implementation of a sustainable groundwater quality

AN INVESTIGATION OF DATA COLLECTION METHODS APPLICABLE IN GROUNDWATER RESEARCH IN RURAL REGIONS OF DEVELOPING NATIONS

A Dissertation

Submitted to the Graduate School of the University of Notre Dame in Partial Fulfillment of the Requirement for the Degree of

Doctor of Philosophy

by

Pamela E. Crane

____________________________ Stephen E. Silliman, Director

Graduate Program in Civil Engineering and Geological Sciences Notre Dame, Indiana December 2007

AN INVESTIGATION OF DATA COLLECTION METHODS APPLICABLE IN GROUNDWATER RESEARCH IN RURAL REGIONS OF DEVELOPING NATIONS

Abstract

by

Pamela E. Crane

Monte Carlo studies and a case study in Bénin investigated data collection methods applicable in rural regions of developing nations. These studies were cased within the framework of three progressions: the progression of analytical methods (POAM), the progression of sampling strategies (POSS), and the progression of expertise (POE). These studies demonstrated the strength of using the full range of each of the progressions. The Monte Carlo studies were based on use of five scenarios of contamination in water derived from a groundwater well in combination with five instruments and six sampling strategies. These studies were applied to assessment of three parameters: mean concentration, max concentration, and total mass load. A wide range of observations regarding use of instrument and expertise were derived from the results. Significant among these is the observation that analytical methods at the low end of the POAM combined with sampling by personnel at the low end of the POE can provide high-quality

Pamela E. Crane estimates of measures (e.g., nitrate or uranium) and parameters (e.g., mean concentration) of groundwater quality. The goal of the Bénin case study was to compliment the Monte Carlo studies by introducing the realities of field work. Following a regional exploratory study, the case study focused on characterization of nitrate and uranium in groundwater in south-central Bénin. The nitrate portion of the study included collaboration with local populations for the sampling effort. These local populations were sampled at high frequency with a uniform sampling quality over a period in excess of a year, thus demonstrating the potential to use the low end of the POE in combination with the low ends of the POAM and POSS. Overall, these studies demonstrated the value of data gained using the less complex portions of the progressions. Further benefit is obtained by using different portions of the progressions at different stages of, or for different aspects of, research projects. It is hoped that these results will be applied to address the growing disconnect between the current direction of groundwater research and the great need for such research and monitoring in rural regions of developing nations.

CONTENTS

Tables .................................................................................................................... vi Figures.................................................................................................................. vii Acknowledgments ................................................................................................ ix Chapter 1: Introduction ........................................................................................1 1.1. Introduction of Motivations........................................................................2 1.2. Research Motivations .................................................................................4 1.3. Working Concept........................................................................................5 1.4. Overview of Approaches ............................................................................6 Chapter 2: Background and Literature Review .................................................9 2.1. Progressions................................................................................................9 2.1.1. Brief Comment on Terminology....................................................11 2.1.2. Progression of Analytical Methods................................................11 2.1.3. Progression of Sampling Strategies ...............................................15 2.1.3.1. Sampling Strategies as explored in Surface Water Literature...................................................................16 2.1.3.2. Relation of the Surface Water Literature to Present Research..................................................................18 2.1.4. Progression of Expertise ................................................................19 2.1.5. Sociological Considerations and Development Literature ............21 2.2. Sampling Strategies and Groundwater Research: A Perspective ............22 2.3. Evaluation of Research Strategies ............................................................25 2.3.1. Monte Carlo Studies ......................................................................26 2.3.2. Use of Case Study..........................................................................27

ii

Chapter 3: Case Study: Exploratory Research in Bénin .................................31 3.1. Field Site Specifics ..................................................................................31 3.2. Regional Groundwater Study ...................................................................36 3.2.1. Methods .........................................................................................38 3.2.2. Results............................................................................................45 3.2.3. Analysis..........................................................................................51 3.2.4. Measures of Further Interest ..........................................................55 Chapter 4: Case Study: Measure Specific Methods .........................................57 4.1. Measure: Nitrate .......................................................................................57 4.1.1. Available Analytical Methods .......................................................60 4.1.1.1. Common Practice in Developing Countries ........................62 4.1.2. Methods Utilized in This Research................................................63 4.1.2.1. Low Frequency Sampling: Methods....................................64 4.1.2.2. Low Frequency Sampling: Field Application .....................65 4.1.2.3. High Frequency Sampling: Methods...................................67 4.1.2.4. High Frequency Sampling: Development ...........................69 4.1.2.5. High Frequency Sampling: Initial Implementation .............71 4.1.2.6. High Frequency Sampling: Extension .................................71 4.2. Measure: Uranium ....................................................................................73 4.2.1. Available Analytical Methods .......................................................75 4.2.2. Methods..........................................................................................76 4.2.2.1. Field Collection Methods ....................................................77 4.2.2.2. Laboratory Methods ............................................................79 Chapter 5: Monte Carlo Studies: Methods .......................................................83 5.1. Generating Population Data Sets .............................................................84 5.1.1. Five Model Scenarios ....................................................................85 5.1.2. Assessment of Study Parameters ...................................................88 5.2. Simulating “Sampled Data”......................................................................89 5.2.1. Definitions of Sampling Periods....................................................89 5.2.2. Definition of Sampling Length .....................................................91 5.2.3. Sub-sampling .................................................................................91 5.2.4. Adding Sampling Errors ................................................................92 5.2.5. Notation..........................................................................................95 5.3. Evaluation of Sampled Data ....................................................................96 5.3.1. Calculation of Parameters of Interest.............................................96 5.3.2. Calculating Parameter Error .........................................................98

iii

Chapter 6: Results and Analysis.......................................................................100 6.1. Results.....................................................................................................100 6.1.1. Case Study: Nitrate ......................................................................100 6.1.2. Case Study: Uranium ...................................................................107 6.1.3. Monte Carlo Studies ....................................................................113 6.2. Measure Specific Analysis From the Bénin Case Study ........................124 6.2.1. Analysis of Nitrate Results .........................................................124 6.2.1.1. Use of the Nitrate Progressions .........................................125 6.2.1.2. Interpretation of Potential Nitrate Sources for Adourékoman Based on Nitrate Data ................................126 6.2.1.3. Support of the Source Identification with Sociological Methods ........................................................127 6.2.1.4. Results From Application of Progressions Leading to New Characterization Goals ......................................................130 6.2.1.5. Progressions Leading to Conclusions................................134 6.2.2. Analysis of Uranium Results .......................................................135 6.2.2.1. Use of the Uranium Progressions ......................................135 6.2.2.2. Interpretation of the Uranium Data in Bénin.....................136 6.2.2.3. Implications for Future Uranium Sampling Efforts ..........137 6.3. Analysis of Monte Carlo Results............................................................138 6.3.1. General Observations...................................................................138 6.3.2. Observations on Use of the Low End of the POE and POAM ....140 6.3.3. Observations on Instruments / Operators with Larger e and b ...146 6.3.4. Trade-offs Among Sampling Period, Instrument and MSE.........147 6.3.5. Observations on Time-Variable Sampling Periods .....................149 6.3.6. Parameter Sensitivity to Instrument.............................................151 6.4. Common Observations to the Case Study and Monte Carlo Studies .....153 Chapter 7: Conclusions .....................................................................................159 7.1. Return to Progressions............................................................................159 7.1.1. Monte Carlo Results ....................................................................161 7.1.2. Case Study Results.......................................................................161 7.2. Assessment of the Working Concept......................................................163 7.2.1. Support of the Working Concept .................................................163 7.2.2. Suggested Modification to Original Working Concept ...............166 7.3. Future Research ......................................................................................166

iv

Appendix 1: Complete Results of the Monte Carlo Studies...........................173 Appendix 2: Complete Data Collected by Local Sampling Groups in Adourékoman.............................................................235 Appendix 3: Complete Results for the Uranium Study..................................244 References...........................................................................................................255

v

TABLES

Table 3.1 Table 3.2

Table 5.1 Table 5.2 Table 5.3 Table 5.4 Table 6.1 Table 6.2 Table 6.3

Table 6.4 Table 6.5

Elements for which analytical results were used in the statistical analyses. ...........................................................................................42 Results of the Principal Component Analysis – First Three Components...................................................................50 Calculated “known” values for assessment parameters for the five data sets....................................................................................89 e and b in the five instrument cases used in the Monte Carlo simulation...............................................................................94 Symbols representing sampling periods over a period of N years...95 Coefficients of parameters that change with sampling period ........97 Concentrations of uranium in the dissolved and colloidal phases of samples collected during the summer of 2006. .........................109 Symbols representing sampling periods over a period of N years.113 Groundwater nitrate data as collected from wells Ayewa-Okouta, 2 filled-in hand dug (HD) wells, and 3 open hand dug wells in Adourékoman.................................................................................131 Evaluation of numerical options to represent discrete ranges of test strip results. .........................................................................143 Results from Monte Carlo Study using uniform test strip intervals ...........................................................................145

Table A1

Complete results of the Monte Carlo Studies. Table arranged by scenario, then sampling period and sampling lengths, and grouped by parameter...................................................................................174

Table A2

Complete water quality results as collected by the three local sampling teams in Adourékoman...................................................236

Table A3

Complete ICP-MS and ICP-OES results for the uranium study. ...245

vi

FIGURES

Figure 1.1

Flowchart of dissertation research ....................................................7

Figure 2.1

Flow chart outlining an expert’s choices in creating a research plan ....................................................................................10 Generic POAM ................................................................................12 POAM for nitrate .............................................................................13 POSS ................................................................................................19 POE ..................................................................................................20

Figure 2.2 Figure 2.3 Figure 2.4 Figure 2.5 Figure 3.1 Figure 3.2 Figure 3.3 Figure 3.4

Figure 3.5

Figure 3.6 Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4

Figure 4.5 Figure 4.6

Figure 5.1

Political map of Bénin .....................................................................32 Monthly distribution of mean annual precipitation of Bénin by latitude ........................................................................................33 The numbers represent approximate location of wells sampled with the number representing the year of sampling .........................37 Indicator variable analysis illustrating consistent clustering of high pH (left), high conductivity (middle), and low phosphorus (right) in south-central portion of study area (indicated in figure). .46 Results from cluster analysis showing locations of the 70 wells with the smallest separation distances (x) versus those wells that fell outside of this group (solid circles)..................................................48 Political departments of Bénin ........................................................56 Select historical nitrate data in the Colline Department of Bénin ...59 POAM for nitrate .............................................................................60 Nitrogen and oxygen isotopes of nitrate in groundwater samples from the Colline department taken during the summer of 2003. .....66 Nitrogen and oxygen isotopes of nitrate in groundwater samples from the Colline Department taken during the summer of 2003 (*), February of 2007 (O), and May of 2007 (∆)....................................67 Nitrate standards as analyzed in initial implementation of high frequency sampling by groups in Adourékoman. ...........................72 Examples of black particles that developed during the dissolution process in ~80% of samples using the initial samples preparation method.............................................................81 Scenario 1; Q has a constant mean within the time series, and C is correlated with Q...............................................................86

vii

Figure 5.2 Figure 5.3 Figure 5.4

Figure 5.5 ]

Figure 6.1

Figure 6.2 Figure 6.3 Figure 6.4 Figure 6.5 Figure 6.6 Figure 6.7

Figure 6.8 Figure 6.9

Figure 6.10 Figure 6.11 Figure 6.12 Figure 6.13

Figure 6.14

Figure 6.15

Scenario 2; Q varies seasonally within the time series, and C is correlated with Q................................................................86 Scenario 3; C and Q have constant means within the time series, C is independent of Q.......................................................................87 Scenario 4; C varies seasonally within the time series, C is independent of Q, and Q has constant mean within the time series...................................................................................87 Scenario 5; C linearly increases across the time series, C is independent of Q and Q has a constant mean within the time series...................................................................................88 (a) Hand drawn map of Adourékoman as represented by local population. (b) Spot image of Adourékoman as provided by DH..............................................................................101 N and O isotopes of nitrate from the three wells in Adourékoman. Dates of samples as indicated.. ......................................................102 Nitrate data from well Ayewa-Okouta in Adourékoman as 104 collected by the local sampling group............................................104 Nitrate data from well Agbo in Adourékoman as collected by the local sampling group. ...............................................................104 Nitrate data from well Ayewa in Adourékoman as collected by the local sampling group. ...............................................................105 Nitrate data from well Ayewa in Adourékoman as collected by the local monitoring group using a colorimeter. ............................106 MSE and precision for TML of scenario 3. (A) Original results, and (B) 10 year length substituted with results using 100 year data set............................................................................................121 Well Ayewa-Okouta and a filled-in hand dug well. ......................130 N and O isotopes of nitrate from hand dug wells and streams in Adourékoman superimposed on Figure 6.1. Dates of sample are as indicated.....................................................................................133 Evolution of MSE and precision for parameter mean over an increase in length of sampling period ............................................141 Evolution of MSE and precision for parameter TML over an increase in length of sampling period ............................................142 Precision and bias2 of all instruments used weekly or semi-annual for one year. The total height of each bar is equal to the MSE......147 Precision and bias2 of one year lengths for weekly, monthly, and semi-annual sampling periods grouped by instrument. The total height of each bar is equal to the MSE. .........................................148 MSE and precision of time-variable sampling as compared to time-uniform weekly sampling. A) season dependent sampling period, B) parameter dependent sampling period. ........................150 Evolution of MSE and precision for parameter max over an increase in length of sampling period. ..........................................153

viii

ACKNOWLEDGMENTS

I would like to acknowledge the following individuals and organizations that have played key roles throughout the pursuit and completion of this research: The West Foundation, the Asnate Foundation, and the Environmental Molecular Science Institute at Notre Dame for funding this research. My committee members, Drs. Clive Neal, Peter Burns, and Susan Sakimoto for their time and effort spent providing research recommendations and help in the defense of this dissertation. Dr. Moussa Bourkari, my research advisor in Bénin, who has been key in enabling the research in Bénin to occur. His gentle spirit is a continual blessing and knowledge of Bénin is essential to our work. Dr. Stephen E. Silliman, my advisor, who gave me the freedom to pursue this research and the necessary support and guidance to see it to its conclusion. Exceeding far beyond what is contained in this dissertation is a body of work that is evidence of thinking outside the box and pushing the limits—it has been an honor to be participate in this work. Caitlin Erkins Rackish, whose work has enabled the further expansion of this research to include anthropological research, a dream of mine that I believe will be valuable to the groundwater and development communities. Sarah A. Runger, Lauren K. Shuttleworth, and Rachel L. Cota, “my girls”, who participated in many aspects of this research enabling it to be completed in a reasonable time frame. Beyond research, these ladies were willing to live in a village in the African bush and let the experience transform their lives; it has been a blessing to be a part of their journeys. My family, whose support has been unwavering; they are best family one could hope for. When I have been tired, they have helped me hold on to and move towards my dreams.

ix

CHAPTER 1:

INTRODUCTION

The technology for sampling of groundwater quality has progressed rapidly over the past few decades with advances ranging from the advent of direct-push methods for gaining access to samples (e.g., www.geoprobe.com), to the development of more sophisticated analytical equipment (e.g., Inductively Coupled Plasma Mass Spectrometry; Weight and Sonderegger 2001), to the development of advanced probabilistic methods for both sampling design and data analysis (e.g., using a combination of tracer elements, multivariate statistics, and geographical information systems to decipher groundwater flow systems; Farnham et al. 2000). While these advances have provided unique capabilities for field characterization of groundwater in situations where access to expert field technicians and rapid return of samples to analytical laboratories is compatible with local personnel and financial resources, a wide array of field characterization efforts are limited in terms of both personnel and access to analytical laboratories. In particular, characterization of many groundwater quality scenarios in developing countries precludes sole reliance on these more advanced methodologies. A classic example of such a field scenario is presented in the next section in the form of review of characterization efforts of the groundwater arsenic contamination problem in Bangladesh. This example is used to motivate the need for a new outlook on

1

more integrated characterization strategies for use in developing countries. From this motivation, the primary hypothesis for this dissertation is introduced and the combined numerical and case study approach (case study based on Bénin, West Africa) is outlined. The remainder of this dissertation provides an assessment of the suggested integration of analytical capability and expertise in field characterization.

1.1 Introduction of Motivations Bangladesh, like many developing nations, is facing serious problems concerning their public drinking water supply. In an effort to reduce diseases associated with using surface water, wells were drilled throughout the country, and it is estimated that there are between 8 and 12 million wells in the country (WHO 2001). While this has greatly decreased public health issues commonly connected to contaminated water supplies such as diarrheal diseases, a significant health problem was detected in 1993 when arsenic was discovered in well water in many regions at concentrations above international standards (UNICEF 2006). In these regions of Bangladesh, the drilling of the wells led to replacing waterborne diseases common to surface water with long-term potential for arsenic poisoning. Recent estimates suggest that approximately 25 million people in Bangladesh are at risk (Chowdhury et al. 1999). A 2006 literature review of research conducted on the Bangladeshi groundwater arsenic problem indicates that, although significant amounts of funding have supported research related to the problem, critical gaps in knowledge remain (Hossain 2006). For instance, inconsistencies in measured arsenic concentrations have been identified between field and lab results for groundwater samples, and almost no time series data

2

exist to show temporal trends in the groundwater arsenic; the longest time-series data that Hossain found were only two years in duration. Although some have interpreted the available data to suggest that the arsenic problem is worsening, Hossain argues that the suggestion of trends in arsenic concentrations might predominantly be a function of the inconsistent monitoring that has occurred. Determination of temporal trends in arsenic concentrations would require development of time-series data sets for individual wells. However, if the above stated estimates of the number of wells affected by the arsenic are remotely accurate, the shear magnitude of considering the establishment of a regular, high-frequency monitoring effort based on professional technicians represents an enormous undertaking. If, on the other hand, a monitoring program could be developed based on local populations using simple analytical methods (e.g., arsenic test strips, Hach Company product number 28000-00), temporal and spatial data sets could potentially be compiled with relatively minimal financial resources. Although these data sets would likely be subject to greater measurement error than data taken by lab technicians, they have the potential to greatly reduce current gaps in knowledge. The significance of considering the Bangladesh arsenic example is its pertinence to discussion of the need to seek alternative sampling strategies to enable collection of scientifically useful, but difficult to collect, data sets. It is this motivation that underlies the research for this dissertation. In considering alternative strategies for sampling, one significant opportunity arises through consideration of reliance on both varying degrees of complexity in the technology and varying levels of expertise in the person performing the sampling. Specifically, moving beyond exclusive use of complex analytical methods

3

leads to the exploration of the full range or progression of available analytical methods, sampling strategies, and personnel involved in data collection. However, in assessing these progressions, questions related to maintaining the integrity of groundwater data necessitate understanding the impact of these progressions on the quality of the resulting data. The aim of this dissertation is to serve as a foundational work in the exploration of these progressions.

1.2 Research Motivation As described in the Bangladesh arsenic example, there tends to be, in groundwater research, a focus on the exclusive use of sophisticated methods that we herein term complex methods; this focus is further detailed in the literature review found in Chapter 2. Unfortunately, this focus is occurring in opposition to the significant need for groundwater research in developing regions. Although the relationship between drinking water quality and mortality rates has been established (e.g., Sepúlveda, et al. 2006, and Schoenen 2002) and World Health Organization guidelines for drinking water quality have been established (WHO 2006), a significant need for research and monitoring of water quality in developing nations continues to exist. A recent article published by the United Nations Development Program demonstrates this when they stated that, “Across much of the developing world, unclean water is an immeasurably greater threat to human security than violent conflict” (UNDP 2006). Based on the disconnect between the need for, and difficulty of, performing water quality research in rural regions of developing nations, the intent of this research is to provide a foundation from which to expand groundwater research past the limitations that

4

are imposed by the current emphasis on complex data collection methods and reliance on expert technicians. This intent is addressed through the exploration of the use of the progressions of analytical methods, sampling strategies and expertise in groundwater research as applied to rural regions of developing nations in order to develop data sets which are useful for many common water quality assessment projects. In the above example of arsenic in Bangladesh, for example, use of a full progression of methodologies might enable collection of the critical, high-frequency temporal data necessary to fully characterize temporal variation in arsenic concentrations.

1.3 Working Concept The objective of this research is to assess whether integration of a spectrum of field techniques (including both analytical methods and expertise of the field personnel) as applied to the characterization of groundwater contamination scenarios may provide unique capability in terms of developing water quality data sets. The spectrum used ranges from those considered state-of-the-art to those that may be considered, by many scientists, to be of secondary utility due to limits in precision but which are consistent with use in low technology applications. It is critical to the following discussion to note that the term “technique” as used here includes the combination of analytical method, sampling strategy, and sampling personnel used to collect a data set. The working concept of this research is stated as: A progression of analytical methods, as guided by prior knowledge and data analysis combined with expertise provided by stakeholders, will provide for the characterization of measures of groundwater quality in

5

rural regions of developing countries. The combination of analytical methods, stakeholder expertise, sampling strategy and data analysis most appropriate for a given characterization scenario will depend, among other things, on the intended use of the characterization and the critical measures identified.

The intent of this dissertation is to explore the working concept through multiple avenues of investigation; specifically, the concept is explored through both numerical experiments and a field case study. It is argued that the rigors of numerical studies combined with the reality of field work will enable an assessment of the working concept valuable to both the scientific community and the development community as both continue their respective research and monitoring of groundwater in rural regions of developing countries.

1.4 Overview of Approaches The first necessary step in this research is defining and exploring the concept of using the full spectrum of analytical methods, sampling strategies, and personnel or expertise available. This concept, herein termed “progressions” of analytical device, sampling strategy and expertise, is explored in Chapter 2 through a review of the literature on groundwater sampling and reliance on technical expertise. A framework is established within that chapter that is used to support the remainder of the research described in this dissertation. Summarized by the flow chart provided in Figure 1.1, the

6

framework is based on two different, though related avenues: a field case study and Monte Carlo studies.

Working Concept

Progression of Analytical Methods

Progression of Sampling Strategies

Monte Carlo Studies

Progression of Expertise

Case Study: Bénin Groundwater Sampling Frequency

Exploratory

Precision / Bias

Nitrate

Mean Conc. / Max Conc. / Total Mass Load

Uranium

Data Utility?

Conclusions

Figure 1.1: Flowchart of dissertation research

7

Discussion of the working concept is considered, as described in Chapters 3 and 4, via a case study in Bénin, West Africa. As discussed in Chapter 3, country-wide exploratory research on groundwater quality in Bénin provided a basic data set from which to initiate water quality research on specific measures of interest (nitrate and uranium) in south-central Bénin (Silliman et al. 2007). The work focused in south-central Bénin, as described in Chapter 4, involved personnel from (i) Notre Dame, (ii) the national university in Bénin, (iii) the Bénin government agency charged with groundwater development, (iv) a Bénin NGO focused on education, and (v) the local population of a rural village. The Bénin case study, with its integration across a progression of field techniques, focuses on assessment of the source(s) of elevated nitrate and uranium concentrations within this region and provides the opportunity to explore the efficacy of integration of multiple combinations of analytical method, data analysis, expertise, and sampling strategy. The working concept is also addressed numerically, as described in Chapter 5, through Monte Carlo studies on sub-sampling from hypothetical groundwater quality data sets. These studies provide the ability to assess a wide array of sampling techniques, including variable sampling frequency, length of sampling period, and instrument / operator characteristics that might be encountered in a variety of field situations. As discussed in Chapter 6, the combination of observations collected via both the field case study and the Monte Carlo analysis helps to more fully inform our analysis of the importance of the working concept to field characterization of groundwater in developing countries. The final conclusions from this research effort, in combination with suggestions for continuing studies, are presented in Chapter 7.

8

CHAPTER 2:

BACKGROUND AND LITERATURE REVIEW

2.1 Progressions Applied and/or fundamental research programs related to environmental hydrology commonly follow a structure similar to that found in Figure 2.1. Specifically, existing knowledge or results from prior research or project efforts leads to identification of a new hypothesis or project objective. From such identification, the expert defines the scope of a new project designed to respond to the project objective. Assuming that the new project involves collection of data, the expert will be required to create a sampling design that includes choices in three areas: (i) analytical method used, (ii) expertise required of the person(s) collecting the data, and (iii) distribution of samples in space and time (herein referred to as sampling strategy). Following these choices, the expert is then able to actively pursue data collection and analysis. Within this dissertation, primary interest is placed on the third step in the process as displayed in the figure; specifically, choice of analytical method, sampling strategy, and expertise of the field personnel. Although a range of choices, herein termed “progressions”, is available in each of the three areas, final choice of analytical method, sampling strategy, and expertise of sampling personnel is often dictated by: specificity of the data required, availability of analytical methods, precision / bias of the measurement required to address the project

9

Expert & Prior Analysis

Project Objective

Progression of Analytical Methods

Progression of Sampling Strategies

Progression of Expertise

Expert’s Choices

Data Collection

Data Analysis Leading to Conclusion Regarding Objective or Hypothesis Figure 2.1: Flow chart outlining an expert’s choices in creating a research plan

objective, budget available for sampling, and the intended data analysis methods to be employed. As detailed below, there has been a tendency in the groundwater literature towards greater reliance on complex analytical methods as well as a broader reliance on technical staff for collection of samples. Unfortunately, such reliance may be ineffective in rural regions of developing countries due to barriers related to access to analytical facilities, limited availability of technical staff, and logistical issues related to access to sampling locations. Due to these barriers combined with increasing interest and need for detailed water-quality data sets in developing regions, our interest is focused on a range of analytical methods, sampling strategies, and field expertise in order to collect viable groundwater quality data sets.

10

Study of the efficacy of using progressions of analytical methods, sampling strategies, and expertise is the focus of this dissertation. Hence, the current chapter provides an overview of our definition of these “progressions” and then presents review of the literature related to prior discussions of sampling and sampling strategies.

2.1.1 Brief Comment on Terminology In order to facilitate the following discussion, two terms are herein defined to differentiate between the value of an individual entry in a data base and values derived through analysis of a series of individual data entries. For example, the difference between an individual measure of the pH of a water sample and the mean pH as determined by appropriate analysis of a group of individual pH measures. The individual measure is herein termed “measure”, whereas a numerical result derived from analysis of a group of “measures” is herein termed a “parameter”. As outlined below, the primary measures of interest to our work are measurements of the concentration of chemical species in a groundwater sample. The primary parameters of interest will include the mean and maximum of these individual concentration measurements as well as total mass load (as defined below).

2.1.2 Progression of Analytical Methods Central to this dissertation research is the selection of methods to be used in water quality analysis. In considering the available methods, it is observed that, for many chemical species, a range of analytical methods exist for measurement of the concentration of those species. We herein refer to this range of methods as the

11

Progression of Analytical Methods (POAM); a POAM is represented as a linear axis of increasing complexity of the method (as shown in Figure 2.2) and methods are classified within this progression based on a range of factors including precision, bias, and degree of expertise required to use the method. It should be noted that, for all progressions presented in this dissertation, no units are displayed as the progressions do not use a specific scale; rather, spacing is used to aid the reader in understanding the concepts behind the progressions.

Increasing complexity Methods with highest level of complexity

Methods with lowest level of complexity

Figure 2.2: Generic POAM

Rather than trying to further define the POAM in general terms that apply to all chemical species, we herein introduce the basic concepts of the POAM through the specific case of measurement of the concentration of nitrate in a water sample. Eaton et al. (2005) provide an overview of standard lab methods that are suggested for nitrate assessment in high technology settings. Included among these are isotopic ratio methods, automated hydrazine reduction, automated cadmium reduction, and manual cadmium reduction. These methods involve use of complex laboratory methods (e.g., cadmium reduction methods require creation and use of a reduction column, and the method includes multiple steps that all require high levels of precision), addition of hazardous or carcinogenic chemicals to the sample as part of the method (e.g., addition of hydrochloric

12

acid in the ultraviolet spectrophotometric screening method or hydrazine sulfate in the hydrazine reduction method; Eaton et al. 2005), and all of these methods use calibration standards to calibrate the instrument each time it is used. The complexity of these methods, as expressed through the number of steps in the method, necessary precision, the use of hazardous chemicals and use of calibration standards with each analysis necessitates the use of an expert to perform these methods. These standard methods are shown on the left portion of Figure 2.3, a POAM for nitrate methods.

Figure 2.3: POAM for nitrate

Towards the middle of this scale are methods such as cadmium reduction spectrophotometry and colorimetry. These methods are often based on addition of prepackaged reagent materials, require lower levels of sample preparation and preservation, therefore requiring lower levels of expertise in the technician performing the analytical methods. In trade-off for these conveniences, these methods tend to have limited ranges of application, higher detection limits and lower precision. For example, the colorimeter used in our field work has a range of 1 – 33 mg/l NO3-N with a reported

13

precision of +/- 3mg/L (as opposed to 0.005 mg/L NO3-N for the cadmium reduction method using a reduction column or a 2.5% reported precision for the nitrate electrode method; Eaton et al. 2005). At the right end of this POAM are test methods which are extremely simple to use, require no reagents, and can be performed by a trained lay person. For the least complex of these methods, test strips, a significant trade-off is that the resulting measure of concentration is reported in terms of a discrete interval. For example, the test strips for nitrate that we use in the Bénin case study provide measures in the following discrete steps: 0, 0.5, 2, 5, 10, 20, 50 mg/L NO3-N, with the company suggested precision of plus or minus one interval (Industrial Test Systems Technician, personal communication, 2005). The nitrate example demonstrates both the basic concept of a POAM and provides the foundation for discussion of the potential importance of considering the entire range of methods represented by the POAM in application to water quality analysis in a developing country. Specifically, while the more complex side on a POAM might be expected to provide low detection limits, increased precision, and unbiased measure of the concentration of nitrate, it commonly requires application of relatively complex analytical techniques by experts in analytical chemistry (and/or water analysis). It also often requires sample preservation in the field using chemicals that are frequently complicated or dangerous for non-experts to handle in uncontrolled settings. As a result, these methods will likely not be the methods of choice for use in rural regions of developing countries when asking local experts and/or local populations to perform the analysis.

14

In contrast, the methods represented at the other extreme on the POAM are commonly subject to limited range of application, high detection limits, low precision, and biased results, but are substantially less complex to perform. While not often considered when performing characterization of chemical species such as nitrate for the purposes of research or health assessment, these methods may find appropriate use in settings where water professionals untrained in analytical methods and/or local populations are being asked to measure chemical species found in groundwater (for examples and an overview of volunteer monitoring systems, see section 2.1.3 of Crane 2006). As argued below, consideration of the entire range of the POAM may provide opportunities for pursuing research in groundwater quality, particularly in rural regions of developing countries, under conditions where research based solely on the more complex end of the POAM might not be possible due to field conditions or limitations on personnel and resources.

2.1.3 Progression of Sampling Strategies Review of the literature covering temporal sampling strategies illustrates that, based on the eventual use of a database, a range of strategies exist for populating that database. In terms of sampling frequency over time, the least complex strategy is uniform time intervals between samples with the interval chosen independent of the value(s) of previously recorded measures. In contrast, the most complex is a strategy involving variable time intervals between samples with the length of each interval dependent on previously recorded values of either the primary or a secondary measure. This second

15

type of sampling is often used, for example, in surface water research when attempting to focus sampling on high flow events during the investigation of total mass loads of nutrients (e.g., Kronvang and Bruhn 1996). In the following sections, brief review is provided of the surface water literature in time variable sampling. This is followed by discussion of implications of this literature for definition of a progression of sampling strategies as applied to characterization of groundwater systems.

2.1.3.1 Sampling Strategies as Explored in Surface Water Literature The literature most closely related to our interest in groundwater quality involves the assessment of water quality in rivers, with their similar interests in mean concentrations, maximum concentrations, and total mass loads. Total mass load is the total mass of the measure of interest over a period of time (e.g., total mass of nitrate in a year). In groundwater quality, this could be of interest when investigating bioaccumulating contaminants. Although significant research investigating how to best optimize sampling strategies for estimation of parameters such as total mass loads has been reported in this surface water literature (e.g., Kronvang and Bruhn 1996, Moosmann et al. 2005, Haraldson and Stalnacke 2006, and others), very limited literature on this topic has been identified involving assessment of groundwater quality. As such, a brief overview of the surface water literature is provided. In research on surface water literature, a diverse group of sampling strategies has been employed for the characterization of a variety of parameters. The sampling strategies used in the literature reviewed below include uniform time intervals of various frequencies (e.g., weekly or monthly), uniform intervals for various time periods (e.g., one, two and five years),

16

intervals that are dependent on the observed concentration of the measure of interest, and composite sampling in which uniform sampling is supplemented during peak flow events with increased sampling frequency. The efficacy of these sampling strategies in estimating parameters of interest was explored due to the cost and time demands of highfrequency or continuous sampling; due to limited available resources to perform sampling, the surface water research was aimed at the development of sampling strategies that best estimated the parameter of interest using the least available resources. Kronvang and Bruhn (1996) used two years of nearly continuous data to evaluate seven sampling strategies to estimate the annual total mass loading of nutrients in streams. They found that the most accurate estimate of annual total mass load was derived from a sampling strategy involving constant intervals consisting of two week periods between samples. These results suggest that relatively high-frequency data provided better statistical results than time-variable composite sampling. Moosmann et al. (2005) worked to find good compromises between the need for high-frequency monitoring and high monitoring costs in long-term projects. They found that: (i) estimates of the annual total mass load of soluble reactive phosphorus required composite sampling to minimize bias in the total mass load estimate, and (ii) that increasing the duration of the monitoring program (e.g., from three to five years) significantly reduced high-frequency sampling needs (to obtain similar statistics on the quality of the mass load estimate) while enabling the identification of long-term trends in phosphorus concentrations. Harladsen and Stalnacke (2006) built on Moosmann’s experience as they explored sampling strategies for estimation of monthly total mass loads of nutrients in agricultural catchments in Nordic countries. In their studies, time uniform sampling was

17

found to underestimate nutrient loads; again, composite sampling was required to more thoroughly characterize total mass load during high stream flow events. A significant observation described by these authors and others is that sampling for total mass loads of nutrients in streams requires different sampling techniques than monitoring for peak nutrient flow or trends in concentration. This implies that, in designing sampling plans for monitoring nutrients in surface water systems, the sampling plan should be directly tied to the parameters of interest. For example, if the project objective was the identification of trends in the nutrient loads of a stream, the literature suggests that uniform interval sampling is most suited, while composite sampling would provide the best estimates of annual total mass loads of the nutrients. Whereas this requires a greater understanding of the implications of specific sampling plans to the resulting estimation of specific parameters, it enables a more efficient use of financial, personnel, and equipment resources.

2.1.3.2 Relation of Surface Water Literature to Present Research The knowledge gained by surface water scientists can be interpreted within the context of groundwater quality monitoring. Sampling for trends in groundwater quality is often done by uniform time intervals (e.g., Dukes and Evans 2006) and some research has addressed the optimization of the time interval for specific field cases based on geostatistics (e.g., Cameron and Hunger 2002). However, as suggested from the surface water literature, uniform time intervals will likely be insufficient in certain sampling scenarios such as when total mass is the parameter to be estimated. Hence, we herein refer to the range of sampling strategies that may be of interest in addressing groundwater

18

quality characterization as the Progression of Sampling Strategies (POSS) and classify strategies within this progression primarily based on the complexity of the strategy. Figure 2.4, for example, shows the POSS considered in this dissertation to examine the possible impact of the POSS on three groundwater quality parameters; clearly, a broader range of sampling scenarios are possible and definition of a POSS may be case and parameter specific.

Figure 2.4: POSS

2.1.4 Progression of Expertise The third component of a strategy to collect field data is the selection of personnel participating in data collection based on the available progression of expertise (POE); as with the other progressions, the definition of a POE will depend on specifics of a characterization effort. For the present research, we utilize the POE illustrated in Figure 2.5, which serves as one possible POE, and is considered consistent with work in many rural regions of developing countries.

19

Figure 2.5: POE

The POE in Figure 2.5 considers the expertise of engineers / scientists and trained local populations that could be involved in the data collection. Of these two personnel groups, engineers / scientists with the highest level of expertise include those familiar not only with the project objectives, but also with the analytical equipment and the sampling strategy being utilized. Familiarity with these aspects of the project, in combination with their professional training and experience, enables these individuals to make informed decisions throughout the data collection process. It is recognized that there can be a considerable range of expertise represented here dependent on the skill sets of the individual engineer / scientist and their familiarity with the specific requirements of a given project; the left end of the POE is therefore represented by three descriptions of engineers / scientists. The second group of personnel, trained local populations, includes individuals trained to use specific analytical methods or sampling strategies, and is divided between those individuals that are or are not familiar with the project objectives. Similar to the engineers / scientists involved in data collection, trained members of the local

20

populations familiar with the project objective are considered to have greater expertise for the specific sampling plan as they have the potential to make decisions during the data collection process that take into consideration the intended objectives of the project.

2.1.4 Sociological Considerations and Development Literature In addressing the concepts of the POAM, POSS, and POE, an additional consideration may come into play. In the Master’s thesis work which provides the foundation for this dissertation (Crane 2006), we demonstrated that establishing a sampling program within a local population required not only selection of field methods consistent with capabilities of those who would perform the measurements, but also required understanding of the sociological aspects of implementing a project in collaboration with the population of stakeholders whom would be performing the sampling. Further, it required the creation of open communication between all involved in the project. In particular, this earlier work implied that use of a group of appropriate sociological methods was critical to the success of the monitoring program. For reasons detailed in Crane (2006), focus groups and population surveys, as well as direct observation of the population and its reaction to the ongoing project, were critical to the success of the sampling program. This observation is consistent with the sociological and anthropological development literature (e.g., Katz and Sara 1998, and Huby and Stevensons 2003). Specifically, Katz and Sara’s research on water projects in developing nations found that project success was higher when household members, rather than local leaders or outside experts, expressed the need for a specific project. Furthering this idea, Huby and

21

Stevensons suggest that although programs can be structured regionally, systems need to be in place for program decisions to be made at a village level. Both examples suggest that sociological research techniques and communication structures must be part of the project design in order to promote overall success. Similar arguments have been presented over the past decade in the hydrologic literature (e.g., Silliman et al. 2007, Journal of Hydrologic Engineering, in press). As our research continues to rely substantially on cooperation with, and sampling by, local populations in rural Bénin, the sociological methods outlined by Crane (2006) will continue to be employed. These will focus, to a large degree, on direct observation, focus groups, and surveying. Further, we have employed an anthropologist to assess the impact of our presence and our project in the study region. Assessment by this anthropologist has provided substantial insight and guidance relative to both technical and non-technical aspects of the local measurement program. However, as the sociological and anthropological methods employed are not a focus of this dissertation, they are not further detailed.

2.2 Sampling Strategies and Groundwater Research: A Perspective In the United States, groundwater research and monitoring has become increasingly reliant on state-of-the-art (usually highly complex) analytical methods combined with carefully designed, high-frequency (in terms of time or space) sampling strategies, that demand high levels of expertise. This focus on the left side of the progressions is a significant shift from forty years ago when less sophisticated analytical

22

methods and sampling strategies were the norm and cost combined with ability to recreate sampling instruments were important considerations in sampling design. A brief review of the journal Ground Water since its inception in 1963 demonstrates this shift. In Ground Water’s first fifteen years, articles discussing field research referred to analytical methods and sampling strategies toward the right, or less complicated side of the current POAM and POSS. Several examples of this include: (i) investigation of a plume through weekly monitoring of temperature using thermometers with 2oF divisions and monthly phosphorus samples (Sampayo and Wilke 1963), (ii) use of peizometers in combination with simple geochemical parameters for determination of artificial recharge to an aquifer (Kelly 1967), and (iii) characterizing bacteria at liquid waste injection sites through culturing water samples on agar plates with different media (Willis et al. 1975). By 1981 articles in Ground Water begin to demonstrate the increasing focus on the more complex side of the progressions as they report on the use of digital data recorders to collect high-frequency data, (e.g., Getzen 1981, and Buss and Bandt 1981). After this point, and in the following twenty-five years, there was a move from use of existing wells and field adaptation of equipment for specific projects to the use of complex sampling equipment and data recording devices, high-frequency or complex sampling strategies, and an increasing use of modeling as opposed to field research. Examples of this include: (i) identification of a nitrate plume from a septic system that used 500 sampling points, nine multilevel piezometer bundles, and isotopic ratios (Aravena, Evans and Cherry 1993), (ii) use of flow-cells to measure chemical parameters and perform biofilm sampling at municipal wells to analyze for iron-related biofouling

23

(Tuhuela, Smith and Tuovinen 1993), and (iii) vacuum filtering of groundwater and epifluorescent-microscopy enumeration in the field of bacterial tracers previously isolated, stained, and reintroduced to the groundwater system (Becker et. al. 2003). It is also important to note that during this time Ground Water articles no longer contained discussion of either the cost of new methods or the relative ease of creating sampling devices / transporting them to the field. This distinction between the early and recent focus of field methods in groundwater research is further demonstrated in the language used to describe field methods. For example, in 1963, the methods section of a paper in Ground Water began, “The writer has developed an inexpensive and relatively simple method for making microtime water level measurements in a well equipped with a recording gage” (Walton 1963). Here, the author viewed it as important to immediately draw the reader’s attention to the low cost and low complexity of the developed method. In contrast to this is the opening sentence of the methods section of a paper published this year in the same journal, “Pore water samples for DOC analyses were collected from eight piezometers installed in the upper oxidized (1.2 and 2.2 m BG) and lower unoxidized till (3.7, 5.3, 6.9, 8.4, 9.7, and 11.2 m BG) on November 4, 2004” (Reszat and Hendry 2007). Here, the reader’s attention is drawn to the complexity and thoroughness of the field sampling method. Although these are just two examples, they are characteristic of the focus of groundwater field sampling methods at the time each was published. As groundwater research has moved towards predominant use of complex methods, the literature suggests that it has simultaneously moved away from use of results derived using many of the older, less complex, methods that may still prove

24

extremely effective in select field situations including rural regions of developing nations due to the operational conditions of the field sites. In rural regions of developing countries, these constraints include poor infrastructure (little to no electricity or regional laboratories), poor road systems (difficult and time consuming to reach remote locations), and limited financial and personnel resources. Given these operational constraints and the current emphasis on complex techniques, research in these regions is either (i) not performed, (ii) when completed, not published, or (iii) when completed, published in journals not commonly cited by the mainstream professional community (e.g., Journal of African Earth Sciences). Given this disconnect between the direction of groundwater research and the need for, and difficulty of, performing water quality research in rural regions of developing nations, our research explores the use the POAM, POSS, and POE in groundwater research in rural regions of developing nations in order to develop data sets which are useful for many common water quality assessment projects.

2.3 Evaluation of Research Strategies For the purposes of this research, two research techniques, Monte Carlo numerical studies and field case studies, were selected to assist in the exploration of the use of the three progressions. The specific measures and parameters chosen for this research are not intended to be an exhaustive list of those available, nor are they intended to represent the full extent of the wide array of groundwater monitoring or research needs in developing nations. Rather, they were chosen to provide an assessment of the working concept. It is anticipated that further research will stem from the results contained in this dissertation that will further explore the wide array of available measures and parameters. The

25

following sections provide brief overview of the research methods, measures and parameters chosen for this dissertation research.

2.3.1 Monte Carlo Studies Monte Carlo studies have been used in a wide array of scientific and engineering fields, including extensive use in groundwater research. (For several recently published examples see: Kupfersberger and Blöschl 1995, Yenigül, et al. 2005, and Gallagher and Doherty 2007.) Monte Carlo studies were chosen for this research to provide numerical analysis of the working concept. Monte Carlo methods are based on numerically producing a series of equally possible realizations of a system (based on a probabilistic description of that system). Analysis of these realizations provides the researcher with the ability to study the range of possible behaviors that may occur in a system subject to uncertainty. As such, as the number of realizations becomes large, Monte Carlo methods can provide probability-based estimates of the behavior of the system under study (Haldar and Mahadevan 2000). For our efforts, the concentration of a generic chemical will be simulated over time based on a series of probability models (various mean, variance, correlation, and drift structures). This set of concentration values will then be subsampled to form the series of sample measures. These measures will then be used to assess three parameters: mean concentration, maximum concentration, and total mass load (TML, through cosimulation of the volumetric rate of production of water from a well). Mean concentration is selected as representative of situations either where the average concentration of an analyte is of interest or when investigating temporal drift in the data.

26

Maximum concentration is selected as representative of situations in which potential violation of maximum concentration constraints is being investigated (see, for example, limits placed by the EPA on coliform bacteria in a drinking water source, EPA 2001). As above discussed, TML is employed as representative of situations in which bioaccumulation of a particular chemical compound is of concern. These three parameters represent a selection of parameters that could be used in groundwater quality research, and parameters for which statistical analyses can be easily performed. As such, they provide a reasonable set of parameters from which to initiate our study of the working concept through Monte Carlo studies. The specifics of the methods used in the Monte Carlo analysis are presented in Chapter 5; results and analysis are found in Chapter 6.

2.3.2 Use of Case Study Coupling a case study alongside Monte Carlo studies can significantly bolster the strength of a research program. For this dissertation, the case study is based on ongoing groundwater research in Bénin, West Africa. This particular case study provides the opportunity to observe the impact on quality of data collection efforts of numerous critical constraints on the methods, sampling frequencies, and field personnel involved in the collection effort. These constraints are similar to those experienced in projects initiated in other developing nations (such as the Bangladesh arsenic example, Hossain 2006) and provided the original motivation for pursing this research. The case study allows assessment of the working concept related to a number of factors impacting groundwater research in Bénin, including:

27

•

the region used in the case study is a significant distance (approximately 5 hours of driving) from the water-quality laboratory available in Cotonou, Bénin (the nearest location with reasonable analytical laboratory resources), thus preventing collection of water samples on regular intervals from multiple villages in the study area as it is prohibitive from the view point of both logistics and resources,

•

the water laboratory available in-country is not equipped to perform chemical analyses that require analytical methods that are found on the more complex side of the POAM (e.g., trace-metals analysis, analysis for most organics, isotopic analysis) – thus requiring water samples to be shipped to laboratories in other countries (specifically, the analytical laboratories at Notre Dame, the isotope laboratory at Waterloo, Canada, and the USGS Stable Isotope Laboratory at Reston, VA),

•

select analyses (e.g., nitrate) are time sensitive and therefore cannot be preserved for analysis in Cotonou or at Notre Dame without use of chemicals considered dangerous for handling by the local population,

•

basic infrastructure (electricity, running water, access to computers) is not available in the field except through equipment carried to the field by project personnel,

•

limitations on financial and personnel resources in Bénin preclude extensive professional sampling efforts in the region of our case study, and

•

since1998, Notre Dame has collaborated on a number of groundwater studies in Bénin with multiple partners including the national university in Bénin (Universite d’Abomey-Calavi – UAC), a West African educational NGO (Centre

28

Afrika Obota – CAO), and the government water agency (Direction l’Hydraulique – DH) – thus providing a strong working relationship from which to pursue this research.

The collaboration between UAC, DH, and the University of Notre Dame (UND) began in 1998 when Dr. Silliman was invited to drill several wells in southern Bénin. Due to an overlap of research interests of the involved parties, exploratory groundwater research began in 2000. Since the original sampling trip, methods have been solidified and there have been an additional five sampling trips, which have been the foundation of two master’s theses and one research paper (Galbis-Reig 2002, Roope 2003, and Silliman et al. 2007). In addition to the exploratory groundwater research, the collaboration has grown to include groundwater modeling and education efforts that, because they are beyond the scope of this case study, are not further discussed herein. One of the observations derived from recent exploratory data analyses was a geochemically unique region in south-central Bénin (Silliman et al. 2007); details that lead to this observation are found in Chapter 3. In parallel to the exploratory research was the ongoing monitoring of wells by DH that lead to the identification of numerous handpumped wells with elevated nitrate concentrations. Combining the knowledge derived from the DH monitoring and the results of the exploratory analyses lead to research in south-central Bénin in which nitrate and, more recently, uranium have been the primary foci of the continued study; both topics were identified by DH due to their concern for the health of local populations.

29

The foundational work for the case study in south-central Bénin can be found in Crane’s master’s thesis (2006). One of the primary conclusions derived from the thesis was that a set of methods does exist that allows regular monitoring of nitrate by local populations having minimal education and no scientific background. The current research in Bénin is based on that success, as well as the prior regional geochemical analyses of groundwater. The current research includes the following: (i) examination of data collection methods for characterization of nitrate concentration by local populations in five villages, (ii) periodic (annual or semi-annual) sampling using more sophisticated analytical methods in these five villages, and (iv) extension of this work to the characterization of uranium contamination of certain wells in the dissolved and particulate form in the groundwater. The methods used in this case study are found in Chapter 4; results and analysis of the case study are found in Chapter 6 of this dissertation.

30

CHAPTER 3:

CASE STUDY: EXPLORATORY RESEARCH IN BÉNIN

This chapter contains the preliminary, or exploratory research, that was performed as part of the Bénin case study. Returning to Figure 2.1 and associated discussion outlining how an expert creates a research plan, this chapter is the first step, or “Expert & Prior Analysis.” As such, all information included in this chapter is intended to support the selection of specific parameters and regions of interest, and provide background information on the field site that then becomes the primary focus of the case study. The majority of this chapter has been published (Silliman, et al. 2007), and significant portions are taken verbatim from the article; details not included in this chapter can be found in the published article. It should be noted that the techniques used in this chapter generally fall under the more complex side of the three progressions (i.e., POAM, POSS, and POE).

3.1 Field Site Specifics The study area for the exploratory research, central Bénin, West Africa, is bounded by latitude ~7o N through ~11o30’ N and longitude ~1o20’ E through ~3o E (an area of approximate dimension 500 km by 180 km; see Figure 3.1). Based on the work of LeLong (1963) and Azonsi and Adjomayi (2005), it is known that Bénin is an

31

intertropical savanna with between 900 and 1500 mm of annual precipitation. The climate progresses from sub-equatorial in the south (with two rainy seasons: April through July and September through November) to tropical in the north (with one rainy season: May through September); see Figure 3.2. Annual potential evapotranspiration is estimated to vary spatially from a mean value of 1350 mm in the south to 1700 mm in the north. Rainfall exceeds evapotranspiration for approximately 6 months out of the year in the south and 3 months out of the year in the north, thus providing the opportunity for recharge of the groundwater. The details of the hydrologic cycle, particularly in the study area, remain poorly defined.

Figure 3.1: Political map of Bénin (taken from the University of Texas at http://www.lib.utexas.edu/maps/cia07/Bénin_sm_2007.gif)

32

Figure 3.2: Monthly distribution of mean annual precipitation of Bénin by latitude (taken from Galbis-Reig 2002)

The geology of Bénin is largely composed of igneous and metamorphic Precambrian rocks (approximately 80% of the country including the entire study area) (Boukari 1980, 1982, 1989). Sedimentary sequences (approximately 20% of the country) can be found in the coastal basin in southern Bénin (Mesozoic and Cenozoic), the Paleozoic basin of Kandi in northeastern Bénin, and the Cambrian basin of Pendjari in northwestern Bénin (Pougnet 1955, Affaton 1975, Insitute Ricerche Breda 1982). Within the study area, the metamorphic formations consist primarily of complex, undifferentiated migmatites but also include granulitic complexes and diverse gneisses, with less abundant amphibolites, mica schists, pyroxenites, serpentinites, quartzites and

33

marbles (Affaton 1975, BRGM 1978, and Insitute Ricerche Breda 1982, 1987). The igneous formations are primarily composed of complexes of syn-tectonic to post-tectonic intrusions (granite with a medium to fine grain, granodiorite, monzodiorite, diorites, gabbros, diabase, and basalt) but also include some volcanic rocks (basalt and rhyolite). Volcanic sequences are also present and consist of felsic to intermediate rocks (Boussari, 1975) with veins of quartz, pegmatite, and diabase in various combinations. The hydrogeology within the study region is based on fractured rock formations with minimal matrix porosity such that groundwater is derived primarily from regions of alteration or fractures. Alteration is most common in the upper 10-20 meters of the subsurface and is associated either with shallow fracture networks or zones of disaggregation leading to formation of sand lenses and/or saprolites. Fractures are present beneath the zone of weathering, with fractures and joints forming potential groundwater flow routes. The fractures are generally sub-vertical, but variable in orientation. Based on this geology, water supplies within the crystalline rocks of Bénin are commonly developed from three ranges of depth that are differentiated by available porosity: •

Superficial aquifers in the shallow, altered rock. These aquifers exhibit an interstitial porosity and are commonly exploited as water supplies based on largediameter, hand-dug wells.

•

Deep fractured rock, with groundwater flow reliant solely on the porosity of the fractures. Development is pursued through drilling, typically with air hammers, to median depths of 60 meters (Guiraud 1975, Engalenc 1978). These wells are commonly equipped either with submersible pumps (where electricity is available

34

and demand is sufficient to warrant construction of a centralized supply) or with manual pumps (hand or foot pumps). •

Fracture zones at the transition between the superficial aquifers and the deep aquifers. While these depths may exhibit behavior associated with both fracture flow and interstitial porosity, their behavior tends towards that of fractured media. These zones have been exploited both through construction of hand-dug wells and through drilling.

Land use in the region includes small acreage agriculture, rural residential, small urban, and undeveloped land. Faced with the scarcity and poor quality of surface water, much of the population originally invested in large-diameter, hand-dug wells that penetrated only into the upper, weathered zones. Difficulties with these wells were observed both in terms of maintaining water quality and of declining water levels between rainy seasons and during periods of drought. In the 1980’s, programs were initiated to alleviate the water problems via installation of small-diameter, drilled wells penetrating into the deep fractured zones. As these wells penetrated to the deep fractures, it was anticipated that there would be the additional benefit of lower risk of direct anthropogenic contamination. The majority of these drilled wells have a relatively small capacity (0.7 to 5 m3/hr), and are therefore equipped with manual (hand or foot) pumps. Select deep wells with larger capacity were equipped with submersible pumps in order to provide water supply for towns and villages with higher population density. Only small capacity drilled wells equipped with manual pumps are included in this study.

35

3.2 Regional Groundwater Study A preliminary field sampling campaign was conducted in the year 2000 and included sampling of 20 wells. Although the resulting data base was considered unreliable due to difficulties in sampling and analysis protocols, this initial sampling effort guided the design of the long-term sampling strategy (Galbis-Reig 2001). Based on results derived from this initial effort, three field campaigns (summers of 2002 – 2004) were pursued and have provided data for this study. With sampling locations as shown in Figure 3.3, these three field campaigns resulted in collecting samples from 37, 28, and 43 groundwater wells, respectively. Field methods are discussed in the following section. Based on these field campaigns, samples were returned to the University of Notre Dame for analysis via a number of analytical devices including ICP-MS (inductively coupled plasma – mass spectrometry), ICP-OES (inductively coupled plasma – optical emission spectrometry), and specific ion electrodes. The distribution of sampling points in Bénin has involved a compromise between sampling for specific purposes (e.g., analysis of elevated concentration of nitrate and uranium in the southern portion of the study site) and sampling for the purpose of developing estimates of spatial structure (e.g., estimation of the variogram). While sampling to characterize specific contamination scenarios leads to a desire to concentrate sampling in specific regions, optimizing the estimate of the variogram leads to a desire to distribute sampling throughout the region of study (e.g., Journel and Huijbregts 1978, Conwell et al. 1997, and others). With respect to Bénin, assessment of the distribution of data following the 2003 sampling campaign demonstrated that the data were heavily clustered in the south-central region. In order to provide a distribution of samples more

36

Figure 3.3: The numbers represent approximate location of wells sampled with the number representing the year of sampling (2, 3, and 4 representing, respectively, 2002, 2003, and 2004).

consistent with estimation of spatial statistics, the locations of samples collected in the summer of 2004 were based on a sampling design dedicated to increasing the utility of the data set for spatial statistical analysis.

37

3.2.1 Methods The methods used for the exploratory research in Bénin are divided into three sections for clarity: field methods, analytical methods, and data analysis methods.

Field Methods The wells sampled in this study are all in the deep aquifer and are equipped with manual pumps. Each was drilled by air-hammer (with rotary drilling occasionally employed to penetrate altered rock near the surface) and was screened at the depth of the most productive fracture zone. Casing with grout was used to seal the well above the screen. Depths range from ~50 meters in the southern portion of the study area to ~70 meters in the northern portion of the region. By focusing solely on wells drilled by air hammer, screened at the zone of production, and equipped with manual pumps, the variation in well chemistry related to variable methods of drilling, construction, or sampling (e.g., possible variation from high-capacity pumps found in the larger cities to buckets used in large-diameter wells throughout the country) is substantially reduced, thus reducing one critical variable in the analysis of the resulting data. As a result of the fact that the wells sampled within this study were actively in use by local populations, it was difficult to follow standard sampling procedure that requires purging multiple well bore volumes prior to collecting a sample. However, substantial effort was made to sample in a manner that would produce reliable data. The following summarizes the field methods employed: 1. As noted above, only drilled wells equipped with manual (hand or foot) pumps were included in this analysis. Thus data from hand-dug wells and/or wells

38

equipped with submersible pumps were not included in the data analysis discussed below. 2. All wells sampled had been in active production by the local population within the previous 24 hours. Further, the majority of wells were in active production when approached for sampling. Hence, concern regarding clearing of multiple well volumes prior to sampling was considered minimal as the water within the wells is refreshed (by pumping) on a nearly continuous basis. Despite this, our field procedure required, for any well not actively being pumped when we arrived for sampling, pumping a minimum of 100 cycles prior to collecting the sample. This level of cycling produces an estimated 100 liters of water, thus it is recognized that it does not ensure complete flushing of the wellbore volume (with an estimated volume between 200 to 800 liters). However, this cycling did ensure flushing of the pump cylinder and drop pipe (the pipe leading from the pump cylinder to the surface), which has a total estimated volume of approximately 15 liters. Hence, it was assumed that the regular use of these wells combined with the 100 cycles pre-sampling was an adequate compromise between following standard development procedures and managing time spent at each sampling location (as well as concern by the local population regarding our monopolizing their pump). 3. Immediately after sample collection, 60 mL of water was filtered through a 0.45 micron filter and acidified with 1.5 mL of concentrated nitric acid. A second 60 mL sample was collected without filtration or acidification. These samples were returned to the University of Notre Dame for elemental analyses. Due to the

39

remote field conditions, these samples could not be refrigerated prior to return to the United States with lag periods as long as ~14 days. 4. An additional container of approximately 3 liters of well water was collected to be tested at the site for temperature, pH, conductivity, and turbidity. As our equipment has improved over the three years of sampling, dissolved oxygen and salinity were added to the parameters sampled in the field, as was nitrate during the 2003 campaign. For consistency, only pH and conductivity are assessed from these parameters in the spatial analysis discussed below.

Analytical Methods The non-acidified / unfiltered samples were analyzed at Notre Dame for chloride and fluoride using specific ion electrodes which were calibrated against standards ranging from 1.7 mg/L to 500 mg/L for chloride and 0.85 mg/L to 49.5 mg/L for fluoride. The acidified / filtered samples were analyzed for both major and minor elements. Major elements were determined using ICP-OES. Minor and trace elements were determined using ICP-MS. In addition to well samples, samples of the nitric acid used during the field sample preservation process (diluted into distilled water) were analyzed. Although samples from 2002 and 2003 were originally analyzed in the fall of those years, respectively, using a Fisons quadrupole PQII STE ICP-MS (Roope 2003), a new Thermofinnegan Element 2 high resolution magnetic sector ICP-MS was brought on-line in the fall of 2004. This unit has allowed greater precision in the measurement of trace elements. Hence, all samples were reanalyzed on this new unit and it is the results of analysis from this newer unit that are discussed below. Samples from 2002 and 2003

40

were kept refrigerated and acidified prior to analysis using the Element 2. The resulting list of parameters used in our analysis is provided in Table 3.1. Additional elements were analyzed, but were eliminated from the data analysis due to excessive numbers of nondetects (e.g., Al, Cs, Re, Sb, Bi, Th) or mass interference without an alternative isotope for analysis on the ICP-MS (e.g, Sc, As, Se, following on the work of Stetzenbach et al. 1994). Internal standards (Be, Ge, Y, In, Tb, and Tl) were used in the ICP-MS analyses of the water samples to correct for machine drift. Based on these estimates of drift on the internal standards, measured intensities for all samples were corrected for drift, background (blank) intensity, and then assessed against calibration curves based on the standards prepared from stock solutions of known concentration. All dilutions were conducted gravimetrically for greater accuracy. Finally, where possible, multiple isotopes were run for the same element in an effort to identify and avoid mass interferences on the ICP-MS. Calibration standards and reference materials SLRS-3 (Riverine Water, NRC Canada; McLaren 1994) and NIST 1643d (Trace Metals in Water) were included at several points in each analytical run on the ICP-MS in order to determine precision and accuracy of the results. All samples were also analyzed on a Perkin Elmer 3300 XL ICP-OES for major ions. Following a similar method to that detailed above for the ICP-MS, internal standards (Y and In) were used to correct for machine drift. Final measured counts were corrected for drift, background (blank) intensity, and then assessed against calibration curves based on the standards prepared from stock solutions of known concentration.

41

TABLE 3.1 ELEMENTS FOR WHICH ANALYTICAL RESULTS WERE USED IN THE STATISTICAL ANALYSES. ELEMENTS ANALYZED BY ICP-MS INCLUDE THE MASS ANALYZED WHEREAS THOSE ANALYZED BY ICP-OES INCLUDE THE WAVELENGTH ANALYZED.

Element or Parameter

Mean / Standard Deviation / Estimated Detection Limits of Results With Units Indicated Parameters Included in Final Analysis Ti49 1.7 / 2.9 / 0.7ppb V51 11.8 / 11.5 / 0.2 ppb Cr52 3.3 / 7.8 / 0.7 ppb Co59 0.52 / 0.62 / 0.02 ppb Ni62 9.4 / 56 / 1.0 ppb Cu63 3.1 / 4.5 / 0.2 ppb Mo95 2.7 / 3.9 / 0.03 ppb Sn118 0.06 / 0.13 / 0.04 ppb Pb206 0.38 / 0.39 / 0.02 ppb U238 12.8 / 59 / 0.003 ppb Mn55 106 / 165 / 0.03 ppb Zn64 75.7 / 158 / 0.4 ppb Sr88 0.39 / 1.01 / 0.06 ppm Ba137 0.15 / 0.15 / 0.04 ppm F¯ 0.54 / 0.72 / ~0.1 ppm Cl¯ 24.5 / 44.4 / ~1.0 ppm Na330.237 24.9 / 20.9 / 2.4 ppm Mg285.213 21.2 / 17.1 /0.40 ppm Al308.215 0.04 / 0.06 / 0.12 ppm Si251.611 37.6 / 14.4 / 0.62 ppm P213.617 0.06 / 0.09 / 0.02 ppm S181.975 4.5 / 11.4 / 0.07 ppm K776.490 9.6 / 12.6 / 0.26 ppm Ca317.933 48.3 / 44.2 / 1.41 ppm Fe238.204 0.39 / 1.00 / 0.01 ppm pH 6.25 / 0.54 Conductivity 434 / 364 μS/cm Latitude N/A Longitude N/A

42

Comments

Analysis Method ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS and ICP-OES ICP-MS Specific Ion Electrode Specific Ion Electrode ICP-OES ICP-OES ICP-OES ICP-OES ICP-OES ICP-OES ICP-OES ICP-OES ICP-OES * Field Probe Field Probe N/A N/A

All dilutions were conducted gravimetrically for greater accuracy. Multiple wavelengths were run for each element to allow comparison of calibration curves and consistency with reported values for the reference solutions. As with the ICP-MS, calibration standards and reference materials SLRS-3 and NIST 1643d were included at select points in each analytical run in order to determine precision and accuracy of the results.

Data Analysis Methods As above mentioned, use of multiple data analysis techniques can provide substantial insight into the information contained in a data base of water quality data (e.g., Farnham et al. 2000, and others). In the present study, analysis of the final element concentrations (as well as the field values for conductivity, pH, latitude, longitude and select concentration ratios) was performed in multiple stages including exploratory data analysis, indicator variable analysis, variogram analysis, cluster analysis and principal component analysis. These five levels of analysis allowed delineation of: (i) spatial patterns in single parameters, (ii) relationships among pairs of parameters (or parameter ratios), (iii) statistical (spatial) structure in single parameters (or parameter ratios), (iv) clusters of samples of similar chemical composition, and (v) groupings of parameters which describe the overall variability observed in the data. The combination provided a multifaceted analysis strategy. Given the focus of this dissertation, only the statistical analyses that directly support the ongoing case study through the identification of the parameters and regions of interest are included in this chapter. Spatial patterns in single parameters were investigated through use of indicator variables determined for multiple target levels for each of the measured parameters.

43

Specifically, for each target level, each sample was assigned either as 0 (sample value below the target level for the parameter of interest) or as 1 (sample value greater than or equal to the target level for the parameter of interest). Spatial plots were then generated contrasting the locations at which the indicator variable was equal to one versus equal to zero (for a particular parameter and a particular target level), thus providing the ability to highlight spatial patterns for each of the parameters. In order to investigate clustering among the samples and patterns among the parameters (excluding latitude and longitude), cluster analysis and principal component analysis were applied to the data (e.g., Salman and Ruka’h 1999). Cluster analysis (e.g., Kaufman 1990, Aldenderfer and Blashfield 1984) was applied to the original data without correction for spatial drift. Within our work, cluster analysis involved estimation of the Euclidian distance, Dm,n, between two sample points, xm and xn, defined as the square root of the sum of the square differences of each of the K parameter (concentration) values, Pγ, for those samples: K

Dm , n = {∑ [ Pk ( x m ) −Pk ( x n )] 2 }0.5

(3.1)

Dm,n = { Σ [ Pγ(xm) - Pγ(xn) ]2 }0.5

(3.2)

k =1

Cl

uster analysis allowed identification of groups of sample points that have similar chemical compositions based on this Euclidian distance. Principal component analysis allowed identification of structure or

groupings among the parameters (e.g., Joliff 2003 and Farnham, et al. 2003). Although various approaches are available for principal component analysis, the choice was made for the present analysis to transform each parameter to a standardized normal distribution.

44

The analysis was then applied to these transformed data. The results provide indications of groups of parameters that explain variability within the transformed data set.

3.2.2. Results Summary statistics and estimates of detection limits for each parameter are found in Table 3.1. In the following paragraphs, the general results for each analysis technique are presented. Specific results are then integrated in the subsequent analysis. Indicator Variable Analysis: Indicator variable analysis was utilized to explore the data for spatial structure within individual variables. This analysis led to a number of observations that were consistent across multiple elements. Figure 3.4, for example, shows all sample locations with the locations of high values for pH (left), conductivity (center), and P (right), plotted with dots and low values plotted as x’s. The threshold value for pH was 6.7, conductivity was 0.55 mS/cm, and phosphorus was 0.05 mg/L. It is noted that there is close spatial correspondence in southern Bénin of sample locations with high values of conductivity and high values of pH (as indicated by the samples enclosed by the box). This region also corresponds to sample locations showing low concentrations of P (as indicated by the preponderance of x’s in this region). Elevated values in this region were observed for a number of parameters, including: conductivity, pH, Fe, Ca, S, Mg, Na, Cl ¯, F¯, Zn, Mn, and Mo. Parameters showing relatively low concentrations in this region include: P, Sn and, to a lesser degree, Si. In addition, a number of parameter ratios showed decreased magnitude in this region: examples include Si/Cl ¯, Si/Conductivity and P/U. As discussed below, this region is identified by a

45

number of the analysis techniques as being a region with an elemental composition

Latitude

substantially different than that observed for the majority of the study region.

11.5

11.5

11.5

11

11

11

10.5

10.5

10.5

10

10

10

9.5

9.5

9.5

9

9

9

8.5

8.5

8.5

8

8

8

7.5

7.5

7.5

7

7

7

6.5

1

2

3

6.5

1

2

3

6.5

1

2

3

Longitude

Figure 3.4: Indicator variable analysis illustrating consistent clustering of high pH (left), high conductivity (middle), and low phosphorus (right) in south-central portion of study area (indicated in figure). High values are indicated by SOLID CIRCLES whereas low values are plotted with x’s.

A second observation identified using indicator variables was the difference in chemical composition of two wells in the northwest portion of the country. These two wells are shown by the oval in the left image of Figure 3.4. Although not obvious from these plots, these two points had extremely low values for conductivity and pH as well as very low concentrations for a number of major and minor elements (including Si and P in addition to Ca, Na, etc.). Examination of the geology indicates that these two wells are drilled into a substantially different lithology than the other wells sampled in this study 46

(the region is dominated by quartzite). Hence, water from these wells exhibits a substantially different element composition. Cluster Analysis: Cluster analysis was performed on standardized parameter values (value minus the mean and divided by the standard deviation) excluding latitude and longitude and using Euclidian distance as the measure of difference between sample locations. No corrections were made for drift, nor were the variables transformed into log space. Results were assessed using the resulting dendrogram and plots of the spatial distribution of groups. An example of a graph of spatial groups is shown in Figure 3.5. Specifically, Figure 3.5 provides a spatial map of the location of the 70 sample locations, shown by the ‘x’ symbols, that have the most similar parameter values (smallest separation distances on the dendrogram). The remaining sample locations that demonstrate larger variation in parameter values (larger separation distances) are represented by the dots on the figure. The largest cluster outside of the main cluster of 70 sample locations involves only 3 wells, hence all points outside the large cluster have relatively distinct chemistries. Two primary observations are made with respect to these results. First, there is a cluster of wells in south-central Bénin (Figure 3.5) where the chemistry is highly variable as indicated by the number of points in this region which are not in the large cluster. This region coincides with the region of elevated concentration of several elements identified through the parameter plots and as identified with the indicator variable analysis

47

Figure 3.5: Results from cluster analysis showing locations of the 70 wells with the smallest separation distances (x) versus those wells that fell outside of this group (solid circles).

(compare with Figure 3.4); further discussion is found below. Second, two wells in the northwestern portion of the sampling region form a separate cluster (identified by the oval in Figure 3.5), consistent with the results obtained through the exploratory data analysis and indicator variable analysis indicating that these two wells have element chemistries substantially different than the majority of other wells (compare with Figure 3.4). Principal Component Analysis (PCA): For PCA, all parameters with the exception of pH were first transformed into log space and then standardized (through subtraction of the transformed mean and division by the transformed standard deviation). Those parameters judged to have excessive values below detection limit were removed from the

48

analysis. Latitude and longitude were not included in the analysis. The resulting data set included 23 parameters. The results of the PCA are provided in Table 3.2 for the first three components; the score is the value associated with the individual parameter for the given component and represents the directional cosine of rotation in parameter space. The weight is the square of the score and represents the percent contribution of the parameter value to the resulting component. This shows that the first three components explained 35%, 13%, and 10% of the total variance, respectively. Hence, approximately 60% of the variability in the transformed data set is explained by the first three components. Looking at the weights within the individual components, it is observed that the first component is dominated by positive loadings with approximately uniform weights (~10-11% each) from Na, Mg, Ca, and Conductivity. A second group of elements each contributes weights of ~6-8% (Cl ¯, F¯, Sr, and K). Hence, the first component reflects, to a large degree, major elements within the groundwater system. The second component is dominated by several trace metals (Cu, Pb, Ni, Zn, Co, Cr), thus implying that the variation in groundwater chemistry is described, independent of the major elements, by variation in the dissolved trace metals. The third component is dominated by Si, Sn, Cr, and V. These results imply nearly complete separation of large loadings on the parameters in the first three components (e.g., Cr is the only parameter that has high loading in more than one component) with no significant negative loadings of the major contributors. One interpretation of this result, consistent with the results from the other

49

TABLE 3.2 RESULTS OF THE PRINCIPAL COMPONENT ANALYSIS – FIRST THREE COMPONENTS

Parameter

% Variance Explained

Cl ¯ F¯ Ti V Cr Co Ni Cu Mo Sn Pb U Mn Zn Sr Ba Na Mg Si K Ca pH Cond

Principal Component 1

Principal Component 2

Principal Component 3

34.9

12.7

9.6

Loading 0.25 0.26 0.19 0.17 -0.02 0.10 0.04 -0.03 0.22 -0.04 0.02 0.22 0.17 0.04 0.29 0.19 0.32 0.32 0.02 0.28 0.33 0.22 0.33

Weight 0.06 0.07 0.03 0.03 0.00 0.01 0.00 0.00 0.05 0.00 0.00 0.05 0.03 0.00 0.08 0.03 0.10 0.10 0.00 0.08 0.11 0.05 0.11

Loading 0.08 -0.02 0.13 0.01 0.29 0.31 0.38 0.45 0.09 -0.07 0.44 -0.02 0.08 0.34 -0.14 0.08 0.05 -0.06 0.06 0.07 -0.10 -0.22 -0.09

50

Weight 0.01 0.00 0.02 0.00 0.09 0.10 0.15 0.20 0.01 0.00 0.20 0.00 0.01 0.12 0.02 0.01 0.00 0.00 0.00 0.00 0.01 0.05 0.01

Loading -0.14 0.05 0.23 0.39 0.30 -0.20 -0.21 0.08 -0.12 0.39 0.10 -0.17 -0.216 -0.15 0.00 0.15 -0.11 0.12 0.48 0.17 0.04 -0.11 -0.02

Weight 0.02 0.00 0.05 0.15 0.09 0.04 0.05 0.01 0.01 0.15 0.01 0.03 0.05 0.02 0.00 0.01 0.01 0.01 0.23 0.03 0.00 0.01 0.00

analysis techniques and classical interpretation of principal components in mixed systems, is that these three components represent different, but not competing, sets of processes or mechanisms that have impacted the groundwater chemistry (i.e., there is no indication of mixing).

3.2.3 Analysis The most significant result of the statistical analyses was the identification of the region in south-central Bénin (identified by the indicator variable and cluster analyses as outlined in Figures 3.3 and 3.5) for which the groundwater exhibits an element composition that is both substantially different than the average composition observed in the remainder of the study region and variable (spatially) from well to well within this region. This region had not been previously identified through hydraulic or chemical methodologies as being unique or exceptional. Two analysis methods were used to further explore the data from this region. Both methods involved separating the total data set into two subsets based on the cluster analysis as shown in Figure 3.5. Of the 34 samples excluded from the primary cluster, 25 were located within the region of interest. While these 25 samples could have been assessed separately based solely on geographic proximity, a decision was made to base separation of the data on a consistent measure, in this case the cluster analysis, rather than an arbitrary choice based solely on geographic location. These two data sets were then assessed via exploratory data analysis. The results of this analysis demonstrated two primary features. First, the points included in the smaller data set, particularly those identified to be from the region of interest, were

51

significantly over-represented in the higher concentrations of the major elements. Second, the data in the smaller data set demonstrated a higher variation than the data in the larger group of samples. Both of these observations indicate that the water in this region is significantly different, in terms of concentrations, than the water in the larger region of Bénin. However, the analysis also indicated that the data in both groups tended to follow very similar trends between parameters. The second assessment of these data was performed using principal component analysis as applied to the same two subsets of the entire data set. Component weights were compared for: (i) the entire data set, (ii) the larger data subset (those points within the large cluster), and (iii) the smaller data subset (those points excluded from the large cluster). Consistent with the discussion provided above for the entire data set, the first component was reflective of the major elements for all three data sets. Further, consistent across the three data sets, the first significant appearance of Si was in the third component. Review of the parameter weights in the first components demonstrates the similarities and differences between the three data sets. For example, all three data sets have significant weights on conductivity, Ca, Na, Mg and Cl ¯ (i.e., several of the major elements), but demonstrate significant differences for select elements, including F¯, Sr, Ti, and K. Further, the larger subset generally shows more consistency with the entire data set than does the smaller subset. Hence, these data indicate that there is consistency among the entire data set and two subsets as expressed in the major elements. At the same time, the two subsets show significant difference in terms of the contribution of the secondary elements.

52

This difference among the three data sets is further highlighted in the third component (the first component with significant contribution from Si for each of the data sets). Although all three show significant weight on Si, there are substantial differences in the remaining parameter weights. As compared to the weights for the full data set, for example, the smaller subset shows significantly increased weights on F¯, Cu, K and pH with reduced weights on V and Cr (as well as a lower weight on Si). The larger subset shows greater weight on Ti and lower weight on Si (as compared to the full data set). Although not based in geochemical analysis, these results imply that there is a significant difference between the two subsets of the data regarding the interplay among the parameter groupings. In attempting to identify unique controls on elemental composition consistent with the original assessment of the complete data set and the analysis of the data subsets, patterns in geology, land-use, and precipitation were considered. Following on Tardy (1969), it is anticipated that each of these may impact groundwater chemistry and variability in this chemistry. Reviewing the hydrogeology of Bénin (Carte Hydrogeologique 1985) and associated discussion in Boukari (1982), it is observed that the south-central region of the study area that was identified via data analysis (Figure 3.4 and 3.5), more so than other regions within the study area, is characterized by relatively rapid changes in rock type, with variation among gabbros, gneisses, and granites and associated significant variations in mineralogy. While this variability may not provide an immediate explanation for the elevated concentrations of major elements in this region, it does provide potential insight into increased variability of both the major and trace elements.

53

Previous assessments (e.g., Carte Hydrogeologique 1985) also indicate that the reliability of water supply is substantially lower in the south-central region than it is elsewhere in the study area. Reliability, in this environment, is closely related to water storage capacity of the fractured rock and the overlying depth of weathered rock. Lower reliability, as measured by critical drop in water level during droughts, is an indication that there is minimal storage within this particular region, suggesting a thin weathered zone and minimal storage within the fractures (Boukari 1982). This environment results in greater sensitivity and response to water quality variations of recharge waters. Further, in this setting, opportunity for dissolution of major and trace elements may vary spatially due both to variation in residence time in the weathered zone and fractured zones (resulting from spatial variation in depth of weathering) as well as minimal mixing and more rapid response to local variation in quality of recharge waters. As such, impact of application of agricultural chemicals such as lime and fertilizers (as well as sensitivity to other anthropogenic contaminants) may be more significant in this region than in regions where a deeper weathered zone would tend both to minimize the impact of these chemicals and smooth the spatial / temporal variation in recharge chemistry related to application of these chemicals. Finally, review of precipitation patterns (Figure 3.2) in Bénin illustrates that this region represents a region of transition from the southern sub-equatorial to the northern tropical climates. Hence, the variability in timing and, more importantly, chemical composition of precipitation may be greater in this region than in regions further to the south or further to the north.

54

3.2.4 Measures of Further Interest This discussion of the relatively unique element chemistry in the identified region of south-central Bénin indicates that this region is more complex than the remainder of the study area in terms of groundwater quality. The importance of this complexity is highlighted by the observations both that all of the wells for which elevated uranium concentrations have been observed (with a maximum observed concentration of 491 ppb) are located within this region and that the government agency in Bénin responsible for water quality issues, Direction de l’Hydraulique (DH), has consistently reported nitrate concentrations above health standards in this general region of Bénin. Conversations with DH both during and after the completion of this exploratory regional study of groundwater in Bénin lead to the decision to pursue ongoing research of the measures nitrate and uranium within Colline Department (Figure 3.6), which approximates the south-central region of Bénin identified through this regional study. These measures were chosen due to the potential health implications of either the specific measure or its source. Measure details and the research performed are found in Chapters 4 and 6 of this dissertation.

55

Figure 3.6: Political departments of Bénin

56

CHAPTER 4

CASE STUDY: MEASURE SPECIFC RESEARCH IN BÉNIN

4.1 Measure: Nitrate Nitrate was chosen as one of the two measures of interest for the Bénin case study due to its health implications and its historical presence in elevated concentrations in the groundwater of the south-central region of Bénin. Prior to discussion of the available analytical methods and the techniques used in this case study, a brief review of the health implications of elevated concentrations of nitrate and historical nitrate data in Bénin is here provided. Normal levels of naturally derived nitrate in groundwater are less than 2 mg/L; the nitrate used by plants for their nitrogen needs due to its high solubility (Mueller and Helsel 1996, and Masoner and Mashburn 2004). The World Health Organization (WHO) and the U.S. Environmental Protection Agency (EPA) have set drinking water standard for nitrate at 50 mg/L NO3 (equivalent to 11.3 mg/l NO3-N) and 10 mg/L NO3-N, respectively (USEPA 2003 and WHO 2004). These limits result from health concerns associated with nitrate in drinking water, the primary concern of which is methaemoglobinaemia, or blue baby syndrome, which reduces the transportability of oxygen in the blood of infants (Knobeloch et. al. 2000). Nitrate has also been associated with miscarriages, thyroid disease, central nervous system cancers, and has been linked to

57

an increased risk of non-Hodgkin’s lymphoma (Canter 1997, L’Hirondel and L’Hirondel 2002, and Nolan, et. al. 1997). Possible sources for elevated nitrate concentrations in groundwater are essentially all linked to human activity including: fertilizers from home or agriculture use, leakage from sewage lines or on-site sewage disposal, animal waste, and industrial sources (Wakida and Lerner 2005). Significantly, each of these nitrate sources will commonly be associated with additional pollutants such as bacterial or organic contamination associated with human or animal waste, and toxic chemical pollutants associated with fertilizers (and pesticides normally applied in the same locations as the fertilizers) or industrial waste (Wakida and Lerner 2005, and Somasundaram, et. al. 1993). Hence, the presence of nitrate represents a direct medical threat and also serves as an indicator of other possible contaminants. Given that nitrate can be relatively easily monitored in field situations (as per analytical methods discussed in the following sections), nitrate has the ability to serve as an indicator of potentially more serious contamination in the groundwater that is more difficult (due to time, money, or available instruments) to monitor. The Bénin government agency responsible for monitoring water quality in drilled wells found in the study area is named Direction de l’Hydraulique (DH). Results from their monitoring efforts demonstrate that, in select wells, nitrate concentrations are consistently above WHO standards and are highly variable in time (e.g., Figure 4.1). Unfortunately, these data were taken at low-frequency (both in time and space) and do not provide geographical relation of wells to village activities, or other evidence of the source of nitrate. DH (Felix Azonsi, personal communication, 2005) has indicated that

58

the sampling design establishing the temporal / spatial frequency of sampling was largely determined by the financial and personnel limitations of DH. Hence, although the elevated concentrations of nitrate were identified in multiple wells, the full extent of the problem remains unknown such that location of the source(s) (and therefore remediation actions) cannot be taken, and additional contaminants potentially accompanying the nitrate in the groundwater are both unknown and untreated.

80.0

Nitrate (mg/L NO3-N)

70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 6/19/1997

1/5/1998

7/24/1998

2/9/1999

8/28/1999 3/15/2000 10/1/2000 4/19/2001 11/5/2001 5/24/2002

Date Moumoudji

Mahu

Adourekoman

Sowe

Figure 4.1: Select historical nitrate data in the Colline Department of Bénin (Data from DH.) One region of Bénin where the nitrates have been observed to be consistently elevated is located in the Colline Department. Although identified through local expertise independent of our regional sampling, this region coincides with the region identified through geostatistical assessment as discussed in Chapter 3. Hence, it was concluded that additional research investigating nitrate in the Colline Department would provide several

59

levels of benefit: it provides valuable information for DH and local populations concerning the source of nitrate, and enables an exploration of the use of the three progressions under realistic field conditions.

4.1.1 Available Analytical Methods The available analytical methods for measurement of concentration of nitrate cover a wide range in terms of complexity and expertise required. The POAM for nitrate was discussed in Chapter 2; the associated figure has been replicated in Figure 4.2 and the discussion is summarized here. In Figure 4.2, the complexity of a given method increases from the right to the left hand side of the POAM. Complexity was determined based on a range of factors including precision, bias, required equipment and chemicals, and degree of expertise required to use the method.

Figure 4.2: POAM for nitrate

60

Included among the most complex methods (found at the left hand side) of the POAM for nitrate are isotopic ratio methods, automated hydrazine reduction, automated cadmium reduction, and manual cadmium reduction. These methods involve use of complex laboratory methods and careful addition of hazardous or carcinogenic chemicals to the sample as part of the method (e.g., addition of hydrochloric acid in the ultraviolet spectrophotometric screening method or hydrazine sulfate in the hydrazine reduction method; Eaton et al. 2005). Towards the middle of this scale are methods such as cadmium reduction spectrophotometry and colorimetry. These methods are often based on addition of prepackaged reagent materials, require lower levels of sample preparation and preservation, and do not require that the technician be an expert at the chosen analytical method. In trade-off for these conveniences, these methods tend to have higher detection limits and lower precision. For example, the colorimeter used in our field work has a range of 1 – 33 mg/L NO3-N with a reported precision of +/- 3mg/L. At the right end of this POAM are test methods such as test strips that are extremely simple to use, require no reagents, and can be performed by a trained lay person. For the least complex of these methods, such as test strips, this comes at the cost of continuous scales. For example, the test strips for nitrate that we use in the field provide measures only in the following discrete steps: 0, 0.5, 2, 5, 10, 20, 50 mg/L NO3-N; company suggested precision is plus or minus one interval (Industrial Test Systems Technician, personal communication, 2005).

61

4.1.1.1 Common Practice in Developing Countries In reviewing the use of nitrate methods in developing countries, we start with our experience in Bénin where DH has historically relied on laboratory analysis using the cadmium reduction spectrophotometer method (DH Laboratory Manager, personal communication, 2005). DH has used this method on a select number of wells located throughout the Colline Department; samples were collected and analyzed by technicians employed by DH on an annual or semi-annual basis. Although the literature is limited with respect to applications in other regions of developing nations, a brief review of groundwater research of rural regions suggests that DH applies methods of groundwater quality monitoring on a more regular basis over longer time periods than is common in other studies. For example: •

Research in Burkina Faso used colorimetric methods to measure nitrate in 159 villages, but applied the technology at only one point in time (Groen, Schuchmann and Geirneaert 1988),

•

Research in India used specific ion electrodes to measure nitrate at 50 sites, but at two points in time, 1987 and 1989 (Somasundaram, Ravindran and Tellam 1993),

•

Research in Niger collected samples from potential nitrate sources and water samples from aquifers and used spectrophotometers and mass spectrophotometers to measure nitrate and nitrogen isotopes of N but limited their research to one point in time (Girard and Hillaire-Marcel 1997), and

•

Research in a village in South Africa used photometric methods (similar to spectrophotometric or colorimetric methods) to perform weekly measurements of

62

the nitrate concentrations in five water sources found in a single village, but only for 13 weeks (Nevondo and Cloete 1999). In these examples, the middle and complex portions of the POAM for nitrate were utilized, sampling was performed at relatively low-frequency in time and/or space, and sampling was performed by experts. At the time that this dissertation was written, no article was found in which a method from the right (less complex) side of the POAM for nitrate was utilized in research involving rural regions of developing nations, nor were studies reported involving long-term monitoring of nitrate in such regions. This serves to demonstrates that the full POAM is not currently being used in groundwater research in developing nations, and thus there is no model on which to base selection of techniques in the Bénin case study.

4.1.2 Methods Utilized in This Research In order to identify potential analytical methodologies allowing for time series sampling in rural field situations, we explored the full spectrum of the POAM for nitrate (i.e., those methods listed in Figure 4.2). In order to provide the possibility of identifying the source(s) of elevated nitrate concentrations in the Colline Department, it was determined that a combination of high-frequency and low-frequency sampling would provide a reasonable sampling strategy with sampling based on techniques derived from opposite ends of each of the POAM, POSS and POE: isotopic analysis of low-frequency samples performed by experts and high-frequency sampling performed by lay participants using colorimeters and test strips. These two groups of techniques are described below.

63

4.1.2.1 Low-frequency Sampling: Methods The low-frequency sampling methods involve sampling and testing the nitrogen and oxygen isotopes of nitrate in the groundwater of Bénin. Only drilled wells equipped with manual (hand or foot) pumps were sampled. All wells sampled had been in active production by the local population within the previous 24 hours. Further, the majority of wells were in active production when approached for sampling. For any well not actively being pumped when we arrived for sampling, a minimum of 100 cycles of pumping was performed prior to collecting the sample producing an estimated 100 liters of water. The remainder of the sample collection method is divided between the 2003 and 2007 samples. In 2003, samples were collected in bottles ranging in size from 125 mL to 1 L; the size of the sample collected was determined by the concentration of nitrate in the groundwater (determined by colorimeter methods described below) such that the necessary quantity of nitrate was collected for the laboratory analysis. At the end of a day of sampling the samples were preserved through the addition of mercuric chloride; thus there was a lag time between 2 and 10 hours between sample collection and preservation. Samples were transported by sampling personnel to the United States and shipped to the University of Waterloo isotope laboratory for analysis. In 2007, each sample was filtered with a 0.45 micron filter directly into a 125 mL bottle. Several drops of concentrated NaOH were then added to raise the sample pH to approximately 10 as per laboratory requirements for sample preservation. Samples were transported by sampling personnel to the United States and shipped to the USGS Stable Isotope Laboratory in Reston, Virginia for analysis.

64

4.1.2.2 Low-frequency Sampling: Field Application The low-frequency isotopic analysis has occurred in three stages. The first stage of isotopic analysis occurred in 2003 and was performed on 11 groundwater samples collected in the Colline Department. This sampling stage served as the preliminary portion of the nitrate study, and sample locations were identified by DH as wells having historically high concentrations of nitrate; this was verified during the field sampling. The range of isotopic values obtained from the Bénin wells were compared with previous characterizations of nitrate signatures (e.g., Groen, Schuchmann and Geirneaert 1988, and Clark and Fritz 1997). This comparison, illustrated in Figure 4.3, leads to an interpretation of these data in which the sources of nitrate in each well is derived from human and/or animal wastes (and limiting the possibility that an agricultural source is responsible for the nitrate contamination). Further, the data suggest that the nitrate may have undergone varying degrees of denitrification (Rock and Mayer 2002, Clark and Fritz 1997, and Roadcap, et al. 2001). Although this first stage of sampling did not identify specific sources of the nitrate contamination, it did narrow the list of potential sources and provided direction for the ongoing research. The second and third stages of low-frequency isotopic sampling were designed to support, in combination with the high-frequency data, the interpretation of the nitrate contamination and to aid in identification of specific sources of nitrate within a village. To this end, the second stage samples were collected in the primary study village, Adourékoman, in February of 2007. These samples were collected from the three hand pump wells and four hand dug wells located in the village. The third stage isotopic sampling occurred during May/June of 2007 and built on the second stage by including

65

Figure 4.3: Nitrogen and oxygen isotopes of nitrate in groundwater samples from the Colline Department taken during the summer of 2003. The line shown in the figure has a slope of ~0.5, indicative of behavior consistent with denitrification. Source compositions from Rock and Mayer (2002). repeat samples from the three hand pump wells in Adourékoman. In addition to those samples, samples were collected from two small seasonal streams in Adourékoman as well as from hand pump wells in the four villages participating in the expansion of this project. The goals of this stage of sampling were: 1. To enable comparison of the nitrate isotopes in the dry and rainy seasons. This goal stems from the difference that occurs between the rainy and dry seasons in Bénin including pump usage, animal watering patterns, and rainwater infiltration that could affect the presence and/or denitrifcation of nitrate in the groundwater. 2. To provide isotope analyses for all wells included in the high-frequency sampling. 3. To enable comparison of isotopic signatures found in streams (located near potential sources) to the signatures in the wells the primary village.

Figure 4.4 shows these newer samples, those collected in February and May of 2007, overlaid on the data from 2003. These data indicate general consistency in trend

66

among the various data sets. Interpretation of the data pertinent to the nitrate case study are discussed in Chapter 6.

Figure 4.4: Nitrogen and oxygen isotopes of nitrate in groundwater samples from the Colline Department taken during the summer of 2003 (*), February of 2007 (O), and May of 2007 (∆). The line shown in the figure has a slope of ~0.5, indicative of behavior consistent with denitrification. Source compositions from Rock and Mayer (2002).

4.1.2.3 High-frequency Sampling: Methods The high-frequency sampling methods include a single parameter colorimeter and test strips. As with the other sampling methods, only drilled wells equipped with manual (hand or foot) pumps were sampled. All wells sampled had been in active production by the local population within the previous 24 hours. Further, the majority of wells were in active production when approached for sampling. For any well not actively being pumped when we arrived for sampling, pumping a minimum of 100 cycles of pumping was performed prior to collecting the sample producing an estimated 100 liters of water. Samples were collected in dedicated plastic containers and divided between the sample needs of the colorimeter and the test strips.

67

The single parameter colorimeter (Hach Nitrate Pocket Colorimeter II) used a cadmium reduction method. As per company instructions, this method uses the following steps: (i)

addition of 10 mL of sample water into a glass sample vial and blank vial,

(ii)

addition of one packet of nitrate reagent (Hach product number 2106169) to the sample vial,

(iii) shake the sample vial for 1 minute, (iv) wait for 5 minutes, (v)

clean the exterior of each vial with water and Kimwipes,

(vi) zero the colorimeter with the blank vial, and (vii) measure the color change of the sample vial, which the colorimeter converts to a concentration.

This method was modified, followed laboratory tests, for the high-frequency sampling that used trained local populations to perform the method. This method replaced the third and fourth steps of the above method such that the sample shaking period was 2 minutes in duration and the wait period was 4 minutes. These methods were adapted to ensure an adequate shaking period; it also enabled the use of a 2 minute sand timer to replace a stop watch. Details of this method, associated lab experiments, and methods used in training of local populations are found in Crane (2006). The nitrate test strip (Industrial Test Systems product 481109) method used the following steps (no modifications to the test strip method were made):

68

(i)

place the test strip in sample water for 2 seconds moving the test strip in the sample,

(ii)

remove the test strip from the sample,

(iii) wait for 1 minute, and (iv) compare the color of the test strip to color chart associated with the specific method and batch of test strips.

4.1.2.4 High-frequency Sampling: Development The high-frequency sampling is based upon, and serves as an extension of, the work discussed in Crane (2006). The high-frequency sampling occurred in three stages in order to develop the methods, implement them in a test location, and finally expand to additional locations. The three stages of implementation are here termed: (i) development stage, (ii) initial implementation, and (iii) second level implementation. The development stage consisted of selecting the analytical methods most appropriate for use by lay individuals in Bénin and establishing techniques for training these individuals to reliably use the selected analytical methods; the development stage was the focus of the Crane (2006). As discussed in that thesis, the analytical methods selected for the high-frequency sampling include a nitrate single-parameter colorimeter and nitrate / nitrite test strips. These two analytical methods were chosen based on the following desired characteristics: •

Field portable and does not require electricity (except through standard batteries)

•

Requires neither significant handling of hazardous chemicals nor precision measurement of chemical mass or volume

69

•

Durable instrument which is either factory calibrated or has automated calibration not requiring the preparation / use of calibration fluids

•

Reagents can be safely delivered to country (or carried on commercial aircraft)

•

Continuous values are preferable (versus multiple, discrete ranges of values)

•

Sufficient level of bias / precision in combination with appropriate detection limits and concentration bounds

•

Availability of, or ability to, create training materials consistent with a population of low literacy or speaking a local dialect not conducive to translation of written materials

•

Reasonable cost relative to the local population’s or local government’s ability to pay

Following selection of these methods, significant effort was invested in understanding their strengths and limitations. This is particularly of note with the nitrate colorimeter. Initial evaluation of this instrument, a single-parameter nitrate colorimeter using the cadmium reduction method, was conducted under laboratory conditions and the impact of all reported interferences was investigated and determined to be minimal under anticipated field conditions. However, in application in the field in Bénin, it was noted that the colorimeter indicated non-detect or extremely low concentrations at certain wells despite other measures (specifically the test strips) indicating true concentrations above both instrument detection limit and WHO health standards. Further, while this underestimation was observed on select (seemingly random) days, the instrument

70

provided higher concentrations consistent with the strip tests (the strip tests provided consistent concentrations over several days). Tests to ascertain whether an interference was causing the underestimation of the concentration were performed in both the field and laboratory. These tests lead to the identification of a previously unknown interference: particulate calcium carbonate apparently interferes with a necessary cadmium reaction and, thereby, the necessary color development. This occurred regardless of the concentration of nitrate in the water or the concentration of calcium carbonate so long as particulate calcium carbonate was present. Specific results and additional discussion can be found in sections 2.3 and 5.1.1 of Crane’s master’s thesis (2006).

4.1.2.5 High-frequency Sampling: Initial Implementation The second stage of high-frequency testing, the initial implementation, included the implementation of the high-frequency monitoring in one village in the Colline Department of Bénin, Adourékoman. There are three wells in Adourékoman (named Ayewa, Ayewa-Okouta and Agbo; well latitudes, longitudes, and a village map found in Chapter 6); trainees from the local population were formed into three teams, with each of the teams trained and then assigned to sample the well in their region of the village on a weekly basis. During the remainder of the initial implementation, the three teams then performed weekly sampling for periods between 5 and 12 months. In order to demonstrate that the trainees, who had no scientific background and little formal education, could accurately perform these methods (both immediately after being trained and throughout the period of sampling), each team was required to analyze

71

individual nitrate standards (with concentrations unknown to the sampling teams) on a monthly basis. Results from these standard measures (Figure 4.5) indicate that, while the results were subject to bias (at a level similar to that observed when the colorimeter was used by professionals), there is no obvious temporal trend in this bias or in the concentrations recorded for similar samples. This is critical as it indicates that the local measurement teams, using the colorimeter, can provide consistent measures of nitrate concentration over time, thus allowing for identification of temporal trends in nitrate concentrations. 35.0

Nitrate Concentration (mg/L NO3-N)

30.0

25.0

20.0

15.0

10.0

5.0

0.0 17-Feb-05 28-May-05

5-Sep-05

14-Dec-05

24-Mar-06

2-Jul-06

10-Oct-06

18-Jan-07

Date 1 mg/L

5 mg/L

10 mg/L

20 mg/L

Figure 4.5: Nitrate standards as analyzed in initial implementation of high-frequency by groups in Adourékoman. The initial implementation stage of the high-frequency monitoring not only established that the local lay teams were able to consistently measure the quality of the

72

water over significant periods of time, but also provided measures of temporal variation of water quality of the groundwater in the three wells in Adourékoman. These results, combined with the isotopic results from the same wells, are the focus of the nitrate results and analysis as presented in the Chapter 6.

4.1.2.6 High-frequency Sampling: Extension The third stage of the high-frequency monitoring involved expansion of the monitoring to an additional ten wells in four new villages during the summer of 2006. Although the details of this expansion is beyond the scope of this dissertation, it is noted that the goal was to test different training techniques and models of involvement with the local communities, and the expansion was built upon the success in the second stage efforts. As with the initial implementation efforts, preliminary results from this expansion indicate local participants’ ability to accurately perform the analytical methods through the regular analysis of blind standards and the collection of high-frequency data over a period of one year. Results from the second level implementation will be the focus of future papers from this project and are not further discussed in this dissertation.

4.2 Measure: Uranium The second measure of interest in the Bénin case study, uranium, was originally identified through the exploratory field sampling campaign that relied predominantly on ICP-MS and ICP-OES. Uranium was chosen for this case study due both to potential medical implications of uranium in drinking water and to substantial interest by the local professional scientists at DH; it serves as a recent extension to the Bénin case study. Prior

73

to discussion of the available analytical methods and the techniques used in this case study, a brief review of the health implications of elevated concentrations of uranium and historical data concerning uranium in Bénin are provided. Natural sources of uranium are found in granite and other mineral deposits (WHO 2004), and enter groundwater systems through leaching of these natural deposits (EPA 2003). The WHO and EPA set the limit for uranium in drinking water, respectively, at 15 μg/L and 30 μg/L (WHO 2004 and EPA 2003). Although only sparse data exists implicating uranium with significant health problems, these limits were set due to an increased risk of cancer and kidney toxicity. A detailed discussion of the available research and reasons for the limits was published by the World Health Organization (WHO 2004). The regional study in Bénin, discussed in Chapter 3, found that 36 of the 108 groundwater samples tested had dissolved uranium concentrations greater than 1 ppb but less than the WHO standard for uranium in drinking water of 15 ppb, and an additional 13 of the samples exceeded the WHO standards, with a maximum concentration of 491 ppb. Two possible sources of the elevated concentrations of uranium are natural geologic deposits and nuclear waste (Ewing 1999). Although extensive geological studies with the goal of identifying uranium deposits have not been performed within Bénin (see TED 2002 and Kusnir and da Fonseca 1975 as examples of the limited published literature discussing uranium sources in Bénin), the statistical analyses performed as part of the regional groundwater study support the uranium being natural in origin. In addition to the dissolved phase of uranium, uranium may also be present in the colloidal phase. Colloids are defined as “suspended particles in the submicrometer size

74

range” (Kersting, et al. 1999) and occur naturally in groundwater. For the purposes of this research, colloids are defined as particles between 16 and 0.45 microns, and the dissolved phase is here defined as all liquid and particles smaller than 0.45 microns in diameter; these definitions are based on the analytical methods selected for this research as described in section 4.2.2. Although previously shown to effect the transport of contaminants in groundwater (e.g., McCarthy and Degueldre 992), recent research has specifically demonstrated the effect of colloids on the transport of elements in the actinide series (e.g., Kersting, et al. 1999, Honeyman 1999, and Novikov, et al. 2006). Hence, methodologies were investigated for measuring both the dissolved and colloidal phases of uranium.

4.2.1 Available Analytical Methods At the time that this research began, no methods for analysis of either the colloidal or dissolved uranium phases were identified that met any of the following criteria: field portable, did not require the use of dangerous chemicals, and did not necessitate the use of electricity, dedicated laboratory space and highly trained personnel. Existing methods for uranium measurement, which did not meet the above criteria, include fluorometric methods, radiochemical methods, alpha spectrometry, and inductively-coupled plasma mass spectrometry. Thus, for uranium, available methodologies exist only at the more complex end of the POAM. Given the lack of available methodology that could be brought into the field, much less used at high frequency by local populations, methods of higher complexity for field sample collection were selected for dissolved uranium, and new field sample

75

collection methods were established for the colloidal phase. By necessity of their application in rural Bénin, these methods involve preserving the samples collected in the field for laboratory analysis. Further, as detailed below, both methods require high levels of expertise due to the complexity of the method and/or the materials handled as part of the method (i.e., concentrated acid). Thus sampling was restricted to experts and, due to resource limitations, was therefore limited solely to low-frequency sampling. Given the significant limitations encountered with the analysis of uranium in terms of available analytical methods and associated necessary levels of expertise, it serves, in this case study, as an example of application under circumstances where the available portions of the progressions are extremely limited.

4.2.2 Methods This section presents the analytical methods, sampling strategy, and personnel utilized in the portion of case study devoted to the measurement of uranium. This section is further divided between field and laboratory methods. Only the results which pertain to the selection or support of these techniques are included in this section; all other results and associated analysis are found in Chapter 6.

76

4.2.2.1 Field Collection Methods The sample collection methods for the dissolved phase of uranium are the same as those used in the regional sampling study; these methods are detailed in section 3.2.1 of this dissertation with brief review here provided: 1. Only drilled wells equipped with manual (hand or foot) pumps were sampled 2. All wells sampled had been in active production by the local population within the previous 24 hours. Further, the majority of wells were in active production when approached for sampling. For any well not actively being pumped when we arrived for sampling, pumping a minimum of 100 cycles of pumping was performed prior to collecting the sample producing an estimated 100 liters of water. 3. Immediately after sample collection, 60 mL of water was filtered through a 0.45 micron filter and acidified with 1.5 mL of concentrated nitric acid. Due to the remote field conditions, these samples could not be refrigerated prior to return to the United States with lag periods as long as ~14 days.

The sample collection method for the colloidal phase of uranium was established during the spring of 2006. A water sample was collected from the well in a dedicated plastic bucket with a lid to reduce the opportunity for contamination from airborne particulates. When possible, two colloid samples were collected: one was pre-filtered, and one was not pre-filtered. Following the filtration process, the same methods were used for both samples. The following methods were used for the filtration process:

77

1. Between 1 and 3 liters of well water were pre-filtered through quantitative filter paper with a pore size of 16 microns (Whatman 43 Filter Paper) using field portable 150 mL analytical filter units (Nalgene brand 130-4045). This step is omitted for colloidal samples collected without pre-filtering. 2. The water was then filtered through a 0.45 micron cellulose membrane filter (Fisher brand 09-719-1D) using field portable Nalgene filter units. 3. In order to facilitate the filtration processes, a vacuum was pulled on the base of the filter unit both during the prefiltration and the filtration process. 4. The quantity of water filtered was dependent on the quantity of colloids collected as determined by coloration of the filter material and rate of flow of the sample water through the 0.45 micron filter under vacuum. 5. Immediately following completion of the filtration process, dedicated forceps rinsed with the filtered water were used to fold the filter material into quarters. The folded material was then placed in plastic bags that were then sealed, folded, and taped to prevent unnecessary movement of the sample within the bag. All samples were kept in hard-sided plastic containers for return to University of Notre Dame laboratories for analysis.

The sample collection methods for both the dissolved and colloidal phases of uranium necessitate that an individual with a high level of expertise perform the sample collection due to the complexity of the methods. For the dissolved phase, the process of collecting the sample and filtering it is something that our experience suggests could be taught to a local participant. However, the addition of concentrated nitric acid to the

78

sample excludes this possibility as we believe the dangers associated with concentrated acid outweigh the benefits of the high-frequency sampling. For the colloidal phase, the process of collecting the sample is complex and requires a high level of expertise to maintain the required precision and to prevent contamination of the sample. As such, it was necessary, in both instances, to use personnel with high levels of expertise; in this case we used trained graduate and undergraduate researchers actively participating in the research project.

4.2.2.2 Laboratory Methods Upon returning to the University of Notre Dame, all samples were analyzed on the ICP-MS and ICP-OES for uranium and a selection of additional elements following methods established in the regional sampling study that are detailed in section 3.2.1. While the dissolved samples were able to follow these methods explicitly, the colloid samples required additional sample preparation in order to dissolve the filter and collected colloids prior to following the established methods for the dissolved sample. Initial methods were established in the summer of 2006. However, due to significant problems encountered, these methods are under continual development. The initial methods and a brief review of both the problems and ensuing experiments are here included. All sample preparation work has been performed in a class-1000 clean lab. The initial method for the dissolution of the filter and colloids followed this procedure: 1. A filter paper with colloids was removed from the plastic storage bag using sterile Teflon forceps and placed in a sterile Teflon beaker.

79

2. 100 drops of concentrated, double distilled nitric acid was added to the beaker. The beaker was capped and placed on a hot plate at 100°F for 24 hours. 3. The sample was then dried down at 150°F, a process taking approximately 4 hours. 4. Steps 2 and 3 were then repeated two additional times in an effort to ensure complete dissolution. 5. As the final step in sample preparation, 75 drops (approx. 2.5 mL) of concentrated nitric acid was add to the sample, which was then poured into a sterile plastic bottle. 18 mega ohm water was used to rinse the beaker and beaker lid to prevent a loss of sample; the final quantity of the solution was approximately 50 mL; preparation of the solution was done gravimetrically for greater accuracy.

In addition to analyzing the colloid samples, several filters blanks and acid blanks were analyzed. The filter blanks remained in the filter units throughout the field sampling campaign and were placed in plastic bags following the above method at the end of the field sampling campaign. The filter blanks were then submitted to the same dissolution process as all other samples. The acid blanks were created by following the dissolution process in a beaker without the addition of a sample. By analyzing both filter blanks and acid blanks elemental concentrations due to either were accounted for thus eliminating two possible sources of sample contamination. The above process was developed with the help of Dr. Shaefer, Notre Dame’s ICP-MS lab manager, and is based on established methods used for analyzing rock specimens. Unfortunately, a significant number, approximately 80%, of samples

80

developed small black particles during the first dissolution cycle that, in some cases, stuck to the outside of the Teflon beakers, and, in all cases, did not completely dissolve through the remainder of the dissolution process (see Figure 4.6). Further, a significant portion of the colloids collected during the February of 2007 sampling trip did not dissolve using the initial method. In order to address these issues, a series of experiments were conducted.

Figure 4.6: Examples of the black particles that developed during the dissolution process in ~80% of samples using the initial sample preparation method. Photo on left demonstrates a ring of particles attaching to the Teflon beaker. Experiments conducted to modify the colloid sample preparation method to prevent the development of black particles and to dissolve all colloids used the following variables: (i) drying down at a lower temperature, 75°F, (ii) elimination of the dry down process, (iii) use of supersonic nebulizer to break apart particles, and (iv) substituting nitric acid with sulfuric acid, hydrogen peroxide, or hydrofluoric acid. These experiments were conducted with unused filter paper, samples collected locally, and samples collected in Bénin. Only the experiment using hydrofluoric acid was able to prevent the

81

development of black particles and completely dissolve the colloids. In this experiment, 50 drops of nitric acid followed by 50 drops of hydrofluoric acid was added to the samples, the samples were capped after the initial reaction and placed on the hotplate at 150°F for 3 days. The samples were then dried down. The last step of the initial method was then used to prepare the sample for analysis. While this modified method prevented the development of black particles and facilitated the complete dissolution of the colloids from the February 2007 samples, only 4 samples were prepared in this manner. As such, this method serves as a preliminary revised method and requires further testing with additional samples for confirmation of the method.

82

CHAPTER 5

MONTE CARLO STUDIES: METHODS

For this research, Monte Carlo studies were performed in order to numerically evaluate different groundwater monitoring strategies. The Monte Carlo studies were aimed at the assessment of three parameters: mean concentration (MeanC), maximum concentration (MaxC), and total mass load (TML) of a contaminant. Although these parameters are not inclusive of all used in groundwater research, they provide a reasonable set of parameters for initial evaluation of the working concept of this research. Following assessments conducted in surface water systems (e.g., Preston 1989, and Robertson 2003), these studies began with the simulation of a series of hypothetical, high-frequency realizations. In order to study the variability likely to be included in the flow rate and concentration at a pumping well, data sets were generated for each of the scenarios described below that included 100 realizations, each 11 years in length. The length of these data sets enabled the use of different sub-sampling strategies for various durations. These simulations included both volume of pumping, Q(i), and concentration, C(i), at time intervals of 0.25 days. For this study, these simulated data sets represent the population values in the aquifer being sampled. The databases were then sub-sampled using four time-uniform sampling strategies (weekly, biweekly, monthly, and semi-

83

annual), and two time-variable sampling strategies (season dependent and concentration dependent); these sampling strategies represent a reasonable POSS for groundwater sampling. Further, in order to explore the range of analytical methods and expertise as expressed through the POAM and POE, various levels of instrument and operator error (e) and bias (b) were added to the sub-sampled data in order to simulate what might be reasonably experienced in rural regions of developing nations under a range of circumstances. Details of the methods are included in the subsequent sections of this chapter.

5.1 Generating Population Data Sets The data sets used for this Monte Carlo study were created by Silliman (Stephen Silliman, personal communication, 2007) as part of the project that expands beyond the scope of this dissertation; a brief overview of the methods used is here provided. In order to generate the populatoin data at intervals of 0.25 days for both flow rate and concentration of a chemical parameter at a hand-pump well, an Autoregressive, Integrated Moving-Average (ARIMA) model, with temporal drift added, was used to simulate both an autocorrelated series of flow rates, Q(i), and a concentration time series, C(i), including both autocorrelation and correlation with Q(i) (Box and Jenkins 1976, and Wei 1990). The units for this study, which are based on our field work, are L/day for Q and mg/L for C. The integrated portion of the ARIMA prevents the simulated data from demonstrating a persistent bias from the pre-determined mean. The temporal drift provides for simulation of scenarios in which the flow rate at the pump varies seasonally.

84

5.1.1 Five Model Scenarios As stated above, each realization (for each model) ran for a period of eleven years; specifically, each individual realization involves two time series, Q(i) and C(i), i = 1 : 16060 (4 points per day over eleven years). For each model, the population data set was generated through generation of 100 realizations based on the same model parameters. For each data set, six sampling strategies were used to sub-sample from the realizations. There are five different scenarios that were used to create data sets consisting of 100 realizations each. The five scenarios have different characteristics intended to simulate a possible situation that could occur in groundwater situations: •

Scenario 1: Q has a constant mean within the time series, and C is correlated with Q

•

Scenario 2: Q varies seasonally within the time series, and C is correlated with Q

•

Scenario 3: C and Q have constant means within the time series, C is independent of Q

•

Scenario 4: C varies seasonally within the time series, C is independent of Q, and Q has constant mean within the time series

•

Scenario 5: C linearly increases across the time series, C is independent of Q, and Q has a constant mean within the time series

Figures 5.1-5.5 depict portions of one realization from each of the five scenarios as examples of the data sets created.

85

4

x 10

18

1.6

16

1.4

14

1.2

12

1

10

0.8

0.6

C (mg/L)

Q (L/day)

1.8

8

0

50

100

150

200 250 Time (days)

300

350

6 400

Figure 5.1: Scenario 1; Q has a constant mean within the time series, and C is correlated with Q

4

2

x 10

30

20

1

10

C (mg/L)

Q (L/day)

1.5

0.5 0

50

100

150

200 250 Time (days)

300

350

0 400

Figure 5.2: Scenario 2; Q varies seasonally within the time series, and C is correlated with Q

86

4

x 10

20

1

0

10

0

50

100

150

200 250 Time (days)

300

350

C (mg/L)

Q (L/day)

2

0 400

Figure 5.3: Scenario 3; C and Q have constant means within the time series, C is independent of Q

4

2

x 10

30

20

1

10

C (mg/L)

Q (L/day)

1.5

0.5 0

100

200

300

400 500 Time (days)

600

700

Figure 5.4: Scenario 4; C varies seasonally within the time series, C is independent of Q, and Q has constant mean within the time series

87

0 800

4

x 10

20

1

0

10

0

200

400

600 Time (days)

800

1000

C (mg/L)

Q (L/day)

2

0 1200

Figure 5.5: Scenario 5; C linearly increases across the time series, C is independent of Q, and Q has a constant mean within the time series

5.1.2 Assessment of Study Parameters Based on these data sets, the impacts of the POSS, POE, and POAM on the MeanC, MaxC and TML were assessed. The population values of these three parameters were determined for a period of T days (starting at sample point j within a realization and extending to sample point j + 4*T in that same realization) from the relationships: j + 4*T

∑ C (i)

MeanC =

i= j

(5.1)

4 *T

j + 4*T

MaxC = Max (C (i ))

(5.2)

i= j

j + 4*T

TML =

∑ Q(i)C (i) * 0.25

(5.3)

i= j

T / 365

88

The averages for each of these parameters for the 100 realizations of each of the five data sets for the full 11 years are found in Table 5.1.

TABLE 5.1 CALCULATED “KNOWN” VALUES FOR ASSESSMENT PARAMETERS FOR THE FIVE DATA SETS Data Set Mean Concentration Max Concentration TML

mean variance mean variance mean variance

1 10.0 0.2 20.3 1.1 4.17E+07 3.33E+12

2 10.5 0.4 45.5 9.4 4.53E+07 1.02E+13

3 10.0 0.0 23.6 1.4 4.13E+07 1.03E+12

4 10.1 0.2 28.4 2.6 4.16E+07 3.03E+12

5 15.0 0.0 34.2 1.3 6.20E+07 7.68E+11

5.2 Simulating “Sampled Data” The time series, or population data sets, were sub-sampled in order to simulate various points on the POE, POSS, and POAM (including variable sampling strategy and precision / bias of instrument). Here, the methods used to simulate sampling interval and sampling precision / bias are discussed.

5.2.1 Definitions of Sampling Periods Sampling periods were simulated simply by sub-sampling at appropriate intervals and for appropriate lengths from each realization. For example, in order to simulate weekly sampling, the time series provided by a given realization was sub-sampled at intervals of N=28; in other words, the series was sub-sampled at i and then at i+28 and then at i + 2* 28, etc. Three time-uniform sampling periods were used: weekly, biweekly,

89

monthly, and semi-annual. Two time-variable sampling periods were used: season dependent and parameter dependent. In order automate the sub-sampling process, operational definitions of each sampling periods were created (some are obvious and some are compromises to facilitate the numerical effort): •

A week is 7 days.

•

Biweekly is 2 weeks (14 days).

•

A month is 4 weeks (28 days); there are therefore 13 months in our simulated year.

•

Semi-annual is 182.5 days.

•

Season dependent sampling divides the year into 2 approximately equal seasons during which one season uses weekly sampling, and the other, monthly sampling. o Weekly sampling is performed during the simulated dry season during which the pumping rate is greater. In each year, there are 24 weeks of weekly sampling. o Monthly sampling is performed during the simulated rainy season during which the pumping rate is lower. In each year, there are 7 months (28 weeks) of monthly sampling.

•

Parameter dependent sampling frequency is dependent upon the concentration of the measure being sampled. If the concentration of a sample has increased by more than a critical value, the next sample will occur 1 week later. In all other circumstances (including significant decrease in concentration), the next sample will occur one month later. In this research, the critical value is defined as an increase in concentration of 4 mg/L.

90

5.2.2 Definition of Sampling Length Sampling lengths were simulated at 1, 2, 5, and 10 years. These lengths were chosen to represent short, medium, and long-term monitoring projects. The first term to be sub-sampled in any given realization was never less than 60 days (i.e., 240 points) after the start of the realization as the first portion of a data realization has the potential to be uncharacteristic of the series as a whole. The first and last terms in the sub-sampling were determined based on the desired length of the time series in terms of years; for example, an annual sampling set based on weekly sampling would include only those sub-samples in the interval 240 < i + sampling length*365*4 < 16,060.

5.2.3 Sub-sampling For each of the five scenarios, each of the 100 realizations was sub-sampled 100 times (10,000 total sample sets) for each sample length and each instrument case (detailed below). For sampling lengths of 1, 2, and 5 years, the first sample was randomly selected to occur no earlier than 60 days into the realization, but early enough within the realization to ensure that the full sampling length was available. For example, the starting points for the 100 sub-samples for a sampling length of 2 years were randomly selected between the data points of 240 and 16,060 - 2*365*4. In contrast, the 10 year sample series consisted of fewer than 100 realizations (due to limits on sampling of the data set without repetition). Specifically, these series were generated starting at point 241 and repeated with starting points at as many successive points as possible without resampling. (E.g., The first sample series would

91

start at point 241, the second at point 242, and the third at point 243.) Specifically, the following was done for each sampling period: •

Weekly sampling: starting at point 241, sample series were generated 28 different times with each sample series starting at the proceeding data point (241 through 268).

•

Biweekly sampling: starting at point 241, sample series were generated 56 different times with each sample series starting at the proceeding data point (241 through 296).

•

Monthly sampling: starting at point 241, sample series were generated 100 different times with each sample series starting at the proceeding data point (241 through 340).

•

Semi-annual sampling: sample series were generated 100 different times with each sample series starting at a different point between points 241 and 1460 (first point of day 365).

•

Season dependent sampling: starting at point 241, sample series were generated 28 different times with each sample series starting at the proceeding data point.

•

Parameter dependent sampling: starting at point 241, sample series were generated 56 different times with each sample series starting at the proceeding data point.

5.2.4 Adding Sampling Errors In order to include instrument and operator errors present in simulated analytical methods, error was added to each sampled C(i) during sub-sampling. For this research, instrument error includes instrument error and bias, e and b, herein defined as:

92

•

The instrument error is herein considered to be normally distributed with a standard deviation equal to “e”; instrument error here includes both error introduced by the sampling device and operator error. e is represented by a percent of parameter concentration (that is, the standard deviation of instrument error is a linear function of the concentration).

•

The instrument bias is symbolized as “b”; as with instrument error, instrument bias includes bias in the sampling device and bias introduced by the operator. Bias is represented by a fixed error in concentration that is independent of parameter concentration.

During the sub-sampling, bias was added to the original simulated value, C(i), and e was simulated to be normally distributed with the appropriate standard deviation. (i.e., C*(i) = C(i) + b + e*N(0,1), where C* is the recorded sample value used in assessing the parameters). In order to perform simulations that cover a range of e and b that would likely be experienced with equipment used in field conditions in developing nations and the range of expertise available, the nitrate standards analyzed by local participants using nitrate colorimetry in the Bénin case study were analyzed and used as guides in selecting both b and e. These data suggest that, with a well trained local population, e and b for the nitrate colorimeter are likely to be in the range of 20% and 4 mg/L, respectively. For the same instrument, using trained local participants with less experience, our data suggest that e and b can be as large as 60% and 5.5 mg/L. For this research, five cases were chosen. Table 5.2 provides the precision (standard deviation, e) and bias (b) introduced in order to simulate the precision and bias

93

anticipated from instrumentation similar to the nitrate colorimetry (first case), substantially higher quality data than we have experienced (second case), lower quality data than we have experienced (third case), and a mixed quality case with e similar to our experience and b better than our experience (fourth case). In addition to the nitrate colorimetry, local participants in the Bénin case study used test strips to analyze nitrate. As test strips are available for a wide range of analytes, and are the least complicated analytical instrument on the POAM, they are assessed as the fifth case used in this simulation. To simulate test strips, a sampled value was first determined from the realization and then classified within an appropriate, discrete concentration range. That discrete range was then recorded in three ways: the bottom of the range (TS1), the mean value of the range (TS2), and the high end of the range (TS3). Thus, three complete data sets are generated and used to assess the parameters. For example, if the sampled value was 6 mg/L, and fell into the range of 5 to 10 mg/L, TS1 would be 5 mg/L, TS2 would be 7.5 mg/L, and TS3 would be 10 mg/L. In this study, the discrete intervals used for the test strip case were based on the test strips used in the case study; specifically, they were: 0, 0.5, 2, 5, 10, 20, and 50 mg/L.

TABLE 5.2 e AND b IN THE FIVE INSTRUMEN CASES USED IN THE MONTE CARLO STUDIES Case 1 (nitrate colorimetry) Bias (mg/L) 4 Precision (%) 20%

Case 2 (higher quality data) 0.4 5%

94

Case 3 (lower quality data) 8 60%

Case 4 (mix quality data) 0.4 20%

Case 5 (test strips) 0 0%

5.2.5 Notation Throughout the remainder of this dissertation, notation has been created in order to easily refer to the different sampling periods, lengths of sampling period, and the instruments simulated in the Monte Carlo studies. This notation applies a symbol to each sampling period and a subscript for the length of the sampling period (Table 5.3). For example, a weekly sampling period for 1 year would be W1, and a semi-annual sampling period for lengths of 1 year, 2 years, and 5 years are denoted as SA1, SA2, and SA5. Reference to instrument characteristics lists the instrument precision, e, and then bias, b.. For example an instrument for which e is 5% and b is 0.4 mg/L would be 5%/0.4. Due to the discrete nature of test strips, no e or b are added to the sample value. Hence, the only specification on these results is whether the value reported is bottom (TS1), middle (TS2), or top (TS3) of the concentration range containing the sample value.

TABLE 5.3 SYMBOLS REPRESENTING SAMPLING FREQUENCIES OVER A PERIOD OF N YEARS Symbol

Sampling Frequency

WN

Weekly

BN

Biweekly

MN

Monthly

SAN

Semi-annual

SDN

Season dependent

PDN

Parameter dependent

B

95

5.3 Evaluation of Sampled Data 5.3.1 Calculation of Parameters of Interest MeanC, MaxC, and TML were estimated for each of the sub-sampled data sets using adjusted forms of the formulas as given above for the population value. For both MeanC* and MaxC*, this is a straightforward calculation of the mean or maximum of the sub-sampled data (equations 5.4 and 5.5). Table 5.4 further details parameter coefficients that change with sampling period.

N

MeanC * =

∑ c * (i ) i =i

(5.4)

n N

MaxC * = Max (c * (i ))

(5.5)

i =1

⎡ nm c *ijm ⎤ q ∑ ∑ jm ⎢∑ ⎥ m =1 j =1 i =1 n m ⎦ ⎣ TML1 = P

Y

TML 2 = c* c*ijm i j ly m nm N Nm P qjm qy Qy Y

Nm

⎛ly

∑ ⎜⎜ y =1

⎝

⎞ ⎟Q qy ⎟ y ⎠

Y

(5.7)

Y

concentration (concentration plus instrument error and bias) concentration at sample point i, day j, and period m point sampled day in time interval average load sampling period (month or half year) number of points sampled in sampling period m total number of points in the sub-sample total number of days in sample period m number of sampling periods in a year sampled flow for day j in sample period m average flow for year y total flow for year y total number of years

96

(5.6)

TABLE 5.4 COEFFICIENTS OF PARAMTERS THAT CHANGE WITH SAMPLING PERIOD Time-uniform sampling periods

Coefficient

Time-variable sampling periods Season Parameter dependent dependent month month

Weekly

Biweekly

Monthly

Semi-annual

m

month

month

month

half year

nm

4

2

1

1

1 or 4

dependent

N

52 * Y

26 * Y

13 * Y

2*Y

31 * Y

dependent

Nm

28

28

28

182.5

28

28

For TML, a wide array of estimators have been evaluated in the surface water literature (e.g., Robertson 2003, and Preston et al. 1989). These papers suggest specific estimators for streams with different flow characteristics (e.g., streams with slow or fast responses to precipitation events) or when using specific sampling strategies. As this research used scenarios with different characteristics for Q and C as well as numerous sampling periods and lengths, it was impossible to select a single estimator that would best fit all situations. Rather, two estimators, a monthly averaging estimator (TML1, equation 5.6), and ratio estimator (TML2, equation 5.7) were adapted from Preston et al. (1989). These estimators were chosen because they use daily flow with concentration averages, which Preston suggests provides lower error when daily concentrations measurements are not available. The availability of measurements of pump flow in rural regions of developing country is a reasonable assumption based on our experience in Bénin as the pump water is sold at a cost per quantity with the total daily income recorded. From these records of daily income, daily flow can be extrapolated. However, daily measurement of concentration is not a reasonable requirement at a hand-pump in a

97

rural region. Hence, this scenario fits reasonably to the conditions discussed by Preston. For TML1, which uses monthly averages of concentration, equation 5.6 was modified for sampling periods where sampling was performed monthly or semi-annually. In these circumstances, no monthly average concentration can be determined. Rather, the concentration recorded at the single sub-sample point is used with the appropriate length of flow rates. For example, if semi-annual sampling was performed, the single sample concentration would be coupled with a half year of flow rates. In sampling periods with time variable sampling (i.e., season dependent and parameter dependent), the equation was modified to be appropriate for the specific number of sub-samples taken in a given month.

5.3.2 Calculating Parameter Error To enable comparison among the cases and across sampling periods and sampling lengths, MeanC*, MaxC*, TML1 and TML2 were each compared with their population values via the mean squared error (MSE, following Preston et al 1989, Roberston 2003, and Moosman et al. 2005). The information contained in the MSE was separated into squared bias and variance (or precision) in order to determine the dominant term in the MSE. Further, average bias was calculated to determine if the bias was positive or negative as this information is lost when calculating the square bias. For this research, the statistics calculated are herein defined: •

Precision is the variance of the sub-sample (with a given sampling period, length of sampling period, and instrument characteristics) around the population value

98

of the parameter of interest (i.e., MeanC, MaxC or TML) for the particular subsample length. •

Bias is the average of the difference between the sub-sample (with a given sampling period, length of sampling period, and instrument characteristics) and the population value for the particular sub-sample length.

•

Bias squared (bias2) is the average of the squared distance of the sub-sample (with a given sampling period, length of sampling period, and instrument characteristics) from the population value for the particular sub-sample length.

•

Mean square error (MSE) is the precision plus the squared bias for the particular sub-sample.

As noted above, each of the above statistics was calculated using 100 sub-samples of the 100 realizations of each scenario. Therefore, the sample size used in each of these calculations is 10,000. The exception for this is the 10 year samples; details of the sample size for the different sampling periods for this sampling length were outlined in section 5.2. Analysis of the results (raw data provided in Appendix 1) was used to determine the effect of sampling periods, sampling length, and the e and b of analytical methods on the ability to reproduce the known mean concentration, maximum concentration, and total mass load, and can be found in Chapter 6.

99

CHAPTER 6

RESULTS AND ANALYSIS

This chapter is divided into two sections. In the first, summaries of the results obtained from the nitrate case study, the uranium case study, and the Monte Carlo studies are presented. The goal of these summaries is to provide an overview of the data available for analysis. The second portion of the chapter then builds upon these three sets of results to complete our analysis of the working concept.

6.1 Results 6.1.1 Case Study: Nitrate As indicated in Chapter 4, the primary focus of the analysis reported in this dissertation (for nitrate) are data gathered in Adourékoman. The second level implementation of the high-frequency sampling and associated low-frequency sampling, while important to the overall Bénin project, is not central to this dissertation and is therefore not discussed herein. Given the focus of this case study on data gathered from the village of Adourékoman, a brief review of the layout of the village is here provided. The village has three drilled, hand pump wells, three open, or functioning, hand dug wells, and two filled-in, or no longer open / functioning, hand dug wells. The three hand pump wells are

100

named by the region of the village in which they are found: Ayewa-Okouta (N 7º 55.142’, E 2º 16.300’), Ayewa (N 7º 54.931, E 2º 16.508’), and Agbo (N 7º 54.814’, E 2º 16.404’). A hand drawn and satellite image of the village are provided in Figure 6.1. N

(b)

+ MC C W1 W2 W3 FHD HD

(a)

KEY Church Medical Clinic Cemetery Well Ayewa-Okouta Well Ayewa Well Agbo Filled-in Hand Dug Well Hand Dug Well Road Main walking path Main housing area Bush

Figure 6.1: (a) Hand drawn map of Adourékoman as represented by local population. Approximate width of map is 600m. (b) Spot image of Adourékoman as provided by DH. The nitrate sampling was divided between low-frequency and high-frequency sampling. The low-frequency sampling consisted of review of historical nitrate data from Direction de l’Hydraulique and characterization of the nitrogen and oxygen isotopes of

101

groundwater nitrate in the three wells in Adourékoman (Figure 6.2, this is a subset of the isotope data shown in Figure 4.4). The two time points were selected to enable testing during both the dry (February 2007) and rainy (May 2007) seasons. In addition to these two sampling periods, well Agbo was sampled during the 2003 isotope sampling campaign and is presented as a green diamond in the figure.

20 18 2-2007 Aywea-Okouta 5-2007 Ayewa-Okouta 2-2007 Ayewa 5-2007 Ayewa 6-2003 Agbo 2-2007 Agbo 5-2007 Agbo

16 14

18

δ O

12 10 8 6 4 2 0 0

10

20

30

40

15

δ N

Figure 6.2: N and O isotopes of nitrate from the three wells in Adourékoman. Dates of samples as indicated. The line shown in the figure has a slope of ~0.5, indicative of behavior consistent with denitrification. Figure 6.2 demonstrates isotopic consistency over time within each well with the three wells demonstrating three different isotopic signatures (with Ayewa being substantially different than the other two). As is further discussed in the analysis, it is noted that, although these three wells are located within one relatively small village, they demonstrate clear isotopic differentiation. Additional observations of these data indicate that they fall along the same trend as observed in the larger isotopic data set (Figure 4.4),

102

leading to inferences regarding both potential sources and denitrification of the nitrate in this region (see analysis section for details). The second portion of the nitrate data involved high-frequency sampling by the local population using test strips and a single-parameter nitrate colorimeter. As discussed in the Chapter 4, the nitrate standards analyzed on a monthly basis using the colorimeter indicate that there is a bias in the data, but demonstrate no temporal trends suggesting that any temporal trends in the data are real and not a function of the sampling instrument / operator. In addition to analyzing for nitrate, these sampling groups used test strips to analyze for a small group of additional parameters. Although discussion of these data is beyond the scope of this dissertation, the data can be found in Appendix 2. The high-frequency nitrate data for each well are displayed in Figures 6.3 – 6.5. In viewing each of these figures, it is noted that the test strip data recorded as a range are reported as the mean of the range while the colorimeter data were recorded as singular values (e.g., 5 mg/L NO3-N) are represented as the value, not a range. The discrete ranges of the test strip are: 0.5, 2, 5, 10, 20, and 50 mg/L NO3-N. In contrast, the colorimeter data involve a continuous scale with a maximum of 33.0 mg/L NO3-N. The monitoring groups did not dilute samples if they were at the top of range for the specific analytical instrument. As such, test strip readings of 50 mg/L NO3-N and colorimeter readings of 33.0 mg/L NO3-N represent ‘above range’ for their respective instruments.

103

50.0 45.0

Nitrate (mg/L NO3-N)

40.0 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 28-May05

5-Sep-05 14-Dec-05 24-Mar-06 2-Jul-06 10-Oct-06 18-Jan-07 28-Apr-07 6-Aug-07 Date Colorimeter

Test Strips

Figure 6.3: Nitrate data from well Ayewa-Okouta in Adourékoman as collected by the local sampling group. 50 45

Nitrate (mg/L NO3-N)

40 35 30 25 20 15 10 5 0 14-Jan-04

1-Aug-04

17-Feb-05

5-Sep-05

24-Mar-06

10-Oct-06

28-Apr-07

Date Colorimeter

Test Strips

Figure 6.4: Nitrate data from well Agbo in Adourékoman as collected by the local sampling group.

104

14-Nov-07

20 18

Nitrate (mg/L NO3-N)

16 14 12 10 8 6 4 2 0 23-Apr- 1-Aug04 04

9-Nov- 17-Feb- 28-May- 5-Sep- 14-Dec- 24-Mar- 2-Jul-06 10-Oct- 18-Jan04 05 05 05 05 06 06 07 Date Colorimeter

Test Strips

Figure 6.5: Nitrate data from well Ayewa in Adourékoman as collected by the local sampling group. Figure 6.3 displays the data for well Ayewa. Two observations are made regarding preliminary assessment of these data. First, while random variability in the nitrate data are apparent in the colorimeter data, no temporal trend is observed. Further, the concentrations measured with the test strips are consistently higher than those collected with the colorimeter. Figure 6.4 displays the data for well Agbo. It is noted that the colorimeter data demonstrate a significant increase between the spring and summer of 2006. Again, the test strip data has significantly higher concentrations than the colorimeter. However, it should be noted that, beginning in the summer of 2006, both the colorimeter and test strip results regularly exceed the range of the measurement device. Figure 6.5 displays the nitrate data for well Ayewa. For this well, the data collected using both the colorimeter and test strips are consistent (with the exception of

105

two data points – it is later suggested that these two points may have been recorded erroneously). If the colorimeter data are plotted alone, therefore decreasing the span of the y-scale, display of data is expanded as shown in Figure 6.6. This figure demonstrates consistent variation in the nitrate concentration in this well, with the suggestion of cyclical variation.

. 7

Nitrate (mg/L NO3-N)

6 5 4 3 2 1 0 23-Apr- 1-Aug04 04

9-Nov- 17-Feb- 28-May- 5-Sep- 14-Dec- 24-Mar- 2-Jul-06 10-Oct- 18-Jan04 05 05 05 05 06 06 07 Date

Figure 6.6: Nitrate data from well Ayewa in Adourékoman as collected by the local sampling group using a colorimeter.

106

6.1.2 Case Study: Uranium As indicated in Chapter 4, the uranium research is a recent extension of the case study. For the purposes of this dissertation, the primary point of interest is the illustration of the use of another set of the three progressions. Although the ongoing work associated with the elevated levels of uranium is a significant aspect of the Bénin project, only the preliminary results, those collected during the summer of 2006, are included herein. While it was hoped that relatively simple methods would be identified for field measurement of uranium or a surrogate compound, the lack of less complex methods resulted in our sampling method for the dissolved phase of uranium requiring the addition of concentrated acid to the water sample for preservation and return of the sample to Notre Dame for analysis (see section 4.2.2.1). Due to the dangers associated with concentrated acid, we believed it inappropriate to involve local personnel with lower levels of expertise in such sampling processes; all sampling for the dissolved phase was completed by ND personnel. Similarly, for the sampling of the colloidal phase, the methods required high levels of expertise to reliably perform the method, in part because sampling personnel were required to make decisions in the field concerning the final total quantity of water sampled (between 1 and 3 liters) based on the quantity of colloids collected. Such decisions represent a level of expertise that would likely be difficult to teach a non-expert in the field; again, all sampling for uranium was conducted by ND personnel. Locations sampled as part of the uranium research during the summer of 2006 were selected for one of two reasons: (i) the location showed high concentrations of dissolved uranium during the earlier regional exploratory research, or (ii) the location

107

was part of the ongoing nitrate research which occurred in the same region. This allowed for hand pumps historically known to have high uranium in the groundwater as well as additional hand pumps in the region to be included in the study. The results of the dissolved and colloidal phases of uranium as collected during the summer of 2006 are found in Table 6.1. The uranium results are expressed in terms of μg/L relative to the original water from which they were collected; this is different than the notation used in Chapter 3. In Table 6.1, the samples are grouped by well and then by date to facilitate the comparison of the two uranium phases for each location and the comparison of variation in concentration over time at a location. When possible, the name associated with the hand pump well that was sampled is provided in order to facilitate discussion of the samples. Finally, a table with the additional elemental analyses performed on the ICP-MS and ICP-OES for each sample is found in Appendix 3.

108

TABLE 6.1 CONCENTRATIONS OF URANIUM IN THE DISSOLVED AND COLLOIDAL PHASES OF SAMPLES COLLECTED DURING THE SUMMER OF 2006. THE PHASE OF EACH SAMPLE IS INDICATED IN THE TALBE: DISSOLVED (D) OR COLLOIDAL (C); THE COLLOIDAL PHASE IS FURTHER DIVIDED INTO UNFILTERED (U) OR FILTERED (F). “ND” 109

INDICATES THE CONCENTRATION OF WAS NOT DETECTED BY THE ANALYTICAL INSTRUMENT USED (ICP-MS).

109

TABLE 6.1 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF Uranium (μg/L)

13-Jun-06 C UF 0.0157

Adourékoman: Agbo 7.9052 2.2734 4-Jul17-Jun-06 11-Jul-06 06 D C C D C F F F 30.8 ND ND 27.2 ND

Well Name Latitude 110

Longitude Date Phase: D or C C: F or UF Uranium (μg/L)

17-Jun-96 D C F 19.0 0.0150

Adourékoman: Ayewa-Okouta 7.919 2.2717 3-Jul-06 5-Jul-06 12-Jul-06 D C D C D C C F F F UF 0.0163 25.4 20.0 0.0032 19.02 0.0083 0.0149 Adourékoman: Ayewa 7.9155 2.2751

Well Name Latitude Longitude Date Phase: D or C C: F or UF Uranium (μg/L)

C UF ND

17-Jun-06 D C F 30.6 0.0103

3-Jul06 C F 0.0131

110

11-Jul-06 D C C F UF 25.9 0.0023 0.0025

18-Jul-06 D C C F UF 31.8 ND ND

TABLE 6.1 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF Uranium (μg/L)

Kpakpazoume: A 7.9278 2.2492 D 261

14-Jun-06 C F 0.231

C UF 1.29

D 240

13-Jul-06 C F 0.159

C UF 1.39

Kpakpazoume: B 7.9278 2.2492 14-Jun-06 14-Jul-06 D C D C C UF F UF 127 0.0113 83.3 0.0172 0.0170

Date Phase: D or C C: F or UF Uranium (μg/L)

Sowe: 2 7.9757 2.1676 24-Jun-06 D C F 0.0111 4.69

Sowe: 1 7.978 2.1659 25-Jun-06 D C F 0.0113 2.31

Agouagon: A 7.9803 2.2932 25-Jun-06 D C F 0.0114 2.25

Agouagon: B Agouagon: C 7.9793 7.9792 2.2987 2.296 1-Jul-06 29-Jun-06 D C D C F F 0.0105 0.0124 0.28 1.95

Well Name Latitude Longitude Date Phase: D or C C: F or UF Uranium (μg/L)

Moumoudji:A 7.796 2.1756 7-Jul-06 16-Jul-06 D C D C F F 0.0148 4.83 10.8 0.0204

Moumoudji:B 7.8004 2.1814 7-Jul-06 D C F 0.0101 2.67

Moumoudji: C 7.7999 2.1762 2-Jul-06 15-Jul-06 D C D C C F F UF 6.54 0.0080 8.01 0.0100 0.0107

Well Name Latitude Longitude 111

111

Mahou: 1 7.8257 2.1129 2-Jul-06 D C F 1.94 0.0076

TABLE 6.1 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF Uranium (μg/L)

Unknown 1 7.8986 2.2579 14-Jun-06 D C F 20.8 0.0006

D 0.32

Unknown 2 7.759 2.3242 15-Jun-06 C C F UF 0.0260 0.0237

112 112

Unknown 3 7.7824 2.2622 15-Jun-06 D C C F UF 2.01 ND ND

Unknown 4 7.8335 2.2776 15-Jun-06 D C F 0.51 ND

Unknown 5 7.8979 2.3208 16-Jun-06 D C F 27.6 0.0404

6.1.3 Monte Carlo Studies The results of the Monte Carlo (MC) studies include those for the parameters MeanC, MaxC, and TML. As noted in Chapter 5, the MC studies allowed comparison of the impact of different sampling designs (sampling period and choice of instrument) on these parameters. Hence, the results are presented as a series of comparisons of different sampling designs. Due to the complexity and wide array of results obtained from these studies, it is impossible to present a complete discussion of all of the insights gained with respect to sampling. Rather, the results are summarized below as a series of direct comparisons of pairs of sampling strategies. These results serve as examples of the complete results, which are provided in Appendix 1. While these results initially appear somewhat disjointed, the synthesis of these results as provided in sections 6.3 and 6.4 illustrates some of the important features of the MC studies. As described in section 5.2.5, notation has been created to aid in the discussion of the results of the MC studies. A brief review of the symbols representing sampling frequencies and sampling lengths is found in Table 6.2.

TABLE 6.2 SYMBOLS REPRESENTING SAMPLING PERIODS OVER A PERIOD OF N YEARS Symbol

Sampling Frequency

WN

Weekly

BN

Biweekly

MN

Monthly

SAN

Semi-annual

B

SDN

Season dependent

PDN

Parameter dependent

113

Representative results from the Monte Carlo studies of the estimate of MeanC, MaxC, and TML are herein provided without generalization or further analysis (prior to section 6.3) within 5 general categories including: general observations, impact of sampling period, impact of sampling instrument, comparison of test strips with continuous scale instruments, and variation across scenarios.

Summary of Results for Mean Concentration •

General Observations: o For nearly all sampling periods, lengths of sampling periods, and instruments, the MSE is dominated by the bias squared term rather than precision. The occasional exception to this is occurred when the instrument b was very low (i.e., instruments 5%/0.4 and 20%/0.4). o For all sampling periods, precision increases as the length of sampling period increases. No trend with length of sampling period was identified for bias2.

•

Impact of Sampling Period: o Weekly versus biweekly and monthly: Comparing W1 through W10 to B1 through B10 and M1 through M10, using the same instruments led to the observation that all three frequencies result in approximately equal bias, bias2, and MSE; however, the precision obtained from biweekly or monthly sampling frequencies, even for the longest sampling period (10 years)

114

remained lower than the precision obtained after only one year using weekly sampling (W1). o Monthly versus Semi-annual: In comparing the semiannual (SA) and monthly (M) sampling periods, it is found that the MSE of semiannual sampling periods for lengths of five years or less is generally greater than the MSE of monthly sampling for a length of one year (M1), and the MSE of SA10 is approximately equal to the MSE of M1; however, the precision of SA for all periods was generally worse than the precision of M1. o Variable frequency sampling: The statistical results for the SD1 through SD10 sampling lengths are generally similar to those for the B1 through B10 respective of the instrument used. The statistical results of the PD1 through PD10 are generally similar to those for M1 through M10 for the same instrument. •

Impact of Sampling Instrument: o Instrument behavior for SA: Examination of the semiannual sampling period (SA1 through SA10) showed that the precision is approximately the same for instruments 5%/0.4, 20%/0.4 and 20%/4. o Comparing the statistical results of instrument 20%/0.4 using W1 to instrument 5%/0.4 using sampling periods and lengths M1 through M10, it is found that while instrument 20%/0.4 using W1 provides higher MSE’s as compared to M1 or M2, it provides approximately equal results to 5%/0.4 using M5 or M10 depending on the specific scenario.

115

•

Comparison of Test Strips with Continuous Scale Instruments: o Weekly sampling with test strips versus semi-annual sampling with a high quality continuous scale instrument: In examining the statistical results of test strips using weekly sampling with instruments 20%/0.4 and 5%/0.4 using semi-annual sampling, the following observations are identified: ƒ

The MSE of test strips using W1 are approximately equal to instruments 20%/0.4 and 5%/0.4 using SA2 or SA5 depending on the specific scenario.

ƒ

The precision of test strips using W1 is better than instruments 20%/0.4 and 5%/0.4 for all semi-annual sampling periods (i.e., SA1,2,5,10).

ƒ

The exception to these results is scenario 5, the only scenario for which there is a continuous temporal trend in the mean concentration.

o Test strips versus low-quality continuous scale instrument for the same sampling period: The MSE and precision for test strips is better than that for instrument 60%/8. For example, the MSE and precision for test strips using W1 is better than the MSE and precision for instrument 60%/8 using W1. The exception to this is scenario 5 for which this statement remains true for precision, but better MSE is provided by instrument 60%/8. o Test strips against different instruments: Comparing test strips to the range of instruments (i.e., 5%/0.4, 20%/0.4, 20%/4 and 60%/8) within a given sampling period and length of sampling period, the MSE for test strips is generally lower than the MSE for instruments 20%/4 and 60%/8. The test strips generally have higher MSE than instruments 20%/0.4 and 5%/0.4. Comparison of the relative level of precision is significantly more variable.

116

•

Variation Across Scenarios: o Comparing similar instruments, sampling period, and length of sampling period across the five scenarios: ƒ

Scenario 2 (Q varies seasonally and C is correlated with Q) generally had lower precision than all other scenarios.

ƒ

Except for the following situations, no trends were identified in MSE or bias2. For seasonally dependent sampling period, scenarios 2 and 4 had worse MSE, bias2, and precision than the other three scenarios; again, scenario 2 had the lowest precision as well as the highest MSE and bias2. For scenario 5, the MSE and bias2 increase significantly and the precision decreases significantly for monthly and parameter sampling frequencies at 10 years length of sampling.

o Comparing the statistical results of instrument 20%/0.4 using W1 to instrument 5%/0.4 using frequencies M1 through M10, the following is found: ƒ

For scenarios 2 and 4, in which concentration varies seasonally, instrument 20%/0.4 using W1 provides approximately equal results as instrument 5%/0.4 used M5.

ƒ

For scenarios 1, 3 and 5, for which concentration either has a constant mean or increases across the entire time series, instrument 20%/0.4 using W1 provides results that are equal to, or better than, the results obtained with instrument 5%/0.4 using the longest sampling period (M10)..

117

Parameter: Max Concentration •

General Observations: o For nearly all sampling periods, lengths of sampling periods, and instruments, the MSE is dominated by bias2. The occasional exception to this occurred when the instrument b was very low (i.e., instruments 5%/0.4 and 20%/0.4). o Precision varies but does not trend (i.e., consistent increase or decrease) with increase in length of sampling period. The exception to this is test strips, which have significantly better precision at 10 years length of sampling period than any other length of sampling period. No trend was identified for bias2.

•

Impact of Sampling Frequency: o Weekly versus monthly sampling: Comparing the results of W1 to M1 for the same instrument, the M1 statistical results are generally better for instruments 20%/0.4, 20%/4, and 60%/8; that is, less frequent sampling provided better results for all but the best instrument. The exception to this is scenario 2. o Weakness in semiannual sampling: Statistical results for semiannual sampling (SA) vary greatly with increased length of sampling period (i.e., from 1 year to 2 years to 5 years to 10 years), and, across all instruments, have very large errors.

•

Impact of Sampling Instrument: o The instrument 60%/8 performed very poorly (i.e., high MSE, bias, bias2, and precision) across all sampling periods, lengths of sampling periods, and scenarios.

118

•

Comparison of Test Strips with Continuous Scale Instruments: o Comparison of results obtained from test strips against results obtained from continuous scale instruments (for the same sampling period and length) is dependent on the specific scenario studies. ƒ

For scenarios 1 and 3 (C has a constant mean across the time series in these scenarios), test strips provide similar precision and MSE to the instrument 20%/4.

ƒ

For scenario 5 (concentration constantly increasing across the time series), test strips provide similar statistical results in terms of MSE and precision to those obtained with instrument 20%/0.4 for the W, B and SD sampling frequencies.

ƒ

For scenario 4 (C varies seasonally and Q has a constant mean), using weekly sampling period, test strips provide results, in terms of MSE and precision, which fall between those obtained from the instrument 20%/4 and those obtained from instrument 20%/0.4 for periods W2,5,10.

ƒ

For all other sampling periods and lengths of sampling periods for scenario 4 as well as for scenario 2 (Q varies seasonally and C is correlated to Q), test strips provided results similar to, or worse than, those obtained using instrument 60%/8.

119

•

Variation Across Scenarios: o Comparing similar instruments, sampling period, and length of sampling period across the five scenarios: ƒ

For weekly and biweekly sampling periods, scenario 2 had the lowest precision. For all other sampling periods, scenario 2 had the lowest precision as well as the highest MSE and bias2.

ƒ

For monthly, semi-annual and parameter dependent sampling periods, scenarios 4 and 5 trend together in terms of MSE, bias2 and precision. For these scenarios and sampling periods, the statistical results for scenarios 4 and 5 are better than scenario 2, but not as good as those for scenarios 1 and 3.

Parameter: Total Mass Load •

General Observations: o The MSE for all sampling periods (W, B, M, SA, SD, and PC) decreased with increased length of sampling period for lengths from 1 through 5 years (Figure 6.7a). o The MSE for the 10 year length for all sampling periods (i.e., W10, B10, M10, SA10, SD10, and PD10), increased significantly as compared to all other periods. It is suspected that this was due to the design of the Monte Carlo study. A brief study was performed in which a 100 year (versus 11 year) data set was created with 100 realizations, thus allowing random start times for the ten year sub-samples. The characteristics of this study were the same as those

120

1.00E+18 1.00E+17

miligrams

2

1.00E+16 1.00E+15 1.00E+14 1.00E+13 1.00E+12 1.00E+11 0

2

4

6

8

10

12

Length of Sampling Period (years) Precision: Weekly, TS2 MSE: Weekly, TS2

Precision: Semi-annual, 5%/0.4 MSE: Semi-annual, 5%/0.4

(a)

1.00E+18 1.00E+17

miligrams

2

1.00E+16 1.00E+15 1.00E+14 1.00E+13 1.00E+12 1.00E+11 0

2

4

6

8

10

12

Length of Sampling Period (years) Precision: Weekly, TS2 MSE: Weekly, TS2

Precision: Semi-annual, 5%/0.4 MSE: Semi-annual, 5%/0.4

(b)

Figure 6.7: MSE and precision for the TML of scenario 3. (a) Original results, and (b) 10 year length substituted

121

used for scenario 3; C and Q both have constant means and are not correlated. The results (Figure 6.7b) suggest that the MSE for the TML10 in the original simulations would have continued to decline as compared to the results for the TML5 if the length of the simulated “reality” had allowed sub-sampling of independent, ten-year sequences. As such, the TML10 results are considered to be biased by our method and are therefore not discussed in the remainder of this dissertation. o For nearly all sampling periods, lengths of sampling period, and instruments, the MSE is dominated by bias2 rather than precision. The occasional exception to this occurred when the instrument b was very low (i.e., instruments 5%/0.4 and 20%/0.4), and even then typically only for 1 and 2 year lengths or for any length of sampling period with the semiannual sampling (i.e., SA1,2,5,10). o Examining the two ways to calculate TML (i.e., TML1 and TML2), TML1 is, predominately, the better way to calculate TML. The primary exception to this is with TS1. •

Impact of Sampling Period: o In general, W and B periods provide the best statistical results, with W providing equal bias, but slightly better precision, as compared to B. For example, the precision and MSE obtained with the dependent sampling frequencies were generally worse than those obtained with weekly or biweekly sampling for a given length of sampling period. The exception to this observation was with scenario 3 (i.e., C and Q are constant and independent of each other), for which the MSE only varied slightly with

122

different sampling periods; precision remained the best with W and B sampling periods. •

Comparison of Test Strips with Continuous Scale Instruments: o Comparing test strips to the different continuous scale instruments (i.e., instruments 5%/0.4, 20%/0.4, 20%/4, and 60%/8) within a given sampling period and length of sampling period, the MSE for test strips are, generally, similar to instrument 20%/4. This does not hold true for precision; precision for test strips is generally worse than instrument 20%/4 and, in some cases, worse than 60%/8. o Comparing test strips using W1 to instrument 20%/4 using M1, M2 and M5, the length of sampling period for which they have approximately equal statistical results is either two years or five years (i.e., M2 or M5), dependent on the specific scenario. o Comparing test strips using W1 to instrument 5%/0.4 using SA1 through SA5, the length of sampling period at which they have approximately equal statistical results is between two years and five years (i.e., SA2 or SA5) with using 100 year data set.

•

Variation Across Scenarios: o Comparing similar instruments, sampling period, and length of sampling period across the five scenarios: In general, no trends were observed in the results for MSE and bias2. However, in terms of precision, scenario 1 generally had the highest precision.

123

o Comparing W1 using instrument 20%/0.4 to M1,2,5,10 using instrument 5%/0.4, the length of sampling period at which they provide approximately equal statistical results is dependent on the characteristics of the scenario. Specifically: ƒ

For scenarios 2 and 4, in which C varies seasonally, instrument 20%/0.4 using W1 has approximately equal statistical results as instrument 5%/0.4 using M5 .

ƒ

For scenarios 1, 3 and 5, in which C does not vary seasonally, instrument 20%/0.4 using W1 has either approximately equal statistical results as instrument 5%/0.4 using M10, or this point of approximate equality is not reached in the lengths of sampling periods used in this study.

6.2 Measure Specific Analysis From The Benin Case Study 6.2.1 Analysis of Nitrate Results The nitrate results, as detailed below, made use of the full range of the progressions to investigate the elevated concentrations of nitrate in the groundwater of Adourékoman. The following sections detail the use of the progressions, analysis of the results, and the expansions of the case study directly stemming from the data analysis.

124

6.2.1.1 Use of the Nitrate Progressions As will be demonstrated through the discussion in the following sections, the nitrate portion of the case study was significantly facilitated through the use of the progressions; indeed, the analysis contained herein would have been impossible without use of the full range of the progressions. As such, the nitrate case study was able to demonstrate the plausibility and strength of using the progressions in certain field research situations. The analysis portion of this chapter will discuss that the combination of the highfrequency and low-frequency data, and thus the use of the full range of the progressions, allows for an understanding of the groundwater nitrate problem in Adourékoman that extends far beyond that which would be possible by using just one level of the progressions. This is demonstrated through the research cycle used in this research. Specifically, methods at the most complex side of the progressions (isotopic analyses) lead to the use of less complex methods at higher frequencies. The analysis of these results then lead to the implementation of complex methods requiring expert personnel as is detailed below in the discussion on the sampling using direct push technology. In addition to the new complex sampling methods, this research has lead to new opportunities with DH and with the local population by which to continue and expand the research to pursue long-term monitoring of the groundwater and to develop solutions to the groundwater quality problems. This research cycle has used both ends of the progressions in order create a combined data set to understand the nitrate problem, a problem that, though previously identified, had not been further researched due to the limitations of working in rural regions of developing nations.

125

6.2.1.2 Interpretation of Potential Nitrate Sources for Adourékoman Based on Nitrate Data The low-frequency and high-frequency analyses of nitrate, in combination with the sociological techniques employed throughout the project, lead to observations concerning possible sources of the elevated nitrate in the groundwater as well as identification of several additional field efforts designed to provide additional information concerning these possible sources. The first step towards identifying these sources was analysis of the regional isotopic data. The analysis of the regional isotopic data suggested human and/or animal waste as the primary sources of nitrate in the regional groundwater. This was supported through the isotopic data taken in Adourékoman in February and May of 2007. These data follow the same patterns of the regional samples in their indication of original nitrate sources as well as the probable occurrence of dentrification in the groundwater. These data also indicate clear distinction among the three wells in Adourékoman in terms of their isotopic composition (and therefore possible differences in source and/or denitrification within the groundwater). The second step towards identifying the nitrate sources was analysis of the highfrequency data collected by the local sampling groups. While not a focus of this dissertation, brief mention of data collected using the additional test strip methods (pH, total metals, ammonia, and phosphate) is here made in combination with general observations concerning the nitrate concentrations in each well; tables of the full data set can be found in Appendix 2. Observations based on these data include:

126

•

Well Ayewa-Okouta has the highest concentrations of metals of the three wells, nearly always above the concentrations of the other two wells. It regularly has elevated nitrates.

•

Well Agbo demonstrated pH, total hardness, and nitrate (from test strips) that were consistently slightly above the values recorded at the other two wells.

•

Well Ayewa regularly has the lowest concentration of nitrate (concentrations below WHO health standards for drinking water)—as measured by both the colorimeter and the test strips. However, the field results showed a distinct temporal trend in this well (see Figure 6.6).

Assessment of the trend in nitrate in the well Ayewa and the distinct differences among the three wells, in terms of isotopes, nitrates, and test strip data, suggests that the three wells are impacted by distance from the source and/or different sources. This variability across the three wells is consistent with the tremendous local variability identified in the regional sampling effort. This is a significant initial observation that was beyond the assessment possible based solely on the low-frequency isotopic analyses.

6.2.1.3 Support of Source Identification with Sociological Methods In addition to these technological approaches to the problem, sociological methods of surveys, focus groups, and observation were used to locate concentrated sources of human or animal activity (and wastes) throughout the village. Although four latrines were identified in the village, it was determined that they were primarily used by the elderly or reserved for respected guests. Common practice of the majority of the population involved defecation in the region of a group of teek trees at the edge of the

127

village; there is no specific location for urination. As such, the normal toilet practices of the local population represent a potentially significant distributed source of nitrate (and other contaminants). Further, the latrines remain potential point sources of concern due to observed free-standing fluids within the latrines. Animal shelters are used to shelter cows and oxen when they are not being used for work purposes at the local farms, and thus there is a potential source of contamination associated with the animal shelters. Specifically, there is an animal shelter approximately 150 feet from well Agbo that can house up to 20 cows, though only six have been observed at any point in time. In addition to the historical animal shelters located near well Agbo, animal shelters have been built close to well Ayewa during the tenure of this project. Between the summers of 2005 and 2006, the region between the village and well Ayewa (located on a hill and about 100 by 100 meters in size) was excavated replacing the natural bush with agricultural fields and an animal shelter for pigs. Several pigs were present in the summer of 2006, but none were present during the summer of 2007. Between the summer of 2006 and 2007, additional extensive excavations were made in the area downhill from well Ayewa that came to within several meters of the well. Specifically, by February 2007, the brush had been burned and the land cleared, and, by May 2007, the land was planted with a variety of crops. Again, a large animal shelter was built on this land in close proximity to the well (10 – 20 meters from the well) with the intent of housing pigs, although no pigs had yet been purchased by the land owner. Given that the nitrate sources are likely human or animal waste, these land changes provide an increase of animal and human waste either in close proximity to or directly uphill from well Ayewa.

128

Another possible source of contamination is the use of garbage piles / trash dumps. Part of the daily routine in the village is to sweep both the house and the yard area each morning and remove all garbage (e.g., leaves, goat or sheep droppings, and food waste) to local garbage piles. In addition to the animal waste and plant decomposition that is part of these garbage piles, young children use the garbage piles as toilets. While there are multiple garbage piles in each region of the village, none are located in close proximity to the village wells. The final potential nitrate source of primary concern in Adourékoman is related to the practice of filling in old hand-dug wells with various materials ranging from wood, to sands, to charcoal, to garbage. The greatest density of hand dug wells is found in the region of well Ayewa-Okouta. Some of the hand dug wells are still operational, or ‘open’, but several have been partially or completely filled in. One of these hand dug wells is located approximately ten feet from well Ayewa-Okouta (Figure 6.8). Conversations with villagers suggest that, rather than sealing the well or filling it will soil, the hand dug well was treated as a garbage pile implying that materials disposed of into the well might include: plant debris, animal or human waste, metal from tin cans and roofing materials, charred pieces of wood (from wood stoves) and soil. Given the close proximity of this hand dug well with well Ayewa-Okouta, it is considered a reasonable probability that the hand dug and drilled, or hand pump, well intersect the same fracture set (albeit at different depths and within regions of different degrees of weathering) thus implying that the garbage found in the hand dug well may act as a contaminant source to the drilled well. If this is found to be true, the presence of the “filled-in” hand dug well could help explain the higher levels of metals in well Ayewa-Okouta.

129

Figure 6.8: Well Ayewa-Okouta and a filled-in hand dug well. The arrow on the picture to the left indicates the location of filled-in hand dug well; well Ayewa-Okouta is at the left of the photo. The picture to the right is a top view of the filled-in hand dug well. 6.2.1.4 Results From Application of Progressions Leading to New Characterization Goals As a continuation to the case study on nitrate, additional measures were taken in an effort to ascertain which, if any, of these potential nitrate sources are contaminating the wells in Adourékoman. In order to do this, we returned to the use of complicated analytical methods used at very low-frequency (in this situation, sampling several locations in space, but only once in time). These efforts include the use of direct push (dp) sampling and isotopic sampling; although these efforts did not conclusively determine the contaminant sources, they provide continuing evidence towards the identification of the sources of nitrate within Adourékoman. In order to determine if the filled-in hand dug well located in close proximity to well Ayewa-Okouta was acting as a contaminant source, a dp was purchased (www.geoprobe.com) with the intent of testing the groundwater in the filled-in hand dug well for comparative purposes; if nitrate was present, isotopic analysis of this water

130

would then be performed. There are two filled-in hand dug wells in Adourékoman, both were sampled using the dp in February of 2007. To sample these filled-in hand dug wells, the dp was placed as close to the center of the well as possible. It took 51/3 – 6½ sections of probe rode, or ~6.5 – 8.0 meters, to reach bottom hole and groundwater in these two wells. Groundwater was then extracted and, due to the significant quantities of sediment in the water, was placed in beakers and left undisturbed for 10-20 minutes. The water was then poured off the top of the beakers in order to leave the sediment in the beakers, and was tested for nitrate using both test strips and the colorimeter. In both wells, the nitrate concentrations were below the detection limits of the instruments (Table 6.3).

TABLE 6.3 GROUNDWATER NITRATE DATA AS COLLECTD FROM WELL AYEWAOKOUTA, 2 FILLED-IN HAND DUG (HD) WELLS, AND 3 OPEN HAND DUG WELLS IN ADOURÉKOMAN. ALL DATA WAS COLLECTED ON FEBRUARY 9 OR 10, 2007, USING TEST STRIPS AND A COLORIMETER. A RESULT OF ZERO INDICATES NON-DETECT ON THE INSTRUMENT.

Well Ayewa-Okouta Filled-in HD1 Filled-in HD2 Open HD1 Open HD2 Open HD3

Test Strips Nitrate Nitrite (NO3-N) (NO2-N) 20-50 0 0 50 50 20-50

131

0 0 0 0 0 0

Colorimeter Nitrate (NO3-N) 25.4 0 0 51 28.7 26.3

In order to compliment the data from the filled-in hand dug wells, the three open hand dug wells in the village were sampled and tested for nitrate over the same time period as the dp experiments. These samples were collected via the buckets and rope units already present at the hand dug wells and used by the local population. The first two open hand-dug wells (HD1 and HD2) are located across the main road from well AyewaOkouta. The third open hand-dug well (HD3) is located in the center of the village, not far from the second filled-in hand-dug well, and between wells Ayewa and Agbo. Each open hand dug well had high concentrations of nitrate that were significantly above the WHO standards (Table 6.3). The combination of the nitrate data from the filled-in and open hand dug wells indicate the presence of elevated nitrate in open hand dug wells, but not in filled-in hand dug wells. Although the small sample size makes this observation quite preliminary, it suggests that filled-in hand dug wells may not be a source of nitrate contamination in Adourékoman. The reasons for this remain to be investigated. Following the sampling of the filled-in hand dug wells, it was hoped that the groundwater below / in close proximity to the animal shelter near well Agbo could be sampled using the dp. If nitrate was present in the groundwater, isotopic analyses were to be performed. Unfortunately, after significant effort including participation by villagers, this sampling had to be abandoned as we were unable to penetrate to the level of the groundwater with the manual dp. Other sampling efforts made to date to provide additional support for the identification of the nitrate sources include isotopic analysis of the samples collected from the open hand dug wells (locations as above discussed) and from two small streams

132

(Figure 6.9). The first stream sample (stream1) was taken uphill, and upstream, from well Ayewa, but down stream from a significant garbage pile. The second stream sample (stream2) was taken close to well Ayewa, though not far from stream1. Both are small seasonal streams; their presence is dependent on recent rainfall events.

20 18 2-2007 Open HD1 2-2007 Open HD2 2-2007 Open HD3 5-2007 Stream1 5-2007 Stream2

16 14

18

δ O

12 10 8 6 4 2 0 0

10

20

30

40

15

δ N

Figure 6.9: N and O isotopes of nitrate from hand dug wells and streams in Adourékoman superimposed on figure 6.2. Dates of samples as indicated. The isotopes of all open hand dug wells and streams tested follow the same trends as those displayed by the wells Ayewa-Okouta, Ayewa, and Agbo suggesting human or animal waste as the source of nitrate. These data also suggest the possible presence of dentrification in the system. HD1 and HD2, those closest to well Ayewa-Okouta, have similar, though slightly lower, isotopic values than well Ayewa-Okouta. The isotopes of HD3, located in the center of the village, are between the values of all three wells.

133

The stream isotopes, which represent rain water runoff and associated surface contamination, support the argument that both the local toilet practices and garbage piles serve as sources of nitrate. One interpretation of the stream isotope samples is that stream2 has experienced greater amounts of dentrification than stream1. One plausible explanation for this is that stream2 is not far from a garbage pile. It is possible that the garbage pile rapidly becomes anaerobic due to the presence of high levels of nutrients— both biomatter and human / animal waste. As such, water flowing through and past the garbage pile could experience high levels of dentrification.

6.2.1.5 Progressions Leading to Conclusions The progressions for nitrate, in combination with the sociological methods, have provided a combined data set for nitrate that has allowed us to investigate possible sources of nitrate contamination identified in Adourékoman (local hygiene / toilet practices, garbage piles / trash dumps, animal shelters, open hand-dug wells, and filled-in hand dug wells). Based on the data collected through use of the progressions, two sources (agriculture and filled-in hand-dug wells) were eliminated and confidence was increased that the other possible sources contribute to the groundwater contamination. As such, four likely sources of nitrate in the village of Adourékoman have been identified through this case study: (i) latrines and common toilet practices, (ii) garbage piles / trash dumps, (iii) animal shelters, and (iv) open hand-dug wells. Additionally, the observations concerning the filled-in hand dug wells provide potentially important observations for the region and practices concerning how to close or fill-in no longer used hand dug wells.

134

6.2.2 Analysis of Uranium Results The uranium results, in combination with the sampling experience, leads to three observations. First, and most importantly for this dissertation, the results suggest that none of the available sampling methods are consistent with being used by the low end of the POE. Second, the colloidal phase represents a relatively small percentage of the uranium present in the groundwater sampled. Third, temporal changes in uranium concentration were identified providing justification for ongoing monitoring.

6.2.2.1 Use of the Uranium Progressions The sampling method used in this research is considered inappropriate for use by sampling personnel with lower levels of expertise. This observation comes from the concentrated acids used in the sampling of the dissolved phase, as well as the precision required to perform the sampling of the colloidal phase of uranium. Significantly, review of the available methods for analysis of uranium leads to the conclusion that there are no available methods that are appropriate for use by sampling personnel with lower levels of expertise. Hence, while the nitrate portion of the project demonstrated the plausibility of working with the lower expertise end of the POE to perform high-frequency sampling using the lower end of the POAM, the uranium effort demonstrates that limits on the POAM may result, at the present time, in the inability to use the lower end of the POE.

135

6.2.2.2 Interpretation of the Uranium Data in Bénin Comparison of the dissolved and colloidal phases of uranium across the sample locations indicate that the vast majority of the uranium is found in the dissolved phase. Although this is evident in all samples, it is particularly evident at wells Kpakpazoume A and Kpakpazoume B (data in Table 6.1). These wells are located at the edge of the village of Kpakpazoume with approximately 20 meters of horizontal distance separating the two wells (vertical separation is unknown). In both wells the dissolved phase of uranium is well above WHO and EPA standards for uranium in drinking water; Kpakpazoume A has dissolved concentrations around 250 μg/L (specifically, 261 μg/L on 14 June 2006 and 240 μg/L on 13 July 2006) and Kpakpazoume B has dissolved concentrations around 100 μg/L (specifically, 127 μg/L on 14 June 2006 and 83..3 μg/L on 14 July 2006). Yet, the colloidal concentration of uranium remains well below 1 μg/L in both wells (Kpakpazoume A: 0.231 μg/L on 14 June 2006 and 0.1589 μg/L on 13 July 2006; Kpakpazoume B: 0.0172 μg/L on 14 July 2006). The unfiltered colloidal samples for Kpakpazoume A, those representing all particles in the water greater than 0.45 microns in size, had concentrations above 1 μg/L (1.29 μg/L on 14 June 2006 and 1.39 μg/L on 13 July 2006). While the concentration in all particles was nearly ten fold higher as compared to the colloidal particles, it represents less than 0.5% of the uranium observed in the dissolved phase. With respect to interpretation of the dissolved-phase data, the Kpakpazoume A and Kpakpazoume B wells produced the highest concentrations of uranium sampled (>100 μg/L). Further, the three wells in Adourékoman (Agbo, Ayewa-Okouta and Ayewa) all have concentrations above WHO standards (15 ppb) and approximately at the

136

EPA standards (30 ppb) for uranium in drinking water. Hence, the field sampling illustrates that there is a relatively persistent problem with uranium in groundwater in this region of Bénin. With respect to temporal trends, all of the locations sampled at more than one point in time demonstrate variation in uranium concentration. Two examples of this are Kpakpazoume B and Adourékoman Ayewa. The difference between the dissolved phase uranium samples at Kpakpazoume B taken one month apart is 44 μg/L, or 35% of the first sample. At Adourékoman Ayewa, three samples were taken over a period of one month. The first and last samples show similar concentrations (approximately 1.2 μg/L different), but the middle sample is 5 μg/L, or 17%, lower than these two samples. While the lack of high-frequency data make it difficult to identify trends or to correlate with other environmental factors, these data suggest both that the concentration of uranium is actively fluctuating over time and that high-frequency sampling for uranium has the potential to provide important insight into the source and persistence of this hazard.

6.2.2.3 Implications for Future Uranium Sampling Efforts The observation of minimal concentration in the colloidal phase has significant importance to the future of the sampling program in Bénin. Specifically, although reasons for distribution of uranium between the dissolved and colloidal phases remain to be investigated, the results suggest that continued sampling of the groundwater colloidal or particulate phases for uranium is likely not warranted from the standpoint of assessment of potential health impacts of uranium contained in the groundwater.

137

The uranium data also suggest that the development of high-frequency sampling techniques for the dissolved phase of uranium consistent with use by the lower end of the POE would be of significant value. In particular, such techniques could provide valuable information regarding the temporal variation of uranium in the groundwater. While this would require the development of new sampling techniques, the nitrate portion of this case study suggests that, if less complex analytical methods were developed, collaborations with local populations could enable the high-frequency sampling desired.

6.3 Analysis of Monte Carlo Results The results of the Monte Carlo studies lead to a series of observations concerning the impact of the sampling period, length of sampling period, instrument / operator, and the scenarios used to create the data sets. The analysis, which supports the working concept and use of the full range of the progressions, begins with general observations and continues to specific observations of the above variables.

6.3.1 General Observations The results of the Monte Carlo studies demonstrate that different combinations of sampling instrument (considering both the analytical method and expertise of the user) and sampling strategy can lead to equivalent MSE for the parameters studied. This implies that a desired level of MSE in the final estimate of a parameter can be achieved when different portions of the POE, POAM and POSS are selected based on specific needs of a particular application. It also implies that the low end of the POAM and POE progressions can, under appropriate conditions, be combined to provide equivalent MSE

138

on parameters such as MeanC and TML as would be obtained using low-frequency application of the upper end of these two progressions, thus supporting the working concept of this dissertation. An example of different progressions leading to the same MSE (discussed in more detail below) is found in comparing, for parameter MeanC, test strips used weekly to instrument 5%/0.4 used semi-annually. In this situation, test strips represent the least complex end of the POAM using the least expert portion of the POE, and instrument 5%/0.4 represents an instrument on the complex portion of the POAM and a high level of expertise. Using these combinations of the POAM and POE with their respective sampling periods (weekly versus semi-annually), approximately equal MSE is achieved in the final estimates of MeanC. The Monte Carlo results also demonstrate that arbitrary choice of analytical method, level of expertise, and sampling strategy will not, in general, ensure achievement of a desired level of MSE in the parameter of interest. Hence, as suggested in the working concept, the selection of the appropriate positions on the POE, POAM and POSS requires knowledge, within the expert, of the intended use of the data analysis and opportunities / limitations of the field sampling scenario. An example of this is found in comparing, for parameter MeanC, instrument 20%/4 used weekly to instrument 5%/0.4 used semiannually. In this situation, instrument 20%/4 represents an instrument on the less complex portion of the POAM (though more complex than test strips) being used by someone on the low side of the POE, while instrument 5%/0.4 represents the complex portion of the POAM and a high level of expertise. Using these combinations of the POAM and POE with their respective sampling periods (weekly and semi-annually), it is better, in terms of MSE, to use instrument 20%/4 regardless of the length of the sampling period. That is,

139

by choosing the instrument and operator combination with the lowest e and b, but limiting the operator to semi-annual sampling, the sampling design resulted in higher MSE than the alternative design using the lower quality instrument.

6.3.2 Observations on Use of the Low End of the POE and POAM 1-For select sampling situations, test strips can represent valuable instrument options for high-frequency sampling scenarios, particularly those using the low end of the POE. Following on the previous paragraph, these studies suggest that a combination of high-frequency sampling using test strips (an analytical method at the low end of the POAM) in combination with expertise effectively from any portion of the POE can represent a valuable field research approach to characterization of groundwater quality. This is particularly valuable for situations in which limitations on analytical methods, sampling strategy, or expertise indicate that the e and b of continuous scale instruments cannot be minimized to the point of providing acceptable MSE of an intended parameter. Figures 6.10 and 6.11, for example, demonstrate the strength of test strips when they are used in combination with high-frequency sampling for parameters MeanC and TML. In both figures the precision and bias2 of test strips (TS2) using W1 through W10 are compared to the precision and bias2 of instrument 5%/0.4 using SA1 through SA10; these figures show the evolution of these data over an increase in length of sampling period. It is noted that for both parameters and sampling lengths of 2 years or less, test strips (W1 and W2) have significantly better precision and lower MSE than instrument 5%/0.4 (SA1 and SA2). For most of the scenarios simulated, and for sampling periods of 2 years or less, similar results were observed when comparing test strip data using biweekly or

140

monthly sampling frequency against instrument 5%/0.4 semi-annual sampling. As such, these observations suggest that it may be better, in certain realistic situations, to use test strips with sampling periods of greater frequency (weekly to monthly depending on the specific scenario), than to use a high quality instrument at substantially lower frequency (e.g., semi-annually) for lengths of time less than 5 years for characterization of parameters MeanC and TML. 7.000 6.000

(miligrams / liter)

2

5.000 4.000 3.000 2.000 1.000 0.000 0

2

4

6

8

10

12

Length of Sampling Period (years) Precision: Weekly, TS2 MSE: Weekly, TS2

Precision: Semi-annual, 5%/0.4 MSE: Semi-annual, 5%/0.4

Figure 6.10: Evolution of MSE and precision for parameter mean over an increase in length of sampling period

141

1.20E+14

1.00E+14

(miligrams)

2

8.00E+13

6.00E+13

4.00E+13

2.00E+13

0.00E+00 0

1

2

3

4

5

6

Length of Sampling Period (years) Precision: Weekly, TS2 MSE: Weekly, TS2

Precision: Semi-annual, 5%/0.4 MSE: Semi-annual, 5%/0.4

Figure 6.11: Evolution of MSE and precision for parameter TML over an increase n length of sampling period

2- Continuing research required to improve use of test strips Although test strips have been shown to be a potentially valuable alternative to continuous-scale methods for high-frequency sampling of certain parameters, the choice of how to report the result of a test strip (in terms of a single value from within the reported range) is complex and dependent on the characteristics of the specific measure / scenario as well as the potential interpretation of the data. For example, in the scenarios used in this study, the minimum MSE obtained from TS1 versus TS2 was dependent on the parameter of interest and the characteristics of the data set. At the same time, in situations where health is a critical concern, the researcher might prefer to use TS3 as it overestimates parameter concentration thus maximizing the probability hat the actual

142

concentration or TML is below that estimated from the data. Table 6.4 shows a summary of the evaluation of TS1 and TS2 based on the precision and MSE. Based on these evaluations for each of the scenarios and parameters, we were able to identify the option that generally provided the best results across all instruments, sampling frequencies, and length of sampling period. The table also identifies whether the bias for the selected option was negative or positive, and whether the absolute value of the bias was the smaller of the two options (i.e., TS1 or TS2). Review of this table demonstrates the complexity of selecting from these two values in reporting test strip data.

TABLE 6.4 EVALUATION OF NUMERICAL OPTIONS TO REPRESENT DISCRETE RANGES OF TEST STRIP RESULTS Parameter

Mean

Max

TML

Scenario 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

Test Strip TS1 TS2 X X X X X X X X* X X X X X* X X

Bias ( - or + ) positive/negative smallest magnitude (Y/N) positive Y negative Y positive Y positive Y negative Y or N negative or positive Y negative Y complex negative Y or N negative Y positive Y negative N complex positive Y negative Y or N

*The decision between TS1 and TS2 was a difficult decision due to significant variation amongst the different sampling periods, lengths of sampling period and instrument.

143

These same Monte Carlo studies indicate that TS3 is a substantially poorer measure than either TS1 or TS2 in terms of the parameters assessed in our study, perhaps implying that TS3 should not be used to report test strip data. However, review of the details of these simulations leads to the observation that the poor performance is a function of a combination of the mean concentration value used in the simulations (10 mg/l – relative to the concentration steps on the test strip; 0, 0.5, 2, 5, 10, 20, and 50 mg/l) combined with the concentration ranges used on the test strips (based on the ranges available on the nitrate test strips). A brief simulation study was performed using a scenario for which the concentration ranges on the tests strips were changed from those used on the actual nitrate test strips to uniform ranges (i.e., 0, 4, 8, 12, 16, 20, and 24) such that the mean of the data sets (10) fell within the middle of one of the ranges rather than at the limit of one of these ranges. Table 6.5 shows that, in this case, all three striptest measures (TS1, TS2, and TS3) perform approximately equally in terms of MSE. These results support the observation that the poor performance of TS3 was a result of the specific characteristics of the simulations and the ranges used for the test strips rather than an inherent problem with using the maximum value within the concentration range in recording the data (i.e., using TS3). While this result is reassuring in terms of explaining our specific result, it illustrates a further complexity in choosing amongst TS1, TS2 and TS3 for a field scenario in which the magnitude of the mean concentration will not be known a priori.

144

TABLE 6.5 RESULTS FROM MONTE CARLO STUDY USING UNIFORM TEST STRIP INERVALS (INTERVALS USED: 0, 4, 8, 12, 16, 20, AND 24) Scenario: #1 Sampling Years Period Weekly 1

Biweekly

1

Monthly

1

Semiannual

1

Season dependent

1

Parameter dependent

1

Mean Concentration 1σ & Bias ts v1 ts v2 ts v3 ts v1 ts v2 ts v3 ts v1 ts v2 ts v3 ts v1 ts v2 ts v3 ts v1 ts v2 ts v3 ts v1 ts v2 ts v3

Prec. 0.039 0.039 0.039 0.072 0.072 0.072 0.163 0.163 0.163 2.192 2.192 2.192 0.295 0.295 0.295 0.185 0.185 0.185

Bias -2.005 -0.005 1.995 -2.001 -0.001 1.999 -2.008 -0.008 1.993 -1.970 0.031 2.031 -2.015 -0.015 1.986 -1.816 0.184 2.184

Bias2 4.060 0.039 4.019 4.076 0.072 4.069 4.193 0.163 4.133 6.071 2.193 6.316 4.354 0.295 4.237 3.483 0.219 4.955

MSE 4.099 0.078 4.058 4.148 0.144 4.141 4.356 0.326 4.296 8.263 4.385 8.508 4.649 0.590 4.532 3.667 0.403 5.140

Given the complexity of the decision regarding which number to use to represent the results as measured by test strips, a suggestion for future researchers is to use all three options—if one of the options is significantly different than the other two options, the reasons for this difference can be investigated, and will likely provide insight into the interactions between the test strip ranges, characteristics of the data set, and parameter of interest.

145

6.3.3 Observations on Instruments / Operators with Larger e and b These studies suggest that instruments with very large e and b will result in high MSE and low precision of parameter estimates, regardless of the sampling period or length of the sampling period. In these studies, the use of instrument 60%/8 resulted in estimates of the parameters with extremely large errors, and these errors were, generally, significantly worse than those experienced with test strips. Figure 6.12 demonstrates the large error experienced by instrument 60%/8 for W1 and SA1 sampling periods. Given the observations above concerning test strips, it is interesting to note that it is worse, in terms of MSE, to utilize instrument 60%/8 using W1 than test strips using SA1. Further, within any sampling period, the errors associated with instrument 60%/8 are significantly worse than those experienced by all other instruments. The decrease of errors associated with instrument 60%/8 over an increase in length of sampling period does not justify, in our studies, the use of such instruments as the errors remain significantly greater than those of test strips or other instruments used for much shorter lengths of time. Thus, if a project is working with a local population and it is likely that the instrument will have significant e and b due to either the instrument and/or the operator, it is, for all parameters used in this study, better to use test strips. In addition, this suggests that, if working with local populations, extra time spent training the local population or creating tools to minimize e and b would be time and energy well spent in terms of reduced error in the final parameter estimates.

146

180.000 160.000

(miligrams / liter)

2

140.000 120.000 100.000 80.000 60.000 40.000 20.000 0.000 W1 TS2

W1 5% / 0.4

W1 W1 W1 20% / 20% / 60% / 0.4 4 8

SA1 TS2

SA1 5% / 0.4

SA1 SA1 SA1 20% / 20% / 60% / 0.4 4 8

Instrument & Sampling Period Precision

Bias

2

Figure 6.12: Precision and bias2 of all instruments used weekly or semi-annually for one year. The total height of each bar is equal to the MSE. 6.3.4 Trade-offs Among Sampling Period, Instrument and MSE These studies suggest that, with a decrease in the frequency of sampling (i.e., W to B to M to SA), there is an increase in the error (i.e., MSE increases and precision decreases) of the results regardless of the instrument used. Figure 6.13 demonstrates this tradeoff between sampling period and resulting MSE. In this figure instruments used in this study are grouped to demonstrate the significant increase in error from W1 to M1 to SA1. (Given the above discussion on instrument 60%/8, it was not included in this figure.) It should be noted that precision increased and MSE decreased significantly with an increase of the sampling period, although a greater portion of the MSE for each instrument and sampling period is attributed to bias2. However, as is demonstrated in

147

Figures 6.10 and 6.11, this effect decreases with an increase in the length of the sampling period. As such, if a short study or an initial study is being performed, it is preferable to use sampling periods with a higher frequency. 35.000

(miligrams / liter)

2

30.000 25.000 20.000 15.000 10.000 5.000 0.000 W1 TS2

M1 TS2

SA1 W1 M1 SA1 W1 M1 SA1 W1 M1 SA1 TS2 5% / 5% / 5% / 20% / 20% / 20% / 20% / 20% / 20% / 0.4 0.4 0.4 0.4 0.4 0.4 4 4 4 Instrument and Sampling Period Precision

Bias 2

Figure 6.13: Precision and bias2 of one year lengths for weekly, monthly, and semi-annual sampling periods grouped by instrument. The total height of each bar is equal to the MSE. An additional observation from Figure 6.13 is drawn from the comparison of instrument 20%/0.4 using W1 to instrument 5%/0.4 using M1 or SA1. In both circumstances, instrument 20%/0.4 using W1 provides a lower MSE and better precision. In situations in which using a local population, as opposed to an expert technician, to perform a method does not influence the instrument b, but increases instrument e, these data suggest that using a local population to sample at a high-frequency (weekly)

148

provides results with less error than obtained from an expert sampling at a lower frequency (monthly or semi-annually).

6.3.5 Observations on Time-variable Sampling Periods These studies suggest that time-variable sampling periods do not generally yield lower error in terms of precision or MSE than uniform sampling periods. It was anticipated, due to surface water literature, that time-variable sampling would decrease the error of TML estimates, but this was not experienced in this study. Figure 6.14 compares TML as derived from W1, W2, and W5 with TML as derived from SD1, SD2, and SD5 (Figure 6.14a) or PD1, PD2, and PD5 (Figure 6.14b); instrument 20%/0.4 is used for all sub-sampling in this figure. The 10 year length of sampling period is excluded from this figure due to the above discussed anomalies associated with parameter TML and this length of sampling period within this study. Figures 6.14a and 6.14b demonstrate that, for sampling periods less than 5 years, it is better to use weekly sampling than either time-variable sampling strategy. At 5 years, the MSE of W5 is approximately equal to SD5, and is better than PD5, but the precision of W5 is better than both SD5 or PD5. These observations are further supported by the results of parameters MeanC and MaxC. These data (for which the 10 year length of sampling period is included in the analysis) suggest that, for many scenarios, PD1 through PD10 provides approximately equal statistical results to M1 through M10. Further, SD1 through SD10 provides approximately equal statistical results to B1 through B10 or M1 through M10 (dependent on scenario and parameter), respective of the length of sampling period. Hence, given the added complexity of training a local population to consider either SD or PD sampling

149

1.60E+13 1.40E+13 1.20E+13

(miligrams)

2

1.00E+13 8.00E+12 6.00E+12 4.00E+12 2.00E+12 0.00E+00 0

1

2

3

4

5

6

Length of Sampling Period (years) Precision: Weekly, 20%/0.4 MSE: Weekly, 20%/0.4

Precision: Season dependent, 20%/0.4 MSE: Season dependent, 20%/0.4

(a)

1.60E+13 1.40E+13 1.20E+13

(miligrams)

2

1.00E+13 8.00E+12 6.00E+12 4.00E+12 2.00E+12 0.00E+00 0

1

2

3

4

5

6

Length of Sampling Period (years) Precision: Weekly, 20%/0.4 MSE: Weekly, 20%/0.4

Precsion: Parameter dependent, 20%/0.4 MSE: Parameter dependent, 20%/0.4

(b)

Figure 6.14: MSE and precision of time-variable sampling as compared to time-uniform weekly sampling of parameter TML. a) Season dependent sampling period, b) parameter dependent sampling period.

150

strategies, it is not clear that there is an adequate return on “investment” to pursue these more sophisticated sampling schemes, particularly when working with the low end of the POE. This conclusion might be due to the lack of a “wash out” effect in the scenarios used in these studies; that is, there were no substantial changes in C related to the pumping rate or season. If there were a “wash out” effect, it is possible that this conclusion would change.

6.3.6 Parameter Sensitivity to Instrument The results of the Monte Carlo studies suggest that the parameter of interest should directly effect the instrument / operator chosen for a specific study. Even though the Monte Carlo studies performed are limited in that only three parameters were utilized, the difference in the sampling design needed to achieve desired error of the estimations of the parameters is evident in the data. For example, in these studies, if the parameter of interest was either MeanC or TML, high-frequency sampling periods (even with lower quality instruments) provide better statistical results than low-frequency sampling periods; this remains true even when instruments with higher e and b are used to perform the high frequency sampling. This is depicted in Figure 6.10, which compares test strips using W1,2,5,10 and instrument 5%/0.4 using SA1,2,5,10, and suggests that at lengths less than 5 years, test strips using weekly sampling periods provides the better results. Therefore, if the parameter of interest is MeanC or TML, it is often better to choose an instrument / operator combination that allows for greater sampling frequency than one with lower e and b that would necessitate sampling at a lower frequency.

151

In our studies, the instrument / operator needs of parameter MaxC is quite different than that of MeanC and TML. The results suggest that, if the parameter of interest is MaxC, instruments with low e and b need to be used if the project is to be longer than 2 years in length. This is demonstrated in Figure 6.15 through the comparison of test strips using W1,2,5,10 with instrument 5%/0.4 using SA1,2,5,10. In this example, the error associated with test strips, though they are used with a higher frequency sampling period than instrument 5%/0.4, increases with increasing length of sampling period (i.e., 1 to 2 to 5 to 10 years), while instrument 5%/0.4, though used with a lower frequency sampling period, has decreasing error with increasing length of sampling period. Thus, if the parameter of interest for a project is MaxC, an instrument with low e and b, which we generally associate as being used by an expert, should be utilized, even if this necessitates a lower frequency of sampling period.

152

60.000

(miligrams / liter)

2

50.000

40.000

30.000

20.000

10.000

0.000 0

2

4

6

8

10

12

Length of Sampling Period (years) Precision: Weekly, TS2 MSE: Weekly, TS2

Precision: Semi-annual, 5%/0.4 MSE: Semi-annual, 5%/0.4

Figure 6.15: Evolution of MSE and precision for parameter max over an increase in length of sampling period

6.4 Common Observations to the Case Study and Monte Carlo Studies The analysis sections of this chapter provided observations and analysis of the measure specific results of the case study and the results of the Monte Carlo studies as they related to the working concept of this dissertation. In addition to these analyses specific to the research tools we utilize, a series of observations discussing the significant overlap between these analyses is here provided in the following five points:

153

1. Benefit of high-frequency sampling even if it is based on the low ends of the POAM and POE In the comparison of the Monte Carlo analysis with our field experience as expressed through the nitrate case study, both studies demonstrate the benefit of highfrequency sampling, even if it is based on the low end of the POAM and the POE. In the Monte Carlo studies it was generally found that using instrument 20%/4 or test strips at high sampling frequencies (i.e., weekly or biweekly) provided lower error in the estimation of parameters MeanC and TML than using an instrument with lower e and b at low sampling frequencies (i.e., 5%/0.4 at monthly or semi-annual sampling frequencies). This was experienced in the nitrate case study in which a colorimeter (with e and b of ~20%/4 when used by the trained local population) and test strips were used by the local populations to sample at high frequencies. Despite the noise in the data as expressed through the high e and b of the analytical methods as used by the local population, these high-frequency samples allowed for a significantly greater understanding of the nitrate problem than was possible through exclusive use of low-frequency sampling as performed by DH. As described in the case study analysis, the use of this combination of the low end of the POAM and POE enabled better understanding of the nitrate problem in Adourékoman, allowed for identification of temporal trends, and lead to additional research to identify sources of nitrate contamination.

154

2. Potential contribution of test strip methods for high-frequency sampling In considering the relationship between test strips and continuous scale instruments, as determined from combination of the Monte Carlo analysis and our field experience, the best e and b that we have experienced working with local populations is approximately 20%/4 for a colorimetric instrument. This level of field precision and bias, combined with the Monte Carlo results, implies that for parameters MeanC and TML, it would be better for the local population to use test strips than the colorimeter for shorter studies (i.e., 1 to 2 years in length). Our field experience also suggests that it is easier to reliably train local populations to use test strips than more complex analytical instruments making this an appealing option in terms of the time commitment to train the local population to use the instrument. Further, the low per sample costs, low initial monetary expenditure, and lack of required MSDS (material safety data sheets) suggest that test strips are an excellent analytical option for select parameters of interest when working in rural regions of developing nations.

3. Demonstration of potential utility of using the entire range of POE, POAM, and POSS Both the Monte Carlo studies and the nitrate portion of the case study support the potential utility of using the entire range of the three progressions in creating a sampling design. The Monte Carlo studies demonstrate the possibility of coupling sampling designs using the low end of each of the progressions with designs using the high end of each of the progressions. For example, test strips used on a weekly basis by a local population over 2 years could be complemented with complex methods used on a semi-

155

annual basis by experts over 10 years. By coupling these two sampling designs, information (such as temporal variation in measures at temporal scales of weeks to months) could be gathered over the first 2 years that would be difficult to ascertain with semi-annual sampling, while the continued sampling using complex methods would provide the ability to observe long-term trends in the parameter of interest after completion of the high-frequency sampling. Alternatively, high-frequency sampling could be employed as a two-year supplement during an extended, semi-annual sampling program if the low-frequency data indicate that measure variability is likely to be happening between the semi-annual samples. The results of the Monte Carlo study suggest that, for some parameters, this would minimize the error of the estimates of the parameters of interest throughout the duration of the project. Further, this agrees with the surface water literature which suggests that increased length of sampling period decreases the need for high-frequency sampling (Moosmann et al. 2005). This combination of using the entire range of the progressions was demonstrated through the nitrate portion of the case study as seen through the research cycle in which different portions different sampling designs were used at different points along the project. By using isotopes (low-frequency sampling), colorimetric / test strip measures of nitrate concentration (high-frequency sampling), followed by additional isotopic sampling (low frequency), the strengths of each type of sampling were maximized and the water quality problem further understood. As such, both studies support sampling designs that utilize the full range of the three progressions.

156

4. Potential value of developing uranium methods consistent with high-frequency sampling At the present time, no methods exist at the lower end, or the less complex portion, of the POAM for uranium thus preventing collaboration with local populations to perform high-frequency sampling. The above points that demonstrate the value of highfrequency sampling, even if it necessitates using instruments with higher e and b, leads to the suggestion that there is potential value in the development of uranium methods that are consistent with high-frequency sampling. As with the nitrate case study, it is not suggested that these methods be used to the exclusion of the complex methods currently available, but rather in parallel with them with appropriate sampling periods. By doing this, the benefits of both sampling designs could be bolstered thus providing the possibility of significant new insights into the source and temporal variation in the uranium threat.

5. Benefits of time-variable sampling do not generally outweigh the potential problems The results of the Monte Carlo study, as supported by our field experience through the case study, suggest that potential problems with using the time-variable sampling periods generally outweigh the benefits. The Monte Carlo studies suggest that time-variable sampling does not provide better estimates (those with less error) than timeuniform sampling periods with higher frequencies. Although these results could be a direct result of behaviors simulated in the scenarios or sampling periods used, these periods and scenarios were selected because they were similar to what has either been experienced in field situations, or, in the case of the sampling periods, what would be

157

possible for either local populations or experts to perform. Given our field experience regarding the complexity associated with teaching a local population to reliably perform time-variable sampling, we suggest that, in many cases, time-variable sampling should not be used. However, in some situations, season dependent sampling periods might be beneficial for longer studies as the time commitment for the local population or the technician is less demanding during certain times of the year than with time-uniform sampling periods with shorter periods. This could be valuable in situations where sampling frequency could be in-sync with other work demands (e.g., harvest demands of farmers).

158

CHAPTER 7:

CONCLUSIONS

7.1 Return to Progressions The need for this dissertation research has been expressed in the increasing dichotomy between the current direction of groundwater research which increasingly relies on the use of highly complex methods and sampling strategies that necessitate the use of experts, and the significant need for groundwater research and monitoring in rural regions of developing countries where operational constraints including remote field sites without infrastructure (e.g., lack of electricity or running water) as well as personnel and financial limitations that exclude the use of highly complex techniques. At the time that this dissertation was written, our research remained unique in its focus on whether less complicated methods, less complicated sampling strategies, and lower levels of expertise can produce valuable data sets and have appropriate contributions to research involving groundwater in rural regions of developing countries. The working concept introduced in Chapter 1 is: A progression of analytical methods, as guided by prior knowledge and data analysis combined with expertise provided by stakeholders, will provide for the characterization of measures of groundwater quality in rural regions of developing countries. The combination of analytical

159

methods, stakeholder expertise, sampling strategy and data analysis most appropriate for a given characterization scenario will depend, among other things, on the intended use of the characterization and the critical measures identified.

Fundamental to assessment of this working concept is identification and use of the three progressions related to data collection — the progression of analytical methods (POAM), the progression of sampling strategies (POSS), and the progression of expertise (POE). These progressions were used to assess the working concept through application in both Monte Carlo studies and the nitrate / uranium case study in Bénin. The Monte Carlo studies provided an opportunity to assess these progressions as applied to a relatively wide range of theoretical conditions involving five scenarios on the structure of the data, various sampling periods, various lengths of sampling period, and various qualities of the instrument (as expressed through a combination of the analytical method and the expertise of the person collecting the data). The Bénin case study allowed examination of the value of incorporating a range of methods, sampling strategies, and expertise for assessment of two field contamination problems incorporating the realities of field work in a rural region of a developing nation. Although neither the range of conditions studied through the Monte Carlo simulations nor the field conditions experienced in Bénin are all encompassing of conditions one might encounter in work within developing nations, this study provides a reasonable foundation for consideration of the potential benefit of incorporating the concept of progressions in the design of field data collection efforts.

160

The assessment of the working concept based on the Monte Carlo and case study results is presented within this chapter.

7.1.1 Monte Carlo Results The Monte Carlo studies used in this research provide support of the working concept as they support the use of the full range of the three progressions and demonstrate the importance of knowing the parameter(s) of interest in a specific project when designing the sampling strategy. Specifically, the results demonstrate that: (i) different combinations of sampling instrument (considering both the analytical method and the expertise of the user, or the POAM and POE) and sampling strategy (i.e, the POSS) can lead to equivalent MSE for the parameters studied, (ii) arbitrary choice of analytical method, level of expertise, and sampling strategy will not, in general, ensure achievement of a desired level of MSE in the estimation of the parameters of interest, (iii) the parameter of interest should directly effect the analytical method and sampling personnel combination chosen for a specific study, and (iv) for select situations, test strips, representing the least complex side of the POAM, can serve as a valuable instrument option for high-frequency sampling scenarios.

7.1.2 Case Study Results In the Bénin case study, exploratory regional research performed by experts using highly complex techniques (high end of the POE and POAM combined with the low end of the POSS) was used to identify the region (south-central Bénin) and parameters (nitrate and uranium) of interest for the case study. The data collected during the nitrate

161

research was based on the entire range of all three progressions — the most complex methods / highest level of expertise (high end of the POAM and POE) were used in the low-frequency sampling for isotopes (low end of the POSS), while the least complex methods / lowest levels of expertise (low end of the POAM and POE) were used in the high-frequency sampling for nitrate concentrations (mid range of the POSS). The initial implementation of this nitrate research, which is the focus of this dissertation, was implemented in the village of Adourékoman over the years 2004 – 2007. Combined with the sociological techniques used, this approach (i) demonstrated the ability of local monitoring groups to accurately and consistently perform the low POAM methods over an extended period of time, (ii) allowed identification of four ikely sources of nitrate (i.e., six were initially identified, two of which were eliminated through additional analyses), (iii) allowed identification of trends in the nitrate concentrations in at least one of the wells, and (iv) demonstrated the utility of combining a range of methods, sampling frequencies, and sampling expertise in developing a consistent and useful data base for assessment of groundwater quality at a rural field site. Although beyond the scope of this dissertation, it is worth noting that the nitrate characterization effort has experienced continuing success in its second level implementation in which it was expanded to include four additional villages. The original intent of the uranium research was to perform high-frequency monitoring of uranium sampling including both the dissolved and colloidal phases. However, this proved impossible due to lack of available methods on the low end of the POAM that are consistent with use by a local population. The field sampling methods established for both the dissolved and colloidal phases of uranium therefore required high

162

levels of expertise due to either dangerous chemicals used or the necessary levels of precision; as such, we were limited to use of experts (the high end of the POE) and, due to resource limitations on these experts, to low-frequency sampling (the low end of the POSS). The uranium research provided several insights relative to both the working concept and the uranium present at this field site: (i) the uranium characterization efforts demonstrated that there will be certain field measures of interest for which the concept of extension into the low end of the POAM and the POE will not be feasible due to technical limitations on the available analytical methods, (ii) the colloidal phase represents a relatively small percentage of the uranium present in the groundwater sampled, and (iii) temporal changes in uranium concentration, as represented in the lowfrequency sampling, are present thus providing justification for ongoing monitoring, preferably at higher frequency.

7.2 Assessment of the Working Concept 7.2.1 Support of the Working Concept Support of the working concept is found in the combination of the Monte Carlo and case studies. This support can be briefly summarized as: (i) The Monte Carlo studies suggest the value of using less complex analytical methods, particularly when combined with the added value of high-frequency sampling. This suggests a significant role for trained members of local populations in contributing to collection of data sets for assessment of groundwater quality. This result from the Monte Carlo studies is support by results from the nitrate case study in which it was demonstrated that local

163

stakeholders are able to consistently and reliably perform high-frequency monitoring over significant periods of time using less complex methods. This case study, for example, demonstrated the value of the resulting data in the identification of temporal trends. (ii) The Monte Carlo studies demonstrated the potential complementary benefits obtained from the combination of low-frequency sampling with sophisticated equipment (implying both complex analytical method and expertise in sampling) and high-frequency sampling (even when using less complex methods or lower level of expertise). Once again, the nitrate case study pointed to this same potential through the results obtained for Adourékoman through combination of complex methods at low sampling frequency (including the regional sampling effort to identify the study region, the use of nitrogen and oxygen isotopes for identification of possible sources of nitrate, and the direct-push sampling) and less complex methods at high sampling frequency (the colorimeter and test-strip data collected by the trained members of the local population). Significantly, the uranium case study further supports this conclusion through a negative outcome whereby the details of the temporal variability could not be delineated because of the lack of availability of low end POE and low end POAM techniques, thus eliminating the potential for high-frequency sampling of uranium.

In continued support of the working concept, both studies demonstrated that the first clause (i.e., that the expertise of stakeholders helps guide the use of the progressions) is of significant importance to the final success of a sampling campaign. For example, the

164

Monte Carlo studies demonstrated that when sampling for a parameter such as maximum concentration of a specific measure (e.g., nitrate), the use of analytical methods and sampling personnel that increase bias and decrease precision in characterizing the measure may not provide data that are sufficient to reliably estimate the parameter and may therefore not represent a viable sampling strategy. In contrast, when sampling for the same measure, but for other parameters (e.g., total mass load or mean concentration), these same analytical and sampling strategies involving increased bias and decreased precision in the measure may represent reliable sampling strategies for estimation of these new parameters. Hence, the expert must consider both the progressions available for characterizing the measure, and the application of the data collected in terms of the estimation of a particular parameter. Similarly, within the Bénin case study, the project lead was able to identify that a combination of methods, sampling strategies, and expertise were available to address the characterization of nitrate. Further, the lead was able to integrate information from these different combinations so as to provide a relatively complete assessment of the potential sources of, and likely variation in, nitrate contamination in the groundwater of Adourékoman. In contrast, the project lead was able to identify both the limitations on the methods available for assessment of uranium and the inconsistency of these methods with the desired use of low-expertise personnel to complete the samples. Hence, the expert was able to identify, without expenditure in the field on high-frequency, but poor utility, data that there were significant limitations on our current ability to characterize the temporal variation in uranium concentrations in this groundwater.

165

7.2.2 Suggested Modification to Original Working Concept Given the substantial support for the working concept provided through this research, we believe that the working concept is effective with only slight modification to the original wording. The Monte Carlo studies demonstrated that the parameter(s) of interest should directly impact the choice of techniques (the combination of analytical methods, stakeholder expertise, and sampling strategy) selected for a given research project or characterization scenario. As such, we believe that adding “parameters of interest” to the factors affecting the selection of techniques bolsters the strength of the working concept. The modified working concept reads as follows: A progression of analytical methods, as guided by prior knowledge and data analysis combined with expertise provided by stakeholders, will provide for the characterization of measures and parameters related to groundwater quality in rural regions of developing countries. The combination of analytical methods, stakeholder expertise, sampling strategy and data analysis most appropriate for a given characterization scenario will depend, among other things, on the intended use of the characterization, as well as the critical measures and parameters identified.

7.3 Future Research This research, through the Bénin case study and the Monte Carlo studies, has demonstrated the strength of a data set that can be obtained by building that data set through contributions from the full range of personnel from experts to local populations

166

using appropriate analytical methods and sampling strategies. While this work is foundational to demonstrating the merit of the working concept, it also directly lends itself to several avenues of future research that can be divided into six categories: expansion of the numerical studies, development of analytical methods, groundwater quality research in developing countries, groundwater quality research in developed countries, expansion to other environmental or public health fields, and sociological / anthropological research. The following is a brief overview of each of the six categories. While providing substantial insight into the sampling process, the Monte Carlo studies addressed only a limited range of parameters (MeanC, MaxC and TML) and underlying scenarios (variation in C and Q). Given the substantial difference in conclusions for MaxC versus MeanC and TML, it seems that a reasonable next step of the numerical research would be to expand the studies to the assessment of other parameters and/or other scenarios. Among these might be both inclusion (as a scenario) and assessment (as a parameter) of temporal trends in the data and assessment (as a parameter) of the probability of exceedance of a set concentration constraint. Such assessment would both provide further insight into the sensitivity of various parameters to sampling design and the utility of progressions across a wider range of sampling scenarios. The uranium portion of the case study demonstrated the need for less complicated analytical methods to be developed for measures of interest. The Bangladesh example discussed at the beginning of this dissertation is one example where less complex analytical methods have been developed in response to specific needs—arsenic test strips that do not require the use of dangerous chemicals have been developed (Hach Company

167

product number 28000-00). Given that our research has demonstrated the possibility of working with local stakeholders to sample groundwater at higher frequency when less complex analytical methods are available, the development of less complex analytical methods for a broader range of measures of interest has the potential to significantly enhance our ability to characterize groundwater quality in rural regions of developing countries. Hence, it is suggested that valuable research can and should be focused on development of new methodologies for select contaminants such as key heavy metals, pathogenic organisms, agricultural chemicals, or industrial chemicals. A second area of research and development of analytical methods that needs to be addressed is that of the short shelf life and narrow range of temperature stability of chemicals used in test kits. Although not discussed in this dissertation, significant issues were encountered with the shelf life of test strips and chemicals used with the nitrate colorimeter. Given the lack of infrastructure when working in rural regions of developing nations, analytical equipment and associated chemicals cannot be stored in artificially cooled rooms (e.g., air conditioned rooms) thus making them subject to ambient air temperatures. In Bénin, the daily temperatures regularly exceeded 30°C, the maximum storage temperature for the analytical methods used, thus reducing the shelf life of the test strips and chemicals. Our experience suggests that, for some test strips, shelf life as short as 6 months will be the reality. This decreased shelf life creates significant problems with equipment distribution given the distance to field sites (both from location of production of the equipment as well as from a capital or main cities within a country). As Benin is characteristic, in terms of climate and available resources, of many developing countries, it is argued that the development of analytical methods with longer

168

shelf life and a wider range of temperature sustainability would significantly facilitate the use of these methods by local populations to perform high-frequency sampling. Moving beyond specific aspects of either the Monte Carlo studies or the Bénin case study, the principles behind the working concept have a wide applicability that opens up significant areas of research that were previously limited due to difficulties accessing regions of interest, lack of sampling personnel, limited finances, or a lack of analytical methods available suitable to the available infrastructure. Concerning groundwater researching in developing nations, the ability to use the full range of the POAM, POSS, and POE enables sampling designs to be created that suit the specific situations thus providing novel ways to monitor water quality where resources are extremely limited. Although the options are extensive, several reasonable ways with which to begin this research expansion is to work with organizations that already have access to the rural regions, have developed relationships with local populations, and understand the associated socio-cultural environment. Two such organizations are the Peace Corps (www.peacecorps.gov) and GLOBE (www.globe.gov). Peace Corps volunteers serve for two years in one region of their host country during which time they work to implement projects in their sector (e.g., health, environment, or education). By collaborating with groups of these volunteers over a wide range of locations, a greater understanding of the strengths and limitations of the working concept could be ascertained as it is applied to different situations while working with vastly different people groups. The second organization, GLOBE (Global Learning and Observation to Benefit the Environment), works in science education in primary and secondary schools around the world, part of which includes regularly performing environmental

169

measurements (e.g., rainfall, cloud cover, or temperature). Given their work in training students to perform these measures, collaboration with GLOBE would provide significant insights into effective methods of training local populations. Further, this would allow the exploration of utilizing students and their teachers to perform high-frequency sampling. Although these are only two organizations, they provide a window into the possibilities of additional groundwater quality research and further exploration of the working concept that would be possible in rural regions of developing countries given collaborations with established organizations. Although the focus of this research has primarily been the implementation of groundwater research in rural regions of developing nations, we believe that it is also applicable in developed nations. While researchers in developed countries typically have greater access to infrastructure and resources (i.e., financial, personnel, and to analytical equipment), limitations on resources do exist. Thus, we believe that using combinations of the three progressions may provide the possibility of enabling expansion of current research or development of new research programs in developed countries. For example, Granger, Indiana, is a rapidly developing region northeast of the University of Notre Dame. This area continues to rely predominantly on use of private wells and septic systems for their water and waste needs. Given the increasing density of the housing in this area, significant questions arise concerning the degradation of groundwater quality. As the monitoring of private wells and the regular cleaning of septic systems is the sole responsibility of the owner, no public record of results from such monitoring is available (further, there is significant doubt that either form of monitoring is regularly performed by any significant proportion of the population). In this situation, collaboration of public

170

health professionals with private well owners could enable the development of a highfrequency, in terms of time and space, water quality database for this region. This could provide significant insight into ongoing spatial / temporal changes in water quality and the effects of high density use of septic tanks and private wells. Such collaboration would provide the added benefit of engaging community members such that they might work to improve their water quality should contamination issues be identified. Taking one step further from the specifics of the dissertation research, we are able move beyond groundwater research to the wider applications of the working concept thus applying it to other environmental and public health fields. For example, at the present time, significant efforts are being invested to monitor the incidents rates and spread of diseases (e.g., malaria, tuberculosis, and HIV/AIDS) as well as to administer treatment for the diseases. Although a significant step beyond the current research, we believe that appropriate use of the progressions (as defined for these health issues) could enable the expansion of current medical monitoring programs and/or the administration of drug regimens. One organization already seeking, and implementing, new ways to administer drug regimens and monitor patient progress is Doctors without Borders (www.doctorswithoutborders.com). Through the work of such organizations, the use of the full range of progressions will help to expand monitoring capabilities, therefore allowing the collection of higher frequency medical information thus providing the data required to enable and expand treatment programs, based in part on local empowerment. Lastly, brief mention is here made of the significant amounts of sociological and anthropological research stemming from the working concept. Although not a focus of this dissertation, such research is a part of the larger Bénin project and is directly

171

connected to the working concept as use of the full range of the POE includes working with local populations. Paramount among the sociological and anthropological research are two questions: (i) how to best work with local populations to move those populations beyond simply serving as sampling personnel (where they only take measurements) to serve as collaborators in the monitoring process (whereby they would be able to interpret the collected data), and (ii) how to best implement future research programs (from the design of sampling programs, to interactions with local populations, to collaboration techniques) to promote program sustainability. Addressing these questions would then enable, through sustainable monitoring programs, significant additional areas of research focusing on empowerment of local populations to not only measure their water quality but also to understand water quality issues and make changes, through active monitoring of their water quality, to positively impact the quality of their water. While this dissertation research has demonstrated the power and efficacy of the working concept thus enabling the expansion to a wide array of fields, the primary focus remains groundwater quality research and development in rural regions of developing nations. It is the hope of the author of this dissertation that this research will significantly impact the focus and manner in which new and ongoing groundwater quality research and development projects are conducted in these settings through the use of the full range of the three progressions and incorporating collaboration with local populations.

172

APPENDIX 1 COMPLETE RESULTS OF THE MONTE CARLO STUDIES

173

TABLE A1 COMPLETE RESULTS OF THE MONTE CARLO STUDIES. 174

TABLE ARRANGED BY SCENARIO, THEN SAMPLING PERIOD AND SAMPLING LENGTHS, AND GROUPED BY PARAMETER.

174

TABLE A1 (contd.) Scenario: #1 Period

Mean Concentration Years

1

weekly

175

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.013 0.006 0.055 0.007 0.109 0.484 1.484 0.003 0.003 0.025 0.005 0.053 0.263 0.830 0.002 0.001 0.007 0.002 0.022 0.112 0.354 0.002 0.001 0.008 0.002 0.664 18.047 13.138

Bias 3.756 0.424 8.165 0.349 -2.521 1.231 4.981 3.904 0.359 8.544 0.144 -2.519 1.227 4.972 3.779 0.391 7.278 0.239 -2.522 1.227 4.976 4.068 0.368 8.264 0.468 -4.317 4.055 7.933

Bias2 14.118 0.186 66.712 0.129 6.461 1.998 26.297 15.244 0.132 73.020 0.026 6.397 1.768 25.553 14.283 0.154 52.973 0.059 6.380 1.618 25.113 16.550 0.137 68.290 0.221 19.301 34.487 76.071

Max Concentration MSE 14.131 0.192 66.766 0.136 6.570 2.482 27.781 15.247 0.135 73.045 0.030 6.451 2.031 26.383 14.285 0.155 52.981 0.061 6.402 1.730 25.467 16.552 0.139 68.298 0.223 19.964 52.534 89.209

Prec. 1.442 0.433 14.208 1.630 2.007 2.309 2.325 1.993 0.489 14.473 1.389 1.566 2.008 1.870 2.874 0.425 14.991 2.050 1.175 1.640 1.497 5.726 0.793 32.881 5.726 9.611 136.273 96.020

Bias 5.574 -0.156 20.027 2.004 -4.733 0.265 5.255 6.139 -0.083 24.313 2.160 -5.454 -0.441 4.557 7.572 -0.032 24.409 3.502 -6.325 -1.312 3.686 10.252 -0.085 33.445 6.652 -6.162 19.278 20.190

Bias2

MSE

32.505 0.457 415.241 5.646 24.404 2.380 29.934 39.681 0.496 605.560 6.055 31.313 2.203 22.631 60.198 0.426 610.732 14.313 41.175 3.361 15.083 110.825 0.800 1151.328 49.969 47.580 507.871 503.633

33.948 0.891 429.449 7.276 26.411 4.689 32.259 41.674 0.985 620.033 7.444 32.879 4.211 24.500 63.072 0.851 625.724 16.363 42.350 5.000 16.580 116.550 1.594 1184.209 55.695 57.190 644.144 599.652

TABLE A1 (contd.) Scenario: #1 Period

Mean Concentration Years

1

biweekly

176

2

5

10

1σ & Bias

Prec.

Bias

Bias2

Max Concentration MSE

Prec.

Bias

Bias2

MSE

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

0.022 0.008 0.050 0.021 0.147 0.555 1.603 0.012 0.009 0.036 0.014 0.073 0.302 0.891 0.004 0.003 0.015 0.005 0.029 0.129 0.385 0.002 0.001 0.006 0.002 0.014 0.071

4.177 0.405 8.011 0.388 -2.527 1.225 4.978 3.785 0.493 7.621 0.155 -2.520 1.228 4.977 4.096 0.369 8.753 0.738 -2.524 1.224 4.972 4.092 0.443 8.310 0.424 -2.526 1.226

17.466 0.172 64.226 0.172 6.533 2.056 26.382 14.336 0.251 58.114 0.038 6.422 1.810 25.655 16.776 0.139 76.618 0.549 6.396 1.627 25.102 16.741 0.198 69.049 0.182 6.393 1.575

17.489 0.181 64.276 0.193 6.680 2.611 27.985 14.348 0.260 58.150 0.052 6.495 2.112 26.546 16.781 0.142 76.633 0.555 6.425 1.756 25.487 16.743 0.199 69.055 0.184 6.406 1.646

1.676 0.494 8.812 2.586 2.009 2.102 2.282 2.786 0.657 9.836 1.722 1.547 1.547 1.547 2.557 0.525 8.627 3.926 1.176 1.270 1.242 2.314 0.577 11.098 2.183 1.124 1.638

4.764 -0.491 18.246 2.168 -4.758 0.225 5.209 5.530 -0.221 18.942 1.570 -5.452 -0.451 4.549 6.641 -0.437 21.483 3.737 -6.338 -1.335 3.665 6.651 -0.197 23.020 3.145 -6.917 -1.903

24.373 0.735 341.689 7.285 24.644 2.153 29.409 33.361 0.706 368.601 4.187 31.266 1.751 22.241 46.656 0.716 470.110 17.887 41.339 3.052 14.675 46.548 0.616 540.935 12.074 48.959 5.258

26.049 1.229 350.501 9.871 26.652 4.256 31.691 36.147 1.363 378.437 5.910 32.814 3.298 23.788 49.213 1.241 478.737 21.813 42.516 4.322 15.916 48.862 1.193 552.032 14.257 50.082 6.896

ts v3

0.222

4.979

25.003

25.225

1.482

3.096

11.063

12.545

TABLE A1 (contd.) Scenario: #1 Period

Mean Concentration Years

1

monthly

177

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.081 0.063 0.175 0.099 0.254 0.791 2.014 0.033 0.029 0.057 0.033 0.124 0.415 1.094 0.012 0.010 0.040 0.013 0.048 0.170 0.455 0.007 0.005 0.018 0.006 0.024 0.094 0.262

Bias 4.206 0.479 8.200 1.589 -2.524 1.229 4.983 4.234 0.458 7.540 0.462 -2.519 1.230 4.978 3.629 0.407 7.609 0.414 -2.519 1.230 4.978 3.524 0.390 7.753 0.778 -2.525 1.227 4.979

Bias2 17.772 0.293 67.401 2.625 6.626 2.301 26.839 17.956 0.239 56.896 0.247 6.469 1.927 25.877 13.181 0.175 57.928 0.185 6.390 1.682 25.234 12.423 0.156 60.126 0.611 6.398 1.599 25.053

Max Concentration MSE 17.854 0.356 67.577 2.724 6.880 3.092 28.853 17.989 0.268 56.952 0.280 6.592 2.342 26.971 13.193 0.185 57.968 0.198 6.438 1.852 25.689 12.429 0.161 60.144 0.617 6.422 1.693 25.314

Prec. 2.365 0.812 12.420 2.085 1.966 2.217 2.668 1.830 0.982 6.998 3.839 1.548 1.553 1.561 2.730 0.777 14.670 2.930 1.134 1.134 1.134 1.990 0.803 15.466 2.684 1.115 1.308 1.249

Bias 4.027 -0.963 18.639 1.492 -4.790 0.170 5.130 4.580 -0.696 14.653 2.002 -5.481 -0.481 4.519 5.063 -0.739 20.247 1.672 -6.308 -1.308 3.693 5.126 -0.700 22.550 2.569 -6.923 -1.917 3.082

Bias2 18.583 1.740 359.790 4.310 24.906 2.246 28.984 22.801 1.466 221.686 7.848 31.586 1.784 21.977 28.366 1.323 424.571 5.725 40.924 2.844 14.769 28.262 1.293 523.921 9.284 49.036 4.984 10.748

MSE 20.948 2.552 372.210 6.395 26.872 4.464 31.652 24.630 2.448 228.684 11.687 33.133 3.337 23.538 31.096 2.099 439.242 8.655 42.058 3.978 15.903 30.252 2.095 539.388 11.968 50.151 6.291 11.998

TABLE A1 (contd.) Scenario: #1 Period

Mean Concentration Years

1

semi-annual

178

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 1.198 1.456 2.400 1.066 2.487 5.624 10.436 0.499 0.715 0.780 0.856 1.163 2.674 5.034 0.232 0.234 0.527 0.263 0.429 1.019 1.960 0.114 0.111 0.169 0.118 0.207 0.499 0.978

Bias 1.907 0.220 1.052 -1.613 -2.519 1.233 4.985 1.681 0.704 6.161 1.608 -2.519 1.231 4.980 3.811 0.532 8.445 0.905 -2.512 1.241 4.993 4.122 0.593 8.893 0.707 -2.517 1.236 4.988

Bias2 4.834 1.504 3.508 3.667 8.831 7.145 35.288 3.323 1.210 38.737 3.441 7.507 4.188 29.833 14.751 0.517 71.841 1.081 6.737 2.558 26.888 17.105 0.462 79.250 0.617 6.540 2.026 25.855

Max Concentration MSE 6.031 2.960 5.908 4.733 11.318 12.769 45.724 3.822 1.925 39.517 4.297 8.670 6.861 34.867 14.983 0.750 72.367 1.344 7.166 3.577 28.848 17.219 0.573 79.418 0.735 6.746 2.524 26.833

Prec. 3.455 2.590 5.162 4.763 4.699 9.582 16.841 2.717 2.153 11.643 2.803 2.483 4.142 6.594 2.801 1.675 18.968 3.193 1.155 1.170 1.193 3.352 1.640 9.147 2.972 1.126 1.126 1.126

Bias -1.357 -3.281 3.356 -3.336 -5.991 -1.628 2.734 -1.237 -2.485 9.138 -1.219 -5.814 -0.984 3.847 1.920 -2.347 19.180 -0.826 -6.316 -1.317 3.682 3.348 -1.969 12.954 -0.654 -6.910 -1.910 3.091

Bias2 5.296 13.357 16.427 15.891 40.593 12.234 24.318 4.249 8.327 95.132 4.288 36.284 5.110 21.390 6.488 7.181 386.795 3.876 41.043 2.905 14.749 14.560 5.516 176.927 3.399 48.875 4.773 10.677

MSE 8.752 15.948 21.589 20.653 45.291 21.816 41.160 6.966 10.480 106.775 7.091 38.767 9.252 27.984 9.289 8.856 405.762 7.068 42.197 4.075 15.942 17.912 7.157 186.075 6.371 50.001 5.899 11.803

TABLE A1 (contd.) Scenario: #1 Period

Mean Concentration Years

1

season dependent

179

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.261 0.240 0.302 0.230 0.357 1.012 2.403 0.127 0.130 0.135 0.127 0.185 0.539 1.297 0.059 0.054 0.059 0.057 0.073 0.228 0.560 0.032 0.032 0.035 0.032 0.041 0.127 0.317

Bias 3.915 0.209 8.253 -0.017 -2.523 1.222 4.966 3.769 0.526 7.848 0.169 -2.523 1.225 4.973 4.318 0.345 7.493 0.456 -2.529 1.215 4.959 4.098 0.400 8.087 0.448 -2.524 1.231 4.985

Bias2 15.590 0.283 68.411 0.231 6.719 2.505 27.062 14.333 0.406 61.720 0.155 6.549 2.040 26.026 18.704 0.173 56.203 0.265 6.469 1.703 25.146 16.820 0.192 65.404 0.232 6.408 1.642 25.163

Max Concentration MSE 15.852 0.523 68.713 0.461 7.076 3.517 29.465 14.460 0.536 61.855 0.282 6.734 2.579 27.322 18.764 0.227 56.261 0.322 6.542 1.931 25.706 16.853 0.223 65.439 0.264 6.448 1.769 25.480

Prec. 2.429 0.770 7.216 2.787 2.035 2.104 2.236 1.735 0.569 9.502 1.735 1.542 1.542 1.542 2.162 0.587 8.771 3.561 1.159 1.346 1.289 3.773 0.619 10.078 2.307 1.117 1.460 1.355

Bias 5.120 -0.708 16.955 1.091 -4.723 0.265 5.253 5.390 -0.348 19.512 1.790 -5.474 -0.474 4.527 6.864 -0.432 20.107 3.662 -6.325 -1.320 3.680 7.423 -0.257 22.380 3.440 -6.918 -1.907 3.093

Bias2 28.642 1.271 294.651 3.977 24.341 2.174 29.825 30.787 0.690 390.198 4.939 31.505 1.766 22.033 49.272 0.773 413.029 16.968 41.167 3.088 14.826 58.852 0.685 510.759 14.134 48.953 5.094 10.919

MSE 31.071 2.041 301.866 6.764 26.375 4.278 32.061 32.522 1.259 399.700 6.674 33.046 3.308 23.574 51.434 1.361 421.800 20.529 42.327 4.435 16.115 62.624 1.304 520.837 16.441 50.071 6.554 12.275

TABLE A1 (contd.) Scenario: #1 Period

Mean Concentration Years

180

parameter dependent

1

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.058 0.058 0.155 0.065 0.255 0.574 1.438 0.034 0.026 0.071 0.031 0.133 0.321 0.824 0.015 0.011 0.041 0.014 0.050 0.153 0.407 0.007 0.005 0.024 0.007 0.024 0.083 0.232

Bias 3.717 0.700 7.400 0.661 -2.417 1.390 5.196 3.819 0.359 6.616 -0.190 -2.496 1.262 5.021 3.908 0.432 7.193 0.151 -2.490 1.274 5.038 3.813 0.396 7.380 0.866 -2.478 1.297 5.072

Bias2 13.873 0.548 54.908 0.502 6.097 2.504 28.432 14.618 0.154 43.835 0.067 6.364 1.914 26.028 15.288 0.197 51.776 0.037 6.251 1.775 25.784 14.543 0.163 54.476 0.757 6.166 1.765 25.957

Max Concentration MSE 13.931 0.606 55.064 0.567 6.352 3.078 29.870 14.652 0.180 43.906 0.098 6.497 2.236 26.852 15.303 0.208 51.817 0.051 6.301 1.928 26.191 14.550 0.168 54.500 0.764 6.190 1.848 26.189

Prec. 1.962 0.680 11.198 2.074 2.109 2.297 2.279 1.608 0.827 4.799 2.703 1.601 1.689 1.662 3.276 0.703 7.434 1.964 1.178 1.364 1.307 2.395 0.673 10.290 3.132 1.128 1.702 1.527

Bias 3.407 -0.627 18.241 0.552 -4.734 0.268 5.264 4.435 -0.818 12.795 0.688 -5.475 -0.472 4.528 5.733 -0.730 16.648 1.330 -6.337 -1.331 3.668 5.340 -0.775 18.820 2.916 -6.919 -1.903 3.094

Bias2 13.567 1.074 343.911 2.379 24.517 2.369 29.987 21.277 1.496 168.486 3.176 31.573 1.912 22.163 36.140 1.236 284.556 3.733 41.328 3.136 14.762 30.904 1.274 364.436 11.636 48.993 5.324 11.101

MSE 15.529 1.754 355.109 4.453 26.627 4.666 32.266 22.885 2.323 173.285 5.879 33.174 3.601 23.824 39.416 1.938 291.991 5.697 42.506 4.500 16.069 33.299 1.947 374.726 14.768 50.121 7.027 12.628

TABLE A1 (contd.) Scenario: #1 Period

TML2

TML v1 Years

1

weekly

181

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

4.88E+11 1.84E+11 1.70E+12 2.96E+11 3.74E+13 1.27E+15 9.28E+14 2.32E+11 8.99E+10 2.66E+12 3.14E+11 2.88E+13 8.23E+14 5.88E+14 4.10E+10 1.35E+10 1.66E+11 2.34E+10 2.88E+11 3.83E+12 3.10E+12 5.54E+10 1.02E+10 2.63E+11 2.05E+10 1.74E+11 1.13E+12 3.80E+12

Bias 1.62E+07 1.05E+06 3.23E+07 9.62E+05 -1.69E+07 2.48E+07 3.83E+07 1.58E+07 2.02E+06 2.85E+07 -2.51E+04 -1.72E+07 2.24E+07 3.71E+07 1.60E+07 1.73E+06 3.31E+07 1.55E+06 -1.60E+07 2.16E+07 3.54E+07 1.59E+07 1.42E+06 3.27E+07 1.09E+06 -1.06E+07 4.96E+06 2.05E+07

Bias2 2.63E+14 1.29E+12 1.04E+15 1.22E+12 3.24E+14 1.88E+15 2.39E+15 2.50E+14 4.16E+12 8.14E+14 3.15E+11 3.24E+14 1.32E+15 1.97E+15 2.57E+14 3.02E+12 1.10E+15 2.42E+12 2.57E+14 4.69E+14 1.25E+15 2.54E+14 2.03E+12 1.07E+15 1.21E+12 1.12E+14 2.58E+13 4.25E+14

MSE 2.63E+14 1.48E+12 1.05E+15 1.52E+12 3.61E+14 3.15E+15 3.32E+15 2.50E+14 4.25E+12 8.17E+14 6.30E+11 3.53E+14 2.15E+15 2.55E+15 2.57E+14 3.03E+12 1.10E+15 2.44E+12 2.58E+14 4.72E+14 1.26E+15 2.54E+14 2.04E+12 1.07E+15 1.23E+12 1.13E+14 2.69E+13 4.28E+14

Prec. 6.52E+11 1.13E+11 3.41E+12 1.73E+11 2.07E+12 8.48E+12 2.66E+13 2.95E+11 5.44E+10 1.75E+12 1.06E+11 9.92E+11 4.64E+12 1.50E+13 1.23E+11 2.03E+10 4.76E+11 4.26E+10 3.93E+11 1.99E+12 6.50E+12 1.13E+11 2.93E+10 5.35E+11 6.07E+10 1.13E+13 3.10E+14 2.26E+14

Bias 1.56E+07 1.87E+06 3.39E+07 1.56E+06 -1.03E+07 5.36E+06 2.10E+07 1.62E+07 1.60E+06 3.54E+07 6.98E+05 -1.03E+07 5.35E+06 2.10E+07 1.57E+07 1.73E+06 3.02E+07 1.10E+06 -1.03E+07 5.35E+06 2.10E+07 1.71E+07 1.76E+06 3.44E+07 2.16E+06 -1.76E+07 1.70E+07 3.30E+07

Bias2

MSE

2.45E+14 3.61E+12 1.15E+15 2.60E+12 1.08E+14 3.72E+13 4.69E+14 2.64E+14 2.61E+12 1.26E+15 5.92E+11 1.07E+14 3.32E+13 4.56E+14 2.48E+14 3.02E+12 9.11E+14 1.25E+12 1.07E+14 3.07E+13 4.49E+14 2.91E+14 3.11E+12 1.18E+15 4.73E+12 3.23E+14 5.99E+14 1.32E+15

2.45E+14 3.72E+12 1.15E+15 2.77E+12 1.10E+14 3.93E+13 4.95E+14 2.65E+14 2.66E+12 1.26E+15 6.98E+11 1.08E+14 3.42E+13 4.71E+14 2.48E+14 3.04E+12 9.12E+14 1.29E+12 1.07E+14 3.10E+13 4.55E+14 2.91E+14 3.14E+12 1.19E+15 4.79E+12 3.34E+14 6.10E+14 1.54E+15

TABLE A1 (contd.) Scenario: #1 Period

TML2

TML v1 Years

1

biweekly

182

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

1.97E+12 6.29E+11 3.95E+12 3.44E+12 3.95E+13 1.04E+15 7.73E+14 5.26E+11 3.00E+11 3.09E+12 4.79E+11 2.97E+13 7.99E+14 5.87E+14 1.98E+11 1.18E+11 9.02E+11 1.47E+11 1.15E+13 2.76E+14 2.03E+14 9.08E+10 5.36E+10 2.64E+11 6.75E+10 4.41E+11 5.40E+12

1.90E+07 1.29E+06 3.05E+07 -3.26E+06 -1.68E+07 2.33E+07 3.71E+07 1.59E+07 1.36E+06 2.77E+07 2.57E+06 -1.64E+07 2.01E+07 3.48E+07 1.55E+07 1.23E+06 2.95E+07 1.57E+06 -1.81E+07 1.70E+07 3.30E+07 1.54E+07 1.81E+06 3.15E+07 2.16E+06 -1.61E+07 2.15E+07

3.61E+14 2.29E+12 9.32E+14 1.41E+13 3.22E+14 1.58E+15 2.15E+15 2.54E+14 2.14E+12 7.72E+14 7.09E+12 3.00E+14 1.20E+15 1.80E+15 2.39E+14 1.63E+12 8.70E+14 2.61E+12 3.38E+14 5.66E+14 1.29E+15 2.37E+14 3.31E+12 9.94E+14 4.71E+12 2.60E+14 4.67E+14

3.63E+14 2.92E+12 9.36E+14 1.75E+13 3.62E+14 2.62E+15 2.92E+15 2.55E+14 2.44E+12 7.75E+14 7.57E+12 3.30E+14 2.00E+15 2.38E+15 2.40E+14 1.75E+12 8.71E+14 2.76E+12 3.50E+14 8.42E+14 1.50E+15 2.37E+14 3.37E+12 9.95E+14 4.78E+12 2.60E+14 4.73E+14

9.21E+11 1.50E+11 3.36E+12 4.93E+11 2.67E+12 9.71E+12 2.87E+13 4.46E+11 1.55E+11 1.86E+12 2.92E+11 1.35E+12 5.29E+12 1.59E+13 1.86E+11 5.49E+10 8.69E+11 1.15E+11 5.16E+11 2.28E+12 7.04E+12 8.19E+10 2.65E+10 4.13E+11 4.76E+10 2.44E+11 1.29E+12

ts v3

4.31E+12

3.53E+07

1.25E+15

1.25E+15

4.08E+12

Bias

Bias2

MSE

Prec.

Bias2

MSE

1.74E+07 1.79E+06 3.32E+07 1.72E+06 -1.03E+07 5.33E+06 2.10E+07 1.58E+07 2.15E+06 3.16E+07 7.41E+05 -1.03E+07 5.35E+06 2.10E+07 1.71E+07 1.64E+06 3.63E+07 3.18E+06 -1.03E+07 5.34E+06 2.10E+07 1.70E+07 1.95E+06 3.45E+07 1.86E+06 -1.03E+07 5.34E+06

3.03E+14 3.34E+12 1.11E+15 3.43E+12 1.10E+14 3.81E+13 4.70E+14 2.49E+14 4.79E+12 1.00E+15 8.40E+11 1.08E+14 3.39E+13 4.58E+14 2.91E+14 2.73E+12 1.32E+15 1.02E+13 1.07E+14 3.08E+13 4.48E+14 2.90E+14 3.83E+12 1.19E+15 3.52E+12 1.07E+14 2.98E+13

3.04E+14 3.49E+12 1.11E+15 3.93E+12 1.12E+14 4.08E+13 4.98E+14 2.49E+14 4.95E+12 1.00E+15 1.13E+12 1.09E+14 3.53E+13 4.74E+14 2.91E+14 2.78E+12 1.32E+15 1.03E+13 1.08E+14 3.13E+13 4.56E+14 2.90E+14 3.86E+12 1.19E+15 3.57E+12 1.07E+14 3.01E+13

2.10E+07

4.46E+14

4.50E+14

Bias

TABLE A1 (contd.) Scenario: #1 Period

TML2

TML v1 Years

1

monthly

183

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

3.83E+12 2.54E+12 2.88E+13 5.16E+12 4.48E+13 1.29E+15 9.48E+14 1.58E+12 1.24E+12 7.03E+12 2.20E+12 3.48E+13 7.71E+14 5.57E+14 6.31E+11 4.93E+11 1.20E+12 6.48E+11 1.42E+13 3.29E+14 2.44E+14 1.31E+13 1.38E+13 1.23E+13 1.39E+13 1.49E+13 1.96E+13 1.72E+13

Bias 1.84E+07 1.21E+06 1.84E+07 5.26E+06 -1.72E+07 2.45E+07 3.80E+07 1.47E+07 8.47E+05 3.77E+07 -2.07E+06 -1.75E+07 1.96E+07 3.46E+07 1.50E+07 1.09E+06 3.23E+07 -8.16E+05 -1.83E+07 1.94E+07 3.49E+07 2.49E+07 1.13E+07 3.82E+07 1.14E+07 -6.50E+06 3.11E+07 4.48E+07

Bias2 3.44E+14 4.00E+12 3.67E+14 3.29E+13 3.42E+14 1.89E+15 2.39E+15 2.19E+14 1.95E+12 1.42E+15 6.49E+12 3.40E+14 1.15E+15 1.76E+15 2.26E+14 1.68E+12 1.05E+15 1.31E+12 3.49E+14 7.06E+14 1.46E+15 6.31E+14 1.42E+14 1.47E+15 1.45E+14 5.71E+13 9.84E+14 2.03E+15

MSE 3.48E+14 6.54E+12 3.96E+14 3.80E+13 3.87E+14 3.17E+15 3.34E+15 2.20E+14 3.19E+12 1.43E+15 8.69E+12 3.75E+14 1.92E+15 2.31E+15 2.26E+14 2.18E+12 1.05E+15 1.96E+12 3.64E+14 1.04E+15 1.71E+15 6.44E+14 1.56E+14 1.48E+15 1.59E+14 7.20E+13 1.00E+15 2.05E+15

Prec. 2.04E+12 1.13E+12 6.93E+12 1.94E+12 4.58E+12 1.39E+13 3.61E+13 9.28E+11 5.32E+11 2.37E+12 7.01E+11 2.19E+12 7.33E+12 1.97E+13 3.17E+11 1.82E+11 1.38E+12 2.71E+11 8.50E+11 3.00E+12 8.22E+12 1.69E+11 8.57E+10 6.91E+11 1.33E+11 4.18E+11 1.68E+12 4.78E+12

Bias 1.75E+07 2.09E+06 3.40E+07 6.72E+06 -1.03E+07 5.34E+06 2.10E+07 1.76E+07 2.00E+06 3.12E+07 2.02E+06 -1.03E+07 5.34E+06 2.10E+07 1.51E+07 1.78E+06 3.15E+07 1.82E+06 -1.03E+07 5.36E+06 2.10E+07 1.47E+07 1.72E+06 3.21E+07 3.34E+06 -1.03E+07 5.33E+06 2.10E+07

Bias2

MSE

3.08E+14 5.48E+12 1.16E+15 4.71E+13 1.11E+14 4.24E+13 4.78E+14 3.11E+14 4.53E+12 9.78E+14 4.79E+12 1.09E+14 3.59E+13 4.61E+14 2.28E+14 3.37E+12 9.97E+14 3.57E+12 1.07E+14 3.17E+13 4.51E+14 2.15E+14 3.03E+12 1.03E+15 1.13E+13 1.07E+14 3.01E+13 4.46E+14

3.10E+14 6.61E+12 1.17E+15 4.91E+13 1.16E+14 4.70E+13 5.14E+14 3.12E+14 5.06E+12 9.81E+14 5.49E+12 1.11E+14 3.81E+13 4.81E+14 2.29E+14 3.55E+12 9.98E+14 3.84E+12 1.08E+14 3.25E+13 4.59E+14 2.15E+14 3.12E+12 1.03E+15 1.14E+13 1.08E+14 3.05E+13 4.51E+14

TABLE A1 (contd.) Scenario: #1 Period

TML2

TML v1 Years

1

semi-annual

184

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

4.27E+13 3.55E+13 5.64E+13 4.23E+13 3.81E+14 2.39E+14 2.42E+14 2.18E+13 1.60E+13 7.58E+13 2.57E+13 2.86E+14 1.84E+14 1.85E+14 6.53E+12 6.31E+12 9.23E+12 6.95E+12 1.37E+14 8.72E+13 8.93E+13 2.89E+14 2.91E+14 2.94E+14 2.90E+14 4.09E+14 3.93E+14 3.94E+14

Bias 1.92E+07 9.02E+05 3.98E+07 4.40E+06 -5.57E+07 -4.95E+07 -4.74E+07 1.88E+07 -8.33E+05 5.26E+07 6.90E+06 -5.72E+07 -5.13E+07 -4.90E+07 1.28E+07 -4.76E+05 3.26E+07 -4.00E+05 -5.52E+07 -5.03E+07 -4.78E+07 -5.44E+08 -5.60E+08 -5.26E+08 -5.59E+08 -6.18E+08 -6.12E+08 -6.09E+08

Bias2 4.13E+14 3.63E+13 1.64E+15 6.16E+13 3.49E+15 2.69E+15 2.49E+15 3.77E+14 1.67E+13 2.84E+15 7.33E+13 3.55E+15 2.81E+15 2.59E+15 1.70E+14 6.53E+12 1.07E+15 7.11E+12 3.18E+15 2.62E+15 2.38E+15 2.96E+17 3.13E+17 2.77E+17 3.13E+17 3.82E+17 3.74E+17 3.72E+17

MSE 4.56E+14 7.18E+13 1.69E+15 1.04E+14 3.87E+15 2.93E+15 2.73E+15 3.98E+14 3.27E+13 2.91E+15 9.90E+13 3.84E+15 3.00E+15 2.77E+15 1.76E+14 1.28E+13 1.08E+15 1.41E+13 3.32E+15 2.70E+15 2.47E+15 2.96E+17 3.14E+17 2.77E+17 3.13E+17 3.82E+17 3.75E+17 3.72E+17

Prec. 2.14E+13 2.52E+13 5.77E+13 2.14E+13 4.29E+13 9.72E+13 1.82E+14 8.87E+12 1.24E+13 1.90E+13 1.49E+13 2.02E+13 4.63E+13 8.77E+13 4.28E+12 4.09E+12 1.28E+13 4.71E+12 7.42E+12 1.77E+13 3.42E+13 2.85E+14 2.88E+14 2.82E+14 2.87E+14 2.92E+14 2.67E+14 2.45E+14

Bias 7.78E+06 8.48E+05 4.18E+06 -6.73E+06 -1.05E+07 5.13E+06 2.07E+07 6.84E+06 2.86E+06 2.54E+07 6.62E+06 -1.05E+07 5.13E+06 2.07E+07 1.57E+07 2.14E+06 3.49E+07 3.69E+06 -1.04E+07 5.17E+06 2.08E+07 -3.58E+08 -3.72E+08 -3.38E+08 -3.72E+08 -3.85E+08 -3.69E+08 -3.54E+08

Bias2

MSE

8.20E+13 2.59E+13 7.52E+13 6.67E+13 1.52E+14 1.24E+14 6.11E+14 5.57E+13 2.06E+13 6.64E+14 5.87E+13 1.30E+14 7.26E+13 5.17E+14 2.51E+14 8.68E+12 1.23E+15 1.83E+13 1.17E+14 4.44E+13 4.66E+14 1.28E+17 1.39E+17 1.14E+17 1.38E+17 1.49E+17 1.37E+17 1.25E+17

1.03E+14 5.11E+13 1.33E+14 8.81E+13 1.95E+14 1.67E+14 7.93E+14 6.45E+13 3.30E+13 6.83E+14 7.36E+13 1.50E+14 9.28E+13 6.05E+14 2.55E+14 1.28E+13 1.24E+15 2.30E+13 1.24E+14 5.18E+13 5.01E+14 1.28E+17 1.39E+17 1.15E+17 1.39E+17 1.49E+17 1.37E+17 1.26E+17

TABLE A1 (contd.) Scenario: #1 Period

TML2

TML v1 Years

1

season dependent

185

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

7.88E+12 3.02E+12 5.23E+13 2.17E+13 3.87E+13 1.33E+15 9.51E+14 4.17E+12 1.61E+12 5.69E+13 4.42E+12 3.31E+13 8.09E+14 5.80E+14 2.04E+12 5.52E+11 6.61E+12 1.63E+12 1.28E+13 2.02E+14 1.53E+14 2.97E+11 2.52E+11 6.49E+11 2.60E+11 8.53E+11 9.98E+12 8.00E+12

Bias 1.83E+07 8.31E+05 2.68E+07 3.49E+06 -1.69E+07 2.71E+07 4.08E+07 1.56E+07 1.52E+06 3.51E+07 3.32E+06 -1.71E+07 1.93E+07 3.45E+07 1.47E+07 1.90E+06 3.97E+07 9.61E+05 -1.79E+07 1.46E+07 3.10E+07 1.76E+07 2.05E+06 3.18E+07 1.33E+06 -1.59E+07 2.18E+07 3.56E+07

Bias2 3.43E+14 3.71E+12 7.73E+14 3.39E+13 3.26E+14 2.06E+15 2.62E+15 2.48E+14 3.92E+12 1.29E+15 1.55E+13 3.27E+14 1.18E+15 1.77E+15 2.17E+14 4.16E+12 1.59E+15 2.55E+12 3.32E+14 4.14E+14 1.12E+15 3.10E+14 4.47E+12 1.01E+15 2.04E+12 2.55E+14 4.84E+14 1.27E+15

MSE 3.51E+14 6.73E+12 8.25E+14 5.56E+13 3.64E+14 3.39E+15 3.57E+15 2.52E+14 5.54E+12 1.34E+15 1.99E+13 3.60E+14 1.99E+15 2.35E+15 2.19E+14 4.71E+12 1.59E+15 4.18E+12 3.45E+14 6.17E+14 1.27E+15 3.10E+14 4.72E+12 1.01E+15 2.30E+12 2.56E+14 4.94E+14 1.28E+15

Prec. 4.99E+12 4.14E+12 7.75E+12 4.03E+12 6.33E+12 1.76E+13 4.25E+13 2.41E+12 2.25E+12 3.51E+12 2.23E+12 3.27E+12 9.45E+12 2.32E+13 1.10E+12 9.26E+11 1.38E+12 1.00E+12 1.31E+12 4.01E+12 1.00E+13 5.61E+11 5.43E+11 7.27E+11 5.40E+11 7.38E+11 2.26E+12 5.68E+12

Bias 1.63E+07 9.56E+05 3.42E+07 2.24E+04 -1.03E+07 5.31E+06 2.09E+07 1.57E+07 2.28E+06 3.25E+07 7.91E+05 -1.03E+07 5.33E+06 2.10E+07 1.80E+07 1.53E+06 3.11E+07 1.99E+06 -1.04E+07 5.29E+06 2.09E+07 1.70E+07 1.76E+06 3.35E+07 1.96E+06 -1.03E+07 5.35E+06 2.10E+07

Bias2

MSE

2.70E+14 5.05E+12 1.18E+15 4.03E+12 1.13E+14 4.57E+13 4.81E+14 2.48E+14 7.46E+12 1.06E+15 2.85E+12 1.10E+14 3.79E+13 4.64E+14 3.24E+14 3.26E+12 9.67E+14 4.97E+12 1.09E+14 3.20E+13 4.49E+14 2.91E+14 3.64E+12 1.12E+15 4.36E+12 1.08E+14 3.09E+13 4.48E+14

2.75E+14 9.19E+12 1.19E+15 8.06E+12 1.19E+14 5.20E+13 5.23E+14 2.51E+14 9.70E+12 1.06E+15 5.08E+12 1.13E+14 4.11E+13 4.87E+14 3.25E+14 4.19E+12 9.68E+14 5.97E+12 1.10E+14 3.33E+13 4.59E+14 2.91E+14 4.18E+12 1.13E+15 4.90E+12 1.08E+14 3.16E+13 4.54E+14

TABLE A1 (contd.) Scenario: #1 Period

Years

1

186

parameter dependent

TML2

TML v1

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

3.70E+12 2.63E+12 3.59E+13 4.18E+12 4.07E+13 1.35E+15 9.79E+14 3.19E+12 1.31E+12 9.31E+12 3.01E+12 3.34E+13 1.01E+15 7.31E+14 8.28E+11 5.07E+11 2.64E+12 8.08E+11 1.37E+13 2.44E+14 1.83E+14 1.31E+13 1.41E+13 1.43E+13 1.41E+13 1.49E+13 1.91E+13 1.68E+13

Bias 1.71E+07 1.31E+06 3.14E+07 -1.17E+06 -1.71E+07 2.47E+07 3.84E+07 1.71E+07 4.77E+05 3.71E+07 2.24E+06 -1.73E+07 2.28E+07 3.71E+07 1.40E+07 7.47E+05 3.26E+07 -8.12E+05 -1.82E+07 1.51E+07 3.13E+07 2.56E+07 1.08E+07 9.85E+06 1.06E+07 -6.57E+06 3.08E+07 4.47E+07

Bias2 2.97E+14 4.34E+12 1.02E+15 5.55E+12 3.32E+14 1.96E+15 2.46E+15 2.97E+14 1.53E+12 1.39E+15 8.02E+12 3.33E+14 1.53E+15 2.11E+15 1.98E+14 1.07E+12 1.07E+15 1.47E+12 3.43E+14 4.72E+14 1.16E+15 6.66E+14 1.30E+14 1.11E+14 1.27E+14 5.80E+13 9.70E+14 2.01E+15

MSE 3.00E+14 6.97E+12 1.06E+15 9.73E+12 3.73E+14 3.31E+15 3.44E+15 3.00E+14 2.84E+12 1.40E+15 1.10E+13 3.66E+14 2.54E+15 2.84E+15 1.99E+14 1.57E+12 1.07E+15 2.27E+12 3.57E+14 7.16E+14 1.35E+15 6.79E+14 1.44E+14 1.26E+14 1.41E+14 7.29E+13 9.89E+14 2.03E+15

Prec. 2.59E+12 1.86E+12 6.55E+12 2.12E+12 4.59E+12 9.99E+12 2.57E+13 3.71E+12 3.21E+12 5.34E+12 3.46E+12 4.61E+12 6.84E+12 1.36E+13 2.38E+12 2.40E+12 2.87E+12 2.34E+12 2.79E+12 4.32E+12 7.98E+12 1.76E+12 1.68E+12 2.17E+12 1.73E+12 1.93E+12 2.87E+12 4.95E+12

Bias 1.66E+07 2.89E+06 3.09E+07 3.81E+06 -9.88E+06 6.00E+06 2.19E+07 1.68E+07 1.67E+06 2.75E+07 -6.97E+05 -1.03E+07 5.43E+06 2.11E+07 1.62E+07 1.99E+06 2.97E+07 7.18E+05 -1.02E+07 5.55E+06 2.13E+07 1.59E+07 1.84E+06 3.05E+07 3.98E+06 -1.01E+07 5.65E+06 2.14E+07

Bias2

MSE

2.78E+14 1.02E+13 9.62E+14 1.66E+13 1.02E+14 4.60E+13 5.05E+14 2.87E+14 6.02E+12 7.62E+14 3.94E+12 1.10E+14 3.63E+13 4.59E+14 2.65E+14 6.37E+12 8.83E+14 2.86E+12 1.07E+14 3.51E+13 4.61E+14 2.55E+14 5.04E+12 9.29E+14 1.76E+13 1.04E+14 3.48E+13 4.64E+14

2.80E+14 1.21E+13 9.68E+14 1.87E+13 1.07E+14 5.06E+13 5.30E+14 2.91E+14 9.23E+12 7.67E+14 7.40E+12 1.14E+14 4.09E+13 4.73E+14 2.67E+14 8.77E+12 8.86E+14 5.21E+12 1.09E+14 3.79E+13 4.69E+14 2.56E+14 6.72E+12 9.31E+14 1.93E+13 1.06E+14 3.67E+13 4.69E+14

TABLE A1 (contd.) Scenario: #2 Period

Mean Concentration Years

1

weekly

187

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.077 0.021 0.625 0.051 0.381 4.574 3.792 0.020 0.010 0.274 0.023 0.197 2.468 2.062 0.009 0.004 0.053 0.018 0.081 1.011 0.828 0.004 0.002 0.017 0.004 0.039 0.532 0.425

Bias 3.754 0.413 9.752 0.547 -2.969 3.378 6.045 4.052 0.303 7.964 0.129 -2.963 3.395 6.059 3.978 0.414 8.426 0.249 -2.976 3.424 6.093 4.098 0.427 8.188 0.449 -2.982 3.372 6.039

Bias2 14.172 0.191 95.714 0.350 9.194 15.983 40.333 16.437 0.101 63.700 0.040 8.974 13.992 38.776 15.832 0.175 71.045 0.080 8.938 12.736 37.955 16.790 0.184 67.035 0.205 8.926 11.901 36.881

Max Concentration MSE 14.249 0.212 96.339 0.401 9.575 20.556 44.124 16.457 0.111 63.974 0.064 9.171 16.460 40.838 15.841 0.179 71.098 0.097 9.020 13.748 38.782 16.794 0.186 67.052 0.209 8.966 12.433 37.307

Prec. 8.203 2.017 96.962 10.847 14.368 71.023 52.039 10.985 2.081 86.617 8.343 13.971 19.830 17.805 11.898 2.229 45.366 11.945 11.667 11.667 11.667 11.127 2.412 59.234 20.970 9.364 9.364 9.364

Bias 5.645 -0.701 26.317 3.039 -7.986 19.929 20.347 6.632 -0.840 27.991 2.386 -9.468 20.320 20.363 8.096 -0.453 30.763 4.439 -12.156 17.847 17.847 9.526 -0.136 32.629 7.648 -13.860 16.150 16.150

Bias2

MSE

40.061 2.508 789.500 20.085 78.137 468.161 465.992 54.968 2.786 870.051 14.034 103.607 432.686 432.411 77.442 2.434 991.635 31.652 159.427 330.140 330.140 101.839 2.431 1123.482 79.440 201.402 270.108 270.108

48.264 4.525 886.462 30.932 92.505 539.184 518.030 65.953 4.866 956.668 22.377 117.578 452.516 450.217 89.340 4.662 1037.001 43.596 171.094 341.807 341.807 112.966 4.843 1182.716 100.410 210.766 279.471 279.471

TABLE A1 (contd.) Scenario: #2 Period

Mean Concentration Years

1

biweekly

188

2

5

10

1σ & Bias

Prec.

Bias

Bias2

Max Concentration MSE

Prec.

Bias

Bias2

MSE

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

0.140 0.089 0.651 0.242 0.482 5.369 4.459 0.080 0.040 0.313 0.061 0.246 2.887 2.408 0.031 0.016 0.151 0.031 0.100 1.148 0.946 0.011 0.007 0.050 0.011 0.048 0.606

3.956 0.351 10.445 0.513 -2.974 3.367 6.033 3.731 0.398 7.234 0.527 -2.974 3.399 6.071 3.796 0.528 7.388 0.195 -2.976 3.400 6.073 4.072 0.401 8.529 0.471 -2.981 3.372

15.787 0.212 109.734 0.505 9.327 16.706 40.857 13.996 0.199 52.640 0.339 9.087 14.440 39.261 14.438 0.295 54.721 0.069 8.959 12.703 37.826 16.589 0.168 72.782 0.233 8.931 11.976

15.927 0.300 110.385 0.747 9.809 22.076 45.316 14.076 0.239 52.953 0.400 9.332 17.327 41.669 14.468 0.310 54.872 0.099 9.059 13.851 38.771 16.600 0.175 72.832 0.245 8.979 12.583

13.360 3.455 49.732 13.257 14.235 104.252 74.775 12.091 3.286 54.951 10.541 14.121 31.802 25.932 13.092 3.274 62.652 13.092 11.787 11.937 11.887 14.838 3.875 103.130 14.838 9.355 9.355

4.774 -1.902 21.624 -0.246 -8.477 18.213 18.875 4.673 -1.712 19.252 1.383 -9.633 19.772 19.892 5.457 -1.244 23.884 1.857 -12.092 17.906 17.907 7.958 -1.050 35.180 4.358 -13.861 16.144

36.148 7.071 517.283 13.318 86.088 435.919 430.994 33.926 6.218 425.563 12.454 106.909 422.707 421.577 42.868 4.820 633.061 16.539 157.979 332.545 332.531 78.164 4.976 1340.566 33.825 201.460 269.933

49.508 10.526 567.014 26.576 100.323 540.170 505.769 46.017 9.504 480.514 22.994 121.030 454.508 447.510 55.960 8.094 695.713 29.631 169.766 344.482 344.418 93.002 8.851 1443.697 48.663 210.815 279.288

ts v3

0.489

6.038

36.942

37.430

9.355

16.144

269.933

279.288

TABLE A1 (contd.) Scenario: #2 Period

Mean Concentration Years

1

monthly

189

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.847 0.417 1.456 0.556 0.810 7.690 6.416 0.235 0.179 0.600 0.265 0.383 4.072 3.393 0.089 0.065 0.277 0.087 0.147 1.571 1.303 0.040 0.033 0.105 0.038 0.072 0.804 0.658

Bias 4.891 0.454 8.702 0.631 -2.974 3.367 6.033 4.424 0.354 9.284 0.195 -2.956 3.462 6.128 4.036 0.450 7.072 0.090 -2.976 3.409 6.082 4.167 0.435 7.904 0.113 -2.979 3.369 6.037

Bias2 24.771 0.624 77.169 0.953 9.653 19.023 42.811 19.806 0.304 86.781 0.303 9.122 16.057 40.941 16.372 0.267 50.280 0.095 9.001 13.194 38.284 17.404 0.223 62.576 0.050 8.948 12.154 37.099

Max Concentration MSE 25.617 1.041 78.625 1.509 10.463 26.713 49.227 20.041 0.482 87.381 0.567 9.505 20.130 44.334 16.461 0.332 50.556 0.182 9.148 14.764 39.587 17.444 0.256 62.682 0.088 9.020 12.957 37.757

Prec. 21.743 6.902 57.456 14.842 15.894 162.747 115.920 14.845 5.683 63.052 14.474 14.043 59.277 44.766 18.004 5.741 49.615 14.940 11.473 12.404 12.098 18.004 5.509 59.665 24.386 9.351 9.351 9.351

Bias 3.999 -3.234 12.340 -1.791 -9.416 14.931 16.060 3.146 -3.320 16.842 -2.034 -10.065 18.487 18.777 4.223 -2.852 18.628 -0.445 -12.078 17.895 17.901 5.478 -2.665 21.770 2.200 -13.862 16.141 16.141

Bias2 37.736 17.362 209.720 18.050 104.539 385.647 373.831 24.744 16.705 346.681 18.611 115.346 401.028 397.324 35.832 13.876 396.588 15.138 157.343 332.594 332.503 48.005 12.609 533.547 29.226 201.486 269.857 269.857

MSE 59.479 24.263 267.176 32.892 120.433 548.394 489.751 39.589 22.388 409.733 33.085 129.389 460.305 442.090 53.837 19.617 446.203 30.078 168.817 344.997 344.600 66.009 18.118 593.212 53.611 210.837 279.208 279.208

TABLE A1 (contd.) Scenario: #2 Period

Mean Concentration Years

1

semi-annual

190

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 13.272 12.857 22.890 14.958 9.409 65.586 57.447 7.413 5.528 3.503 4.874 4.391 31.406 27.603 2.066 2.140 4.339 2.819 1.621 13.111 11.065 1.242 1.148 1.674 1.242 0.864 8.006 6.428

Bias 3.337 0.597 11.678 1.363 -2.934 3.471 6.125 5.202 0.388 4.604 -0.797 -3.053 3.067 5.806 3.570 0.283 11.133 0.118 -3.018 3.212 5.924 4.278 0.433 8.932 0.678 -2.982 3.278 5.978

Bias2 24.408 13.213 159.247 16.815 18.014 77.636 94.956 34.469 5.678 24.698 5.509 13.709 40.812 61.315 14.812 2.220 128.281 2.833 10.729 23.427 46.150 19.546 1.336 81.440 1.702 9.755 18.748 42.165

Max Concentration MSE 37.681 26.070 182.137 31.773 27.423 143.222 152.403 41.882 11.205 28.200 10.383 18.100 72.218 88.917 16.878 4.359 132.620 5.653 12.350 36.538 57.214 20.788 2.484 83.115 2.944 10.619 26.755 48.593

Prec. 51.249 38.714 61.215 43.369 32.930 240.563 193.265 35.744 27.736 32.840 29.946 31.738 283.435 212.334 32.912 23.528 77.194 57.400 28.623 257.564 190.498 28.916 20.364 74.151 28.916 19.279 139.327 104.431

Bias -7.863 -10.377 2.279 -9.166 -15.492 -4.672 -1.334 -4.178 -9.624 -6.384 -11.631 -15.672 -1.020 1.962 -5.092 -9.118 9.017 -6.610 -15.163 7.144 8.683 -3.275 -8.682 8.688 -6.875 -15.087 11.808 12.429

Bias2 113.069 146.391 66.410 127.374 272.909 262.387 195.046 53.201 120.349 73.597 165.220 277.331 284.475 216.181 58.835 106.658 158.486 101.094 258.512 308.594 265.877 39.639 95.734 149.619 76.177 246.883 278.739 258.907

MSE 164.319 185.105 127.625 170.744 305.839 502.951 388.312 88.945 148.085 106.438 195.166 309.070 567.910 428.515 91.747 130.186 235.681 158.494 287.134 566.157 456.375 68.555 116.098 223.769 105.094 266.162 418.066 363.338

TABLE A1 (contd.) Scenario: #2 Period

Mean Concentration Years

1

season dependent

191

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 1.357 1.238 3.122 1.201 1.047 18.025 14.286 0.728 0.615 0.769 0.643 0.503 8.954 7.171 0.258 0.249 0.281 0.615 0.198 3.524 2.812 0.161 0.160 0.227 0.166 0.118 2.054 1.641

Bias 6.640 3.075 13.461 2.211 -0.973 7.750 10.668 7.234 3.044 10.774 3.233 -0.960 7.770 10.703 6.744 3.100 10.047 3.044 -0.972 7.767 10.699 6.901 3.188 10.694 3.169 -0.945 7.757 10.701

Bias2 45.438 10.695 184.311 6.089 1.994 78.084 128.083 53.055 9.880 116.829 11.096 1.425 69.321 121.718 45.741 9.857 101.213 9.880 1.143 63.839 117.267 47.761 10.323 114.550 10.204 1.012 62.207 116.109

Max Concentration MSE 46.795 11.933 187.433 7.290 3.041 96.109 142.369 53.784 10.495 117.599 11.740 1.928 78.276 128.889 45.998 10.106 101.493 10.495 1.342 67.363 120.080 47.922 10.483 114.776 10.370 1.130 64.261 117.751

Prec. 12.336 2.650 77.942 12.414 14.158 83.935 60.700 8.820 2.266 57.772 13.577 14.292 21.747 19.178 16.626 2.223 54.641 2.266 11.884 11.884 11.884 17.973 3.010 57.041 11.950 9.364 9.364 9.364

Bias 5.724 -0.925 29.103 1.954 -8.088 19.328 19.845 7.321 -0.820 24.655 3.966 -9.522 20.208 20.263 8.843 -0.567 28.994 -0.820 -12.163 17.840 17.840 10.846 -0.390 30.423 4.625 -13.860 16.150 16.150

Bias2 45.099 3.507 924.865 16.231 79.563 457.451 454.488 62.406 2.938 665.605 29.304 104.953 430.085 429.722 94.811 2.544 895.225 2.938 159.804 330.124 330.124 135.571 3.162 982.285 33.334 201.402 270.108 270.108

MSE 57.435 6.157 1002.807 28.646 93.721 541.386 515.188 71.227 5.204 723.378 42.881 119.246 451.832 448.899 111.437 4.767 949.866 5.204 171.688 342.008 342.008 153.544 6.173 1039.327 45.284 210.766 279.471 279.471

TABLE A1 (contd.) Scenario: #2 Period

Mean Concentration Years

192

parameter dependent

1

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.756 0.531 1.594 0.645 1.739 9.972 8.490 0.377 0.272 0.811 0.421 0.747 5.376 4.525 0.134 0.111 0.398 0.158 0.285 2.339 1.922 0.060 0.054 0.150 0.066 0.114 1.167 0.884

Bias 5.232 1.278 8.557 1.062 -2.223 5.119 7.808 5.407 1.327 6.937 1.165 -2.150 5.333 8.037 4.629 1.281 8.041 1.232 -2.187 5.231 7.958 4.878 1.360 9.032 1.166 -1.903 5.929 8.630

Bias2 28.125 2.165 74.813 1.772 6.679 36.173 69.456 29.606 2.033 48.929 1.777 5.370 33.811 69.110 21.556 1.752 65.056 1.677 5.068 29.697 65.241 23.850 1.903 81.722 1.425 3.735 36.315 75.345

Max Concentration MSE 28.881 2.696 76.406 2.417 8.418 46.146 77.946 29.983 2.305 49.740 2.198 6.117 39.187 73.635 21.690 1.863 65.454 1.836 5.353 32.036 67.163 23.910 1.957 81.873 1.491 3.849 37.482 76.229

Prec. 11.315 5.227 117.364 15.950 17.021 97.436 71.789 16.878 4.958 36.250 15.600 15.136 28.212 23.974 13.040 5.267 47.839 17.511 11.812 11.887 11.862 15.485 5.126 55.197 14.888 9.351 9.351 9.351

Bias 3.315 -2.432 17.558 -0.413 -8.066 19.259 19.795 6.053 -2.242 13.462 -0.071 -9.584 20.007 20.089 4.600 -2.087 20.850 1.115 -12.153 17.847 17.848 6.361 -1.684 27.855 2.732 -13.862 16.141 16.141

Bias2 22.300 11.139 425.627 16.120 82.077 468.311 463.577 53.511 9.986 217.449 15.605 106.972 428.446 427.516 34.199 9.624 482.530 18.755 159.503 330.373 330.366 55.948 7.961 831.036 22.349 201.486 269.857 269.857

MSE 33.615 16.366 542.991 32.070 99.098 565.747 535.366 70.389 14.945 253.698 31.205 122.108 456.658 451.489 47.239 14.890 530.369 36.266 171.315 342.259 342.227 71.433 13.087 886.233 37.237 210.837 279.208 279.208

TABLE A1 (contd.) TML1

Scenario: #2 Period

Years

1

weekly

193

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

2.88E+12 4.72E+11 3.51E+13 1.27E+12 1.41E+13 1.44E+14 1.30E+14 1.46E+12 2.04E+11 1.15E+13 6.67E+11 7.39E+12 7.83E+13 7.10E+13 5.39E+11 7.54E+10 3.43E+12 5.04E+11 3.07E+12 3.24E+13 2.89E+13 2.68E+11 3.76E+10 1.50E+12 9.35E+10 1.49E+12 1.72E+13 1.50E+13

Bias 1.45E+07 9.12E+05 4.03E+07 1.53E+06 -1.45E+07 1.68E+07 2.82E+07 1.59E+07 3.81E+05 3.20E+07 -3.89E+05 -1.44E+07 1.68E+07 2.83E+07 1.56E+07 9.19E+05 3.41E+07 1.37E+05 -1.45E+07 1.70E+07 2.85E+07 1.62E+07 1.06E+06 3.30E+07 1.17E+06 -1.44E+07 1.68E+07 2.83E+07

TML2 Bias2

MSE

Prec.

2.15E+14 1.30E+12 1.66E+15 3.60E+12 2.23E+14 4.25E+14 9.25E+14 2.55E+14 3.49E+11 1.03E+15 8.18E+11 2.16E+14 3.62E+14 8.70E+14 2.43E+14 9.20E+11 1.17E+15 5.23E+11 2.14E+14 3.21E+14 8.39E+14 2.62E+14 1.16E+12 1.09E+15 1.46E+12 2.10E+14 3.00E+14 8.13E+14

2.17E+14 1.78E+12 1.69E+15 4.88E+12 2.37E+14 5.69E+14 1.06E+15 2.56E+14 5.53E+11 1.04E+15 1.48E+12 2.23E+14 4.40E+14 9.41E+14 2.44E+14 9.95E+11 1.17E+15 1.03E+12 2.17E+14 3.53E+14 8.67E+14 2.62E+14 1.20E+12 1.09E+15 1.55E+12 2.12E+14 3.17E+14 8.28E+14

3.09E+12 4.66E+11 3.71E+13 1.38E+12 1.37E+13 1.53E+14 1.36E+14 1.61E+12 1.82E+11 1.21E+13 6.70E+11 7.17E+12 8.28E+13 7.44E+13 5.90E+11 7.11E+10 3.65E+12 5.15E+11 2.97E+12 3.41E+13 3.02E+13 2.94E+11 3.60E+10 1.62E+12 1.04E+11 1.46E+12 1.80E+13 1.56E+13

Bias 1.55E+07 1.84E+06 4.15E+07 2.48E+06 -1.38E+07 1.83E+07 2.98E+07 1.69E+07 1.30E+06 3.30E+07 5.18E+05 -1.37E+07 1.83E+07 2.99E+07 1.65E+07 1.85E+06 3.52E+07 1.07E+06 -1.38E+07 1.85E+07 3.01E+07 1.72E+07 2.02E+06 3.41E+07 2.12E+06 -1.37E+07 1.84E+07 2.99E+07

Bias2

MSE

2.43E+14 3.84E+12 1.76E+15 7.50E+12 2.03E+14 4.87E+14 1.03E+15 2.87E+14 1.88E+12 1.10E+15 9.39E+11 1.96E+14 4.19E+14 9.66E+14 2.74E+14 3.50E+12 1.24E+15 1.66E+12 1.95E+14 3.76E+14 9.33E+14 2.96E+14 4.12E+12 1.17E+15 4.60E+12 1.90E+14 3.55E+14 9.09E+14

2.46E+14 4.31E+12 1.79E+15 8.88E+12 2.17E+14 5.00E+14 1.16E+15 2.89E+14 2.07E+12 1.11E+15 1.61E+12 2.03E+14 4.26E+14 1.04E+15 2.75E+14 3.57E+12 1.25E+15 2.18E+12 1.98E+14 3.79E+14 9.64E+14 2.97E+14 4.16E+12 1.17E+15 4.70E+12 1.92E+14 3.56E+14 9.24E+14

TABLE A1 (contd.) Scenario: #2 Period

TML1 Years

1

biweekly

194

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

4.40E+12 1.76E+12 3.90E+13 6.46E+12 1.64E+13 1.62E+14 1.44E+14 2.35E+12 7.88E+11 1.07E+13 1.37E+12 8.55E+12 8.79E+13 7.87E+13 8.81E+11 3.07E+11 4.65E+12 7.13E+11 3.55E+12 3.54E+13 3.14E+13 4.23E+11 1.45E+11 2.24E+12 2.57E+11 1.70E+12 1.87E+13

ts v3

1.63E+13

TML2 Bias2

MSE

Prec.

1.53E+07 4.33E+05 4.33E+07 1.14E+06 -1.46E+07 1.64E+07 2.78E+07 1.43E+07 6.44E+05 2.84E+07 1.27E+06 -1.46E+07 1.65E+07 2.80E+07 1.45E+07 1.26E+06 2.91E+07 -2.91E+05 -1.46E+07 1.65E+07 2.80E+07 1.60E+07 7.61E+05 3.44E+07 1.16E+06 -1.46E+07 1.65E+07

2.38E+14 1.95E+12 1.91E+15 7.77E+12 2.30E+14 4.30E+14 9.17E+14 2.06E+14 1.20E+12 8.18E+14 2.99E+12 2.23E+14 3.62E+14 8.63E+14 2.12E+14 1.89E+12 8.52E+14 7.98E+11 2.18E+14 3.09E+14 8.16E+14 2.55E+14 7.24E+11 1.19E+15 1.60E+12 2.14E+14 2.90E+14

2.42E+14 3.72E+12 1.95E+15 1.42E+13 2.46E+14 5.91E+14 1.06E+15 2.08E+14 1.99E+12 8.29E+14 4.36E+12 2.31E+14 4.49E+14 9.42E+14 2.13E+14 2.20E+12 8.57E+14 1.51E+12 2.22E+14 3.44E+14 8.48E+14 2.56E+14 8.68E+11 1.19E+15 1.86E+12 2.16E+14 3.09E+14

5.09E+12 1.83E+12 4.21E+13 6.88E+12 1.61E+13 1.73E+14 1.53E+14 2.56E+12 8.29E+11 1.20E+13 1.56E+12 8.31E+12 9.39E+13 8.33E+13 9.71E+11 3.29E+11 5.12E+12 7.56E+11 3.44E+12 3.78E+13 3.32E+13 4.94E+11 1.50E+11 2.42E+12 3.09E+11 1.66E+12 1.99E+13

2.79E+07

7.95E+14

8.11E+14

1.72E+13

Bias

Bias2

MSE

1.64E+07 1.54E+06 4.47E+07 2.26E+06 -1.38E+07 1.82E+07 2.97E+07 1.54E+07 1.76E+06 2.95E+07 2.38E+06 -1.38E+07 1.84E+07 2.99E+07 1.57E+07 2.39E+06 3.03E+07 8.03E+05 -1.38E+07 1.84E+07 2.99E+07 1.71E+07 1.90E+06 3.58E+07 2.29E+06 -1.37E+07 1.83E+07

2.75E+14 4.20E+12 2.04E+15 1.20E+13 2.06E+14 5.04E+14 1.04E+15 2.39E+14 3.93E+12 8.85E+14 7.20E+12 1.99E+14 4.31E+14 9.79E+14 2.47E+14 6.03E+12 9.21E+14 1.40E+12 1.95E+14 3.75E+14 9.29E+14 2.94E+14 3.77E+12 1.28E+15 5.57E+12 1.91E+14 3.56E+14

2.80E+14 6.03E+12 2.08E+15 1.89E+13 2.23E+14 5.20E+14 1.19E+15 2.42E+14 4.76E+12 8.97E+14 8.75E+12 2.07E+14 4.39E+14 1.06E+15 2.48E+14 6.36E+12 9.26E+14 2.16E+12 1.98E+14 3.78E+14 9.62E+14 2.95E+14 3.92E+12 1.28E+15 5.88E+12 1.92E+14 3.58E+14

2.99E+07

9.09E+14

9.26E+14

Bias

TABLE A1 (contd.) Scenario: #2 Period

TML1 Years

1

monthly

195

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

2.37E+13 8.35E+12 4.59E+13 1.21E+13 2.36E+13 2.11E+14 1.83E+14 6.12E+12 3.48E+12 2.18E+13 5.92E+12 1.15E+13 1.13E+14 9.90E+13 2.31E+12 1.26E+12 7.82E+12 1.95E+12 4.55E+12 4.41E+13 3.86E+13 1.07E+12 6.46E+11 3.26E+12 7.88E+11 2.24E+12 2.28E+13 1.97E+13

Bias 1.90E+07 3.69E+05 3.46E+07 1.18E+06 -1.50E+07 1.54E+07 2.68E+07 1.69E+07 -9.64E+04 3.73E+07 -8.57E+05 -1.50E+07 1.59E+07 2.73E+07 1.51E+07 3.42E+05 2.71E+07 -1.32E+06 -1.51E+07 1.57E+07 2.71E+07 1.58E+07 2.52E+05 3.09E+07 -1.10E+06 -1.50E+07 1.55E+07 2.69E+07

TML2 Bias2

MSE

Prec.

3.85E+14 8.49E+12 1.24E+15 1.35E+13 2.50E+14 4.49E+14 9.04E+14 2.92E+14 3.49E+12 1.42E+15 6.66E+12 2.36E+14 3.67E+14 8.46E+14 2.31E+14 1.37E+12 7.43E+14 3.70E+12 2.31E+14 2.89E+14 7.73E+14 2.49E+14 7.10E+11 9.59E+14 2.00E+12 2.27E+14 2.64E+14 7.45E+14

4.09E+14 1.68E+13 1.29E+15 2.57E+13 2.73E+14 6.60E+14 1.09E+15 2.98E+14 6.96E+12 1.44E+15 1.26E+13 2.48E+14 4.80E+14 9.45E+14 2.33E+14 2.63E+12 7.51E+14 5.65E+12 2.36E+14 3.33E+14 8.11E+14 2.50E+14 1.36E+12 9.62E+14 2.79E+12 2.29E+14 2.87E+14 7.65E+14

2.60E+13 9.11E+12 5.10E+13 1.33E+13 2.37E+13 2.35E+14 2.02E+14 6.95E+12 3.73E+12 2.49E+13 6.36E+12 1.14E+13 1.26E+14 1.09E+14 2.62E+12 1.34E+12 9.02E+12 2.10E+12 4.46E+12 4.90E+13 4.22E+13 1.18E+12 6.79E+11 3.93E+12 8.74E+11 2.18E+12 2.52E+13 2.14E+13

Bias 2.08E+07 2.02E+06 3.64E+07 2.86E+06 -1.38E+07 1.81E+07 2.97E+07 1.87E+07 1.54E+06 3.93E+07 7.71E+05 -1.38E+07 1.86E+07 3.02E+07 1.68E+07 1.99E+06 2.87E+07 2.68E+05 -1.38E+07 1.84E+07 2.99E+07 1.75E+07 1.88E+06 3.27E+07 5.16E+05 -1.38E+07 1.83E+07 2.98E+07

Bias2

MSE

4.59E+14 1.32E+13 1.38E+15 2.15E+13 2.15E+14 5.64E+14 1.08E+15 3.56E+14 6.10E+12 1.57E+15 6.96E+12 2.01E+14 4.73E+14 1.02E+15 2.86E+14 5.29E+12 8.33E+14 2.17E+12 1.96E+14 3.86E+14 9.38E+14 3.08E+14 4.23E+12 1.07E+15 1.14E+12 1.92E+14 3.59E+14 9.10E+14

4.85E+14 2.23E+13 1.43E+15 3.48E+13 2.38E+14 5.88E+14 1.28E+15 3.63E+14 9.84E+12 1.59E+15 1.33E+13 2.13E+14 4.85E+14 1.13E+15 2.88E+14 6.63E+12 8.42E+14 4.27E+12 2.00E+14 3.91E+14 9.80E+14 3.10E+14 4.91E+12 1.08E+15 2.01E+12 1.94E+14 3.61E+14 9.31E+14

TABLE A1 (contd.) Scenario: #2 Period

TML1 Years

1

semi-annual

196

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

2.43E+14 2.41E+14 4.34E+14 2.77E+14 1.41E+14 1.45E+14 1.39E+14 1.38E+14 1.04E+14 6.40E+13 9.41E+13 7.65E+13 7.32E+13 7.08E+13 3.98E+13 4.11E+13 8.77E+13 5.34E+13 3.06E+13 2.81E+13 2.69E+13 1.43E+15 1.46E+15 1.39E+15 1.45E+15 1.69E+15 1.65E+15 1.64E+15

Bias 7.85E+06 -3.37E+06 4.23E+07 -8.02E+04 -4.46E+07 -4.03E+07 -3.86E+07 1.58E+07 -4.35E+06 1.29E+07 -9.40E+06 -4.51E+07 -4.11E+07 -3.94E+07 8.87E+06 -4.51E+06 4.02E+07 -5.57E+06 -4.52E+07 -4.12E+07 -3.94E+07 -4.37E+08 -4.53E+08 -4.18E+08 -4.52E+08 -4.94E+08 -4.90E+08 -4.88E+08

TML2 Bias2

MSE

Prec.

3.04E+14 2.53E+14 2.22E+15 2.77E+14 2.13E+15 1.77E+15 1.63E+15 3.88E+14 1.23E+14 2.31E+14 1.83E+14 2.11E+15 1.76E+15 1.62E+15 1.18E+14 6.14E+13 1.70E+15 8.44E+13 2.07E+15 1.72E+15 1.58E+15 1.92E+17 2.06E+17 1.76E+17 2.05E+17 2.45E+17 2.41E+17 2.40E+17

5.47E+14 4.94E+14 2.66E+15 5.55E+14 2.27E+15 1.91E+15 1.77E+15 5.26E+14 2.28E+14 2.95E+14 2.77E+14 2.19E+15 1.84E+15 1.69E+15 1.58E+14 1.02E+14 1.79E+15 1.38E+14 2.10E+15 1.75E+15 1.61E+15 1.94E+17 2.08E+17 1.77E+17 2.07E+17 2.47E+17 2.43E+17 2.41E+17

3.10E+14 2.88E+14 5.15E+14 3.34E+14 2.23E+14 1.73E+15 1.45E+15 1.66E+14 1.25E+14 8.02E+13 1.14E+14 1.04E+14 8.26E+14 6.98E+14 5.02E+13 5.28E+13 1.11E+14 7.11E+13 4.08E+13 3.67E+14 3.01E+14 1.40E+15 1.43E+15 1.37E+15 1.42E+15 1.50E+15 1.38E+15 1.32E+15

Bias 1.20E+07 1.45E+06 4.86E+07 5.00E+06 -1.45E+07 1.64E+07 2.78E+07 2.05E+07 6.75E+05 1.62E+07 -4.99E+06 -1.54E+07 1.39E+07 2.57E+07 1.33E+07 -5.86E+04 4.61E+07 -1.17E+06 -1.51E+07 1.49E+07 2.66E+07 -4.32E+08 -4.48E+08 -4.13E+08 -4.47E+08 -4.63E+08 -4.33E+08 -4.22E+08

Bias2

MSE

4.53E+14 2.90E+14 2.88E+15 3.59E+14 4.35E+14 2.00E+15 2.23E+15 5.87E+14 1.26E+14 3.44E+14 1.39E+14 3.43E+14 1.02E+15 1.36E+15 2.27E+14 5.28E+13 2.24E+15 7.25E+13 2.70E+14 5.88E+14 1.01E+15 1.88E+17 2.02E+17 1.72E+17 2.01E+17 2.16E+17 1.89E+17 1.79E+17

7.64E+14 5.79E+14 3.39E+15 6.92E+14 6.57E+14 2.22E+15 3.68E+15 7.53E+14 2.51E+14 4.24E+14 2.52E+14 4.47E+14 1.12E+15 2.06E+15 2.77E+14 1.06E+14 2.35E+15 1.44E+14 3.11E+14 6.29E+14 1.31E+15 1.90E+17 2.04E+17 1.73E+17 2.03E+17 2.18E+17 1.91E+17 1.80E+17

TABLE A1 (contd.) Scenario: #2 Period

TML1 Years

1

season dependent

197

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

1.99E+13 9.34E+12 1.80E+14 2.34E+13 2.30E+13 2.52E+14 2.14E+14 9.69E+12 4.75E+12 2.85E+13 8.42E+12 1.14E+13 1.30E+14 1.12E+14 3.58E+12 1.72E+12 1.45E+13 4.75E+12 4.45E+12 5.22E+13 4.46E+13 1.27E+12 8.53E+11 4.37E+12 1.28E+12 2.45E+12 2.55E+13 2.15E+13

Bias 1.71E+07 1.84E+06 4.69E+07 -2.89E+06 -1.38E+07 1.87E+07 3.01E+07 1.87E+07 1.85E+06 3.38E+07 2.39E+06 -1.38E+07 1.88E+07 3.02E+07 1.73E+07 1.94E+06 3.27E+07 1.85E+06 -1.38E+07 1.89E+07 3.04E+07 1.78E+07 1.63E+06 3.35E+07 2.30E+06 -1.38E+07 1.86E+07 3.00E+07

TML2 Bias2

MSE

Prec.

3.12E+14 1.27E+13 2.38E+15 3.18E+13 2.15E+14 6.02E+14 1.12E+15 3.59E+14 8.17E+12 1.17E+15 1.41E+13 2.01E+14 4.82E+14 1.02E+15 3.04E+14 5.47E+12 1.08E+15 8.17E+12 1.94E+14 4.10E+14 9.67E+14 3.19E+14 3.50E+12 1.13E+15 6.57E+12 1.92E+14 3.70E+14 9.19E+14

3.32E+14 2.21E+13 2.56E+15 5.52E+13 2.38E+14 8.54E+14 1.33E+15 3.69E+14 1.29E+13 1.20E+15 2.25E+13 2.12E+14 6.12E+14 1.14E+15 3.08E+14 7.19E+12 1.10E+15 1.29E+13 1.98E+14 4.62E+14 1.01E+15 3.20E+14 4.36E+12 1.13E+15 7.86E+12 1.94E+14 3.95E+14 9.41E+14

2.78E+13 1.96E+13 9.95E+13 1.88E+13 1.77E+13 4.56E+14 3.78E+14 1.60E+13 9.47E+12 2.26E+13 1.05E+13 8.59E+12 2.30E+14 1.93E+14 5.29E+12 3.81E+12 7.52E+12 9.47E+12 3.50E+12 9.05E+13 7.57E+13 3.10E+12 2.44E+12 5.85E+12 2.60E+12 1.95E+12 5.09E+13 4.21E+13

Bias 2.67E+07 1.20E+07 5.60E+07 8.02E+06 -6.15E+06 3.59E+07 4.80E+07 2.93E+07 1.18E+07 4.38E+07 1.26E+07 -6.12E+06 3.59E+07 4.81E+07 2.71E+07 1.20E+07 4.05E+07 1.18E+07 -6.19E+06 3.59E+07 4.81E+07 2.78E+07 1.25E+07 4.33E+07 1.24E+07 -5.98E+06 3.59E+07 4.81E+07

Bias2

MSE

7.40E+14 1.63E+14 3.23E+15 8.31E+13 5.55E+13 1.74E+15 2.68E+15 8.74E+14 1.48E+14 1.94E+15 1.70E+14 4.61E+13 1.52E+15 2.51E+15 7.39E+14 1.49E+14 1.64E+15 1.48E+14 4.18E+13 1.38E+15 2.39E+15 7.78E+14 1.58E+14 1.88E+15 1.56E+14 3.77E+13 1.34E+15 2.36E+15

7.67E+14 1.83E+14 3.33E+15 1.02E+14 7.32E+13 1.76E+15 3.06E+15 8.90E+14 1.58E+14 1.96E+15 1.80E+14 5.46E+13 1.53E+15 2.70E+15 7.44E+14 1.52E+14 1.65E+15 1.58E+14 4.53E+13 1.38E+15 2.46E+15 7.81E+14 1.61E+14 1.88E+15 1.58E+14 3.96E+13 1.34E+15 2.40E+15

TABLE A1 (contd.) Scenario: #2 Period

TML1 Years

198

parameter dependent

1

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

1.90E+13 1.18E+13 7.76E+13 1.67E+13 2.48E+13 2.48E+14 2.14E+14 1.22E+13 6.27E+12 3.07E+13 1.14E+13 1.29E+13 1.34E+14 1.14E+14 3.38E+12 2.24E+12 1.44E+13 4.19E+12 5.29E+12 5.56E+13 4.70E+13 1.70E+12 1.10E+12 6.99E+12 1.61E+12 2.64E+12 2.94E+13 2.44E+13

Bias 1.69E+07 5.76E+05 3.08E+07 -3.65E+05 -1.49E+07 1.58E+07 2.71E+07 1.79E+07 7.92E+05 2.33E+07 1.84E+05 -1.47E+07 1.67E+07 2.80E+07 1.44E+07 6.32E+05 2.84E+07 4.74E+05 -1.46E+07 1.70E+07 2.84E+07 1.54E+07 9.69E+05 3.31E+07 1.58E+05 -1.44E+07 1.70E+07 2.84E+07

TML2 Bias2

MSE

Prec.

3.03E+14 1.21E+13 1.03E+15 1.68E+13 2.46E+14 4.99E+14 9.49E+14 3.34E+14 6.90E+12 5.73E+14 1.14E+13 2.28E+14 4.12E+14 9.00E+14 2.10E+14 2.64E+12 8.18E+14 4.41E+12 2.18E+14 3.43E+14 8.52E+14 2.38E+14 2.04E+12 1.10E+15 1.64E+12 2.10E+14 3.20E+14 8.33E+14

3.22E+14 2.39E+13 1.11E+15 3.35E+13 2.71E+14 7.47E+14 1.16E+15 3.46E+14 1.32E+13 6.04E+14 2.29E+13 2.41E+14 5.45E+14 1.01E+15 2.13E+14 4.89E+12 8.33E+14 8.60E+12 2.24E+14 3.99E+14 8.99E+14 2.40E+14 3.15E+12 1.11E+15 3.25E+12 2.13E+14 3.49E+14 8.57E+14

3.78E+13 4.79E+13 8.53E+13 3.27E+13 4.27E+13 2.68E+14 2.30E+14 2.61E+13 2.49E+13 3.74E+13 2.62E+13 2.92E+13 1.28E+14 1.10E+14 1.12E+13 1.39E+13 1.36E+13 1.67E+13 1.61E+13 5.76E+13 4.88E+13 5.32E+12 5.69E+12 6.51E+12 6.19E+12 7.49E+12 2.65E+13 2.15E+13

Bias 2.53E+07 1.13E+07 4.49E+07 5.96E+06 -1.06E+07 2.60E+07 3.75E+07 2.45E+07 7.22E+06 3.20E+07 5.27E+06 -1.03E+07 2.72E+07 3.87E+07 1.94E+07 6.05E+06 3.26E+07 5.39E+06 -1.05E+07 2.66E+07 3.81E+07 2.25E+07 8.15E+06 3.92E+07 7.13E+06 -8.97E+06 3.04E+07 4.17E+07

Bias2

MSE

6.76E+14 1.77E+14 2.10E+15 6.83E+13 1.55E+14 9.46E+14 1.63E+15 6.24E+14 7.70E+13 1.06E+15 5.40E+13 1.34E+14 8.68E+14 1.61E+15 3.89E+14 5.05E+13 1.08E+15 4.58E+13 1.27E+14 7.63E+14 1.50E+15 5.11E+14 7.21E+13 1.54E+15 5.71E+13 8.79E+13 9.48E+14 1.76E+15

7.14E+14 2.24E+14 2.18E+15 1.01E+14 1.97E+14 9.89E+14 1.86E+15 6.50E+14 1.02E+14 1.10E+15 8.02E+13 1.64E+14 8.98E+14 1.72E+15 4.01E+14 6.44E+13 1.09E+15 6.25E+13 1.43E+14 7.79E+14 1.55E+15 5.16E+14 7.78E+13 1.55E+15 6.33E+13 9.54E+13 9.56E+14 1.78E+15

TABLE A1 (contd.) Scenario: #3 Period

Mean Concentration Years

1

weekly

199

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.017 0.012 0.052 0.013 0.062 0.165 0.438 0.007 0.005 0.022 0.007 0.030 0.081 0.217 0.003 0.002 0.011 0.003 0.012 0.032 0.086 0.002 0.001 0.006 0.002 0.006 0.017 0.049

Bias 4.251 0.338 8.198 0.250 -2.552 1.186 4.924 3.836 0.336 7.701 0.144 -2.556 1.178 4.910 3.984 0.406 8.037 0.424 -2.554 1.181 4.913 4.023 0.370 8.108 0.528 -2.556 1.182 4.920

Bias2 18.088 0.126 67.253 0.076 6.574 1.573 24.677 14.720 0.118 59.327 0.028 6.563 1.468 24.322 15.877 0.167 64.603 0.183 6.536 1.426 24.225 16.182 0.138 65.723 0.281 6.537 1.414 24.241

Max Concentration MSE 18.105 0.138 67.305 0.089 6.636 1.738 25.114 14.727 0.123 59.349 0.036 6.593 1.550 24.539 15.880 0.169 64.614 0.186 6.548 1.458 24.311 16.183 0.139 65.728 0.282 6.543 1.431 24.290

Prec. 2.416 0.752 15.156 2.474 1.486 3.363 2.758 3.206 0.694 26.658 2.193 1.282 6.133 4.596 2.572 0.759 11.422 2.920 1.570 18.248 13.039 4.363 0.761 14.841 3.439 1.726 23.259 16.616

Bias 5.585 -0.610 20.955 1.475 -6.285 -1.227 3.762 5.975 -0.575 24.139 1.621 -6.922 -1.779 3.193 6.367 -0.315 23.128 2.984 -7.570 -2.092 2.813 8.243 -0.255 25.553 3.891 -8.053 -2.452 2.430

Bias2 33.609 1.124 454.242 4.651 40.986 4.869 16.907 38.902 1.025 609.288 4.821 49.196 9.299 14.789 43.101 0.858 546.294 11.821 58.864 22.623 20.953 72.290 0.826 667.542 18.574 66.548 29.271 22.517

MSE 36.025 1.876 469.398 7.125 42.472 8.232 19.665 42.108 1.719 635.946 7.013 50.479 15.432 19.385 45.673 1.616 557.716 14.741 60.433 40.871 33.992 76.653 1.586 682.383 22.014 68.274 52.530 39.133

TABLE A1 (contd.) Scenario: #3 Period

Mean Concentration Years

1

biweekly

200

2

5

10

1σ & Bias

Prec.

Bias

Bias2

Max Concentration MSE

Prec.

Bias

Bias2

MSE

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

0.059 0.011 0.170 0.017 0.121 0.300 0.674 0.022 0.019 0.050 0.024 0.059 0.146 0.329 0.008 0.007 0.025 0.009 0.023 0.056 0.129 0.004 0.003 0.014 0.004 0.011 0.029

4.783 0.326 4.796 0.155 -2.555 1.181 4.915 3.950 0.462 7.490 0.674 -2.553 1.186 4.923 3.814 0.390 8.127 0.250 -2.554 1.182 4.917 4.069 0.374 7.624 0.477 -2.556 1.182

22.937 0.117 23.171 0.041 6.648 1.694 24.824 15.626 0.233 56.143 0.478 6.576 1.552 24.559 14.552 0.159 66.063 0.071 6.543 1.454 24.299 16.558 0.144 58.126 0.232 6.541 1.426

22.996 0.128 23.341 0.058 6.769 1.994 25.498 15.647 0.251 56.193 0.502 6.634 1.697 24.888 14.561 0.167 66.088 0.080 6.566 1.511 24.428 16.562 0.147 58.140 0.236 6.552 1.455

5.671 0.779 13.335 2.324 1.539 2.584 2.245 2.208 1.033 7.810 2.492 1.349 4.261 3.350 3.033 1.139 10.163 2.981 1.469 8.497 6.297 2.915 1.053 12.996 2.566 1.522 12.448

6.789 -0.609 15.486 1.436 -6.285 -1.252 3.742 4.488 -0.884 16.077 1.287 -6.958 -1.875 3.109 5.151 -0.856 19.849 1.875 -7.672 -2.472 2.489 6.040 -0.898 23.353 2.387 -8.171 -2.871

51.764 1.150 253.140 4.384 41.039 4.152 16.245 22.346 1.815 266.260 4.147 49.755 7.776 13.016 29.565 1.872 404.090 6.496 60.328 14.606 12.490 39.393 1.860 558.255 8.261 68.274 20.688

57.435 1.929 266.476 6.708 42.577 6.736 18.490 24.554 2.848 274.070 6.639 51.104 12.037 16.366 32.598 3.012 414.252 9.476 61.797 23.102 18.788 42.308 2.913 571.251 10.828 69.796 33.136

ts v3

0.070

4.919

24.259

24.329

9.081

2.070

13.366

22.446

TABLE A1 (contd.) Scenario: #3 Period

Mean Concentration Years

1

monthly

201

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.176 0.154 0.285 0.189 0.299 0.698 1.380 0.090 0.079 0.161 0.088 0.146 0.341 0.675 0.031 0.030 0.064 0.034 0.056 0.130 0.258 0.017 0.015 0.029 0.016 0.028 0.066 0.135

Bias 4.479 0.351 5.936 0.749 -2.550 1.188 4.924 4.123 0.543 8.611 0.770 -2.551 1.183 4.915 3.715 0.352 7.616 0.499 -2.558 1.175 4.906 4.018 0.354 7.444 0.363 -2.557 1.180 4.915

Bias2 20.233 0.277 35.514 0.750 6.803 2.108 25.624 17.090 0.373 74.299 0.682 6.651 1.741 24.828 13.833 0.154 58.054 0.283 6.597 1.510 24.321 16.156 0.140 55.442 0.148 6.564 1.459 24.295

Max Concentration MSE 20.409 0.430 35.799 0.939 7.103 2.806 27.004 17.180 0.452 74.460 0.770 6.796 2.082 25.503 13.864 0.184 58.117 0.317 6.652 1.640 24.579 16.174 0.155 55.471 0.164 6.592 1.525 24.430

Prec. 3.452 1.538 10.915 3.593 1.531 2.044 1.895 4.287 1.655 18.463 3.178 1.328 2.609 2.203 3.185 1.463 9.718 3.118 1.439 4.913 3.830 4.763 1.442 10.553 4.151 1.432 6.658 5.055

Bias 3.536 -1.923 13.242 0.137 -6.292 -1.277 3.719 4.403 -1.359 17.712 0.408 -6.949 -1.911 3.082 4.110 -1.606 16.977 0.814 -7.713 -2.615 2.366 5.810 -1.569 19.265 1.400 -8.233 -3.093 1.880

Bias2 15.953 5.236 186.244 3.611 41.116 3.675 15.726 23.668 3.503 332.162 3.344 49.614 6.262 11.700 20.073 4.043 297.916 3.781 60.922 11.750 9.428 38.514 3.903 381.668 6.110 69.215 16.223 8.588

MSE 19.405 6.774 197.159 7.204 42.647 5.719 17.621 27.955 5.158 350.625 6.521 50.942 8.871 13.903 23.258 5.506 307.634 6.899 62.360 16.664 13.258 43.277 5.344 392.221 10.261 70.647 22.880 13.643

TABLE A1 (contd.) Scenario: #3 Period

Mean Concentration Years

1

semi-annual

202

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 3.112 2.749 7.961 1.843 3.100 6.844 12.175 1.092 1.250 2.449 1.529 1.535 3.425 6.094 0.495 0.498 0.684 0.551 0.585 1.323 2.347 0.257 0.241 0.334 0.257 0.284 0.642 1.153

Bias 4.809 0.595 12.288 -1.298 -2.522 1.229 4.980 3.176 0.354 11.601 1.467 -2.549 1.190 4.927 3.945 0.451 8.252 0.701 -2.554 1.185 4.918 4.305 0.486 8.888 0.704 -2.550 1.188 4.925

Bias2 26.233 3.103 158.934 3.527 9.462 8.354 36.970 11.178 1.375 137.014 3.681 8.030 4.841 30.368 16.054 0.701 68.780 1.042 7.105 2.727 26.531 18.786 0.477 79.327 0.754 6.784 2.054 25.405

Max Concentration MSE 29.346 5.851 166.895 5.369 12.562 15.198 49.145 12.270 2.625 139.463 5.210 9.565 8.267 36.462 16.549 1.199 69.463 1.593 7.689 4.050 28.878 19.044 0.718 79.661 1.011 7.068 2.696 26.558

Prec. 5.808 4.749 32.201 3.818 5.557 10.943 18.578 4.432 3.358 7.803 5.285 2.654 4.575 6.970 5.830 2.941 19.626 6.213 1.469 3.036 2.602 4.215 2.634 11.146 4.215 1.338 2.245 1.964

Bias 0.118 -4.299 15.015 -6.377 -7.478 -3.065 1.348 -0.724 -4.144 8.913 -2.294 -7.276 -2.423 2.424 1.526 -3.487 11.794 -1.147 -7.744 -2.703 2.286 1.916 -3.206 11.917 -1.684 -8.280 -3.254 1.741

Bias2 5.822 23.225 257.632 44.484 61.466 20.334 20.395 4.956 20.527 87.238 10.546 55.588 10.447 12.843 8.160 15.098 158.722 7.527 61.430 10.344 7.828 7.887 12.909 153.142 7.051 69.884 12.834 4.996

MSE 11.629 27.974 289.833 48.303 67.022 31.277 38.973 9.389 23.886 95.041 15.831 58.242 15.022 19.813 13.989 18.039 178.348 13.740 62.899 13.381 10.431 12.103 15.543 164.288 11.266 71.222 15.079 6.960

TABLE A1 (contd.) Scenario: #3 Period

Mean Concentration Years

1

season dependent

203

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.189 0.184 0.286 0.184 0.220 0.518 1.060 0.100 0.091 0.109 0.095 0.105 0.253 0.523 0.039 0.034 0.049 0.036 0.042 0.098 0.203 0.021 0.018 0.024 0.018 0.023 0.055 0.114

Bias 3.978 0.383 8.872 0.310 -2.568 1.161 4.888 4.182 0.357 7.651 0.454 -2.563 1.170 4.901 4.248 0.340 8.061 0.251 -2.565 1.164 4.890 4.036 0.344 7.369 0.528 -2.566 1.168 4.901

Bias2 16.009 0.331 78.985 0.280 6.815 1.866 24.953 17.585 0.218 58.637 0.301 6.676 1.621 24.543 18.086 0.150 65.020 0.099 6.618 1.452 24.117 16.304 0.136 54.304 0.296 6.605 1.418 24.123

Max Concentration MSE 16.199 0.515 79.271 0.463 7.035 2.384 26.013 17.685 0.309 58.746 0.397 6.781 1.874 25.066 18.125 0.185 65.068 0.135 6.660 1.551 24.321 16.325 0.154 54.328 0.314 6.628 1.473 24.237

Prec. 4.861 1.393 12.227 3.458 1.553 2.900 2.460 3.833 1.163 9.517 4.650 1.292 3.383 2.716 2.990 1.203 10.259 3.002 1.395 8.068 5.950 2.947 1.199 18.276 2.944 1.312 10.031 7.297

Bias 5.428 -0.891 19.794 1.286 -6.298 -1.255 3.737 6.148 -0.988 17.948 2.379 -6.973 -1.910 3.078 6.098 -0.821 19.660 2.205 -7.696 -2.497 2.464 6.287 -0.808 24.244 3.142 -8.192 -2.940 2.012

Bias2 34.316 2.187 404.001 5.112 41.212 4.474 16.425 41.624 2.140 331.618 10.311 49.909 7.031 12.189 40.176 1.877 396.741 7.864 60.610 14.305 12.018 42.457 1.852 605.855 12.810 68.396 18.672 11.342

MSE 39.177 3.581 416.228 8.570 42.765 7.374 18.886 45.457 3.302 341.135 14.960 51.200 10.414 14.905 43.166 3.080 406.999 10.866 62.004 22.373 17.968 45.405 3.051 624.131 15.754 69.708 28.703 18.639

TABLE A1 (contd.) Scenario: #3 Period

Mean Concentration Years

204

parameter dependent

1

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.150 0.161 0.453 0.169 0.265 0.522 0.979 0.083 0.072 0.309 0.075 0.130 0.267 0.508 0.035 0.032 0.064 0.035 0.050 0.108 0.211 0.018 0.016 0.028 0.018 0.025 0.056 0.113

Bias 3.721 0.654 7.828 0.945 -2.336 1.509 5.353 4.205 0.440 8.649 -0.339 -2.398 1.414 5.224 3.881 0.576 7.329 0.280 -2.428 1.371 5.166 3.936 0.529 7.496 0.740 -2.390 1.429 5.247

Bias2 13.991 0.590 61.731 1.061 5.722 2.799 29.629 17.761 0.265 75.107 0.190 5.879 2.267 27.798 15.092 0.363 53.766 0.114 5.944 1.986 26.900 15.508 0.296 56.214 0.566 5.739 2.098 27.638

Max Concentration MSE 14.141 0.751 62.184 1.230 5.987 3.322 30.608 17.844 0.337 75.416 0.266 6.009 2.534 28.305 15.128 0.395 53.830 0.149 5.993 2.094 27.112 15.526 0.312 56.242 0.584 5.764 2.154 27.751

Prec. 3.026 1.628 14.670 3.441 1.538 2.265 2.030 7.260 1.475 11.490 4.277 1.329 3.433 2.772 3.004 1.391 9.320 3.004 1.516 7.349 5.537 3.263 1.436 14.644 4.543 1.515 10.052 7.433

Bias 3.102 -1.408 16.413 0.098 -6.293 -1.270 3.726 5.027 -1.574 15.799 -0.127 -6.955 -1.895 3.094 4.376 -1.153 16.223 0.776 -7.692 -2.529 2.439 5.035 -1.111 19.607 2.678 -8.197 -2.967 1.988

Bias2 12.646 3.609 284.032 3.450 41.130 3.876 15.915 32.524 3.951 261.067 4.293 49.701 7.022 12.342 22.153 2.721 272.495 3.606 60.674 13.743 11.486 28.608 2.670 399.047 11.712 68.706 18.854 11.383

MSE 15.671 5.236 298.702 6.891 42.669 6.141 17.944 39.783 5.426 272.557 8.570 51.030 10.455 15.115 25.157 4.112 281.814 6.611 62.190 21.093 17.022 31.871 4.106 413.691 16.255 70.221 28.906 18.816

TABLE A1 (contd.) TML1

Scenario: #3 Period

Years

1

weekly

205

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

7.91E+11 2.09E+11 2.74E+12 2.51E+11 1.23E+12 2.88E+12 8.24E+12 3.24E+11 9.61E+10 1.26E+12 1.35E+11 6.25E+11 1.40E+12 4.04E+12 1.42E+11 3.81E+10 5.68E+11 5.63E+10 2.49E+11 5.55E+11 1.63E+12 6.86E+10 1.93E+10 2.59E+11 3.12E+10 1.15E+11 2.91E+11 9.00E+11

Bias 1.75E+07 1.40E+06 3.38E+07 1.03E+06 -1.05E+07 4.89E+06 2.03E+07 1.58E+07 1.38E+06 3.17E+07 5.94E+05 -1.05E+07 4.86E+06 2.02E+07 1.64E+07 1.67E+06 3.31E+07 1.75E+06 -1.05E+07 4.87E+06 2.03E+07 1.66E+07 1.53E+06 3.34E+07 2.18E+06 -1.05E+07 4.87E+06 2.03E+07

TML2 Bias2

MSE

Prec.

3.08E+14 2.16E+12 1.14E+15 1.31E+12 1.12E+14 2.68E+13 4.19E+14 2.50E+14 2.01E+12 1.01E+15 4.87E+11 1.12E+14 2.50E+13 4.14E+14 2.70E+14 2.84E+12 1.10E+15 3.12E+12 1.11E+14 2.43E+13 4.12E+14 2.75E+14 2.35E+12 1.12E+15 4.80E+12 1.11E+14 2.40E+13 4.12E+14

3.08E+14 2.36E+12 1.15E+15 1.57E+12 1.13E+14 2.97E+13 4.28E+14 2.51E+14 2.11E+12 1.01E+15 6.22E+11 1.12E+14 2.64E+13 4.18E+14 2.70E+14 2.88E+12 1.10E+15 3.18E+12 1.11E+14 2.48E+13 4.14E+14 2.75E+14 2.37E+12 1.12E+15 4.83E+12 1.11E+14 2.43E+13 4.13E+14

8.00E+11 2.06E+11 2.86E+12 2.52E+11 1.24E+12 2.93E+12 8.35E+12 3.30E+11 9.39E+10 1.30E+12 1.38E+11 6.29E+11 1.43E+12 4.10E+12 1.45E+11 3.73E+10 5.94E+11 5.86E+10 2.51E+11 5.65E+11 1.66E+12 7.22E+10 1.90E+10 2.66E+11 3.24E+10 1.15E+11 3.01E+11 9.26E+11

Bias 1.77E+07 1.51E+06 3.40E+07 1.14E+06 -1.04E+07 5.01E+06 2.04E+07 1.60E+07 1.50E+06 3.19E+07 7.08E+05 -1.05E+07 4.98E+06 2.04E+07 1.66E+07 1.79E+06 3.33E+07 1.87E+06 -1.05E+07 5.00E+06 2.04E+07 1.67E+07 1.64E+06 3.36E+07 2.31E+06 -1.04E+07 5.00E+06 2.04E+07

Bias2

MSE

3.13E+14 2.49E+12 1.16E+15 1.56E+12 1.10E+14 2.81E+13 4.26E+14 2.55E+14 2.35E+12 1.02E+15 6.39E+11 1.10E+14 2.62E+13 4.21E+14 2.75E+14 3.25E+12 1.11E+15 3.55E+12 1.09E+14 2.55E+13 4.19E+14 2.80E+14 2.72E+12 1.13E+15 5.35E+12 1.09E+14 2.53E+13 4.19E+14

3.14E+14 2.70E+12 1.16E+15 1.81E+12 1.11E+14 2.93E+13 4.35E+14 2.56E+14 2.44E+12 1.02E+15 7.77E+11 1.11E+14 2.69E+13 4.25E+14 2.76E+14 3.28E+12 1.11E+15 3.61E+12 1.10E+14 2.58E+13 4.21E+14 2.80E+14 2.74E+12 1.13E+15 5.38E+12 1.09E+14 2.54E+13 4.20E+14

TABLE A1 (contd.) Scenario: #3 Period

TML1 Years

1

biweekly

206

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

1.65E+12 1.96E+11 3.88E+12 3.12E+11 2.28E+12 5.19E+12 1.22E+13 6.01E+11 3.32E+11 1.69E+12 4.32E+11 1.11E+12 2.51E+12 5.98E+12 2.24E+11 1.27E+11 8.29E+11 1.55E+11 4.36E+11 9.68E+11 2.36E+12 1.11E+11 5.88E+10 4.37E+11 7.56E+10 2.06E+11 4.95E+11

ts v3

1.26E+12

TML2 Bias2

MSE

Prec.

1.97E+07 1.35E+06 1.98E+07 6.41E+05 -1.05E+07 4.87E+06 2.03E+07 1.63E+07 1.91E+06 3.09E+07 2.78E+06 -1.05E+07 4.89E+06 2.03E+07 1.57E+07 1.61E+06 3.35E+07 1.03E+06 -1.05E+07 4.88E+06 2.03E+07 1.68E+07 1.54E+06 3.14E+07 1.95E+06 -1.05E+07 4.87E+06

3.90E+14 2.00E+12 3.94E+14 7.22E+11 1.13E+14 2.89E+13 4.23E+14 2.66E+14 3.97E+12 9.55E+14 8.14E+12 1.12E+14 2.64E+13 4.18E+14 2.48E+14 2.72E+12 1.12E+15 1.22E+12 1.11E+14 2.48E+13 4.14E+14 2.81E+14 2.44E+12 9.85E+14 3.90E+12 1.11E+14 2.42E+13

3.92E+14 2.20E+12 3.98E+14 1.03E+12 1.15E+14 3.41E+13 4.35E+14 2.67E+14 4.31E+12 9.57E+14 8.57E+12 1.13E+14 2.89E+13 4.24E+14 2.48E+14 2.84E+12 1.12E+15 1.38E+12 1.12E+14 2.57E+13 4.16E+14 2.81E+14 2.49E+12 9.86E+14 3.97E+12 1.11E+14 2.47E+13

1.70E+12 1.91E+11 4.22E+12 3.23E+11 2.30E+12 5.29E+12 1.25E+13 6.22E+11 3.35E+11 1.81E+12 4.49E+11 1.12E+12 2.55E+12 6.08E+12 2.31E+11 1.28E+11 8.83E+11 1.61E+11 4.39E+11 9.87E+11 2.40E+12 1.17E+11 5.99E+10 4.29E+11 7.86E+10 2.08E+11 5.10E+11

2.03E+07

4.12E+14

4.13E+14

1.29E+12

Bias

Bias2

MSE

1.99E+07 1.46E+06 1.99E+07 7.52E+05 -1.05E+07 4.99E+06 2.04E+07 1.64E+07 2.03E+06 3.11E+07 2.90E+06 -1.04E+07 5.01E+06 2.05E+07 1.59E+07 1.72E+06 3.37E+07 1.15E+06 -1.04E+07 5.00E+06 2.04E+07 1.69E+07 1.66E+06 3.16E+07 2.07E+06 -1.04E+07 5.00E+06

3.97E+14 2.33E+12 4.01E+14 8.89E+11 1.12E+14 3.02E+13 4.30E+14 2.71E+14 4.44E+12 9.68E+14 8.83E+12 1.10E+14 2.77E+13 4.25E+14 2.53E+14 3.10E+12 1.14E+15 1.48E+12 1.10E+14 2.60E+13 4.20E+14 2.86E+14 2.82E+12 9.98E+14 4.38E+12 1.09E+14 2.55E+13

3.99E+14 2.52E+12 4.06E+14 1.21E+12 1.14E+14 3.25E+13 4.42E+14 2.72E+14 4.77E+12 9.69E+14 9.28E+12 1.11E+14 2.88E+13 4.31E+14 2.53E+14 3.23E+12 1.14E+15 1.64E+12 1.10E+14 2.64E+13 4.23E+14 2.87E+14 2.88E+12 9.98E+14 4.45E+12 1.10E+14 2.57E+13

2.04E+07

4.19E+14

4.20E+14

Bias

TABLE A1 (contd.) Scenario: #3 Period

TML1 Years

1

monthly

207

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

3.56E+12 2.67E+12 6.43E+12 3.33E+12 5.30E+12 1.21E+13 2.45E+13 1.84E+12 1.38E+12 4.11E+12 1.56E+12 2.59E+12 5.88E+12 1.20E+13 6.30E+11 5.28E+11 1.53E+12 6.06E+11 1.02E+12 2.24E+12 4.54E+12 3.34E+11 2.65E+11 7.24E+11 2.98E+11 5.07E+11 1.16E+12 2.42E+12

Bias 1.85E+07 1.45E+06 2.45E+07 3.09E+06 -1.05E+07 4.91E+06 2.03E+07 1.70E+07 2.24E+06 3.55E+07 3.18E+06 -1.05E+07 4.89E+06 2.03E+07 1.53E+07 1.45E+06 3.14E+07 2.06E+06 -1.05E+07 4.85E+06 2.02E+07 1.66E+07 1.47E+06 3.07E+07 1.51E+06 -1.05E+07 4.87E+06 2.03E+07

TML2 Bias2

MSE

Prec.

3.44E+14 4.78E+12 6.04E+14 1.29E+13 1.16E+14 3.61E+13 4.37E+14 2.91E+14 6.40E+12 1.26E+15 1.17E+13 1.13E+14 2.98E+13 4.23E+14 2.35E+14 2.64E+12 9.88E+14 4.85E+12 1.12E+14 2.58E+13 4.14E+14 2.74E+14 2.41E+12 9.43E+14 2.56E+12 1.11E+14 2.48E+13 4.13E+14

3.48E+14 7.45E+12 6.11E+14 1.62E+13 1.21E+14 4.82E+13 4.61E+14 2.93E+14 7.78E+12 1.27E+15 1.32E+13 1.16E+14 3.57E+13 4.35E+14 2.36E+14 3.17E+12 9.90E+14 5.45E+12 1.13E+14 2.80E+13 4.19E+14 2.75E+14 2.68E+12 9.44E+14 2.86E+12 1.12E+14 2.60E+13 4.15E+14

3.61E+12 2.69E+12 6.99E+12 3.39E+12 5.36E+12 1.22E+13 2.48E+13 1.87E+12 1.38E+12 4.35E+12 1.57E+12 2.61E+12 5.96E+12 1.21E+13 6.38E+11 5.26E+11 1.60E+12 6.14E+11 1.02E+12 2.26E+12 4.58E+12 3.35E+11 2.64E+11 7.67E+11 2.98E+11 5.04E+11 1.16E+12 2.44E+12

Bias 1.86E+07 1.57E+06 2.46E+07 3.20E+06 -1.04E+07 5.03E+06 2.05E+07 1.72E+07 2.36E+06 3.57E+07 3.30E+06 -1.04E+07 5.01E+06 2.04E+07 1.55E+07 1.57E+06 3.16E+07 2.17E+06 -1.05E+07 4.97E+06 2.04E+07 1.67E+07 1.58E+06 3.09E+07 1.63E+06 -1.05E+07 4.99E+06 2.04E+07

Bias2

MSE

3.50E+14 5.14E+12 6.13E+14 1.37E+13 1.14E+14 3.75E+13 4.44E+14 2.96E+14 6.95E+12 1.28E+15 1.25E+13 1.11E+14 3.10E+13 4.30E+14 2.40E+14 2.98E+12 1.00E+15 5.34E+12 1.11E+14 2.70E+13 4.21E+14 2.79E+14 2.75E+12 9.55E+14 2.94E+12 1.10E+14 2.60E+13 4.19E+14

3.54E+14 7.84E+12 6.20E+14 1.71E+13 1.19E+14 4.29E+13 4.68E+14 2.98E+14 8.33E+12 1.28E+15 1.40E+13 1.14E+14 3.36E+13 4.42E+14 2.41E+14 3.51E+12 1.00E+15 5.96E+12 1.12E+14 2.80E+13 4.25E+14 2.80E+14 3.02E+12 9.56E+14 3.24E+12 1.10E+14 2.65E+13 4.22E+14

TABLE A1 (contd.) Scenario: #3 Period

TML1 Years

1

semi-annual

208

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

5.36E+13 4.68E+13 1.41E+14 3.15E+13 8.69E+12 9.54E+12 1.11E+13 1.88E+13 2.13E+13 4.36E+13 2.62E+13 4.41E+12 4.81E+12 5.58E+12 8.53E+12 8.51E+12 1.22E+13 9.43E+12 1.80E+12 1.93E+12 2.20E+12 8.97E+13 9.16E+13 8.81E+13 9.14E+13 1.05E+14 1.04E+14 1.03E+14

Bias 1.98E+07 2.46E+06 5.06E+07 -5.34E+06 -3.65E+07 -3.41E+07 -3.17E+07 1.31E+07 1.47E+06 4.78E+07 6.05E+06 -3.65E+07 -3.41E+07 -3.18E+07 1.63E+07 1.87E+06 3.40E+07 2.89E+06 -3.65E+07 -3.42E+07 -3.18E+07 -3.53E+08 -3.69E+08 -3.34E+08 -3.68E+08 -4.08E+08 -4.05E+08 -4.03E+08

TML2 Bias2

MSE

Prec.

4.46E+14 5.29E+13 2.70E+15 6.00E+13 1.34E+15 1.17E+15 1.02E+15 1.90E+14 2.35E+13 2.33E+15 6.28E+13 1.34E+15 1.17E+15 1.02E+15 2.73E+14 1.20E+13 1.17E+15 1.78E+13 1.34E+15 1.17E+15 1.01E+15 1.25E+17 1.36E+17 1.12E+17 1.36E+17 1.66E+17 1.64E+17 1.62E+17

5.00E+14 9.97E+13 2.84E+15 9.15E+13 1.35E+15 1.18E+15 1.03E+15 2.09E+14 4.48E+13 2.38E+15 8.89E+13 1.34E+15 1.18E+15 1.02E+15 2.82E+14 2.05E+13 1.18E+15 2.72E+13 1.34E+15 1.17E+15 1.02E+15 1.25E+17 1.36E+17 1.12E+17 1.36E+17 1.66E+17 1.64E+17 1.62E+17

5.42E+13 4.73E+13 1.54E+14 3.19E+13 5.36E+13 1.18E+14 2.11E+14 1.92E+13 2.15E+13 4.40E+13 2.64E+13 2.65E+13 5.93E+13 1.06E+14 8.65E+12 8.60E+12 1.28E+13 9.70E+12 1.02E+13 2.30E+13 4.09E+13 8.98E+13 9.17E+13 8.83E+13 9.15E+13 9.47E+13 9.25E+13 9.31E+13

Bias 2.00E+07 2.57E+06 5.08E+07 -5.25E+06 -1.03E+07 5.20E+06 2.07E+07 1.32E+07 1.58E+06 4.81E+07 6.18E+06 -1.04E+07 5.03E+06 2.05E+07 1.64E+07 1.98E+06 3.42E+07 3.02E+06 -1.05E+07 5.01E+06 2.04E+07 -3.53E+08 -3.69E+08 -3.34E+08 -3.68E+08 -3.81E+08 -3.66E+08 -3.51E+08

Bias2

MSE

4.53E+14 5.39E+13 2.74E+15 5.95E+13 1.60E+14 1.45E+14 6.40E+14 1.94E+14 2.40E+13 2.35E+15 6.46E+13 1.35E+14 8.46E+13 5.25E+14 2.78E+14 1.25E+13 1.19E+15 1.88E+13 1.19E+14 4.81E+13 4.59E+14 1.25E+17 1.36E+17 1.12E+17 1.35E+17 1.46E+17 1.34E+17 1.23E+17

5.07E+14 1.01E+14 2.89E+15 9.14E+13 2.13E+14 1.99E+14 8.51E+14 2.14E+14 4.55E+13 2.40E+15 9.10E+13 1.62E+14 1.11E+14 6.31E+14 2.87E+14 2.11E+13 1.20E+15 2.85E+13 1.30E+14 5.83E+13 5.00E+14 1.25E+17 1.36E+17 1.12E+17 1.36E+17 1.46E+17 1.34E+17 1.23E+17

TABLE A1 (contd.) Scenario: #3 Period

TML1 Years

1

season dependent

209

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

6.07E+12 2.96E+12 5.11E+13 6.06E+12 5.23E+12 1.21E+13 2.44E+13 5.18E+12 1.60E+12 2.25E+13 2.44E+12 2.59E+12 5.88E+12 1.19E+13 1.36E+12 5.28E+11 3.96E+12 1.27E+12 1.02E+12 2.40E+12 4.91E+12 3.67E+11 2.51E+11 7.57E+11 3.50E+11 4.73E+11 1.19E+12 2.55E+12

Bias 1.58E+07 1.68E+06 3.47E+07 1.47E+06 -1.05E+07 4.87E+06 2.03E+07 1.67E+07 1.63E+06 3.06E+07 1.96E+06 -1.05E+07 4.86E+06 2.03E+07 1.77E+07 1.54E+06 3.28E+07 8.84E+05 -1.05E+07 4.86E+06 2.02E+07 1.63E+07 1.52E+06 3.05E+07 3.71E+06 -1.05E+07 4.93E+06 2.04E+07

TML2 Bias2

MSE

Prec.

2.56E+14 5.80E+12 1.26E+15 8.22E+12 1.16E+14 3.58E+13 4.35E+14 2.85E+14 4.27E+12 9.62E+14 6.27E+12 1.14E+14 2.95E+13 4.22E+14 3.15E+14 2.90E+12 1.08E+15 2.05E+12 1.12E+14 2.60E+13 4.15E+14 2.66E+14 2.57E+12 9.29E+14 1.41E+13 1.11E+14 2.55E+13 4.17E+14

2.62E+14 8.76E+12 1.31E+15 1.43E+13 1.21E+14 4.79E+13 4.59E+14 2.90E+14 5.87E+12 9.84E+14 8.71E+12 1.16E+14 3.54E+13 4.34E+14 3.16E+14 3.43E+12 1.08E+15 3.31E+12 1.13E+14 2.84E+13 4.20E+14 2.66E+14 2.82E+12 9.29E+14 1.44E+13 1.11E+14 2.67E+13 4.19E+14

3.71E+12 3.17E+12 7.49E+12 3.20E+12 4.02E+12 9.01E+12 1.90E+13 1.96E+12 1.57E+12 2.86E+12 1.67E+12 1.97E+12 4.40E+12 9.28E+12 7.43E+11 5.98E+11 1.15E+12 6.33E+11 7.94E+11 1.71E+12 3.59E+12 3.67E+11 3.07E+11 5.56E+11 3.10E+11 4.48E+11 9.45E+11 1.98E+12

Bias 1.66E+07 1.70E+06 3.68E+07 1.39E+06 -1.05E+07 4.92E+06 2.03E+07 1.74E+07 1.59E+06 3.17E+07 1.99E+06 -1.05E+07 4.95E+06 2.04E+07 1.77E+07 1.52E+06 3.35E+07 1.15E+06 -1.05E+07 4.93E+06 2.04E+07 1.68E+07 1.54E+06 3.06E+07 2.30E+06 -1.05E+07 4.94E+06 2.04E+07

Bias2

MSE

2.78E+14 6.05E+12 1.36E+15 5.15E+12 1.14E+14 3.32E+13 4.32E+14 3.05E+14 4.11E+12 1.01E+15 5.63E+12 1.12E+14 2.89E+13 4.25E+14 3.13E+14 2.91E+12 1.12E+15 1.95E+12 1.11E+14 2.60E+13 4.18E+14 2.82E+14 2.67E+12 9.35E+14 5.58E+12 1.10E+14 2.54E+13 4.17E+14

2.82E+14 9.22E+12 1.37E+15 8.35E+12 1.18E+14 3.72E+13 4.51E+14 3.07E+14 5.68E+12 1.01E+15 7.30E+12 1.14E+14 3.09E+13 4.34E+14 3.14E+14 3.51E+12 1.12E+15 2.58E+12 1.12E+14 2.68E+13 4.21E+14 2.83E+14 2.98E+12 9.36E+14 5.89E+12 1.11E+14 2.58E+13 4.19E+14

TABLE A1 (contd.) Scenario: #3 Period

TML1 Years

210

parameter dependent

1

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

3.63E+12 2.88E+12 2.29E+13 3.20E+12 5.42E+12 1.22E+13 2.47E+13 1.85E+12 1.31E+12 8.48E+12 1.56E+12 2.67E+12 6.00E+12 1.22E+13 7.26E+11 5.45E+11 2.09E+12 6.58E+11 1.01E+12 2.29E+12 4.72E+12 3.47E+11 2.62E+11 8.90E+11 3.67E+11 5.06E+11 1.22E+12 2.59E+12

Bias 1.43E+07 1.62E+06 3.13E+07 2.64E+06 -1.09E+07 4.34E+06 1.96E+07 1.63E+07 8.28E+05 3.48E+07 -2.30E+06 -1.08E+07 4.45E+06 1.97E+07 1.51E+07 1.46E+06 2.93E+07 2.38E+05 -1.07E+07 4.60E+06 1.99E+07 1.53E+07 1.30E+06 3.00E+07 2.15E+06 -1.07E+07 4.65E+06 2.00E+07

TML2 Bias2

MSE

Prec.

2.07E+14 5.51E+12 1.00E+15 1.02E+13 1.24E+14 3.11E+13 4.07E+14 2.68E+14 1.99E+12 1.22E+15 6.86E+12 1.20E+14 2.58E+13 4.00E+14 2.28E+14 2.69E+12 8.61E+14 7.14E+11 1.16E+14 2.35E+13 4.01E+14 2.36E+14 1.94E+12 8.99E+14 5.01E+12 1.14E+14 2.28E+13 4.01E+14

2.11E+14 8.39E+12 1.03E+15 1.34E+13 1.29E+14 4.33E+13 4.32E+14 2.70E+14 3.29E+12 1.23E+15 8.42E+12 1.22E+14 3.18E+13 4.12E+14 2.29E+14 3.23E+12 8.63E+14 1.37E+12 1.17E+14 2.57E+13 4.06E+14 2.36E+14 2.20E+12 9.00E+14 5.37E+12 1.15E+14 2.40E+13 4.04E+14

3.30E+12 2.96E+12 2.10E+13 3.53E+12 4.76E+12 9.18E+12 1.78E+13 2.48E+12 2.00E+12 4.02E+12 2.04E+12 2.75E+12 4.62E+12 8.33E+12 1.08E+12 1.02E+12 1.61E+12 1.02E+12 1.20E+12 1.96E+12 3.46E+12 6.45E+11 5.92E+11 9.01E+11 6.28E+11 6.98E+11 1.10E+12 1.89E+12

Bias 1.43E+07 2.37E+06 3.73E+07 4.51E+06 -9.57E+06 6.30E+06 2.22E+07 1.82E+07 2.31E+06 3.30E+07 -6.37E+05 -9.80E+06 5.96E+06 2.17E+07 1.63E+07 2.61E+06 3.06E+07 1.37E+06 -9.93E+06 5.78E+06 2.15E+07 1.66E+07 2.28E+06 3.19E+07 2.85E+06 -9.78E+06 5.99E+06 2.18E+07

Bias2

MSE

2.09E+14 8.55E+12 1.41E+15 2.39E+13 9.63E+13 4.89E+13 5.09E+14 3.33E+14 7.35E+12 1.09E+15 2.45E+12 9.87E+13 4.02E+13 4.80E+14 2.65E+14 7.84E+12 9.39E+14 2.89E+12 9.97E+13 3.54E+13 4.65E+14 2.75E+14 5.78E+12 1.02E+15 8.74E+12 9.64E+13 3.69E+13 4.75E+14

2.12E+14 1.15E+13 1.43E+15 2.74E+13 1.01E+14 5.37E+13 5.27E+14 3.35E+14 9.34E+12 1.10E+15 4.49E+12 1.01E+14 4.29E+13 4.88E+14 2.66E+14 8.86E+12 9.40E+14 3.91E+12 1.01E+14 3.66E+13 4.68E+14 2.76E+14 6.37E+12 1.02E+15 9.37E+12 9.71E+13 3.76E+13 4.77E+14

TABLE A1 (contd.) Scenario: #4 Period

Mean Concentration Years

1

weekly

211

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.052 0.014 0.204 0.044 0.208 0.594 0.681 0.013 0.006 0.047 0.012 0.108 0.285 0.334 0.004 0.003 0.019 0.005 0.042 0.117 0.139 0.002 0.001 0.010 0.002 0.023 0.056 0.073

Bias 4.362 0.348 8.530 0.720 -2.928 1.008 4.502 3.952 0.334 8.311 0.543 -2.936 1.002 4.493 3.910 0.430 8.055 0.428 -2.938 1.007 4.496 3.991 0.380 8.316 0.273 -2.948 0.988 4.482

Bias2 19.081 0.135 72.963 0.562 8.781 1.611 20.943 15.626 0.118 69.118 0.306 8.728 1.289 20.518 15.294 0.188 64.898 0.189 8.670 1.130 20.348 15.921 0.145 69.144 0.077 8.713 1.032 20.151

Max Concentration MSE 19.133 0.150 73.166 0.606 8.989 2.205 21.624 15.638 0.124 69.165 0.318 8.836 1.574 20.852 15.298 0.191 64.917 0.194 8.712 1.246 20.486 15.923 0.147 69.154 0.079 8.736 1.087 20.224

Prec. 6.181 1.117 28.817 7.190 10.802 214.426 151.917 13.487 0.998 33.268 5.074 13.915 254.983 181.969 6.757 1.127 21.554 5.418 8.388 119.969 86.386 7.136 1.297 22.619 6.809 3.430 18.822 14.197

Bias 6.547 -0.600 17.312 3.487 -7.064 5.952 9.350 8.886 -0.658 22.879 3.069 -6.177 12.450 14.725 7.385 -0.243 24.242 3.710 -4.420 22.348 22.995 7.557 -0.377 28.074 4.390 -4.317 25.301 25.380

Bias2 49.045 1.477 328.480 19.347 60.698 249.851 239.323 92.444 1.432 556.681 14.491 52.070 409.960 398.768 61.283 1.187 609.151 19.179 27.921 619.345 615.099 64.217 1.440 810.477 26.070 22.057 658.741 658.097

MSE 55.226 2.594 357.297 26.538 71.499 464.277 391.240 105.931 2.430 589.949 19.566 65.985 664.943 580.737 68.040 2.314 630.705 24.597 36.308 739.314 701.485 71.354 2.737 833.096 32.879 25.487 677.563 672.294

TABLE A1 (contd.) Scenario: #4 Period

Mean Concentration Years

1

biweekly

212

2

5

10

1σ & Bias

Prec.

Bias

Bias2

Max Concentration MSE

Prec.

Bias

Bias2

MSE

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

0.072 0.050 0.560 0.072 0.263 0.803 0.905 0.029 0.020 0.077 0.027 0.133 0.402 0.460 0.011 0.008 0.033 0.282 0.051 0.161 0.187 0.006 0.004 0.016 0.006 0.028 0.079

4.430 0.494 9.214 0.829 -2.928 1.008 4.500 4.096 0.316 7.455 0.006 -2.937 1.000 4.491 3.701 0.352 8.552 2.261 -2.938 1.000 4.487 4.029 0.433 7.988 0.444 -2.948 0.988

19.694 0.294 85.444 0.760 8.836 1.819 21.150 16.807 0.120 55.643 0.027 8.758 1.401 20.624 13.704 0.132 73.169 5.391 8.684 1.161 20.314 16.240 0.192 63.813 0.203 8.717 1.054

19.766 0.344 86.004 0.833 9.099 2.623 22.054 16.836 0.141 55.720 0.055 8.891 1.803 21.084 13.716 0.141 73.203 5.673 8.735 1.322 20.502 16.246 0.195 63.829 0.209 8.745 1.133

5.172 1.750 26.553 5.172 9.815 181.865 128.956 4.529 1.503 20.955 4.915 14.209 254.342 181.706 5.880 1.500 17.675 5.994 12.455 192.185 138.345 5.662 1.687 30.289 6.027 5.728 61.480

4.712 -1.077 17.939 1.111 -7.821 3.306 7.081 4.745 -1.330 18.387 0.644 -7.254 8.680 11.493 5.381 -1.134 19.522 3.451 -5.452 18.716 19.883 6.274 -0.790 23.296 2.870 -4.746 23.800

27.369 2.910 348.313 6.407 70.970 192.791 179.091 27.040 3.271 358.988 5.329 66.830 329.672 313.794 34.834 2.785 398.756 17.900 42.174 542.427 533.631 45.020 2.312 572.882 14.264 28.245 627.799

32.541 4.660 374.866 11.579 80.785 374.656 308.046 31.569 4.773 379.942 10.244 81.038 584.015 495.500 40.714 4.286 416.431 23.894 54.629 734.612 671.976 50.682 3.999 603.170 20.290 33.973 689.279

ts v3

0.097

4.481

20.171

20.268

44.830

24.092

625.131

669.962

TABLE A1 (contd.) Scenario: #4 Period

Mean Concentration Years

1

monthly

213

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.322 0.210 0.605 0.210 0.421 1.538 1.701 0.178 0.085 0.352 0.112 0.212 0.718 0.807 0.042 0.035 0.145 0.044 0.081 0.312 0.345 0.020 0.017 0.046 0.020 0.042 0.146 0.168

Bias 5.035 0.688 6.804 -0.096 -2.938 1.006 4.498 4.790 0.128 5.736 0.715 -2.926 1.005 4.501 3.711 0.460 7.815 0.215 -2.937 1.004 4.491 3.947 0.394 7.115 0.345 -2.946 0.990 4.483

Bias2 25.675 0.684 46.895 0.219 9.051 2.549 21.930 23.124 0.102 33.245 0.623 8.770 1.727 21.067 13.813 0.247 61.213 0.090 8.704 1.321 20.512 15.596 0.173 50.668 0.139 8.722 1.127 20.265

Max Concentration MSE 25.996 0.894 47.501 0.429 9.472 4.087 23.631 23.302 0.187 33.598 0.735 8.982 2.445 21.874 13.855 0.282 61.358 0.135 8.784 1.633 20.858 15.616 0.190 50.714 0.159 8.764 1.273 20.434

Prec. 6.830 2.965 20.301 5.208 8.808 137.547 98.000 8.842 2.833 52.877 5.706 13.465 220.170 157.823 6.144 2.482 35.556 8.414 17.777 264.588 191.186 6.127 2.291 29.369 5.757 11.831 158.160 114.789

Bias 4.522 -1.645 7.485 -2.472 -8.651 0.482 4.656 5.432 -2.503 12.765 0.435 -8.597 3.906 7.406 3.883 -1.729 19.492 0.908 -7.137 12.818 14.827 4.749 -1.710 17.746 0.740 -5.862 19.893 20.743

Bias2 27.272 5.672 76.321 11.318 83.642 137.779 119.678 38.341 9.097 215.801 5.895 87.371 235.429 212.671 21.220 5.470 415.463 9.237 68.708 428.863 411.013 28.676 5.213 344.245 6.304 46.192 553.850 545.000

MSE 34.102 8.636 96.622 16.525 92.450 275.326 217.678 47.182 11.930 268.678 11.600 100.836 455.600 370.494 27.364 7.952 451.019 17.651 86.485 693.451 602.199 34.803 7.504 373.614 12.062 58.024 712.010 659.789

TABLE A1 (contd.) Scenario: #4 Period

Mean Concentration Years

1

semi-annual

214

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 5.772 4.750 4.364 4.996 3.809 14.332 16.620 1.729 2.438 4.227 2.471 1.774 7.217 8.203 1.165 0.841 1.458 0.890 0.650 2.532 2.945 0.452 0.420 0.607 0.489 0.325 1.252 1.464

Bias 4.841 0.162 7.258 -0.165 -2.918 1.021 4.528 2.659 0.739 11.251 1.008 -2.938 1.012 4.502 5.197 0.397 9.498 0.457 -2.926 0.984 4.492 4.324 0.492 8.943 1.005 -2.927 0.994 4.500

Bias2 29.209 4.776 57.034 5.024 12.324 15.374 37.123 8.799 2.984 130.799 3.486 10.405 8.242 28.473 28.176 0.999 91.664 1.100 9.207 3.500 23.122 19.150 0.662 80.573 1.499 8.894 2.240 21.708

Max Concentration MSE 34.981 9.526 61.398 10.020 16.133 29.706 53.743 10.529 5.422 135.026 5.956 12.180 15.459 36.677 29.341 1.840 93.123 1.990 9.857 6.031 26.068 19.602 1.083 81.180 1.988 9.218 3.491 23.171

Prec. 14.973 11.786 12.163 14.938 9.442 42.133 39.709 8.242 10.327 26.028 12.083 8.197 67.603 51.457 15.775 8.037 30.829 10.066 11.708 128.599 94.305 10.731 7.107 28.985 10.461 18.167 214.388 157.149

Bias -1.644 -6.866 0.084 -7.142 -10.849 -5.528 -1.071 -4.289 -5.429 7.808 -4.820 -11.152 -4.703 -0.070 1.437 -5.399 10.062 -4.854 -11.873 -3.755 0.622 -0.060 -5.306 10.861 -3.142 -11.814 -0.930 2.893

Bias2 17.675 58.918 12.171 65.936 127.124 72.688 40.857 26.635 39.801 86.983 35.312 132.546 89.723 51.462 17.840 37.186 132.053 33.623 152.668 142.697 94.691 10.734 35.255 146.942 20.335 157.724 215.254 165.520

MSE 32.648 70.703 24.334 80.874 136.566 114.821 80.566 34.877 50.128 113.011 47.395 140.743 157.325 102.919 33.615 45.222 162.882 43.688 164.376 271.296 188.996 21.465 42.362 175.927 30.795 175.891 429.642 322.669

TABLE A1 (contd.) Scenario: #4 Period

Mean Concentration Years

1

season dependent

215

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.544 0.482 0.679 0.467 0.436 2.349 2.210 0.251 0.262 0.456 0.279 0.227 1.133 1.095 0.114 0.105 0.153 0.109 0.093 0.477 0.464 0.058 0.059 0.067 0.065 0.053 0.232 0.236

Bias 5.540 1.801 8.387 1.349 -1.833 2.813 6.731 5.241 2.064 10.675 2.304 -1.829 2.803 6.726 5.858 2.003 9.578 2.141 -1.820 2.823 6.746 5.753 2.057 9.800 2.356 -1.818 2.828 6.763

Bias2 31.228 3.725 71.005 2.288 3.794 10.260 47.513 27.719 4.522 114.390 5.588 3.573 8.987 46.330 34.421 4.118 91.875 4.692 3.406 8.446 45.967 33.142 4.289 96.066 5.613 3.356 8.228 45.956

Max Concentration MSE 31.773 4.207 71.684 2.755 4.230 12.609 49.723 27.970 4.783 114.846 5.867 3.800 10.120 47.425 34.534 4.223 92.027 4.801 3.499 8.923 46.432 33.200 4.349 96.133 5.678 3.409 8.460 46.192

Prec. 5.526 1.525 14.476 7.620 10.976 214.504 152.072 5.849 1.325 47.062 5.764 14.182 256.787 183.378 4.721 1.287 22.402 4.590 9.110 133.268 95.946 6.459 1.207 30.642 5.589 3.701 24.886 18.518

Bias 5.466 -0.973 12.893 2.581 -7.129 5.847 9.253 4.982 -0.376 24.835 3.976 -6.290 12.057 14.388 7.229 -0.398 24.621 3.554 -4.584 21.761 22.493 8.058 -0.200 29.497 5.009 -4.377 25.089 25.198

Bias2 35.401 2.471 180.700 14.280 61.794 248.691 237.674 30.666 1.467 663.767 21.571 53.741 402.142 390.376 56.972 1.446 628.557 17.222 30.119 606.780 601.829 71.363 1.247 900.402 30.673 22.855 654.097 653.206

MSE 40.927 3.996 195.176 21.900 72.770 463.195 389.747 36.515 2.792 710.829 27.334 67.923 658.929 573.754 61.693 2.733 650.959 21.812 39.229 740.048 697.775 77.822 2.455 931.044 36.262 26.557 678.983 671.723

TABLE A1 (contd.) Scenario: #4 Period

Mean Concentration Years

216

parameter dependent

1

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.315 0.231 0.315 0.303 0.608 2.123 2.106 0.148 0.106 0.358 0.121 0.281 1.004 1.011 0.046 0.043 0.126 0.061 0.108 0.392 0.396 0.024 0.022 0.054 0.028 0.052 0.199 0.194

Bias 5.004 0.816 5.004 0.555 -2.569 1.669 5.271 4.342 0.757 7.325 0.315 -2.658 1.497 5.072 4.083 0.778 7.561 0.658 -2.661 1.492 5.076 4.117 0.861 8.206 1.120 -2.568 1.663 5.275

Bias2 25.354 0.897 25.354 0.611 7.206 4.907 29.885 19.001 0.679 54.009 0.220 7.345 3.245 26.730 16.713 0.648 57.283 0.494 7.188 2.618 26.157 16.971 0.763 67.393 1.283 6.647 2.963 28.020

Max Concentration MSE 25.668 1.128 25.668 0.914 7.814 7.030 31.991 19.149 0.785 54.367 0.341 7.626 4.250 27.741 16.759 0.691 57.409 0.555 7.296 3.011 26.553 16.995 0.785 67.448 1.311 6.699 3.162 28.214

Prec. 13.174 2.341 13.174 7.878 14.167 205.857 148.063 6.503 2.309 23.910 4.822 15.941 255.060 183.238 5.400 2.110 21.610 11.284 15.879 230.694 166.817 5.579 2.064 34.565 9.192 7.383 85.192 62.085

Bias 5.596 -1.770 5.596 0.178 -7.676 3.938 7.616 3.775 -1.764 16.204 -0.627 -7.629 7.367 10.368 4.168 -1.520 17.803 1.990 -6.097 16.461 17.950 4.494 -1.183 23.750 4.357 -4.985 22.963 23.374

Bias2 44.488 5.473 44.488 7.909 73.079 221.362 206.057 20.755 5.419 286.442 5.216 74.143 309.329 290.731 22.771 4.420 338.537 15.244 53.043 501.622 488.983 25.771 3.463 598.594 28.176 32.231 612.428 608.365

MSE 57.662 7.814 57.662 15.787 87.246 427.219 354.119 27.258 7.728 310.352 10.038 90.085 564.389 473.969 28.171 6.530 360.147 26.528 68.921 732.316 655.800 31.350 5.526 633.159 37.368 39.613 697.620 670.451

TABLE A1 (contd.) Scenario: #4 Period

TML1 Years

1

weekly

217

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

1.22E+12 2.50E+11 4.68E+12 7.89E+11 3.79E+12 1.03E+13 1.20E+13 3.67E+11 1.17E+11 1.44E+12 2.21E+11 1.94E+12 4.94E+12 5.95E+12 1.17E+11 4.77E+10 5.52E+11 1.08E+11 7.71E+11 2.01E+12 2.46E+12 5.65E+10 2.21E+10 2.35E+11 3.73E+10 4.17E+11 9.49E+11 1.28E+12

Bias 1.80E+07 1.44E+06 3.52E+07 2.96E+06 -1.21E+07 4.15E+06 1.86E+07 1.63E+07 1.38E+06 3.43E+07 2.24E+06 -1.21E+07 4.15E+06 1.85E+07 1.61E+07 1.77E+06 3.32E+07 1.77E+06 -1.21E+07 4.16E+06 1.86E+07 1.65E+07 1.61E+06 3.43E+07 1.17E+06 -1.21E+07 4.12E+06 1.85E+07

TML2 Bias2

MSE

Prec.

3.24E+14 2.32E+12 1.24E+15 9.57E+12 1.50E+14 2.75E+13 3.56E+14 2.66E+14 2.03E+12 1.18E+15 5.23E+12 1.48E+14 2.22E+13 3.50E+14 2.60E+14 3.20E+12 1.10E+15 3.23E+12 1.48E+14 1.93E+13 3.47E+14 2.72E+14 2.60E+12 1.18E+15 1.41E+12 1.47E+14 1.79E+13 3.44E+14

3.26E+14 2.57E+12 1.25E+15 1.04E+13 1.54E+14 3.78E+13 3.68E+14 2.66E+14 2.15E+12 1.18E+15 5.45E+12 1.50E+14 2.71E+13 3.56E+14 2.60E+14 3.24E+12 1.11E+15 3.34E+12 1.48E+14 2.14E+13 3.49E+14 2.72E+14 2.62E+12 1.18E+15 1.45E+12 1.48E+14 1.89E+13 3.46E+14

1.27E+12 2.66E+11 4.90E+12 8.49E+11 3.79E+12 1.05E+13 1.22E+13 4.01E+11 1.20E+11 1.62E+12 2.41E+11 1.95E+12 5.05E+12 6.03E+12 1.27E+11 5.05E+10 6.14E+11 1.17E+11 7.74E+11 2.05E+12 2.50E+12 6.08E+10 2.34E+10 2.88E+11 4.18E+10 4.09E+11 9.74E+11 1.31E+12

Bias 1.81E+07 1.55E+06 3.54E+07 3.08E+06 -1.20E+07 4.29E+06 1.87E+07 1.65E+07 1.50E+06 3.45E+07 2.36E+06 -1.20E+07 4.29E+06 1.87E+07 1.63E+07 1.89E+06 3.34E+07 1.89E+06 -1.20E+07 4.30E+06 1.87E+07 1.67E+07 1.73E+06 3.46E+07 1.30E+06 -1.20E+07 4.26E+06 1.87E+07

Bias2

MSE

3.30E+14 2.68E+12 1.26E+15 1.03E+13 1.48E+14 2.89E+13 3.63E+14 2.71E+14 2.38E+12 1.19E+15 5.79E+12 1.46E+14 2.34E+13 3.57E+14 2.66E+14 3.64E+12 1.12E+15 3.68E+12 1.46E+14 2.06E+13 3.53E+14 2.77E+14 3.01E+12 1.19E+15 1.72E+12 1.45E+14 1.91E+13 3.51E+14

3.32E+14 2.94E+12 1.26E+15 1.12E+13 1.51E+14 3.27E+13 3.75E+14 2.72E+14 2.50E+12 1.19E+15 6.03E+12 1.48E+14 2.54E+13 3.63E+14 2.66E+14 3.69E+12 1.12E+15 3.80E+12 1.46E+14 2.13E+13 3.56E+14 2.77E+14 3.04E+12 1.19E+15 1.76E+12 1.45E+14 1.95E+13 3.52E+14

TABLE A1 (contd.) Scenario: #4 Period

TML1 Years

1

biweekly

218

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

1.58E+12 8.69E+11 1.13E+13 1.29E+12 4.69E+12 1.38E+13 1.58E+13 6.30E+11 3.59E+11 1.82E+12 4.80E+11 2.37E+12 6.98E+12 8.16E+12 2.35E+11 1.42E+11 8.47E+11 3.25E+12 9.31E+11 2.79E+12 3.30E+12 1.32E+11 6.49E+10 4.33E+11 9.65E+10 4.99E+11 1.34E+12

ts v3

1.70E+12

TML2 Bias2

MSE

Prec.

1.83E+07 2.03E+06 3.80E+07 3.43E+06 -1.21E+07 4.16E+06 1.85E+07 1.69E+07 1.31E+06 3.07E+07 3.22E+04 -1.21E+07 4.14E+06 1.85E+07 1.53E+07 1.45E+06 3.53E+07 2.16E+06 -1.21E+07 4.14E+06 1.85E+07 1.67E+07 1.83E+06 3.29E+07 1.87E+06 -1.21E+07 4.12E+06

3.35E+14 4.99E+12 1.45E+15 1.31E+13 1.50E+14 3.11E+13 3.60E+14 2.86E+14 2.07E+12 9.47E+14 4.81E+11 1.49E+14 2.41E+13 3.52E+14 2.33E+14 2.25E+12 1.25E+15 7.90E+12 1.48E+14 1.99E+13 3.46E+14 2.77E+14 3.40E+12 1.09E+15 3.60E+12 1.47E+14 1.83E+13

3.37E+14 5.86E+12 1.47E+15 1.43E+13 1.55E+14 4.50E+13 3.76E+14 2.86E+14 2.43E+12 9.49E+14 9.61E+11 1.51E+14 3.11E+13 3.60E+14 2.33E+14 2.39E+12 1.25E+15 1.11E+13 1.49E+14 2.27E+13 3.50E+14 2.78E+14 3.47E+12 1.09E+15 3.70E+12 1.48E+14 1.96E+13

1.77E+12 1.00E+12 1.22E+13 1.48E+12 4.76E+12 1.44E+13 1.63E+13 7.24E+11 4.24E+11 2.13E+12 5.58E+11 2.41E+12 7.26E+12 8.42E+12 2.74E+11 1.69E+11 9.57E+11 4.98E+12 9.48E+11 2.87E+12 3.39E+12 1.54E+11 7.78E+10 5.10E+11 1.19E+11 5.02E+11 1.41E+12

1.85E+07

3.45E+14

3.46E+14

1.78E+12

Bias

Bias2

MSE

1.84E+07 2.15E+06 3.82E+07 3.55E+06 -1.20E+07 4.29E+06 1.87E+07 1.70E+07 1.43E+06 3.09E+07 1.48E+05 -1.20E+07 4.29E+06 1.87E+07 1.54E+07 1.57E+06 3.55E+07 9.45E+06 -1.20E+07 4.27E+06 1.87E+07 1.68E+07 1.95E+06 3.32E+07 1.99E+06 -1.20E+07 4.26E+06

3.41E+14 5.61E+12 1.47E+15 1.41E+13 1.48E+14 3.28E+13 3.67E+14 2.91E+14 2.47E+12 9.59E+14 5.80E+11 1.47E+14 2.57E+13 3.59E+14 2.38E+14 2.63E+12 1.26E+15 9.43E+13 1.46E+14 2.11E+13 3.53E+14 2.83E+14 3.88E+12 1.10E+15 4.08E+12 1.45E+14 1.95E+13

3.43E+14 6.61E+12 1.48E+15 1.55E+13 1.53E+14 3.76E+13 3.83E+14 2.92E+14 2.90E+12 9.61E+14 1.14E+12 1.49E+14 2.81E+13 3.68E+14 2.38E+14 2.80E+12 1.26E+15 9.93E+13 1.47E+14 2.20E+13 3.56E+14 2.83E+14 3.96E+12 1.10E+15 4.20E+12 1.46E+14 2.00E+13

1.87E+07

3.51E+14

3.53E+14

Bias

TABLE A1 (contd.) Scenario: #4 Period

TML1 Years

1

monthly

219

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

6.09E+12 3.68E+12 1.13E+13 3.69E+12 7.45E+12 2.66E+13 2.97E+13 3.39E+12 1.49E+12 6.63E+12 1.97E+12 3.74E+12 1.24E+13 1.41E+13 7.72E+11 6.25E+11 2.84E+12 8.01E+11 1.46E+12 5.41E+12 6.06E+12 3.83E+11 3.08E+11 9.70E+11 3.66E+11 7.49E+11 2.55E+12 2.99E+12

Bias 2.08E+07 2.84E+06 2.81E+07 -3.98E+05 -1.21E+07 4.15E+06 1.85E+07 1.98E+07 5.34E+05 2.36E+07 2.96E+06 -1.21E+07 4.15E+06 1.86E+07 1.53E+07 1.89E+06 3.22E+07 8.93E+05 -1.21E+07 4.16E+06 1.85E+07 1.63E+07 1.67E+06 2.94E+07 1.47E+06 -1.21E+07 4.13E+06 1.85E+07

TML2 Bias2

MSE

Prec.

4.38E+14 1.18E+13 7.99E+14 3.85E+12 1.54E+14 4.38E+13 3.74E+14 3.94E+14 1.78E+12 5.66E+14 1.07E+13 1.49E+14 2.97E+13 3.59E+14 2.35E+14 4.21E+12 1.04E+15 1.60E+12 1.48E+14 2.27E+13 3.50E+14 2.67E+14 3.09E+12 8.63E+14 2.52E+12 1.47E+14 1.96E+13 3.46E+14

4.44E+14 1.54E+13 8.10E+14 7.55E+12 1.62E+14 7.03E+13 4.03E+14 3.97E+14 3.27E+12 5.72E+14 1.27E+13 1.53E+14 4.21E+13 3.73E+14 2.36E+14 4.83E+12 1.05E+15 2.40E+12 1.50E+14 2.81E+13 3.56E+14 2.67E+14 3.40E+12 8.64E+14 2.89E+12 1.48E+14 2.22E+13 3.49E+14

6.73E+12 4.17E+12 1.22E+13 4.10E+12 7.74E+12 2.76E+13 3.10E+13 3.73E+12 1.70E+12 7.27E+12 2.27E+12 3.89E+12 1.31E+13 1.48E+13 8.75E+11 7.05E+11 3.11E+12 9.08E+11 1.52E+12 5.70E+12 6.38E+12 4.31E+11 3.50E+11 1.15E+12 4.23E+11 7.81E+11 2.70E+12 3.16E+12

Bias 2.09E+07 2.97E+06 2.83E+07 -2.84E+05 -1.20E+07 4.26E+06 1.87E+07 1.99E+07 6.40E+05 2.38E+07 3.08E+06 -1.20E+07 4.29E+06 1.87E+07 1.55E+07 2.01E+06 3.24E+07 1.00E+06 -1.20E+07 4.29E+06 1.87E+07 1.65E+07 1.79E+06 2.96E+07 1.59E+06 -1.20E+07 4.26E+06 1.87E+07

Bias2

MSE

4.45E+14 1.30E+13 8.10E+14 4.18E+12 1.53E+14 4.58E+13 3.80E+14 4.01E+14 2.11E+12 5.74E+14 1.18E+13 1.47E+14 3.15E+13 3.66E+14 2.40E+14 4.76E+12 1.05E+15 1.91E+12 1.46E+14 2.41E+13 3.56E+14 2.72E+14 3.54E+12 8.75E+14 2.95E+12 1.45E+14 2.08E+13 3.53E+14

4.52E+14 1.71E+13 8.22E+14 8.27E+12 1.60E+14 5.35E+13 4.11E+14 4.04E+14 3.81E+12 5.81E+14 1.40E+13 1.51E+14 3.54E+13 3.81E+14 2.41E+14 5.46E+12 1.06E+15 2.82E+12 1.48E+14 2.56E+13 3.63E+14 2.72E+14 3.89E+12 8.77E+14 3.37E+12 1.46E+14 2.16E+13 3.56E+14

TABLE A1 (contd.) Scenario: #4 Period

TML1 Years

1

semi-annual

220

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

9.90E+13 8.11E+13 7.51E+13 8.59E+13 2.26E+13 2.43E+13 2.43E+13 2.96E+13 4.18E+13 7.34E+13 4.22E+13 1.11E+13 1.18E+13 1.16E+13 2.00E+13 1.44E+13 2.53E+13 1.53E+13 4.67E+12 4.77E+12 4.63E+12 2.68E+14 2.70E+14 2.66E+14 2.66E+14 3.16E+14 3.11E+14 3.09E+14

Bias 2.00E+07 6.65E+05 2.99E+07 -6.88E+05 -3.70E+07 -3.45E+07 -3.23E+07 1.09E+07 3.05E+06 4.64E+07 4.14E+06 -3.70E+07 -3.45E+07 -3.23E+07 2.14E+07 1.62E+06 3.92E+07 1.88E+06 -3.70E+07 -3.46E+07 -3.23E+07 -3.56E+08 -3.72E+08 -3.37E+08 -3.70E+08 -4.11E+08 -4.09E+08 -4.06E+08

TML2 Bias2

MSE

Prec.

4.97E+14 8.15E+13 9.71E+14 8.63E+13 1.39E+15 1.22E+15 1.07E+15 1.49E+14 5.11E+13 2.22E+15 5.94E+13 1.38E+15 1.20E+15 1.06E+15 4.79E+14 1.71E+13 1.56E+15 1.88E+13 1.38E+15 1.20E+15 1.05E+15 1.27E+17 1.39E+17 1.14E+17 1.37E+17 1.69E+17 1.67E+17 1.66E+17

5.96E+14 1.63E+14 1.05E+15 1.72E+14 1.42E+15 1.24E+15 1.09E+15 1.79E+14 9.29E+13 2.30E+15 1.02E+14 1.39E+15 1.22E+15 1.07E+15 4.99E+14 3.15E+13 1.59E+15 3.40E+13 1.38E+15 1.20E+15 1.05E+15 1.28E+17 1.39E+17 1.14E+17 1.37E+17 1.70E+17 1.68E+17 1.66E+17

1.04E+14 8.55E+13 7.93E+13 9.02E+13 6.72E+13 2.51E+14 2.93E+14 3.10E+13 4.39E+13 7.73E+13 4.44E+13 3.17E+13 1.28E+14 1.46E+14 2.10E+13 1.50E+13 2.69E+13 1.59E+13 1.15E+13 4.45E+13 5.20E+13 2.68E+14 2.70E+14 2.66E+14 2.66E+14 2.84E+14 2.72E+14 2.62E+14

Bias 2.01E+07 7.94E+05 3.01E+07 -5.63E+05 -1.19E+07 4.34E+06 1.88E+07 1.11E+07 3.16E+06 4.66E+07 4.28E+06 -1.20E+07 4.32E+06 1.88E+07 2.16E+07 1.74E+06 3.94E+07 2.01E+06 -1.20E+07 4.17E+06 1.87E+07 -3.56E+08 -3.72E+08 -3.37E+08 -3.70E+08 -3.86E+08 -3.70E+08 -3.55E+08

Bias2

MSE

5.10E+14 8.62E+13 9.87E+14 9.05E+13 2.10E+14 2.70E+14 6.48E+14 1.54E+14 5.39E+13 2.25E+15 6.27E+13 1.76E+14 1.46E+14 4.97E+14 4.87E+14 1.80E+13 1.58E+15 2.00E+13 1.55E+14 6.19E+13 4.01E+14 1.27E+17 1.39E+17 1.14E+17 1.37E+17 1.49E+17 1.37E+17 1.27E+17

6.15E+14 1.72E+14 1.07E+15 1.81E+14 2.77E+14 3.37E+14 9.40E+14 1.85E+14 9.79E+13 2.33E+15 1.07E+14 2.08E+14 1.78E+14 6.43E+14 5.08E+14 3.30E+13 1.61E+15 3.59E+13 1.67E+14 7.35E+13 4.53E+14 1.27E+17 1.39E+17 1.14E+17 1.37E+17 1.50E+17 1.37E+17 1.27E+17

TABLE A1 (contd.) Scenario: #4 Period

TML1 Years

1

season dependent

221

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

9.50E+12 3.97E+12 3.06E+13 8.82E+12 7.49E+12 2.93E+13 3.24E+13 3.53E+12 1.92E+12 1.90E+13 5.28E+12 3.76E+12 1.41E+13 1.60E+13 1.29E+12 7.77E+11 5.01E+12 9.76E+11 1.43E+12 5.78E+12 6.62E+12 5.47E+11 4.61E+11 1.35E+12 5.78E+11 7.95E+11 2.98E+12 3.62E+12

Bias 1.71E+07 1.16E+06 2.99E+07 5.08E+05 -1.18E+07 4.61E+06 1.91E+07 1.56E+07 2.24E+06 3.85E+07 3.14E+06 -1.19E+07 4.52E+06 1.90E+07 1.75E+07 1.92E+06 3.34E+07 2.50E+06 -1.19E+07 4.57E+06 1.91E+07 1.64E+07 2.24E+06 3.48E+07 4.36E+06 -1.19E+07 4.49E+06 1.90E+07

TML2 Bias2

MSE

Prec.

3.03E+14 5.31E+12 9.25E+14 9.08E+12 1.48E+14 5.06E+13 3.97E+14 2.48E+14 6.92E+12 1.50E+15 1.51E+13 1.45E+14 3.45E+13 3.78E+14 3.07E+14 4.45E+12 1.12E+15 7.24E+12 1.42E+14 2.67E+13 3.70E+14 2.68E+14 5.49E+12 1.21E+15 1.96E+13 1.42E+14 2.31E+13 3.64E+14

3.13E+14 9.29E+12 9.56E+14 1.79E+13 1.55E+14 7.99E+13 4.30E+14 2.51E+14 8.84E+12 1.52E+15 2.04E+13 1.48E+14 4.86E+13 3.94E+14 3.08E+14 5.23E+12 1.12E+15 8.21E+12 1.43E+14 3.24E+13 3.77E+14 2.69E+14 5.95E+12 1.21E+15 2.02E+13 1.43E+14 2.61E+13 3.67E+14

9.90E+12 8.44E+12 1.33E+13 8.24E+12 7.69E+12 4.12E+13 3.91E+13 4.60E+12 4.59E+12 9.14E+12 4.93E+12 4.02E+12 1.99E+13 1.93E+13 2.02E+12 1.85E+12 2.96E+12 1.90E+12 1.71E+12 8.43E+12 8.18E+12 9.98E+11 9.98E+11 1.38E+12 1.14E+12 9.54E+11 4.07E+12 4.09E+12

Bias 2.30E+07 7.56E+06 3.48E+07 5.68E+06 -7.47E+06 1.18E+07 2.79E+07 2.18E+07 8.65E+06 4.42E+07 9.63E+06 -7.45E+06 1.17E+07 2.79E+07 2.43E+07 8.39E+06 3.97E+07 8.97E+06 -7.42E+06 1.18E+07 2.80E+07 2.39E+07 8.66E+06 4.07E+07 9.91E+06 -7.35E+06 1.19E+07 2.81E+07

Bias2

MSE

5.39E+14 6.55E+13 1.22E+15 4.05E+13 6.34E+13 1.79E+14 8.20E+14 4.78E+14 7.94E+13 1.97E+15 9.77E+13 5.96E+13 1.57E+14 8.00E+14 5.94E+14 7.23E+13 1.58E+15 8.23E+13 5.68E+13 1.48E+14 7.94E+14 5.74E+14 7.60E+13 1.66E+15 9.94E+13 5.50E+13 1.45E+14 7.96E+14

5.49E+14 7.40E+13 1.24E+15 4.87E+13 7.11E+13 1.87E+14 8.59E+14 4.83E+14 8.40E+13 1.97E+15 1.03E+14 6.36E+13 1.61E+14 8.19E+14 5.96E+14 7.41E+13 1.58E+15 8.42E+13 5.85E+13 1.50E+14 8.02E+14 5.75E+14 7.70E+13 1.66E+15 1.00E+14 5.60E+13 1.46E+14 8.00E+14

TABLE A1 (contd.) Scenario: #4 Period

TML1 Years

222

parameter dependent

1

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

6.36E+12 3.96E+12 6.36E+12 6.94E+12 7.71E+12 2.60E+13 2.90E+13 3.41E+12 1.93E+12 1.04E+13 2.42E+12 3.83E+12 1.39E+13 1.54E+13 9.63E+11 7.56E+11 3.68E+12 1.21E+12 1.60E+12 5.59E+12 6.10E+12 5.13E+11 3.79E+11 1.51E+12 6.24E+11 8.32E+11 2.77E+12 3.09E+12

Bias 1.84E+07 1.53E+06 1.84E+07 5.43E+05 -1.24E+07 3.60E+06 1.79E+07 1.64E+07 1.46E+06 2.92E+07 -3.79E+05 -1.22E+07 3.94E+06 1.83E+07 1.52E+07 1.52E+06 2.97E+07 1.02E+06 -1.22E+07 3.98E+06 1.84E+07 1.54E+07 1.92E+06 3.20E+07 2.99E+06 -1.22E+07 4.04E+06 1.84E+07

TML2 Bias2

MSE

Prec.

3.44E+14 6.29E+12 3.44E+14 7.23E+12 1.62E+14 3.90E+13 3.50E+14 2.71E+14 4.07E+12 8.62E+14 2.56E+12 1.53E+14 2.94E+13 3.50E+14 2.32E+14 3.06E+12 8.85E+14 2.25E+12 1.51E+14 2.15E+13 3.43E+14 2.37E+14 4.05E+12 1.03E+15 9.56E+12 1.49E+14 1.91E+13 3.43E+14

3.51E+14 1.03E+13 3.51E+14 1.42E+13 1.69E+14 6.50E+13 3.79E+14 2.75E+14 6.00E+12 8.72E+14 4.98E+12 1.57E+14 4.33E+13 3.65E+14 2.33E+14 3.82E+12 8.88E+14 3.46E+12 1.52E+14 2.71E+13 3.49E+14 2.37E+14 4.43E+12 1.03E+15 1.02E+13 1.49E+14 2.19E+13 3.46E+14

9.66E+12 9.45E+12 9.66E+12 1.28E+13 1.10E+13 3.89E+13 3.88E+13 6.33E+12 6.20E+12 9.35E+12 7.09E+12 7.17E+12 1.81E+13 1.85E+13 3.20E+12 3.37E+12 3.82E+12 3.90E+12 3.36E+12 7.30E+12 7.63E+12 1.96E+12 1.80E+12 2.56E+12 1.80E+12 2.01E+12 3.97E+12 4.01E+12

Bias 2.20E+07 4.27E+06 2.20E+07 3.36E+06 -1.05E+07 7.06E+06 2.19E+07 1.77E+07 3.36E+06 2.92E+07 1.43E+06 -1.08E+07 6.43E+06 2.12E+07 1.75E+07 2.93E+06 3.19E+07 2.74E+06 -1.08E+07 6.41E+06 2.13E+07 1.82E+07 4.36E+06 3.54E+07 5.38E+06 -1.04E+07 7.17E+06 2.21E+07

Bias2

MSE

4.96E+14 2.77E+13 4.96E+14 2.41E+13 1.21E+14 8.87E+13 5.20E+14 3.20E+14 1.75E+13 8.61E+14 9.13E+12 1.24E+14 5.94E+13 4.69E+14 3.10E+14 1.20E+13 1.02E+15 1.14E+13 1.20E+14 4.84E+13 4.60E+14 3.31E+14 2.08E+13 1.26E+15 3.08E+13 1.10E+14 5.54E+13 4.94E+14

5.05E+14 3.71E+13 5.05E+14 3.70E+13 1.32E+14 9.97E+13 5.59E+14 3.26E+14 2.37E+13 8.71E+14 1.62E+13 1.31E+14 6.66E+13 4.87E+14 3.13E+14 1.53E+13 1.03E+15 1.53E+13 1.24E+14 5.18E+13 4.68E+14 3.33E+14 2.26E+13 1.26E+15 3.26E+13 1.12E+14 5.74E+13 4.98E+14

TABLE A1 (contd.) Scenario: #5 Period

Mean Concentration Years

1

weekly

223

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.012 0.010 0.031 0.014 2.148 74.472 54.538 0.006 0.005 0.131 0.017 1.658 48.429 34.557 0.001 0.001 0.004 0.001 0.016 0.227 0.175 0.001 0.001 0.004 0.001 0.010 0.063 0.208

Bias 3.924 0.252 7.829 0.232 -4.104 6.016 9.278 3.825 0.484 6.905 -0.012 -4.169 5.422 8.997 3.866 0.395 8.010 0.352 -3.910 5.205 8.551 3.911 0.388 7.978 0.308 -2.526 1.226 4.979

Bias2 15.412 0.074 61.319 0.068 18.993 110.666 140.610 14.634 0.239 47.808 0.017 19.039 77.819 115.493 14.942 0.157 64.148 0.125 15.295 27.306 73.261 15.288 0.151 63.633 0.096 6.391 1.567 24.993

Max Concentration MSE 15.424 0.083 61.350 0.083 21.141 185.138 195.148 14.640 0.244 47.939 0.035 20.697 126.248 150.050 14.943 0.157 64.152 0.126 15.311 27.533 73.436 15.289 0.152 63.637 0.097 6.401 1.630 25.201

Prec. 6.541 0.905 19.939 4.761 8.336 176.583 118.723 4.895 1.107 57.058 3.656 6.700 186.996 126.305 3.913 1.108 18.839 2.969 0.946 0.946 0.946 3.639 0.480 8.038 2.012 1.142 2.171 1.858

Bias 7.607 -0.067 19.523 3.843 -5.680 11.989 14.423 6.869 0.250 26.321 2.889 -6.212 12.812 15.008 9.138 0.280 40.373 5.913 -8.350 21.661 21.661 7.806 0.089 24.250 3.851 -6.910 -1.882 3.115

Bias2

MSE

64.402 0.910 401.039 19.525 40.600 320.300 326.712 52.070 1.170 749.760 12.002 45.285 351.138 351.527 87.395 1.186 1648.252 37.919 70.644 469.963 469.963 64.555 0.487 595.879 16.834 48.880 5.711 11.555

70.944 1.815 420.978 24.286 48.936 496.883 445.435 56.965 2.277 806.819 15.658 51.986 538.134 477.832 91.308 2.294 1667.092 40.888 71.589 470.909 470.909 68.194 0.967 603.917 18.846 50.022 7.882 13.413

TABLE A1 (contd.) Scenario: #5 Period

Mean Concentration Years

1

biweekly

224

2

5

10

1σ & Bias

Prec.

Bias

Bias2

Max Concentration MSE

Prec.

Bias

Bias2

MSE

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

0.092 0.036 0.147 0.194 2.273 61.161 45.352 0.022 0.017 0.148 0.027 1.711 46.986 34.455 0.009 0.007 0.041 0.008 0.666 16.226 11.908 0.004 0.003 0.009 0.004 0.024 0.320

4.594 0.310 7.387 -0.794 -4.079 5.642 8.996 3.855 0.323 6.720 0.618 -3.992 4.872 8.430 3.737 0.285 7.138 0.367 -4.390 4.114 7.996 3.710 0.413 7.619 0.498 -3.926 5.186

21.191 0.132 54.705 0.825 18.912 92.992 126.279 14.881 0.121 45.303 0.408 17.643 70.716 105.512 13.972 0.088 50.983 0.143 19.932 33.147 75.832 13.763 0.174 58.042 0.252 15.438 27.212

21.283 0.168 54.853 1.019 21.185 154.152 171.631 14.903 0.138 45.451 0.434 19.355 117.701 139.967 13.981 0.095 51.024 0.150 20.599 49.373 87.741 13.767 0.177 58.050 0.256 15.463 27.532

7.682 1.164 106.369 7.843 7.342 182.143 124.124 4.962 1.075 26.293 5.040 7.161 185.499 125.016 4.258 1.295 32.746 4.036 9.360 150.030 105.438 6.816 1.464 15.809 4.105 0.944 0.944

7.297 -0.655 29.647 2.375 -5.918 10.872 13.432 6.620 -0.885 19.760 3.173 -6.554 9.905 12.613 6.985 -0.511 28.748 4.247 -6.590 18.220 19.259 10.642 -0.092 38.627 7.497 -8.373 21.632

60.916 1.593 985.201 13.485 42.366 300.334 304.527 48.783 1.858 416.721 15.105 50.110 283.594 284.089 53.043 1.556 859.110 22.071 52.778 481.981 476.312 120.038 1.473 1507.611 60.301 71.039 468.820

68.599 2.758 1091.571 21.327 49.708 482.477 428.651 53.745 2.934 443.013 20.145 57.271 469.093 409.105 57.301 2.850 891.856 26.106 62.138 632.011 581.750 126.854 2.937 1523.421 64.406 71.983 469.764

ts v3

0.246

8.532

73.021

73.267

0.944

21.632

468.820

469.764

TABLE A1 (contd.) Scenario: #5 Period

Mean Concentration Years

1

monthly

225

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.202 0.146 1.635 0.295 2.571 75.494 55.650 0.084 0.071 0.320 0.123 2.004 45.315 32.702 0.032 0.028 0.049 0.035 0.818 19.337 14.317 0.438 0.439 0.445 0.443 0.455 0.907 0.799

Bias 4.472 0.289 4.459 1.273 -4.181 5.942 9.211 3.566 0.199 9.126 -0.507 -4.242 4.737 8.392 3.624 0.251 7.824 -0.210 -4.449 4.698 8.450 7.558 4.280 10.795 4.301 -0.045 9.056 12.401

Bias2 20.195 0.230 21.516 1.915 20.050 110.801 140.487 12.799 0.111 83.599 0.380 19.998 67.752 103.122 13.164 0.091 61.254 0.080 20.614 41.403 85.717 57.562 18.757 116.974 18.937 0.457 82.909 154.579

Max Concentration MSE 20.397 0.376 23.151 2.209 22.621 186.295 196.137 12.883 0.182 83.919 0.503 22.002 113.067 135.824 13.196 0.118 61.303 0.115 21.432 60.740 100.034 58.001 19.195 117.418 19.380 0.912 83.816 155.377

Prec. 2.350 1.491 29.712 5.795 8.850 188.707 132.648 4.268 1.540 69.354 3.158 7.447 193.100 131.141 4.749 1.706 19.562 4.688 8.135 170.377 118.193 12.431 11.067 21.399 11.388 9.351 9.351 9.351

Bias 3.226 -1.566 16.163 1.333 -6.666 9.324 11.911 3.877 -1.586 28.312 0.399 -6.856 9.997 12.623 6.148 -0.879 21.362 1.852 -7.522 15.717 17.070 3.039 -5.627 20.513 -2.463 -13.862 16.141 16.141

Bias2 12.755 3.944 290.939 7.573 53.277 275.628 274.497 19.300 4.055 870.853 3.317 54.446 293.036 290.466 42.549 2.477 475.862 8.116 64.718 417.386 409.548 21.668 42.728 442.120 17.454 201.486 269.857 269.857

MSE 15.105 5.435 320.651 13.368 62.127 464.335 407.145 23.568 5.595 940.207 6.475 61.892 486.136 421.606 47.298 4.183 495.425 12.804 72.853 587.763 527.741 34.100 53.796 463.519 28.842 210.837 279.208 279.208

TABLE A1 (contd.) Scenario: #5 Period

Mean Concentration Years

1

semi-annual

226

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 2.483 2.086 3.201 2.483 5.880 109.868 83.264 1.265 0.937 4.332 1.508 3.915 69.579 51.668 0.379 0.369 0.511 0.406 1.446 20.227 15.552 0.198 0.186 0.290 0.198 0.404 4.338 3.445

Bias 4.663 0.215 9.639 1.063 -4.340 5.526 8.753 4.561 -0.208 12.734 1.666 -4.643 4.608 8.168 3.082 -0.128 7.883 -0.110 -4.629 3.065 6.988 3.837 0.025 8.607 0.237 -4.347 5.339 8.634

Bias2 24.226 2.132 96.111 3.612 24.711 140.400 159.875 22.062 0.981 166.466 4.283 25.468 90.809 118.385 9.876 0.386 62.646 0.418 22.874 29.623 64.383 14.922 0.186 74.355 0.254 19.299 32.844 77.986

Max Concentration MSE 26.709 4.219 99.312 6.095 30.591 250.269 243.139 23.328 1.918 170.798 5.791 29.383 160.388 170.052 10.255 0.755 63.157 0.823 24.320 49.850 79.935 15.120 0.372 74.646 0.452 19.702 37.182 81.431

Prec. 4.820 3.730 10.666 4.820 9.379 168.370 124.680 6.489 4.042 11.292 4.478 10.136 191.676 137.153 7.376 3.232 22.017 9.783 10.947 227.534 159.338 6.525 3.312 9.699 6.525 1.243 1.654 1.532

Bias 0.532 -4.095 7.561 -3.068 -8.780 2.944 6.032 1.844 -3.431 11.187 -1.714 -8.965 4.544 7.603 2.742 -4.250 19.056 0.349 -9.612 7.007 9.684 3.273 -3.691 17.000 -0.327 -9.023 20.970 20.972

Bias2 5.103 20.493 67.829 14.233 86.453 177.035 161.067 9.891 15.815 136.424 7.417 90.497 212.324 194.957 14.897 21.292 385.118 9.905 103.325 276.632 253.111 17.238 16.936 298.675 6.632 82.641 441.371 441.333

MSE 9.923 24.223 78.496 19.053 95.833 345.405 285.747 16.380 19.856 147.716 11.895 100.633 404.000 332.111 22.273 24.524 407.135 19.688 114.271 504.165 412.449 23.763 20.248 308.374 13.156 83.884 443.025 442.864

TABLE A1 (contd.) Scenario: #5 Period

Mean Concentration Years

1

season dependent

227

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.182 0.137 0.311 0.211 2.040 75.620 53.844 0.090 0.085 0.106 0.125 1.822 46.328 33.221 0.053 0.052 0.158 0.050 0.713 11.882 8.959 0.010 0.011 0.018 0.012 0.030 0.334 0.262

Bias 4.437 0.216 6.703 1.100 -4.110 6.559 9.904 3.772 0.314 7.922 1.003 -4.176 4.620 8.307 3.776 0.397 9.530 0.286 -4.353 3.503 7.489 4.304 0.549 7.801 0.678 -3.746 5.652 8.975

Bias2 19.872 0.183 45.238 1.422 18.927 118.641 151.925 14.320 0.184 62.850 1.130 19.263 67.667 102.214 14.307 0.210 90.972 0.132 19.657 24.150 65.038 18.531 0.312 60.859 0.471 14.061 32.270 80.786

Max Concentration MSE 20.054 0.320 45.549 1.633 20.967 194.261 205.769 14.410 0.269 62.956 1.254 21.085 113.994 135.434 14.360 0.262 91.130 0.183 20.370 36.032 73.998 18.542 0.323 60.877 0.483 14.091 32.604 81.048

Prec. 6.686 1.317 50.166 12.143 7.478 183.692 124.150 7.183 1.278 25.479 6.468 7.885 196.189 133.545 3.839 1.320 54.984 3.792 8.677 155.639 109.260 4.780 1.388 17.403 3.340 0.946 0.946 0.946

Bias 7.684 -0.638 22.950 6.176 -6.352 10.835 13.368 7.760 -0.492 21.811 3.995 -6.320 11.532 13.962 6.947 -0.318 34.727 3.327 -6.343 18.237 19.321 10.141 0.395 38.159 6.387 -8.350 21.661 21.661

Bias2

MSE

65.727 1.724 576.817 50.282 47.824 301.073 302.843 67.398 1.520 501.156 22.423 47.826 329.151 328.456 52.091 1.421 1260.840 14.857 48.912 488.178 482.530 107.593 1.544 1472.962 44.120 70.644 469.963 469.963

72.414 3.041 626.983 62.425 55.303 484.764 426.993 74.580 2.798 526.634 28.891 55.711 525.340 462.001 55.929 2.740 1315.823 18.650 57.589 643.818 591.790 112.373 2.932 1490.366 47.460 71.589 470.909 470.909

TABLE A1 (contd.) Scenario: #5 Period

Mean Concentration Years

228

parameter dependent

1

2

5

10

1σ & Bias 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

Prec. 0.164 0.140 1.016 0.177 2.412 72.714 51.721 0.126 0.071 0.352 0.125 2.250 49.762 35.022 0.042 0.033 0.062 0.042 0.790 15.516 11.294 0.451 0.452 0.456 0.454 0.481 1.029 0.901

Bias 4.380 0.542 8.273 -0.063 -3.814 6.885 10.176 4.350 0.318 9.364 0.716 -3.933 6.274 9.698 3.571 0.349 8.113 -0.029 -4.013 4.849 8.680 7.921 4.333 4.118 4.306 0.742 11.519 14.612

Bias2 19.348 0.434 69.444 0.181 16.961 120.109 155.261 19.044 0.172 88.022 0.638 17.715 89.121 129.063 12.796 0.155 65.871 0.043 16.889 39.028 86.632 63.189 19.227 17.412 18.990 1.031 133.693 214.386

Max Concentration MSE 19.512 0.574 70.460 0.359 19.373 192.823 206.981 19.170 0.242 88.374 0.762 19.965 138.883 164.085 12.839 0.189 65.932 0.086 17.679 54.544 97.926 63.639 19.679 17.868 19.444 1.513 134.722 215.287

Prec. 4.570 1.371 59.612 5.259 14.993 203.079 141.523 5.734 1.663 30.672 14.960 20.689 188.899 135.117 4.109 1.583 23.334 4.109 11.284 82.176 60.120 16.695 10.199 12.690 12.915 9.351 9.351 9.351

Bias 4.429 -1.248 23.524 0.633 -5.248 13.809 15.974 6.006 -1.215 20.048 6.748 -4.587 18.428 19.826 5.451 -1.251 24.421 1.850 -5.161 22.687 23.118 5.233 -6.492 -0.387 0.424 -13.862 16.141 16.141

Bias2 24.187 2.928 612.935 5.660 42.536 393.741 396.672 41.796 3.141 432.568 60.497 41.723 528.467 528.146 33.814 3.148 619.648 7.532 37.913 596.833 594.519 44.078 52.341 12.840 13.095 201.486 269.857 269.857

MSE 28.757 4.298 672.547 10.919 57.528 596.820 538.195 47.530 4.804 463.239 75.458 62.413 717.367 663.263 37.923 4.731 642.982 11.640 49.197 679.009 654.639 60.773 62.540 25.530 26.011 210.837 279.208 279.208

TABLE A1 (contd.) TML1

Scenario: #5 Period

Years

1

weekly

229

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

4.88E+11 1.84E+11 1.70E+12 2.96E+11 3.74E+13 1.27E+15 9.28E+14 2.32E+11 8.99E+10 2.66E+12 3.14E+11 2.88E+13 8.23E+14 5.88E+14 4.10E+10 1.35E+10 1.66E+11 2.34E+10 2.88E+11 3.83E+12 3.10E+12 5.54E+10 1.02E+10 2.63E+11 2.05E+10 1.74E+11 1.13E+12 3.80E+12

Bias 1.62E+07 1.05E+06 3.23E+07 9.62E+05 -1.69E+07 2.48E+07 3.83E+07 1.58E+07 2.02E+06 2.85E+07 -2.51E+04 -1.72E+07 2.24E+07 3.71E+07 1.60E+07 1.73E+06 3.31E+07 1.55E+06 -1.60E+07 2.16E+07 3.54E+07 1.59E+07 1.42E+06 3.27E+07 1.09E+06 -1.06E+07 4.96E+06 2.05E+07

TML2 Bias2

MSE

Prec.

2.63E+14 1.29E+12 1.04E+15 1.22E+12 3.24E+14 1.88E+15 2.39E+15 2.50E+14 4.16E+12 8.14E+14 3.15E+11 3.24E+14 1.32E+15 1.97E+15 2.57E+14 3.02E+12 1.10E+15 2.42E+12 2.57E+14 4.69E+14 1.25E+15 2.54E+14 2.03E+12 1.07E+15 1.21E+12 1.12E+14 2.58E+13 4.25E+14

2.63E+14 1.48E+12 1.05E+15 1.52E+12 3.61E+14 3.15E+15 3.32E+15 2.50E+14 4.25E+12 8.17E+14 6.30E+11 3.53E+14 2.15E+15 2.55E+15 2.57E+14 3.03E+12 1.10E+15 2.44E+12 2.58E+14 4.72E+14 1.26E+15 2.54E+14 2.04E+12 1.07E+15 1.23E+12 1.13E+14 2.69E+13 4.28E+14

5.47E+11 1.61E+11 2.15E+12 3.38E+11 3.71E+13 1.28E+15 9.37E+14 2.44E+11 1.11E+11 2.81E+12 2.92E+11 2.85E+13 8.30E+14 5.94E+14 5.05E+10 1.43E+10 2.27E+11 2.98E+10 2.86E+11 3.87E+12 3.13E+12 5.74E+10 1.06E+10 2.82E+11 2.28E+10 1.82E+11 1.15E+12 3.84E+12

Bias 1.64E+07 1.22E+06 3.26E+07 1.15E+06 -1.68E+07 2.51E+07 3.86E+07 1.60E+07 2.19E+06 2.87E+07 1.43E+05 -1.71E+07 2.26E+07 3.74E+07 1.62E+07 1.90E+06 3.34E+07 1.72E+06 -1.59E+07 2.18E+07 3.56E+07 1.63E+07 1.72E+06 3.31E+07 1.38E+06 -1.03E+07 5.34E+06 2.10E+07

Bias2

MSE

2.70E+14 1.65E+12 1.06E+15 1.65E+12 3.19E+14 1.91E+15 2.42E+15 2.56E+14 4.92E+12 8.28E+14 3.12E+11 3.20E+14 1.34E+15 1.99E+15 2.64E+14 3.63E+12 1.11E+15 2.97E+12 2.53E+14 4.79E+14 1.27E+15 2.64E+14 2.96E+12 1.09E+15 1.94E+12 1.07E+14 2.97E+13 4.46E+14

2.70E+14 1.81E+12 1.06E+15 1.99E+12 3.56E+14 1.94E+15 3.36E+15 2.57E+14 5.04E+12 8.31E+14 6.04E+11 3.48E+14 1.37E+15 2.59E+15 2.64E+14 3.65E+12 1.11E+15 3.00E+12 2.54E+14 4.79E+14 1.28E+15 2.65E+14 2.97E+12 1.09E+15 1.96E+12 1.07E+14 2.99E+13 4.50E+14

TABLE A1 (contd.) Scenario: #5 Period

TML1 Years

1

biweekly

230

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2

1.97E+12 6.29E+11 3.95E+12 3.44E+12 3.95E+13 1.04E+15 7.73E+14 5.26E+11 3.00E+11 3.09E+12 4.79E+11 2.97E+13 7.99E+14 5.87E+14 1.98E+11 1.18E+11 9.02E+11 1.47E+11 1.15E+13 2.76E+14 2.03E+14 9.08E+10 5.36E+10 2.64E+11 6.75E+10 4.41E+11 5.40E+12

ts v3

4.31E+12

TML2 Bias2

MSE

Prec.

1.90E+07 1.29E+06 3.05E+07 -3.26E+06 -1.68E+07 2.33E+07 3.71E+07 1.59E+07 1.36E+06 2.77E+07 2.57E+06 -1.64E+07 2.01E+07 3.48E+07 1.55E+07 1.23E+06 2.95E+07 1.57E+06 -1.81E+07 1.70E+07 3.30E+07 1.54E+07 1.81E+06 3.15E+07 2.16E+06 -1.61E+07 2.15E+07

3.61E+14 2.29E+12 9.32E+14 1.41E+13 3.22E+14 1.58E+15 2.15E+15 2.54E+14 2.14E+12 7.72E+14 7.09E+12 3.00E+14 1.20E+15 1.80E+15 2.39E+14 1.63E+12 8.70E+14 2.61E+12 3.38E+14 5.66E+14 1.29E+15 2.37E+14 3.31E+12 9.94E+14 4.71E+12 2.60E+14 4.67E+14

3.63E+14 2.92E+12 9.36E+14 1.75E+13 3.62E+14 2.62E+15 2.92E+15 2.55E+14 2.44E+12 7.75E+14 7.57E+12 3.30E+14 2.00E+15 2.38E+15 2.40E+14 1.75E+12 8.71E+14 2.76E+12 3.50E+14 8.42E+14 1.50E+15 2.37E+14 3.37E+12 9.95E+14 4.78E+12 2.60E+14 4.73E+14

2.27E+12 6.55E+11 5.13E+12 3.43E+12 3.93E+13 1.05E+15 7.81E+14 5.92E+11 3.12E+11 3.51E+12 5.87E+11 2.95E+13 8.06E+14 5.92E+14 2.19E+11 1.22E+11 1.06E+12 1.83E+11 1.14E+13 2.79E+14 2.05E+14 1.06E+11 5.95E+10 4.10E+11 8.15E+10 4.45E+11 5.49E+12

3.53E+07

1.25E+15

1.25E+15

4.38E+12

Bias

Bias2

MSE

1.92E+07 1.47E+06 3.07E+07 -3.10E+06 -1.67E+07 2.35E+07 3.74E+07 1.61E+07 1.52E+06 2.80E+07 2.74E+06 -1.63E+07 2.03E+07 3.50E+07 1.57E+07 1.40E+06 2.97E+07 1.75E+06 -1.79E+07 1.72E+07 3.33E+07 1.56E+07 1.98E+06 3.18E+07 2.31E+06 -1.60E+07 2.17E+07

3.70E+14 2.81E+12 9.49E+14 1.31E+13 3.18E+14 1.60E+15 2.18E+15 2.61E+14 2.62E+12 7.86E+14 8.10E+12 2.96E+14 1.22E+15 1.82E+15 2.46E+14 2.09E+12 8.85E+14 3.26E+12 3.33E+14 5.76E+14 1.31E+15 2.43E+14 3.97E+12 1.01E+15 5.44E+12 2.56E+14 4.77E+14

3.72E+14 3.47E+12 9.54E+14 1.65E+13 3.57E+14 1.64E+15 2.96E+15 2.62E+14 2.93E+12 7.90E+14 8.69E+12 3.26E+14 1.25E+15 2.41E+15 2.46E+14 2.21E+12 8.86E+14 3.44E+12 3.45E+14 5.88E+14 1.52E+15 2.44E+14 4.03E+12 1.01E+15 5.52E+12 2.56E+14 4.78E+14

3.56E+07

1.27E+15

1.27E+15

Bias

TABLE A1 (contd.) Scenario: #5 Period

TML1 Years

1

monthly

231

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

3.83E+12 2.54E+12 2.88E+13 5.16E+12 4.48E+13 1.29E+15 9.48E+14 1.58E+12 1.24E+12 7.03E+12 2.20E+12 3.48E+13 7.71E+14 5.57E+14 6.31E+11 4.93E+11 1.20E+12 6.48E+11 1.42E+13 3.29E+14 2.44E+14 1.31E+13 1.38E+13 1.23E+13 1.39E+13 1.49E+13 1.96E+13 1.72E+13

Bias 1.84E+07 1.21E+06 1.84E+07 5.26E+06 -1.72E+07 2.45E+07 3.80E+07 1.47E+07 8.47E+05 3.77E+07 -2.07E+06 -1.75E+07 1.96E+07 3.46E+07 1.50E+07 1.09E+06 3.23E+07 -8.16E+05 -1.83E+07 1.94E+07 3.49E+07 2.49E+07 1.13E+07 3.82E+07 1.14E+07 -6.50E+06 3.11E+07 4.48E+07

TML2 Bias2

MSE

Prec.

3.44E+14 4.00E+12 3.67E+14 3.29E+13 3.42E+14 1.89E+15 2.39E+15 2.19E+14 1.95E+12 1.42E+15 6.49E+12 3.40E+14 1.15E+15 1.76E+15 2.26E+14 1.68E+12 1.05E+15 1.31E+12 3.49E+14 7.06E+14 1.46E+15 6.31E+14 1.42E+14 1.47E+15 1.45E+14 5.71E+13 9.84E+14 2.03E+15

3.48E+14 6.54E+12 3.96E+14 3.80E+13 3.87E+14 3.17E+15 3.34E+15 2.20E+14 3.19E+12 1.43E+15 8.69E+12 3.75E+14 1.92E+15 2.31E+15 2.26E+14 2.18E+12 1.05E+15 1.96E+12 3.64E+14 1.04E+15 1.71E+15 6.44E+14 1.56E+14 1.48E+15 1.59E+14 7.20E+13 1.00E+15 2.05E+15

4.09E+12 2.62E+12 3.09E+13 5.53E+12 4.47E+13 1.30E+15 9.58E+14 1.67E+12 1.27E+12 8.63E+12 2.27E+12 3.46E+13 7.79E+14 5.64E+14 6.76E+11 5.03E+11 1.56E+12 6.83E+11 1.40E+13 3.32E+14 2.47E+14 1.32E+13 1.38E+13 1.25E+13 1.40E+13 1.50E+13 2.01E+13 1.76E+13

Bias 1.87E+07 1.37E+06 1.86E+07 5.43E+06 -1.71E+07 2.48E+07 3.83E+07 1.49E+07 1.02E+06 3.79E+07 -1.90E+06 -1.74E+07 1.98E+07 3.49E+07 1.52E+07 1.27E+06 3.26E+07 -6.47E+05 -1.82E+07 1.97E+07 3.52E+07 2.51E+07 1.15E+07 3.84E+07 1.16E+07 -6.38E+06 3.13E+07 4.51E+07

Bias2

MSE

3.52E+14 4.50E+12 3.78E+14 3.50E+13 3.38E+14 1.91E+15 2.42E+15 2.25E+14 2.32E+12 1.44E+15 5.87E+12 3.36E+14 1.17E+15 1.78E+15 2.32E+14 2.11E+12 1.06E+15 1.10E+12 3.45E+14 7.18E+14 1.48E+15 6.41E+14 1.46E+14 1.49E+15 1.49E+14 5.57E+13 9.98E+14 2.05E+15

3.56E+14 7.11E+12 4.08E+14 4.05E+13 3.82E+14 1.95E+15 3.38E+15 2.26E+14 3.59E+12 1.45E+15 8.14E+12 3.70E+14 1.21E+15 2.35E+15 2.33E+14 2.61E+12 1.06E+15 1.78E+12 3.59E+14 7.32E+14 1.73E+15 6.54E+14 1.60E+14 1.50E+15 1.63E+14 7.07E+13 1.01E+15 2.07E+15

TABLE A1 (contd.) Scenario: #5 Period

TML1 Years

1

semi-annual

232

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

4.27E+13 3.55E+13 5.64E+13 4.23E+13 3.81E+14 2.39E+14 2.42E+14 2.18E+13 1.60E+13 7.58E+13 2.57E+13 2.86E+14 1.84E+14 1.85E+14 6.53E+12 6.31E+12 9.23E+12 6.95E+12 1.37E+14 8.72E+13 8.93E+13 2.89E+14 2.91E+14 2.94E+14 2.90E+14 4.09E+14 3.93E+14 3.94E+14

Bias 1.92E+07 9.02E+05 3.98E+07 4.40E+06 -5.57E+07 -4.95E+07 -4.74E+07 1.88E+07 -8.33E+05 5.26E+07 6.90E+06 -5.72E+07 -5.13E+07 -4.90E+07 1.28E+07 -4.76E+05 3.26E+07 -4.00E+05 -5.52E+07 -5.03E+07 -4.78E+07 -5.44E+08 -5.60E+08 -5.26E+08 -5.59E+08 -6.18E+08 -6.12E+08 -6.09E+08

TML2 Bias2

MSE

Prec.

4.13E+14 3.63E+13 1.64E+15 6.16E+13 3.49E+15 2.69E+15 2.49E+15 3.77E+14 1.67E+13 2.84E+15 7.33E+13 3.55E+15 2.81E+15 2.59E+15 1.70E+14 6.53E+12 1.07E+15 7.11E+12 3.18E+15 2.62E+15 2.38E+15 2.96E+17 3.13E+17 2.77E+17 3.13E+17 3.82E+17 3.74E+17 3.72E+17

4.56E+14 7.18E+13 1.69E+15 1.04E+14 3.87E+15 2.93E+15 2.73E+15 3.98E+14 3.27E+13 2.91E+15 9.90E+13 3.84E+15 3.00E+15 2.77E+15 1.76E+14 1.28E+13 1.08E+15 1.41E+13 3.32E+15 2.70E+15 2.47E+15 2.96E+17 3.14E+17 2.77E+17 3.13E+17 3.82E+17 3.75E+17 3.72E+17

4.36E+13 3.61E+13 6.20E+13 4.32E+13 1.02E+14 1.90E+15 1.44E+15 2.29E+13 1.65E+13 7.80E+13 2.64E+13 6.78E+13 1.20E+15 8.88E+14 7.16E+12 6.47E+12 1.30E+13 7.66E+12 2.49E+13 3.49E+14 2.69E+14 2.88E+14 2.91E+14 2.95E+14 2.90E+14 3.58E+14 3.20E+14 3.10E+14

Bias 1.95E+07 1.07E+06 4.01E+07 4.57E+06 -1.77E+07 2.31E+07 3.64E+07 1.91E+07 -6.63E+05 5.29E+07 7.08E+06 -1.90E+07 1.92E+07 3.40E+07 1.30E+07 -3.05E+05 3.28E+07 -2.19E+05 -1.89E+07 1.29E+07 2.91E+07 -5.44E+08 -5.60E+08 -5.25E+08 -5.59E+08 -5.80E+08 -5.40E+08 -5.26E+08

Bias2

MSE

4.22E+14 3.72E+13 1.67E+15 6.41E+13 4.17E+14 2.43E+15 2.76E+15 3.86E+14 1.70E+13 2.87E+15 7.66E+13 4.29E+14 1.56E+15 2.04E+15 1.76E+14 6.56E+12 1.09E+15 7.71E+12 3.84E+14 5.15E+14 1.12E+15 2.96E+17 3.13E+17 2.76E+17 3.12E+17 3.36E+17 2.92E+17 2.77E+17

4.66E+14 7.33E+13 1.73E+15 1.07E+14 5.19E+14 2.53E+15 4.20E+15 4.09E+14 3.35E+13 2.95E+15 1.03E+14 4.97E+14 1.63E+15 2.93E+15 1.83E+14 1.30E+13 1.10E+15 1.54E+13 4.09E+14 5.40E+14 1.39E+15 2.96E+17 3.14E+17 2.77E+17 3.13E+17 3.37E+17 2.92E+17 2.77E+17

TABLE A1 (contd.) Scenario: #5 Period

TML1 Years

1

season dependent

233

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

7.88E+12 3.02E+12 5.23E+13 2.17E+13 3.87E+13 1.33E+15 9.51E+14 4.17E+12 1.61E+12 5.69E+13 4.42E+12 3.31E+13 8.09E+14 5.80E+14 2.04E+12 5.52E+11 6.61E+12 1.63E+12 1.28E+13 2.02E+14 1.53E+14 2.97E+11 2.52E+11 6.49E+11 2.60E+11 8.53E+11 9.98E+12 8.00E+12

Bias 1.83E+07 8.31E+05 2.68E+07 3.49E+06 -1.69E+07 2.71E+07 4.08E+07 1.56E+07 1.52E+06 3.51E+07 3.32E+06 -1.71E+07 1.93E+07 3.45E+07 1.47E+07 1.90E+06 3.97E+07 9.61E+05 -1.79E+07 1.46E+07 3.10E+07 1.76E+07 2.05E+06 3.18E+07 1.33E+06 -1.59E+07 2.18E+07 3.56E+07

TML2 Bias2

MSE

Prec.

3.43E+14 3.71E+12 7.73E+14 3.39E+13 3.26E+14 2.06E+15 2.62E+15 2.48E+14 3.92E+12 1.29E+15 1.55E+13 3.27E+14 1.18E+15 1.77E+15 2.17E+14 4.16E+12 1.59E+15 2.55E+12 3.32E+14 4.14E+14 1.12E+15 3.10E+14 4.47E+12 1.01E+15 2.04E+12 2.55E+14 4.84E+14 1.27E+15

3.51E+14 6.73E+12 8.25E+14 5.56E+13 3.64E+14 3.39E+15 3.57E+15 2.52E+14 5.54E+12 1.34E+15 1.99E+13 3.60E+14 1.99E+15 2.35E+15 2.19E+14 4.71E+12 1.59E+15 4.18E+12 3.45E+14 6.17E+14 1.27E+15 3.10E+14 4.72E+12 1.01E+15 2.30E+12 2.56E+14 4.94E+14 1.28E+15

3.80E+12 2.37E+12 6.63E+12 4.07E+12 3.53E+13 1.30E+15 9.25E+14 1.72E+12 1.48E+12 2.88E+12 2.32E+12 3.14E+13 7.95E+14 5.71E+14 9.54E+11 9.03E+11 3.40E+12 8.86E+11 1.22E+13 2.05E+14 1.55E+14 2.12E+11 1.92E+11 5.01E+11 2.35E+11 5.52E+11 5.83E+12 4.70E+12

Bias 1.85E+07 1.09E+06 2.79E+07 4.74E+06 -1.68E+07 2.73E+07 4.12E+07 1.58E+07 1.49E+06 3.29E+07 4.33E+06 -1.71E+07 1.93E+07 3.46E+07 1.58E+07 1.87E+06 3.97E+07 1.41E+06 -1.78E+07 1.47E+07 3.12E+07 1.81E+07 2.55E+06 3.25E+07 3.07E+06 -1.52E+07 2.37E+07 3.74E+07

Bias2

MSE

3.47E+14 3.55E+12 7.85E+14 2.65E+13 3.18E+14 2.04E+15 2.62E+15 2.51E+14 3.71E+12 1.09E+15 2.11E+13 3.23E+14 1.17E+15 1.77E+15 2.52E+14 4.40E+12 1.58E+15 2.88E+12 3.29E+14 4.22E+14 1.13E+15 3.27E+14 6.67E+12 1.06E+15 9.66E+12 2.32E+14 5.66E+14 1.40E+15

3.51E+14 5.92E+12 7.92E+14 3.06E+13 3.53E+14 2.08E+15 3.54E+15 2.53E+14 5.19E+12 1.09E+15 2.34E+13 3.55E+14 1.20E+15 2.34E+15 2.53E+14 5.30E+12 1.58E+15 3.76E+12 3.41E+14 4.34E+14 1.28E+15 3.27E+14 6.87E+12 1.06E+15 9.89E+12 2.33E+14 5.67E+14 1.41E+15

TABLE A1 (contd.) Scenario: #5 Period

TML1 Years

234

parameter dependent

1

2

5

10

1σ & Bias

Prec.

20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3 20% / 4 5% / 0.4 60% / 8 20% / 0.4 ts v1 ts v2 ts v3

3.70E+12 2.63E+12 3.59E+13 4.18E+12 4.07E+13 1.35E+15 9.79E+14 3.19E+12 1.31E+12 9.31E+12 3.01E+12 3.34E+13 1.01E+15 7.31E+14 8.28E+11 5.07E+11 2.64E+12 8.08E+11 1.37E+13 2.44E+14 1.83E+14 1.31E+13 1.41E+13 1.43E+13 1.41E+13 1.49E+13 1.91E+13 1.68E+13

Bias 1.71E+07 1.31E+06 3.14E+07 -1.17E+06 -1.71E+07 2.47E+07 3.84E+07 1.71E+07 4.77E+05 3.71E+07 2.24E+06 -1.73E+07 2.28E+07 3.71E+07 1.40E+07 7.47E+05 3.26E+07 -8.12E+05 -1.82E+07 1.51E+07 3.13E+07 2.56E+07 1.08E+07 9.85E+06 1.06E+07 -6.57E+06 3.08E+07 4.47E+07

TML2 Bias2

MSE

Prec.

2.97E+14 4.34E+12 1.02E+15 5.55E+12 3.32E+14 1.96E+15 2.46E+15 2.97E+14 1.53E+12 1.39E+15 8.02E+12 3.33E+14 1.53E+15 2.11E+15 1.98E+14 1.07E+12 1.07E+15 1.47E+12 3.43E+14 4.72E+14 1.16E+15 6.66E+14 1.30E+14 1.11E+14 1.27E+14 5.80E+13 9.70E+14 2.01E+15

3.00E+14 6.97E+12 1.06E+15 9.73E+12 3.73E+14 3.31E+15 3.44E+15 3.00E+14 2.84E+12 1.40E+15 1.10E+13 3.66E+14 2.54E+15 2.84E+15 1.99E+14 1.57E+12 1.07E+15 2.27E+12 3.57E+14 7.16E+14 1.35E+15 6.79E+14 1.44E+14 1.26E+14 1.41E+14 7.29E+13 9.89E+14 2.03E+15

6.24E+12 5.60E+12 1.21E+13 8.67E+12 4.18E+13 1.25E+15 8.91E+14 4.88E+12 5.95E+12 9.49E+12 3.90E+12 4.00E+13 8.32E+14 5.86E+14 2.20E+12 1.90E+12 2.78E+12 2.20E+12 1.41E+13 2.62E+14 1.91E+14 1.40E+13 1.48E+13 1.48E+13 1.48E+13 1.59E+13 2.08E+13 1.85E+13

Bias 1.98E+07 3.37E+06 3.81E+07 1.83E+06 -1.56E+07 2.86E+07 4.22E+07 1.78E+07 2.94E+06 4.51E+07 2.35E+06 -1.61E+07 2.61E+07 4.03E+07 1.66E+07 3.20E+06 3.48E+07 1.72E+06 -1.64E+07 2.03E+07 3.62E+07 2.94E+07 1.42E+07 1.33E+07 1.38E+07 -3.30E+06 4.09E+07 5.38E+07

Bias2

MSE

4.00E+14 1.69E+13 1.47E+15 1.20E+13 2.85E+14 2.07E+15 2.68E+15 3.23E+14 1.46E+13 2.04E+15 9.42E+12 2.99E+14 1.52E+15 2.21E+15 2.78E+14 1.21E+13 1.21E+15 5.15E+12 2.82E+14 6.75E+14 1.50E+15 8.76E+14 2.16E+14 1.92E+14 2.06E+14 2.68E+13 1.70E+15 2.91E+15

4.06E+14 2.25E+13 1.48E+15 2.07E+13 3.27E+14 2.11E+15 3.57E+15 3.28E+14 2.06E+13 2.05E+15 1.33E+13 3.39E+14 1.56E+15 2.80E+15 2.80E+14 1.40E+13 1.22E+15 7.35E+12 2.96E+14 6.90E+14 1.69E+15 8.90E+14 2.30E+14 2.06E+14 2.21E+14 4.27E+13 1.71E+15 2.93E+15

APPENDIX 2 COMPLETE DATA COLLECTED BY LOCAL SAMPLING GROUPS IN ADOUREKOMAN

235

TABLE A2 COMPLETE WATER QUALITY RESULTS AS COLLECTED BY 236

THE THREE LOCAL SAMPLING TEAMS IN ADOUREKOMAN.

236

TABLE A2 (contd.)

237

Village: Adourekoman Date Calendar pH Date 11-Jul-05 7.5 16-Jul-05 7.5 24-Jul-05 7.5 31-Jul-05 7.5 8-Aug-05 7.5 11-Aug-05 7.5 25-Aug-05 7.0 27-Aug-05 7.5 4-Sep-05 7.5 12-Sep-05 7.5 18-Sep-05 7.0 25-Sep-05 7.5 8-Oct-05 7.5 15-Oct-05 7.5 22-Oct-05 7.0 28-Oct-05 7.5 5-Nov-05 7.5 11-Nov-05 7.5 18-Nov-05 7.0 27-Nov-05 7.5 4-Dec-05 7.5 18-Dec-05 7.5 25-Dec-05 7.0 31-Dec-05 7.5 6-Jan-06 7.5 14-Jan-06 7.5 22-Jan-06 7.5 30-Jan-06 7.5

Total Hardness (ppm) 180 120 120 - 180 120 180 120 - 180 120 120 - 180 120 180 120 - 180 120 120 180 180 120 - 180 120 180 120 - 180 120 120 180 180 120 80 80 180 120

Well: Ayewa-Okouta Test Strips Nitrate Metals Check (NO3-N) (ppb) 1000 50 1000 20 1000 20 1000 20 400 50 1000 20 1000 20 1000 20 100 20 100 20 100 20 100 20 100 20 200 20 100 20 400 20 100 20 400 20 100 20 100 20 100 20 200 20 100 20 100 20 1000 20 1000 20 100 20 1000 20

Nitrite (NO2-N) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Ammonia (NH3-N) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Phosphate (ppm) 0.5 0.5 5 5 5 0-5 0-5 0.5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

Colorimeter Nitrate (NO3-N) 7.1 4.2 0.8 1.5 1.2 2.1 4.2 1.5 3.2 4.1 6.0 5.3 2.9 2.2 5.6 3.1 5.2 3.2 6.1 4.2 5.3 2.1 5.2 2.9 6.3 3.2 4.1

TABLE A2 (contd.)

238

Village: Adourekoman Date Calendar pH Date 5-Feb-06 7.5 11-Feb-06 7.5 19-Feb-06 7.5 27-Feb-06 7.5 4-Mar-06 7.5 12-Mar-06 7.5 20-Mar-06 7.5 30-Mar-06 7.5 1-Apr-06 7.5 8-Apr-06 7.5 16-Apr-06 7.5 28-Apr-06 7.5 7-May-06 7.0 14-May-06 7.5 21-May-06 7.5 28-May-06 7.5 6-Aug-06 7.5 12-Aug-06 7.5 21-Aug-06 7.5 9-Aug-06 7.5 3-Sep-06 7.5 10-Sep-06 7.5 18-Sep-06 7.5 25-Sep-06 7.0 1-Oct-06 7.5 9-Oct-06 7.5 15-Oct-06 7.5 28-Oct-06 7.0

Total Hardness (ppm) 180 120 120 120 180 - 120 180 180 120 120 120 180 180 120 120 180 120 120 120 - 180 180 180 180 180 120 120 180 180 180 180

Well: Ayewa-Okouta Test Strips Nitrate Metals Check (NO3-N) (ppb) 1000 20 1000 20 100 20 100 20 100 20 100 50 100 20 100 20 1000 20 1000 20 1000 50 100 50 1000 20 1000 20 1000 20 1000 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20

Nitrite (NO2-N) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Ammonia (NH3-N) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Phosphate (ppm) 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

Colorimeter Nitrate (NO3-N) 2.1 3.2 4.3 5.1 4.3 6.1 5.2 4.3 6.1 2.1 3.2 5.2 3.1 2.0 5.1 4.2 3.2 5.1 7.2 4.5 1.4 4.2 3.5 2.3 6.1 2.5 4.3 5.4

TABLE A2 (contd.)

239

Village: Adourekoman Date Calendar pH Date 2-Nov-06 7.0 11-Nov-06 7.5 13-Nov-06 7.5 27-Nov-06 7.0 3-Dec-06 7.5 12-Dec-06 7.5 18-Dec-06 7.0 30-Dec-06 7.0 7-Jan-07 7.5 14-Jan-07 7.5 21-Jan-07 7.5 28-Jan-07 7.0 4-Feb-07 7.5 10-Feb-07 7.5 18-Feb-07 7.5 25-Feb-07 7.5 4-Mar-07 7.5 11-Mar-07 7.5 19-Mar-07 7.0 29-Mar-07 7.5

Total Hardness (ppm) 120 120 120 120 180 180 120 180 180 180 120 180 180 180 120 180 120 120 -180 120 120

Well: Ayewa-Okouta Test Strips Nitrate Metals Check (NO3-N) (ppb) 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20 100 20

Nitrite (NO2-N) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Ammonia (NH3-N) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Phosphate (ppm) 5 5 0 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

Colorimeter Nitrate (NO3-N) 5.3 6.1 2.3 4.1 6.2 5.1 4.3 1.2 5.2 2.1 3.4 5.3 2.1 3.1 4.2 5.2 3.2 5.3 4.1 2.3

TABLE A2 (contd.)

240

Village: Adourekoman Date Calendar pH Date 7-Jul-05 7.0 17-Jul-05 7.5 23-Jul-05 7.5 30-Jul-05 8.0 5-Nov-05 7.0 12-Nov-05 7.5 19-Nov-05 7.5 4-Dec-05 7.0 11-Dec-05 7.0 18-Dec-05 7.5 25-Dec-05 7.5 1-Jan-06 7.5 8-Jan-06 7.0 15-Jan-06 7.5 22-Jan-06 7.5 29-Jan-06 7.5 5-Feb-06 7.5 12-Feb-06 7.0 19-Feb-06 7.5 16-Feb-06 7.0 5-Mar-06 7.5 12-Mar-06 7.5 19-Mar-06 7.5 26-Mar-06 7.0 2-Apr-06 7.5 9-Apr-06 7.5 16-Apr-06 7.0 23-Apr-06 7.5

Well: Ayewa Total Hardness (ppm) 120 - 180 120 - 180 120 - 180 120 120 -180 180 120 - 180 120 120 180 120 0 0 10 10 100 120 - 180 120 120 120 - 180 120 120 - 180 120 - 180 120 120 - 180 120 - 180 120 120 - 180

Test Strips Nitrate Metals Check (NO3-N) (ppb) 20 -50 5 - 10 50 -100 5 10 10 10 10 50 5 100 50 50 10 100 0 10 0 10 2 10 2-5 100 2 100 2.5 50 0.2 50 0.2 50 0.2 50 5 100 5 100 50 50 5 50 5 50 2 50 5 50 5 50 5 50 5 50 5 50 5

Nitrite (NO2-N) 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Ammonia (NH3-N) 0 0.25 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0

Phosphate (ppm) 5 5 5 5 5 5 5 5 0-5 5 5 0-5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

Colorimeter Nitrate (NO3-N) 1.2 0.9 1.4 1.0 1.5 0.8 1.5 0.5 3.3 0.6 0.9 2.0 1.9 2.1 1.9 1.8 0.9 1.9 1.7 1.3 1.5 1.2 1.8 1.9 3.0 2.1 1.9 2.6

TABLE A2 (contd.)

241

Date Calendar Date 30-Apr-06 7-May-06 14-May-06 21-May-06 28-May-06 6-Aug-06 13-Aug-06 20-Aug-06 27-Aug-06 3-Sep-06 10-Sep-06 17-Sep-06 24-Sep-06 1-Oct-06 8-Oct-06 15-Oct-06 22-Oct-06 5-Nov-06 12-Nov-06 19-Nov-06 25-Nov-06

pH

Total Hardness (ppm)

7.5 7.5 7.5 7.5 7.0 7.0 7.0 7.5 7.0 7.0 7.0 7.5 7.0 7.5 7.0 7.5 7.5 7.0 7.0 7.5 7.0

120 - 180 120 - 180 120 - 180 120 - 180 120 80 - 120 80 - 120 120 120 0 - 40 0 - 40 120 - 180 0 - 40 120 180 0 - 40 0 - 40 0 - 40 0 0 - 40 0 - 40

Test Strips Nitrate Metals Check (NO3-N) (ppb) 50 5 50 5 50 5 50 5 50 5 50 5 50 5 100 5 50 5 50 5 50 5 50 10 50 5 50 10 50 10 100 5 100 5 50 5 50 5 100 5 50 5

Nitrite (NO2-N) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Ammonia (NH3-N) 0 0 0 0 0 0.25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Phosphate (ppm) 5 5 5 5 5 0-5 0-5 0-5 5 0-5 0-5 5 5 5 5 5 5 5 15 15 15

Colorimeter Nitrate (NO3-N) 2.9 3.9 1.9 2.6 3.0 4.5 4.3 3.1 2.1 1.9 3.2 3.0 3.0 4.1 5.3 4.8 1.2 0.8 1.2 0.6

TABLE A2 (contd.)

242

Village: Adourekoman Date Calendar pH Date 3-Jul-05 7.5 10-Jul-05 7.5 17-Jul-05 7.0 31-Jul-05 7.0 7-Aug-05 7.0 14-Aug-05 7.0 21-Aug-05 7.5 28-Aug-05 7.0 4-Sep-05 7.0 11-Sep-05 7.0 18-Sep-05 7.5 25-Sep-05 7.0 2-Oct-05 7.5 9-Oct-05 7.0 15-Oct-05 7.5 20-Oct-05 7.5 5-Nov-05 7.5 12-Nov-05 7.0 13-Nov-05 7.0 19-Nov-05 7.0 4-Dec-05 7.0 11-Dec-05 7.0 18-Dec-05 7.5 25-Dec-05 7.5 5-Feb-06 7.5 12-Feb-06 7.5 19-Feb-06 7.5 16-Feb-06 7.0

Well: Agbo Total Hardness (ppm) 180 - 250 180 - 250 120 - 180 180 - 250 180 - 250 120 - 180 120 - 180 180 - 250 180 - 250 120 - 180 120 - 180 120 - 180 120 - 180 120 - 180 180 - 250 120 - 180 120 - 180 180 - 250 180 - 250 120 - 180 120 - 180 180 - 250 180 - 250 120 - 180 180 - 250 120 - 180 120 - 180 180 - 250

Metals Check (ppb) 50 50 50 50 50 50 50 50 50 50 50 50 50 50 100 0.5 50 50 50 50 50 50 50 50 100 0.5 50 50

Test Strips Nitrate (NO3-N) 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50

Nitrite (NO2-N) 0 0 0 0 0 0 0 0 0.15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Ammonia (NH3-N) 0 0 0 0 0.25 0 0 0 0 0 0 0 0 0.25 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Phosphate (ppm) 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 5 - 15 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5

Colorimeter Nitrate (NO3-N) 3.1 6.0 4.4 14.1 1.5 6.2 4.1 8.2 3.9 2.1 2.4 3.1 5.9 3.0 2.7 5.7 4.1 8.2 1.5 4.4 4.4 14.4 3.1 6.0 2.7 5.7 4.1 8.2

TABLE A2 (contd.)

243

Village: Adourekoman Date Calendar pH Date 5-Mar-06 7.5 12-Mar-06 7.5 19-Mar-06 7.0 26-Mar-06 7.0 6-Aug-06 7.5 13-Aug-06 7.5 20-Aug-06 7.5 27-Aug-06 7.5 3-Sep-06 7.5 10-Sep-06 7.0 17-Sep-06 7.0 24-Sep-06 7.5 4-Oct-06 7.5 12-Oct-06 7.5 8-Nov-06 7.0 22-Nov-06 7.5 8-Dec-06 7.5 24-Dec-06 7.0 10-Jan-07 7.5 22-Jan-07 7.0 8-Mar-07 7.0 22-Mar-07 7.5 5-May-07 7.5 26-Apr-07 7.0

Well: Agbo Total Hardness (ppm) 180 - 250 180 - 250 120 - 180 180 - 250 120 -180 120 - 180 120 - 180 180 - 250 40 - 80 40 - 80 40 - 80 120 - 180 120 - 180 40 - 80 120 - 180 120 - 180 120 - 180 120 - 180 120 - 180 120 - 180 120 - 180 120 - 180 120 - 180 120 - 180

Metals Check (ppb) 50 50 50 50 50 50 50 50 100 100 100 50 100 100 50 50 50 50 50 50 50 50 50 50

Test Strips Nitrate (NO3-N) 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50

Nitrite (NO2-N) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Ammonia (NH3-N) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Phosphate (ppm) 0-5 0-5 0-5 0-5 5 5 5 5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5 0-5

Colorimeter Nitrate (NO3-N) 3.1 6.0 4.4 14.1 33.0 2.4 31.1 28.4 33.0 30.2 32.1 29.5 30.2 33.0 2.4 31.2 29.4 32.1 30.1 32.3 30.1 31.2 29.7

APPENDIX 3 COMPLETE RESULT FOR THE URANIUM STUDY

244

TABLE A3 COMPLETE ICP-MS AND ICP-OES RESULTS FOR THE URANIUM STUDY. THE PHASE OF EACH SAMPLE IS INDICATED IN THE TALBE: DISSOLVED (D) OR COLLOIDAL (C); THE COLLOIDAL PHASE IS FURTHER DIVIDED INTO UNFILTERED (U) OR FILTERED (F). ICP-MS RESULTS ARE IN PPB, AND ICPOES RESULTS ARE IN PPM. THE MASS OR WAVELENGTH USED FOR THE ANALYSIS IS FOUND IN THE ELEMENT LIST. “ND” INDICATES NON-DETECT ON THE INSTRUMENT OF ANALYSIS.

245

TABLE A3 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF

D

11-Jul-06 C F

C UF

0.0145

1.52

ND

0.0040

0.0012

1.31

0.0016

0.0032

51

0.0142

24.77

ND

0.0670

0.0207

19.31

0.0251

0.0110

53

0.0591

4.73

ND

0.0608

0.1014

7.82

0.0930

0.0724

59

0.0037

1.71

0.0003

0.0022

0.0004

1.43

0.0006

0.0011

60

0.0127

5.83

0.0097

0.0138

0.0086

10.49

0.0082

0.0195

63

0.1225

5.49

ND

0.1281

0.0420

3.15

0.1391

0.0697

Cr

Co Ni

Cu

66

0.4565

14.04

ND

0.3366

0.1509

3.60

0.3190

0.3291

95

0.0011

1.25

ND

0.0016

0.0002

0.98

0.0001

0.0009

111

0.0019

12.77

ND

0.0023

0.0001

8.32

0.0005

0.9127

Sn

118

0.0083

0.07

ND

0.0040

0.0015

0.02

0.0020

0.0018

Sb

121

0.0021

0.02

ND

0.0006

0.0003

0.00

0.0001

0.0002

Cs

133

0.0010

0.03

ND

0.0009

ND

0.08

ND

ND

185

0.0009

0.02

ND

0.0009

ND

0.04

ND

ND

206

0.0302

0.40

ND

0.0120

0.0053

0.27

0.0101

0.0099

209

0.0036

0.01

ND

0.0009

ND

0.09

0.0004

0.0000 0.0003

Zn

Mo Cd ppb

17-Jun-06 D C F

Ti47 V

Re

Pb Bi

Th U

232

238

Mn

55

88

Sr

279.077

Mg

0.08

ND

ND

0.09

0.0002

0.0157

30.80

ND

0.0126

ND

27.17

ND

ND

0.0171

567.91

ND

0.0200

0.0089

427.70

0.0092

0.0132

0.0709

14.03

ND

0.0733

0.0811

3.60

0.0724

0.0775 0.0065

0.0079

63.96

0.0056

0.0062

0.0065

66.28

0.0059

0.0080

0.08

0.0018

0.0022

0.0019

0.11

0.0019

0.0021

212.412

0.0085

30.42

0.0015

0.0022

0.0030

32.90

0.0009

0.0015

181.975

0.0196

19.60

0.0116

0.0131

0.0109

20.08

0.0109

0.0118

404.721

Si S

0.0072

0.0037

308.215

Al

ppm

13-Jun-06 C UF

Adourekoman: Agbo 7.9052 2.2734 3-Jul-06 4-Jul-06 C C F F

K

0.0093

0.20

0.0027

0.0165

0.0015

0.22

0.0015

0.0029

317.933

0.0167

123.55

0.0108

0.0119

0.0123

127.57

0.0115

0.0135

239.562

0.0081

0.01

0.0017

0.0168

0.0006

0.01

0.0005

0.0018

233.527

0.0195

20.16

0.0143

0.0143

0.0144

20.44

0.0142

0.0144

0.0015

77.34

0.0058

0.0017

0.0070

79.21

0.0071

0.0080

Ca Fe

Ba Na330.237

246

TABLE A3 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF

12-Jul-06 C F

C UF

1.33

0.0025

1.37

0.0071

1.26

0.0035

1.23

0.0030

0.0086

V51

27.42

0.5981

18.85

18.87

0.3385

18.17

1.1025

1.9198 0.0276

53

0.62

0.0406

8.42

1.3446 0.0262

3.96

0.0239

4.08

0.1055

59

3.39

0.0028

1.45

0.0027

1.97

0.0015

2.15

0.0044

0.0021

60

0.89

0.0087

11.22

0.0073

6.96

0.0048

6.93

0.0375

0.0056

Co Ni

63

1.89

0.1050

3.28

0.0590

1.21

0.0458

2.41

0.0823

0.0774

66

45.83

0.4445

3.42

0.1996

18.52

0.4670

1.0832

1.40

0.0006

0.91

1.3784 0.0008

19.34

95

1.36

0.0002

1.37

0.0008

0.0004

111

10.03

0.0018

7.64

0.0048

10.38

0.0012

7.76

0.0317

1.4716

0.05

0.0138

0.02

0.02

0.0082

0.03

0.0313

0.0353 0.0012

Cu Zn

Mo Cd ppb

D

Ti47 Cr

Sn118 Sb121

0.02

0.0007

0.01

2.3227 0.0008

0.00

0.0002

0.03

0.0012

Cs133

0.03

0.0008

0.08

0.0007

0.08

ND

0.08

ND

ND

Re185

0.01

0.0008

0.04

0.0010

0.03

ND

0.03

ND

ND

Pb206

0.29

0.0488

0.36

0.0738

0.29

0.3558

0.5631

0.01

0.0012

0.09

0.2555 0.0012

0.26

Bi209

0.08

ND

0.08

0.0005

0.0009

Th232

0.07

0.0057

0.09

0.0037

0.08

0.0007

0.08

0.0002

0.0003

19.02

0.0150

25.38

0.0163

20.00

0.0032

19.02

0.0083

0.0149

283.80

0.0145

403.55

0.0179

239.43

0.0113

241.95

0.0245

0.0161

45.83

0.1104

3.42

0.0918

19.34

0.0697

18.52

0.0988

0.0864

U

238

Mn

55

88

Sr

279.077

Mg

34.24

0.0050

67.24

0.0050

37.02

0.0041

36.67

0.0043

0.0043

Al

0.09

0.0023

0.11

0.0024

0.11

0.0021

0.11

0.0020

0.0021

Si212.412

29.52

0.0105

33.42

0.0183

29.17

0.0136

29.21

0.0212

0.0277

S181.975

9.06

0.0155

20.39

0.0127

9.65

0.0105

9.61

0.0108

0.0099

K

1.27

0.0910

0.18

0.1036

0.24

0.0728

0.16

0.1248

0.1734

Ca317.933

57.82

0.0118

128.49

0.0149

68.69

0.0117

66.72

0.0136

0.0165

Fe239.562

1.14

0.0957

0.00

0.1085

0.07

0.0759

0.01

0.1317

0.1902

Ba233.527 Na330.237

18.74

0.0290

20.73

0.0290

18.30

0.0241

18.49

0.0300

0.0352

74.32

0.0032

79.85

0.0046

78.93

0.0063

78.68

0.0079

0.0065

308.215

ppm

17-Jun-06 D C F

Adourekoman: Ayewa-Okouta 7.9190 2.2717 3-Jul-06 5-Jul-06 D C D C F F

404.721

247

TABLE A3 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF

D

18-Jul-06 C F

C UF

1.33

0.0023

0.0102

1.29

0.0042

0.0132

1.46

0.0045

0.0054

V51

62.27

0.0120

0.0282

47.32

0.0109

0.0133

65.95

0.0110

0.0117

53

0.85

0.0775

0.0809

5.30

0.0545

0.0541

2.27

0.0549

0.0720

59

0.52

0.0143

0.0027

0.64

0.0010

0.0015

0.67

0.0006

0.0016

60

ND

0.0092

0.0138

7.36

0.0078

0.0101

0.13

0.0089

0.0100

63

7.92

0.0864

0.1143

4.17

0.0805

0.1557

6.10

0.0724

0.0843

Co Ni

Cu

66

18.67

0.2233

0.2667

6.12

0.0199

0.1217

34.03

0.1709

0.1776

95

2.08

0.0010

0.0014

1.60

0.0014

0.0013

2.09

0.0002

0.0034

111

10.43

0.0010

0.0035

7.99

0.0008

2.8453

9.55

1.1002

0.1554

0.06

0.0038

0.0092

0.06

0.0046

0.0055

0.06

0.0025

0.0033 0.0008

Zn

Mo Cd ppb

3-Jul-06 C F

Ti47 Cr

Sn118 Sb121

0.02

0.0005

0.0012

0.03

0.0014

0.0014

0.02

0.0006

Cs133

0.03

0.0007

0.0010

0.08

0.0001

0.0001

0.03

ND

ND

Re185

0.02

0.0007

0.0009

0.04

0.0002

ND

0.02

ND

ND

Pb206

0.36

0.0077

0.0165

0.68

0.0105

0.0168

0.38

0.0121

0.0164

Bi209

0.01

0.0010

0.0009

0.12

0.0029

0.0046

0.01

0.0003

0.0006

232

0.07

0.0051

0.0039

0.08

0.0044

0.0046

0.07

ND

ND

30.59

0.0103

0.0131

25.86

0.0023

0.0025

31.81

ND

ND

319.87

0.0082

0.0168

291.70

0.0088

0.0145

419.58

0.0095

0.0173

18.67

0.0476

0.0636

6.12

0.0147

0.0165

15.24

0.0430

0.0434

Mg

29.89

0.0035

0.0041

30.75

0.0032

0.0034

30.99

0.0031

0.0070

Al308.215

0.09

0.0024

0.0045

0.12

0.0027

0.0032

0.12

0.0021

0.0062

Si212.412

29.21

0.0027

0.0046

31.15

0.0030

0.0037

31.27

0.0019

0.0068

S181.975

8.05

0.0119

0.0128

7.64

0.0120

0.0113

8.01

0.0102

0.0256

Th

U238 Mn

55

88

Sr

279.077

ppm

17-Jun-06 D C F

Adourekoman: Ayewa 7.9155 2.2751 11-Jul-06 D C C F UF

404.721

K

0.23

0.0013

0.0053

0.17

0.0014

0.0021

0.12

0.0022

0.0103

Ca317.933

72.46

0.0080

0.0105

73.10

0.0075

0.0090

73.82

0.0074

0.0191

Fe239.562

0.04

0.0008

0.0049

0.01

0.0011

0.0017

0.01

0.0009

0.0084

Ba233.527 Na330.237

9.97

0.0136

0.0145

10.26

0.0139

0.0139

10.32

0.0142

0.0348

56.86

0.0013

0.0016

57.03

0.0031

0.0020

56.44

0.0052

0.0106

248

TABLE A3 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF

14-Jun-06 C F

C UF

D

13-Jul-06 C F

C UF

0.66

0.0523

0.3196

0.75

0.0125

0.0230

1.07

0.0049

V51

15.72

0.7203

2.7592

9.53

1.1439

7.69

0.2080

0.6330

4.96

4.3190 0.0368

8.30

53

3.5793 0.0378

11.97

0.1960

59

1.20

0.0093

0.0173

1.21

0.0082

0.0087

0.84

0.0050

60

27.93

0.1453

0.2393

26.03

0.1725

14.27

0.0327

1.91

0.1917

Cr Ni

63

0.95

0.5962

1.1368

0.93

0.1906

0.6017 0.2553

66

31.74

1.8219

4.8423

17.57

0.0230

0.0312

25.67

4.9209 0.0057

17.8481

23.87

4.2661 0.0054

637.19

95

4.68

0.0034

111

7.67

0.0016

0.0046

8.18

0.0293

8.12

ND

Sn118

0.02

0.0363

0.0669

0.01

0.0083

1.6937 0.0104

0.02

0.0187

Sb121

0.17

0.0023

0.0028

0.18

0.0006

0.0007

0.17

0.0015

133

0.10

0.0011

0.0032

0.10

0.0010

0.0011

0.09

ND

Re185

0.39

0.0004

0.0005

0.40

0.0010

0.0010

0.05

ND

Pb206

0.16

0.1348

0.1616

0.19

0.4097

0.08

0.0033

0.0024

0.08

0.2465 0.0011

1.12

Bi209

0.1685 0.0010

0.09

ND

0.0049

0.09

ND

27.58

0.0404

Cu Zn

Mo Cd ppb

D

Unknown 7.8979 2.3208 16-Jun-06 D C F

Ti47

Co

Cs

232

0.08

0.0742

0.2695

0.08

0.0041

261.17

0.2312

1.2924

239.75

0.1586

0.2017

84.96

0.0384

0.0748

92.12

0.0407

0.0455

84.65

0.0399

31.74

0.0823

0.1067

17.57

0.3346

0.3310

637.19

0.1737

Mg

28.00

0.0107

0.0156

28.11

0.0054

0.0058

44.06

0.0142

Al308.215

0.07

0.0333

0.1015

0.11

0.0080

0.0119

0.07

0.0063

Si212.412

18.07

0.0519

0.1731

17.87

0.0733

0.0886

23.86

0.0481

S181.975

45.10

0.0461

0.0469

43.98

0.0157

0.0153

26.89

0.0363

Th

U238 Mn

55

88

Sr

279.077

ppm

Kpakpazoume: A 7.9278 2.2492

404.721

K

0.48

0.0884

0.3411

0.17

0.4497

0.6033

0.91

0.4388

Ca317.933

77.08

0.0316

0.0424

76.06

0.0364

0.0459

135.07

0.0549

Fe239.562

0.26

0.0918

0.3571

0.01

0.5085

0.6792

0.72

0.4711

Ba233.527 Na330.237

5.19

0.0445

0.0530

5.17

0.0338

0.0402

4.75

0.0475

56.25

0.0036

0.0054

56.35

0.0041

0.0027

72.07

0.0282

249

TABLE A3 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF

Sowe: 2 7.98 2.17 24-Jun-06 D C F

1.30

0.0029

1.07

0.0065

0.0164

1.10

0.0066

1.07

0.0129

V51

33.66

0.0732

21.39

0.0114

0.0150

6.70

0.2509

11.66

0.0131

3.53

0.0589

0.0613

ND

0.0834

5.20

0.1421

53

7.76

0.3156

59

1.11

0.0009

0.66

0.0022

0.0031

0.51

0.0025

0.86

0.0020

60

24.74

0.0313

12.67

0.0143

0.0205

ND

0.0101

6.94

0.0128

63

1.78

0.0517

1.07

0.2778

0.3201

1.86

0.3285

2.02

0.1430

16.03

0.1926

Co Ni

Cu

66

10.47

0.4241

5.44

95

12.32

0.0020

111

12.88

0.0001

0.03

0.0023

Zn

Mo Cd ppb

C UF

Sowe: 1 7.9780 2.1659 25-Jun-06 D C F

Ti47 Cr

Sn118 Sb121

0.07

0.4750 0.0022

600.21

7.82

0.1893 0.0021

4.3943

0.89

0.0016

0.84

0.0024

7.84

0.0067

9.18

0.0036

9.18

0.0020

0.02

0.0044

0.8609 0.0047

0.05

0.0261

0.07

0.0043

0.04

0.0006

0.0009

0.04

0.0009

0.02

0.0009

ND

0.12

0.0010

0.0009

0.03

0.0010

0.03

0.0010

0.0004

133

0.17

185

0.18

ND

0.11

0.0010

0.0009

0.01

0.0010

0.02

0.0010

Pb206

0.33

0.0077

0.22

0.0182

0.72

0.3746

0.29

0.0107

Bi209

0.11

ND

0.12

0.0156 0.0014

0.0022

0.01

0.0009

0.01

0.0008

0.07

0.0037

0.0037

0.06

0.0038

0.07

0.0038

Cs

Re

Th

232

238

0.11

0.0005

126.70

0.0113

83.27

0.0172

0.0170

2.31

0.0113

4.69

0.0111

186.43

0.0102

130.45

0.0101

0.0189

204.82

0.0160

230.28

0.0089

10.47

0.1534

5.44

0.0544

0.0542

592.46

0.1002

16.03

0.0937

Mg

22.28

0.0072

22.59

0.0024

0.0026

28.03

0.0031

45.13

0.0044

Al308.215

0.08

0.0051

0.11

0.0032

0.0044

0.09

0.0025

0.13

0.0037

Si212.412

25.68

0.0035

26.94

0.0036

0.0049

22.40

0.0120

20.51

0.0039

S181.975

18.25

0.0339

16.82

0.0115

0.0126

4.17

0.0124

10.29

0.0119

U

Mn

55

88

Sr

279.077

ppm

Kpakpazoume: B 7.9278 2.2492 14-Jun-06 14-Jul-06 D C D C UF F

404.721

K

0.20

0.0041

0.16

0.0018

0.0036

0.34

0.0782

0.20

0.0029

Ca317.933

62.73

0.0188

62.28

0.0062

0.0074

80.77

0.0113

147.98

0.0138

Fe239.562

0.01

0.0001

0.00

0.0013

0.0033

0.16

0.0823

0.01

0.0025

Ba233.527 Na330.237

6.55

0.0418

6.62

0.0135

0.0136

17.01

0.0200

2.17

0.0136

43.12

0.0110

43.05

ND

ND

32.13

0.0003

50.43

0.0005

250

TABLE A3 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF

Agouagon: C 7.9792 2.2960 29-Jun-06 D C F

Mahou: 1 7.8257 2.1129 2-Jul-06 D C F

1.52

0.0039

1.64

0.0157

1.49

0.0044

0.87

0.0045

V51

8.42

0.0051

4.86

0.0601

17.19

0.0076

1.76

0.0030

53

2.52

0.0505

2.27

0.0627

1.51

0.0869

3.03

0.0736

59

0.22

0.0013

0.14

0.0016

0.43

0.0021

0.50

0.0012

60

ND

0.0075

2.95

0.0095

2.27

0.0089

7.82

0.0182

63

7.68

0.0393

1.31

0.2257

2.70

0.0325

0.71

0.0898

66

2.75

0.1580

79.58

0.5900

14.66

0.2140

1.45

0.0013

6.98

0.0022

2.25

0.1766 0.0016

8.92

95

7.57

0.0014

111

9.17

0.0029

8.31

0.0023

10.79

0.0023

8.69

0.0014

Sn118

0.08

0.0019

0.03

0.0149

0.06

0.0022

0.02

0.0034

Sb121

0.02

0.0012

0.00

0.0005

0.01

0.0020

0.01

0.0004

Cs133

0.02

0.0009

0.07

0.0011

0.02

0.0010

0.08

0.0006

Re185

0.02

0.0010

0.03

0.0009

0.01

0.0011

0.07

0.0007

Pb206

0.34

0.0040

0.30

0.0966

0.34

0.27

0.0095

Bi209

0.01

0.0008

0.08

0.0009

0.01

0.0152 0.0009

0.09

0.0009

Th232

0.07

0.0038

0.07

0.0048

0.06

0.0042

0.08

0.0046

2.25

0.0114

0.28

0.0105

1.95

0.0124

1.94

0.0076

0.46

0.0032

37.92

0.0071

125.28

0.0084

417.18

0.0126

20.73

0.0335

79.58

0.0265

14.66

0.0379

8.92

0.0714

Mg

14.64

0.0016

8.06

0.0016

19.24

0.0021

23.48

0.0028

Al308.215

0.09

0.0021

0.10

0.0124

0.10

0.0023

0.10

0.0023

Si212.412

37.28

0.0018

43.36

0.0229

37.49

0.0035

16.90

0.0023

S181.975

13.26

0.0126

1.90

0.0129

7.75

0.0111

9.09

0.0110

Cr Ni

Cu Zn

Mo Cd ppb

Agouagon: B 7.9793 2.2987 1-Jul-06 D C F

Ti47

Co

U

238

Mn

55

88

Sr

279.077

ppm

Agouagon: A 7.98 2.29 25-Jun-06 D C F

404.721

K

0.18

0.0011

0.35

0.0519

0.18

0.0009

0.20

0.0017

Ca317.933

42.54

0.0039

20.90

0.0038

38.54

0.0039

89.47

0.0110

Fe239.562

0.00

0.0004

0.19

0.0544

0.00

0.0004

0.01

0.0008

Ba233.527 Na330.237

3.00

0.0133

2.26

0.0143

5.06

0.0134

4.78

0.0134

35.06

ND

19.02

ND

33.05

ND

45.66

0.0010

251

TABLE A3 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF

1.25

0.0106

1.81

0.0049

1.13

0.0178

V51

31.50

58.96

1.4678

1.97

0.0096

11.79

0.0808

2.15

0.1307

53

5.90

3.3854 0.0830

59

2.35

0.0054

4.34

0.0046

0.31

0.0020

60

6.84

0.0161

8.29

0.0169

4.52

0.0285

63

1.41

0.2475

3.79

0.4798

1.32

0.1672

66

20.30

51.75

0.5015

12.82

0.2264

Co Ni

Cu Zn

95

3.53

1.2640 0.0027

6.68

0.0042

4.79

0.0022

111

11.39

0.0041

19.14

0.0043

8.91

0.0092

0.02

0.0263

0.15

0.0527

0.02

0.0025

Mo Cd ppb

Moumoudji:B 7.8004 2.1814 7-Jul-06 D C F

Ti47 Cr

Sn118 Sb121

0.01

0.0013

0.17

0.0024

0.01

0.0009

133

0.08

0.0008

0.06

0.0007

0.08

0.0009

185

0.07

0.0009

0.11

0.0007

0.03

0.0009

Pb206

0.24

1.61

0.2616

0.48

0.0893

Bi209

0.09

0.1756 0.0008

0.04

0.0015

0.08

0.0007

0.08

0.0034

0.13

0.0051

0.08

0.0041

Cs

Re

Th

232

238

4.83

0.0148

10.78

0.0204

2.67

0.0101

663.97

0.0700

1206.14

0.0580

47.20

0.0186

20.30

0.0805

51.75

0.0585

12.82

0.0357

Mg

36.33

0.0050

35.98

0.0033

9.45

0.0014

Al308.215

0.11

0.0030

0.15

0.0022

0.11

0.0042

Si212.412

30.16

0.0311

29.67

0.0148

28.61

0.0041

S181.975

15.02

0.0133

15.00

0.0101

4.19

0.0118

U

Mn

55

88

Sr

279.077

ppm

Moumoudji:A 7.7960 2.1756 7-Jul-06 16-Jul-06 D C D C F F

404.721

K

0.42

0.2000

0.22

0.1398

0.18

0.0035

Ca317.933

55.81

0.0150

54.67

0.0097

53.63

0.0056

Fe239.562

0.27

0.2203

0.11

0.1545

0.00

0.0031

Ba233.527 Na330.237

14.76

0.0345

14.37

0.0228

2.81

0.0136

45.57

ND

45.27

0.0008

44.90

ND

252

TABLE A3 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF

C F

Ti47

1.44

0.0032

1.63

0.0050

0.0091

1.31

0.0027

V51

36.05

0.0068

34.63

0.0109

0.0142

17.50

0.4995

0.0826

53

10.41

0.0572

9.91

0.0836

6.04

0.2244

59

1.87

0.0015

1.98

0.0024

0.0038

1.23

0.0023

60

10.43

0.0086

10.10

0.0102

0.0124

12.01

0.0325

63

2.96

0.0794

3.35

0.0623

0.1269

1.62

11.0228

66

3.29

0.1782

4.07

58.56

5.5398

2.78

0.0007

3.35

0.2490 0.0015

0.2051

95

0.0037

1.16

0.0010

111

7.02

0.0011

9.27

0.0041

0.0075

7.82

0.0005

Sn118

0.02

0.0032

0.02

0.0019

0.0036

0.02

0.0283

0.0006

Cr Ni

Cu Zn

Mo Cd ppb

C UF

Unknown 7.8986 2.2579 14-Jun-06 D C F

D

Co

Sb121

0.01

0.0004

0.01

0.0007

0.04

0.0019

Cs133

0.11

0.0007

0.13

0.0009

0.0009

0.11

ND

Re185

0.06

0.0007

0.07

0.0008

0.0009

0.05

ND

Pb206

0.22

0.0075

0.25

0.0047

0.0097

0.28

0.1407

Bi209

0.08

0.0009

0.14

0.0016

0.0036

0.09

0.0021

0.0035

232

0.08

0.0055

0.09

0.0032

0.09

0.0006

6.54

0.0080

8.01

0.0100

0.0107

20.83

0.0006

699.81

0.0135

754.89

0.0225

0.0376

204.79

0.0249

3.29

0.0392

4.07

0.0444

0.0442

58.56

0.3215

Mg

57.53

0.0052

57.39

0.0054

0.0057

36.89

0.0107

Al308.215

0.12

0.0024

0.10

0.0025

0.0032

0.07

0.0054

Si212.412

33.29

0.0030

32.90

0.0031

0.0043

29.33

0.0170

S181.975

21.29

0.0118

21.44

0.0127

0.0135

11.06

0.0333

Th

U238 Mn

55

88

Sr

279.077

ppm

2-Jul-06

Moumoudji: C 7.7999 2.1762 15-Jul-06 D C F

404.721

K

0.18

0.0010

0.17

0.0011

0.0019

0.36

0.0672

Ca317.933

88.01

0.0078

87.59

0.0082

0.0090

82.64

0.0261

Fe239.562

0.00

0.0004

0.01

0.0009

0.0016

0.17

0.0675

Ba233.527 Na330.237

18.38

0.0139

18.13

0.0138

0.0145

29.89

0.0558

65.12

0.0021

65.75

0.0015

0.0011

70.14

0.0182

253

TABLE A3 (contd.) Well Name Latitude Longitude Date Phase: D or C C: F or UF

D

C UF

Unknown 7.8335 2.2776 15-Jun-06 D C F

1.61

0.0088

0.0063

1.51

0.0079

0.0251

1.52

0.0046

V51

10.17

0.1486

0.1695

3.87

0.0651

0.0121

19.71

2.7683

53

2.40

0.2068

0.4974

1.50

0.3320

0.1926

8.70

0.4460

59

0.31

0.0043

0.0079

0.26

0.0029

0.0055

0.90

0.0070

60

5.98

0.0337

0.1567

6.27

0.0356

0.0415

8.52

0.0697

63

1.68

1.6567

0.6603

2.20

0.2848

0.3184

0.76

0.6566

66

634.73

3.6795

3.6539

20.89

0.4889

0.8076

1111.91

22.4726

95

0.21

0.0025

0.0061

3.24

0.0020

0.0025

0.59

0.0113

111

8.29

0.0026

0.0069

7.52

ND

0.0028

8.56

0.0032

Sn118

0.02

0.0462

0.0517

0.01

0.0033

0.0049

0.02

0.0648

Sb121

0.07

0.0038

0.0044

0.00

0.0004

0.0007

0.05

0.0081

Cs133

0.07

0.0019

0.0019

0.08

ND

ND

0.08

ND

Re185

0.03

0.0022

0.0020

0.03

ND

ND

0.04

ND

Pb206

0.32

0.2271

0.1950

0.26

0.0153

0.0222

1.13

1.0112

Bi209

0.08

0.0036

0.0029

0.09

ND

0.0006

0.09

ND

232

0.08

0.0165

0.0143

0.08

0.0006

0.0008

0.09

ND

0.32

0.0260

0.0237

2.01

ND

ND

0.51

ND

Co Ni

Cu Zn

Mo Cd ppb

C UF

Unknown 7.7824 2.2622 15-Jun-06 C F

Ti47 Cr

Th

U238 Mn

55

88

Sr

279.077

Mg

10.52

0.0062

0.0437

3.72

0.0646

0.1338

220.16

0.0943

634.73

0.0531

0.0536

20.89

0.0762

0.0814

1111.90

0.2932 0.0119

13.47

0.0048

0.0049

19.63

0.0067

0.0069

33.67

Al

0.07

0.0071

0.0063

0.07

0.0058

0.0074

0.07

0.0056

Si212.412

41.29

0.0121

0.0146

35.70

0.0097

0.0096

35.01

0.0988

S181.975

1.19

0.0385

0.0342

0.94

0.0296

0.0273

15.41

0.0350

308.215

ppm

D

Unknown 7.7590 2.3242 15-Jun-06 C F

404.721

K

0.31

0.0418

0.0827

0.19

0.0051

0.0067

0.78

0.4961

Ca317.933

39.09

0.0123

0.0127

50.80

0.0141

0.0138

65.93

0.0284

Fe239.562

0.15

0.0426

0.0866

0.00

0.0012

0.0026

0.57

0.5385

Ba233.527 Na330.237

3.27

0.0408

0.0407

5.59

0.0422

0.0414

8.99

0.0560

10.20

ND

ND

35.64

0.0141

0.0123

43.32

0.0267

254

REFERENCES

Affaton, P. 1975. Etude géologique et structurale du Nord-Ouest du Dahomey, du Nord Togo et du Sud-Est de la Haute Volta. Trav. Lab. Sci. Terre St-Jérôme, Marseille, B, 10, 201 p., 96 fig. Aldenderfer, M.S. and R.K. Blashfield. 1984. Cluster Analysis. Newbury Park: Sage Publications, 88 pp. Aravena, R. , M. L. Evans, and J. A. Cherry. 1993. “Stable Isotopes of Oxygen and Nitrogen in Source Identification of Nitrate from Septic Systems.” Ground Water 31 (2): 180-186. Azonsi F. and A. Adjomayi. 2005. Rapport général sur le réseau piézométrique en République du Béni . Rap. SEH-DRE-DGH. Cotonou. Becker, Matthew W., David W. Metge, Samantha A. Collins, Allen M. Shapiro, and Ronald W. Harvey. 2003. “Bacterial Transport Experiments in fractured Crystalline Bedrock.” Ground Water 41 (5): 682-689. Boukari, M. 1980. Contribution à l’étude hydrogéologique des régions de socle de l’Afrique occidentale : mise au point des connaissances relatives à la République Populaire du Bénin. Rap. Dépt. Géol. No 4. Nouv. Série, Mémoire de DEA, Univ. de Dakar, Sénégal. 141 pp., 13 tableaux ann., 1 carte hors texte. Boukari, M. 1982. Contribution à l’étude hydrogéologique des régions de socle de l’Afrique Intertropicale : l’hydrogéologie de la région de Dassa-Zoumé (Bénin). Thèse de 3ème cycle, Univ. de Dakar, 142 pp. & Annexes. Boukari, M.1989. Etude de synthèse et de cartographie relative au cadre physique et aux ressources naturelles de la région nord du Bénin : rapport de synthèse et notices explicatives des cartes. SERHAU, Cotonou, Bénin. 68 pp., 6 Ann. 5 cartes hors texte. Boussari, W.T. 1975. Contribution à l’étude géologique du socle cristallin de la zone mobile Pan-africaine (région central du Dahomey). Thèse de 3ème cycle, Univ. Besançon, 236, 105 pp.

255

Box, George E. P. and Gwilym M. Jenkins. 1976. Time Series Analysis. San Francisco: Holden-Day, 575 pp. BRGM. 1978. Carte géologique au 1/200 000 du Bénin et du Togo entre les 9ème et 10ème degré de latitude Nord. Obemines. Cotonou, Bénin. 78 RDM 055 AF conseil Entente. Obemines, Cotonou. 2 feuilles, 1 notice explicative. British Geological Survey (BGS). 2000. “Executive Summary of the Main Report of Phase I, Groundwater Studies of As Contamination in Bangladesh.” Electronic resource. (http://www.bgs.ac.uk/arsenic/Bangladesh/Reports/ExecSummary.pdf). Buss, Daniel F. and Keith E. Bandt. 1981. “Field Report: An All-Teflon Bailer and an Air-Driven Pump for Evacuating Small-Diameter Ground-Water Wells.” Ground Water 19 (4): 429-431. Cameron, Kirk and Philip Hunger. 2002. “Using spatial models and kriging techniques to optimize long-term ground-water monitoring networks: a case study.” Environmetrics 13: 629-656. Canter, Larry W. 1997. Nitrates in Groundwater. New York: Lewis Publisher, 263 pp. Carte Géohydraulique. 1985. Carte hydrogéologique du Bénin à l’échelle de 1/500 000. DH. Cotonou, Bénin. 1 carte, 1 notice explicative. Chowdhury, T.R., G.K. Basu, B.K. Mandal, G. Samanta, U.K. Chowdhury, C.R. Chanda, D. Lodh, S Lal Roy, K.C. Saha, and S. Roy. 1999. “As Poisoning in the Ganges delta.” Nature 401: 545-546. Clark, Ian and Peter Fritz. 1997. Environmental Isotopes in Hydrology. New York: Lewis Publishers, 328 pp. Conwell, P.M., S.E. Silliman, and L. Zheng. 1997. “Design of a piezometer network for estimation of the variogram of the hydraulic gradient: The role of the instrument”, Water Resources Research 33(11): 2489-2494. Crane, Pamela. 2006. Implementation of a Sustainable Groundwater Quality Monitoring Program in Rural Benin, West Africa. Master’s Thesis. University of Notre Dame. Duarte, E.A., I. Neto, M. Alegrias, and R. Barroso. 1998. “‘Appropriate Technology’ for pollution control in corrugated board industry — The Portuguese case.” Water Science and Technology 38 (6): 45-53. Dudley, J.R. Research methods in social work: becoming consumers and producers of research. Allyn & Backon: web publication, http://www.safarix.com/0205365299.

256

Dukes, Michael D. and Robert O. Evans. 2006. “Impact of Agriculture on Water Quality in the North Carolina Middle Coastal Plain.” Journal of Irrigation and Drainage Engineering 132 (3): 250-262. Eaton, A.D., L.S. Clesceri and A.E. Greenberg, eds. 1995. Standard Methods for the Examination of Water and Wastewater, 19th ed. Washington, D.C.: American Public Health Association, 1100 pp. EPA. 2001. “Total Coliform Rule: A Quick Reference Guide.” EPA 816-F-01-035. web publication, www.epa.gov/safewater . EPA. 2003. National Drinking Water Standards. Web publication: http://www.epa.gov/safewater/contaminants/index.html . Engalenc. 1978. Les modalités de la recherche d’eau dan les roches cristallines fracturées de l’Afrique de l’Ouest. Bull. CIEH, nº 33-34, Ouagadougou, p. 22-30. Ewing, Rodney C. 1999. “Radioactivity and the 20th Century.” Uranium: Mineralogy, Geochemistry and the Environment. Reviews in Mineralogy 38: 1-21. Farnham, I.M., K.H. Johannesson, A.K. Singh, V.F. Hodge and K.J. Stetzenbach. 2003. “Factor analytical approaches for evaluating groundwater trace element chemistry data.” Analytica Chimica Acta 490(1-2): 123-138. Farnham, I.M., A.K. Singh, K.J. Stetzenbach, K.H. Johannesson. 2002. “Treatment of nondetects in multivariate analysis of groundwater geochemistry data.” Chemometrics and Intelligent Systems 60: 265-281. Farnham, I.M., K.J. Stetzenbach, A.K. Singh, and K.H. Johannesson. 2000. “Deciphering groundwater flow systems in Oasis Valley, Nevada, using tracer element chemistry, multivariate statistics, and geographical information systems.” Mathematical Geology 32 (8): 943-968. Galbis-Reig, Clara. 2002. Analytical and Statistical Analysis of Groundwater for the Determination of Atmospheric, Geological, and Anthropogenic Signatures: Benin, Africa. Master’s Thesis. University of Notre Dame. Gallagher, Mark and John Doherty. 2007. “Parameter estimation and uncertainty analysis for a watershed model.” Environmental Modelling & Software 22: 1000-1020. Gauch, H.G., Jr. 2003. Scientific Method in Practice. Cambridge, UK: Cambridge University Press, 435 pp. Getzen, Rufus T. 1981. “Field Report: Digital Recorder for Aquifer Tests.” Ground Water 19 (1): 88-92.

257

Girard, Pierre and Claude Hillaire-Marcel. 1997. “Determining the source of nitrate pollution in the Niger discontinuous acquifers using the natural 15N/14N ratios.” Journal of Hydrology 199: 239-251. Groen, J., J. B. Schuchmann and W. Geirnaert. 1988. “The occurrence of high nitrate concentration in groundwater in villages in Northwestern Burkina Faso.” Journal of African Earth Science 7 (7/8): 999-1009. Guiraud, R. 1975. Eléments pour une orientation nouvelle de la recherche des eaux souterraines dans les régions à substratum métamorphique ou éruptif de l’Afrique Occidentale. Réun. Porto-Alegre, Mém. AIH, vol XI p. 15-19. Haraldsen, T.K. and P. Stalnacke. 2006. “Methods for water quality sampling and load estimation in monitoring of Norwegian agricultural catchments.” Nordic Hydrology 37 (1): 81-92. Haldar, Achintya and Sankaran Mahadevan. 2000. Probability, Reliability and Statistical Methods in Engineering Design. New York: John Wiley & Sons, Inc., 304 pp. Honeyman, Bruce D. 1999. “Colloidal culprits in contamination.” Nature 397: 23-24. Hossain, M.F. 2006. “Arsenic contamination in Bangladesh — An overview.” Agriculture, Ecosystems and Environment 113: 1-16. Huby, Meg and Stuart Stevenson. 2003. “Meeting need and achieving sustainability in water project interventions.” Progress in Development Studies 3 (3): 196-209. Istituto Ricerche Breda. 1982. Etude de cartographie et de prospection minière de reconnaissance au Nord du 11ème parallèle. Projet FED No 4105-011-13-20. Projet FED No 4105-011-13-20., Obemines. Cotonou, Bénin. 1 carte impr.,1 rap. de synt. (55 pp. et 20 annexes), 6 documents annexes. Istituto Ricerche Breda. 1987. Etude de cartographie et de prospection minière de reconnaissance au Sud du 9ème parallèle. Projet FED No 5100-11-13-015., Obemines. Cotonou, Bénin. 3 cartes impr., 1 rap. de synt. (80 p.) + annexes. Isaaks, E.H. and R.M. Srivastava. 1990. An Introduction to Applied Geostatistics. New York: Oxford University Press, 592 pp. Jolliffe, I.T. 2002. Principal Component Analysis. New York: Springer, 487 pp. Journel, A.G. and Ch.J. Huijbregts. 1978. Mining Geostatistics. New York: Academic Press, 600 pp. Katz, Travis and Jennifer Sara. 1998 “Making Rural Water Supply Sustainable: Recommendations from a Global Study.” UNDP-World Bank Water and

258

Sanitation Program in web publication, http://www.wsp.org/publications/global_ruralstudy.pdf . Kaufman, L. 1990. Finding Groups in Data: An Introduction to Cluster Analysis. New York: John Wiley, 342 pp. Kelly, T. E. 1967. “Artificial Recharge at Valley City, North Dakota, 1932 to 1965.” Ground Water 5 (2): 20-25. Kersting, A. B., D. W. Efurd, D. L. Finnegan, D.J. Rokop, D. K. Smith and J. L. Thompson. 1999. “Migration of plutonium in groundwater at the Nevada Test Site.” Nature 397: 56-59. Khare, Deepak, M.K. Jat and Ediwahyunan. 2006. “Assessment of counjunctive use planning options: A case study of Sapon irrigation command area of Indonesia.” Journal of Hydrology 328: 767-777. Kim, Gyoo-Bum, Kang-kun Lee, Jin-Yong Lee and Myeong-Jae Yi. 2007. “Case study for determination of a water level monitoring frequency for nationwide groundwater monitoring networks in Korea.” Journal of Hydrology 342: 223-237. Knobeloch, L., et. al., 2000. “Blue Babies and Nitrate-Contaminated Well Water.” Environmental Health Perspectives 108 (7), 675-9. Kraemer, D.K., I. Rabinowitz, K.H. Johannesson, and K.J. Stetzenback. 1996. “Trace element geochemistry in water from selected springs in Death Valley National Park, California.” Ground Water 34 (1): 95-103. Kronvang, B. and A.J. Bruhn. 1996. “Choice of sampling strategy and estimation method for calculating nitrogen and phosphorus transport in small lowland streams.” Hydrological Processes 10: 1483-1501. Kupfersberger, H. and G. Blöschl. 1995. “Estimating aquifer transmissivities — on the value of auxiliary data.” Journal of Hydrology 165: 85-99. Kusnier, I. and E. C. da Fonseca. 1975. Rapport: Sur les Travaux de Follow—UP Effectues dans la Zone IX—Lanta, Subzone Djidja. Republic du Dahomey, Porgramme des Nations Unies Pour le Developpement, Project: Recherches Minieres au Dahomey, DAH 73/006. Lelong, F. 1963. Nouvelles données sur les “nappes d’arène” à la suite d’une reconnaissance hydrogéologique du Centre-Nord Dahomey (région de Parakou et de Nikki). Publ. CIEH, 37 pp. Ouagadougou, Burkina Faso. L’Hirondel, J. and J.-L. L’Hinrondel. Nitrate and Man: Toxic, Harmless or Beneficial. New York: CABI Publishing, 168pp.

259

Masoner, Jason R. and Shana L. Mashburn. 2004. Water Quality and Possible Sources of Nitrate in the Cimarron Terrace Aquifer, Oklahoma, 2003.” USGS/SIR 20045221. McCarthy, J.F. and C. Degueldre. 1993. “Sampling and characterization of colloids and particles in groundwater for studying their role in contaminant transport.” Environmental Particles, vol. 2. Buffle, J. and H.P. van Leeuwen, eds. Lewis Publishers: Boca Ranton, 247-315. McLaren, J.W., 1994. Certificate of Analysis, SLRS-3 Riverine Water Reference Material for Trace Metals. NRC, Ottawa. Mende, Andreas, Allan Astorga and Detlev Neumann. 2007. “Strategy for groundwater management in developing countries: A case study in northern Costa Rica.” Journal of Hydrology 334: 109-124. Moosmann, L., B. Muller, R. Gachter, A. Wuest, E. Butscher, and P. Herzog. 2005. “Trend-oriented sampling strategy and estimation of soluble reactive phosphorus loads in streams.” Water Resources Research 41, W01020. Mueller, D.K. and D.R. Helsel. 1996. Nutrients in the Nation’s Waters—To Much of a Good Thing?. Reston, VA: U.S. Geological Survey, Circular 1136. Nolan, Bernard T. , Barbara C. Ruddy, Kerie J. Hitt, and Dennis R. Helsel. 1997. “Risk of Nitrate in Groundwaters of the United States—A National Perspective.” Environmental Science and Technology 31: 2229-2236. Novikov, Alexander P., Stepan N. Kalmykov, Satoshi Utsunomiya, Rodney C. Ewing, François Horreard, Alex Merkulov, Sue B. Clark, Vladimir V. Tkachev, and Boris F. Myasoedov. 2006. “Colloid Transprot of Plutonium in the Far-Field of the Mayak Production Association, Russia.” Nature 314: 638-641. Novondo, T.S. and T.E. Cloete. 1999. “Bacterial and chemical quality of water supply in the Dertig village settlement.” Water SA 25 (2): open access online at http://www.wrc.org/za. Preston, Stephen D., Victor J. Bierman, Jr, and Stephen E. Silliman. 1989. “An Evaluation of Methods for the Estimation of Tributary Mass Loads.” Water Resources Research 25 (6): 1379-1389. Pougnet, R.,1955. Le Précambrien du Dahomey. Thèse Univ. Clermont-Ferrand et Bull. Dir. Féd. Min. Afr. Occ. Fr., Dakar, 22, 186 p. Sénégal. Reszat, T. N. and M. J. Hendry. 2007. “Complexation of Aqueous Elements by DOC in a Clay Aquitard.” Ground Water 45 (5): 542-553.

260

Roadcap, George S., Keith C. Hackley, Hue-Hwa Hwang, Thomas M. Johnson. 2001. “Application of Nitrogen and Oxygen Isotopes to Identify Sources of Nitrates.” Illinois Groundwater Consortium Conference, web publication, www.siu.edu/worda/igc/proceedings/01/roadcap.pdf . Robertson, Dale M. 2003. “Influence of Different Temporal Sampling Strategies n Estimating Total Phosphorus and Suspended Sediment Concentrations and Transport in Small Streams.” Journal of the American Water Resources Association 39 (5): 1281-1308. Rock, L. and B. Mayer. 2002. “Isotopic Assessment of Sources and Processes Affecting Sulfate and Nitrate in Surface Water and Groundwater of Luxembourg.” Isotopes Environ. Health Stud. 38 (4): 191-206. Roope, Benjamin. 2003. Analysis of Elevated Uranium and Impact of the Cotton Industry on Groundwater in Benin, Africa. Master’s Thesis. University of Notre Dame. Salman, S.R. and A. Ruka’h. 1999. “Multivariate and principal component statistical analysis of contamination in urban and agricultural soils from north Jordan”, Environmental Geology 38(3), 265-270. Sampayo, F. F. and H. R. Wilke. 1963. “Temperature and Phosphates as Ground-Water Tracers.” Ground Water 1 (4): 36-38. Scott, Andrew. 1996. “Appropriate Technology: Is it ready for—and relevant to—the Millennium?” Appropriate Technology 23 (3): 1-4. Schoenen, D. 2002. “Role of disinfection in suppressing the spread of pathogens with drinking water: possibilities and limitations.” Water Research 36: 3874-3888. Sepulveda, Jaime, Jose Luis Valdespino, and Lourdes Garcia-Garcia. 2006. “Cholera in Mexico: The paradoxical benefits of the last pandemic.” International Journal of Infectious Diseases 10: 4-13. Silliman, S. E., M. Boukari, P. Crane, F. Azonsi, and C. R. Neal. 2007. “Observations on Elemental Concentrations of Groundwater in Central Benin.” Journal of Hydrology 335 (3-4): 374-388. Silliman, S.E. C. Hamlin, P.E. Crane, and M. Boukari. 2007. “International Collaborations and Incorporating the Social Sciences in Research in Hydrology and Hydrologic Engineering.” Journal of Hydrologic Engineering. In press. Somasundaram, M.V., G. Ravindran, and J. H. Tellam. 1993. “Ground-Water Pollution of the Madras Urban Aquifer, India.” Ground Water 31 (1): 4-11.

261

Stetzenbach, K.J., M. Amano, D.K. Kreamer, and V.F. Hodge. 1994. “Testing the limits of ICP-MS: Determination of trace elements in ground water at the parts-pertrillion level.” Ground Water 32 (6), 976-985. Stetzenbach, K.J., I.M. Farnham, V.F. Hodge, and K.H. Johannesson. 1999. “Using multivariate statistical analysis of groundwater major cation and trace element concentrations to evaluate groundwater flow in a regional aquifer.” Hydrological Processess 13, 2655-2673. Tardy Y. 1969. Géochimie des altérations, étude des arènes et des eaux de quelques massifs cristallins d’Europe et d’Afrique. Mém. Serv. Cart. Géol. AlsaceLorraine, nº 31. TED (American University Trade and Environment Database). 2002. Benin Hazardous Waste. Web publication: http://www.american.edu/projects/mandala/TED/benin.htm . Tuhela, Laura, Stuart A. Smith, and Olli H. Tuovinen. 1993. “Microbiological Analysis of Iron-Related Biofouling in Water Wells and a Flow-Cell Apparatus for Field and Laboratory Investigations.” Ground Water 31 (6): 982-988. Tukey, J.W. 1977. Exploratory Data Analysis. Reading, Massachusetts: Addison-Wesley Publishing Company, 688 pp. U.S. Environmental Protection Agency (USEPA). June 2003. “List of Contaminants & their MCLs; EPA 816-F-03-016.” Web publication: http://www.epa.gov/safewater/mcl.html . UNDP. 2006. “World water and sanitation crisis urgently needs a Global Action Plan.” United Nations Development Programme, web publication: http://content.undp.org/go/newsroom/november-2006/hdr-water-20061109.en . UNICEF. 2006. “Aresenic Mitigation in Bangladesh.” United Nations Children’s Fund, web publication: http://www.unicef.org/Bangladesh/Arsenic.pdf . Wakida, Fernando T., and David N. Lerner. 2005. “Non-agricultural sources of groundwater nitrate: a review and case study.” Water Research 39: 3-16. Walton, W. C. 1963. “Microtime Measurements of Ground-Water Level Fluctuations.” Ground Water 1 (2): 18-19. Wei, William W. 1990. Time Series Analysis: Univariate and Multivariate Methods. Redwood City, California: Addison-Wesley Publishing Company, Inc., 478 pp. Weight, W.D. and J.L. Sonderegger. 2001. Manual of Applied Field Hydrology. McGraw-Hill: New York, 609 pp.

262

Willis, C. J., G. H. Elkan, E. Horvath and K. R. Dail. 1975. “Bacterial Flora of Saline Aquifers.” Ground Water 13 (5): 406-409. WHO. 2001. “Arsenic in Drinking Water.” World Health Organization: Fact Sheet No210, web publication: http://www.who.int/mediacentre/factsheets/fs210/en/ . WHO. 2004. Uranium in Drinking-water: Background document for development of WHO Guidelines for Drinking-water Quality. World Health Organization, 15 pp. WHO. 2006. Guidelines for Drinking-water Quality: incorporating first addendum, 3 ed. Volume 1: Recommendations. World Health Organization, web publication: (http://www.who.int/water_sanitation_health/dwq/gdwq3rev/en/index.html . World Health Organization (WHO). 2004. “Water, Sanitation and Hygiene Links to Health Facts and Figures—Updated November 2004.” Yenigül, N. Buket, Amro M. M. Elfeki, Johannes C. Gehrels, Cees van den Akker, André T. Hensbergen and F. Michel Dekking. 2005. “Reliability assessment of groundwater monitoring networks at landfill sites.” Journal of Hydrology 308: 117.

263