of variance, t-tests or the calculation of confidence intervals would all be affected by fewer degrees of freedom at lower sampling frequencies. For example, the apparent 95% confidence intervals for mean pH and aluminium across the 72 sites would have been increased by 0.84 pH units and 0.11 mg litre -~, respectively, as sampling frequency was reduced from weekly to bimonthly (Table 3). Such patterns form part of the trade-off governing logistical considerations of sampling frequency.

ACKNOWLEDGEMENTS

The authors thank Drs Neil Weatherly and Nigel Milner from the Environment Agency, Simon Bareham from Countryside Council for Wales and Havard Prosser from the Welsh Office for encouraging this piece of work in preparation for the 1995 acid waters survey of Wales.

REFERENCES

Adams, J. R. W., Dolby, J. C. & Williams, P. N. (1992). National Rivers Authority initiatives in the development of portable monitoring equipment. J. Inst. Water Environ. Manage., 6, 64-71. Brakke, D. F., Baker, J. P., Bohmer, J., Hartmann, A., Havas, M., Jenkins, A., Kelly, C., Ormerod, S. J., Paces, T., Putz, R., Rosseland, B. O., Schindler, D. W. & Segner, H. (1994). Physiological and ecological effects of acidification on aquatic biota. In Acidification of Freshwater Ecosystems: Implications for the Future, ed. C. E. W. Steinberg & R. F. Wright. John Wiley and Sons, Chichester, pp 275312. Casey, D., Nemetz, P. N. & Uyeno, D. H. (1983). Sampling frequency for water quality monitoring: measures of effectiveness. Water Resour. Res., 19, 1107-1110. Cosby, B. J., Bulger, A. J. & Wright, R. F. (1994). Predicting recovery of freshwater ecosystems: trout in Norwegian lakes. In Acidification of Freshwater Ecosystems: Implications for the Future, ed. C. E. W. Steinberg & R. F. Wright. John Wiley and Sons, Chichester, pp 355-373. Davies, T. D., Tranter, M., Wigington, P. J. & Eshleman, K. N. (1992). 'Acidic episodes' in surface waters in Europe. J. Hydrol., 132, 25-69. Edwards, R. W., Gee, A. S. & Stoner, J. H. (1990). Acid Waters in Wales. Kluwer, Dordrecht, 337 pp. Hemond, H. F. (1990). Acid neutralization capacity, alkalinity and acid-base status of natural waters containing organic acids. Env. Sci. Technol., 24, 1486-1489. Hulme, P., Johnson, P. C., Walker, D. G. & Ellis, J. C. (1985). Industrial effluent charging--how many samples? Wat. Pollut. Control., 84, 486-501. Kreiser, A. M., Patrick, S. T. & Battarbee, R. W. (1993). Critical loads for UK freshwaters: introduction, sampling strategy and use of maps. In Critical Loads--Concepts and Applications, ed. M. Hornung and R. A. Skefflngton. HMSO, London, pp 94-98. Muniz, I. (1991). Freshwater acidification: its effects on species and communities of freshwater microbes, plants and animals. Proc. Roy. Soc. Edinburgh, 97B, 227-254. Neal, C., Reynolds, B., Stevens, P. A. & Hornung, M. (1992). Factors influencing streamwater chemistry in the Welsh uplands: examples from Plynlimon. Report of the Welsh Soils Discussion Group Nos. 27-29, pp 7-24. Nisbet, T. R., Fowler, D. & Smith, R. I. (1995). An investigation of the impact of afforestation on stream water chemistry in the Loch Dee Catchment, SW Scotland. Environ. Pollut., 90, 111-120. Ormerod, S. J., Boole, P., McCahon, C. P., Weatherley, N. S., Pascoe, D. & Edwards, R. W. (1987). Short-term acidification of a Welsh stream: comparing the biological effects of hydrogen ions and aluminium. Freshwat. Biol., 17, 341-356. Ormerod, S. J., Donald, A. P. & Brown, S. J. (1989). The influence of plantation forestry on the pH and aluminium concentration of upland Welsh streams: a re-examination. Environ. Pollut., 62, 47-62. Ormerod, S. J. & Jenkins, A. (1994). The biological effects of acid episodes. In: Acidification of Freshwater Ecosystems." Implications for the Future, ed. C. E. W. Steinberg & R. F. Wright, John Wiley and Sons, Chichester, pp 259-272. Reynolds, B. R., Neal, C., Hornung, M. & Stevens, P. A. (1986). Baseflow buffering of streamwater acidity in five mid-Wales catchments. J. Hydrol., 87, 167-185. Royal Commission on Environmental Pollution (1992). Sixteenth Report: Freshwater Quality. HMSO, London, 291 pp. Rundle, S. D., Weatherley, N. S. & Ormerod, S. J. (1995). The effects of catchment liming on the chemistry and biology of upland Welsh streams: testing predictions from empirical models. Freshwat. Biol., 34, 165-176.

Sampling frequency and stream chemistry Rutt, G. P., Weatherley, N. S. & Ormerod, S. J. (1990). Relationships between the physicochemistry and macroinvertebrates of British upland streams: the development of modelling and indicator systems for predicting fauna and detecting acidity. Freshwat. Biol., 24, 463-480. Sanders, T. G. & Adrian. D. D. (1978). Sampling frequency for river quality monitoring. Water Resour. Res., 14. 569-576.

157

Wade, K. R., Ormerod. S. J. & Gee, A. S. (1989). Ordination and classification of macroinvertebrate assemblages to predict stream acidity in upland Wales. Hydrobiol., 171, 59-78. Weatherley, N. S. & Ormerod, S. J. (1991). The importance of acid episodes in determining faunal distributions in Welsh streams. Freshwat. Biol., 25, 71-84.