Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Hydrogeochemical site characterization and groundwater flow modeling of the arsenic-contaminated Gotra… Koenig, Cassandra E.M. 2011

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2011_spring_koenig_cassandra.pdf [ 14.05MB ]
Metadata
JSON: 24-1.0053170.json
JSON-LD: 24-1.0053170-ld.json
RDF/XML (Pretty): 24-1.0053170-rdf.xml
RDF/JSON: 24-1.0053170-rdf.json
Turtle: 24-1.0053170-turtle.txt
N-Triples: 24-1.0053170-rdf-ntriples.txt
Original Record: 24-1.0053170-source.json
Full Text
24-1.0053170-fulltext.txt
Citation
24-1.0053170.ris

Full Text

  HYDROGEOCHEMICAL SITE CHARACTERIZATION AND GROUNDWATER FLOW MODELING OF THE ARSENIC-CONTAMINATED GOTRA AQUIFER WEST BENGAL, INDIA  by  Cassandra E.M. Koenig Hon. B.Sc., University of Toronto, 2005  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES  (Geological Sciences)  THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)  April, 2011  © Cassandra E.M. Koenig, 2011 Abstract ii ABSTRACT Groundwater geochemistry and flow have been studied in Gotra, West Bengal, India, where geogenic arsenic contaminates groundwater at levels above World Health Organization limits. The village is situated upon the natural levee of an abandoned channel, which terminates a fluvio-deltaic depositional sequence.  The formerly prograding meander bend deposited point- bar sands during the Holocene that now comprise the 30m-thick shallow aquifer, while incising deeper Pleistocene sands and a shallow floodplain sequence.  Hand-pumped tubewells are completed in point-bar sands, whereas irrigation wells are generally screened within Pleistocene materials.  Hydraulic conductivity of the point-bar aquifer is approximately 5x10-4 m/s.  A leaky- confining layer of levee/crevasse splay deposits overlies this aquifer beneath crop fields.  Near the village, the shallow aquifer is confined by channel-fill silts that have an estimated conductivity of 1x10-7 m/s.  The paleohorizon separating Holocene and Pleistocene sediments is marked locally by fine-grained material comprised of organic material and a hard clay. Groundwater elevations vary 7m over the course of a given year due to monsoonal climate and groundwater extraction.  Convergent flow towards shallow pumping wells is a salient feature during irrigation season, and persistent localized downward gradients suggest that meteoric recharge and pumping are the predominant groundwater sources and sinks.  A northeast trending geochemical gradient in the point-bar aquifer suggests that local flows transporting arsenic are directed away from the channel-fill silt.  Concentration gradients of terminal electron acceptors and redox reaction products in the shallow aqueous profile also coincide with a flowpath originating from this unit.  A numerical model of groundwater flow was developed based on a conceptual model derived from field observations to investigate controls of arsenic transport and to provide a context in which to interpret geochemical data.  Short- (7-day) and long- (3-year) term transient simulations were implemented to simulate groundwater flowpaths, water balance, and average residence times.  Modeling results support field observations that contamination is related to the channel-fill unit deposit, and suggest that the water balance has been significantly altered compared to pre-irrigation conditions.  Results also suggest contaminant flushing will not occur over a human timescale.  Table of Contents iii TABLE OF CONTENTS ABSTRACT ............................................................................................................................ ii TABLE OF CONTENTS ......................................................................................................... iii LIST OF TABLES ...................................................................................................................vi LIST OF FIGURES ................................................................................................................ vii ACKNOWLEDGEMENTS....................................................................................................... x 1.0 INTRODUCTION ............................................................................................................ 1 1.1. Purpose and Objectives .................................................................................................. 2 1.2. Thesis Structure .............................................................................................................. 3 2.0 ARSENIC CONTAMINATION OF THE BENGAL DELTA PLAIN AQUIFERS .................. 5 2.1. Arsenic Occurrence and Provenance .............................................................................. 6 2.2. Arsenic Mobilization within BDP Aquifers ........................................................................ 8 2.3. Depositional History of the Bengal Delta Plain ................................................................ 9 2.3.1. Groundwater Arsenic Occurrence in Relation to Geology................................ 11 2.4. Hydrogeology of the Bengal Delta Plain ........................................................................ 13 2.4.1. Aquifer Properties ........................................................................................... 14 2.4.2. Hydrostratigraphy and Aquifer Fabric .............................................................. 15 2.4.3. Aquifer Flooding and Meteoric Recharge ........................................................ 16 2.4.4. Regional Flow and Delta Flushing ................................................................... 17 2.4.5. Intermediate and Local Scale Flow ................................................................. 18 2.4.6. Irrigation Coverage and Localized Flow Cells.................................................. 20 2.5. Implications of BDP Hydrogeology for Arsenic Distribution and Transport ..................... 21 2.5.1. Geological and Geomorphological Manifestations ........................................... 21 2.5.2. Source of Organic Substrate ........................................................................... 23 2.5.3. Downward Migration of Arsenic to Deep Aquifers ........................................... 24 2.6. Previous Groundwater Modeling in the BDP and the Gotra Research Site .................... 24 3.0 SITE CHARACTERIZATION METHODS ...................................................................... 26 3.1. Site Location and Previous Work .................................................................................. 26 3.2. Gotra Wells and Piezometers ........................................................................................ 27 3.2.1. Domestic and Irrigation Wells .......................................................................... 27 3.2.2. Observation Wells ........................................................................................... 28 3.2.3. Small-Diameter Piezometers ........................................................................... 29 3.3. Hydrologic and Hydrostratigraphic Characterization ...................................................... 29 3.3.1. Groundwater Levels ........................................................................................ 29 3.3.2. Piezometer Tests ............................................................................................ 30 3.3.3. Pumping Tests and Superposition Modeling ................................................... 30 3.3.4. Sediment Sampling for Grain Size Analysis .................................................... 31 3.4. Geochemical Data Collection ........................................................................................ 31 3.4.1. Electrical Conductivity, pH and Temperature Measurement ............................ 32 3.4.2. Dissolved Oxygen ........................................................................................... 32 3.4.3. Alkalinity .......................................................................................................... 32 Table of Contents iv 3.4.4. Dissolved Ammonia and Ferrous Iron ............................................................. 33 3.4.5. Groundwater Collected for Laboratory Analysis .............................................. 33 4.0 SITE CHARACTERIZATION AND CONCEPTUAL MODEL DEVELOPMENT .............. 47 4.1. Site Overview ................................................................................................................ 47 4.1.1. Physiography and Vegetation ......................................................................... 47 4.1.2. Meteorology .................................................................................................... 48 4.1.3. Hydrology ........................................................................................................ 48 4.1.4. Geology .......................................................................................................... 49 4.2. Hydrogeology ................................................................................................................ 52 4.2.1. Aquifer Parameters ......................................................................................... 52 4.2.2. Hydrostratigraphy ............................................................................................ 54 4.3. Piezometry .................................................................................................................... 55 4.3.1. Time-Series Observations ............................................................................... 55 4.3.2. Response to Pumping and Local Connectivity of Aquifer ................................ 56 4.3.3. Groundwater Flow Directions .......................................................................... 57 4.3.4. Piezometric Data Summary ............................................................................. 59 4.4. Monthly Water Budget ................................................................................................... 60 4.4.1. Groundwater Extracted through Irrigation Pumping (Q) ................................... 60 4.4.2. Net Recharge (G) ............................................................................................ 63 4.4.3. Other Potential Recharge Sources .................................................................. 67 4.4.4. Water Budget Summary .................................................................................. 68 4.5. Hydrogeochemistry ....................................................................................................... 69 4.5.1. Groundwater Quality and Contaminant Distribution ......................................... 70 4.5.2. Geochemical Controls on Dissolved Arsenic in Gotra ..................................... 71 4.5.3. Relationship between Arsenic Contamination and Hydrogeology .................... 72 4.5.4. Hydrogeochemistry Summary ......................................................................... 75 4.6. Conceptual Model and Numerical Modeling Goals ........................................................ 75 4.6.1. Conceptual Hydrogeologic and Arsenic Transport Model ................................ 76 4.6.2. Specific Goals of Numerical Modeling ............................................................. 77 5.0 NUMERICAL GROUNDWATER FLOW MODEL ......................................................... 120 5.1. Model Description and Inverse Modeling Stress Schedules ........................................ 120 5.1.1. Grid Design ................................................................................................... 121 5.1.2. Model Hydrostratigraphy and Hydraulic Parameter Values ........................... 122 5.1.3. Initial Conditions ............................................................................................ 123 5.1.4. Boundary Conditions ..................................................................................... 124 5.1.5. Time Discretization ........................................................................................ 127 5.2. Model Calibration ........................................................................................................ 128 5.3. Current Groundwater Flow System ............................................................................. 130 5.3.1. Water Table Configuration and Village Flow Directions ................................. 130 5.3.2. Groundwater Budget and Recharge Sources ................................................ 132 5.3.3. Groundwater Residence Times ..................................................................... 134 5.4. Sensitivity Analysis ...................................................................................................... 134 5.4.1. Effect on Parameter Variation on Model Calibration ...................................... 135 5.4.2. Impacts of Parameter Variation to the Gotra Groundwater Flow System ....... 135 Table of Contents v 5.5. Pre-Development Scenario ......................................................................................... 138 5.5.1. Principal Groundwater Flow paths and Travel Times .................................... 139 5.5.2. Hydrologic Balance and Recharge Sources .................................................. 140 5.5.3. Groundwater Residence Times and Aquifer Flushing .................................... 140 5.6. Implications for Arsenic Contamination of Gotra Groundwater .................................... 141 5.6.1. Principal Groundwater Flow Paths and Travel Times .................................... 141 5.6.2. Hydrologic Balance and Recharge Sources .................................................. 143 5.6.3. Groundwater Residence Times and Aquifer Flushing .................................... 144 5.7. Model Limitations and Opportunities for Improvement ................................................. 145 5.7.1. Boundary Conditions ..................................................................................... 146 5.7.2. Pumping Well Locations and Schedules ....................................................... 146 5.7.3. Regional Hydrostratigraphy ........................................................................... 146 5.7.4. Pond Recharge Contribution and Unsaturated Flow ...................................... 147 5.8. Summary of Numerical Model Results ........................................................................ 147 6.0 SUMMARY AND CONCLUSION ................................................................................. 188 REFERENCES ................................................................................................................... 191 APPENDIX A – BOREHOLE LOGS, GRAIN SIZE DATA AND WELL COMPLETION DETAILS ..................................................................................................................... 201 APPENDIX B – PIEZOMETRIC MONITORING .................................................................. 225 APPENDIX C – ESTIMATION OF HYDROGEOLOGIC PARAMETERS ............................ 229 C.1.0 Piezometer Tests ...................................................................................................... 229 C.1.1 Cooper-Bredehoeft-Papadopulos (CBP) Analysis .............................................. 230 C.1.2 Hvorslev Analysis .............................................................................................. 232 C.1.3 Bouwer-Rice Analysis ........................................................................................ 234 C.1.4 Summary of Slug Results .................................................................................. 234 C.2.0 Time-Drawdown Analysis and Superposition Modeling ............................................. 235 C.2.1 Least Squares Fitting to Theis Solution – Superposition Modeling ..................... 236 C.2.2 Semi-Log Methods............................................................................................. 239 C.2.3 Summary and Discussion of Time-Drawdown Analysis ..................................... 241 C.3.0 Grain Size Methods .................................................................................................. 243 APPENDIX D – CLIMATE DATA AND ESTIMATION OF HYDROLOGIC PARAMETERS . 272 D.1.0 Evapotranspiration .................................................................................................... 272 D.1.1 Calculation of Reference Evapotranspiration, ETo ............................................. 272 D.2.2 Calculation of Crop Evapotranspiration, ETc...................................................... 277 D.1.3 Calculation of Potential Evaporation, Ep ............................................................ 279 D.2.0 Surface Runoff .......................................................................................................... 280 APPENDIX E – GEOCHEMISTRY ..................................................................................... 288 APPENDIX F – NUMERICAL MODEL ADDENDA ............................................................. 294 APPENDIX G – DISCRETE REMOVAL OF ARSENIC CONSIDERING SORPTION ......... 298 List of Tables vi LIST OF TABLES Table 4.1: Geometric averages of parameter estimates. .................................................... 116 Table 4.2: Summary of vertical gradients ........................................................................... 116 Table 4.3: Horizontal hydraulic gradients calculated between pairs of wells. ...................... 117 Table 4.4: Monthly evapotranspiration and evaporation estimates. .................................... 118 Table 4.5: Monthly runoff estimates. ................................................................................... 118 Table 4.6: Monthly estimates of recharge to the water table. .............................................. 119 Table 5.1: Statistical analysis of model calibration .............................................................. 182 Table 5.2:  Errors for simulation with a low-K paleosol ....................................................... 182 Table 5.3: Calibrated Hydrogeologic Dataset. .................................................................... 182 Table 5.4: Distribution of Recharge Sources for the Model Domain and Gotra Area........... 183 Table 5.5: Base case groundwater budget. ........................................................................ 184 Table 5.6: Groundwater budget for scenario without pumping ............................................ 185 Table 5.7: Irrigation Return Flow to the Model Domain and Gotra Area. ............................. 186 Table 5.8: Sensitivity cases that improve the calibration of the LTT Model. ........................ 186 Table 5.9: Percent changes to groundwater fluxes through the Gotra area and simulated residence times for sensitivity tests. ................................................................................... 187  List of Figures vii LIST OF FIGURES Figure 3.1: General location of field site. .............................................................................. 34 Figure 3.2: Google Earth Image showing all wells in the study area. .................................... 35 Figure 3.3: Names and locations of domestic and observation wells. ................................... 36 Figure 3.4: Photographs of domestic wells in Gotra.............................................................. 38 Figure 3.5: Irrigation wells within the village and visible in the surrounding fields. ................ 39 Figure 3.6: Preliminary assessment of arsenic contamination of Gotra domestic wells. ........ 40 Figure 3.7: Hand-flapper drilling technique. .......................................................................... 41 Figure 3.8: Collection of sediment from drill pipes. ............................................................... 41 Figure 3.9: Hand-flapper drill setup. ..................................................................................... 42 Figure 3.10: Photograph of drill pipe. .................................................................................... 42 Figure 3.11: Observation well screens and silt traps. ............................................................ 43 Figure 3.12: Well installation. ............................................................................................... 43 Figure 3.13: Multilevel sampling ports. ................................................................................. 44 Figure 3.14: Well completion. ............................................................................................... 44 Figure 3.15: Standpipe piezometer. ...................................................................................... 45 Figure 3.16: Depth to water measurement. .......................................................................... 45 Figure 3.17: Slug test setup at well GSI0609. ....................................................................... 46 Figure 4.1: Google Earth Image of Chakdaha and surrounding region. ................................ 79 Figure 4.2: Weather stations and Triberni gauging station. ................................................... 80 Figure 4.3: Site climate data. ................................................................................................ 81 Figure 4.4: Historical precipitation data ................................................................................ 82 Figure 4.5: Google Maps image showing surface hydrologic features. ................................. 83 Figure 4.6: Classical depositional model of a meandering stream. ....................................... 84 Figure 4.7: Stratigraphic succession resulting from a meandering river depositional environment.  ........................................................................................................................ 84 Figure 4.8: Geological cross-section locations. ..................................................................... 85 Figure 4.9: Longitudinal cross section AA‟ showing abandoned channel sequence incising older alluvial deposits. ................................................................................................................... 86 Figure 4.10: Cross-section BB‟ running parallel to the meander scar. .................................. 87 Figure 4.11: Cross-section CC‟ ............................................................................................ 88 List of Figures viii Figure 4.12: Depositional events leading to the development of the stratigraphy in Gotra. ... 89 Figure 4.13: Summary of computed hydraulic conductivities. ............................................... 90 Figure 4.14: Stratigraphic interpretation and division of Gotra into hydrostratigraphic units. . 91 Figure 4.15: Hydraulic head measured in Gotra from 2006 to 2009. ..................................... 92 Figure 4.16: Response of potentiometric levels to pumping over the course of a day. .......... 93 Figure 4.17: Intensive water level measurement .................................................................. 94 Figure 4.18: Hydrochemical data collected over the 2006-2007 period. ............................... 95 Figure 4.19: Head correlation matrices for observation wells. ............................................... 96 Figure 4.20: Potentiometric response to pumping. ............................................................... 97 Figure 4.21: Vertical hydraulic gradient observed between GSI0603 and GSI0604. ............. 98 Figure 4.22: Vertical hydraulic gradient observed in the channel-fill silt near GSI0605. ........ 99 Figure 4.23: Potentiometric map of Gotra during a load-shedding event in March of 2007. .. 99 Figure 4.24: Potentiometric map of Gotra during a load-shedding event in February of 2008. ..  ...................................................................................................................... 100 Figure 4.25: Potentiometric map of Gotra during the 2007 monsoon. ................................. 101 Figure 4.26: Schematic representation of flow directions from gradient interpretations ....... 102 Figure 4.27: Hooghly River stage data. .............................................................................. 103 Figure 4.28: Daily irrigation pumping duration. ................................................................... 104 Figure 4.29: Gotra pumping rates. ...................................................................................... 105 Figure 4.30: Monthly water balance at the water table ........................................................ 106 Figure 4.31: ICPMS results for groundwater arsenic. ......................................................... 107 Figure 4.32: Geochemical depth profiles of groundwater in Gotra, ..................................... 108 Figure 4.33: Major element geochemistry of domestic and observation wells in Gotra. ...... 109 Figure 4.34: Profile of arsenic contamination. ..................................................................... 109 Figure 4.35: Relationship of arsenic to dissolved iron and manganese .............................. 110 Figure 4.36: Relationship of iron to dissolved sulfate and phosphate. ................................ 110 Figure 4.37: Mineral saturation controls of iron in Gotra groundwater. ................................ 111 Figure 4.38: Concentrations of dissolved organic carbon (DOC) in shallow aqueous samples.  ...................................................................................................................... 111 Figure 4.39: Profile contours of selected reactants and products of TEAPs. ....................... 112 Figure 4.40: Schematic SW-NE section through the village showing interpreted redox zonation.  ...................................................................................................................... 113 Figure 4.41: Schematic of major flow paths during the wet and dry seasons. ..................... 114 List of Figures ix Figure 4.42: Conceptual model of arsenic release and transport in Gotra. ......................... 115 Figure 5.1: Plan view of model domain extent. ................................................................... 151 Figure 5.2: Model discretization. ......................................................................................... 152 Figure 5.3: Model hydrostratigraphy. .................................................................................. 153 Figure 5.4: Model boundary conditions. .............................................................................. 154 Figure 5.5: DEM image showing physical basis for head dependent boundaries. ............... 155 Figure 5.6: Irrigation wells shown with model hydrostratigraphy. ........................................ 156 Figure 5.7: Distribution of residual errors in calibrated models. .......................................... 157 Figure 5.8: Observed and simulated heads in the calibrated STT model. ........................... 158 Figure 5.9: Observed and simulated heads in the calibrated LTT model. ........................... 161 Figure 5.10: Potentiometric maps and groundwater flow directions during the monsoon. ... 165 Figure 5.11: Potentiometric maps and groundwater flow directions during the pumping season.  ...................................................................................................................... 166 Figure 5.12: Comparison of observed to simulated vertical gradients. ................................ 167 Figure 5.13: Sectional view of simulated pathlines and well capture. .................................. 168 Figure 5.14: Global mass balance over the observation period. ......................................... 169 Figure 5.15: Mass balance for the Gotra area over the observation period. ........................ 170 Figure 5.16: Effect of parameter variation on model calibration. ......................................... 171 Figure 5.17: Scaled sensitivities of modeled groundwater fluxes between the Gotra area and surrounding domain............................................................................................................ 172 Figure 5.18: Scaled sensitivities of modeled residence times in the channel-fill silt and point-bar sand. ...................................................................................................................... 173 Figure 5.19: Sectional view of pathlines for scenario with Kh/Kv = 10 .................................. 174 Figure 5.20: Pathlines for scenario with low conductivity layer on the paleosol horizon. ..... 175 Figure 5.21: Sectional view of pathlines for no-pumping scenario. ..................................... 176 Figure 5.22: Global mass balance for scenario with no pumping ........................................ 177 Figure 5.23: Mass balance for the Gotra area with no pumping. ......................................... 178 Figure 5.24: Mass balance comparison between simulations with and without pumping for the full domain. ...................................................................................................................... 179 Figure 5.25: Mass balance comparison between simulations with and without pumping for the Gotra sub-zone. ................................................................................................................. 180 Figure 5.26: Predicted flushing of arsenic from the channel-fill sediments considering desorption from sediments and irrigation pumping. .............................................................................. 181 Acknowledgements x ACKNOWLEDGEMENTS Financial support for this work was provided by NSERC and the Environment and Health Program of Natural Resources Canada, as part of a joint research project between the Geological Survey of Canada and the Geological Survey of India entitled “Development of a mitigation strategy to manage risk from arsenic toxicity in groundwater of West Bengal, India”. 1.0 Introduction 1 1.0 INTRODUCTION A considerable amount of groundwater used for domestic and agricultural purposes across Southeast Asia is pumped from aquifers that are contaminated with geogenic arsenic. Currently, more than 60 million people across the sub-continent are experiencing adverse health effects as a result of prolonged consumption of groundwater containing arsenic in excess of the World Health Organization (WHO) standard of 10 µg/L [Berg et al., 2007; Dhar et al., 1997; M.  Hossain, 2006; Mandal et al., 1996; Polya and Charlet, 2009].  Severely impacted populations include residents of Bangladesh [BGS and DPHE, 2001], West Bengal (India) [Mandal et al., 1996], Cambodia [Polya et al., 2005], Vietnam and Myanmar [Winkel et al., 2008]. Traditionally, scientific research has sought to understand the biogeochemical source and delivery of arsenic to Southeast Asian groundwaters.  Oxidative weathering of arsenic-bearing sulphides is believed to have effectively transferred arsenic to the surfaces of iron-oxides that simultaneously precipitated at the time of sediment deposition [Fendorf et al., 2003]. Subsequent desorption of arsenic from these mineral surfaces and generation of contemporary arsenic enrichment is understood to be facilitated by anaerobic microbial reduction (e.g. Nickson et al., [1998]).  Iron-reducing bacteria have been identified as the principal drivers of these reactions, fuelled by organic matter [M.  Hossain, 2006]. In contrast to biogeochemical studies, little work has focussed on the role of groundwater flow in arsenic distribution.  However, there has been rising interest in the research community concerning groundwater flow with respect to the transport of arsenic itself and solutes that control its mobility.  For example, recent evidence has suggested that human exploitation of groundwater resources through irrigation practices may have enhanced the mobilization of the contaminant and increased the risk of exposure to large populations [Harvey et al., 2002].  Two growing concerns that relate directly to irrigation use and groundwater flow include the potential for increasing contamination in shallow groundwater [Harvey et al., 2006], and the migration of arsenic to deeper resources, which are largely arsenic-free [Burgess et al., 2002]. The merit of detailed hydrogeologic characterization and flow models to elucidate spatial and temporal distributions of arsenic in groundwater has been widely recognized in the research community [Polya and Charlet, 2009].  Consideration of alternate water resources must be carried out on a site-by-site basis, as unique combinations of geology, microbiology and 1.0 Introduction 2 hydrogeology conspire to produce high variability of the contaminant in groundwater (e.g. Weinman et al., [2008]; Harvey et al., [2006].  Such characterization and modeling thus are of principal importance for assessing risks of future engineering endeavours aimed at developing contaminant-free, sustainable aquifer resources for currently affected communities. 1.1. Purpose and Objectives Our study site is located in Nadia District of West Bengal, in the farming village of Gotra.  The village lies within a severely arsenic affected zone where groundwater is heavily pumped for both irrigation and domestic purposes.  Since 1988, the Geological Survey of India (GSI) has carried out investigations in the village area, which have helped delineate areas that have arsenic-contaminated groundwater and recognize its geological cause.  In April 2003, Natural Resources Canada and the Ministry of Mines of the Republic of India initiated a joint project between the Geological Survey of Canada (GSC), the GSI and the University of British Columbia (UBC), aimed at identifying the mechanisms responsible for arsenic release from sediments into groundwater.  The purpose of this thesis is to supplement the project goals by investigating the occurrence of dissolved arsenic and its relationship to groundwater flow in the Gotra region.  Specifically, the goals of this work are to develop a conceptual hydrogeologic model for arsenic release to the village aquifer based on hydrogeochemical field observations, and to investigate potential advective controls on its distribution with a numerical groundwater model under current and pre-irrigation conditions.  Particular focuses of numerical modeling include:  The delineation of principal groundwater flow paths and associated travel times in the village area.  This will provide insight to the specific source of contaminant, or solutes facilitating release.  Associated travel times will reveal the amount of time required for advective transport between localities of interest, namely from origin to the contaminated or susceptible zones.  Comparison to a pre-development case will reveal whether principal flow paths and travel times have been impacted by irrigation.  The quantification of the site water balance under current and pre-irrigation conditions. This may reveal whether recent perturbations to recharge and discharge have impacted arsenic distribution at the site.  Two specific areas of interest are the recharge to the ground surface and flow of shallow groundwater to the deep aquifer, where the majority of irrigation wells are screened. 1.0 Introduction 3  Bulk residence times of groundwater within the Holocene materials.  The amount of time required to flush solid and aqueous phase arsenic from these materials with groundwater is of interest because it provides an indication of when arsenic should be exhausted from the aquifer.  Again, comparison of pre-development to current conditions may reveal agricultural practices have impacted this process. The modeling results will be used to direct future research and planning efforts towards identifying a sustainable drinking water source for the village. 1.2. Thesis Structure This thesis consists of six chapters, including the current introductory chapter, and seven appendices.  Chapters 2 through 6 are organized as follows: Chapter 2 reviews aspects of Bengal Basin geology and hydrogeology, as well as arsenic release hypotheses in the literature that form the conceptual basis for this work.  This chapter focuses on highlighting characteristics of groundwater flow and aquifer systems to guide conceptual model development for Gotra.  Pertinent theories regarding the influence of groundwater flow on arsenic distribution are also emphasized in the review. Chapter 3 provides a summary field and analytical methods used to characterize the stratigraphy, hydrogeology and hydrochemistry of the village area.  This begins with a detailed introduction and discussion of work previously conducted at the site.  Preliminary surveys of contamination of domestic wells are presented in this chapter as they guided the emplacement of observation wells for this project.  Drilling techniques, piezometer and datalogger installation, hydrogeologic testing and geochemical sampling and analyses that have been conducted as a part of the current work in Gotra are all described in this chapter. Assimilation of field data into a conceptual groundwater flow and arsenic transport model for Gotra is carried out in Chapter 4.  This includes a compilation of existing physiographic, meteorological and hydrologic information as well as the interpretation of borehole stratigraphy from drilled holes to develop an overview of the site.  This compilation is then used to guide the interpretation of hydrogeologic test data and to establish the hydrostratigraphic context for the site.  Analysis of potentiometric data is carried out subsequently, highlighting the seasonality of groundwater levels and prominent hydraulic gradients between observation wells at the site.  A combination of piezometric and meteorological data is then used to calculate a site water 1.0 Introduction 4 balance.  This water balance is intended to provide insight into the fate of precipitation as well as estimate of input recharge to initiate numerical model calibration.  Subsequently, aqueous geochemistry observations are interpreted in terms of potential arsenic release mechanisms and relationship to groundwater flow directions.  At the end of this chapter, the conceptual model of site hydrogeology and arsenic transport is presented, and the key numerical modeling objectives are summarized. The numerical groundwater flow model is developed, calibrated, tested and interrogated in Chapter 5.  Direct implementation of the conceptual model and its transience into the 3D- finite difference code MODFLOW is carried out through consideration of grid design, the assignment of hydrostratigraphic units, initial and boundary conditions, as well as time discretization.  The model is calibrated in transient mode iterating between two independent stress schedules, such that both provide acceptable fits to piezometric data.  The base case simulated flow system is then presented, highlighting principal flow directions and travel times of interest, water balance components and bulk groundwater residence times through Holocene materials.  This is followed by an evaluation of the model through a comprehensive sensitivity analysis, which further constrains input parameters, and documents the specific impacts of uncertainties to groundwater flow in the Gotra area.  The model is then interrogated to simulate a pre- development scenario without irrigation pumping to investigate the impact that pumping has had on principal flow directions and travel times, water balance components and bulk groundwater residence times.  The key implications for arsenic transport and distribution are then discussed in light of current stressors on the flow system, compared to the pre-development case. Following this discussion, important model limitations and opportunities for improvement are noted. Finally, a summary of the major conclusions arrived at from this work is presented in Chapter 6. This chapter includes a recap of key field observations, numerical modeling results and implications for arsenic release at the site. 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 5 2.0 ARSENIC CONTAMINATION OF THE BENGAL DELTA PLAIN AQUIFERS Since the 1970‟s, a large proportion of the rural population of Bangladesh and West Bengal (India) switched its water supply from surface to groundwater to avoid water-borne disease (e.g. Yu et al., [2003]).  The high-yielding alluvial aquifers of the Bengal Delta Plain (BDP) were subsequently tapped by over 10 million hand-pumped tubewells to satisfy the domestic water demands of hundreds of millions of people [Nath et al., 2008b].  The relatively low cost of materials and labour for well emplacement also made it economically feasible for mass installation of irrigation wells throughout the region.  This helped make Bangladesh and West Bengal largely self-sufficient in food production over the past several decades [M. Hossain et al., 2003; A A Rahman and Ravenscroft, 2003].  The now accessible groundwater aids in the irrigation of boro- (dry-season) rice, which has increased cultivatable land within Bangladesh by 45% and the ability to support a population that has tripled in size over 40 years [Harvey et al., 2006].  Similarly, the past 40 years has observed an increase of irrigation coverage from ~31% to ~62% in West Bengal, which improves the agricultural productivity and sustainability of population growth within the state.  To accommodate the projected population increase by the year 2050, West Bengal irrigation coverage must continue to increase so that food grain output exceeds 400 million tonnes, from the current ~210 million tonnes of grain produced annually [Gupta, 2006].  Clearly, the groundwater tapped from the aquifers of the BDP has become a fundamental resource for Bangladesh and West Bengal, necessary to supply inhabitants with coliform-free drinking water, as well as supplementing agricultural yields. Tragically, significant quantities of the groundwater in the aquifers of the BDP are contaminated by geogenic arsenic.  This problem was not initially recognized during tubewell emplacement, as wells were generally drilled to minimum depths where acceptable yields and salinity levels of groundwater could be obtained [BGS and DPHE, 2001].  Elevated arsenic concentrations in groundwater were first identified in West Bengal in the late 1970's [Acharyya et al., 2000] then in Bangladesh in the 1990's [BGS and DPHE, 2001].  As such, arsenic contamination has resulted in excessive exposure to over 40 million people for more than 3 decades [Smedley and Kinniburgh, 2002].  The problem has been called the „largest poisoning of a population in history‟ [Smith et al., 2000], with an estimated 25 to 33% of wells emplaced across the basin contaminated beyond the World Health Organization (WHO) permissible limit of 10 µg/L [Horneman et al., 2004; McArthur et al., 2004].  Ironically, the assumed „safe‟ alternative of groundwater has plagued the population with a slow acting poison, in place of the original acute threat of pathogenic infection.  Prolonged exposure due to increased groundwater consumption 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 6 has increased death rates due to cancer and the development of severe health problems such as skin lesions and arsenicosis [Anawar et al., 2002; Karim, 2000; M M Rahman et al., 2001; Smith et al., 2000; Yu et al., 2003].  These health effects have been attributed directly to mass arsenic exposure, as detailed records and statistics from the region are generally lacking. 2.1. Arsenic Occurrence and Provenance The most severely arsenic-affected regions in India and Bangladesh are shown in Figure 1 of Acharyya et al., [2000].  These areas are generally bound in the west by the River Hooghly- Bhagirathi, in the east by the Tripura Hills, in the north by the River Padma, and terminate in the south approximately 100km inland from the Bay of Bengal.  In the eastern Indian affected area, contamination follows the River Meghna north to the Shillong Plateau, and extends approximately 50km on the eastern and western flanks [Acharyya et al., 2000; M M Rahman et al., 2001].  The total polluted area in the Basin is approximately 240,000 km2 [M M Rahman et al., 2001]. A general southward enrichment of arsenic in tube-wells has been observed regionally along the Ganges-Brahmaputra-Meghna complex [BGS and DPHE, 2001], as well as in the fluvial drainages of West Bengal [Sengupta et al., 2004].  However, arsenic distribution commonly exhibits high degrees of spatial heterogeneity over much smaller scales throughout the delta. Contamination has been described by several researchers as „patchy‟ [BGS and DPHE, 2001; Ravenscroft et al., 2005; Yu et al., 2003], as order of magnitude spatial variations in groundwater arsenic are common over lateral scales of approximately 100m [McArthur et al., 2004; Nath et al., 2008b; van Geen et al., 2003b; Weinman et al., 2008], and metre scales in the vertical dimension [Harvey et al., 2006; Harvey et al., 2002; McArthur et al., 2004]. Currently, there is no universal agreement on the cause of this variability, although links to site- specific combinations of sedimentological features, groundwater hydrology and biogeochemical mechanisms have been postulated by several research groups [BGS and DPHE, 2001; Harvey and Beckie, 2005; Nath et al., 2008b; Weinman et al., 2008]. Over regional (i.e. 102 km) scales, the vertical distribution of dissolved arsenic is consistent, with high concentrations occurring in shallow (<100m deep) sediments and concentrations below detection limits at greater (>100m) depths [BGS and DPHE, 2001; Harvey et al., 2002; Ravenscroft et al., 2001; van Geen et al., 2003b].  In many cases, vertical arsenic profiles appear to exhibit a distinct „bell-shaped‟ pattern, with peak concentrations (approximately 1100μg/L) are observed at depths from 20-70m [Harvey and Beckie, 2005; McArthur et al., 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 7 2001; Nickson et al., 2000; Ravenscroft et al., 2001].  Some researchers, however, have observed considerable scatter in vertical profiles of arsenic in groundwater between 10 and 40m depths [Nath et al., 2008b].  Other exceptions to this pattern have been observed in West Bengal, where dissolved arsenic can exceed WHO standards at depths greater than 100m [A Mukherjee and Fryar, 2008]; although the general trend of decreasing arsenic with depth is still observed. There is a general consensus within the literature that the arsenic polluting the BDP aquifers was originally derived from iron sulphides (i.e. arsenopyrite or arsenian pyrite (FeAsS)) hosted in the Himalayan Mountains [Acharyya et al., 1999; BGS and DPHE, 2001; Harvey et al., 2006; McArthur et al., 2004].  Holocene-aged (i.e. 10 ka) alluvial sediments deposited by the Ganges- Brahmaputra river system are thought to be the major source of contemporary arsenic pollution found in the delta [Acharyya and Shah, 2007; Acharyya et al., 2005; BGS and DPHE, 2001; Bhattacharya et al., 1997], although high arsenic concentrations are also observed in deeper, Pleistocene (i.e. 1.6 Ma) aquifers (e.g. Pal and Mukherjee, [2009]). Oxidation of arsenopyrite at the ground surface during the early accretion of the delta is the most probable cause of initial arsenic mobilization from host minerals [Acharyya et al., 1999; Anawar et al., 2002; Harvey et al., 2002; McArthur et al., 2001; Nickson et al., 2000]. Processes such as sorption onto colloids, or secondary mineral precipitates (e.g. hydrous ferric oxides, (HFOs) or manganese oxides), during sediment deposition are likely to have scavenged arsenic from surface waters, concentrating it in the delta sediments [Acharyya et al., 2005; Fendorf et al., 1997; McArthur et al., 2004; Mok et al., 1988; Ravenscroft et al., 2001].  The onset of strongly reducing conditions following sediment burial is believed to have liberated the arsenic by means of microbially mediated reductive-desorption of arsenic from mineral surfaces, followed by the reductive-dissolution of HFO or manganese oxide minerals [BGS and DPHE, 2001; Harvey et al., 2002; Nickson et al., 1998; Smedley and Kinniburgh, 2002].  If some arsenic-bearing pyrite minerals were able to reach the Ganges delta and were incorporated into the Holocene aquifers, contamination may also have developed from mineral weathering caused by redox cycling associated with annual water table fluctuations [Harvey et al., 2006; Polizzotto et al., 2006].  Diagenetic iron sulphide may also have formed in the aquifers from microbial sulphate reduction in groundwater, providing a sink for dissolved arsenic at the time [McArthur et al., 2004], and potentially re-releasing the contaminant due to subsequent oxidation.  In general, the agreed-upon mechanism among researchers for current arsenic enrichment in groundwater involves the reduction of iron and manganese (oxy)hydroxides, based on affected aquifers being virtually devoid of pyrite (FeS2) or arsenopyrite and dissolved 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 8 sulphate (e.g. Nath et al., [2008a]) as well as exhibiting strongly reducing groundwaters [Harvey et al., 2002]. 2.2. Arsenic Mobilization within BDP Aquifers Much research on arsenic mobilization and transport in the aquifers of the BDP focussed almost exclusively on groundwater and sediment geochemistry  (e.g. Bhattacharya et al., [1997]; Nickson et al., [2000]; BGS and DPHE, [2001]; Smedley and Kinniburgh, [2002]; McArthur et al., [2004]), with little to no consideration for groundwater flow until recent years (e.g. Harvey et al., [2002]).  However, the ubiquity of arsenic in the BDP groundwaters in the absence of a wide- spread point source [McArthur et al., 2004; Nickson et al., 1998; Swartz et al., 2004], indicates that observed present-day accumulations are likely the result of interacting reactive and transport processes that have yet to flush the contaminant from groundwater systems [Harvey et al., 2006; Harvey et al., 2005; Smedley and Kinniburgh, 2002].  In addition, the fact that microbial and geochemical redox processes driving arsenic release can occur over timescales as short as hours [Polizzotto et al., 2006; van Geen et al., 2003b] suggests that advective transport of arsenic itself may play a key role in producing the observed contaminant distribution in groundwater.  Similarly, the transport of solutes (i.e. dissolved organic carbon) that create the necessary geochemical conditions for arsenic release will also be determined by groundwater flow [Harvey et al., 2006].  Evidently, the most powerful insight towards understanding the widely observed „patchiness‟ of dissolved arsenic in the BDP involves quantifying, and successfully coupling flow characteristics such as residence times and groundwater flow paths with biogeochemical theories.  This would also help to mitigate the risk of exposure. The need for a clear understanding of the nature of groundwater flow in order to decipher the arsenic problem within the BDP is now well recognized.  Several research groups have implemented approaches to couple interpretations of contaminant persistence to hydrogeology through site characterization and numerical modeling [Aziz et al., 2008; Harvey et al., 2002; 2005; Michael and Voss, 2009a; b; A Mukherjee et al., 2007a; Neumann et al., 2010; Polizzotto et al., 2008; Ravenscroft and McArthur, 2004; Ravenscroft et al., 2005; Stute et al., 2007]. Current unresolved issues regarding arsenic release from a hydrogeological standpoint include the source of organic substrate required for microbial metabolism, the location of release in the sediment profile, and the role of irrigation pumping in contaminant transport [Polizzotto et al., 2008].  This lack of universal agreement for explaining the specific evolution of arsenic contamination is likely a reflection of variable hydrological, biogeochemical and geological contributors between sites.  In addition, many of these hydrology studies have been conducted 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 9 at scales that are too large (e.g. Mukherjee et al., [2007a]) to investigate site specific processes responsible for creating individual arsenic hot-spots. In addition to controls of groundwater flow, the evolution of dissolved arsenic throughout the BDP is known to be closely related to Quaternary stratigraphy (e.g. Nath et al., [2008b]; Weinman et al., [2008]).  As groundwater flow is ultimately constrained by depositional environments, reviewing recent basin evolution will aid in aquifer conceptualization within the delta.  A brief summary of the BDP depositional history, followed by possible associations between stratigraphy and spatial patterns of arsenic are therefore presented in Section 2.3, before examining current hydrogeological ideas and the modeling efforts made towards understanding the arsenic problem.  A detailed review of groundwater hydrology is presented in Section 2.4, which provides context for conceptual and numerical models of arsenic transport and sequestration in the delta groundwaters discussed in Sections 2.5 and 2.6. 2.3. Depositional History of the Bengal Delta Plain The Bengal Basin is drained by the Ganges, Meghna and Brahmaputra/Jamuna rivers and forms the world‟s largest fluvio-deltaic system [Alam et al., 2003; Coleman, 1981].  The current delta is composed of approximately 5x105 km3 of sediment, and has been prograding since the Miocene (i.e. 24 Ma) from a northeast-trending hinge line across West Bengal and Bangladesh, [Goodbred and Kuehl, 2000; Johnson, 1994].  The nature of sedimentation throughout the BDP has been sensitive to sea level fluctuation over geologic time.  Rapid aggradation occurs when sea levels are low from active deposition of fluvial channel sediment and the formation of lobate deltas.  During sea level highs, predominantly fine-grained materials are deposited in large estuaries [Dowling et al., 2003].  The current estimate of the total drainage area of the BDP is approximately 2x106 km2 [Coleman, 1969; Holeman, 1968] which delivers a sediment load of approximately one-billion tons (i.e. 9.1x1011 kg) per year to the Bay of Bengal [Milliman and Syvitski, 1992].  The floodplain area of the delta within India and Bangladesh is estimated to cover 2.5x105 km2 [Dowling et al., 2003], extend to depths greater than 20 km [Shamsuddin et al., 2002] and experiences sedimentation rates of approximately 1.2 mm/yr [Worm et al., 1998]. The BDP aquifers consist of Quaternary sediments that range in age from Pleistocene (1.6 Ma - 10 ka) to Holocene (10 ka - present).  These sediments were deposited via fluvial processes beginning at the onset of the last global glacial maximum, when the regional base sea level was approximately 100m below present mean sea level (msl) [Islam and Tooley, 1999; Morgan and McIntire, 1959].  During this period, the proto-Meghna and Ganges-Brahmaputra rivers incised 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 10 deeply into the existing Pleistocene sediments, depositing coarse-grained materials at river bases.  The Pleistocene sediments themselves consist of sands and silts occurring in characteristically fining-up sequences that result from fluvial-deltaic deposition and are typically found at depths >100 m below ground level (bgl).  As a result of the thick unsaturated zone and steep hydraulic gradients that persisted at the time, the existing Pleistocene sediments were also exposed to extensive sub-aerial weathering and groundwater flushing [Kinniburgh et al., 2003].  Consequently, Pleistocene materials commonly exhibit a distinctive brown, or yellow- orange hue [Morgan and McIntire, 1959] from an iron oxide coat (e.g. Pal and Mukherjee, [2009]), indicative of sub-aerial weathering and the oxic conditions under which they were deposited. An oxidized clay unit caps much of the Pleistocene sands in areas throughout the Basin [Goodbred and Kuehl, 2000].  The recent work of McArthur et al., [2008] suggests that the fine- grained unit is widely continuous throughout the Bengal Basin due to a glacioeustatic-driven sea level fall during the Pleistocene.  McArthur and colleagues [2004] suggest that the clay formed in paleointerfluvial areas, and should be pervasive throughout the delta, with the exception of areas affected by erosion or an absence of depositional processes.  This inferred clay cap continuity is identified as a paleosol horizon unit, which marks the Last Glacial Maximum and hence named the Last Glacial Maximum Paleosol (LGMP) [McArthur et al., 2008].  In southern West Bengal the LGMP occurs ~ 21-24 m bgl [McArthur et al., 2008] and has been observed at depths of 16-40m elsewhere in the south-central Bengal Basin [Goodbred and Kuehl, 2000]. The present day Ganges-Meghna-Brahmaputra delta began to prograde at the onset of the Holocene (~11-10 ka) into the Bay of Bengal [Goodbred and Kuehl, 2000].  As deltaic advance was met with rapid glacio-eustatic marine transgression from 10-7 ka, deposition of finer grained sediments occurred to the south [Goodbred and Kuehl, 2000; Islam and Tooley, 1999]. During this time, the lower parts of the proto-Meghna and Ganges-Brahmaputra developed into marshes and lagoons that deposited silt and mud units rich in organic matter.  These deposits were also interbedded with sand lenses from small tributaries.  High sedimentation rates concurrent with elevated temperatures and rapid sea level rise (i.e. from 7 - 5.5 ka) facilitated the development of tidal mangrove and paludal basins in the southern regions of the delta and created ideal conditions for peat accumulation.  If sea level estimates are correct and delta progradation occurred simultaneously, depths of paludal deposits are expected to increase inland [Ravenscroft et al., 2001].  The fluvio-deltaic Holocene sediments deposited to the north are dominated by medium to coarse grained sediments resulting from the overlapping of numerous sub-deltas [Morgan and McIntire, 1959].  Rapid avulsion of tributaries and 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 11 distributaries of the Ganges River until the present day has built layers of upward-fining sequences of alternating overbank silts and incised channel sands that extend tens of metres deep [Coleman, 1969; Goodbred and Kuehl, 2000; Nickson et al., 2000]. The sedimentation patterns occurring in the Bengal Basin have resulted in limited lateral continuity of stratigraphy throughout India and Bangladesh.  In fact, large-scale spatial heterogeneity of the BDP has been confirmed by a recently compiled basin-wide driller database [DPHE, 2006], except for a persistent fine grained unit, possibly the aforementioned LGMP.  These sediments are often simply classified in the literature as either Pleistocene or Holocene deposits due to their complex history.  This generalization has been extended to the interpretation of arsenic distribution and generation, since much of the arsenic affected areas occur within the highstand deposits with radiocarbon dates less than 10 ka (e.g. BGS and DPHE, [2001]).  However, it is known that arsenic contamination is not limited to aquifers of Holocene age [BGS and DPHE, 2001; Pal and Mukherjee, 2009; Zheng et al., 2004], and may best be interpreted in terms of sedimentary depositional environments.  This is discussed in the following section. 2.3.1. Groundwater Arsenic Occurrence in Relation to Geology The BDP aquifers have habitually been grouped into two categories according to age, appearance and arsenic geochemistry.  These groups include the brown Pleistocene sands and grey Holocene sands [BGS and DPHE, 2001; Harvey et al., 2006; Harvey et al., 2002; McArthur et al., 2004; van Geen et al., 2003b].  Dissolved arsenic within the brown sands is typically low, whereas higher concentrations are often observed in the groundwaters within the grey sands. The difference can largely be attributed to the adsorption of arsenic to iron (oxy)hydroxides, which are prevalent in the brown Pleistocene sands [McArthur et al., 2004] and generally absent within the grey Holocene sands [Harvey et al., 2002; Nath et al., 2008b; Polizzotto et al., 2006]. However, it has also been suggested that the grey sands contain sparsely distributed iron oxides, as a result of rapid transport of immature sediments during the Holocene period [McArthur et al., 2004].  Indeed, abundant muscovite and biotite, along with a lack of fine- grained material found in the Holocene sediments support the theory that the intensity of chemical weathering at the time of deposition was low compared to physical weathering [McArthur et al., 2004] and effectively limits the formation of iron (oxy) hydroxide precipitates. This sparse distribution of oxide precipitates, coupled with the current reducing conditions in the associated groundwaters of the grey sands, promotes the observed higher arsenic concentrations [Harvey et al., 2002].  However, recent reductive dissolution of iron (oxy) 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 12 hydroxides within Holocene sediments may be partly responsible for the high dissolved arsenic concentrations. The reductive dissolution can be attributed to redox cycling caused by seasonal water table fluctuation, which would imply that the groundwater system is currently in a state where the arsenic has not yet been flushed [Harvey et al., 2006]. Despite the customary association of arsenic enrichment with Holocene aquifers, it is important to note that contamination is not necessarily constrained by sediment age.  For example, [Zheng et al., 2005] noted that although arsenic concentrations in deeper aquifers are often lower than what is observed in shallow aquifers, the sediments can be of either Pleistocene and Holocene origin.  Furthermore, [Pal and Mukherjee, 2009] observed that specific aquifer characteristics in West Bengal determine regions of arsenic contamination rather than aquifer age.  Their studies suggest that affected areas typically occurred in aquifers that are leaky to semi-confined, and capped by a grey, soft clay unit rich in organics.  Materials of both Pleistocene and Holocene origin comprised these affected aquifers in their study.  It is thus more precise to note that arsenic contamination is ubiquitous in shallow aquifers, which generally consist of, but are not restricted to, Holocene materials. This observation is consistent with research by Weinman et al., [2008], who suggest that patterns of dissolved arsenic in shallow aquifers are best interpreted through reconstruction of local depositional environments.  They stress that local aquifer depositional history is especially relevant because village locations, and therefore exposure resulting from tube-well emplacement in contaminated areas, are closely linked with elevated surface landforms (e.g. former levees and point bars).  From their 25-km2 systematic study of 200 shallow tubewells in Araihazar, Bangladesh, Weinman et al., [2008] revealed that elevated geomorphologic features form under high-energy conditions (e.g. levees and point bars) and consist of permeable sands, These features are typically associated with low levels of groundwater arsenic, presumably as a result of enhanced recharge.  Conversely, deposits that form under low-energy meandering conditions tend to be lower in elevation, comprised of fine grained sediment in the upper few metres of stratigraphy (e.g. scroll bars) and are associated with high groundwater arsenic concentrations.  Comparatively lower elevations of these deposit types combined with their typically fine-grained fabric favour lower recharge rates and hence less flushing of arsenic from these areas. The preceding discussion illustrates how age, depositional history and geomorphology play important roles in creating the arsenic distribution observed in BDP groundwaters.  The importance of variable recharge as an agent for arsenic transport through the subsurface 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 13 warrants a closer review of hydrogeology.  Accordingly, review of the current understanding of hydrogeology of the Bengal Basin is presented in the following section. 2.4. Hydrogeology of the Bengal Delta Plain The depositional history of the BDP that has created complicated inter-layering of coarse and fine grained sediments throughout the delta has generated an equally complex regime for groundwater flow [Dowling et al., 2003].  Ravenscroft et al., [2005] propose that groundwater flow in the BDP occurs on three distinct but simultaneously operating scales.  These include local, or village-scale flow systems that discharge groundwater several kilometres from recharge zones, intermediate scale flow systems (~10km flow paths) and a basinal-scale flow system (>102 km lateral flowpath).  Localized „flow-cells‟ generated by widespread irrigation may be a significant process disrupting local-scale flow [Harvey et al., 2006; Ravenscroft et al., 2005] by accelerating shallow groundwater cycling, and removing controls of local geomorphologic features (i.e. floodplains, levees, minor rivers) [Harvey, 2002].  Intermediate and basin-wide flow systems on the other hand likely remain undisturbed by anthropogenic activities and experience residence times of 102-103 and >104 years, respectively [Ravenscroft et al., 2005], compared to residence times on the order of decades estimated for systems dominated by flow-cells [Harvey et al., 2006]. In general, the BDP aquifers are highly productive, and the water table is usually located within 15 m bgl (Mukherjee et al., [2007a] and references therein).  Extensive irrigation practices combined with the monsoonal climate of the region are the principal controls on groundwater levels.  Groundwater elevations are drawn down with the progression of irrigation season each year, and tend to return to aquifer-full levels after the monsoon, suggesting considerable meteoric recharge, or excess recharge capacity [BGS and DPHE, 2001].  Major rivers and ponds may also contribute significantly to aquifer recharge, depending on the season [Harvey et al., 2006], and flow system scale. A review of the literature demonstrates that hydrogeological and groundwater modeling studies within the BDP have been conducted on various scales, with the first extensive investigation performed by the British Geological Survey [BGS and DPHE, 2001].  Major contributing hydrogeological studies that have been conducted on intermediate and village scales in Bangladesh include work by Ravenscroft et al., [2005], Harvey et al., [2005] and [2006], Stute et al, [2007] and Neumann et a [2010].  Hydrogeological investigations that have extended to West Bengal include studies on both basinal scales (~104 km2) [A Mukherjee et al., 2007a] 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 14 intermediate scales (~10 km2) [Nath et al., 2008b; Nath et al., 2005], as well as on the village scale ([R Beckie et al., 2006] (i.e. part of GSC-GSI-UBC project investigations)).  The general lack of aquifer continuity beyond kilometre scales has necessitated the application of simplified, or upscaled flow system interpretations in several groundwater modeling studies (e.g. ~21,000 km2 in [A Mukherjee et al., 2007a]; ~300,000 km2 [Michael and Voss, 2009a; b]).  Although detailed village-scale modeling studies have also been conducted [Ashfaque, 2007; Neumann et al., 2010], they are generally underrepresented in the literature, despite known connections between arsenic distribution and local-scale flow processes (e.g. Neumann et al., [2010]).  The following section reviews the delta hydrogeology in detail, in terms of aquifer parameters and fabric, recharge and scale dependent flow.  Where possible, emphasis will be placed on hydrogeological research conducted in the Nadia District of West Bengal (Chakdaha Municipality), as the Gotra research site is located within this region. 2.4.1. Aquifer Properties A number of published accounts of regional aquifer properties have been summarized by BGS and DPHE [2001] and indicate relationships between age, grain size and degree of weathering. Regional hydrogeologic properties may also vary as a result of basin development and physiographic location.  For example, sediment deposited in areas of the delta that are fluvially dominated will likely be coarser grained than in floodplain regions [Goodbred et al., 2003]. Other researchers have proposed that the more recently deposited grey Holocene sediments generally have higher conductivity values than brown-red sands deposited in the Pleistocene (e.g., BGS and DPHE, [2001]; Horneman et al., [2004]; Ravenscroft et al., [2005]; von Bromssen et al., [2007]).  Hydraulic conductivity values summarized by BGS and DPHE [2001] for grey Holocene sands range from 4.6×10−6 to 1.2×10−3 m/s, whereas values of the Pleistocene brown-red sands are estimated between 2.3×10−6 to 5.8×10−4 m/s.  Floodplain materials in the BGS and DPHE [2001] review were also documented to have higher characteristic hydraulic conductivity values than materials deposited in terraces.  Storage coefficients of sands (i.e. storativity values) are generally less than 10-3, and specific yield values range from ~0.05 to 0.3 showing a positive relationship with increasing grain size [BGS and DPHE, 2001].  Typical specific yield estimates for shallow, fine-grained units are ~0.02 to 0.03; whereas deeper (possibly Pleistocene) aquitards have values that are lower [Ravenscroft et al., 2005].  Aquifer porosity ranges between 0.05 and 0.2, and is estimated at ~0.1 in the lower Ganges area (i.e. near the study site) [BGS and DPHE, 2001].  Porosity is assumed to increase with depth, as sediments progressively become coarser [BGS and DPHE, 2001]. 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 15 Regional scale (i.e. equivalent) hydraulic properties of the BDP sediments were recently estimated by Michael and Voss, [2009a] using both statistical and numerical methods.  Good model fits were obtained when specific storage was fixed at 9.4×10-5 m-1; specific yield, horizontal (Kh) and vertical (Kv) hydraulic conductivities were automatically estimated at 0.11, 5.2×10−4 m/s, and 1.4×10−8 m/s, respectively.  The Kh and Kv values quoted for the Central Floodplain region in West Bengal (i.e. closest to Gotra) are 1.9×10−4 and 1.8×10−8 m/s, respectively.  Best-guess values estimated from this study for the large-scale equivalent properties of the entire system are Kh=5×10 −4 m/s and Kv=5×10 −8 m/s, (i.e. Kh/Kv=10,000).  It is worth noting that this high anisotropy is an artifact of extreme upscaling necessary to represent stratified sequences of aquifers and aquitards that comprise the basin regionally. 2.4.2. Hydrostratigraphy and Aquifer Fabric Permeable sediments in the BDP extend to depths ranging from 100 to 3000m below the ground surface, where they either overlie the Precambrian basement or a consistent basal shale layer [Michael and Voss, 2009a].  In Nadia District, this basal unit is thought to exist at 170 to 200m below ground surface [A Mukherjee et al., 2007a].  The lateral discontinuity of aquifers and confining units resulting from dynamic fluvial deposition have produced a highly heterogeneous aquifer framework, which is not correlatable beyond ~20km scales [Michael and Voss, 2009b].  Despite the lack of regional continuity, pumping tests throughout Bangladesh indicate that the top ~200m of sediment responds like a hydraulically connected, layered aquifer system [MPO, 1987].  In fact, several researchers have considered the sediments of the BDP to behave regionally as a single aquifer [BGS and DPHE, 2001; Michael and Voss, 2009a; b; A Mukherjee et al., 2007a] and have found that deep and regional behaviour may best be conceptualized as homogeneous and highly anisotropic [Michael and Voss, 2009a; b].  Shallow and local scale flows, on the other hand are better understood in light of detailed stratigraphic descriptions and aquifer fabric.  For example, statistical analyses of driller log data by Michael and Voss, [2009a] reveal that effective aquifer anisotropy and hydraulic conductivities are highly sensitive to horizontal length scales less than 1km.  In the western part of the BDP, village-scale stratigraphic studies reveal the existence of multiple stacked aquifers that are often separated from one another by clay or silt units from underlying depositional cycles [Pal and Mukherjee, 2009]. Although references to shallow versus deep aquifer systems are common in the literature, there is no concrete distinction between these systems in terms of physical or hydraulic boundaries. In fact, shallow and deep systems are likely to be connected in some areas due to cross-cutting 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 16 of depositional sequences, which could be relevant on the village scale [A Mukherjee et al., 2007a].  In general, „shallow‟ (<100m) aquifers are described as unconfined, whereas deeper aquifers often exhibit semi-confined behaviour [Nath et al., 2008a].  Overall, aquifer tests conducted in Bangladesh tend to show leaky-confined responses in the short term, but over long periods of time (i.e. weeks to months), may exhibit unconfined or semi-confined behaviour [BGS and DPHE, 2001; Ravenscroft et al., 2005].  This is consistent with the conceptual model of regionally disconnected, stacked fluvial deposits, where silt / clay layers may act as local, but not regionally confining units. 2.4.3. Aquifer Flooding and Meteoric Recharge The annual hydrologic cycle in the BDP can be divided into distinct wet and dry seasons, which span from June to October and November to May, respectively.  Groundwater flow is heavily influenced by monsoon rainfall, which contributes to extensive flooding during months of intense precipitation [Allison, 1998; Harvey et al., 2006; A Mukherjee et al., 2007a; Stute et al., 2007]. Approximately 30% the land surface (mostly crop-lands) floods during the wet season [JICA, 2002], which is the combined result of increased river flow from melt water in the Himalayas and tidal increase in addition to monsoon rain [BGS and DPHE, 2001].  Seasonal flooding does not drive significant groundwater flow on any scale because lateral gradients are suppressed when the ground is inundated [Harvey, 2002].  Vertical flow induced by surface ponding is thought to be negligible under these circumstances [Harvey et al., 2006], although shallow downward hydraulic gradients have been noted following major precipitation events [Stute et al., 2007]. Rain during the monsoon does however contribute to large increases in river stage elevations as well as aquifer heads [Harvey et al., 2006; Harvey et al., 2002; Stute et al., 2007], which can fluctuate 4-6m in the absence of irrigation pumping [Ravenscroft et al., 2005].  These variations may cause lateral flux reversals between aquifers and surface water bodies prior to and following the rainy season [Harvey et al., 2006; Nath et al., 2008a].  Post-monsoon groundwater levels generally lag river levels until the onset of the following wet-season [Stute et al., 2007] and regional horizontal gradients that materialize are largely determined by the mild relief throughout the region [Harvey, 2002]. The mean annual precipitation in West Bengal ranges from 1295 to 3945 mm/yr (Nath et al., [2008a] and references therein) and falls predominantly during the monsoon season (i.e. June to October).  Less than 5% of mean annual rainfall occurs from November to March, and 10% falls from April to May [BGS and DPHE, 2001].  Throughout the entire delta, annual evapotranspiration generally exceeds precipitation [Allison, 1998].  Meteoric recharge is 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 17 estimated at ~15% of total precipitation [SWID, 1998]; and also disproportionally occurs during the monsoon [A Mukherjee et al., 2007a], as aquifer-full conditions develop in the majority of the Bengal Basin [BGS and DPHE, 2001; Burgess et al., 2002; Michael and Voss, 2009b; Ravenscroft et al., 2005].  Accordingly, surface runoff varies as the monsoon season progresses [BGS and DPHE, 2001], and is also a function of catchment area, ground surface slope and soil type [Dingman, 2002].  Annual runoff estimates for catchments in Eastern India range from 130 to 3100 mm [Kothyari and Garde, 1991].  Monthly runoff estimates for the River Mayurakshi catchment (i.e. ~250km northeast of the Gotra site in Chakdaha Municipality) vary from 40 to 140 mm between July and October, with peak values occurring in September [Kothyari, 1995]. 2.4.4. Regional Flow and Delta Flushing The gentle topography of the BDP implies that regional groundwater flow is driven by very shallow gradients from north to south.  In the absence of any applied stress, the overall southward gradient should only allow for local variations near major rivers.  In fact, regional gradients are low, ranging from 0.001 in the north to 0.00001 in the south [BGS and DPHE, 2001] and are likely to influence groundwater flow only at depths of 100m and greater [Michael and Voss, 2009b].  The modeling study of [A Mukherjee et al., 2007a] suggests that prior to the onset of widespread irrigation in the 1970s, regional gradients dominated shallower flow systems as well.  Present day hydraulic gradients in the Ganges-area are estimated at approximately 0.00008, whereas early Holocene gradients are believed to have been slightly higher (0.00028, BGS and DHPE, 2001). The regional seepage velocity in the Ganges area under current gradients is on the order of ~103 m/d, which translates to an approximately 12 ka for one pore-volume to flush through the system [BGS and DPHE, 2001].  This is important because it suggests that Holocene aquifers, which are often classified as arsenic-affected zones, have been subjected to relatively little flushing since deposition; some shallow sediments are even less than 5 ka [Smedley and Kinniburgh, 2002].  Furthermore, since the reaction times of geochemical and microbial processes releasing arsenic into groundwater are comparatively instantaneous, there is sufficient time for arsenic to accumulate in younger sediments.  Pleistocene aquifers on the other hand were deposited prior to 10 ka and consequentially have experienced more extensive flushing since deposition, especially under steeper (i.e. approximately 10 times) historical hydraulic gradients. 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 18 Submarine groundwater discharge (SGD) is likely the most significant regional flux of groundwater to surface in the BDP, since deeper groundwater has little interaction with surface water bodies [A Mukherjee et al., 2007b].  Mukherjee et al. [2007b] estimate total annual SGD of 5.9x107 m3 through a 110km front (i.e. 20%) of the delta to the Bay of Bengal, representing a 4% decrease since pumping began in the 1970s.  Similarly, Michael and Voss, [Michael and Voss, 2009b] estimate a value of SGD to the Bay of Bengal of approximately 108 m3/yr for pre- development conditions.  Both estimates assume flux to the Bay is the result of deeper flow driven by regional gradients, since local flow systems dominate at shallow depths.  Other mass balance calculations have estimated SGD to be approximately 3 orders of magnitude higher (e.g. Basu et al., [2001]; Dowling et al., [2003]), although such a flux would require unrealistic artesian conditions 5 km inland, which do not exist [Harvey, 2002]. 2.4.5. Intermediate and Local Scale Flow Although SGD constitutes a considerable volume of groundwater exiting the BDP each year, most significant groundwater-surface water interactions generally occur within local flow systems [A Mukherjee et al., 2007a; A Mukherjee et al., 2007b].  The regional modeling study of Mukherjee et al., [2007a] suggests that recharge and discharge at depths greater than 200m are minimal.  In addition, intermediate and local scale flow systems have become increasingly important with the advent of irrigation pumping in the past few decades [Harvey, 2002], with potential implications for patterns of dissolved arsenic [Harvey et al., 2006]. In shallow flow systems within the delta, groundwater flows under the influence of water table gradients are limited to short lateral distances, ranging from metre to tens of kilometre scales [Michael and Voss, 2009b].  As with precipitation, the contribution to groundwater recharge from rivers, lakes and ponds exhibits a strong seasonality [Harvey et al., 2006].  During dry periods in particular, major rivers, lakes, small distributaries and ponds will contribute considerably to kilometre-scale groundwater flow systems [Michael and Voss, 2009b]. Local rivers and streams have been identified as significant conduits for both recharge and discharge to and from aquifers depending on the season [Harvey et al., 2006; A Mukherjee et al., 2007a; Nath et al., 2008b; Ravenscroft et al., 2005; Stute et al., 2007], and therefore complicate shallow hydrology.  For example, the Bhagirathi-Hooghly River, which runs through central West Bengal and west of the Gotra research area, recharges the nearby shallow aquifers prior to the monsoon [A R Ghosh and Mukherjee, 2002] and also functions as a discharge zone following the monsoon period [Nath et al., 2008a].  Another example is the 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 19 Ichamati River (east of the Gotra research site), which exhibits similar flux reversals in response to seasonal precipitation and irrigation pumping [Harvey et al., 2006]. In the case of the Ichamati, the aquifer drains into the river during and immediately following the monsoon rain, whereas immediately prior, cumulative agricultural abstraction facilitates a drop in groundwater level below the river stage and hence recharge to the aquifer [Ashfaque, 2007].  For example, modeling studies suggest that, during the wet season, the specific discharge from aquifers to rivers is on the order of 1.5 m/day. At the height of the irrigation season, the groundwater flow direction is reversed and there is a net specific recharge of 2m/day from the rivers to the aquifers.  However, this study also suggests that dual behaviour of the Ichamati does not predate the explosive development of irrigation systems beginning in the 1970s, and that shallow aquifers in Bangladesh were earlier drained year-round by the rivers. Excavated ponds are common in rural areas throughout Bangladesh and India and have been hypothesized by some researchers to provide shallow aquifers with a significant source of recharge [Ashfaque, 2007; Harvey et al., 2006; Neumann et al., 2010].  The majority of these ponds are believed to have been dug within the past 50 years in order to provide clay/silt material for village construction over monsoon flooding levels [Harvey et al., 2006].  Remnant ponds from oxbow lakes are also ubiquitous ([Desbarats et al., 2009], i.e. the Gotra site), and as such have been present in the delta long before village construction and agricultural development.  Ponds generally extend several metres deep, are often lined with low permeability clay bottoms, and tend to have higher water levels than groundwater during the dry season [Ashfaque, 2007; Harvey et al., 2006], or may even be perched above the water table [McArthur et al., 2004].  During monsoon season, ponds may be submersed by floodwaters and become hydraulically connected to the shallow groundwater system [Ashfaque, 2007].  Harvey et al., [2006] estimated the contribution of pond water to surrounding aquifers during the irrigation season to be approximately 1 cm/day, with older ponds contributing less to recharge as a result of sediment settling on the bottom over time. While rivers and their distributaries, lakes and ponds are important features for local scale groundwater flow, irrigation pumping during the dry season has also been shown to drive a significant amount of groundwater flow in the BDP [Harvey et al., 2002; 2005].  Vigorous pumping from irrigation tubewells creates spatially distinct sinks that not only influence head distribution, but alter recharge sources, and decrease average groundwater residence times in agricultural areas [Harvey et al., 2006].  These effects are discussed in the next section. 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 20 2.4.6. Irrigation Coverage and Localized Flow Cells Irrigable lands cover up to 50% of land surface in many regions of the BDP, with an average of ~25% of total district area [WARPO, 2000].  Estimates of irrigated land near the Gotra site in Chakdaha, West Bengal, have doubled from 1991 to 2001 and have led to groundwater usages amounting to ~70-80% of annual recharge values.  For comparison, current estimates of irrigation usage for the entire BDP are slightly lower, ranging between approximately 40-60% of annual recharge values [Nath et al., 2008a].  Throughout most of the basin, the amount of water extracted from aquifers for irrigation significantly exceeds the quantity pumped for domestic and industrial purposes [Michael and Voss, 2009b].  In 4 districts of West Bengal, total annual irrigation abstraction is estimated at 8.7x109 m3/yr (i.e. 414 mm/yr), 89% of which is pumped from shallow (~10-70 m bgl) wells, and the remaining 11% from deep (~60-180 m bgl) wells [A Mukherjee et al., 2007a].  Shallow tube wells usually are privately-owned, have low to medium yields (10-20 m3/h) and are operated by diesel-powered centrifugal or submersible pumps.  The deeper wells tend to be government-owned, are operated by electric turbine or submersible pumps, and have yields of ~75 to 150 m3/h [A Mukherjee et al., 2007a]. Irrigation pumping has significantly impacted shallow groundwater flow paths by inducing strong vertical gradients that can overwhelm any existing lateral flow [Harvey et al., 2002].  Isolated groundwater sinks resulting from widespread pumping have created self-contained, localized flow-cells that cycle shallow groundwater throughout the delta.  Steep vertical gradients, combined with re-infiltrated water (i.e. irrigation return-flow) have been shown to decrease vertical travel times compared to those preceding the advent of crop irrigation [Harvey et al., 2006; Harvey et al., 2002].  In addition, pumping has been shown to have substantially shifted local and regional water balances from pre-development scenarios, by reducing natural discharge to (or increasing recharge from) rivers, increasing recharge from ponds and lakes, and increasing salt water intrusion near the coast [Harvey et al., 2006; Michael and Voss, 2009b; A Mukherjee et al., 2007a].  Re-infiltrated irrigation water (return flow) has also been shown to contribute considerably to aquifer recharge [Harvey et al., 2006], and pumping from deep wells may increase downward flow of shallow groundwater to deeper aquifers [Michael and Voss, 2009b; A Mukherjee et al., 2007a]. In response to the anticipated population growth of West Bengal [Gupta, 2006], it is likely that the impact of pumping on groundwater budgets, flow paths and residence times will continue to grow, and disrupt flow systems at increasingly larger scales (e.g. Mukherjee et al., [2007a]). The implications of this dynamic water balance for groundwater arsenic distribution and 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 21 transport, along with additional hydrogeological controls, are therefore discussed in the following sections. 2.5. Implications of BDP Hydrogeology for Arsenic Distribution and Transport The preceding discussions highlight the complexity of groundwater flow in the BDP.  It follows that advective transport of dissolved arsenic should be equally complex, particularly in light of perturbations to water balances in recent decades.  Irrigation pumping has likely induced chemical and physical disequilibria in groundwater systems, which is manifested by the patchy occurrence of groundwater arsenic.  More specifically, contaminant distribution of groundwater is likely the result of wide-spread (but locally distributed) three dimensional transient flow systems that have altered natural recharge and discharge zones [Harvey et al., 2006; Harvey et al., 2002].  These anthropogenic disturbances have made the determination of baseline hydrogeological controls on contaminant distribution from current conditions in the BDP difficult. However, hydrogeologic studies in the relatively undisturbed (but arsenic-contaminated) Red River and Mekong Deltas (e.g. Berg et al., [2001]; [2007]; Benner et al., [2008]; Polizzotto et al., [2008]) in Cambodia and Vietnam may serve as pre-disturbance analogues for India and Bangladesh.  These deltas exhibit similarities in geologic deposition and aquifer sediment source, as well as regional hydraulic gradients [Berg et al., 2001; Berg et al., 2007]. In the following section, a review of the central concerns regarding arsenic release and persistence in the BDP is presented in the context of system hydrogeology.  This involves consideration of groundwater recharge and discharge, as well as the potential alteration of hydraulic regimes as a result of irrigation practices.  As baseline studies in the BDP are lacking, the geochemical and hydrogeological similarities of groundwater arsenic in the Southeast Asian deltas will be drawn upon to exemplify the possible implications of irrigation practices on arsenic mobilization and transport where applicable.  Primary issues to be discussed include geological and geomorphological controls, sources of biologically available organic carbon to contaminated areas, as well as the susceptibility of clean aquifers to contamination in the future. 2.5.1. Geological and Geomorphological Manifestations An explanation for the local heterogeneity of dissolved arsenic in shallow aquifers may be that recharge and flow patterns are constrained by geomorphology and shallow stratigraphy [Nath et al., 2008a; Weinman et al., 2008].  For example, it is commonly stated that arsenic accumulation in areas overlain by, or within, ~10m of thick surficial silt and clay deposits is a 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 22 result of increased average residence times, or restricted recharge due to lower hydraulic conductivities of the fine grained material [Nath et al., 2008a; Stute et al., 2007; van Geen et al., 2006a].  Similarly, it is common to observe lower arsenic concentrations in association with proximal coarse-grained material that may susceptible to locally enhanced flushing rates due to higher relative permeability [Stute et al., 2007; van Geen et al., 2006a].  In the case that these higher conductivity materials also occur as elevated geomorphic features (i.e. levees and bars), resultant potentiometric surfaces will tend to be higher than in adjacent zones, effectively preventing flow of contaminated water into the area [Aziz et al., 2008; Weinman et al., 2008]. Permeable, elevated areas naturally comprise an ideal setting for enhanced groundwater recharge [Hubbert, 1940].  Such a geological configuration may conceivably reduce arsenic contamination through dilution [Aziz et al., 2008; Nath et al., 2008a], and the inhibition redox processes via infiltrating oxic rainwater [Aziz et al., 2008; Wang and Mulligan, 2006].  On the other hand, recharge through thick, fine-grained sediment sequences (e.g. channel-fill silts) will be slower [Weinman et al., 2008] and allow for prolonged water-rock interactions [Nath et al., 2008a]. These conditions may act to facilitate the persistence of reducing conditions, which result in arsenic liberation from sediments. The relationships of dissolved arsenic with stratigraphy imply that erratic spatial distributions of the contaminant may well be explained through reconstructions of local depositional history. However, it is worth noting that the aforementioned associations have been documented in areas where irrigation is prevalent, and therefore cannot be explicitly distinguished from this impact.  For a baseline comparison, the Cambodian Mekong Delta reveals that similar geologic and geomorphologic features are important contributors to dissolved arsenic patterns [Benner et al., 2008; Kocar et al., 2008].  Additionally, adequate flushing of this delta since deposition (i.e. 3-30 pore volumes over 6000 years), suggests that an upstream source of the contaminant may exist [Polizzotto et al., 2008].  However, observed steep vertical gradients between surface and ground water that coincide with chemical gradients and biogeochemical signatures below near- surface features suggest that arsenic release is hydrology-driven [Polizzotto et al., 2008], therefore arsenic mobility will be sensitive to irrigation disturbances.  Two examples of alteration to shallow aquifer redox conditions (and hence arsenic release) as a result of irrigation pumping are the creation of stagnant zones that may allow prolonged groundwater interaction with buried organic matter, and the mobilization of organic carbon to depth from distant sources.  Both scenarios are discussed in the following section. 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 23 2.5.2. Source of Organic Substrate Dissolved organic carbon (DOC) that fuels microbial metabolism for arsenic release in the BDP may have leached from aquifer sediments or have been mobilized and brought to depth from surface sources.  It is clear that irrigation pumping will influence these processes [Harvey et al., 2002]; however, the extent of the impact at individual sites remains a subject of controversy (e.g. van Geen et al., [2003a]; Aggarwal et al., [2003]; Harvey et al., [2003]). Possible sedimentary sources of DOC include buried peat or detrital carbon [McArthur et al., 2001; Nickson et al., 2000; Ravenscroft et al., 2001], or other organic material co-deposited with aquifer sediments [Meharg et al., 2006].  In order for buried peat to drive redox processes, slow flow or stagnant conditions are required to promote prolonged sediment-water interactions [McArthur et al., 2001; McArthur et al., 2004], which is conceivable in an extensively pumped environment [van Geen et al., 2003b].  However, studies by Pal et al., [2002b], which document occurrences of high dissolved arsenic in aquifers devoid of peat, as well as arsenic-free zones in areas in which peat is abundant, suggest that peat may not be a significant redox driver. Additionally, radioactive carbon signatures of microbial by-products within groundwater measured by Harvey and colleagues [2002]; [2006] support a more recent source of DOC than detrital carbon driving biogeochemical reactions.  Naturally slow residence times within thick accumulations of near-surface sediments rich in organics (i.e. former oxbow lakes or channel bars) may also create reducing waters and concomitant arsenic enriched groundwater (e.g. Polizzotto et al., [2008]; Weinman et al., [2008]). Labile DOC introduced into aquifers with groundwater recharge, and originating from surface waters (e.g. rice paddy fields or ponds), may also create the required reducing conditions for arsenic release (Harvey et al., 2002, Neumann et al., 2010).  This process may occur in minimally disturbed areas (e.g. Polizzotto et al., [2008]) or intensely pumped regions [Harvey et al., 2002], differing mainly in the time span required to transport DOC to sediments containing sorbed arsenic.  In an affected aquifer within the Cambodian Mekong delta, arsenic is released and transported on centennial time scales [Polizzotto et al., 2008], whereas carbon signatures at a heavily irrigated site in Bangladesh indicate that pumping has facilitated release from more recently (i.e. <50 year-old) recharged waters [Harvey et al., 2002]. The idea that enhanced infiltration of reducing waters from near surface sources to depth due to irrigation has been met with much scepticism [Aggarwal et al., 2003; van Geen et al., 2003a].  In addition, irrigation facilitated carbon transport also does not rule out hypotheses where 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 24 indigenous sediments provide DOC for microbial metabolism.  However, it is important to note that the documented effects of pumping on solute concentrations are largely site-specific (e.g. Harvey et al., [2003]).  Realistically, sedimentological features, redox status as well as the availability of organic matter to drive redox reactions each may have a significant impact on arsenic migration and sequestration within the BDP aquifers.  Revealing the true cause for arsenic enrichment requires site-specific consideration of these effects. 2.5.3. Downward Migration of Arsenic to Deep Aquifers It is known that the majority of water used for drinking and irrigation is pumped from shallow (<100m depth) aquifers.  In addition, dissolved arsenic tends to be more prevalent in shallow groundwaters.  Deeper aquifers are often low in arsenic, perhaps as a result of increased oxidized mineral surface area to sequester the contaminant, or even potential physical barriers to vertical infiltration [McArthur et al., 2008].  These observations make tapping deeper aquifers appear to be a logical and simple solution to the current arsenic problem.  However, modeling studies of regional flow have suggested that extraction of large volumes of groundwater from deep aquifers may result in significant downward migration of the contaminant [Michael and Voss, 2009b; A Mukherjee et al., 2007a].  The development of deeper resources therefore must be exercised with caution, for example, by limiting extraction of deep groundwater to non- irrigation needs. The numerous factors dictating how hydraulic regimes may affect arsenic concentrations speak to the urgency of developing hydrologic models to guide future management of groundwater resources.  In particular, site-specific models are necessary to direct mitigation efforts, as unique aspects of geology, geochemistry and hydrology conspire to mobilize and sequester arsenic at local scales across the BDP.  The following section reviews previous efforts that have been made to quantify flow in the Ganges delta through numerical modeling. 2.6. Previous Groundwater Modeling in the BDP and the Gotra Research Site To date, the majority of groundwater modeling studies in the BDP have been conducted in Bangladesh.  Most of these models have focused on local-scale flow, and have considered the effects of pumping on arsenic migration [BGS and DPHE, 2001; Cuthbert et al., 2002; JICA, 2002] as well as shallow, dynamic water balances [Ashfaque, 2007; Harvey et al., 2006; Neumann et al., 2010].  Important groundwater modeling studies in arsenic-affected aquifers that extend to the West Bengal portion of the delta have focused on regional flow [Michael and 2.0 Arsenic Contamination of the Bengal Delta Plain Aquifers 25 Voss, 2009a; b; A Mukherjee et al., 2007b], with some exceptions showing how groundwater irrigation has altered flow dynamics and solute transport at the village scale [N C Ghosh et al., 1999; Mukhejee, 2004]. Local-scale models are useful for understanding shallow flow systems, but they cannot effectively simulate large-scale flows at depth.  By the same token, regional models are useful for simulating and characterizing deep groundwater flow, but fall short of quantifying details of shallow systems relevant on the village scale, or flow patterns and arsenic that might affect individual wells.  However, large scale extraction of groundwater from deep aquifers will expand anthropogenic flow regimes, possibly resulting in the downward migration of contaminated groundwater from shallow zones.  In addition, recognition of potentially arsenic-free shallow zones may be accomplished by classifying shallow flow regimes as they are constrained by geology or geomorphology.  Evidently, mitigation efforts for both shallow and deep resources must be informed by a firm understanding of shallow flow systems, which may be achieved through interrogation of well-calibrated shallow groundwater models. In light of the importance of site-specific hydrogeologic features to the broader picture, this thesis proposes to develop a village-scale numerical model to examine controls of dissolved arsenic at the Gotra site.  Particular features of interest are geochemical gradients observed in groundwater over decameter length-scales (both horizontal and vertical), and their possible association with shallow stratigraphy, groundwater flow directions and recharge sources.  If aqueous chemistry can be traced along flow paths, the location at which groundwater picks up its dissolved arsenic load may be determined.  The impacts of irrigation pumping and seasonal variability on site water balance are also considered.  The numerical model has been constructed based on a conceptual model developed from field observations, and characterization methods employed at the site are summarized in the following chapter.  3.0 Site Characterization Methods 26 3.0 SITE CHARACTERIZATION METHODS The Gotra site was characterized over the course three years to gain an understanding of local hydrogeology and hydrogeochemistry.  Site characterization involved three components: establishing the hydrostratigraphic and hydrogeologic context; potentiometric surveys and monitoring; groundwater chemistry surveys and monitoring.  Borehole stratigraphy was used with the results of aquifer tests and material grain sizes to delineate hydrostratigraphic units, and groundwater levels were monitored in boreholes equipped with piezometers from 2006 to 2009.  Meteorological and river stage data were obtained for this time period to help define the local water balance and flow regime.  Major element geochemistry and arsenious zones within the village were delineated through detailed geochemical sampling of domestic wells in 2006 and 2007, and pH and EC were monitored at select locations from 2006-2008.  All geochemistry and water level data were collected from residential wells, irrigation wells and dedicated wells that were emplaced in 2006 and 2007. 3.1. Site Location and Previous Work Gotra is located in Nadia District of West Bengal, about 60km NE of Kolkata, (Figure 3.1).  The sampling region covers an approximate area of 1km2 (centered on 662500E and 2546750N, WGS1984 UTM Zone 45N) and is situated on a natural levee associated with a NW-SE meander scar which is bound in the north and south by several parallel trending ponds (Figure 3.2 and 3.3).  The ponds are likely remnants of an oxbow lake (i.e. abandoned river channel) which formed by the migration of the Ganges-Hooghly-Bhagirathi River system in the Late Holocene.  Site physiography and hydrology are discussed in greater detail in Section 4.1. Previous work by the GSI [P K Mukherjee et al., 2001; Mukhopadhyay et al., 2006; Pal et al., 2002a; Pal et al., 2002b] identified abrupt contrasts in arsenic concentrations over short distances with no appreciable variation in stratigraphy or sediment chemistry in the greater Gotra area (Chakdaha Municipality).  As a primary objective of this study is to understand hydraulic controls of patchy contaminant distribution over short distances, the high density of domestic wells already in place and available for geochemical sampling in Gotra makes the site particularly appealing (Figure 3.2). The Gotra site was initially selected for study in 2003 as part of a joint initiative between the Geological Survey of Canada (GSC), the Geological Survey of India (GSI) and the University of 3.0 Site Characterization Methods 27 British Columbia (UBC) aimed at identifying arsenic release mechanisms into the groundwaters of West Bengal.  The current project supplements the results of a preliminary field investigation in 2004 [R Beckie et al., 2006], with the objective of identifying key hydrogeological constraints that control arsenic distribution in the groundwater at the Gotra site.  Interpretations and results of modeling studies from this project are intended to aid in the development of arsenic risk management strategies. 3.2. Gotra Wells and Piezometers This section describes the wells used for hydrogeochemical investigation of Gotra.  These include domestic wells, crop irrigation wells, dedicated wells and piezometers installed within the village and surrounding agricultural fields.  Preliminary arsenic testing and the rationale for locations of observation well emplacement are also discussed.  Figure 3.2 shows the locations of the wells used in this study. 3.2.1. Domestic and Irrigation Wells For their potable needs, the population of Gotra relies principally on the groundwater from 54 shallow (<150m deep) low-yielding, hand-pumped tubewells (Figure 3.2, Figure 3.3 and Figure 3.4).  Three wells that also are used regularly for household purposes are equipped with motorized pumps (Figure 3.4).  During the dry (boro) season, groundwater is heavily extracted for crop irrigation by wells in the fields that surround the village (Figure 3.2 and Figure 3.5).  The position of all domestic wells in Gotra and irrigation wells that were visible within ~ 1km of the village centre were determined using a handheld GPS unit. Preliminary arsenic concentrations in domestic wells were determined using a portable Merck Kit [Van Geen et al., 2005] during the initial field visit in 2006.  These tests identified a sharp arsenic concentration gradient running SW-NE, roughly orthogonal to the meander scar at a depth of <30m (100ft), (Figure 3.6).  This apparent geomorphologic association is consistent with the observations of the GSI [Sengupta et al., 2004] as well as Weinman et al., [2008], that high arsenic zones typically occur as narrow sinuous strips that are correlated to channel deposits.  This pattern thus instructed the emplacement of observation wells for the current study (Section 3.2.2). 3.0 Site Characterization Methods 28 3.2.2. Observation Wells To investigate the potential association of arsenic release with the Gotra meander scar, observation wells were installed both parallel and perpendicular to the apparent geochemical gradient at screen depths ranging from 20-30m (Figure 3.3 and Appendix A).  Two wells were installed within 1m horizontal separation of one another (GSI0603 and GSI0604) to assess local vertical head gradients in the village (Section 4.3).  Wells GSI0609 and GSI0610 were installed in the rice paddy fields to expand stratigraphic interpretation perpendicular to the meander scar, as well as to observe hydraulic and geochemical properties of the aquifer away from the village centre.  Several wells were also drilled in the Ghetugatchi area (Figure 3.3), and used mainly for geological interpretation.  The names and locations of all observation wells are shown in Figure 3.4. All boreholes were drilled using the indigenous “hand-flapper” method, [Horneman et al., 2004; Rowland et al., 2005], which is practical at the Gotra site because the water table is near the surface, and the basin sands do not collapse easily.  The technique relies on alternating application of hand suction and release to the top end of a flared drill pipe (Figure 3.10), and the use of pond or surface water as a drill fluid.  During drilling, water travels around the exterior of the pipe and fluidizes the sediment at the base of the borehole.  The sediment is subsequently raised by suction and expelled from the top of the pipe when released at the ground surface (Figure 3.7, Figure 3.8 and Figure 3.9). Lithologies and approximate stratigraphic horizons at each borehole were logged according to visual grain size and colour differences (Appendix A) based upon the composition of the slurry of sediment as it was expelled from the drill pipe (Figure 3.8) and drillers‟ reports of material compliance.  Once the desired depths were reached, the drive casing was removed and well assemblies were installed at 13 of the 17 holes (10 in 2006 and 3 in 2007).  From bottom to top, all well assemblies consist of a silt trap, a well screen and a PVC riser pipe (2” ID for 2006 wells; 3” ID for 2007 wells) that is coupled to a 1.5 m metal riser pipe, which protrudes from the ground surface.  7 wells also are equipped with multilevel ports that allow for the collection of depth specific aqueous samples, in addition to those collected at the main screen.  After each well assembly was placed into its borehole, native sand was shoveled into the hole to float along the annulus between the formation and well assembly (Figure 3.14), forming a permeable sand zone around the well screen.  Above the sand pack and well screen, the formation was allowed to collapse back on the riser pipe.  The metal riser pipe was then cemented in place with a small 3.0 Site Characterization Methods 29 (30 cm x 30 cm x 15 cm) concrete pad in order to prevent future well movement, and ensure that reliable water levels measurements can be made. 3.2.3. Small-Diameter Piezometers In 2007 a total of 4 small-diameter piezometers were installed adjacent to existing wells for characterization of vertical hydraulic and chemical gradients.  The piezometers were completed with 1-inch-OD continuous PVC riser pipes connected to Solinst® standpipe piezometers with pointed tips to provide a filtered inlet point for sampling (Figure 3.15).  Three piezometers were emplaced next to GSI0605 at depths of ~6, 11 and 17 metres (20, 37 and 56 feet) below the ground surface to investigate vertical flow through the soft clay (Appendix A).  The forth piezometer was installed next to GSI0606 within the sand at ~11m (35 feet) below ground surface. 3.3. Hydrologic and Hydrostratigraphic Characterization Methods used to characterize site hydrogeology include the collection and interpretation of time- series hydraulic head data at observation wells, the estimation of hydrologic parameters using slug and time-drawdown tests, as well as grain size analyses.  A compilation of hydrometeorological data was also completed to help constrain potential aquifer recharge. These site characteristics are essential for the development of a conceptual hydrogeological model of the site, and to define parameter ranges for calibration of a numerical model.  A brief summary of each activity is described in this section. 3.3.1. Groundwater Levels Automated time series hydraulic head data were collected at observation wells using various instruments from May 2006 through February 2008.  Several instruments were capable of collecting physico-chemical data (i.e. pH, electrical conductivity and dissolved oxygen) in addition to groundwater level.  Corrections for barometric effects were not necessary for instruments with vented cables (i.e. all the in-Situ instruments).  Data collected by all other instruments were corrected for barometric fluctuations using transducers deployed above the piezometric surface.  At wells equipped with instruments that lacked an internal setting for benchmark elevation, a measurement of the depth to water was taken at the beginning of each logging period using a Solinst® Water Level Meter (Figure 3.16).  This depth-to-water measurement was used as a reference from which the actual water table elevation could be 3.0 Site Characterization Methods 30 calculated using surveyed elevations of observation well casings (Appendix A).  All instrumentation specifications, locations of deployment as well as data collection schemes are summarized in detail in Appendix B. 3.3.2. Piezometer Tests Slug tests [Freeze and Cherry, 1979] were performed to obtain estimates of aquifer conductivity at observation wells GSI0604-10 on February 24 and 25 2007.  The slug was constructed from three segments of standard 1" ID PVC pipe joined by couplers and caps fitted on the ends using PVC glue.  Its dimensions when assembled were: Length = 152cm (5ft) Diameter = 2.9cm (1 1/8" OD).  Rising- and falling-head tests were carried out in either duplicate or triplicate, and water table recovery was monitored using an In-Situ® miniTROLL with vented cable (Figure 3.17).  The data collection scheme was logarithmic and continued for a duration of up to 60 seconds.  Complete recovery of the water table was achieved within 30 seconds in all cases. During the 2008 field visit, bail (i.e. rising-head) tests were performed at the piezometers installed adjacent to GSI0605 to estimate the conductivity in the channel-fill silt unit.  Because hydraulic conductivity is significantly lower in this unit than in underlying the sand, automated measurements of water table recovery after purging were not necessary.  Depths-to-water after purging were therefore measured manually using a Solinst® Water Level Meter.  Full recovery in these tests required approximately one day. Analytical methods employed to estimate aquifer parameters from slug test data include solutions developed by Cooper, Bredehoeft and Papadopulos, [1967] (CPB), Hvorslev, [1951] (Hv) and Bouwer and Rice, [1976] (BR).  All three methods were used on test data collected from wells screened in sands, whereas only Hv and BR methods were used for the silts.  A discussion of the assumptions, advantages and short-comings of each of these models may be found in Appendix C, along with raw data and recommended calibration ranges of parameters for numerical modeling. 3.3.3. Pumping Tests and Superposition Modeling Selected periods of time-series hydraulic head data collected in 2007 and 2008 (1 minute sampling intervals) were used as “pumping test” data to examine the response of the aquifer to irrigation well pumping, as well as to estimate hydrologic parameters.  Assessments of aquifer homogeneity as well as the state of aquifer confinement from these tests are discussed in 3.0 Site Characterization Methods 31 Section 4.0.  To estimated aquifer properties, an automated parameter estimation method minimizing residual error was performed using MATLAB, the results of which are compared to values computed using standard semi-log regression methods (Appendix C). To estimate pumping well extraction rates, Q, the time required to fill a 53L bucket (Figure 3.4) with the outflow at various wells within the village and fields was measured.  The bucket was filled three times at each well visited, and the average pumping rates were deduced from the times recorded. 3.3.4. Sediment Sampling for Grain Size Analysis Intact clay samples were collected from overflowing drill pipes as they were advanced into the ground during drilling (Figure 3.8) for grain size analysis at the Geological Survey of Canada (Ottawa).  It was not possible to obtain a representative sample of coarser materials (i.e. sands) because fine-grained constituents were washed away by the drilling fluid (pond water).  Particle separation methods performed on the fine grained samples in the laboratory are described by Girard et al., [2004]. Grain size data were used to estimate hydraulic conductivity of these materials using the Hazen Method [Hazen, 1911], the Breyer Method [Kresic, 1997], as well as USDA computer program ROSETTA (http://www.ussl.ars.usda.gov, [Schaap et al., 2001]).  Calculation details and procedures for making these estimates are summarized in Appendix C. 3.4. Geochemical Data Collection Groundwater samples were collected from the Gotra site for geochemical characterization in both high and low arsenic regions during May of 2006, February and March of 2007 and February of 2008.  Electrical conductivity, pH, dissolved oxygen, alkalinity and dissolved ammonia were measured in the field.  Samples were also sent to laboratories of the GSC for major cation, anion and dissolved organic carbon analyses, as well as the determination of stable isotopic ratios (δ2H and δ18O). Samples were collected from all domestic, irrigation and observation wells within Gotra and the surrounding countryside (Figure 3.2, Figure 3.3 and Figure 3.5).  All wells were first purged by removing approximately 15 L of water using the pre-existing hand pumps found on all the wells before samples were collected (Figure 3.4).  When sampling observation wells during the year 3.0 Site Characterization Methods 32 of emplacement, the wells were purged before sample collection for 1 hour using a submersible pump at the main well screen, and multilevel ports were purged using a peristaltic pump. Irrigation wells were sampled directly from the outflow, since they run consistently, and purging is not necessary.  Detailed descriptions of individual sampling procedures are outlined in the following paragraphs. 3.4.1. Electrical Conductivity, pH and Temperature Measurement Electrical conductivity (μS/cm) was measured at each well using either a HANNA Instruments Portable Multi-Range Conductivity/TDS Meter or a HACH sensION5 Conductivity Meter and corresponding probes.  pH was measured at each sampling well using and an ORION 250A meter connected to a Denver Instruments temperature-compensated probe.  Temperature values were also measured with the pH probe and meter.  Instruments were calibrated daily. 3.4.2. Dissolved Oxygen Dissolved oxygen (DO) was measured at domestic wells either using a Chemetrics model K- 7553 low-range self-filling Oxygen 3 Vacu-vials® Kits, or K-7501 self-filling ampoules (100 to 1400 ppb range).  Water was pumped into a bucket and ampoule tips were snapped at the bottom to minimize exposure of the sample to atmospheric oxygen.  DO in samples collected using the 7553 model were determined quantitatively by measuring absorbance at 555 nm using a HACH 2400 field spectrophotometer; DO in those collected using the other vials were determined using the accompanying DO comparator vials. 3.4.3. Alkalinity Alkalinity was measured in the field within 20 minutes of sample collection by titration of 25 mL of filtered sample with 0.2 N sulfuric acid using a 2 mL Gilmont microburet.  The titration acid had been prepared at UBC, and standardized by back-titration using Na2CO3. The alkalinity of each sample was determined using the Gran plot method [Stumm and Morgan, 1996] where the straight line portion the Gran function vs. mL of acid added to sample was regressed onto a linear equation.  The equivalence point was determined by solving the regression equation for the y-intercept. 3.0 Site Characterization Methods 33 3.4.4. Dissolved Ammonia and Ferrous Iron Dissolved ammonia was measured in the field within 20 minutes of sampling to avoid degassing.  HACH methods 8155 Powder Pillows, 10031 or 10023 Test „N Tube vials were used for ranges 0.01 to 0.50, 0.4 to 50.0 and 0.02 to 2.50 mg/l NH3 – N, respectively.  Ferrous iron was also measured in the field within 20 minutes of sample collection, using the 1,10 Phenanthroline method.  Concentrations in each sample were determined quantitatively by measuring absorbance with a HACH DR2400 spectrophotometer. 3.4.5. Groundwater Collected for Laboratory Analysis Two distinct 60 mL water samples were collected from each well.  Each sample was filtered to 0.45 µm using 25 mm diameter polysulphonate syringe filters.  One sample was preserved to pH < 2 using trace-element-grade nitric acid for cation analysis.  The second sample was collected for anion analysis without preservation.  Major ion, trace metal and anion analyses were performed at the Geological Survey of Canada (GSC) in Ottawa using inductively coupled plasma atomic emission (ICP-AES), inductively coupled plasma mass spectroscopy (ICP-MS) and ion chromatography (IC) methods, respectively.  Dissolved organic carbon (DOC) was determined using a Shimadzu TOC-VCSH (combustion catalytic method) also at the GSC lab. Stable isotopic ratios for deuterium and oxygen (δ2H and δ18O) were determined at the GSC Delta Stable Isotope Geochemistry Laboratory. 3.0 Site Characterization Methods 34  Figure 3.1: General location of field site. Gotra is located in eastern India (A) within Nadia District of West Bengal (B), approximately 60km NNW of Kolkata (C).  (C) shows the field site superimposed upon an SRTM digital elevation model obtained from East View Cartographic services, (90m, 3 arc-second resolution, collected by NASA and NGA in 2004, www.cartographic.com).  The image clearly shows lineaments, floodplains, meander scrolls and abandoned channels which have formed as a result of the active avulsion of the Hooghly-Bhagirathi and Ichamati Rivers and their respective distributaries.  Darker regions represent lower topography, and brighter regions are higher.  Total relief of the map area ~55m. 3.0 Site Characterization Methods 35  Figure 3.2: Google Earth Image showing all wells in the study area. The current project focuses predominantly on the Gotra area, although geochemical and stratigraphic information was also collected from selected locations in Ghetugatchi.  Domestic wells (white circles) are equipped with hand-pumps and are usually screened between depths of 15 and 150m (~50 to 500ft). Observation wells (green circles) were installed in 2006 and 2007 specifically for potentiometric and geochemical monitoring purposes and are screened at 20-30m (~70-80 and 100 ft) depths. Wells and annotations shown over image from Google Earth, (Image © 2008 DigitalGlobe, © 2007 Europa Technologies)  3.0 Site Characterization Methods 36 (a)  (b)  Figure 3.3: Names and locations of domestic and observation wells.  (a) and (b) show Gotra and (c) shows Ghetugatchi.  The majority of domestic wells are equipped with hand pumps (Figure 3.4a), whereas Well 4 and 33/34 use motorized pumps (Figure 3.4). Wells plotted over image from Google Earth, (Image © 2006 DigitalGlobe, © 2006 Europa Technologies)  3.0 Site Characterization Methods 37 (c)   Figure 3.3 (cont.): Names and locations of domestic and observation wells.  (a) and (b) show Gotra and (c) shows Ghetugatchi.  The majority of domestic wells are equipped with hand pumps (Figure 3.4a), whereas Well 4 and 33/34 use motorized pumps (Figure 3.4). Wells plotted over image from Google Earth, (Image © 2006 DigitalGlobe, © 2006 Europa Technologies) 3.0 Site Characterization Methods 38 (a)  (b)  (c)  Figure 3.4: Photographs of domestic wells in Gotra. A typical hand-pumped well is shown in (a).  Wells 4 and 33 are equipped with electrical pumps, shown in (b) and (c), respectively.  3.0 Site Characterization Methods 39  Figure 3.5: Irrigation wells within the village and visible in the surrounding fields. Pink circles indicate shallow wells (<20m deep) whereas white circles indicate deep (~60m deep) wells. Wells plotted over image from Google Earth, (Image © 2006 DigitalGlobe, © 2006 Europa Technologies)   3.0 Site Characterization Methods 40   Figure 3.6: Preliminary assessment of arsenic contamination of Gotra domestic wells. Shallow (<130ft) wells are shown in (a) and deep (>130ft) wells are shown in (b).  A sharp concentration gradient trends perpendicular to the Gotra meander scar at shallow depths, suggesting that As release and/or transport may be associated with the scroll. Wells plotted over image from Google Earth, (Image © 2006 DigitalGlobe, © 2006 Europa Technologies) 3.0 Site Characterization Methods 41  Figure 3.7: Hand-flapper drilling technique. A shallow (~0.5m deep) hole is dug into the ground and filled with pond water, and the drill pipe is bored into the soil at the desired location of the well.  The pipe top is then covered by hand and lifted from the ground, containing both sediment and sump water (A).  The pipe is then released to expel both sediment and water in a single motion (B).  At all observation wells, samples were collected from the slurry and used to log approximate stratigraphic horizons. (a)  (b)  Figure 3.8: Collection of sediment from drill pipes. (a) Sand is collected as slurry and (b) clay samples remain intact.  Intact clay samples were kept for laboratory analysis.  Coarser grained materials could not retained for representative geochemical analyses as the sediment was washed by pond water and fine components were lost. Sump water (drill lubricant) Drill pipe Sand expelled from drill pipe with slurry and collected to log Intact clay samples collected for analysis and to log stratigraphy  3.0 Site Characterization Methods 42  Figure 3.9: Hand-flapper drill setup. A bamboo lever-fulcrum apparatus aids with raising and lowering the drill pipe into the borehole.  The drill pipe is suspended from the lever by a metal chain over the borehole location.  Typically, 2-3 drillers are required to manage the load.  Figure 3.10: Photograph of drill pipe.  3.0 Site Characterization Methods 43  Figure 3.11: Observation well screens and silt traps.   Figure 3.12: Well installation. Main well screens PVC tubing Well screen and silt trap used for construction of observation wells 3.0 Site Characterization Methods 44 (a)  (b)  (c)  Figure 3.13: Multilevel sampling ports. 6 observation wells in the village were equipped with multilevel ports to allow collection of depth specific aqueous samples.  Multilevel tips were constructed with 1/4 inch ID PVC tubing and 185μm mesh screens.  A ~10cm interval was perforated and covered with mesh to allow flow while preventing clogging (a).  Screens were fitted to 1/8 inch tubing and attached to the main tube well, placing screens at specified depths, (b) and (c).  Figure 3.14: Well completion. The photograph shows displacement of water as sand is added to the annular space between the borehole and formation.  The collapsing of surrounding sediments over time is assumed to isolate the well screen sufficiently so that there will be negligible leakage from the surface.  3.0 Site Characterization Methods 45  Figure 3.15: Standpipe piezometer. Solinst® standpipe piezometer connected to 1-inch OD PVC riser pipe. Piezometers were installed adjacent to observation wells GSI0605 and GSI0606 to investigate geochemical and hydraulic gradients.   Figure 3.16: Depth to water measurement.  3.0 Site Characterization Methods 46  Figure 3.17: Slug test setup at well GSI0609. In this photo, the slug is currently in the well In-Situ® miniTROLL, which is connected to the PC for data downloading.  4.0 Site Characterization and Conceptual Model Development 47 4.0 SITE CHARACTERIZATION AND CONCEPTUAL MODEL DEVELOPMENT Interpretations of field data collected by the various means previously discussed are presented in this chapter.  Brief discussions of local geomorphology as well as the relevant depositional model precede the interpretation of site geology to supply context for the stratigraphic reconstruction based on borehole data (Section 4.1).  Geological interpretation is followed by hydraulic parameter estimates from slug tests, time-series head data, as well as grain size methods (Section 4.2).  Groundwater flow directions and aquifer connectivity are discussed in Section 4.3 from the analysis of piezometric data, and estimates of aquifer recharge based on hydrometeorological data and agricultural land use are presented in Section 4.4.  A qualitative discussion of site groundwater geochemistry is presented in Section 4.5, which will be used in conjunction with hydrological and geological interpretations to construct the preliminary site conceptual model of flow and transport (Section 4.6). For brevity, this section is restricted to the major conclusions arrived at which are most relevant for constructing the site conceptual model.  Raw data, assumptions, calculations, analysis complications and details may be found in Appendices A – E. 4.1. Site Overview This section summarizes site specific physiography, meteorology and hydrology.  Regional features are reviewed extensively in Section 2.4. 4.1.1. Physiography and Vegetation As discussed in Section 3.1, the study site lies within a region of Holocene meander belts scoured by the migrating Hooghly-Bhagirathi River.  Typical geomorphic features of this area include meander scars, oxbow lakes and abandoned channels, as well as wetlands between interdistributary levees (Figure 4.1).  The region is characterized by low relief, with surface elevations ranging between ~4 to 10 m asl in the immediate study area.  Ponds cover approximately 3% of the landscape and may have natural or anthropogenic origins (i.e. abandoned channel versus excavated).  Ponded areas are most common near villages, which are typically established upon natural levees and are heavily vegetated with trees.  The remainder of the landscape is lower-lying and planted to various types of crops including jute, rice, potatoes, mustard seed, cauliflower, cabbage and potatoes [Pal, 2008]. 4.0 Site Characterization and Conceptual Model Development 48 4.1.2. Meteorology As discussed in Section 2.4.3, climate in the region is tropical and can be grouped into distinct wet and dry seasons.  The dry season includes cool winters (November to February) and hot summers with occasional thunderstorms (March to May).  The majority of rain falls during the monsoon, which occurs from June to October [Nath et al., 2008b].  Available meteorological data for the site are limited to temperature and precipitation measurements documented on public websites.  Historical average monthly temperature and precipitation data recorded at local weather stations (Figure 4.2) were obtained from the National Climatic Data Center of the NOAA U.S. Department of Commerce (http://www.ncdc.noaa.gov/oa/climate/ghcn- monthly/index.php).  Climate Normals, including maximum and minimum daily temperatures and mean monthly total rainfall data (Appendix D) were obtained from the India Meteorological Service website (http://www.imd.gov.in/section/climate/kolkataweb.htm), and used for evapotranspiration calculations.  Rainfall data for Nadia district was obtained from the Indian Meteorological Department for the years 2004 to 2008 (http://www.imd.gov.in/section/hydro/distrainfall/webrain/wb/nadia.txt).  The data indicate that average monthly temperature ranges between ~19 and 31ºC (Figure 4.3), with daily minima and maxima of ~14 and 36ºC respectively (Appendix D).  Monthly precipitation in Gotra has ranged from ~0 to 471 mm in recent years (Figure 4.3).  Annual values vary between ~1260 and 1693 mm, and are comparable to historical data obtained from nearby weather stations (Figure 4.4). 4.1.3. Hydrology Nadia district falls within an interfluvial delta plain [A Mukherjee et al., 2007a], bounded by in the west by the river Hooghly and the river Ichamati in the east.  To the north of the study area, the River Churni departs from the Ichamati and flows westward towards the Hooghly.  River stage data obtained from the Tribeni tidal gauging station (Figure 4.2) show that the Hooghly Stage varies between approximately 2.5 and 8 m from dry to wet season, respectively (Figure 4.27). No discharge data is available from stream flow gauging stations or similar, although poor drainage in the immediate study area is suggested by the persistence of wetlands (“Bils”) in the dry season.  Local floodwaters that develop during the monsoon may drain to the oxbow lake via a north-south trending depression in the landscape (Figure 4.5).  Ponds are not hydraulically connected to any drainage system during the dry season, although they may become connected to monsoon floodwaters as they do in Bangladesh [Harvey et al., 2006]. Accordingly, surface runoff from village areas may conceivably drain to distant lakes with runoff from the agricultural areas, but may drain to ponds as well. 4.0 Site Characterization and Conceptual Model Development 49 4.1.4. Geology Upward-fining sequences characteristic to shifting meander belts are observed in the sedimentary record near the site [Pal and Mukherjee, 2009; Pal et al., 2002b], reflecting both the lateral and vertical accretion of sediment as channel and overbank deposits.  Satellite imaging and borehole data within Gotra suggest that the archetypal facies distribution and sedimentary sequences of a meandering stream (Allen [1964], Figure 4.6) is an appropriate context in which to interpret stratigraphy.  The following presents a brief review of the processes that deposit point bar successions and channel fill deposits, as well as general grain size characteristics for each, as described in the text “Depositional Systems” by Davis Jr., [1992]. 4.1.4.1. Point Bar Successions A stratigraphic section of a typical sedimentary succession deposited from a meandering river environment is shown in Figure 4.7 (Dickinson [1975]).  The fining-up sequence is divided into distinct deposit types which include channel lag, lower and upper point bar, and capped by overbank facies, which include levee/crevasse splay and flood basin deposits. Channel lag deposits are coarse grained materials (pebbles) that are generally not transported by meandering streams except during periods of high discharge.  Typically, these sediments are left behind as “lag” as the continually migrating river erodes channel banks. Point bar deposits represent the greatest accumulation of sediment in a meandering environment, and their thickness is indicative of the depth to which the channel extended.  They often display a distinct decreasing trend in grain size from bottom to top, which depends on particle size range and availability. Levee and crevasse splay deposits form on the banks of the channel meander, and consist of finer grained sands and silts that accumulate due to a loss in competence of the river bank. The difference between the two types of deposits is mainly in their respective extent of accumulation; natural levees are typically widespread and uniformly deposited, whereas crevasse splay deposits are the result of sudden, localized developments of fan-shaped “splays” from an overflowing channel that breach the river banks. Flood basin deposits accumulate in swampy areas between fluvial channels of low relief that experience poor drainage and slow sedimentation rates.  Materials are supplied from discrete 4.0 Site Characterization and Conceptual Model Development 50 flooding events and generally consist of fine-grained silts and clays that settle from suspension of floodwaters.  Organic materials such as plant debris (peat) are common in lacustrine environments, as well as bioturbation structures in the sediment.  Thinner accumulations are found in broad, flat areas compared to areas with lakes and swamps. At any given time, each of these environments may be present, from lag deposits up to the muddy overbank layers. 4.1.4.2. Oxbow Lakes and Channel Fill Deposits As the concave banks of a highly sinuous meander erode towards one another, the neck eventually is breached and the channel abruptly takes a new course.  Eddies will plug the abandoned channel, resulting in the gradual formation of an oxbow lake.  Flooding events alone supply sediment to the lake, which settle from suspension.  The materials supplied to the lake consist of fine grained silts that are often rich in organics.  Deposits are commonly stratified, and can preserve the formerly active bedforms that existed on the point bar surface prior to cutoff. 4.1.4.3. Gotra Geologic Model The regional Chakdaha stratigraphy has been identified as stacked sequences of deposit types described in Section 4.1.4.1 [Pal and Mukherjee, 2009; Pal et al., 2002b], and the shallow stratigraphy in Gotra represents an individual fluvial depositional package.  The village itself lies on the natural levee of an abandoned channel, with another abandoned channel to the northeast and low-lying wetlands (Bils) to the south (Figure 3.2).  Based on borehole information (Appendix A), the shallow stratigraphy of the site is understood to have been built through several depositional events in the Holocene.  These include the accretion of point bar sands by an active meander, followed by a subsequent channel shift and oxbow lake formation.  The point bar sands as well as the channel-fill silts were later breached by splay fans that developed during subsequent flooding events.  The evolution of the sequence is illustrated in Figure 4.12. 4.1.4.4. Cross Section Interpretation The strata at the site generally consist of Holocene sediments up to a depth of 25-30m, which are underlain by brownish grey sediments, possibly emplaced during the Pleistocene [Pal and Mukherjee, 2009; Pal et al., 2002b].  As previously mentioned, cross sections drawn through the Gotra area using data from the hand-flapper boreholes, reveal a classic abandoned channel 4.0 Site Characterization and Conceptual Model Development 51 facies within the Holocene sediments.  These are comprised of grey point bar sands and channel fill silts, capped by brown crevasse-splay overbank deposits. Channel-fill silts are found to the southwest of the Gotra meander scar and a series of remnant ponds mark the former location of the oxbow lake (i.e. the abandoned river channel) (Figure 4.8).  These deposits consist of fine-grained, dark-grey, organic-rich sediments and extend to an approximate depth of 20m below the ground surface (Appendix A).  Point-bar sands extend to the north east of the silt plug, indicating a southwesterly migration of the active channel, (Figure 4.9 and Figure 4.11).  Channel lag deposits (coarse sands, Figure 4.7) are rare in the borehole logs, but the point bar sands display a distinct fining-upward trend in grain size typical to this particular depositional environment (Section 4.1.4.1, [Dickinson, 1975]). Shallow sands, silts and clays (~3m below the ground surface) cut into both channel-fill and point-bar deposits, and represent depositional events that occurred subsequent to the silting up of the oxbow lake (Figure 4.9, Figure 4.10 and Figure 4.11).  The scouring of channel-fill and point-bar deposits likely occurred through the development of a splay fan during an extreme flooding event, where sediment-laden water was drained from the eastward shifted channel. Features with positive relief in the Google Earth image oriented roughly east to west and originating from the Naryanpur scroll represent the former channels of splay fans, from which smaller distributaries developed during flooding (Figure 4.9).  These distributaries transported the materials now breaching the surface of the channel-fill and point-bar deposits in the sediment record to their current location.  Internal organization of grain sizes is not as orderly as observed in the other fluvial deposits, indicating that splay sediment accumulated over multiple flooding occasions [Davis Jr., 1992]. The base of the meander sequence is inferred at a depth of approximately 30 m from the incorporation of organic materials (peat, wood) and a hard clayey layer containing calcrete („kankar‟) at some borehole locations (Figure 4.9 and Figure 4.10), representing a paleosol horizon.  These constituents may not be continuous throughout the lithologs due to partial erosion.  In areas where this paleosol exists, the unit grades into brown (oxidized) sands, identified regionally as the “Orange Sand Aquifer”, a potable water resource [Pal and Mukherjee, 2008]. 4.0 Site Characterization and Conceptual Model Development 52 4.2. Hydrogeology The following section focuses on conceptualizing site hydrogeology by translating the geologic model into hydrostratigraphic units.  The hydrostratigraphic context is developed in Section 4.2.2, following the interpretation hydrogeologic data (Section 4.2.1).  Aquifer connectivity is verified in Section 4.3. 4.2.1. Aquifer Parameters This section summarizes the hydraulic parameters computed for point-bar aquifer sands; channel-fill clayey-silts; and levee/crevasse-splay silts.  The main conclusions arrived at through data analysis are also presented.  Analysis of the data is discussed in detail in Appendix C. 4.2.1.1. Slug Test Results Conductivity estimates for the point bar sands from slug test data vary from 1 x 10-4 to 2 x 10-3 m/s (Appendix C).  This is a reasonable range of estimates for a clean sand [Freeze and Cherry, 1979].  Aquifer storativity estimates are highly variable, and range from ~10-2 to 10-7 (Appendix C).  This variability reflects the difficulty in fitting type-curve slopes to the test data. Hydraulic conductivity values calculated for the channel-fill silt range from 7 x 10-8 to 4 x 10-7 m/s (Appendix C), which are also reasonable estimates for the interpreted geological unit [Freeze and Cherry, 1979]. Aquifer hydraulic conductivity calculated using rising head data from slug tests are more reliable than values calculated from falling head data due to oscillations caused by dropping the slug into the well.  In addition, since the Hv and BR methods allow for partial well penetration and leaky-confined conditions that likely exist at the site (see Section 4.3.2 and Figure 4.20), they are considered to have produced superior, more accurate results than the CBP method.  Hv analyses are preferred over BR because no empirical quantities are required for this method [Bouwer and Rice, 1976]. 4.2.1.2. Pumping / Recovery Tests Aquifer hydraulic conductivity estimates made by fitting potentiometric data during irrigation well pumping to the Theis [1935] are approximately 2x10-4 m/s (Appendix C), which is a reasonable range of values for a sandy aquifer [Freeze and Cherry, 1979].  Values calculated using straight 4.0 Site Characterization and Conceptual Model Development 53 line methods are similar, and range from 6x10-4 to 2x10-3 m/s (Appendix C).  Specific storage values on the other hand tend to vary over several orders of magnitude (i.e. 1 x10-5 to 3x10-3 m- 1, Appendix C), suggesting an inconsistency of applying the chosen analytical model to characterize the properties of the Gotra aquifer.  In particular, leaky to unconfined conditions may prevail at the site since the uppermost deposits are fairly silty (Appendix A), and likely do not form an adequate confining upper layer.  This is supported by high storage estimates in some cases, which may be approximating specific yield as the upper units drain. 4.2.1.3. Hydraulic Conductivity Estimates from Grain Size Data Hydraulic conductivity values for fine - grained materials calculated using the Breyer method [Kresic, 1997] are systematically several orders of magnitude lower than those computed using the ROSETTA program [Schaap et al., 2001].  Breyer conductivities vary from 1x10-8 to 8x10-7 m/s (Appendix C) which is in the range of a silt or loess-type deposit [Freeze and Cherry, 1979]. Conductivities calculated using the ROSETTA program, on the other hand, range from 1x10-6 to 1x10-5 m/s (Appendix C), which are more representative of silts to silty sands [Freeze and Cherry, 1979].  Because of the grain - size classification of these materials predominantly as silts (Wentworth, [1922], Appendix A), the values calculated using the Breyer method are preferred. 4.2.1.4. Summary and Aquifer Parameter Calibration Ranges The hydraulic conductivity of the aquifer is high, generally >5x10-4 m/s, but varies within the range quoted by Freeze and Cherry, [1979] for a clean sand (i.e.7x10-5 to 1x10-3 m/s). Conductivities for the channel fill silt are approximately 1 to 2x10-7 m/s, which is within the range for a silt [Freeze and Cherry, 1979], and is consistent with geological interpretation.  Values computed for older alluvium, lenses and overbank deposits fall in a similar range as the channel fill silts; however, data for these units are sparse. Hydraulic parameter estimates for the site are summarized in Table 4.1 and plotted in Figure 4.13  The values recorded in this summary table will be used as best estimates for the different hydrostratigraphic units in the numerical model of the site (Section 5). 4.0 Site Characterization and Conceptual Model Development 54 4.2.2. Hydrostratigraphy Surface geomorphic features (Figure 4.1) combined with an idealized meandering fluvial depositional environment (Figure 4.6 and Figure 4.7) were used as a guide to construct the hypothetical hydrostratigraphy block diagram of Gotra shown in Figure 4.14.  This interpretation is consistent with the reconstructed stratigraphy based on borehole logs (Figure 4.9, Figure 4.10 and Figure 4.11), and the hypothesized sequence of depositional events that built the existing succession (Figure 4.12). It is important to note that the continuity of low-permeability material coincident with the paleosurface suggested by Figure 4.14 is a simplification of site geology.  Although this layer was not intersected by all boreholes drilled in Gotra, it has been included into the block model because several lines of evidence in the literature (e.g. Pal and Mukherjee, [2009]) as well as local knowledge inform us that a low conductivity layer is widely present at a depth of ~30 m. This particular layer has been also suggested to have acted as an impervious boundary for groundwater, preventing downward infiltration of arsenic contaminated groundwater [Pal and Mukherjee, 2008].  The implications of having a low permeability layer coincident with the paleosol horizon for the groundwater flow system will therefore be examined with numerical modeling (Section 5.0). Regions in Figure 4.14 labeled as “Earlier Deposits” represent strata already in place at the time that the meander was active.  As little borehole information was available for materials to the west of the channel fill deposits, these deposits are assumed to be part of an older floodplain sequence consisting of layered silts and sands (Figure 4.9 and Figure 4.10).  Sediments below the base of the paleosol were not intercepted by boreholes in this study, however, previous work by Pal and Mukherjee, [2009] has characterized them as fine- to coarse-grained oxidized sands, which were likely deposited during the Pleistocene. Hydrostratigraphic units are delineated in Figure 4.14 based on the geologic interpretation previously described (Section 4.1.4) as well as results from aquifer tests summarized in Section 4.2.1.  The upper and lower point-bar sands, as well as lag deposits are grouped together to form the main village aquifer, and channel-fill silts are conceptualized as low conductivity units. Finer grained overbank deposits and the paleosol included in this conceptualization may act as upper and lower confining units, respectively.  The floodplain alluvium is divided into layers of low and medium conductivity units, and the Pleistocene sand is represented as a single, homogeneous aquifer. 4.0 Site Characterization and Conceptual Model Development 55 4.3. Piezometry This section summarizes the key site characteristics determined from time-series data collected at the piezometers.  This includes a summary of general physico-chemical observations, followed by inferences made from the data about aquifer connectivity and anisotropy, as well as groundwater flow directions.  Examination of the potential contribution of the Hooghly River may to aquifer recharge is also presented.  Piezometric instrument deployment and specification details are described in Appendix B. 4.3.1. Time-Series Observations Both seasonal and diurnal oscillations of the potentiometric surface are observed at the site Seasonal variations are apparent in Figure 4.15, which shows the complete hydraulic head dataset at all wells over the entire data collection period.  As barometric correction data are not available for 2008-2009, only approximate heads measured at GSI0603 are plotted for this period. Long-term monitoring of the piezometric surface indicates that head in the aquifer varies up to 7m annually.  Water levels rise concurrently with the progression of the monsoon (~June to October), then fall as the rainy season ends and irrigation picks up.  Although there is evidence of aquifer pumping throughout the year, pumps are either shut off during intense rain events, or the drawdown signal is dwarfed by recharge entering the aquifer (Figure 4.15). Expanding the scale of the hydrograph during the 2007 and 2008 field visits reveals a dynamic response of the piezometric surface to pumping (Figure 4.16 and Figure 4.17).  Intermittent irrigation and domestic pumping from Wells 4, 33 and 50 combined with rolling blackouts due to excessive demand for electricity (load shedding), produce an irregular series of drawdown and recovery responses of the piezometric surface (Figure 4.16 and Figure 4.17 ).  Pumping appears to persist in the region until stopped by a load shedding event, which generally occurs in early to mid-afternoon.  The piezometric dataset reveals that there may also be sporadic power cuts responsible for shorter recovery periods.  Alternatively, shorter pumping durations may be also the result of manual pump operation (i.e., a farmer turns pumps on in the morning, off in the evening), although there is no documentation of pumping schedules.  Daily head changes due to pumping range from ~0.1 to 0.5m. 4.0 Site Characterization and Conceptual Model Development 56 The field visits in 2007 and 2008 both occurred during the irrigation (dry) season.  Figure 4.17 (a) displays the data collected during the 2007 visit, approximately one week later in the season than the 2008 sampling interval shown in Figure 4.17 (b), and as such, the overall water levels in (a) are lower than they are in (b).  Both datasets display a gradual overall decline consistent with the advancement of the dry season and cumulative extraction of groundwater from the aquifer, before it is recharged during the monsoon. Temperature measured throughout the year in the aquifer remains relatively constant, between 26 and 27 degrees Celsius (not shown).  Slight and gradual increases are recorded by probes deployed at shallower depths (i.e. GSI0607 and GSI0608 (2006-2007) and GSI0609 (2007- 2008)) that are coincident with the progress of the monsoon season and reflect surface temperatures.  Electrical conductivity (EC) and pH were monitored in wells equipped with In- situ® Troll MPs and Aqua Trolls (Appendix B).  Aside from sensor equilibration periods immediately following instrument deployment, the probes are generally stable from 2006 to 2008 (Figure 4.18).  Electrical conductivity for wells GSI0605, 08 and GSI0714 generally remain at ~1800, 1400 and 1100 µS/cm respectively, and pH in all wells measured approximately 8 for the duration of the monitoring period.  Rapid pH increases may represent episodic degassing of CO2 due to biological activity in the fine grained sediments (i.e. in GSI0605 and GSI0606). 4.3.2. Response to Pumping and Local Connectivity of Aquifer In order to assess the level of hydraulic connectivity between observation wells, and therefore local continuity of the aquifer, head versus head cross-plots were constructed using data from wells that had simultaneous water level measurements during the dry season.  Long-term data sets were excluded from this analysis because the piezometric surface rises uniformly in all wells in response to seasonal stressors (e.g. intense rainfall), and thus does not reveal much about local flow heterogeneities.  Conversely, head data collected at high frequencies are not immediately affected by any major seasonal water level change.  This allows the instantaneous effects of pumping, (which are small compared to seasonal fluctuations), to be compared between observation points. These interference tests are organized into head correlation matrices in Figure 4.19 (a) and (b) using data from 2007 and 2008, respectively, and confirm that there is good correlation of heads between most wells screened within the village.  Good correlation indicates that the aquifer is well connected between observation points, and that any observed drawdown is caused by the same set of pumping wells.  GSI0713 however does not correlate well with other 4.0 Site Characterization and Conceptual Model Development 57 village wells, which suggests that it is screened within a different hydrostratigraphic unit (possibly those denoted as “Earlier Deposits” in Section 4.2.2).  This would cause GSI0713 to respond differently to pumping than its closest neighbors, despite its proximity to common pumping well(s).  GSI0609 also appears to be an outlier, as there is scatter in the data plotted in each of its correlation graphs, including GSI0713.  The depreciated correlation of heads at GSI0609 with the other village wells may be explained by pumping at distant wells (i.e. near Naryanpur), or heterogeneities distinct from those experienced near wells screened immediately within Gotra (i.e. beyond a correlation scale of ~10m). Drawdowns used to estimate aquifer parameters in Section 4.2.1 are plotted with time normalized by the square of the distance from the pumping well are in Figure 4.20.  The parallel nature of these lines for observation wells screened in the village is indicative of a general homogeneous transmissivity.  A slightly flatter slope exhibited in the data collected at GSI0713 and GSI0609 indicate that the sands are slightly less permeable in these zones.  The decreasing trend in drawdown with distance from the pumping well (i.e. separation of lines) suggests that there is leakage from units above the drawdown cone.  This separation may also reflect the variable thickness of the channel-fill unit (i.e. wells such as GSI0603 and 04 screened below thinner successions do not drain as quickly), although this is complicated by wells pumping simultaneously at numerous locations. 4.3.3. Groundwater Flow Directions Time-series water level data used to examine horizontal and vertical flow in the area are discussed the following sections.  The number of dataloggers collecting simultaneous measurements limits well control to few data points for the construction of potentiometric maps, making the interpretation of horizontal flow across the village difficult.  Gradients between pairs of wells were calculated during periods of pumping quiescence (i.e. load shedding or monsoon season), where simultaneous measurements are available to aid with flow direction interpretation. 4.3.3.1. Vertical Flow The vertical component of groundwater flow in the upper point bar sand was investigated using head data collected at wells GSI0603 and GS10604 where simultaneous measurements were available.  Available data and calculated gradients are plotted in Figure 4.21.  The data indicate that there is a consistent downward gradient from GSI0604 to GSI0603, implying that downward 4.0 Site Characterization and Conceptual Model Development 58 flow occurs between these well screens.  Domestic and irrigation pumping complicate the signal, often reversing gradient direction momentarily.  However, when pumps are shut off overnight (Figure 4.21 (a) and (b)) or during the rainy season (Figure 4.21 (c)), the gradients approach a fixed value, which ranges from 0.0004 to 0.002 downward (averages between 0.001 and 0.002). A similar analysis procedure was conducted on the pre-purge data collected from piezometers screened within the channel fill silt near well GSI0605.  Plots of the calculated gradient reveal a more pronounced downward component of flow, which ranges between 0.02 and 1.0 (Figure 4.22).  Average vertical gradients from both units are summarized in Table 4.2. 4.3.3.2. Horizontal Flow An examination of horizontal gradients from piezometric data is presented in this section. Potentiometric maps were constructed by interpolating water level data to illustrate planar groundwater flow directions at various points in time.  Contours plotted in Figure 4.23 and Figure 4.24 were constructed from data collected during the 2007 and 2008 field visits respectively, and represent snapshots in time when pumps were shut off (i.e. Figure 4.21 (a) and (b)).  The resulting potentiometric maps imply that the general flow direction is from the crop fields towards the village.  Additional head data selected for the construction of potentiometric maps were taken from measurements made at midnight on October 1, 2007 (i.e. Figure 4.21 (c)) to compare monsoon flow directions with the pattern observed during the dry season.  The piezometric surface interpreted using these measurement points is shown in Figure 4.25, and implies that there is divergent flow away from the village. Since continuous datasets from all observation wells are not available (i.e. dataloggers are not always deployed in the same wells), horizontal flow conceptualization is difficult.  Contoured surfaces are incomplete and produce bulls-eye patterns around wells reflecting measurement coverage biases. Gradients calculated between pairs of wells (Table 4.3) suggest that horizontal flow in the immediate village area may be less significant than flow in the vertical direction.  However, horizontal gradients between wells involving GSI0609, GSI0713 and 14 are comparable (i.e. ~1 x 10-3) to those calculated between GSI0603 and 04.  In general, flow is directed away from GSI0609 and towards GSI0713 (i.e. northeast to southwest).  Flow is also directed away from 4.0 Site Characterization and Conceptual Model Development 59 GSI0714.  A cross sectional schematic showing flow directions from gradient calculations is shown in Figure 4.26. 4.3.3.3. Potential Regional Gradients Figure 4.27 shows that piezometric levels in the village are consistently lower than the Hooghly River stage, except for short time intervals following the peak of the monsoon.  This shows that aquifers may potentially be recharged by the Hooghly River on a regional scale for the majority of the year, although flow reversals may occur briefly during periods of minimal rainfall. However, as with horizontal gradients within Gotra, gradients between the Hooghly and the village are low (i.e. ~1 x 10-3 Figure 4.27 (b)), and will likely be overwhelmed by local vertical gradients throughout the year. 4.3.4. Piezometric Data Summary The piezometric data reveal substantial information about the aquifer and groundwater flow in the Gotra area.  Oscillations of the potentiometric surface over time indicate that groundwater flow in the area is sensitive to both seasonal and daily stressors, and suggest that vertical flow is significant. Comparison of simultaneous head measurements between wells screened within the main point-bar aquifer indicates that local connectivity of this unit is appreciable.  Reduced correlation between measurements collected at GSI0713, GSI0609 and other wells indicate that these piezometers are screened within separate materials, or are affected by distinct stressors.  In the case of GSI0713, it is likely that the well is screened within an older alluvial unit, as discussed in Sections 4.1.4 and 4.2.2.  As GSI0609 is located closer to crop fields than other piezometers, it is likely that irrigation wells in the fields influence drawdown at this well more than others, thus making it an outlier. An overall homogeneous transmissivity is suggested by drawdowns from “pumping tests” plotted against normalized time for wells within the village.  Flatter slopes observed in similar plots for wells GSI0713 and GSI0609 suggest lower transmissivity measured at these points, consistent with the possible extraneous influences of stratigraphy and pumping discussed previously.  Decreasing trends in drawdown with distance from the respective pumping well may be indicative of vertical leakage, or may also reflect the variable thickness of the overlying channel-fill unit at the various piezometers. 4.0 Site Characterization and Conceptual Model Development 60 Analysis of hydraulic gradients using simultaneous head measurements in the village indicates that horizontal flow in the point bar sand may be less significant than vertical flow.  Vertical gradients in the point bar unit tend to be on the order of 10-3, whereas horizontal gradients in this unit vary from 10-3 to 10-4.  Vertical gradients within the channel-fill silt are greater in magnitude, and vary from 10-2 to 10-1.  Additionally, it is likely that the Hooghly River has little control on flow within Gotra, as imposed regional gradients are comparatively small (~10-4). The piezometric data in Gotra indicate that groundwater flow is controlled heavily by seasonal and anthropogenic stressors, and is dominated by vertical gradients.  Given these characteristics, it follows that meteoric recharge and irrigation pumping should be primary sources and sinks of water to the aquifer, respectively.  However, the contribution to aquifer recharge from regional sources is uncertain.  In order to improve the conceptualization of recharge sources to the Gotra aquifer, it is prudent to compute a water budget of potential groundwater sources and sinks with available data.  This exercise is carried out in the following section. 4.4. Monthly Water Budget The purpose of this section is to develop a rough idea of the fate of precipitation and irrigation water from available data for comparison with the total volume groundwater for irrigation.  As discussed in the previous section, these quantities are considered to be of principal significance on the water budget in the village area.  Total extracted groundwater is estimated using piezometric data, irrigation well densities as well as anecdotal information (Section 4.4.1).  Net recharge to the water table is computed through summation of monthly precipitation, estimated and values of evapotranspiration, irrigation return flow and runoff, under the assumption that changes in storage from month to month are negligible (Section 4.4.2).  Potential contribution to aquifer recharge from regional sources (i.e. the Hooghly River and oxbow lakes) is also considered (Section 4.4.3).  The values computed in this section will aid in boundary condition conceptualization for the numerical flow model. 4.4.1. Groundwater Extracted through Irrigation Pumping (Q) Harvey et al., [2006] used daily oscillations of aquifer head to calculate the number of wells in use each day, and an average daily pumping duration over the course of a measurement period.  Using these two quantities combined with a measured average pumping rate, the 4.0 Site Characterization and Conceptual Model Development 61 amount of water extracted through irrigation within a given agricultural area can be expressed as: Equation 4.1:    t i ii av dn At Q Q 1 where: Q = extracted irrigation water over measurement period [L]; Qav = average pumping rate at individual wells [L 3/T]; A = area of cultivable land [L2]; ni = number of wells pumping in a given day [-]; di = duration of pumping in a given day [T]; t = total number of days in measurement period (i.e. number of days irrigation pumping occurs in a given year) [-] It is important to note that the applicability of this method to our site is limited because it is based on a model that assumes one single value of hydraulic head throughout the entire aquifer.  Consequently, this calculation requires that the drawdown observed from each irrigation well pumping at the same rate is equal (i.e. the medium behaves like an infinitely permeable “tank”).  In our case, the hydrograph at any given observation well is affected most by the irrigation wells that are nearest, more specifically, pumping wells that extract water from shallower depths.  Thus, the total drawdown at an observation well is not the sum of equal drawdowns from each irrigation well, and it is therefore not possible to determine the number of wells pumping at any given time.  However, since the drawdown responses observed in the hydrographs at Gotra are reflective of shallow irrigation schedules (i.e. pumping at wells 33 and 50), they may be used to approximate irrigation water usage over the course of the year (i.e. the duration of time that pumps are on).  Instead of using daily drawdowns to estimate the number of wells pumping, it is assumed that all irrigation wells are pumped for the same proportion of a given day as suggested by the shallow drawdown signal from November to May.  Although this 4.0 Site Characterization and Conceptual Model Development 62 may be an overestimate of daily pumping in the dry season, it is nonetheless reasonable because crop irrigation schedules are largely controlled by load-shedding (Section 4.3); in all likelihood wells will be pumping for irrigation as long as power is available, and this is indicated by the drawdown response observed in the hydrographs.  Equation 4.1 was therefore modified to become: Equation 4.2:    t i iavwells dQQ 1   where: ρwells = well density [L -2] At the Gotra site, 17 deep (>40m) and 2 shallow (<40m) irrigation wells were noted within an approximate radius of 1km from the village.  This translates to a well density of >5 wells per square kilometre, which is similar, but slightly higher than the tally quoted by Nath et al., [2008b] of ~2 wells/km2 of cultivable land in Nadia district from statistical records of land use. The hydrograph of GSI0605 was used to determine daily pumping durations from June 2006 to February 2007.  Seasonal variations were removed from the data to reveal only diurnal head changes by subtracting a 24-hour moving average from the data (Figure 4.28 (a)), which represent daily pumping.  The daily durations of pumping were calculated by determining the elapsed time between maximum head followed by minimum head measurements within a given day.  The estimated number of hours of pumping each day are plotted in Figure 4.28 (b), which shows that the daily duration of pumping (but not necessarily volume extracted), is generally constant throughout the year, at an average of 10.0 +/- 5 hours per day (~42% of the day). To estimate individual well extraction rates, Qav, the time required to fill a 53L bucket with the outflow at various pumping wells within the village and fields was measured.  The bucket was filled three times at each well visited, and the average pumping rates were deduced from the times recorded.  The average calculated pumping rate is 11.3 +/- 1.32 L/s for all the wells measured at the site (Figure 4.29).  This is considerably lower than the value obtained by Harvey et al., [2006] of 24.1 L/s for similar pumping wells in Bangladesh, even when considering the error in our measurements at individual wells.  It is however similar to the value of 15 L/s quoted in older reports of shallow irrigation wells in the area cited by Harvey et al., [2006]. 4.0 Site Characterization and Conceptual Model Development 63 Finally, using Equation 4.2 and assuming the irrigation occurs primarily from November to May [Pal, 2006] the calculated amount of water extracted as a result of irrigation activities in Gotra is approximately 0.43m (433mm) per year.  This result agrees with the total annual abstraction estimated for West Bengal of Mukherjee et al., [2007a], of 438 mm. It is important to note that pumping schedule estimates made contain two possible sources of error: 1- it was assumed that all wells were pumping for the daily durations implied by the hydrograph from November to May, and 2- the number of wells pumping from June to October was assumed to be zero, despite the occurrence of daily oscillations in the hydrograph (although dampened) during this time.  However, in the absence of any documented pumping schedule for the area complete with individual pumping rates, it was considered reasonable to make the said assumptions. 4.4.2. Net Recharge (G) Potential recharge sources to the Gotra aquifer include precipitation, irrigation return flow and inflow from distant sources such as lake seepage or losing stream reaches.  This section addresses potential recharge to the aquifer only through vertical infiltration from the ground surface. The following relationship describes the total amount of water available to recharge the aquifer from the water table: Equation 4.3:  RETIPG  where: P = precipitation, [L]; I = recycled irrigation water, [L]; ET = evapotranspiration, [L]; R = surface runoff, [L]; G = groundwater recharge [L]; 4.0 Site Characterization and Conceptual Model Development 64 Land use at the site is variable (i.e. for habitation, ponds, and various types of agriculture) and as such values of G should reflect these zone differences.  For example, non-zero values of I will only apply to irrigated areas.  Similarly, estimates of ET will depend on crop characteristics and land management [R G Allen et al., 1998]; and R should reflect soil type and land cover [SCS, 1956] in addition to climatic parameters.  Consequently, the various land zones in the vicinity of the research site have been designated as irrigated “High” and “Low” Lands, ponds and village areas, based on knowledge of agricultural use [Pal, 2009] and visible zone boundaries.  This results in calculation of unique values of G for the various specified zones. The total landscape area in the vicinity of the study site is comprised of the aforementioned zones as follows: 29% High Lands; 43% Low Lands; 3% ponds; 24% village.  Computation of the individual contributing components of Equation 4.3 is described in the following sections. 4.4.2.1. Irrigation Return Flow (I) Harvey et al., [2006] showed that irrigation return flow (i.e. recycled groundwater) can constitute a significant amount of recharge to the aquifer at their study site in rural Bangladesh.  Similarly, since crops are heavily irrigated in Gotra, it is prudent to consider the potential contribution of irrigation water to recharge.  Harvey et al., [2006] estimate irrigation efficiency as the water applied to fields (Q), minus water “stored” in the fields (ΔS) and water lost due to evapotranspiration (ET).  As evapotranspiration has been accounted for separately in this work (see Section 4.4.2.2), a modified irrigation return flow term for Gotra is calculated as: Equation 4.4: SQI   If it is assumed that the water column above the ground surface is maintained at 20 cm as it is in Bangladesh [Harvey et al., 2006], the contribution of extracted groundwater to recharge is 0.43m – 0.2m = 0.23m per year in Gotra.  This flux is assumed to be distributed over the months of November to May (i.e. ~33mm /month for 7 months), and applies to agricultural lands only. 4.4.2.2. Evapotranspiration and Evaporation, (ET, Ep) The FAO Penman-Monteith [R G Allen et al., 1998] method was used to calculate evapotranspiration in the village, High Land and Low Land areas.  This procedure requires determination of a monthly reference evapotranspiration, ET0, in addition to unique monthly crop 4.0 Site Characterization and Conceptual Model Development 65 coefficients Kci to obtain distinct ET values for given area.  A separate method was used for estimating evaporation from ponds Ep, which is also known as potential evaporation from a free surface [Maidment, 1993].  These methods and calculations are discussed in detail in Appendix D. Annual ET for High Lands, Low Lands and Village, and Ep for ponded areas are 1330, 1710, 1486 and 1849 mm, respectively.  Monthly evapotranspiration and pond evaporation are shown in Table 4.4.  In general, pond evaporation values are systematically higher than all other values throughout the year (104 to 207 mm/month), and evapotranspiration in the High Lands is less than or equal to that in the Low Lands for the most part (48 to 162 mm/month vs. 93 to 190 mm/month, respectively).  Village evapotranspiration is generally lower than ETc Low Lands, but higher than ETc High Lands (84 to 172 mm/month).  These ranges in values are consistent with evapotranspiration values calculated by Harvey et al., [2006] for their site in Bangladesh. 4.4.2.3. Surface Runoff, R Surface runoff is expected to be negligible during the dry season due to the high evaporation and evapotranspiration rates computed in the previous section.  However, upon the development of aquifer-full conditions during the monsoon, there will be excess water from precipitation that does not recharge the aquifer.  Indeed, in the year 2000, flooding in Nadia district was documented [A Mukherjee et al., 2007a].  It therefore is necessary to consider surface runoff in the water balance for months when P + I – ET >0. Accurate quantification of surface runoff requires long-term hydraulic head data and surface water stage measurements from multiple locations in the study area.  Because such data are not available, the empirical USDA Soil Conservation Service (SCS) Runoff Curve Number (CN) method [SCS, 1956], modified for Indian conditions [SCD, 1972] as used to estimate surface runoff in the area.  A summary of this method, assumptions and calculations is presented in Appendix D. As discussed in Section 4.1.3, crop field flood waters likely drain to distant lakes via ephemeral channels, whereas surface runoff from the village may or may not contribute to pond recharge. For the case where village surface runoff enters ponded areas, the additional contribution to pond recharge (Rp) from village runoff may be calculated as: 4.0 Site Characterization and Conceptual Model Development 66 Equation 4.5: v p v p R f f R   where: fv = fraction of total landscape occupied by villages, [-]; fv = fraction of total landscape occupied by pond, [-]; Rv = village surface runoff, [L] Runoff to ponds is added to potential recharge to the aquifer from ponded areas, in contrast to lowering net recharge in the remaining zones in Equation 4.3. Annual surface runoff for High Lands, Low Lands and village areas are approximately 131, 605 and 229 mm, respectively.  Monthly values (when P + I - ET >0) range from 0.1 to 59, 47 to 216 and 10 to 83 mm, and are shown in Table 4.5.  These values are slightly lower, but comparable to observed annual and monthly runoff values for small catchments in Eastern India, which range from ~200 to 700 mm/yr and 30 to 200 mm/month, respectively [Kothyari, 1995; Kothyari and Garde, 1991].  The maximum additional contribution to pond recharge from the village overflows to the ponds calculated using Equation 4.5 is approximately 1907 mm/yr, and ranges from 80 to 693 mm in flooding months (Table 4.5). 4.4.2.4. Groundwater Recharge, G Returning to Equation 4.3, groundwater recharge was calculated using monthly precipitation data for Nadia District, and the computed values for irrigation return flow, evapotranspiration and evaporation and surface runoff.  During months where ET+R > P+I, aquifer recharge estimates were assigned value of zero.  Using average annual precipitation values, annual recharge for High Lands, Low Lands and village areas are estimated at 489, 0 and 417 mm, respectively.  Pond recharge is bracketed by the cases where village runoff to ponds or to distant lakes, and ranges from 403 to 2310 mm/yr.  Scaling the recharge over the area of interest according to the contributing fractions of each zone to the entire landscape (see Section 4.4.2) produces an overall recharge value of 312 mm/yr when ponds receive village recharge and 257 mm/yr when they do not.  Recharge values are summarized in Table 4.6, and plotted in Figure 4.30 with individual components. 4.0 Site Characterization and Conceptual Model Development 67 A notable feature about the plots in Figure 4.30 is that Low lands are not expected to receive any meteoric or recycled groundwater recharge at any time during the year.  This situation is conceivable, as these areas have comparably low elevations, and thus may even act as discharge zones during the monsoon season.  The plots in Figure 4.30 also illustrate that irrigation return flow contributes very little to recharge, as high evapotranspiration rates prevent the re-infiltration of pumped water during irrigation months.  Indeed, if return flow is neglected, annual meteoric recharge to the area ranges between 305 and 250mm for the two described drainage scenarios.  Furthermore, this additional recharge is expected only to affect High Lands during the month of May, (i.e. the only month during the irrigation season for which ET+R is expected to exceed P+I).  Thus, the overall contribution of recycled groundwater to aquifer recharge in Gotra is insignificant, contrary to the observations of Harvey et al., [2006], at their site in Bangladesh. Estimates of recharge to the water table of the Gotra aquifer are higher than, although consistent with simulated values of meteoric recharge to West Bengal (~183 mm/yr) by Mukherjee et al., [2007a].  However, the approximation of a large number of uncertain parameters (i.e. the Penman-Monteith method), and the use of an empirical model for estimation of evapotranspiration and runoff have likely contributed some error to the calculations for Gotra.  On the other hand, the difference in scale between this study and that of Mukherjee et al., [2007a] may account for some discrepancy, as local conditions (i.e. total rainfall, temperatures, land use patterns) used to predict recharge in Gotra may not necessarily be applied regionally. 4.4.3. Other Potential Recharge Sources As estimates of extracted groundwater exceed the computed value of vertical recharge, it follows that there is an additional source of groundwater that replenishes the Gotra aquifer that has not been considered if annual inflows balance outflows.  In fact, Mukherjee et al., [2007a] estimate that 274 mm/yr is required in addition to meteoric recharge to calibrate to observed heads in their numerical model.  Other studies have estimated average groundwater recharge which includes meteoric, recycled and other sources of water at ~600 mm/year for the entire Bengal Basin [Basu et al., 2001; Dowling et al., 2003]. A potential regional source of groundwater to the Gotra aquifer might be the Hooghly River, since a gradient exists between the river stage and village heads, and is directed towards the village (Figure 4.27).  However, the observed gradients between the river and the heads 4.0 Site Characterization and Conceptual Model Development 68 observed in GSI0605 would require an unrealistically high conductivity (~10-1 m/s) to supply the equivalent aerial flux of ~150mm/yr to the village, assuming an average aquifer thickness of 30m and a 12.5km distance to the river.  Furthermore, exchange between the river and surrounding aquifer is likely limited by the hydraulic conductivity of the river channel sediments. Such behaviour was noted by Harvey et al., [2006], at their site in Bangladesh, where large differences between the Ichamati River stage and aquifer heads were observed, despite the close proximity of piezometers to the river (i.e. tens of metres). Other potential sources of recharge to the Gotra aquifer include the nearby oxbow lakes and wetlands (Figure 4.1, Figure 4.5).  Based on the Google Earth image, these features comprise approximately 8% of the area within a 5km radius of the village.  If average annual water levels in these areas are assumed to equal DEM elevations, and average depths are assumed to be 5-10 m [Pal, 2007], a hydraulic conductivity on the order of 10-7 m/s would allow these sources to supply the necessary flux to the Gotra aquifer to replenish it annually.  This conductivity value is consistent with those estimated for the channel fill silts below the village area (Section 4.2.1), which were deposited in the exact same manner (Section 4.1.4).  Accordingly, it is conceivable that these surface water reservoirs could provide the aquifer with the annual deficit of groundwater predicted by previous water balance calculations.  However, this relies on the assumption that average levels in Gotra are ~3m, and does not consider the transient nature of the system over the course of the year.  In order for lakes to be a net recharge source to the aquifer, lake heads must exceed groundwater levels for the majority of the year, and there are insufficient data to conclude whether or not this is the case. 4.4.4. Water Budget Summary The total amount of groundwater that is extracted annually from the Gotra aquifer through pumping is approximately 433 mm.  This exceeds the amount of water entering the aquifer from meteoric and recycled sources, whether ponds receive surface runoff from village areas or not (i.e. 312 and 257 mm/yr respectively).  If there is no net change in storage over the course of a hydrologic year, non-meteoric recharge must enter the aquifer in order for it to be replenished. Recharge to the water table occurs predominantly within the monsoon months, (June to October), and is expected to be almost exclusively of meteoric origin.  This is the result of high evapotranspiration rates during the irrigation season, which predict that applied irrigation water will not infiltrate the ground surface, and thus will not comprise a significant portion of aquifer recharge.  Land use patterns at the site indicate that recharge sources to the water table will be 4.0 Site Characterization and Conceptual Model Development 69 spatially variable as a result of unique estimates of evapotranspiration, evaporation and runoff. Greater amounts of recharge are expected to enter the aquifer through the village than the agricultural fields.  Pond recharge is expected to either greatly exceed other sources, or be comparable to village and High Land values, depending on the fate of village surface runoff. Low Land areas are not expected to receive any recharge from meteoric sources, as runoff from paddy fields is predicted to be excessive.  Although considerable uncertainty has likely arisen as a result of approximations where data were lacking and the use of empirical models, the computed values are consistent with those of several researchers who have conducted similar water balance analyses in the Bengal Basin. The source of non-meteoric recharge to the Gotra aquifer is uncertain.  It is possible that the Hooghly River contributes inflow to the area, but unlikely as hydraulic connection between the River and surrounding aquifers is probably limited by the low conductivity of riverbed sediments. Wetlands and oxbow lakes may also provide the required volume of water to balance the deficit at the water table predicted by water balance calculations.  However, no hydrologic data from gauging stations are available to constrain potential fluxes that may be relevant to the site. When combined with the conceptualized hydrostratigraphy, examination of recharge contributions through numerical modeling may clarify how geochemical patterns have developed at the site.  Of particular interest will be the resultant flow patterns and flushing times, and their relationship to the arsenic pattern.  Geochemical characteristics of the Gotra groundwater as they relate to the construed understanding of hydrostratigraphy and groundwater flow are discussed in the following section, prior to numerical model development. 4.5. Hydrogeochemistry The purpose of this section is to identify geochemical processes potentially responsible for arsenic release, as well as to speculate on relationships between aqueous chemistry and the hydrogeologic features discussed in the preceding sections of this chapter.  More importantly, this section will interpret geochemical patterns in the context of interpreted groundwater flow. Analytical results for all aqueous samples are provided in Appendix E.  Speciation calculations were computed for samples where alkalinity and field pH were available using PhreeqCI (v. 2.15, Parkhurst and Appelo, [1999]) and the Wateq4f thermodynamic database.  Charge balance errors were generally less than 10%, and are also given in Appendix E. 4.0 Site Characterization and Conceptual Model Development 70 4.5.1. Groundwater Quality and Contaminant Distribution Approximately 80% of the wells sampled exceed the WHO permissible limit (10μg/L), and ~50% exceed the Indian guideline (50μg/L) for arsenic content in drinking water.  The highest observed concentration of dissolved arsenic was 515.9 μg/L, and was measured in domestic well 46 at a depth of 27.4 m.  This sample location is directly below the thickest accumulation of channel-fill deposits.  Laboratory analyses confirm preliminary results from the Merck Kit that contaminant distribution at the site is patchy over short horizontal scales; contamination can vary up to 2 orders of magnitude over distances as short as 20m in the shallow aquifer (Figure 4.31), and 30m below the interpreted paleohorizon (Figure 4.32).  Major ion chemistry of the groundwater in Gotra is dominated by calcium, carbonate and bicarbonate (Figure 4.33). Additionally, it appears that arsenic contamination is associated with the most Ca and HCO3 + CO3 enriched samples. Vertical profiles of water quality (Figure 4.32) show that aqueous samples exhibit comparatively high geochemical variability above a depth of ~30m.  The profiles also indicate that higher concentrations of dissolved arsenic in Gotra are more commonly found in samples collected from shallow depths.  However, concentrations in the 100 to 200 ppb range are also observed in deeper groundwaters (Figure 4.32).  pH ranges from 7-9 (mean 7) throughout the dataset, with no identifiable spatial trends, or relationship to arsenic enrichment (Figure 4.32 (a)). Electrical conductivity also does not correlate with dissolved arsenic, although higher values (up to ~3000 µS/cm) are encountered at shallower depths (Figure 4.32 (b)).  Dissolved iron, ammonia, phosphorous (likely phosphate), and measured alkalinity all appear to be correlated with dissolved arsenic, whereas sulphate exhibits an inverse relationship with the contaminant (Figure 4.32 (c) to (g)).  As discussed in sections that follow, elevated alkalinity and concentrations of dissolved iron and ammonia may be indications of reducing conditions in groundwater that influence arsenic release.  Elevated simulated pCO2 values associated with high arsenic suggest that reducing conditions occur in the most contaminated waters (Figure 4.32 (h)). Despite the geochemical variability in shallow waters, there appears to be a decreasing trend in shallow groundwater-arsenic in a northeast direction, perpendicular to the surface trace of the channel-fill silt (Figure 4.31). Projection of all available data onto a northeast trending profile shows the pattern of contamination with the interpreted location of the scroll in cross section (Figure 4.34).  Figure 4.31 and Figure 4.34 both suggest that the zone of arsenic release in the Gotra area is closely tied to the channel-fill deposit.  This relationship is supported by the 4.0 Site Characterization and Conceptual Model Development 71 association of solid-phase arsenic to this deposit type established through sequential leach and total-digestion analyses documented in companion studies of the site [R Beckie et al., 2006; R D Beckie et al., 2009; Pal et al., 2009]. 4.5.2. Geochemical Controls on Dissolved Arsenic in Gotra The geochemical characteristics of affected waters in Gotra are consistent with those of many arsenic contaminated aquifers observed in the BDP (e.g. [BGS and DPHE, 2001; Bhattacharya et al., 2003; McArthur et al., 2001; Nath et al., 2008a; Nickson et al., 2000; Ravenscroft et al., 2001; Smedley and Kinniburgh, 2002; van Geen et al., 2003b]).  These aqueous characteristics are conducive to reductive arsenic release from iron and manganese oxides into groundwater. Specifically, highly contaminated waters in Gotra tend to have elevated dissolved iron and ammonia, have an inverse relationship with sulfate, and are generally more reducing (i.e. higher simulated pCO2, Figure 4.32).  Dissolved methane is present in contaminated groundwater (gas collected from affected wells was flammable), indicating that affected waters are reducing; however, methane content has not been quantified in the laboratory. No arsenic-bearing minerals were computed to be at saturation in PhreeqCI simulations.  This suggests that dissolved arsenic at the site is controlled at least in part by surface processes such as sorption.  However, phosphate is known to compete strongly with arsenic for sorption sites on oxide mineral surfaces [Murali and Aylmore, 1983], and the weak correlation between dissolved iron and phosphate (mole equivalent of total phosphorous) (Figure 4.36) suggests sorption control of phosphorous by iron oxides as well.  Competition for sites is unlikely in light of the weak correlation between dissolved arsenic and phosphate (r2 = 0.782, not shown), which is consistent with the observations of McArthur et al., [2001].  Therefore, sorption is likely an important process, but may not be the only control on dissolved arsenic at the site. The positive relationship between iron and arsenic concomitant with an inverse relationship between iron and sulfate in aqueous samples (Figure 4.35 (a) and Figure 4.36(a)) preclude a mobilization hypothesis via oxidation of arsenic bearing iron-sulfides [McArthur et al., 2001; McArthur et al., 2004].  Furthermore, sediment extraction data in companion studies [R Beckie et al., 2006; R D Beckie et al., 2009], show appreciable correlation of solid phase arsenic with iron and manganese, and no correlation with sulfur.  However, the weak inverse correlation between dissolved arsenic and manganese Figure 4.35 (b)), suggests that arsenic release from manganese oxides contributes little to contamination in Gotra.  It is also worth noting that a stronger correlation between iron and arsenic might be expected than what is shown by the 4.0 Site Characterization and Conceptual Model Development 72 data if arsenic is released principally from iron oxide dissolution (i.e. if dissolution of iron oxides is and release of iron and arsenic was congruent).  A possible explanation for this disparity might be the consumption of iron in a secondary mineral phase such as siderite (FeCO3) [Horneman et al., 2004; Pedersen et al., 2006; Van Geen et al., 2004; van Geen et al., 2006b] or vivianite (Fe3(PO4)2•8(H2O)) [Harvey et al., 2002].  Indeed, Figure 4.37 shows the control of iron of siderite and vivianite saturation in Gotra groundwater, and Pal et al. [2002b] document the existence of siderite in sediments within the area.  The approach to a stable saturation index above zero as iron concentration increases, suggests that some of the ferrous iron initially released from oxide dissolution may be consumed in these phases. Some researchers have suggested that arsenic may be scavenged from waters as a result of sulfate reduction and co-precipitation with iron sulfides [Chakraborti et al., 2001; Korte and Fernando, 1991; McArthur et al., 2001; Pal et al., 2002b; Polizzotto et al., 2006; Rittle et al., 1995; Swartz et al., 2004].  In fact, Pal et al., [2002b] report framboidal pyrite (FeS2) in sediments from the Gotra-area aquifer sands.  Unfortunately, they did not measure the arsenic- sulfide content of these minerals so could not confirm a co-deposition hypothesis.  In the current dataset, PhreeqCI simulations predict that pyrite is far below saturation, suggesting that precipitation of sulfides is not favourable.  However, dissolved sulfide may not be represented in total sulfur analyses, as its high volatility likely prevented preservation during transport to the laboratory.  On the other hand, siderite has also been documented in Gotra sediments [Pal et al., 2002b], and it is only after dissolved sulfide is exhausted from solution, possibly through the precipitation of pyrite, that siderite becomes stable [Appelo and Postma, 2005].  It is therefore difficult to ascertain whether arsenic is being sequestered into sulfide precipitates in Gotra based on the aqueous data and documented mineral phases that are present in zones of contamination. 4.5.3. Relationship between Arsenic Contamination and Hydrogeology The above discussion indicates that numerous geochemical processes may be responsible for arsenic distribution at the site.  Given the vigour of groundwater flow implied by piezometric monitoring data (Section 4.3), it is intuitive that it would be a critical vehicle for transport of arsenic or solutes facilitating its release creating the geochemical variability at the site.  To obtain a satisfactory understanding of processes occurring in Gotra, aqueous geochemistry should therefore be interpreted in the context of groundwater flow.  This requires the delineation of hydrologic pathways connecting recharge areas to arsenic-rich zones. 4.0 Site Characterization and Conceptual Model Development 73 A notable feature about the Gotra site is that, in spite of the seasonal transience of groundwater flow, the geochemistry of the wells is stable from year to year.  This implies that the geochemical pattern in the subsurface may relate to a prevailing flowpath that is linked either to the contaminant or solute source responsible for release.  As arsenic release into groundwater is driven primarily by redox processes, it follows that the spatial distribution of dissolved arsenic will be representative of redox environments.  Such redox environments typically evolve along groundwater flow paths [Appelo and Postma, 2005], and are commonly considered in terms of microbially catalyzed terminal electron accepting processes (TEAPs) [Lovley and Chapelle, 1995]. In the following sections, site geochemistry will be examined in terms of spatial distributions of the potential reductant source (Section 4.5.3.1) and TEAPs (Section 4.5.3.2) in the aqueous profile to make a connection to groundwater flow. 4.5.3.1. Source of Organic Carbon for Microbial Respiration in Gotra With organic matter as the driving reductant [McArthur et al., 2001], groundwater chemistry may evolve along a flowpath originating from the organic carbon source [Appelo and Postma, 2005]. In the BDP aquifers, dissolved organic carbon may enter groundwater as recharge from reducing surface waters [Harvey et al., 2002], may originate from sediment such as peat [McArthur et al., 2001; McArthur et al., 2004] or may leach from organic matter co-deposited with the sediment [Meharg et al., 2006]. At the Gotra site, limited analyses of dissolved organic carbon (DOC) are available (Appendix E).  However, the existing data show that elevated DOC concentrations do tend to occur in closer proximity to the channel fill silts (Figure 4.38).  This suggests that the reductant may originate from this area, although the specific source is ambiguous.  Indeed, channel-fill sediments have been found to be organic-rich, with Corg up to 1.8 weight percent [Pal et al., 2009], and thus could conceivably supply the required organic matter for microbial metabolism. On the other hand, remnant ponds of the paleochannel overlie this area, which are rich in organics as a result of algae growth as well as from receiving domestic waste.  Remnant ponds above the channel-fill unit therefore provide an alternate source of biologically available organic carbon to the aquifer. Regardless of the particular source of the reductant, examination of redox zonation beyond the region of elevated DOC will be indicative of flow direction, as electron acceptors are 4.0 Site Characterization and Conceptual Model Development 74 sequentially consumed.  The following section therefore examines aqueous chemistry in the context of terminal electron accepting processes (TEAPs). 4.5.3.2. Redox Processes in the Shallow Aqueous Profile (TEAPs) Dissolved organic carbon is oxidized through microbially-mediated anaerobic TEAPs, [Lovley and Chapelle, 1995], either in the order of decreasing energy yield to bacteria, or by the local supply of a given electron acceptor ([Lovley and Chapelle, 1995; Pedersen et al., 2006].  With organic matter as the driving reductant, the evolution of water chemistry along a flowpath may be represented through redox zonation [Appelo and Postma, 2005], revealed through examination of the concentrations of reactants and products of various TEAPs.  The predominant TEAPs occurring in Gotra groundwater may therefore be determined by examining concentrations of electron acceptors (i.e. NO3 -, SO4 2-) or characteristic final products (i.e. NH4 +, Fe2+, Mn2+) within the aqueous profile. As discussed by Beckie et al., [2006], redox processes likely occurring at the site are denitrification, manganese, iron and sulfate reduction and methanogenesis.  Evidence that these processes are occurring in contaminated zones are found through comparison of cross- sectional contours of electron acceptors or redox reaction products (Figure 4.39 (a) to (e)) with the aqueous arsenic profile (Figure 4.34).  Denitrification is suggested by ammonia enrichment concomitant with low concentrations of nitrate in highly contaminated areas (Figure 4.39 (a) and (b).  These areas also have elevated concentrations of dissolved Mn(II) and Fe(II), indicating that reduction of iron and manganese oxides is favoured (Figure 4.39 (c) and (d)). Contaminated waters are effectively devoid of sulfate, suggesting that sulfate reduction has occurred.  Sulfate reduction is also supported by the documentation of framboidal pyrite within sediments in the area [Pal et al., 2002b].   Although not measured in the laboratory, the presence of methane in groundwater at affected wells was detected in the field through the ignition of gases released during sample collection, suggesting methanogenesis.  Additionally, mineral saturation indices of siderite and vivianite, which are characteristic phases of methanic waters [Berner, 1981], suggest that these phases are likely to precipitate from the most contaminated waters (Figure 4.36). As with dissolved arsenic and DOC, the concentrations of dissolved iron, manganese and ammonia all decrease with increasing distance from the scroll.  This suggests that contamination originates in water recharged through the channel-fill deposit, and evolves along a flowpath as organic carbon is consumed.  This is consistent with a source of organic carbon 4.0 Site Characterization and Conceptual Model Development 75 either originating from the remnant ponds of the paleochannel, or from the channel-fill deposit itself.  A schematic of redox fronts that would result under these circumstances is shown in Figure 4.40. 4.5.4. Hydrogeochemistry Summary The groundwater geochemistry in Gotra showed highly variable arsenic values over short distances.  Arsenic enrichment appears to be associated with waters with elevated Ca and HCO3 + CO3.  Dissolved arsenic was also correlated with iron, ammonia and phosphorous, but inversely correlated with sulphate, and only weakly correlated with manganese.  These characteristics of are suggestive of reductive release of arsenic into groundwater via desorption from, or dissolution iron oxide minerals. Documentation of framboidal pyrite in the area suggests that re-release of arsenic upon the introduction of oxic conditions to such assemblages is possible.  However, it has not been confirmed that arsenic is in fact hosted by this mineral phase.  This does not rule out the possibility of arsenic cycling in areas beyond the village (e.g. agricultural fields) that experience extensive water table fluctuation.  Furthermore, the aqueous geochemistry of the suite analyzed cannot unequivocally determine whether this process is active Despite the patchy nature of contamination, there appears to be a relationship between groundwater arsenic enrichment and the channel fill-deposit.  A geochemical gradient evolves in a direction perpendicular to trend of the scroll in the shallow sediments, where arsenic concentrations decrease with increasing distance from the village.  This gradient coincides with an evolving redox front, indicating a potential flowpath originating from the channel fill silt. Elevated concentrations of DOC in close proximity to the channel-fill deposit suggests that recharge through this unit may provide a source of organic carbon driving redox processes that release arsenic. 4.6. Conceptual Model and Numerical Modeling Goals The key hydrogeological and geochemical characteristics of the Gotra site are assembled in this section into a conceptual model of groundwater flow and contaminant transport.  These characteristics were used to guide numerical model development and calibration in Chapter 5, and to interpret modelling results in the context of arsenic transport.  The conceptual 4.0 Site Characterization and Conceptual Model Development 76 model for the site is summarized Section 4.6.1, and the specific modeling goals are discussed in Section 4.6.2. 4.6.1. Conceptual Hydrogeologic and Arsenic Transport Model A schematic of the flow system illustrating conceptualized seasonal flow directions is provided in Figure 4.41.  Groundwater enters the system predominantly as meteoric recharge during the monsoon through the ground surface within the village, agricultural fields and from ponds, and leaves the system through irrigation pumping during the dry season.  An additional recharge flux to the aquifer is likely supplied from oxbow lakes and wetlands, and may provide the aquifer with the volume of water necessary to replenish it each year.  There is potential for exchange of site groundwater with a regional flow system that connects the aquifer to distant surface water sources (i.e. the Hooghly River). The hydrostratigraphy of the immediate study area is characterized by three main units: the channel-fill silt underlying the village area, the main point bar aquifer to the north east and a layered floodplain unit to the south west (Figure 4.14).  The deeper Pleistocene materials have been conceptualized as a homogeneous unit, as site specific data below 30 m are limited.  Fine grained silty levee deposits over the north eastern portion of the study area form a leaky confining layer above the main aquifer.  The paleohorizon separating the Holocene and deeper Pleistocene sediments is marked by finer grained material comprised of organic material and a hard clay (kankar) in areas.  The thickness and continuity of low permeability material along this horizon is uncertain. Some basic characteristics of Gotra potentiometric levels are as follows: 1-village water levels rise simultaneously during monsoon flooding, and decline together as irrigation season progresses; 2-hydraulic gradients between pairs of wells are stronger in the vertical direction than horizontally; 3-water levels in the Hooghly River exceed village groundwater levels except for short periods following the monsoon, where gradients reverse and are directed from the aquifer to the river.  Piezometric data indicate that seasonal fluctuations cause groundwater elevations to vary from approximately -0.5 to 6.5 m asl.  These correspond to depths of 5 to 6 m below ground surface during irrigation season, and less than 1 m or flooding of low lying areas during the monsoon.  No prominent horizontal gradients are discernable from the data; however convergent flow towards shallow pumping wells is a salient feature during irrigation season. Localized downward gradients exist in the point-bar and channel-fill units, as meteoric recharge and pumping are the predominant sources and sinks to the aquifer, respectively. 4.0 Site Characterization and Conceptual Model Development 77 Geochemical data indicate that the contaminant release zone is closely related to the channel- fill deposit.  A northeast trending geochemical gradient in the point bar aquifer suggests that local flows transporting arsenic are directed away from this unit.  Flow in the direction of this gradient is consistent with a potentially evolving redox front originating from the channel-fill silt. However, it is uncertain whether the reductant associated with arsenic release originates from the sediments or remnant ponds of the paleochannel.  A schematic showing this geochemical evolution in relation to irrigation and observation wells, the main hydrostratigraphic units at the site, as well as local ponds is shown in Figure 4.42.  Interpreted flow directions coincident with the geochemical and observed head gradients are shown to illustrate the conceptualized transport of arsenic to the zone of highest concentration. 4.6.2. Specific Goals of Numerical Modeling The overall goal of numerical modeling is to develop a hydrogeologic representation of Gotra that includes the main features of stratigraphy and pumping that influence local arsenic transport.  Development of such a model can be used to interpret the current arsenic distribution patterns, as well as to speculate on the future migration of the contaminant.  The model will also be used to make inferences about the groundwater flow system prior to the establishment of widespread irrigation to highlight potential implications that recent perturbations to the flow system have had on arsenic transport.  Specific aspects of the field site to be examined with the numerical flow model are summarized in the following paragraphs. 4.6.2.1. Principal Groundwater Flow paths and Associated Travel Times Tracing the origin of groundwater from the arsenic enriched zone through simulated flow paths and examining associated travel times may provide insight into the contaminant source or release zone at the site.  Specifically, simulated flow paths will provide evidence as to whether contaminant accumulation in the village is the result of a localized, rather than an up-gradient source.  Examination of associated travel times will indicate the amount of time required for advective transport between localities of interest.  Of particular interest are paths related to the channel-fill silt, ponded areas and observed aqueous geochemical gradients.  Other flow paths of note are those that emerge from ponds to the north of the village, as well as those that travel from the contaminated zone to deeper Pleistocene materials.  All of these flow paths and their respective travel times will be examined under current flow conditions and compared to the pre- development case. 4.0 Site Characterization and Conceptual Model Development 78 4.6.2.2. Hydrologic Balance and Recharge Sources The hydrologic balance and recharge sources to the Gotra aquifer are of interest because they provide indications of potential origins of solutes or the contaminant to zones of accumulation. Comparison of the computed water balance for current conditions to a pre-pumping scenario may reveal whether recent perturbations to recharge and discharge have occurred that might affect arsenic release or transport.  Two particular areas of interest to be examined with the numerical model include the recharge to the ground surface and flow of shallow groundwater to the deep aquifer. 4.6.2.3. Average Residence Times within the Point-Bar Sand and Channel-Fill Silt Groundwater residence times are of interest for interpreting arsenic accumulation because they may indicate the time required for adequate flushing of the contaminant from solid and aqueous phases.  Contaminant flushing may be estimated coupling equilibrium desorption to average times to flush aquifer materials.  Average groundwater residence times will therefore be computed with the numerical model in order to estimate the number of pore-volumes required to flush the Holocene materials of contamination from present day. 4.0 Site Characterization and Conceptual Model Development 79  Figure 4.1: Google Earth Image of Chakdaha and surrounding region. Gotra is situated on a filled abandoned meander loop (chute cutoff) from eastward migration of the paleochannel. Annotations over image from Google Earth, (Image © 2008 DigitalGlobe, © 2008 Europa Technologies, Image © 2008 TerraMetrics)    Figure 4.5 4.0 Site Characterization and Conceptual Model Development 80  Station Name Distance from Gotra site (km) Ranaghat 20 Bonganon 24.5 Haringhata 4 Hooghly 22.5 Kolkata - Dum Dum 37 Kolkata - Alipore 64.5 Triberni  18.6   Figure 4.2: Weather stations and Triberni gauging station. Station coordinates were obtained from the NOAA website.  (http://www.ncdc.noaa.gov/oa/climate/ghcn-monthly/index.php).  4.0 Site Characterization and Conceptual Model Development 81  Figure 4.3: Site climate data. Monthly data for Nadia District range from 2004 to 2008 and historical (~1950-2000) monthly temperatures were measured near Kolkata.  4.0 Site Characterization and Conceptual Model Development 82  Figure 4.4: Historical precipitation data Annual historical precipitation from various weather stations shown with total annual rainfall in Nadia from 2004 to 2008.  4.0 Site Characterization and Conceptual Model Development 83  Figure 4.5: Google Maps image showing surface hydrologic features. Annotations over image from Google Maps, (© 2011 Google - Imagery © 2011 DigitalGlobe, GeoEye, Map data © 2011)   4.0 Site Characterization and Conceptual Model Development 84   Figure 4.6: Classical depositional model of a meandering stream. (after Allen, [1964]) This model was used to interpret stratigraphy in Gotra. The village area is located roughly on top of the thalweg of the former channel, which has been filled with fine grained material following abandonment.   Figure 4.7: Stratigraphic succession resulting from a meandering river depositional environment. (After Dickinson [1975]).  4.0 Site Characterization and Conceptual Model Development 85  Figure 4.8: Geological cross-section locations. Borehole data are provided in Appendix A. Wells plotted over image from Google Earth, (Image © 2006 DigitalGlobe, © 2006 Europa Technologies) Remnant Oxbow Ponds 4.0 Site Characterization and Conceptual Model Development 86  Figure 4.9: Longitudinal cross section AA’ showing abandoned channel sequence incising older alluvial deposits. The point bar and channel-fill sequence indicate southwesterly migration of the former meander. 4.0 Site Characterization and Conceptual Model Development 87  Figure 4.10: Cross-section BB’ running parallel to the meander scar. Former channel migration direction is out of the page  4.0 Site Characterization and Conceptual Model Development 88  Figure 4.11: Cross-section CC’  4.0 Site Characterization and Conceptual Model Development 89   Figure 4.12: Depositional events leading to the development of the stratigraphy in Gotra. (A) shows part of an active channel, where point bar deposits have been built with the increasing sinuosity of the meander.  The channel shift shown in (B) leads to the abandonment of the original bend with the subsequent formation of an oxbow lake (C), and a switch from high to low energy sedimentation.  Settling of suspended silts from the new-formed oxbow lake results in the deposition of finer grained sediments in (C).  (D) represents depositional events occurring after the channel has been filled.  Splay fans that develop under flooding conditions breach the surface of point bar as well as channel fill deposits to shallow depths.  4.0 Site Characterization and Conceptual Model Development 90  Figure 4.13: Summary of computed hydraulic conductivities. The chart includes all values for the aquifer estimated using the Theis solution, an average of values computed using linear regression methods on pump test data, and the Hvorslev slug test method using rising head data.  Channel fill silt values plotted were calculated using Hvorslev analysis of bail tests and Breyer grain size analysis.  Breyer grain size results are plotted for older alluvium (OA), overbank (OB) and lens (L) deposits.  4.0 Site Characterization and Conceptual Model Development 91  Figure 4.14: Stratigraphic interpretation and division of Gotra into hydrostratigraphic units. Block models were developed based on borehole logs and the depositional models of meandering river, oxbow lake and splay fan environments. (a) shows site in plan view, and (b) is a representative cross-section of the geology.  Overburden clays, silts and fine sands (OB/Splay) are deposits associated with splay events, whereas channel-fill silts (CF) that extend to depth were emplaced by low energy deposition from oxbow lakes.  Shallow sands represent upper and lower point-bar deposits (UPB and LPB) from a westward prograding point bar, deposited prior to channel abandonment.  Lag deposits are shown at the base of the sequence above the inferred paleosol layer, and undifferentiated sequences are labeled as earlier deposits.  (c) shows the division of the sediments into hydrostratigraphic units. Annotations shown over image from Google Earth (Image © 2008 DigitalGlobe, © 2007 Europa Technologies) 4.0 Site Characterization and Conceptual Model Development 92  * Approximate head  Figure 4.15: Hydraulic head measured in Gotra from 2006 to 2009. Water level highs correspond with periods of intense rainfall, showing the influence of the monsoon on aquifer recharge.  Lowering of heads at the end of an irrigation season has likely been enhanced due to cumulative extraction of groundwater in a given year.  Diurnal oscillations due to pumping are present at all times except for periods of intense rain.  It is likely that pumping occurs throughout the year, and that pumping signals are dwarfed by infiltrating water after storm events. 4.0 Site Characterization and Conceptual Model Development 93   Figure 4.16: Response of potentiometric levels to pumping over the course of a day. The irrigation schedule appears to be controlled heavily by the availability of electricity, and pumps typically run continuously while power is available.  Load-shedding events occur daily, which allows water levels to recover for short periods of time. 4.0 Site Characterization and Conceptual Model Development 94 (a)   (b)  Figure 4.17: Intensive water level measurement Data collected during 2007 and 2008 are shown in (a) and (b), respectively.  The plots show the daily drawdown and recovery patterns of the hydraulic head in response to irrigation pumping and load-shedding.  4.0 Site Characterization and Conceptual Model Development 95   Figure 4.18: Hydrochemical data collected over the 2006-2007 period. Head data collected at GSI0605 is shown for reference to the hydrogeologic year.  4.0 Site Characterization and Conceptual Model Development 96 (a) 2007 Field Visit (b) 2008 Field Visit  Figure 4.19: Head correlation matrices for observation wells. Data collected during 2007 and 2008 are shown in (a) and (b), respectively.  Scatter in the data is the result of irrigation pumping captured by the high-frequency sampling.  Correlation is generally high among wells in the village but less between village wells and GSI0609 and GSI0713. Correlation between GSI0713 and GSI0609 is also lower than what is observed in the village.  This is likely because groundwater flow near GSI0609 is influenced by the drawdown produced by more distant wells, as well as regional stratigraphy.  GSI0713 believed to be stratigraphically isolated from the village wells, contributing to a lack of correlation.  4.0 Site Characterization and Conceptual Model Development 97   Figure 4.20: Potentiometric response to pumping. Data from Well 4 and Well 50 are shown in (a) and (b) respectively.  Drawdowns are plotted against time normalized by the square of distance from the pumping well.  All data from wells screened within the point bar sand in the village plot on parallel lines, indicating that the aquifer is homogeneous. Response to Pumping at Well 4 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 1.00E-03 1.00E-02 1.00E-01 t/r 2  (min/m 2 ) [H o -H ] (m ) GSI0603 GSI0605 GSI0606 GSI0607 GSI0608 Response to Pumping at Well 50 0.00 0.05 0.10 0.15 0.20 0.25 0.30 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00 t/r 2  (min/m 2 ) [H o -H ] (m ) GSI0603 GSI0604 GSI0607 GSI0609 GSI0713 4.0 Site Characterization and Conceptual Model Development 98 (a)  (b)  (c)    Figure 4.21: Vertical hydraulic gradient observed between GSI0603 and GSI0604. Plots at the top display hydraulic head, and bottom plots show vertical gradients calculated from the data sets above.  Gradient plots show that when pumps are shut off and water levels recover, vertical gradients approach steady values and are directed downwards.  Data plotted in (a) and (b) were collected during the 2007 and 2008 field seasons, respectively, and data in (c) were collected from March 2007 to January 2008. The head versus time plot in (c) only displays data from GSI0603, as the time scale is too large to display differences in head between the two wells.  GSI0603 and GSI0604 Head vs Time During 2007 Field Season 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 01-Mar 01-Mar 02-Mar 02-Mar 02-Mar 02-Mar H e a d  ( m a s l) GSI0603 GSI0604 GSI0603 and GSI0604 Head vs Time During 2008 Field Season 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3 10-Feb 11-Feb 12-Feb 13-Feb 14-Feb 15-Feb 16-Feb 17-Feb H e a d  ( m a s l) GSI0603 GSI0604 GSI0603 Head* vs Time from March 2007 - January 2008 -1.000 0.000 1.000 2.000 3.000 4.000 5.000 6.000 7.000 Mar-07 May-07 Jun-07 Aug-07 Oct-07 Dec-07 H e a d * (m a s l) * Levels not corrected for barometric effects Vertical Hydraulic Gradient from GSI0604 to GSI0603 During 2007 Field Season -0.004 -0.003 -0.002 -0.001 0 0.001 0.002 0.003 0.004 0.005 0.006 1-Mar 12:00 2-Mar 0:00 2-Mar 12:00 H y d ra u lic  G ra d ie n t Vertical Hydraulic Gradient from GSI0604 to GSI0603 During 2008 Field Season -0.004 -0.003 -0.002 -0.001 0 0.001 0.002 0.003 0.004 0.005 0.006 10-Feb 11-Feb 12-Feb 13-Feb 14-Feb 15-Feb 16-Feb 17-Feb H y d ra u lic  G ra d ie n t Vertical Hydraulic Gradient from GSI0604 to GSI0603 from March 2007 - January 2008 -0.004 -0.003 -0.002 -0.001 0.000 0.001 0.002 0.003 0.004 0.005 0.006 Mar-07 May-07 Jun-07 Aug-07 Oct-07 Dec-07 H y d ra u li c  G ra d ie n t Figure 4.23 Figure 4.24 Figure 4.25 4.0 Site Characterization and Conceptual Model Development 99 (a)  (b)  Figure 4.22: Vertical hydraulic gradient observed in the channel-fill silt near GSI0605. (a) shows depth to water measurements made prior to purging for slug tests, and (b) shows the calculated gradients between the various measurement points from the data in (a).  In all cases the calculated gradient is directed downwards.  Figure 4.23: Potentiometric map of Gotra during a load-shedding event in March of 2007. Data from GSI0603, 06, 07, and 09 and reveal a gradient from the NE field towards the scroll as water levels recover from pumping in the evening.  Contour interval = 0.01m. Wells plotted over image from Google Earth, (Image © 2006 DigitalGlobe, © 2006 Europa Technologies) 0 1 2 3 4 5 6 7 D e p th  t o  W a te r (m ) Shallow Medium Deep 0 0.2 0.4 0.6 0.8 1 1.2 D o w n w a rd  G ra d ie n t Shallow to Medium Shallow to Deep Medium to Deep 4.0 Site Characterization and Conceptual Model Development 100  Figure 4.24: Potentiometric map of Gotra during a load-shedding event in February of 2008. Data from GSI0603, 07, 09 and GSI0713 only and reveal a flow direction from east to west. Contour interval = 0.02m.  Wells plotted over image from Google Earth, (Image © 2006 DigitalGlobe, © 2006 Europa Technologies)  2.88 4.0 Site Characterization and Conceptual Model Development 101  Figure 4.25: Potentiometric map of Gotra during the 2007 monsoon. Data from GSI0605, GSI0606, GSI0608 and GSI0714 and uggest that there is divergent flow from the scroll.  Contour interval = 0.01m Wells plotted over image from Google Earth, (Image © 2006 DigitalGlobe, © 2006 Europa Technologies) 4.0 Site Characterization and Conceptual Model Development 102  Figure 4.26: Schematic representation of flow directions from gradient interpretations Wells are projected onto a line directed N45ºE.  4.0 Site Characterization and Conceptual Model Development 103  (a)  (b)  Figure 4.27: Hooghly River stage data. (a) Hooghly River stage compared to piezometric heads in GSI0605, and (b) gradients between the river and village from ~May 2006 to ~March 2008. Monthly precipitation is plotted for seasonal reference.  4.0 Site Characterization and Conceptual Model Development 104 a)  (b)  Figure 4.28: Daily irrigation pumping duration. (a) shows daily fluctuations in head as recorded at GSI0605 from June 2006 to February 2008, calculated by subtracting a 24-hr moving average from the raw data.  Daily pumping durations (b) were calculated using the daily head oscillations displayed in (a).  -0.2 -0.1 0 0.1 0.2 Mar-06 Jun-06 Aug-06 Nov-06 Feb-07 Apr-07 Jul-07 Sep-07 Dec-07 Mar-08 Z e ro  M e a n  H e a d  ( m ) 0 4 8 12 16 20 24 Mar-06 Jun-06 Aug-06 Nov-06 Feb-07 Apr-07 Jul-07 Sep-07 Dec-07 Mar-08 P u m p in g  D u ra ti o n  ( h r) Average = 10 Average +1SD Average -1SD 4.0 Site Characterization and Conceptual Model Development 105   Number of Wells  8 Minimum Rate (L/s)  9.4 Maximum Rate (L/s)  13.6 Mean Rate (L/s)  11.3 Standard Dev. (L/s)  1.32   Figure 4.29: Gotra pumping rates. Rates were determined by monitoring the time required to fill a bucket with the outflow from various wells in the village and surrounding fields.  Pumping wells in the area are of similar 3 inch construction with the same electric submersible pump. Error bars shown in the plot show the variation in measured times at individual wells. 4.0 Site Characterization and Conceptual Model Development 106   Figure 4.30: Monthly water balance at the water table Water balances are calculated for the various land use zones at the site.  Note scale difference for ponds. -200 -100 0 100 200 300 400 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec (m m ) High Lands Precipitation Irrigation Evapotranspiration Runoff Recharge -200 -100 0 100 200 300 400 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec (m m ) Low Lands Precipitation Irrigation Evapotranspiration Runoff Recharge -200 -100 0 100 200 300 400 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec (m m ) Village Precipitation Evapotranspiration Runoff Recharge -300 -200 -100 0 100 200 300 400 500 600 700 800 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec (m m ) Ponds Precipitation Evapotranspiration Runoff Recharge 4.0 Site Characterization and Conceptual Model Development 107 (a)  (b)  Figure 4.31: ICPMS results for groundwater arsenic. Maps show horizontal arsenic variability above (a) and below (b) the paleohorizon separating Holocene and Pleistocene deposits.  Arsenic enrichment in shallower wells (a) appears to be associated with the channel-fill deposit. Wells plotted over image from Google Earth, (Image © 2006 DigitalGlobe, © 2006 Europa Technologies) 4.0 Site Characterization and Conceptual Model Development 108 (a)  (b)  (c)  (d)  (e)  (f)  (g)  (h)  Figure 4.32: Geochemical depth profiles of groundwater in Gotra, Figures also show the relationship to arsenic enrichment.  0 20 40 60 80 100 120 140 160 6 7 8 9 D e p th  ( m ) pH 0 20 40 60 80 100 120 140 160 0 1000 2000 3000 4000 EC (uS/cm) D e p th  ( m ) 0 20 40 60 80 100 120 140 160 0 10 20 Fe (mg/L) D e p th  ( m ) 0 20 40 60 80 100 120 140 160 0 5 10 15 20 NH4 (mg/L) D e p th  ( m ) 0 20 40 60 80 100 120 140 160 0 2 4 P (mg/L) D e p th  ( m ) 0 20 40 60 80 100 120 140 160 0 5 10 15 20 SO4 (mg/L) D e p th  ( m ) 0 20 40 60 80 100 120 140 160 0 10 20 30 Alkalinity (meq/L) D e p th  ( m ) 0 20 40 60 80 100 120 140 160 -2 -1.5 -1 -0.5 0 Log(PCO2) (bar) D e p th  ( m ) .  4.0 Site Characterization and Conceptual Model Development 109  Figure 4.33: Major element geochemistry of domestic and observation wells in Gotra. The plot also shows the relationship of well depth with water type.  Figure 4.34: Profile of arsenic contamination. The plot shows arsenic contamination in relation to geochemical sampling points and approximate location of the channel-fill silt.  Data have been projected on to the same SW-NE trending line as shown in Figure 4.26. The zone of arsenic enrichment appears to be associated with the channel-fill deposit. G S I0 6 0 9 G S I0 7 1 5 G S I0 6 0 3 , 0 4 G S I0 6 0 6 G S I0 6 0 8 G S I0 6 0 7 G S I0 7 1 4 G S I0 6 0 5 G S I0 7 1 3 4.0 Site Characterization and Conceptual Model Development 110  (a)  (b)  Figure 4.35: Relationship of arsenic to dissolved iron and manganese (a) shows that there is a positive relationship between the As and Fe content of Gotra groundwaters, indicating concomitant contaminant release with the dissolution of iron-bearing compounds.  (b) shows that Fe is only weakly correlated with Mn which suggests that arsenic release via reductive desorption from or reductive dissolution of Mn oxides has less of a control on arsenic contamination.  (a)  (b)  Figure 4.36: Relationship of iron to dissolved sulfate and phosphate. Arsenic content is colour-coded according to the legend in Figure 4.34.  Fe is weakly correlated with SO4 (r 2  = 0.02), and weakly correlated with PO4 (r 2  = 0.55).  Wells with high As also have low SO4 and high PO4. R² = 0.694 0 100 200 300 400 500 600 0 5 10 15 20 A s  ( p p b ) Fe (ppm) . 0 100 200 300 400 500 600 0 400 800 1200 1600 2000 2400 2800 A s  ( p p b ) Mn (ppb) . 0 20 40 60 80 100 120 140 0 5 10 15 S O 4  ( p p m ) Fe (ppm) . R² = 0.5508 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0 5 10 15 P O 4  ( p p b ) Fe (ppm) . 4.0 Site Characterization and Conceptual Model Development 111  Figure 4.37: Mineral saturation controls of iron in Gotra groundwater. Plots also show the relationship to arsenic enrichment.  These minerals are characteristic phases of methanic waters.  Figure 4.38: Concentrations of dissolved organic carbon (DOC) in shallow aqueous samples. -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 5 10 15 S I Fe (ppm) Siderite - FeCO3 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 5 10 15 S I Fe (ppm) Vivianite - Fe3(PO4)2:8H2O G S I0 6 0 9 G S I0 7 1 5 G S I0 6 0 3 , 0 4 G S I0 6 0 6 G S I0 6 0 8 G S I0 6 0 7 G S I0 7 1 4 G S I0 6 0 5 G S I0 7 1 3  4.0 Site Characterization and Conceptual Model Development 112 (a)  (b)   (c)  Figure 4.39: Profile contours of selected reactants and products of TEAPs. Mn(II) and Fe(II) concentrations were estimated using PhreeqCI.  G S I0 6 0 9 G S I0 7 1 5 G S I0 6 0 3 , 0 4 G S I0 6 0 6 G S I0 6 0 8 G S I0 6 0 7 G S I0 7 1 4 G S I0 6 0 5 G S I0 7 1 3 G S I0 6 0 9 G S I0 7 1 5 G S I0 6 0 3 , 0 4 G S I0 6 0 6 G S I0 6 0 8 G S I0 6 0 7 G S I0 7 1 4 G S I0 6 0 5 G S I0 7 1 3 G S I0 6 0 9 G S I0 7 1 5 G S I0 6 0 3 , 0 4 G S I0 6 0 6 G S I0 6 0 8 G S I0 6 0 7 G S I0 7 1 4 G S I0 6 0 5 G S I0 7 1 3 4.0 Site Characterization and Conceptual Model Development 113  (d)  (e)  Figure 4.38 (Continued): Profile contours of selected reactants and products of TEAPs. Mn(II) and Fe(II) concentrations were estimated using PhreeqCI.   Figure 4.40: Schematic SW-NE section through the village showing interpreted redox zonation.  G S I0 6 0 9 G S I0 7 1 5 G S I0 6 0 3 , 0 4 G S I0 6 0 6 G S I0 6 0 8 G S I0 6 0 7 G S I0 7 1 4 G S I0 6 0 5 G S I0 7 1 3 G S I0 6 0 9 G S I0 7 1 5 G S I0 6 0 3 , 0 4 G S I0 6 0 6 G S I0 6 0 8 G S I0 6 0 7 G S I0 7 1 4 G S I0 6 0 5 G S I0 7 1 3 G S I0 6 0 9 G S I0 7 1 5 G S I0 6 0 3 , 0 4 G S I0 6 0 6 G S I0 6 0 8 G S I0 6 0 7 G S I0 7 1 4 G S I0 6 0 5 G S I0 7 1 3 Methanogenic / Sulfate Reducing? Fe Reducing Mn Reducing NO  Reducing?3 - ?? 4.0 Site Characterization and Conceptual Model Development 114   Figure 4.41: Schematic of major flow paths during the wet and dry seasons.  4.0 Site Characterization and Conceptual Model Development 115  Figure 4.42: Conceptual model of arsenic release and transport in Gotra.  4.0 Site Characterization and Conceptual Model Development 116  Table 4.1: Geometric averages of parameter estimates. Detailed calculations are found in Appendix C.   Table 4.2: Summary of vertical gradients Values were calculated in the upper point bar sand (UPB) near GSI0603 and GSI0604, and in the channel fill silt (CF) near GSI0605.  Unit Method Kgeo (m/s) St Dev (m/s) Sgeo (m -1 ) St Dev (m -1 ) Aquifer Theis Pumping at W4 5.36E-04 1.07E-04 1.10E-04 2.32E-04 Theis Pumping at W50 5.15E-04 4.45E-04 4.11E-04 1.45E-03 Theis Superposition 8.18E-04 6.17E-04 2.40E-04 2.13E-04 Regression Analysis of Pump Test 1.12E-03 3.37E-04 - - Slug Tests (K from Hv; S from CPB, Rising Head Data) 6.86E-04 3.17E-04 1.22E-06 8.67E-05 Channel Fill Slug Tests (Hv Method) 1.90E-07 1.38E-07 - - Breyer Grain Size 1.02E-07 7.23E-08 - - Older Alluvium Breyer Grain Size 7.87E-08 2.30E-07 - - Overbank Breyer Grain Size 1.17E-07 2.04E-07 - - Lenses Breyer Grain Size 1.79E-07 2.76E-08 - - 4.0 Site Characterization and Conceptual Model Development 117  Table 4.3: Horizontal hydraulic gradients calculated between pairs of wells. Available data for the selected time-snapshots are shown in Figure 4.23, Figure 4.24 and Figure 4.25.  Horizontal gradients are generally lower than vertical gradients except for those involving GSI0609, GSI0713 and 14, which are comparable.  GSI0603 GSI0605 GSI0606 GSI0607 GSI0608 GSI0609 GSI0713 GSI0603 GSI0605 - GSI0606 -1.52E-04 - GSI0607 -3.50E-04 - -1.65E-04 GSI0608 - - - - GSI0609 1.12E-03 - 1.06E-03 8.89E-04 - GSI0713 - - - - - - GSI0714 - - - - - - - GSI0603 GSI0605 - GSI0606 - - GSI0607 6.64E-04 - - GSI0608 - - - - GSI0609 8.80E-04 - - 3.83E-04 - GSI0713 -5.96E-04 - - -1.41E-03 - -7.29E-04 GSI0714 - - - - - - - GSI0603 GSI0605 - GSI0606 - -2.91E-04 GSI0607 - - - GSI0608 - -1.64E-04 2.50E-04 - GSI0609 - - - - - GSI0713 - - - - - - GSI0714 - 1.52E-03 8.27E-04 - 2.31E-03 - - Towards 3 /1 /2 0 0 7 2 /1 3 /2 0 0 8 1 0 /1 /2 0 0 7 Away From 4.0 Site Characterization and Conceptual Model Development 118  Table 4.4: Monthly evapotranspiration and evaporation estimates. Values are in millimetres (mm) and are shown for the various land use regions in the study area.  Calculations are summarized in Appendix D.   Table 4.5: Monthly runoff estimates. Values are in millimetres (mm) and are shown for the various land use regions in the study area.  Calculations are summarized in Appendix D.   Month ETO ETHL ETLL ETvillage Ep Jan 88.10 57.85 101.52 83.69 108.17 Feb 111.83 48.10 134.19 106.24 137.54 Mar 147.98 132.88 177.19 140.58 176.49 Apr 178.44 135.54 169.11 169.52 205.82 May 181.21 84.57 190.27 172.15 206.96 Jun 153.45 162.33 161.21 145.78 180.90 Jul 131.13 154.46 149.68 124.57 158.72 Aug 124.13 148.96 148.96 117.92 151.60 Sep 125.34 125.03 150.35 119.07 151.38 Oct 125.28 78.23 120.38 119.01 144.52 Nov 108.51 113.47 113.93 103.08 122.75 Dec 88.57 89.03 93.08 84.14 104.41 Month ROHL ROLL ROV Rp* Jan 0.00 0.00 0.00 0.00 Feb 58.87 0.00 0.00 0.00 Mar 0.00 0.00 0.00 0.00 Apr 0.00 0.00 0.00 0.00 May 18.12 0.00 0.00 0.00 Jun 0.09 73.80 63.41 257.14 Jul 18.73 179.91 17.77 72.08 Aug 0.18 88.07 54.72 221.92 Sep 31.98 216.41 9.53 38.65 Oct 3.27 47.18 83.08 336.94 Nov 0.00 0.00 0.00 0.00 Dec 0.00 0.00 0.00 0.00 * Surface runoff from village areas to ponds 4.0 Site Characterization and Conceptual Model Development 119  Table 4.6: Monthly estimates of recharge to the water table. Values are in millimetres (mm) and are shown for the various land use regions in the study area  Month GHL GLL GV GP Jan 0.00 0.00 0.00 0.00 Feb 0.00 0.00 0.00 0.00 Mar 0.00 0.00 0.00 0.00 Apr 0.00 0.00 0.00 0.00 May 22.52 0.00 0.00 0.00 Jun 20.54 0.00 0.00 531.30 Jul 136.15 0.00 166.99 298.98 Aug 52.11 0.00 28.59 506.40 Sep 192.89 0.00 221.30 278.06 Oct 65.04 0.00 0.00 695.50 Nov 0.00 0.00 0.00 0.00 Dec 0.00 0.00 0.00 0.00 5.0 Numerical Groundwater Flow Model 120 5.0 NUMERICAL GROUNDWATER FLOW MODEL The three-dimensional finite difference program MODFLOW (Version 2000, Harbaugh et al, [2000]) and postprocessors MODPATH [Pollock, 1989] and ZoneBudget [Harbaugh, 1990] were used to investigate groundwater flow in the study area.  The graphical user interface Visual MODFLOW (Version 2009.1, SWS, [2009]) was used to develop the model and the proprietary WHS solver [WHI, 2003] was used to calculate the flow solution. The model was calibrated in transient mode to available head data through a trial and error process.  This was achieved by varying hydraulic conductivity, specific storage and specific yield of subsurface materials, as well as meteoric recharge applied to the ground surface within expected ranges (Section 4.0).  Calibration was carried out iteratively over two separate simulation periods (i.e. several days and 3 years) in order to ensure internal consistency of the model in both the short and long term. Model development, calibration and simulation of the current groundwater flow system are discussed in this chapter along with a hydrogeologic parameter sensitivity analyses.  Boundary conditions, specific yield, recharge and deep hydrogeology are all major uncertainties in the model because data to constrain these variables are scarce.  Elastic storage estimates from field data, anisotropy of the deep aquifer, as well as the existence of a low permeability layer separating shallow and deep flow systems are also uncertain, and are explored through sensitivity tests.  Finally, the simulation of a pre-development flow system, without irrigation stressors is presented. 5.1. Model Description and Inverse Modeling Stress Schedules The conceptual groundwater flow model for Gotra was developed using piezometric observations, Hooghly River stage data and estimates of potential recharge that incorporate evapotranspiration, irrigation return flow and surface runoff (Section 4.0).  A schematic of the major contributing groundwater sources and sinks is shown in Figure 4.41 for both wet and dry seasons.  As previously discussed, net vertical recharge is estimated between 312 and 257 mm/yr.  Total extraction of groundwater due to pumping is estimated at 433 mm/yr, and is removed from the aquifer over the months of November to May only.  Fluxes to or from rivers, lakes and wetlands in the vicinity are unknown and will be estimated using the calibrated model. Representation of the flow system in the numerical model involves the use of the site 5.0 Numerical Groundwater Flow Model 121 observational data in two inverse modeling scenarios: 1- a short-term transient (STT) model of dry season conditions and 2- a long-term transient (LTT) model that spans three years of available head data.  An internally consistent numerical model will independently provide good fits to observed data in both of these scenarios. The STT model tests the hydrostratigraphy by simulating village-well drawdown during the dry season in response to irrigation pumping.  No recharge from precipitation or runoff occurs during this period.  Head data collected at high frequencies (~ 1-2 minute intervals) from February 10 to 14 of 2008 were used in this scenario to help define hydraulic parameters. Accordingly, the purpose of the STT model was to 1- refine K estimates from field tests; 2- estimate aquifer storage properties; and 3- fine-tune the village hydrostratigraphic model (i.e. the connectivity of the point bar sands with floodplain deposits, and plausibility of a continuous low-permeability layer at the paleosol horizon). Parameters estimated through STT simulations fed into the LTT model, which simulates flow over the full data collection period (i.e. 2006 to 2009).  The final parameter values were determined by iteratively running STT and LTT scenarios until parameter sets that agreed within reasonable ranges of independent estimates (Section 4.0), and acceptable fits to observed head data were obtained in both models.  The LTT scenario contributes to most of our understanding of the flow system through the estimation of recharge and specific yield because 1- it encompasses the monsoon season (i.e. periods of recharge) and 2- spans a period of time sufficiently long to simulate water table fluctuations (i.e. draining of aquifer pore space).  These parameters were adjusted to match the observed seasonal amplitudes of piezometric heads. Groundwater elevations used for head targets in this scenario were daily values collected at the piezometers from May 2006 to March 2009. The overall framework of the numerical model and is summarized in Sections 5.1.1 to 5.1.5. These sections describe grid discretization, hydrostratigraphy, initial and boundary conditions, as well as time discretization.  Specific modifications to model stressors (i.e. recharge and pumping) necessary for each simulation scenario are also discussed. 5.1.1. Grid Design The extent of the model domain is shown in map view in Figure 5.1.  The region encompasses an area of 7.25 km2 (3.8 km x 1.4 km) around the village, is oriented lengthwise N45ºW and is discretized into a mesh of 109 rows and 55 columns. The domain area was designed to be 5.0 Numerical Groundwater Flow Model 122 large compared to the village area (~2.15 km2) in order to include two surface water bodies as local flow boundaries (Figure 5.4).  This approach was adopted in order to avoid imposing artificial local boundary conditions that could be influenced by pumping in the absence of any obvious physical boundaries, strong regional gradients, steep topography or major surface water bodies.  Uniform 20 m by 20 m gridblocks were defined in the immediate village area, and expanded by a factor of 1.2 towards the domain extents to a maximum area of 50m2.  For reference, model cells within the immediate village area are coloured grey in Figure 5.2. The grid extends to an elevation of approximately 240 m below sea level (depth ranging from 245 to 250 m bgl), which is assumed to be significantly deeper than the flow system of the arsenic affected aquifer.  The model bottom elevation is consistent with the lower aquifer boundary inferred from the hydrostratigraphic modeling of Mukherjee et al., [2007a], and with the depth of deltaic sediments to bedrock imposed in the hydrogeologic modeling studies of [Michael and Voss, 2009a; b].  Thirty-seven model layers of variable thickness were used to vertically discretize the domain.  A 1-metre vertical layer spacing was assigned from ground surface to a depth of approximately 30 m (i.e. the depth of the channel-fill, point bar package), and increased by a factor of 1.2 to a maximum spacing of 50m. Ground surface elevations were assigned using data modified from the DEM obtained from East View Cartographic (see Section 3.0, Figure 3.1).  Elevations of areas covered with trees appear to be overestimated in the DEM data compared to total survey station data, so a relationship between true elevation and DEM elevation was devised using a least squares regression to fit surveyed elevations to a logarithmic function, in order to dampen this effect (Appendix F).  The dampened elevation data was then imported into Visual MODFLOW and interpolated with kriging. 5.1.2. Model Hydrostratigraphy and Hydraulic Parameter Values Separate hydraulic conductivity zones for the channel-fill silt, aquifers, floodplain succession, confining unit, overbank deposits, top soil and ponds were assigned to areas of the domain that coincided with observed borehole stratigraphy (Appendix A and Section 4.2.2).  Ponded areas were assigned an arbitrarily high hydraulic conductivity value (i.e. 1 m/s) to allow water to flow freely through these cells.  Hydraulic conductivity zones specified in the model are shown in Figure 5.3.  Initial hydraulic conductivities of the hydrogeologic units where field test data were lacking were assigned based on representative literature values [Freeze and Cherry, 1979]. Specific storage values of 10-5 and 10-3 m-1 were initially assigned to sands and silt/clay units in 5.0 Numerical Groundwater Flow Model 123 the model, respectively, based on values considered to be representative of these materials [Batu, 1998].  The entire model domain was assigned an effective porosity value of 0.2.  This value was selected in order to be consistent with other researchers for these types of aquifers [Harvey, 2002; Michael and Voss, 2009a; b]  due to the lack of site specific porosity estimates. Specific yield for the entire domain was initially assumed to be 0.1 in order to fall within an expected range for porous materials [Freeze and Cherry, 1979]. Short term transient modeling was initiated using the conceptual model of hydrostratigraphy discussed in Section 4.2.2, which included a low hydraulic conductivity layer coincident with the paleosol horizon.  However, simulations using this geologic scenario required that the pumping rates of deeper irrigation wells be doubled from their measured values in order to match the seasonal decline in heads observed in the data.  Subsequent trial and error simulations showed that the same decline could be modeled if the low hydraulic conductivity layer was removed from the domain, and irrigation pumping rates were set to values measured in the field measured.  This geological configuration was therefore preferred over a scenario that includes a low permeability paleosol, since it could simultaneously reproduce heads using reasonable pumping rates.  The consequence of incorporating this low conductivity layer is examined as a sensitivity test in the LTT scenario and are discussed in Section 5.4.2.5. Hydraulic conductivity and recharge were modified within expected ranges based on field measurements, and storage parameters were varied within literature ranges during model calibration (discussed below).  The calibrated hydrogeologic parameter set is summarized in Table 5.3. 5.1.3. Initial Conditions Initial heads equal to the ground surface elevation were assigned to cells in both models.  Since head values at the beginning of the model observation periods do not represent steady state conditions, both simulations were run with a generic pumping schedule until field measured heads were matched.  This “spin-up” period effectively reduces the error that may be associated with selecting inaccurate initial conditions for a system that is not at steady state [Anderson and Woessner, 1991].  In the case of the LTT model, simulations were run using average annual pumping and recharge parameters until a dynamic steady state was reached between one hydrogeologic year and the next.  These dynamic cyclical conditions were also verified in sensitivity tests run with the LTT model to ensure that any measured differences from the base case were solely the result of the change in the specified parameter (i.e. not the adjustment of 5.0 Numerical Groundwater Flow Model 124 the system to initial heads at the beginning of the observation period).  In the short-term model, a baseline condition for the potentiometric surface was simulated until a dynamic steady state was reached for a 24 hour period. 5.1.4. Boundary Conditions No obvious physical or hydraulic boundaries in the Village area could limit the local flow system, so the domain was expanded beyond the village to include distant surface water bodies, as discussed in Section 5.1.1.  The model area extends far enough in the NE direction to include a large wetland (Bil), and in the SW direction to encompass part of an oxbow lake.  Specified head boundaries (Section 5.1.4.1), were applied to represent these bodies of water in the top model layer.  Head-dependent boundaries (Section 5.1.4.2) were assigned to NW and SE sides of the domain to allow for regional flow.  Recharge was applied to the remainder of the ground surface through various zones based on land use (Section 5.1.4.3) during initial calibration and seasonal simulations, and pumping wells were added to extract water from their surveyed locations in the fields, as well as from additional arbitrary locations (Section 5.1.4.4).  The bottom of the model was set as no-flow in all cases, as groundwater flow at depth is assumed to be essentially horizontal.  Specified head, head-dependent and recharge boundaries are shown in Figure 5.4, and pumping wells are shown in Figure 5.6. 5.1.4.1. Specified Head – Surface Water Bodies The large wetland (Bil) to the northeast of the village, and the oxbow lake in the southwest (Figure 5.1) were designated as constant head boundaries in model layer 1 (Figure 5.4).  As wetland and lake stage data were not available, water levels in these zones were estimated from surface topography.  These elevations were considered to represent average water levels of surface water bodies over the course of the year. As a constant head boundary fixes head within a given cell regardless of flow conditions, it may act as an infinite source or sink to the groundwater system.  In order to mitigate potential unrealistic conditions that may result (i.e. surging inflows during the dry season, or excessive discharge during the monsoon), a low conductivity zone was delineated in the cells below these boundaries.  This zone represents the silted bottoms commonly observed in nearby surface water bodies (e.g. Harvey et al., [2002]). 5.0 Numerical Groundwater Flow Model 125 5.1.4.2. Head Dependent Boundaries – NW and SE Edges of Model Domain In order to allow regional flow through the model, the northwest and southeast edges of the model domain were set as head dependent boundaries (Figure 5.4), using the General Head Boundary (GHB) package [Harbaugh et al., 2000].  This package calculates flow into or out of a cell from a source external to the model domain, which is proportional to the difference between the head assigned to the external water body and the head in the cell.  Flux through these boundaries is described by: Equation 5.1: )( hhCQ bbb  where: Qb = groundwater flux across the boundary [L 3/T]; Cb = conductance, [L 2/T]; hb = elevation of surface water body beyond the domain [L]; and h = head in the cell, calculated by MODFLOW [L]. The conductance is a numerical parameter representing resistance to flow between the model domain and external source or sink.  For example, conductance in the x-direction is represented by: Equation 5.2:  b b D zKy C    where: Δy = width of the grid cell in the y direction, [L]; Δz = width of the grid cell in the z direction, [L]; 5.0 Numerical Groundwater Flow Model 126 K = average hydraulic conductivity of the porous material separating the model domain from the external surface water source / sink, [L/T]; and Db = the distance between the model grid and the conceptualized external source / sink, [L]. The surface water bodies used for the external head values for shallow GHBs were the Hooghly River (~12km NW) and an oxbow lake (~5km SE), shown in Figure 5.5.  Head values for the Hooghly model boundary were taken from the average annual river stage measured at the Tribeni tidal gauging station (Section 4.1.3).  Head values for the SE oxbow lake were assumed to equal ground surface elevation of the DEM.  The external head values assigned to general head boundaries remained fixed over the course of the year. 5.1.4.3. Top Boundary – Recharge Four recharge zones were distinguished based on known land use (Appendix D and Section 4.4.2).  These zones are shown in Figure 5.4, and include: village or inhabited areas; ponds or surface water bodies; “High Lands” which are agricultural areas that cultivate numerous types of crops; and “Low Lands”, which are conceptualized as rice paddies exclusively (Appendix D). Delineation of these zones was guided by the Google Earth image as shown in Figure 5.1, where lighter-coloured fields were assumed to host multiple crop types, and the darker green fields assumed to represent rice paddies. The recharge boundary was set to zero in the STT model because the simulation period falls within the dry season.  During this time, the estimated recharge through vertical infiltration is zero, which includes potential irrigation return flow (Section 4.4.2.1).  Input data for recharge in the LTT scenario were based on values calculated from annual averages (Section 4.4.2.4) for a 10-year spin-up.  As monthly precipitation records were available for Nadia district from 2005 to 2009, model input recharge was computed for this period using actual values during this period. Water balance components are summarized in Table 4.4 and Table 4.5.  During model calibration, recharge rates were varied uniformly in order to match observed heads.  Details of calibration trials, including calibrated values are discussed in Section 5.2. 5.0 Numerical Groundwater Flow Model 127 5.1.4.4. Pumping Wells Twenty-five additional deep (>40m) irrigation wells were added to the model area in order to obtain a well density in the domain of ~5 wells per square kilometre in the Gotra area.  This well density was estimated based on the observation of a total of 19 pumping wells noted within a 1km radius of the village (Section 4.4.1) and is consistent with tallies quoted by (Nath et al., [2008b] for the Chakdaha area.  This is roughly consistent with the area of irrigated land in the region with poor well control, and thus simulates reasonable volumes of groundwater extracted from depth.  Additional wells were evenly emplaced (~300m grid) predominantly in the southwestern portion of the domain, where the number of wells was not known.  Figure 5.6 shows the locations of all pumping wells as they were included in model simulations. Pumping schedules for the two modeling scenarios were assigned based on different combinations of assumptions and observations.  For the STT model, observed pumping schedules were determined using observed controls for wells 4, 33 and 50, and an assumed diurnal schedule that coincided with interpreted load-shedding events for the remainder of wells (e.g. Figure 4.16).  In the LTT case, annual pumping schedules were assigned based on the assumptions that: 1- irrigation pumping only occurs in the months of November through May (inclusive); 2- pumps are off from June to October; and 3- the area is irrigated with 0.43m of groundwater annually (Section 4.4.1).  Whereas field-measured pumping rates were applied in the STT scenario, a pro-rated, semi-annual schedule (Qann) of 420 m 3/day was applied to the wells to simulate the steady extraction of this volume during irrigation seasons.  This was calculated as: Equation 5.3:  avann dQQ  where: d = average daily pumping duration (~42% of the day, Section 4.4.1) Qav = average pumping rate at individual wells (~1000 m 3/day, Section 4.4.1) 5.1.5. Time Discretization MODFLOW does not allow explicit specification of the number or length of stress periods in a transient simulation.  Instead, stress periods are automatically defined over time spans in which 5.0 Numerical Groundwater Flow Model 128 all system stressors (i.e. boundary conditions) are constant.  The number of time steps and the time step multiplier(s) however, may be user-defined. The total number of stress periods for the STT and LTT models are 31 and 107, respectively. This includes stress periods generated during model spin-up, in addition to periods where hydraulic head data were available.  No appreciable difference between simulation results (i.e. heads, mass balance) produced with the default value of 10 time steps compared to a solution calculated with 40 time steps was observed in either stress schedule.  Therefore, the default value of 10 time steps was used in each scenario.  The default time step multiplier of 1.2 was also used in both scenarios. 5.2. Model Calibration The model was calibrated to hydraulic head data through trial and error, alternating between the STT and LTT stress schedules.  Hydraulic conductivity and other input parameters (See Appendix F) were varied until the model simulated reasonable matches to observed heads at calibration targets, along with a water table elevation similar to what was observed in the field during the dry season, and anecdotal information during the monsoon.  Parameters that were identified to have significant impact on the flow system are discussed in Section 5.4.  A traditional method of model calibration is to minimize the normalized root mean squared (NRMS) error between observed and simulated target values provided that residual errors are normally distributed [Anderson and Woessner, 1991; Hill, 1998].  Additionally, an acceptable correlation coefficient between observed and simulated heads for a calibrated model should range between 0.7 and 1.0 [Spitz and Moreno, 1996].  An NRMS value of less than 10% indicates that the ratio of RMS error to total head loss in the flow system is small, and is a customary criterion for model calibration (e.g. NBLM, [2006]).  Parameters in the model were therefore altered until an NRMS of less than 10% and correlation coefficient greater than 0.7 were achieved in both modeling scenarios. The NRMS values of the calibrated STT and LTT models are 9.75% and 9.60% respectively. The correlation coefficient in the STT model is 0.81 and 0.96 in the LTT model.  Statistical analyses of residual errors in the calibrated model are summarized in Table 5.1, and residual error distributions are shown in Figure 5.7.  Simulated mass balance error was less than 0.5% for the time span of interest in each case (Appendix F), confirming that specified head change criterion of 0.01m for model convergence was consistently sufficient [Reilly and Harbaugh, 2004]. 5.0 Numerical Groundwater Flow Model 129 Time-series simulated and observed hydraulic heads for STT and LTT stress schedules are shown in Figure 5.8 and Figure 5.9, respectively.  The short term modeling scenario produces a good visual match to observed drawdown in three out of the five wells with available data. Discrepancies between calculated and observed values in GSI0713 and GSI0609 are most likely the result of localized heterogeneities or variable pumping which are not accounted for in the model.  Specifically, the signal in GSI0713 is poorly matched likely because the geology in this area is not well constrained.  In the area of well GSI0609, there may be an influence from pumping events distinct from those nearer to the village.  Accordingly, these events are not simulated with the applied pumping scheme.  Long-term simulations also produce good matches to observed heads.  However, disparity between the model and observation does occur, and appears to be the result of the inferred pumping schedule.  As the actual pumping scheme was not available (or reasonably obtainable), it was necessary to assume that all pumps were switched on and off uniformly at the beginning and end of the annual irrigation season.  Given the limitations of the inferred pumping schedule, the simulated match was considered reasonable.  A better representation of the system might be achieved if true pumping records were applied. The calibrated hydrogeologic parameter set is summarized in Table 5.3.  Hydraulic conductivities for the point-bar and channel-fill units are in agreement with estimates made based on field data (Section 4.2.1), and conductivities for auxiliary units are within expected ranges for their respective material type (i.e. sands or silts/clays, Freeze and Cherry, [1979]). The calibrated horizontal hydraulic conductivity value for the Pleistocene sands (1x10-5m/s) is low in comparison to field conductivity estimates for the shallow sands (Section 4.2.1). However, it falls within the range of regional estimates quoted by BGS and DPHE, [2001], which may also reflect conductivity contributions from layering of fine and coarse grained sediments on a regional scale [Michael and Voss, 2009a].  Calibrated values of specific storage are also within ranges of representative literature values [Batu, 1998].  A specific yield of 0.05 is on the low end of an acceptable range for sands [Freeze and Cherry, 1979], however it is consistent with literature estimates for shallow, fine-grained materials within the Ganges Delta [Ravenscroft et al., 2005], and is therefore considered a reasonable overall value for the package in the Gotra area. The calibrated model includes a highly anisotropic Pleistocene aquifer (Kh/Kv = 10 2) underlying the point-bar and channel-fill succession.  This conceptualization of deep hydrostratigraphy is consistent with that of Michael and Voss, [Michael and Voss, 2009a; b], which represents a large-scale equivalent hydraulic conductivity for layered Pleistocene materials.  On the village 5.0 Numerical Groundwater Flow Model 130 scale, it has been suggested that a low permeability layer separates the shallow meandering fluvial sequence from deep materials [Pal and Mukherjee, 2009].  However, as previously mentioned, model calibration did not require a low permeability unit at the base of the package, as a better match to observed data was achieved in trials by excluding it (Table 5.2).  This is significant because pumping from wells screened deep in the aquifer will impact downward flow from the shallow fluvial sediments. A summary of calibrated meteoric recharge values for the various land use zones is provided in Table 5.4.  Contributions from each zone were calculated for the entire model domain, as well as within the immediate Gotra area as shown in Figure 5.2.  The model predicts that approximately 177 mm of meteoric water recharges the entire domain annually, whereas total recharge to the immediate Gotra area is approximately 206 mm/yr.  These are equivalent to 13 and 15% of the average annual rainfall recorded for Nadia district from 2005 to 2009 respectively, and are consistent with the recharge estimates of 15% for the entire BDP [SWID, 1998]. 5.3. Current Groundwater Flow System The calibrated model simulates a flow system consistent with the conceptual model described in Section 4.6.  Flow directions and boundary fluxes are seasonally variable, driven principally through pumping and meteoric recharge.  Particle tracking with MODPATH [Pollock, 1989] indicates a specific origin of groundwater passing through the arsenic enriched zone, and predicts distinct groundwater travel times through the various hydrostratigraphic units.  Global and sub-zone water budgets computed with ZoneBudget Version 2.0 [Harbaugh, 1990] predict that regional sources of groundwater are important to the model water balance, and that downward flow from the Holocene materials into the deeper aquifer is significant.  Similar bulk residence times (i.e. the time to flush one pore volume of groundwater) are predicted for groundwater passing through the channel-fill and point-bar sands, suggesting comparable flushing rates for these units under the existing flow conditions.  These aspects of the current groundwater flow system are discussed in the subsequent paragraphs. 5.3.1. Water Table Configuration and Village Flow Directions The simulated water table and potentiometric contours for wet and dry seasons are shown in Figure 5.10 and Figure 5.11 respectively.  Unit vectors are also plotted to illustrate the predominant flow directions occurring in each season, and geochemical data are also shown to 5.0 Numerical Groundwater Flow Model 131 provide reference to the arsenic enriched zone.  At the peak of the monsoon, the water table is predicted to be a subdued replica of topography within ~1.5 m of the ground surface.  In the area of interest, groundwater travels downward through the channel-fill silt, and horizontally north-eastward through the point bar sand, then upwards towards local low lying areas. Southwest of the channel fill-silt, groundwater travels vertically downwards through the floodplain succession, and then moves horizontally through the sandy unit either to the northeast, or southwest.  A local flow divide is produced as a result, which is consistent with the conceptualization of shallow recharge to the highest elevation of the levee.  Modeled groundwater velocity in the channel-fill silt is ~0.1 cm/day, and flow through the point bar sands ranges from approximately 1 to 10 cm/day.  Groundwater velocity (i.e. seepage velocity q/n) in the Pleistocene aquifer is computed to be ~0.1 cm/day during the monsoon. During the dry season, the water table recedes to a maximum depth of ~5 m below the ground surface.  The model predicts that hydraulic connection between ponds and the aquifer is not always maintained throughout the dry period.  However, ponds underlain by lower conductivity materials (i.e. the channel-fill silt) do not become desaturated as often as those underlain by sandy materials (i.e. the point-bar sand).  Water levels in ponded areas near the village are typically ~3 to 5 m asl, which at times are predicted to be perched during the dry season. Figure 5.11 shows that vertical groundwater flow through the domain dominates during irrigation season, except on a horizon that is level with the majority of well screen centres.  In the village area, the general direction of flow through the channel-fill silt is downwards, and convergent towards shallow irrigation pumps within the point-bar sands.  This result is also consistent with field observation.  Flow divides are induced at depth as a result of irrigation pumping, generating distinct capture zones.  Maximum flow velocity within the channel-fill unit during this time is similar to the modeled values during the monsoon (~0.1 cm/day).  On the other hand, flow within the sandy units increases up to a factor 10 times the rates modeled during the monsoon. Comparisons of simulated vertical gradients for the channel-fill unit and point bar sand to field observed values over time are shown in Figure 5.12.  Although slightly lower, modeled gradients agree qualitatively with field observations, and predict downward flow at the piezometers nests.  Discrepancies between modeled and observed values are likely the result of applying inferred pumping schedules. MODPATH [Pollock, 1989] was used to track the movement of groundwater entering the model at the water table in the immediate village area over the course of the long-term simulation.  A 5.0 Numerical Groundwater Flow Model 132 cross-sectional schematic illustrating these results is shown in Figure 5.13.  The diagram shows that pumping induced flow paths and resultant capture zones determine the fate of recharge to the village over time.  This is expected given the significant increases in velocity within aquifer units that occur during the pumping season, which spans approximately seven months of every year.  Some noteworthy features of this schematic are that 1- water entering the arsenic enriched zone originates from a path exiting the channel-fill silt; 2- the majority of groundwater within the point-bar aquifer exits to shallow (i.e. ~25m deep) irrigation wells and 3- shallow groundwater that does not report to shallow wells is drawn to the screens of deep (~60 m bgl) irrigation wells. Forward particle tracking with MODPATH allows the computation of groundwater travel times from specified points of origin.  Several pathlines of interest and their respective travel times are highlighted in Figure 5.13.  Summation of pathline segments indicate that between 20 and 26 years is required for groundwater travel through the channel-fill silt, and 4 to 9 years is required within the point-bar sand.  Flow to deep irrigation wells from the paleosol horizon requires approximately 22 years, and between 44 to 48 years from the water table, depending on hydraulic conductivity values of overlying materials. An important feature of note from MODPATH simulations is that the tracking of particles from ponds is discontinuous and limited by the annual water table fluctuation.  Particles near the ground surface were often stopped upon the development of unsaturated conditions and drying of model cells.  Accordingly, the exact flow paths and travel times from these points and through this periodically unsaturated zone are approximate.  In particular, travel times from the pond to permanently saturated areas are likely underestimated. 5.3.2. Groundwater Budget and Recharge Sources A summary of the simulated groundwater budget for the period spanning March 2006 to March 2009 is provided in Table 5.5 for both the full domain and the immediate village area (i.e. as shown in Figure 5.2).  A net loss of groundwater to aquifer storage indicates that the system does not reach a dynamic steady-state during the period over which the annual mass-balance was summed (i.e. the observation period that has a variable recharge schedule).  This indicates that the aquifer is not totally replenished each year as a result of variable precipitation.  The mass balance for the full domain indicates that groundwater removed from the system through pumping (~355 mm/yr) will be balanced by roughly equal amounts meteoric recharge (~180 mm/yr) and groundwater pulled from regional sources (~215 mm/yr) on an annual basis.  Lesser 5.0 Numerical Groundwater Flow Model 133 contributions to groundwater recharge originate from boundaries simulated as surface water bodies (i.e. the Bil and oxbow lake).  In fact, the computed net loss of groundwater into storage is comparable in magnitude to the overall inward flow to the model from constant head boundaries.  The contribution to aquifer recharge from this boundary type is therefore considered negligible in the current flow system.  The mass balance for the Gotra sub-zone also shows that surface recharge and pumping wells are important sources and sinks to the area of interest.  Computations for this sub-zone also indicate that lateral inflow to the area of interest is significant, and that there is considerable flux of groundwater to the deep aquifer.  These results are reflective of a dynamic flow regime driven by irrigation pumping, where model boundary conceptualization at the village scale is consequentially complex. Seasonal flux dynamics shown in Figure 5.14 and Figure 5.15 illustrate how the hydrogeologic year is divided by pumping and meteoric recharge.  In order for the system to achieve mass balance at every time step, water is removed from storage at the beginning of irrigation season, and is returned as the aquifer refills during monsoon periods.  Modeled domain boundaries behave both as groundwater sources and sinks over the course of the hydrogeologic year, providing recharge to the aquifer during irrigation season, and then discharging groundwater to balance infiltrating monsoon rain.  Simulated fluxes between the lateral edges of the village area and the remainder of the domain demonstrate similar dual behaviour throughout the year, acting as both recharge and discharge zones to the village aquifer.  Downward flow of groundwater to the Pleistocene aquifer is most pronounced during irrigation season, concurrent with the widespread operation of deep pumps for agriculture (i.e. most irrigation pumps are screened within this unit). The breakdown of meteoric recharge according to land use zone reported in Table 5.4 indicates that the greatest contribution of rainfall to groundwater recharge to the full domain infiltrates through elevated zones, consistent with water balance estimates in Section 4.4.  These include agricultural fields designated as High Lands (83 mm/yr), and the village areas (63 mm/yr) which are situated on natural levees.  This is also true for recharge to the immediate Gotra area, which receives a total of 81 mm in inhabited areas, and 58 mm in high lands annually.  Less significant contributions to aquifer recharge infiltrate the domain through the Low Lands (i.e. paddy fields, 24 mm/yr) and ponded areas (7 mm/yr).  In the immediate village area, a comparatively greater fraction of area occupied by ponds results in relatively more significant volumes of pond recharge entering the model (i.e. 52 mm/yr) compared to the entire domain.  Similarly, a smaller Low-Land fraction predicts that lower relative recharge fluxes through this zone in the Gotra area (i.e. 15 mm/yr) compared to the full model domain. 5.0 Numerical Groundwater Flow Model 134 The annual amount of recharge applied to the model domain that originated as irrigation water from agricultural lands is approximately 12 mm, or 7% of total recharge (Table 5.7), consisting of 11 mm/yr from High Lands and 1 mm/yr Low Lands.  Within the immediate Gotra area, irrigation return flow is even less significant as a result of comparatively less agricultural lands in this sub-zone. 5.3.3. Groundwater Residence Times Bulk groundwater residence times (i.e. the time required to flush one pore-volume of groundwater) in the channel-fill and point-bar units may be computed by dividing the respective material volumes by the annual volumetric flux of groundwater passing through each [Harvey et al., 2006].  Grid cell volumes comprising each unit were summed in order to obtain material volumes, and ZoneBudget Version 2.0 [Harbaugh, 1990] was used to calculate the yearly groundwater fluxes through both regions.  Average residence times calculated for the channel- fill silts and point-bar sands using the LTT schedule are 7.9 and 4.0 years, respectively (Table 5.9).  Assuming that pumping began in the early 1970s, the maximum amount of times the channel-fill unit could have been flushed is five, and 10 for the point-bar sands. It is important to note that the bulk residence time computed for the channel-fill silt is likely biased towards a high value due to the geometry of the deposit.  Flows through model cells in shallow zones to the north-east of the channel thalweg contribute most to this bias as a result of adjacent cells with higher hydraulic conductivity values (i.e. the point-bar deposits and top soil). Furthermore, a flushing time as it relates to arsenic removal from this unit is best considered in a direction coincident with flowpaths that lead to the arsenious zone.  Using Darcy‟s Law, a vertical conductivity of 2.5x10-8 m/s, a path length of 20 m (i.e. the thalweg thickness) and assuming a downward hydraulic gradient of 0.2, the estimated flushing time through this unit would be closer to 25 years.  This estimate implies that 160 mm of recharge enters the channel- fill, which is consistent with modeling results (see Table 4.1).  In addition, a flushing time of 25 years is consistent with estimates made using particle tracking in MODPATH, as shown in Figure 5.13.  The simulated bulk residence time for the channel-fill silt must therefore be interpreted with caution when considering arsenic accumulation to the specific hotspot in Gotra. 5.4. Sensitivity Analysis The LTT stress schedule was used to examine the model sensitivity by altering various input parameters within reasonable ranges of values.  The impacts of varying each parameter on 5.0 Numerical Groundwater Flow Model 135 calibration statistics (i.e. NRMS, RMS, residual mean, absolute residual mean and r2) were inspected for scenarios that either met or improved the calibration criteria discussed in Section 5.2.  The effects on the system for each of these scenarios were then assessed in terms of changes to groundwater fluxes within the Gotra flow regime (i.e. sub-zone in Figure 5.15), as well as bulk residence times of the channel-fill and point- bar units.  Parameters that were varied include applied recharge, point-bar and channel-fill conductivities and specific storage, deep aquifer conductivity and anisotropy, the conductivity of lake and wetland bottoms, and specific yield.  The effect of including a low conductivity unit coincident with the paleosol horizon was also investigated.  A summary of sensitivity tests is provided in Appendix F along with the range of values considered for each input parameter.  The effects of parameter variation on model calibration and flow system are discussed in Sections 5.4.1 and 5.4.2, respectively. 5.4.1. Effect on Parameter Variation on Model Calibration The effect of varying each of the input parameters listed in Appendix F on modeled heads is shown in Figure 5.16.  The plot indicates that modeled heads are sensitive to most hydrologic parameter values, as the majority of test cases appear to depreciate the goodness of fit to observation data.  Several scenarios however appeared to improve the match to observed heads.  These include variations to channel-fill conductivity and storage, point-bar storage deep aquifer anisotropy and specific yield (Table 5.8).  The hydrogeologic parameters specified in these scenarios are considered as potential alternate calibration sets.  The impacts of these parameters on the groundwater flow system are therefore explored in the following section. Residual statistics for all sensitivity scenarios run with the LTT schedule are shown in Appendix F. 5.4.2. Impacts of Parameter Variation to the Gotra Groundwater Flow System The sensitivity of general flow directions, water balance and groundwater residence times that characterize the base case flow system were assessed using the alternate calibration scenarios presented in the previous section.  Simulated fluxes to lateral boundaries and the bottom of the Gotra sub-zone (Figure 5.2) as well as bulk groundwater residence times were therefore computed for these scenarios and compared to the base case. Scaled sensitivities were computed in order to compare the relative influences of the varied parameters on the calculation of simulated values.  Sensitivities are expressed as percent 5.0 Numerical Groundwater Flow Model 136 changes in simulated values with respect to the percent changes in parameter values from the base case [Hill, 1998]: Equation 5.4: %100 %100      j jk i ik ijk p p x x S Where: Sijk = scaled sensitivity of simulated fluxes or residence times [-]; xi = base case simulated value of the i th quantity (i.e. flux or residence time), [m3/yr; yr]; Δxik = difference between the k th sensitivity simulation and base case simulated value of the ith quantity (i.e. flux or residence time), [m3/yr; yr]; pj  = best fit value of the j th parameter (e.g. specific yield); and Δpjk = difference between the k th sensitivity simulation and best fit value of the jth parameter (e.g. specific yield). Greater absolute values of Sijk indicate a greater influence of varying a given parameter on model results.  A comparison of model sensitivities showing the effect of each parameter is shown in Figure 5.17 for fluxes between the Gotra sub-zone and underlying Pleistocene aquifer and lateral boundaries, and in Figure 5.18  for the effect on bulk residence times.  Percent changes to groundwater fluxes, as well as the simulated residence times are listed in Table 5.9, and a discussion of each of these effects is presented below. 5.4.2.1. Channel-Fill Hydraulic Conductivity and Storage The most important impact to flow in the Gotra area incurred by varying channel-fill hydraulic conductivity and storage is a change in average residence times within this unit (Figure 5.18). 5.0 Numerical Groundwater Flow Model 137 Fluxes to southern lateral extents of the zone are only mildly affected, and even less significant changes in flux are observed at the northern extents and the bottom (Figure 5.17, Table 5.9). An increase in channel-fill conductivity by a factor of 5 decreases groundwater residence times by ~38% (i.e. down to 4.8 years), while decreasing conductivity by a factor of 10 results in a 264% increase, or residence time of 28.7 years.  Increasing specific storage by factors of 5 and 10 lowers average residence times within the silt by ~ 30 and 40% respectively.  Additional slug tests within this unit at separate locations would help to further constrain this estimate. 5.4.2.2. Point-Bar Specific Storage Groundwater flow through the Gotra region appears to be insensitive to an increase in point-bar specific storage by a factor of 5 (Figure 5.17, Figure 5.18, Table 5.9).  Accordingly, the range of values of this parameter permitted by model calibration criteria has negligible consequence for flow predictions in the Gotra area. 5.4.2.3. Deep Aquifer Anisotropy Inverse modeling studies and driller log analyses conducted by Michael and Voss [2009a] have shown that deep sediments (i.e. > 100m ) within the BDP may be characterized by high values of vertical anisotropy (Kh/Kv.~10 4).  Although the major implications of extremely high anisotropy of deep aquifer sediments are expected to affect more regional flow (i.e. over tens to hundreds of kilometres), the sensitivity of the current model to this parameter was examined.  Flow paths for this scenario are plotted in Figure 5.19.  Test results indicate that decreasing the anisotropy of the deep aquifer by increasing Kv increases the magnitude of all fluxes across the extents of the Gotra sub-zone (Table 5.9).  This is because a more isotropic pumped resource allows for increased flow from shallow zones to well screens (which are outside the Gotra sub-zone). Considerably more groundwater enters the village area through the northeast, northwest and southeast edges, which is balanced by greater amounts of water flowing to the Pleistocene aquifer.  Implications of this result are primarily changes to the extent of well capture and groundwater sources to the deep aquifer.  However, there are no appreciable changes in residence times within the channel-fill or point-bar units in this scenario.  Figure 5.19 indicates that the only essential difference in this scenario from the base case is a lowering of travel time from shallow materials to deep irrigation well screens of approximately 25% (i.e. from 22 years in the base case and 14 years in the sensitivity case). 5.0 Numerical Groundwater Flow Model 138 5.4.2.4. Specific Yield Model results are mildly sensitive to increasing specific yield by a factor of 2 from the base case.  Simulated residence times and groundwater fluxes between the Gotra area and the remainder of the domain are not significantly impacted by this variation (Figure 5.17, Figure 5.18, Table 5.9) 5.4.2.5. Effect of Including a Low-Conductivity Layer at the Paleosol Horizon Although the preferred geological representation of the area excludes this layer, a sensitivity test was run to examine potential consequences of having a low permeability unit at the base of the point-bar and channel-fill sediment package.  Flow paths for this scenario are plotted in Figure 5.20.  The model predicts that the existence of such a unit will reduce total flow through shallow materials, as well as inhibit downward flow of groundwater to deep wells.  The bulk residence time calculated for the channel-fill unit in this scenario (6.5 years) is essentially unchanged from the base case, however, flushing of the sands slows down by ~40% (i.e. residence time of 5.6 years).  Figure 5.20 shows that the primary changes to flow paths compared to the base case that result from this scenario are the expansion of well capture in the shallow materials, and an increase in travel times from shallow sediments to screens of deep irrigation wells (i.e. by a factor of approximately 2.5). 5.5. Pre-Development Scenario To simulate the behaviour of groundwater flow without the effects of irrigation, the LTT model was run with pumping wells deactivated.  The recharge schedule applied in the base case was also applied to this scenario in order to isolate the effects of pumping and thereby allow for direct comparison of mass balance results.  Thus, the scenario presented here represents the would-be groundwater flow system for 2006 to 2009, had irrigation agriculture not developed in the area. A major assumption implicit to this scenario is that boundary condition representation in the model with pumping is also valid for a scenario without pumping.  In fact, initial trial and error simulations revealed that heads would reach approximately 1 to 2 metres above the ground surface during the monsoon season.  This result indicates that recharge calibrated to current conditions is excessive for circumstances where the aquifer is not pumped.  In pre-development 5.0 Numerical Groundwater Flow Model 139 times, the model suggests that there would be a lot more rejected recharge (i.e. surface runoff) than under current conditions. To improve simulations of the flow system without pumping, a drain boundary is added to the top model layer to ensure that heads do not exceed the elevation of the ground surface.  The mathematical formulation of a drain boundary is similar to that of a general head boundary described in Equation 5.1, except outflow of groundwater is simulated.  When modeled heads fall below the specified head value at the drain boundary (i.e. the ground surface elevation), no outflow is simulated.  Additionally, conductance is set arbitrarily high to allow groundwater to freely discharge to these boundaries.  Groundwater flux to this boundary is conceptualized as overland flow, or a reduction in meteoric recharge to the area. The following sections present the results obtained for a scenario simulated without pumping. As with the irrigation scenario, these results are discussed in terms of flow paths (Section 5.5.1), groundwater budget and recharge sources (Section 5.5.2), as well as average residence times (Section 5.5.3).  Direct comparisons to the flow system simulated with irrigation are also noted. 5.5.1. Principal Groundwater Flow paths and Travel Times Groundwater flow directions simulated with MODPATH for the pre-development scenario are shown in section view in Figure 5.21, and are similar to the vector field produced under pumped conditions during the monsoon.  Without pumping, the model indicates that groundwater reaching the arsenic enriched zone first has travelled along a downward flowpath through the channel-fill silt and has entered the domain from the ground surface.  Flow through the point-bar sand is generally directed downwards to a depth of about 10 m, then north-eastward through the domain.  Within the floodplain succession, groundwater travels downwards in low permeability materials, and moves in a south-westward direction through sandy materials. From the arsenic hotspot, groundwater is predicted to flow downwards to the Pleistocene aquifer. Figure 5.21 shows that travel times between points of interest have increased considerably in comparison to the scenario with pumping.  In the channel-fill silt, the time required for groundwater to reach the arsenic enriched zones ranges between 30 and 55 years, and 130 years is required for it to reach a depth of approximately 60 m (i.e. the depth of irrigation wells within the Pleistocene aquifer).  These represent increases by factors of approximately 1.5 to 2 5.0 Numerical Groundwater Flow Model 140 and 6, respectively.  Direct comparison of travel times through the point-bar sands cannot be made since simulated flow paths are vastly different from the base case.  However, the computed travel times for groundwater entering from the ground surface as shown in Figure 5.13, and exiting the shallow aquifer to the Pleistocene zone are significantly greater (i.e. a factor of 10) than the time required for groundwater to reach shallow well screens in the pumped case. 5.5.2. Hydrologic Balance and Recharge Sources The net groundwater budget for the modeling scenario without pumping is provided in Table 5.6.  These constituents are also compared graphically to the base case in Figure 5.24 and Figure 5.25 for full domain and Gotra sub-zone, respectively.  The overall mass balance for the scenario without pumping predicts that total flow through the domain and Gotra area are lower the base case.  In the absence of a major sink to pumping, a considerable amount of the recharge flux is rejected at the ground surface and must exit the through drain boundaries in the top model layer to avoid excessive ponding.  For both the full domain and Gotra area, this is equivalent to nearly 50% of the recharge flux to the ground surface (Table 5.6). Examination of global mass balance components over time indicates that both surface water and regional flow boundaries remain discharging throughout the year (Figure 5.22), contrary to the dual recharge/discharge behaviour in the pumped case (Figure 5.14).  In the immediate Gotra area, the most significant change to flux dynamics is a reduction in amount of water drawn down to the Pleistocene aquifer, which appears to be restricted to periods of heavy rainfall (Figure 5.23). 5.5.3. Groundwater Residence Times and Aquifer Flushing Simulation of flow without pumping predicts that the average residence times within the point- bar sand will be approximately 7.3 years, and 10.0 years within the channel-fill silt (Table 5.9). These values correspond to increases in flushing times of 27 and 83%, respectively. Accordingly, the model predicts that the point-bar would have been flushed about 1640 times, and the channel-fill unit approximately 1200 times since the beginning of the Holocene. 5.0 Numerical Groundwater Flow Model 141 5.6. Implications for Arsenic Contamination of Groundwater in Gotra The model results presented in this chapter may be used to interpret current arsenic distribution patterns, as well as to predict future migration of arsenic at the site.  In addition, a comparison of the current simulated flow regime to a system that predates the advent of irrigation highlights several consequences that recent perturbations to the flow system may have had on arsenic transport.  The following sections use model results to briefly speculate on the potential origin of contamination at the site, as well as discuss several implications that pumping may have had for arsenic accumulation. 5.6.1. Principal Groundwater Flow Paths and Travel Times A key result of the modeling study presented in the preceding sections of this chapter is that groundwater flowing to the arsenic-rich zone originates from a local recharge area to the ground surface.  The predominant flow direction also coincides with decreasing geochemical gradients measured in piezometers at the site.  These observations are true for both pumped and pre- pumped cases, and indicate that arsenic and/or the solutes driving its release are provided from a local source rather than a distal source that is hydrologically upgradient. The principal groundwater flow paths entering the contaminated zone have in all cases first passed through the channel-fill silt.  If arsenic release is the result of reduction driven processes, flow to the arsenic hotspot suggests that DOC driving arsenic release could have been derived from this deposit.  Indeed, these silts have been found to contain high amounts of organic matter (1 to 6 wt% [Pal and Mukherjee, 2009]), which is favourable for bacterial population [Meharg and Rahman, 2003] and the generation of reducing groundwater environments that liberate arsenic.  Naturally slow flow through these sediments due to their comparatively low permeability likely contributes to the generation of reducing groundwater in this area, which is supportive of hypotheses of some researchers (e.g. Polizzotto et al., [2008]; Weinman et al., [2008]). The observation of flow emerging from the channel-fill silt to the contaminated zone does not rule out the potential contribution of DOC originating from ponds, which has been shown to drive arsenic release at other sites (e.g. Harvey et al., [2002]).  As discussed in Section 4.1.3, ponds of both natural and of anthropogenic origin are ubiquitous across the landscape.  In addition to DOC from pond algae, the intermittent use of these ponds as domestic waste repositories may also contribute biologically available organic carbon to the aquifer.  In the 5.0 Numerical Groundwater Flow Model 142 Gotra area, several ponds representing the terminus of the migrating paleo-channel are located directly above and supply recharge to the channel-fill deposit.  It is therefore conceivable that these oxbow ponds supplement the DOC required for arsenic release. As carbon isotopic studies such as those conducted by Harvey et al., [2002] have not been carried out in Gotra, it is not possible to identify the specific source of DOC to the arsenic release zone.  In all likelihood the DOC driving arsenic release originates from a combination of channel-fill and pond sources.  Differences in simulated groundwater travel times from the ground surface to the arsenic hotspot between pre- and post-development scenarios may have important biogeochemical consequences for arsenic release for this reason.  The time for groundwater to travel to the arsenic rich zone from the surface ponds under current conditions is approximately 23 years.  This is approximately half the required time estimated for pre- development conditions.  If a principal redox driver for arsenic release prior to irrigation practices was organic carbon derived from ponds, the modeling results for current conditions indicate that pumping would have sped up the transport of reducing solutes from this source, potentially enhancing arsenic release. The acceleration of reducing pond water to depth may also have implications for arsenic contamination of deep water at the site.  This is important because a specific sediment sequence within the Gotra Pleistocene deposits has been identified by Pal and Mukherjee, [2008] as potentially safe groundwater resource for the area.  The authors concede that a particular “orange sand” – which is coated significantly with iron-oxyhydroxides and contain up to 16 mg/kg solid phase arsenic – will provide a sustainable source of potable water simply because it is protected by an overlying low-permeability geologic unit (i.e. coincident with the aforementioned paleohorizon in Section 4.1.4).  If this barrier was continuous in the subsurface, model results predict that downward flow would indeed be negligible (Section 5.4.2.5). However, the model cannot be calibrated if this layer is included (and Table 5.2).  Simulated flow paths predict that contaminated groundwater will therefore travel to deep irrigation wells under current conditions, and to a similar horizon (i.e. ~60m bgl) in the pre-development case. A possible explanation that the groundwater in these areas has remained uncontaminated to date may be re-sorption by abundant iron-oxyhydroxdes, explaining the high observed solid arsenic in these materials.  However, the onset of irrigation and resultant acceleration of reducing pond water to depth may provide the DOC to liberate arsenic lacking under pre- development conditions.  As the travel time along this flowpath for pumped conditions is approximately 22 years (i.e. decreased from the predevelopment case by a factor of ~6), it is 5.0 Numerical Groundwater Flow Model 143 likely that deeper aquifers will soon observe a change in geochemical conditions that could increase aqueous arsenic concentrations. 5.6.2. Hydrologic Balance and Recharge Sources An important outcome of the modeling study which may have consequences for arsenic contamination at the site is that a significant shift in the water balance is observed as a direct result of irrigation practices.  This shift necessitates increased recharge through the ground surface, as well as a significant flux of shallow groundwater to the Pleistocene materials compared to a pre-development scenario.  Other significant differences include decreases in natural discharge, complete flux reversals at surface water boundaries, and the recycling of groundwater through irrigation return flow. An increase in recharge through the ground surface with pumping combined with variable land use could have important biogeochemical consequences in Gotra.  At our focussed research site, the oxbow ponds are of particular interest in this regard as they are potential sources of anaerobic water to the subsurface.  As discussed in Section 5.6.1, these ponds are located above the channel-fill deposit, and likely contribute to groundwater destined to reach the arsenic hotspot.  Enhanced recharge through this zone therefore would result in enhanced loading of labile organic carbon to the accumulation zone, potentially enhancing to arsenic release. Although the numerical simulation of the pre-pumped scenario has not been set up to determine the change in this flux from present-day conditions, increases in pond recharge have been simulated for pumped conditions at other contaminated sites (i.e. Harvey et al., [2006]). Under present-day conditions, the most significant sink of groundwater for the Gotra sub-zone is the downward flux to deep materials.  In the pre-pumping scenario, this flux is predicted to be comparatively negligible.  As groundwater flow paths indicate that downward migration from the specific zone of accumulation in Gotra is likely, the enhanced downward flux of groundwater caused by pumping will result in increased arsenic loadings to depth.  This particular aspect of the pumping-induced water balance shift therefore indicates that irrigation activities in Gotra have directly increased the risk of contaminating deep resources with arsenic.  These results accord with modeling studies of Michael and Voss [2009b], who show that enhanced flux of shallow groundwater to deep aquifers as a result of irrigation pumping is significant on a regional scale. 5.0 Numerical Groundwater Flow Model 144 Decreases in natural discharge and flux reversals and the re-infiltration of groundwater are other notable changes to the water balance in Gotra resulting from pumping.  Both of these changes may conceivably act to alter arsenic concentrations in the subsurface.  In the former case, natural aquifer flushing may be slowed down resulting in attenuation of arsenic.  The application of pumped groundwater to rice fields may reduce arsenic concentrations, as observed by Neumann et al., [2010] through a process analogous to pump-and-treat remediation methods, where considerable arsenic is retained by bund and surface soil of fields. 5.6.3. Groundwater Residence Times and Aquifer Flushing Impacts on the groundwater flow system due to irrigation pumping are not solely manifested as changes to the timing and locations of groundwater fluxes.  Modeling results show that pumping has also coerced an increase in total flow through the system, and consequently higher rates of groundwater flushing.  Whether this will significantly impact the natural removal of arsenic is examined in the following paragraphs. Microbial processes tend to drive geochemical systems towards equilibrium due to their short reaction timescales compared to groundwater flow rates [Harvey et al., 2006].  Accordingly, the rate of arsenic mobilization from sediments via microbially mediated processes must be limited by advective transport of DOC to the zone of release.  As sorptive processes are known to drive arsenic release into West Bengal groundwaters (e.g. Nickson et al. [2000]), it is instructive to couple an equilibrium sorption model for arsenic on iron-hydroxides to flow modeling results to speculate on the state of arsenic flushing from Gotra sediments.  Specifically, the number of required pore volumes to flush aqueous and solid concentrations from the site may be estimated with a zero-dimensional arsenic sorption model and simulated groundwater residence times. The linear-Langmuir sorption isotherm developed by Pierce and Moore [1982] for As(III) (pH = 7) on pure amorphous iron hydroxides was coupled with a discrete flushing  model to estimate required flushing of the system from present-day conditions.  Longer residence times and higher sorbed (and aqueous) concentrations observed in the channel-fill unit in Gotra suggest that rates of arsenic removal will be limited by the flushing of this unit as opposed to the point-bar aquifer.  The initial sorbed concentration used in this calculation were therefore set to 30 mg of arsenic per bulk kg of sediment, which is representative of the arsenic content in these materials [Pal et al., 2002b].  Further details of these calculations are described in Appendix G. 5.0 Numerical Groundwater Flow Model 145 Estimated sorbed and dissolved concentrations of arsenic as a function of pore volume flushed are plotted in Figure 5.26.  The plot shows that 1228 and 7962 pore volumes of groundwater must pass through the subsurface to bring aqueous concentrations down to Indian and WHO standards, respectively.  As the channel-fill residence time for pumped conditions is ~8 years, the model predicts that 12,204 and 65,268 years will elapse before groundwater will reach either of these standards on its own. The above estimated flushing times correspond to decreases of 0.4 and 17.9% respectively, compared to a scenario without pumping.  However, coupling of the sorption isotherm with groundwater residence times indicate that the solid reservoir of arsenic is sufficiently large to prevent the contaminant from being eradicated from the system on a human time scale.  In addition, other geochemical processes may conspire to further retard arsenic such as redox- cycling near the ground surface associated with annual water table fluctuations [Harvey et al., 2006; Polizzotto et al., 2006].  Natural flushing of arsenic from therefore is likely not an effective remediation strategy on its own. 5.7. Model Limitations and Opportunities for Improvement Ideally, numerical modeling results should be used to direct future research and planning towards identifying a sustainable drinking water source for the village.  Several important unknowns regarding the hydrogeology and geochemistry of the site remain that warrant cautious courses of action when attempting to mitigate exposure. Hydrogeochemical characterization and numerical modeling results have indicated that arsenic accumulation at the Gotra site requires a specific combination of geology, hydrology and potentially human influence.  In particular, results have shown that groundwater flow is controlled by features such as ponds and the channel-fill unit, and by irrigation pumping at the site.  It is important to note that several inherent limitations to the modeling effort have resulted from assumptions necessary for its construction.  Principal uncertainties include boundary conditions, far-field pumping well locations and schedules, regional hydrostratigraphy and flow from surface water sources through the vadose zone.  The implications of these uncertainties as they relate to the predictive capability of the model are discussed qualitatively in the following paragraphs.  Opportunities for improvement are also noted. 5.0 Numerical Groundwater Flow Model 146 5.7.1. Boundary Conditions The mild relief of the region does not allow the assignment of boundary conditions to be guided by any major topographic features, such as flow divides.  In addition, no flux measurements are available to constrain the exchange between surface and groundwater in the region.  Thus, the accuracy of boundary condition representation the groundwater flow model is uncertain. However, simulated fluxes between the domain and local surface water boundaries in the calibrated base case range approximately between +/-0.2 mm/day, and fluxes between the domain and regional sources vary within 1.5 and -0.2 mm/day (Figure 5.14).  These values are comparable to the fluxes simulated by Harvey et al., [2006] for similar features in a similar climate, which suggests that boundary condition representation in the current model is reasonable.  Improvements to boundary condition assignment can be made through field measurement of fluxes between aquifer and lakes, which may also serve as additional calibration targets. 5.7.2. Pumping Well Locations and Schedules The majority of pumping wells outside of a 2 km radius centered on Gotra specified in the model have been placed at arbitrary locations based on observed well densities for the area.  Wells in the LTT model have been assigned an annual pumping schedule to extract an equivalent volume of water deduced from piezometric data (Section 4.4.1), and an assumed diurnal schedule was inferred for deep wells in the STT model.  Although the specified pumping schemes in both models simulate observed heads reasonably well, details of well locations and pumping records would improve the representation of this groundwater sink.  Specific implications of this uncertainty are errors in model estimates of recharge and deep aquifer conductivity.  The geometry of well capture and therefore recharge sources to areas of interest are also dependent on pumping well locations. 5.7.3. Regional Hydrostratigraphy Although the calibrated hydraulic conductivity value for the Pleistocene sand is within the range of estimates reported by Michael and Voss, [2009a] from drill log analyses, no local field measurements (i.e. aquifer test data in deep wells) are available to independently verify this estimate.  Additionally, model calibration requires that the floodplain sand intercepted at GSI0713 be hydraulically connected to the point bar sand.  However, the stratigraphic relationship of the floodplain unit to the point-bar aquifer is not well delineated from borehole 5.0 Numerical Groundwater Flow Model 147 data.  Impacts of these uncertainties include the magnitude of downward groundwater flow and sources of groundwater to shallow pumping wells, respectively.  Drilling and slug testing of additional deep (i.e. >30m) boreholes would help to constrain the conductivity of the deep materials, and drilling to the southwest of the channel-fill unit would provide useful stratigraphic information within floodplain deposits. 5.7.4. Pond Recharge Contribution and Unsaturated Flow The fluctuation of the water table in the numerical model complicates the tracking of recharge sources to the aquifer from the ground surface.  For example, not all ponds in the numerical model remain hydraulically connected to the groundwater system throughout the year, and particles simulated with MODPATH are stopped within cells below ponds that become desaturated.  The contribution to recharge from these areas therefore cannot be accurately assessed with the current model.  Currently there are no field data available in the area to ascertain which ponds do indeed maintain hydraulic connection with the groundwater system throughout the year.  In all likelihood, there will be a mixture of perched as well as connected conditions depending on the conductivity of underlying materials (e.g. Neumann et al., [2010]). Long-term piezometric monitoring of heads directly adjacent to ponds, as well as monitoring of pond water levels will improve conceptualization in this respect.  Improvements to system representation through modeling may involve coupling the current model with simulations of unsaturated flow or running the current model with MODFLOW-SURFACT, a proprietary code developed by Hydrogeologic Inc. [1997] that provides improved treatment of partially saturated conditions in MODFLOW. 5.8. Summary of Numerical Model Results The numerical model developed in this chapter simulates the groundwater flow regime of a 7.25 km2 area surrounding the village of Gotra.  The model incorporates the main hydrogeological features observed at the site, including stratigraphic units and their associated hydraulic parameters, lakes and wetlands, precipitation, evapotranspiration and irrigation pumping. Simulations were performed under both short-term and long-term transient stress schedules, both of which independently provided an acceptable fit to observed data after model calibration through iterative parameter adjustments.  Results from the model calibration and sensitivity analysis suggest that input parameters are reasonable and the structure of the model appropriately represents the system. 5.0 Numerical Groundwater Flow Model 148 The calibrated value for aquifer recharge of ~180 mm/yr is equivalent to 13% of the average annual rainfall recorded for Nadia district from 2005 to 2009.  This value is in agreement with the estimates of 15% for the entire BDP  [SWID, 1998].  A high vertical anisotropy (Kh/Kv = 100) of the deep aquifer unit was required for calibration, which is consistent the layered nature of deep sediments that have been noted on regional scales [Michael and Voss, 2009a].  Model calibration did not require an impervious unit at the base of the channel-fill and point-bar succession, since it could simultaneously reproduce heads and the observed secular variation using reasonable pumping rates. The simulated flow system exhibits a seasonal periodicity, consistent with the conceptual model of the study site.  The predominant flow direction in the village area during the monsoon season is downward through the channel-fill silt, and horizontally north-eastward through the point bar sand.  Dry-season horizontal groundwater flow directions in the village area converge towards shallow pumping wells in the point-bar aquifer, which is consistent with field observations.  Post- processing of model results with MODPATH indicates that pumping-induced flow directions as observed in the dry season, determine the fate of groundwater entering the village area from the ground surface.  Some noteworthy features of MODPATH results are that 1- water entering the arsenic enriched zone originates from a path exiting the channel-fill silt; 2- the majority of groundwater within the point-bar aquifer exits to shallow irrigation wells and 3- shallow groundwater that does not report to shallow wells is drawn to the screens of deep irrigation wells.  Approximate travel times computed for the three above mentioned flow paths are 46, 7 and 22 years, respectively. As with flow directions, simulated boundary fluxes are seasonally variable, and are driven principally through dry-season pumping and monsoonal recharge.  Modeled domain boundaries behave both as groundwater sources and sinks over the course of the hydrogeologic year, providing recharge to the aquifer during irrigation season, and then discharging groundwater to balance infiltrating monsoon rain.  In the immediate village area, there is considerable flux from shallow sediments to the Pleistocene aquifer, which is more pronounced during the irrigation season.  Analysis of recharge contributions to the model from specific land use zones indicates that the majority of meteoric recharge infiltrates the ground surface through elevated zones. These include inhabited areas and agricultural “High Lands”.  Less significant fluxes are predicted to recharge the model through “Low Lands” and ponded areas.  This is consistent with a conceptualization of these zones as predominantly as discharge areas.  Irrigation return flow to the full domain is predicted to comprise approximately 7% of applied recharge (i.e. ~ 1% of precipitation).  Even less significant recharge contributions from irrigation water are expected to 5.0 Numerical Groundwater Flow Model 149 infiltrate in the immediate Gotra area.  Bulk average residence times in the channel-fill silt deposits and point-bar sands estimated using ZoneBudget are 7.9 and 4.0 years, respectively. These results indicate that since the advent of irrigation in the 1970s, the point-bar unit has been flushed about 10 times, and channel-fill 5 times. A comprehensive sensitivity analysis shows that modeled heads are sensitive to the variation of several parameters.  Most sensitivity tests worsen the goodness of fit to observation data, although several scenarios also were found to satisfy calibration criteria.  Of these scenarios, only channel-fill conductivity and deep aquifer anisotropy have implications for the flow system in the area of interest.  Predicted residence times within the channel-fill silt may range between 5 and 29 years as a result of varying conductivity and storage of this unit.  The extent of well capture and sources of shallow groundwater to the deep aquifer are most influenced by having a Kh/Kv ratio of 10 for the Pleistocene materials. Simulation of a pre-development scenario by deactivating pumping wells indicates several important changes to the groundwater regime have been incurred since the 1970s.  Without pumping, groundwater flow paths in the village area travel in directions coincident with vector fields produced under pumped conditions during the monsoon.  Simulated travel times along flow paths of interest in this scenario are considerably longer than results for current conditions, presumably because of lower gradients.  A significant shift in the water balance also is simulated, with the principal impacts being a decrease in recharge from the ground surface and a reduction in the flux of groundwater from Holocene to Pleistocene materials.  Rejected recharge at the ground surface is conceptualized as overland flow, and comprises nearly 50% of the calibrated recharge value under pumped conditions.  Other changes to the mass balance regime are that surface water and regional groundwater boundaries remain discharging throughout the year.  Bulk residence times computed for the pre-development scenario are predicted to have been approximately 27% and 83% longer for the point bar and channel-fill units respectively.  These correspond to a total of 1640 and 1200 flushed pore-volumes for these units since deposition. The results of numerical modeling in Gotra under both current and pre-development conditions support a model of arsenic release located near the channel-fill deposit.  Oxbow ponds located above this deposit that coincide with recharge flowpaths to the arsenious zone make the source of DOC to the sediments ambiguous.  However, simulations of current conditions predict accelerated travel times through this unit, which may have resulted in a biogeochemical shift in recent years in favour of arsenic mobilization.  Specifically, increased loadings of DOC from 5.0 Numerical Groundwater Flow Model 150 surface ponds as a result of strong vertical gradients induced through irrigation pumping may have enhanced arsenic release from sediments.  Simulated flow paths also indicate that groundwater flows from the shallow arsenious zones to deeper materials that are largely arsenic-free.  Irrigation pumping has accelerated this downward contribution and suggests that Pleistocene materials may soon observe an increase in contamination. Modeling results also predict a shift in groundwater balance under irrigation stresses will occur. This shift supports accelerated organic carbon infiltration from the surface facilitating shallow arsenic release, as well as potential for enhanced arsenic loadings to the deep resource.  Other impacts to the water balance that could directly affect dissolved arsenic concentrations include flux reversals for half of the year, and application of contaminated water to the ground surface. Examination of simulated bulk residence times in the channel-fill unit coupled to a linear- Langmuir model for arsenic sorption indicates that complete flushing of the contaminant from the system will not occur on a human timescale, even with the effects of irrigation. Several limitations of the numerical model have been revealed through sensitivity testing. These include uncertainties in the assignment of boundary conditions, pumping well locations and schedules, regional hydrostratigraphy and flow from surface water sources through the vadose zone.  Improvements to boundary condition representation in the model can be made through field measurement of fluxes between aquifer and lakes, which may also serve as additional calibration targets.  Additional surveys of pumping rates and well locations, as well as borehole surveys and hydrogeologic testing of auxiliary geologic units would improve the model representation of capture zones and flow between village and surrounding area.  Long-term piezometric monitoring of heads directly adjacent to ponds, as well as monitoring of pond water levels will improve conceptualization of flow through the vadose zone beneath these features. Improvements to numerical modeling of unsaturated flow may also be made using a numerical code that is capable of simulating these conditions.  5.0 Numerical Groundwater Flow Model 151  Figure 5.1: Plan view of model domain extent. Annotations over image from Google Maps, (© 2011 Google - Imagery © 2011 DigitalGlobe, GeoEye, Map data © 2011)  5.0 Numerical Groundwater Flow Model 152   Figure 5.2: Model discretization. Model cells within the Gotra “sub-zone” are coloured grey in the bottom figure. Ground image draped over domain from Google Earth (Image © 2008 DigitalGlobe, © 2007 Europa Technologies) No Vertical Exaggeration 20x Vertical Exaggeration 5.0 Numerical Groundwater Flow Model 153    Figure 5.3: Model hydrostratigraphy.  5.0 Numerical Groundwater Flow Model 154   Figure 5.4: Model boundary conditions.  5.0 Numerical Groundwater Flow Model 155   Figure 5.5: DEM image showing physical basis for head dependent boundaries. The figure shows the distances of the Hooghly River, as well as an oxbow lake to the southwest of the model domain (black rectangle), as well as the input head values. d1 ~12km h1 = 7.78 m d2 ~ 5 km h2 = 5m River Hooghly SE Oxbow Model Domain 5.0 Numerical Groundwater Flow Model 156  Figure 5.6: Irrigation wells shown with model hydrostratigraphy. Figure shows a cross-sectional view of stratigraphy halfway through domain.  Note shallow irrigation wells within the point bar unit. Ground image draped over domain from Google Earth (Image © 2008 DigitalGlobe, © 2007 Europa Technologies) 5.0 Numerical Groundwater Flow Model 157   Figure 5.7: Distribution of residual errors in calibrated models. -1.0 -0.5 0.0 0.5 STT Residual Error (m) 0 100 200 300 400 500 600 700 800 C o u n t 0.0 0.1 0.2 0.3 P ro p o rtio n  p e r B a r -2 -1 0 1 2 LTT Residual Error (m) 0 100 200 300 400 500 C o u n t 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 P ro p o rtio n  p e r B a r 5.0 Numerical Groundwater Flow Model 158 GSI0603  GSI0604  Figure 5.8: Observed and simulated heads in the calibrated STT model. 1.5 2 2.5 3 3.5 H e ad  (m  a sl ) 1.5 2 2.5 3 3.5 H e ad  (m  a sl ) 5.0 Numerical Groundwater Flow Model 159 GSI0607  GSI0609  Figure 5.8 (cont.): Observed and simulated heads in the calibrated STT model. 1.5 2 2.5 3 3.5 H e ad  (m  a sl ) 1.5 2 2.5 3 3.5 H e ad  (m  a sl ) 5.0 Numerical Groundwater Flow Model 160 GSI0713  Figure 5.8 (cont.): Observed and simulated heads in the calibrated STT model. 1.5 2 2.5 3 3.5 H e ad  (m  a sl ) 5.0 Numerical Groundwater Flow Model 161 GSI0603  GSI0604  Figure 5.9: Observed and simulated heads in the calibrated LTT model. -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 H e ad  (m  a sl ) -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 H e ad  (m  a sl ) 5.0 Numerical Groundwater Flow Model 162 GSI0605  GSI0606  Figure 5.9 (cont.): Observed and simulated heads in the calibrated LTT model. -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 H e ad  (m  a sl ) -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 H e ad  (m  a sl ) 5.0 Numerical Groundwater Flow Model 163 GSI0607  GSI0608  Figure 5.9 (cont.): Observed and simulated heads in the calibrated LTT model. -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 H e ad  (m  a sl ) -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 H e ad  (m  a sl ) 5.0 Numerical Groundwater Flow Model 164 GSI0609  GSI0714  Figure 5.9 (cont.): Observed and simulated heads in the calibrated LTT model. -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 H e ad  (m  a sl ) -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 H e ad  (m  a sl ) 5.0 Numerical Groundwater Flow Model 165  Figure 5.10: Potentiometric maps and groundwater flow directions during the monsoon. Head data shown are from Oct 1, 2007.  Geochemistry data from domestic and observation wells are also shown. 5.0 Numerical Groundwater Flow Model 166  Figure 5.11: Potentiometric maps and groundwater flow directions during the pumping season. Head data shown are from March 31, 2007.  Geochemistry data from domestic and observation wells are also shown. 5.0 Numerical Groundwater Flow Model 167 (a)  (b)  Figure 5.12: Comparison of observed to simulated vertical gradients. (a) shows the gradients within the channel-fill silt, and gradients within the point-bar sands are shown in (b). 0 0.1 0.2 0.3 0.4 0.5 C h an n e l F ill  D o w n d ar d  G ra d ie n t Observed Simulated -0.004 -0.003 -0.002 -0.001 0 0.001 0.002 0.003 0.004 0.005 0.006 P o in t B ar  D o w n d ar d  G ra d ie n t Observed Simulated 5.0 Numerical Groundwater Flow Model 168  Figure 5.13: Sectional view of simulated pathlines and well capture. Cross section AA‟ location as shown in Figure 5.10 and Figure 5.11.  Approximate screen locations of irrigation wells 33 and 50 are also shown. 5.0 Numerical Groundwater Flow Model 169  Figure 5.14: Global mass balance over the observation period.  -5 -2.5 0 2.5 5 Fl u x (m m /d ay ) Recharge Wells Storage Regional GW Local Surface Water Bodies 5.0 Numerical Groundwater Flow Model 170  Figure 5.15: Mass balance for the Gotra area over the observation period. “Deep” indicates flux between the village are and the Pleistocene aquifer.  -7.5 -5 -2.5 0 2.5 5 7.5 Fl u x (m m /d ay ) Recharge Wells Storage NW NE SE SW Deep 5.0 Numerical Groundwater Flow Model 171  Figure 5.16: Effect of parameter variation on model calibration. The plot indicates that modeled heads are sensitive to most input hydrologic parameters varied in sensitivity tests.  Most tests depreciate the goodness of fit to observation data, although several scenarios appear to improve the match to observed heads. Improved cases are listed in Table 5.8.  0 10 20 30 40 50 60 70 80 90 100 0.75 0.8 0.85 0.9 0.95 1 N R M S (% ) R 2 R2 Normalized RMS (%) R ec ha rg e Po in t B ar  K C h an n el  F ill  K Po in t B ar  S s C ha nn el  F ill  S s D ee p  A qu if er  K D ee p  A q u if e r A n is o tr o p y K  o f L ak e an d  P o n d B o tt o m s Sp ec if ic Yi el d B as e C as e 5.0 Numerical Groundwater Flow Model 172  Figure 5.17: Scaled sensitivities of modeled groundwater fluxes between the Gotra area and surrounding domain.  -0.80 -0.70 -0.60 -0.50 -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 %  C h an ge  in  F lu x / %  C h an ge  in  P ar am e te r NE SW NW SE Deep C ha nn el  F ill  K Po in t B ar  S s C ha nn el  F ill  S s D ee p  A qu if er  A ni so tr o py Sp ec if ic Yi el d 5.0 Numerical Groundwater Flow Model 173  Figure 5.18: Scaled sensitivities of modeled residence times in the channel-fill silt and point-bar sand.  -3.50 -3.00 -2.50 -2.00 -1.50 -1.00 -0.50 0.00 0.50 %  C h an ge  in  R e si d e n ce  T im e  /  %  C h an ge  in  P ar am e te r Channel Fill Point Bar C ha nn el  F ill  K Po in t B ar  S s C ha nn el  F ill  S s D ee p  A qu if er  A ni so tr o py Sp ec if ic Yi el d 5.0 Numerical Groundwater Flow Model 174  Figure 5.19: Sectional view of pathlines for scenario with Kh/Kv = 10   5.0 Numerical Groundwater Flow Model 175  Figure 5.20: Pathlines for scenario with low conductivity layer on the paleosol horizon. 5.0 Numerical Groundwater Flow Model 176  Figure 5.21: Sectional view of pathlines for no-pumping scenario.  5.0 Numerical Groundwater Flow Model 177  Figure 5.22: Global mass balance for scenario with no pumping  -5 -2.5 0 2.5 5 Fl u x (m m /d ay ) Recharge Regional GW Overland Flow Local Surface Water Bodies Storage 5.0 Numerical Groundwater Flow Model 178  Figure 5.23: Mass balance for the Gotra area with no pumping. “Deep” indicates flux between the village are and the Pleistocene aquifer.  -5 -2.5 0 2.5 5 Fl u x (m m /d ay ) Recharge Overland Flow Storage NW NE SE SW Deep 5.0 Numerical Groundwater Flow Model 179  Figure 5.24: Mass balance comparison between simulations with and without pumping for the full domain.    -400 -200 0 200 400 Meteoric Recharge Wells Regional GW Local Surface Water Bodies Storage Without Pumping With Pumping Flux (mm/yr) Discharging Recharging 5.0 Numerical Groundwater Flow Model 180  Figure 5.25: Mass balance comparison between simulations with and without pumping for the Gotra sub-zone.  -600 -400 -200 0 200 400 600 Meteoric Recharge Wells Deep Aquifer NE Edge SW Edge NW Edge SE Edge Storage Without Pumping With Pumping Flux (mm/yr) Discharging Recharging 5.0 Numerical Groundwater Flow Model 181  Figure 5.26: Predicted flushing of arsenic from the channel-fill sediments considering desorption from sediments and irrigation pumping. Approximately 1228 and 7962 pore volumes of water must be removed for the village groundwater to reach Indian and WHO standards, respectively.  These correspond to 12204 and 65268 years from present day.  Calculations are developed in Appendix G.   5.0 Numerical Groundwater Flow Model 182  Table 5.1: Statistical analysis of model calibration  Table 5.2:  Errors for simulation with a low-K paleosol  Table 5.3: Calibrated Hydrogeologic Dataset. STT LTT Number of Data Points 2528 3255 Minimum  (m) -0.513 -1.900 Abs. Minimum  (m) 1.00E-06 2.56E-04 Maximum (m) 0.289 1.818 Abs. Maximum (m) 0.513 1.900 Mean (m) -0.017 0.208 Abs. Mean (m) 0.017 0.541 RMS (m) 0.081 0.686 NRMS (%) 9.75 9.60 r2 0.81 0.96 Residual Mean 4.182 Abs. Residual Mean 4.187 RMS 4.438 nrms (%) 62.09 r2 0.863 Hydrogeologic Unit Kh (m/s) Kv (m/s) Ss  (m -1) Sy Upper Point Bar Sand 1.00E-04 1.00E-05 5.00E-04 0.05 Lower Point Bar Sand 5.00E-04 5.00E-05 1.00E-03 0.05 Channel Fill Silt 2.50E-07 2.50E-08 1.00E-03 0.05 Older Floodplain Sands 5.00E-04 5.00E-05 1.00E-04 0.05 Older Floodplain Silts and Clays 2.50E-07 2.50E-08 1.00E-03 0.05 Top Soil 3.00E-05 3.00E-06 1.00E-03 0.05 Overbank / Splay Silts and Clays 2.50E-07 2.50E-08 1.00E-03 0.05 Lake / Wetland Bottoms 2.50E-07 2.50E-08 1.00E-03 0.05 Paleosol Layer 1.00E-05 1.00E-06 1.00E-04 0.05 Deep Aquifer 1.00E-05 1.00E-07 1.00E-04 0.05 5.0 Numerical Groundwater Flow Model 183  1. Estimated from water balance calculations in Chapter 4 Table 5.4: Distribution of Recharge Sources for the Model Domain and Gotra Area. The immediate Gotra area is shown in Figure 5.2.  Village Ponds High Lands Low Lands Potential Recharge Estimated For Observation Period (mm/yr)1 436.6 1772.7 (receiving village runoff) 389.9 (no village runoff) 506.7 91.5 Calibrated Recharge (mm/yr) 262.0 234.0 291.8 54.7 Fraction of Domain Area 0.24 0.03 0.29 0.44 Contribution to Total Recharge within Model Domain (mm/yr) 62.9 7.0 83.2 23.8 % of Meteoritic Recharge to Model Domain 35.6 4.0 47.0 13.5 Total Recharge to Model Domain (mm/yr) % of average PPT in Nadia (2005-2009) Fraction of Gotra Area 0.31 0.22 0.20 0.27 Contribution to Total Recharge within Gotra Area (mm/yr) 81.2 51.5 58.4 14.8 % of Meteoritic Recharge to Gotra Area 39.5 25.0 28.4 7.2 Total Recharge to Gotra Area (mm/yr) % of average PPT in Nadia (2005-2009) G o tr a A re a 205.8 14.5% Land Use Zone Fu ll D o m ai n 176.8 12.5% 5.0 Numerical Groundwater Flow Model 184  Table 5.5: Base case groundwater budget. Values have been calculated from model results from 2005 to 2009.  The immediate Gotra area is shown in Figure 5.2. Entering Domain Leaving Domain Net Meteoric Recharge 180.4 0.0 180.4 Regional GW 214.9 25.7 189.2 Local Surface Water Bodies 30.9 11.5 19.4 Wells 0.0 355.3 -355.3 Storage 262.9 296.2 -33.3 Total 689.1 688.6 0.5 Entering Gotra Leaving Gotra Net Meteoric Recharge 209.1 0.0 209.1 Deep Aquifer 1.5 452.0 -450.5 NE Edge 161.6 5.0 156.6 SW Edge 8.1 64.9 -56.8 NW Edge 256.4 17.6 238.8 SE Edge 220.3 16.0 204.3 Wells 0.0 283.7 -283.7 Storage 280.4 296.9 -16.5 Total 1137.5 1136.2 1.3 Source / Sink Flux (mm/yr) Source / Sink Flux (mm/yr) G o tr a A re a Fu ll D o m ai n 5.0 Numerical Groundwater Flow Model 185  Table 5.6: Groundwater budget for scenario without pumping Overland flow refers to water exiting the domain through drain boundaries in the top model layer.  Values have been calculated from model results from 2005 to 2009.  The immediate Gotra area is shown in Figure 5.2.  Entering Domain Leaving Domain Net Meteoric Recharge 178.9 0.0 178.9 Regional GW 0.3 89.1 -88.7 Local Surface Water Bodies 4.3 26.7 -22.4 Overland Flow 0.0 79.6 -79.6 Storage 73.7 62.1 11.7 Total 257.3 257.5 -0.2 Entering Gotra Leaving Gotra Net Meteoric Recharge 253.8 0.0 253.8 Deep Aquifer 0.0 42.7 -42.7 NE Edge 14.2 18.6 -4.3 SW Edge 5.6 26.0 -20.4 NW Edge 73.7 2.7 71.0 SE Edge 11.0 131.7 -120.7 Overland Flow 0.0 162.5 -162.5 Storage 131.5 107.7 23.8 Total 489.9 491.9 -2.0 G o tr a A re a Source / Sink Flux (mm/yr) Fu ll D o m ai n Source / Sink Flux (mm/yr) 5.0 Numerical Groundwater Flow Model 186  Table 5.7: Irrigation Return Flow to the Model Domain and Gotra Area. The immediate Gotra area is shown in Figure 5.2.  Table 5.8: Sensitivity cases that improve the calibration of the LTT Model.  Contribution to Total Recharge within Model (mm/yr) % of Total Recharge to Domain Contribution to Total Recharge within Model (mm/yr) % of Total Recharge to Gotra Area High Lands 39.5 11.3 6.4% 7.9 3.8% Low Lands 2.6 1.1 0.6% 0.7 0.3% 12.4 7.0% 8.6 4.2% 1 From applied recharge Full Domain Gotra Area Total Calibrated Return Flow (mm/yr)1 Recharge Zone Simulation Number Parameter Varied Sensitivity Parameter Value Factor Change from Base Case NRMS (%) RMS (m) Residual Mean (m) Abs. Residual Mean (m) r 2 10 1.25 x10 -6 5 9.449 0.675 0.31 0.554 0.962 11 5 x10 -8 0.20 9.69 0.693 0.13 0.529 0.958 12 2.5 x10 -8 0.10 9.829 0.703 0.10 0.533 0.958 14 UPB and LPB Ss (m-1) 2.5 x10 -3 ; 5x10 -3 5 9.26 0.662 0.10 0.572 0.948 17 1 x10 -2 10 8.648 0.618 0.17 0.489 0.959 18 5 x10 -3 5 9.766 0.698 0.18 0.546 0.958 19 2 x10 -4 0.20 9.743 0.696 0.21 0.55 0.96 20 1 x10 -4 0.10 9.786 0.699 0.21 0.553 0.958 27 Deep Aquifer Anisotropy (Kh/Kv) 10 0.10 9.61 0.687 -0.04 0.518 0.948 33 Sy (-) 0.1 2 8.106 0.579 0.05 0.478 0.957 39 Base Case N/A N/A 9.597 0.686 0.21 0.541 0.959 CF K (m/s) CF Ss  (m -1 ) 5.0 Numerical Groundwater Flow Model 187  Table 5.9: Percent changes to groundwater fluxes through the Gotra area and simulated residence times for sensitivity tests. Percent changes to fluxes and simulated residence times for model runs without pumping and a low permeability layer at the paleosol horizon are also shown. % Change in Deep Flux % Change in NE Flux % Change in SW Flux % Change in NW Flux % Change in SE Flux % Change in Storage Channel Fill Point Bar 1.25 x10 -6 5 0.33 1.57 -9.80 0.60 3.98 90.42 4.85 4.00 5 x10 -8 0.2 -0.43 0.90 6.22 -0.73 -0.70 -17.31 19.26 4.00 2.5 x10 -8 0.1 -0.68 2.03 8.85 0.00 -0.77 1.06 28.65 4.01 UPB and LPB Ss  (m -1 ) 2.5 x10 -3 ; 5x10 -3 5 -0.70 -0.16 -3.75 -0.24 1.78 55.43 8.05 3.69 1 x10 -2 10 0.54 2.20 2.80 -2.42 -3.29 -72.78 4.40 4.16 5 x10 -3 5 0.17 3.83 4.41 -1.04 0.19 12.22 5.32 4.00 2 x10 -4 0.2 0.03 1.42 1.21 0.75 0.22 31.69 8.80 3.98 1 x10 -4 0.1 0.02 -0.18 0.51 -0.32 0.16 0.06 8.81 3.98 Deep Aquifer Anisotropy (Kh/Kv) 10 0.1 63.50 68.05 7.63 40.47 29.19 -136.24 7.82 3.86 Sy (-) 0.1 2 -0.32 -0.83 7.09 -3.98 2.71 -44.07 8.34 4.02 No Pumping N/A N/A -87.57 -124.89 -74.74 -84.08 -165.84 -80.68 9.98 7.29 Lo K Layer N/A N/A -77.15 -81.93 -99.66 -32.85 -107.27 -115.03 6.45 5.57 Base Case N/A N/A N/A N/A N/A N/A N/A N/A 7.88 3.99 CF K (m/s) CF Ss  (m -1 ) Change in Flux from Base Case (%) Simulated Residence Time (yr) Factor Change from Base Case Parameter Varied Sensitivity Parameter Value 6.0 Summary and Conclusion 188 6.0 SUMMARY AND CONCLUSION The Gotra site lies within a severely arsenic contaminated zone where groundwater is heavily pumped for both irrigation and domestic purposes.  Arsenic concentrations in shallow groundwater tapped by some domestic wells in Gotra reach up to 50 times the WHO standard (i.e. 516 µg/L), whereas others provide a potable resource.  Wells extracting deeper groundwater in the village tend to be less contaminated. Several lines of evidence from this study suggest that arsenic contamination at the site is controlled by groundwater flow.  Numerical modeling results support field observations that contamination is related to a specific channel-fill deposit, and suggest that the water balance has been significantly altered compared to pre-irrigation conditions.  Results from the numerical modeling study also suggest that recent perturbations to the groundwater flow system may facilitate the release of arsenic by several means including the transport of the contaminant and/or release agents and the diversion of groundwater fluxes and flowpaths. Borehole data was used to classify shallow materials in the village area into three principal hydrostratigraphic units of Holocene age.  These include a channel-fill silt, the main point-bar aquifer and an older floodplain succession.  The base of the shallow package is marked at a depth of approximately 30m by sparsely distributed deposits of wood, peat and calcrete (kankar).  Below this paleohorizon, sands of Pleistocene age have been intercepted by some boreholes.  These sands are presumed to be regionally continuous, as numerous high-yielding irrigation wells are screened at depths greater than 30 m. Field tests and analysis of time-series piezometric data indicate that hydraulic conductivity of the channel-fill silt is approximately 10-7 m/s and the point-bar sand is 5x10-4 m/s.  Piezometric monitoring also reveals that irrigation pumping and the monsoonal climate of the region are the principal controls on groundwater levels at the site.  Groundwater elevations to vary from approximately -0.5 to 6.5 m asl over the course of the year.  The cyclical recovery of piezometric heads to similar levels in successive years suggests that groundwater is fully recharged following the monsoon.  No prominent horizontal gradients are discernable from the data; however convergent flow towards shallow pumping wells is a salient feature during irrigation season.  Localized downward gradients in the point-bar and channel-fill units are observed throughout the year. 6.0 Summary and Conclusion 189 The geochemical characteristics of affected waters in Gotra are consistent with those of many arsenic contaminated aquifers in the BDP.  These characteristics are conducive to reductive arsenic release from oxide precipitates into groundwater.  Also similar to many sites within the BDP, arsenic distribution in the Gotra area is patchy, varying up to two orders of magnitude over a distance of 20m.  Despite this variability, a distinct geochemical gradient evolves in a direction perpendicular to trend of the channel-fill silt; arsenic concentrations decrease with increasing distance from this deposit.  This gradient may be interpreted as redox fronts of terminal electron accepting processes, where contamination is associated with the most reducing groundwater. Elevated concentrations of DOC, high simulated pCO2 and methanogenic waters in close proximity to the channel-fill deposits all suggest that recharge through this unit is related to the delivery of organic carbon driving arsenic release from iron-oxides.  However, the data collected from the site cannot unequivocally map the source of the DOC to the oxbow ponds overlying these deposits, or the organics within the sediments themselves. A conceptual model of flow and arsenic transport has been developed from field observations at the site.  Numerical implementation of this conceptual model using the finite-difference code MODFLOW simulates a flow system that is consistent with field observation.  Results from the model calibration and sensitivity analysis indicate that input parameters are reasonable and the structure of the model appropriately represents the system.  Under current conditions, water enters the domain principally as meteoric recharge during the monsoon, and is drawn from regional sources during the irrigation season.  Pumping is the principal groundwater sink, although minor discharge fluxes to regional and surface water boundaries also drain the model. The most significant sink of groundwater for the immediate Gotra area is the downward flux to deep aquifers induced by deep irrigation pumping. Simulated flow paths under both current and pre-development conditions indicate that groundwater entering from the surface flows downward through the channel-fill deposits, then horizontally through the point bar sands perpendicular to the trend of the abandoned channel. This path aligns with profiles of geochemical gradients observed between domestic wells, and supports a model of arsenic release from the channel-fill silt.  Modeled flow paths also indicate that there is potential for contaminated groundwater from the zone of arsenic enrichment to migrate to deeper aquifers.  Travel times associated with these flow paths suggest that changes to biogeochemistry may increase loadings of DOC from oxbow ponds as a result of wide-spread irrigation.  Irrigation pumping has also likely influenced biogeochemistry at depth by drawing down solutes to deeper Pleistocene materials. 6.0 Summary and Conclusion 190 Results from the modeling study indicate that the current stresses on the system have reduced natural discharge from a pre-pumped scenario by causing flux reversals for seven months of each year across boundaries that otherwise would be strictly discharging.  However, the most significant changes to the flow regime as a result of irrigation pumping include a doubling of meteoric recharge to the water table, as well as an increased downward flux of shallow groundwater.  These findings support the potential acceleration of arsenic release as a result of increased loading of labile organic carbon from surface ponds for example, and again highlight the current vulnerability of the deep resource. Bulk average residence times simulated for the point-bar sand and channel-fill silt are estimated at less than a decade under current conditions, which is approximately half the value expected for the pre-development case.  Coupling pore-volume flushing of the channel-fill silt to an equilibrium sorption model indicates that adequate flushing of the contaminant these materials (i.e. solid and aqueous phase) will not occur on a human time scale. While the effects of irrigation pumping are readily observed and modeled, the fate of arsenic in groundwater is much more complex.  The time to develop of new geochemical equilibria is limited by advective flow, and the groundwater chemistry Gotra is likely still adjusting to anthropogenic stresses that can alter dissolved arsenic concentrations.  As suggested by modeling results, switching of domestic wells to known uncontaminated zones is not necessarily a simple mitigation option at this time because many unknowns about groundwater flow at the site remain.  Additional site characterization and modeling efforts for the Gotra area are warranted and recommended to improve the potential success of identifying a sustainable drinking water source for the village. References 191 REFERENCES Acharyya, S. K., and B. A. Shah (2007), Arsenic-contaminated groundwater from parts of Damodar fan-delta and west of Bhagirathi River, West Bengal, India: influence of fluvial geomorphology and Quaternary morphostratigraphy, Environmental Geology, 52(3), 489-501. Acharyya, S. K., S. Lahiri, B. C. Raymahashay, and A. Bhowmik (2000), Arsenic toxicity of groundwater in parts of the Bengal basin in India and Bangladesh: the role of Quaternary stratigraphy and Holocene sea-level fluctuation, Environmental Geology, 39(10), 1127-1137. Acharyya, S. K., B. A. Shah, I. D. Ashyiya, and Y. Pandey (2005), Arsenic contamination in groundwater affecting major parts of southern West Bengal and parts of western Chhattisgarh: Source and mobilization process, Current Science, 82(6), 740-744. Acharyya, S. K., P. Chakraborty, S. Lahiri, B. C. Raymahashay, S. Guha, and A. Bhowmik (1999), Arsenic poisoning in the Ganges delta, Nature, 401(6753), 545-545. Aggarwal, P. K., A. R. Basu, and K. M. Kulkarni (2003), Comment on "Arsenic mobility and groundwater extraction in Bangladesh" (I), Science, 300(5619), 584B-U581. Alam, M., M. M. Alam, J. R. Curray, A. L. R. Chowdhury, and M. R. Gani (2003), An overview of the sedimentary geology of the Bengal Basin in relation to the regional tectonic framework and basin-fill history, Sedimentary Geology, 155(3-4), 179-208. Allen, J. R. L. (1964), Studies in Fluviatile Sedimentation - 6 Cyclotherms from the Lower Old Red Sandstone, Anglo-Welsh Basin Sedimentology, 3(3), 163-198. Allen, R. G., L. S. R. Pereira, D., and M. Smith (1998), Crop Evapotranspiration: Guidelines for Computing Crop Water Requirements, Journal of Irrigation and Drainage, 56. Allison, M. A. (1998), Geologic framework and environmental status of the Ganges- Brahmaputra delta, Journal of Coastal Research, 14(3), 826-836. Anawar, H. M., J. Akai, K. M. G. Mostofa, S. Safiullah, and S. M. Tareq (2002), Arsenic poisoning in groundwater - Health risk and geochemical sources in Bangladesh, Environment International, 27(7), 597-604. Anderson, M. P., and W. W. Woessner (1991), Applied Groundwater Modeling: Simulation of Flow and Advective Transport 381 pp., Academic Press, London. Appelo, C. A. J., and D. Postma (2005), Geochemistry, groundwater and pollution, 2nd ed., 649 pp., A.A. Balkema, The Netherlands. Ashfaque, K. (2007), Effect of Hydrological Flow Pattern on Groundwater Arsenic Concentration in Bangladesh, 286 pp, Massachusetts Institute of Technology, Massachusetts. Aziz, Z., et al. (2008), Impact of local recharge on arsenic concentrations in shallow aquifers inferred from the electromagnetic conductivity of soils in Araihazar, Bangladesh, Water Resources Research, 44(7). Basu, A. R., S. B. Jacobsen, R. J. Poreda, C. B. Dowling, and P. K. Aggarwal (2001), Large groundwater strontium flux to the oceans from the bengal basin and the marine strontium isotope record, Science, 293(5534), 1470-1473. Batu, V. (1998), Aquifer Hydraulics: A Comprehensive Guide to Hydrogeologic Data Analysis, 727 pp., John Wiley and Sons Inc., New York. Beckie, R., S. Sengupta, G. Hall, T. Pal, P. Mukherjee, A. Rencz, A. Desbarats, and C. E. M. Koenig (2006), Naturally occurring arsenic in groundwater: a preliminary investigation of References 192 sources and release mechanisms, Gotra, West Bengal, India, edited, p. 31, Geological Survey of Canada Report. Beckie, R. D., A. J. Desbarats, T. Pal, P. K. Mukherjee, and C. E. M. Koenig (2009), The geochemistry of high- and low-arsenic groundwaters and their association with an in-filled abandoned channel at a Field Site in Gotra, Nadia District, West Bengal, India, paper presented at AGU Chapman Conference on Arsenic in Groundwater of Southern Asia, Siem Reap, Cambodia, 24–27 March 2009. Benner, S. G., M. L. Polizzotto, B. D. Kocar, S. Ganguly, K. Phan, K. Ouch, M. Sampson, and S. Fendorf (2008), Groundwater flow in an arsenic-contaminated aquifer, Mekong Delta, Cambodia, Applied Geochemistry, 23(11), 3072-3087. Berg, M., H. Con Tran, T. H. Nguyen, P. H. Viet, R. Schertenleib, and W. Giger (2001), Arsenic Contamination of Groundwater and Drinking Water in Vietnam: A Human Health Threat, Environmental Science & Technology, 35(13), 2621-2626. Berg, M., C. Stengel, P. T. K. Trang, P. H. Viet, M. L. Sampson, M. Leng, S. Samreth, and D. Fredericks (2007), Magnitude of arsenic pollution in the Mekong and Red River Deltas - Cambodia and Vietnam, Science of the Total Environment, 372(2-3), 413-425. Berner, R. A. (1981), A New Geochemical Classification of Sedimentary Environments, Journal of Sedimentary Petrology, 51(2), 359-365. BGS, and DPHE (2001), Arsenic Contamination of Groundwater in Bangladesh (Phase 2), British Geological Survey Technical Rep. WC/00/19. Bhattacharya, P., D. Chatterjee, and G. Jacks (1997), Occurrence of Arsenic Contaminated Groundwater in Alluvial Aquifers from Delta Plains, Eastern India: Options for Safe Drinking Water Supply, International Journal of Water Resources, 13, 79-92. Bhattacharya, P., J. Jana, B. Nath, S. J. Sahu, D. Chatterjee, and G. Jacks (2003), Groundwater As mobilization in the Bengal Delta Plain, the use of ferralite as a possible remedial measure - a case study, Applied Geochemistry, 18, 1435-1451. Boulton, N. S. (1954), Unsteady Radial Flow to a Pumped Well allowing for Delayed Yield from Storage, Institution of Civil Engineers, Journal, 26, 468-482. Bouwer, H., and R. C. Rice (1976), Slug Test for determining Hydraulic Conductivity of Unconfined Aquifers with completely or partially penetrating wells, Water Resources Research, 12(3), 423-428. Burgess, W. G., M. Burren, J. Perrin, and K. M. Ahmed (2002), Constraints on sustainable development of arsenic-bearing aquifers in southern Bangladesh. Part 1: A conceptual model of arsenic in the aquifer, in Sustainable Groundwater Development, edited by K. M. Hiscock, M. O. Rivett and R. M. Davison, pp. 145-163. Chakraborti, D., G. K. Basu, L. K. Biswas, U. K. Chowdhury, M. M. Rahman, K. Paul, T. R. Chowdhury, C. R. Chanda, D. Lodh, and S. L. Ray (2001), Characterization of arsenic-bearing sediments in the gangetic delta of West Bengal, India, 27-52 pp. Coleman, J. M. (1969), Brahmaputra River - Channel Processes and Sedimentation, Sedimentary Geology, 3(2-3), 131-&. Coleman, J. M. (1981), Deltas: Processes of Deposition and Models of Exploration, 2nd ed., 124 pp., Burgess, Minneapolis. Cooper, H. H., and C. E. Jacob (1946), A generalized graphical method for evaluating formation constants and summarizing well field history, American Geophysical Union, Transactions, 27, 526-534. References 193 Cooper, H. H., J. D. Bredehoeft, and S. S. Papadopulod (1967), Response of a finite-diameter well to an instantaneous charge of water, Water Resources Research, 3(1), 263-269. Cuthbert, M. O., W. G. Burgess, and L. Connell (2002), Constraints on sustainable development of arsenic-bearing aquifers in southern Bangladesh. Part 2: Preliminary models of arsenic variability in pumped groundwater, in Sustainable Groundwater Development, edited by K. M. Hiscock, M. O. Rivett and R. M. Davison, pp. 165-179. Davis Jr., R. A. (1992), Depositional Systems - An introduction to Sedimentology and Stratigraphy, Prentice-Hall, New Jersey. Desbarats, A., C. E. M. Koenig, T. Pal, P. K. Mukherjee, and R. D. Beckie (2009), Modeling shallow groundwater flow in an arsenic-impacted aquifer at a field site in Gotra, Nadia District, West Bengal, India, paper presented at AGU Chapman Conference on Arsenic in Groundwater of Southern Asia, Siem Reap, Cambodia, 24–27 March 2009. Dhar, R. K., et al. (1997), Groundwater arsenic calamity in Bangladesh, Current Science, 73(1), 48-59. Dickinson, W. R. (1975), Current concept of depositional systems with applications for petroleum geology, San Joaquin Geological Society Short Course, 124p. Dingman, S. L. (2002), Physical Hydrology, 2nd ed., 646 pp., Prentice Hall, New Jersey. Dixit, S., and J. G. Hering (2003), Comparison of Arsenic(V) and Arsenic(III) Sorption onto Iron Oxide Minerals: Implications for Arsenic Mobility, Environmental Science & Technology, 37, 4182-4189. Dowling, C. B., R. J. Poreda, and A. R. Basu (2003), The groundwater geochemistry of the Bengal Basin: Weathering, chemsorption, and trace metal flux to the oceans, Geochimica Et Cosmochimica Acta, 67(12), 2117-2136. DPHE (2006), Final Report on Development of Deep Aquifer Database and Preliminary Deep Aquifer Map (First Phase), Final Rep., Department of Public Health and Engineering, Government of the People‟s Republic of Bangladesh, Arsenic Policy Support Unit, and JICA, Dhaka. Effron, B., and R. J. Tibshirani (1993), An introduction to the Bootstrap, Monographs on Statistics and Applied Probability, 436 pp., Chapman and Hall, New York. Fendorf, S., M. J. Eick, P. Grossl, and D. L. Sparks (1997), Arsenate and chromate retention mechanisms on goethite, Environmental Science & Technology, 31(2), 315-320. Fendorf, S., M. Polizzotto, M. J. Herbel, B. C. Bostick, and C. Harvey (2003), Arsenic cycling within surface and subsurface environments: Impact of iron mineralogy, Abstracts of Papers of the American Chemical Society, 226, 013-GEOC. Fetter, C. W. (2001), Applied Hydrogeology, 4th ed., 598 pp., Prentice Hall, New Jersey. Freeze, R. A., and J. A. Cherry (1979), Groundwater, 2nd ed., 604 pp., Prentice Hall, New Jersey. Ghosh, A. R., and A. Mukherjee (2002), Arsenic contamination of groundwater and human health impacts in Burdwan District, West Bengal, India, Geological Society of America, Abstracts with Programs 34(2), 107. Ghosh, N. C., B. Chakraborty, and P. K. Mazumder (1999), Groundwater flow model, paper presented at Proceedings of Workshop on Groundwater and Its Protection with Special Reference to Arsenic Contamination, Calcutta. Girard, I., R. A. Klassen, and R. R. Laframboise (2004), Sedimentology Laboratory Manual, Terrain Sciences Division Geological Survey of Canada, Open File Rep. 4823, 137 pp. References 194 Goodbred, S. L., and S. A. Kuehl (2000), The significance of large sediment supply, active tectonism, and eustasy on margin sequence development: Late Quaternary stratigraphy and evolution of the Ganges-Brahmaputra delta, Sedimentary Geology, 133(3-4), 227-248. Goodbred, S. L., S. A. Kuehl, M. S. Steckler, and M. H. Sarker (2003), Controls on facies distribution and stratigraphic preservation in the Ganges-Brahmaputra delta sequence, Sedimentary Geology, 155(3-4), 301-316. Gupta, R. (2006), Development of irrigation in West Bengal and its role in agricultural productivity, ICFAI Journal of Infrastructure, 4(1), 67-78. Hantush, M. S., and C. E. Jacob (1955), Non-steady raidial flow in an infinite leaky aquifer, American Geophysical Union, Transactions, 36, 95-100. Harbaugh, A. W. (1990), A computer program for calculating subregional water budgets using results from the U.S. Geological Survey modular three-dimensional ground-water flow model U.S. Geological Survey Open-File Rep. 90-392, 46 pp. Harbaugh, A. W., E. R. Banta, M. C. Hill, and M. G. McDonald (2000), MODFLOW-2000, the U.S. Geological Survey modular ground-water model user guide to modularization concepts and the ground-water flow process, U.S. Geological Survey Open-File Rep., 121 pp. Harvey, C. F. (2002), Groundwater flow in the Ganges delta, Science, 296(5573). Harvey, C. F., and R. D. Beckie (2005), Arsenic: Its biogeochemistry and transport in groundwater, in Metal Ions in Biological Systems, Vol 44, edited, pp. 145-169. Harvey, C. F., et al. (2006), Groundwater dynamics and arsenic contamination in Bangladesh, Chemical Geology, 228(1-3), 112-136. Harvey, C. F., et al. (2002), Arsenic mobility and groundwater extraction in Bangladesh, Science, 298(5598), 1602-1606. Harvey, C. F., et al. (2003), Response to comments on "Arsenic mobility and groundwater extraction in Bangladesh", Science, 300(5619), 584D-U583. Harvey, C. F., et al. (2005), Groundwater arsenic contamination on the Ganges Delta: biogeochemistry, hydrology, human perturbations, and human suffering on a large scale, Comptes Rendus Geoscience, 337(1-2), 285-296. Hazen, A. (1911), Discussion of “Dams on sand formations,” by A.C. Koenig, Transactions of the American Society of Civil Engineers 73, 199-203. Hill, M. C. (1998), Methods and guidelines for effective model calibration, Water - Resources Investigations Rep. 98-4005, 90 pp. Holeman, J. N. (1968), Sediment yield for major rivers of the world, Water Resources Research, 4(4), 737-&. Horneman, A., et al. (2004), Decoupling of As and Fe release to Bangladesh groundwater under reducing conditions. Part 1: Evidence from sediment profiles, Geochimica Et Cosmochimica Acta, 68(17), 3459-3473. Hossain, M. (2006), Arsenic contamination in Bangladesh--An overview, Agriculture, Ecosystems & Environment, 113(1-4), 1-16. Hossain, M., D. Lewis, M. Bose, and A. Chowdhury (2003), Rice Research: Technological Progress and Impacts on the Poor: The Bangladesh Case, Summary Rep., International Food Policy Research Institute (IFPRI). Hubbert, M. K. (1940), The theory of groundwater motion, Journal of Geology, 48, 785-944. References 195 Hvorslev, M. J. (1951), Time lag and soil permeability in Ground-Water Observations, edited, p. 50, Waterways Experiment Station, U.S. Army Corps of Engineers, Vicksburg. Islam, M. S., and M. J. Tooley (1999), Coastal and sea-level changes during the Holocene in Bangladesh, Quaternary International, 55, 61-75. JICA (2002), The study on the ground water development of deep aquifers for safe drinking water supply to arsenic affected areas in western Bangladesh, Draft Final Rep. 1-3, Japan International Cooperation Agency, Kokusai Kogyo Co. Ltd. and Mitsui Mineral Development Engineering Co. Ltd. . Johnson, M. R. W. (1994), Volume balance of erosional loss and sediment deposition related to Himalayan uplifts, Journal of the Geological Society, 151, 217-220. Karim, M. (2000), Arsenic in groundwater and health problems in Bangladesh, Water Research, 34(1), 304-310. Kinniburgh, D. G., P. L. Smedley, J. Davies, C. J. Milne, I. Gaus, J. M. Trafford, S. Burden, S. M. I. Huq, N. Ahmad, and M. K. Ahmad (2003), The scale and causes of the groundwater arsenic problem in Bangladesh, in Arsenic in groundwater: Geochemistry and occurrence, edited by A. H. Welch and K. G. Stollenwork, pp. 211-257, Kluwer Academic Publishers, Boston. Kocar, B. D., et al. (2008), Integrated biogeochemical and hydrologic processes driving arsenic release from shallow sediments to groundwaters of the Mekong delta, Applied Geochemistry, 23(11), 3059-3071. Korte, N. E., and Q. Fernando (1991), A review of arsenic(III) in groundwater, Critical Reviews in Environmental Control, 21, 1-39. Kothyari, U. C. (1995), Estimation of monthly runoff from small catchments in India, Hydrological Sciences Journal, 40(4), 533-542. Kothyari, U. C., and R. J. Garde (1991), Annual runoff estimation for catchments in India, Journal of Water Resources Planning and Management-ASCE, 117(1), 1-10. Kresic, N. (1997), Quantitative Solutions in Hydrogeology and Groundwater Modelling, 461 pp., CRC Lewis Kumar, P., K. N. Tiwari, and P. D.K. (1997), 19Establishing SCS runoff curve number form IRS digital database, Journal of the Indian Society of Remote Sensing, 19, 246-251. Lovley, D. R., and F. H. Chapelle (1995), Deep subsurface microbial processes, Reviews of Geophysics, 33(3), 365-381. Maidment, R. (1993), Handbook of Hydrology, 1400 pp., McGraw-Hill. Mandal, B. K., T. R. Chowdhury, G. Samanta, G. K. Basu, P. P. Chowdhury, C. R. Chanda, D. Lodh, N. K. Karan, R. K. Dhar, and D. K. Tamili (1996), Arsenic in groundwater in seven districts of West Bengal, India, the biggest arsenic calamity in the world, Current Science, 70(11), 976. McArthur, J. M., P. Ravenscroft, S. Safiulla, and M. F. Thirlwall (2001), Arsenic in groundwater: Testing pollution mechanisms for sedimentary aquifers in Bangladesh, Water Resources Research, 37(1), 109-117. McArthur, J. M., P. Ravenscroft, D. M. Banerjee, J. Milsom, K. A. Hudson-Edwards, S. Sengupta, C. Bristow, A. Sarkar, S. Tonkin, and R. Purohit (2008), How paleosols influence groundwater flow and arsenic pollution: A model from the Bengal Basin and its worldwide implication, Water Resources Research, 44(11). References 196 McArthur, J. M., et al. (2004), Natural organic matter in sedimentary basins and its relation to arsenic in anoxic ground water: the example of West Bengal and its worldwide implications, Applied Geochemistry, 19(8), 1255-1293. Meharg, A. A., and M. Rahman (2003), Arsenic contamination of Bangladesh paddy field soils: implications for rice contribution to arsenic consumption, Environmental Science & Technology, 37, 229-234. Meharg, A. A., C. Scrimgeour, S. A. Hossain, K. Fuller, K. Cruickshank, P. N. Williams, and D. G. Kinniburgh (2006), Codeposition of organic carbon and arsenic in Bengal Delta aquifers, Environmental Science & Technology, 40(16), 4928-4935. Michael, H. A., and C. I. Voss (2009a), Estimation of regional-scale groundwater flow properties in the Bengal Basin of India and Bangladesh, Hydrogeology Journal, 17(6), 1329-1346. Michael, H. A., and C. I. Voss (2009b), Controls on groundwater flow in the Bengal Basin of India and Bangladesh: regional modeling analysis, Hydrogeology Journal, 17(7), 1561-1577. Milliman, J. D., and J. P. M. Syvitski (1992), Geomorphic tectonic control of sediment discharge to the ocean - the importance of small mountainous rivers, Journal of Geology, 100(5), 525-544. Mok, W. M., J. A. Riley, and C. M. Wai (1988), Arsenic speciation and quality of groundwater in a lead-zinc mine, Idaho, Water Research, 22(6), 769-774. Morgan, J. P., and W. G. McIntire (1959), Quaternary geology of the Bengal Basin, East Pakistan and India, Geological Society of America Bulletin, 70(3), 319-342. MPO (1987), The groundwater resource and its availability for development, Technical Rep. 5, Ministry of Irrigation, Water Development and Flood Control, Dhaka. Mualem, Y. (1976), NEW MODEL FOR PREDICTING HYDRAULIC CONDUCTIVITY OF UNSATURATED POROUS-MEDIA, Water Resources Research, 12(3), 513-522. Mukhejee, S. (2004), rerequisite studies for numerical flow modeling to locate safe drinking water wells in the zone of arsenic polluted groundwater in the Yamuna sub-basin, West Bengal, India, paper presented at Proceedings of 32nd International Geologic Congress, Florence, Italy. Mukherjee, A., and A. E. Fryar (2008), Deeper groundwater chemistry and geochemical modeling of the arsenic affected western Bengal basin, West Bengal, India, Applied Geochemistry, 23(4), 863-894. Mukherjee, A., A. E. Fryar, and P. D. Howell (2007a), Regional hydrostratigraphy and groundwater flow modeling in the arsenic-affected areas of the western Bengal basin, West Bengal, India, Hydrogeology Journal, 15(7), 1397-1418. Mukherjee, A., A. E. Fryar, and H. D. Rowe (2007b), Regional-scale stable isotopic signatures of recharge and deep groundwater in the arsenic affected areas of West Bengal, India, Journal of Hydrology, 334(1-2), 151-161. Mukherjee, P. K., T. Pal, S. Sengupta, and S. Shome (2001), Arsenic rich phases in aquifer sediments from southern West Bengal, Journal of the Geological Society of India, 58(2), 173- 176. Mukhopadhyay, B., P. K. Mukherjee, D. Bhattacharya, and S. Sengupta (2006), Delineation of arsenic-contaminated zones in Bengal Delta, India: a geographic information system and fractal approach, Environmental Geology, 49(7), 1009-1020. Murali, V., and L. A. G. Aylmore (1983), Competitive adsorption during solute transport in soils 3: A review of experimental evidende of competitive adsorption and an evaluation of simple competition models, Soil Science, 136(5), 279-290. References 197 Narasimhan, T. N., and M. Zhu (1993), TRANSIENT FLOW OF WATER TO A WELL IN AN UNCONFINED AQUIFER - APPLICABILITY OF SOME CONCEPTUAL MODELS, Water Resources Research, 29(1), 179-191. Nath, B., Z. Berner, D. Chatterjee, S. B. Mallik, and D. Stuben (2008a), Mobility of arsenic in West Bengal aquifers conducting low and high groundwater arsenic. Part II: Comparative geochemical profile and leaching study, Applied Geochemistry, 23(5), 996-1011. Nath, B., D. Stuben, S. B. Mallik, D. Chatterjee, and L. Charlet (2008b), Mobility of arsenic in West Bengal aquifers conducting low and high groundwater arsenic. Part I: Comparative hydrochemical and hydrogeological characteristics, Applied Geochemistry, 23(5), 977-995. Nath, B., Z. Berner, S. B. Mallik, D. Chatterjee, L. Charlet, and D. Stueben (2005), Characterization of aquifers conducting groundwaters with low and high arsenic concentrations: a comparative case study from West Bengal, India, Mineralogical Magazine, 69(5), 841-854. NBLM (2006), Groundwater Modelling Guidance for Mining Activities, Instruction Memorandum NV-2006-065. Neuman, S. P. (1972), THEORY OF FLOW IN UNCONFINED AQUIFERS CONSIDERING DELAYED RESPONSE OF WATER TABLE, Water Resources Research, 8(4), 1031-&. Neuman, S. P. (1974), Effect of Partial Penetration on Flow in Unconfined Aquifers Considering Delayed Gravity Response of the Water Table, Water Resources Research, 10(2), 303-312. Neumann, R. B., K. N. Ashfaque, A. B. M. Badruzzaman, M. A. Ali, J. K. Shoemaker, and C. F. Harvey (2010), Anthropogenic influences on groundwater arsenic concentrations in Bangladesh, Nature Geoscience, 3(1), 46-52. Nickson, R., J. M. McArthur, P. Ravenscroft, W. G. Burgess, and K. M. Ahmed (2000), Mechanism of arsenic release to groundwater, Bangladesh and West Bengal, Applied Geochemistry, 15(4), 403-413. Nickson, R., J. McArthur, W. Burgess, K. M. Ahmed, P. Ravenscroft, and M. Rahman (1998), Arsenic poisoning of Bangladesh groundwater, Nature, 395(6700), 338-338. Pal, T. (2006), Sedimentary Geologist, p. communication. Geological Survey of India. Pal, T. (2007), Sedimentary Geologist, p. communication. Geological Survey of India. Pal, T. (2008), Sedimentary Geologist, p. communication. Geological Survey of India. Pal, T. (2009), Sedimentary Geologist, p. communication. Geological Survey of India. Pal, T., and P. K. Mukherjee (2008), 'Orange sand' - A geological solution for arsenic pollution in Bengal delta, Current Science, 94(1), 31-33. Pal, T., and P. K. Mukherjee (2009), Study of subsurface geology in locating arsenic-free groundwater in Bengal delta, West Bengal, India, Environmental Geology, 56(6), 1211-1225. Pal, T., P. K. Mukherjee, and S. Sengupta (2002a), Nature of arsenic pollutants in groundwater of Bengal basin - A case study from Baruipur area, West Bengal, India, Current Science, 82(5), 554-561. Pal, T., P. K. Mukherjee, S. Sengupta, A. K. Bhattacharyya, and S. Shome (2002b), Arsenic pollution in groundwater of West Bengal, India - An insight into the problem by subsurface sediment analysis, Gondwana Research, 5(2), 501-512. Pal, T., P. K. Mukherjee, A. J. Desbarats, C. E. M. Koenig, and R. D. Beckie (2009), Geological Controls on Groundwater Arsenic: New Insights from a Field Site at Gotra, Nadia District, West Bengal, India, paper presented at AGU Chapman Conference on Arsenic in Groundwater of Southern Asia, Siem Reap, Cambodia, 24–27 March 2009 References 198 Papadopulos, S. S., J. D. Bredehoeft, and H. H. Cooper (1973), ANALYSIS OF SLUG-TEST DATA, Water Resources Research, 9(4), 1087-1089. Parkhurst, D. L. (2010), p. communication. United States Geological Survey. Parkhurst, D. L., and C. A. J. Appelo (1999), User's guide to PHREEQC (version 2)-A computer program for speciation, batch-reaction, one-dimensional transport, and inverse geochemical calculations, U.S. Geological Survey Water-Resources Investigations Rep. 99-4259, 312 pp. Pedersen, H. D., D. Postma, and R. Jakobsen (2006), Release of arsenic associated with the reduction and transformation of iron oxides, Geochimica Et Cosmochimica Acta, 70(16), 4116- 4129. Pierce, M. L., and C. B. Moore (1982), Adsorption of Arsenite and Aresnate and Amorphous Iron Hydroxide, Water Research, 16, 1247-1253. Polizzotto, M. L., B. D. Kocar, S. G. Benner, M. Sampson, and S. Fendorf (2008), Near-surface wetland sediments as a source of arsenic release to ground water in Asia, Nature, 454(7203), 505-U505. Polizzotto, M. L., C. F. Harvey, G. C. Li, B. Badruzzman, A. Ali, M. Newville, S. Sutton, and S. Fendorf (2006), Solid-phases and desorption processes of arsenic within Bangladesh sediments, Chemical Geology, 228(1-3), 97-111. Pollock, D. W. (1989), Documentation of computer programs to compute and display pathlines using results from the U.S. Geological Survey modular three-dimensional finite-difference ground-water flow model, U.S. Geological Survey Open-File Rep. 89-381, 188 pp. Polya, D. A., and L. Charlet (2009), Rising Arsenic Risk?, Nature Geoscience, 2(383-384). Polya, D. A., et al. (2005), Arsenic hazard in shallow Cambodian groundwaters, Mineralogical Magazine, 69(5), 807-823. Rahman, A. A., and P. Ravenscroft (2003), Groundwater resources and development in Bangladesh, background to the arsenic crisis, agricultural potential and the environment, 466 pp., University Press, Dhaka. Rahman, M. M., et al. (2001), Chronic arsenic toxicity in Bangladesh and West Bengal, India - A review and commentary, Journal of Toxicology-Clinical Toxicology, 39(7), 683-700. Ravenscroft, P., and J. McArthur (2004), Mechanism of regional enrichment of groundwater by boron: the examples of Bangladesh and Michigan, USA, Applied Geochemistry, 19(9), 1413- 1430. Ravenscroft, P., J. M. McArthur, and B. A. Hoque (2001), Geochemical and palaeohydrological controls on pollution of groundwater by arsenic, 53-77 pp. Ravenscroft, P., W. G. Burgess, K. M. Ahmed, M. Burren, and J. Perrin (2005), Arsenic in groundwater of the Bengal Basin, Bangladesh: Distribution, field relations, and hydrogeological setting, Hydrogeology Journal, 13(5-6), 727-751. Reilly, T. E., and A. A. Harbaugh (2004), Guidelines for Evaluating Ground-Water Flow Models, U.S. Geological Survey Scientific Investigations Rep. 2004-5038, 30pp. Rittle, K. A., J. I. Drever, and Colberg (1995), Precipitation of arsenic during bacterial sulfate reduction, Geomicrobiological Journal, 13, 1-11. Rowland, H. A. L., A. G. Gault, J. M. Charnock, and D. A. Polya (2005), Preservation and XANES determination of the oxidation state of solid-phase arsenic in shallow sedimentary aquifers in Bengal and Cambodia, Mineralogical Magazine, 69(5), 825-839. SCD (1972), Handbook of Hydrology, Soil Conservation Department, Ministry of Agriculture, New Delhi. References 199 Schaap, M. G., F. J. Leij, and M. T. van Genuchten (2001), ROSETTA: a computer program for estimating soil hydraulic parameters with hierarchical pedotransfer functions, Journal of Hydrology, 251(3-4), 163-176. SCS (1956), National Engineering Handbook, Section 4: Hydrology, USDA, Washington. Sengupta, S., P. K. Mukherjee, T. Pal, and S. Shome (2004), Nature and origin of arsenic carriers in shallow aquifer sediments of Bengal Delta, India, Environmental Geology, 45(8), 1071-1081. Shamsuddin, A. H. M., T. A. Brown, and M. Rickard (2002), Bangladesh gas endowment-1: Resource studies indicate large gas potential in Bangladesh, Oil & Gas Journal, 100(16), 48-52. Smedley, P. L., and D. G. Kinniburgh (2002), A review of the source, behaviour and distribution of arsenic in natural waters, Applied Geochemistry, 17(5), 517-568. Smith, A. H., E. O. Lingas, and M. Rahman (2000), Contamination of drinking water by arsenic in Bangladesh: a public health emergency, World Health Organization, Bulletin, 78, 1093-1103. Spitz, K., and J. Moreno (1996), A Practical Guide to Groundwater and Solute Transport Modeling, John Wiley and Sons, New York. Stumm, W., and J. J. Morgan (1996), Aquatic Chemistry, 3rd ed., 1022 pp., John Wiley and Sons Inc., New York. Stute, M., et al. (2007), Hydrological control of As concentrations in Bangladesh groundwater, Water Resources Research, 43(9). Swartz, C. H., N. K. Blute, B. Badruzzaman, A. Ali, D. Brabander, J. Jay, J. Besancon, S. Islam, H. F. Hemond, and C. F. Harvey (2004), Mobility of arsenic in a Bangladesh aquifer: Inferences from geochemical profiles, leaching data, and mineralogical characterization, Geochimica Et Cosmochimica Acta, 68(22), 4539-4557. SWID (1998), A comprehensive hydrogeological information of Murshidabad district, edited, Geologic Division No. III, State Water Investigation Directorate (SWID), Calcutta. SWS (2009), Visual MODFLOW Version 2009.1 Pro, Waterloo, Ontario, www.swstechnology.com. Theis, C. V. (1935), The relation between the lowering of the piezometric surface and the rate and duration of discharge of a well using groundwater storage, American Geophysical Union, Transactions, 16, 519-524. van Geen, A., Y. Zheng, M. Stute, and K. M. Ahmed (2003a), Comment on "Arsenic mobility and groundwater extraction in Bangladesh" (II), Science, 300(5619), 584C-584C. Van Geen, A., J. Rose, S. Thoral, J. M. Garnier, Y. Zheng, and J. Y. Bottero (2004), Decoupling of As and Fe release to Bangladesh groundwater under reducing conditions. Part II: Evidence from sediment incubations, Geochimica Et Cosmochimica Acta, 68(17), 3475-3486. Van Geen, A., Z. Cheng, A. A. Seddique, M. A. Hoque, A. Gelman, J. H. Graziano, H. Ahsan, F. Parvez, and K. M. Ahmed (2005), Reliability of a commercial kit to test groundwater for arsenic in Bangladesh, Environmental Science & Technology, 39(1), 299-303. van Geen, A., Y. Zheng, Z. Cheng, Y. He, R. K. Dhar, J. M. Garnier, J. Rose, A. Seddique, M. A. Hoque, and K. M. Ahmed (2006a), Impact of irrigating rice paddies with groundwater containing arsenic in Bangladesh, Science of the Total Environment, 367(2-3), 769-777. van Geen, A., et al. (2006b), Preliminary evidence of a link between surface soil properties and the arsenic content of shallow groundwater in Bangladesh, Journal of Geochemical Exploration, 88(1-3), 157-161. References 200 van Geen, A., et al. (2003b), Spatial variability of arsenic in 6000 tube wells in a 25 km(2) area of Bangladesh, Water Resources Research, 39(5). von Bromssen, M., M. Jakariya, P. Bhattacharya, K. M. Ahmed, M. A. Hasan, O. Sracek, L. Jonsson, L. Lundell, and G. Jacks (2007), Targeting low-arsenic aquifers in Matlab Upazila, Southeastern Bangladesh, Science of the Total Environment, 379(2-3), 121-132. Wang, S. L., and C. N. Mulligan (2006), Natural attenuation processes for remediation of arsenic contaminated soils and groundwater, Journal of Hazardous Materials, 138(3), 459-470. WARPO (2000), National Water Management Plan Project: draft development strategy, edited, Ministry of Water Resources, Government of the People‟s Republic of Bangladesh, Dhaka. Weinman, B., S. L. Goodbred, Y. Zheng, Z. Aziz, M. Steckler, A. van Geen, A. K. Singhvi, and Y. C. Nagar (2008), Contributions of floodplain stratigraphy and evolution to the spatial patterns of groundwater arsenic in Araihazar, Bangladesh, Geological Society of America Bulletin, 120(11-12), 1567-1580. Wentworth, C. K. (1922), A scale of grade and class terms for clastic sediments, Journal of Geology, 30(5), 377-392. WHI (2003), Visual MODFLOW User‟s Manual, Waterloo Hydrogeologic, Inc. Winkel, L., M. Berg, M. Amini, S. J. Hug, and C. A. Johnson (2008), Predicting groundwater arsenic contamination in Southeast Asia from surface parameters, Nature Geoscience, 1, 536- 542. Worm, H. U., A. M. M. Ahmed, N. U. Ahmed, H. O. Islam, M. M. Huq, U. Hambach, and J. Lietz (1998), Large sedimentation rate in the Bengal Delta: Magnetostratigraphic dating of Cenozoic sediments from northeastern Bangladesh, Geology, 26(6), 487-490. Yu, W. H., C. M. Harvey, and C. F. Harvey (2003), Arsenic in groundwater in Bangladesh: A geostatistical and epidemiological framework for evaluating health effects and potential remedies, Water Resources Research, 39(6). Zheng, Y., M. Stute, A. van Geen, I. Gavrieli, R. Dhar, H. J. Simpson, P. Schlosser, and K. M. Ahmed (2004), Redox control of arsenic mobilization in Bangladesh groundwater, Applied Geochemistry, 19(2), 201-214. Zheng, Y., et al. (2005), Geochemical and hydrogeological contrasts between shallow and deeper aquifers in two villages of Araihazar, Bangladesh: Implications for deeper aquifers as drinking water sources, Geochimica Et Cosmochimica Acta, 69(22), 5203-5218.  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 201 APPENDIX A – BOREHOLE LOGS, GRAIN SIZE DATA AND WELL COMPLETION DETAILS Borehole logs, grain size analyses and completion details for the observation wells drilled within the village (Figure 3.3) are presented in this section.  Percentages of sand, silt and clay, as well as inorganic and organic carbon from sediment analyses are included in the logs where data were available.  Calculated values of hydraulic conductivity from piezometer tests, pumping tests and from grain size analysis (Appendix C) are also included where available.  Lithologs of boreholes BH 19 and BH 20, which were drilled in 2004, are included as well, since they were used for correlation (Section 4.0).  Average sea level was used as the datum for plotting all borehole logs and cross sections, and ground elevations at the well locations were assumed to be consistently 0.61m (2 ft) below top of casing (TOC) elevations (Table A.1). Borehole sediments were classified numerically into groups according to the log descriptions recorded by the GSI while in the field, and are listed in Table A.2.  These consist of sands, silts, clays of various grain sizes as well as calcrete (“kankar”) and organic matter.  Reduced materials (greyish colour) were distinguished from oxidized sediment (brown or orange colour) by specifying a subclass “g or “b” to the materials, respectively. Intact “clay” samples were kept for laboratory analysis, whereas representative coarse grained materials could not be obtained because fine components were lost due to washing from the drilling fluid (pond water; Section 3.0, Figure 3.8).  Classification of sediments beyond field speculation (i.e. grain size, chemistry) therefore was not possible.  Grain size analyses revealed that the majority of these samples contained more than 50 wt% of silt sized particles (4-63μm), although they were initially identified as clays in the field (Figure A.1 and A.2).  Field descriptions will be used solely for the purpose stratigraphic correlation since a complete grain size analyses for all sediments at the site is not available.  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 202  Table A.1: Elevations of top of well casings (TOC) from GSI survey data (2006, 2007).  These values were used for correction of water level data to obtain static water levels (SWL) at the site.   Table A.2: Gotra field-based sediment classification scheme.  Well Name TOC Elevation (masl) Well Name TOC Elevation (masl) GSI-06-01 6.41275 GSI-06-09 6.90875 GSI-06-02 7.00075 GSI-06-10 7.58475 GSI-06-03 7.83975 GSI0611 6.67775 GSI-06-04 7.84575 GSI0612 8.68275 GSI-06-05 7.99975 GSI-07-13 8.39875 GSI-06-06 8.64875 GSI-07-14 9.28775 GSI-06-07 9.01075 GSI-07-14A 9.28775 GSI-06-08 8.43875 GSI-07-15 8.75575 Unit No. Field Description 1 Coarse Sand 2 Medium Sand 3 Fine Sand 4 Silty Sand 5 Silt 6 Silty Clay 7 Hard Clay 8 Kankar 9 Peat 10 Detrital Wood Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 203    Figure A.1:  Percentages of sand silt and clay within sediment samples collected at the site (a).  Field descriptions of the collected materials often were designated as “clays”; however, grain size data clearly indicate that samples contain grains in the silt category. Materials in this plot are classified according to the Udden-Wentworth standard grain size scale for clastic sediments (Wentworth, [1922]; (b))   (a) (b) Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 204   Figure A.2: Grain-size distribution curves for fine materials collected at GSI- 06-03, 09, 10 and GSI-07-15 (top) and GSI-06-05 (bottom). 0 10 20 30 40 50 60 70 80 90 100 0.0001 0.001 0.01 0.1 1 10 Aperture (mm) %  P a s s e d GSI-06-03 7-17 ft GSI-07-15  61-62 ft GSI-06-09 3-15 ft GSI-06-10 85-86 ft CLAY SILT SAND 0 10 20 30 40 50 60 70 80 90 100 0.0001 0.001 0.01 0.1 1 10 Aperture (mm) %  P a s s e d GSI-06-05 16-38 ft GSI-06-05 38-69 ft GSI-06-05  56 ft GSI-06-05  >56 ft CLAY SILT SAND Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 205   Figure A.2 (continued): Grain-size distribution curves for fine materials collected at GSI-06-06, 08 (top) and GSI-06-07, GSI-07-14, 14A (bottom). 0 10 20 30 40 50 60 70 80 90 100 0.0001 0.001 0.01 0.1 1 10 Aperture (mm) %  P a s s e d GSI-06-06 12-15.5 ft GSI-06-06  20-27 ft GSI-06-06 10-30 ft GSI-06-08 15-31 ft CLAY SILT SAND 0 10 20 30 40 50 60 70 80 90 100 0.0001 0.001 0.01 0.1 1 10 Aperture (mm) %  P a s s e d GSI-06-07 20-55 ft GSI-07-14  7-10 ft GSI-07-14  15-50 ft GSI-07-14A  20-68 ft GSI-07-14A  95-105 ft GSI-07-14A 105-115 ft CLAY SILT SAND Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 206  Figure A.2 (continued): Grain-size distribution curves for fine materials collected at GSI-07-13.  The plots show percentages of material passing against sieve aperture.  Analyses were performed at the Geological Survey of Canada using the method of Girard et al., [2004].  Relative percentages of sand, silt and clay for are plotted in Figure A.1 as well as with borehole logs (below).  Grain size data were also used in several methods for the estimation of hydraulic conductivity (Appendix C).  0 10 20 30 40 50 60 70 80 90 100 0.0001 0.001 0.01 0.1 1 10 Aperture (mm) %  P a s s e d GSI-07-13  10-15 ft GSI-07-13  15-20 ft GSI-07-13  20-30 ft GSI-07-13  30-35 ft GSI-07-13  35-38 ft GSI-07-13  38-43 ft GSI-07-13  46-71 ft GSI-07-13  85-93 ft CLAY SILT SAND Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 207 - Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 208  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 209  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 210  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 211  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 212  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 213  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 214  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 215  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 216  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 217  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 218  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 219  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 220  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 221  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 222  Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 223   Appendix A – Borehole Logs, Grain Size Data and Well Completion Details 224 Appendix B – Piezometric Monitoring 225 APPENDIX B – PIEZOMETRIC MONITORING Groundwater levels and physico-chemical parameters were monitored in-situ within the GSI observation wells from May 2006 to April 2009.  Measurements were collected at the wells over short intervals (1 to 20 minutes) during annual visits to the field site, and over larger intervals (15 minutes to 4 hours) for the remainder of the year.  Table B.1 lists deployment locations, instrument and data collection types, as well as data collection schemes for all loggers from during this period.  This section describes the instrumentation, deployment and physico- chemical data collection schemes at site observation wells.  Some principal observations are also noted and discrepancies discussed as they relate to instrument troubleshooting. Solinst® Leveloggers recorded absolute pressure in the wells, which is the total head of water plus barometric pressure.  A Solinst® Barologger was used to measure barometric pressure at the same recording intervals as the Leveloggers in order to compensate the Levelogger data. Compensations were made using the Solinst® software wizard, which simply subtracts the barometric reading from the Levelogger reading for consistent time stamps, and provides the true net water levels above the measurement points (instrument depths).  A measurement of the depth to water was taken at the beginning of each logging period at all wells equipped with Leveloggers using a Solinst® Water Level Meter.  This depth to water measurement was used as a reference point from which the actual water table elevation could be calculated using surveyed elevations of tops of observation well casings (TOC elevations, Appendix A, Table A.1). Pressure ranges of instruments in several wells were exceeded due to higher than anticipated water level fluctuations during the monsoon of 2006 and 2007 (Table B.1).  As a result, water level measurements in wells GSI0607 and 08 from 2006-2007 and GSI0609 in 2007-2008 are not available during the rainy season.  In addition, the Solinst® Barologger was submerged by groundwater during the monsoon of 2007, making it impossible to apply barometric corrections to any Solinst® data during this interval.  However, data collected at wells GSI0603 and 04 during this period remained useful for the investigation of vertical gradients since only head differences, not absolute values, are necessary.  During the 2008-2009 period, barometric pressure measurements were not collected, and as such transducer measurements documented for GSI-06-03 represent total pressure of the water column in addition to atmospheric fluctuations. Appendix B – Piezometric Monitoring 226 In-situ® Troll MP 8000s (pressure, temperature, EC pH and DO) and Aqua Troll 200s (pressure, temperature and EC) were all equipped with vented cables and as such the data did not require barometric compensation.  Water table reference elevations were noted at the beginning of each logging period, and incorporated into measurement using the In-situ® logger programming software immediately before instrument deployment.  In addition to water level measurements, all In-situ® instruments had the capability of collecting time-series physico-chemical parameters data (EC and pH and DO) as well as water temperature (Table B.1).  Throughout the measurement period, electrical conductivity and pH remained in calibration, however pH measurements likely were affected by degassing of CO2, and therefore reflect borehole rather than formation values.  Measurements of dissolved oxygen made with these sensors are considered unreliable due to drift, and therefore these data are not used in the site characterization. In-situ® Aqua Trolls and Troll MP probes were suspended approximately 20m below the well casings (~ at the well screen depth), whereas Solinst® Leveloggers were deployed at shallower depths.  The probes in GSI0607 and GSI0608 from 2006-2007, as well as GSI0609 in the 2007- 2008 sampling interval collect data at an approximate depth of 7-9m, and the probes in GSI0603 and GSI0604 over the 2007-2008 interval collect measurements between 11 and 12m below the well casings.  Variations in depth of logger deployment between observation wells likely resulted in observed water temperature differences (Section 4). Time-series water level measurements from 2006 to 2009 are plotted in Figure B.1, and show the time span of the hydrologic year (i.e. October to October).  Only approximate heads for GSI0603 are shown, as barometric correction data are not available for 2009. Appendix B – Piezometric Monitoring 227  Table B.1:  Summary of in-situ water level and physico-chemical time-series data collection from Gotra observation wells   Well Instrument Data Sampling Frequency Start End GSI-06-05 In-situ ®  Troll MP 8000 with vented cable Water Level, Temperature, pH, EC, DO 4 Hours 5/22/2006 16:00 2/16/2007 12:00 GSI-06-06 In-situ ®  Troll MP 8000 with vented cable Water Level, Temperature, pH, DO 4 Hours 5/22/2006 16:00 2/16/2007 12:00 Solinst® Levelogger Water Level M5/F15 Temperature Solinst ®  Levelogger Water Level M5/F15 Temperature Barometric Pressure Temperature Well Instrument Data Sampling Frequency Start End Solinst ®  Levelogger Water Level M10/F30 Temperature 15 Minutes 12/8/2007 13:00 1/29/2008 21:15 Solinst ®  Levelogger Water Level M10/F30 Temperature 15 Minutes 12/8/2007 13:00 1/29/2008 21:15 GSI-06-05 In-situ ®  Aqua troll 200 with vented cable Water Level, Temperature, EC 1 Hour 3/1/2007 17:00 2/8/2008 16:00 GSI-06-06 In-situ ®  Troll MP 8000 with vented cable Water Level, Temperature, pH, DO 4 Hours 3/2/2007 16:00 2/2/2008 12:00 Solinst ®  Levelogger Water Level M10/F30 Temperature 15 Minutes 12/8/2007 13:00 1/29/2008 21:15 GSI-06-08 In-situ ®  Aqua troll 200 with vented cable Water Level, Temperature, EC 1 Hour 3/1/2007 17:00 2/16/2008 12:00 Solinst ®  Levelogger Water Level M5/F15 Temperature 20 Minutes 12/8/2007 13:00 1/29/2008 21:20 Barometric Pressure Temperature GSI-07-14 In-situ ®  Troll MP 8000 with vented cable Water Level, Temperature, pH, EC, DO 4 Hours 3/2/2007 20:00 2/2/2008 12:00 GSI-06-09 20 Minutes 3/2/2007 13:40 9/10/2007 22:20 GSI-06-09 Solinst ®  Baralogger 20 Minutes 3/2/2007 13:40 1/29/2008 21:20 GSI-06-04 15 Minutes 3/2/2007 13:45 9/10/2007 22:45 GSI-06-07 15 Minutes 3/2/2007 13:45 9/10/2007 22:45 GSI-06-03 15 Minutes 3/2/2007 13:45 9/10/2007 22:45 Low Frequency Data Collection (2007-2008) GSI-06-08 Solinst ®  Baralogger 2 Hours 5/22/2006 12:00 2/17/2007 10:00 GSI-06-08 2 Hours 5/22/2006 12:00 2/17/2007 10:00 GSI-06-07 2 Hours 5/22/2006 12:00 2/16/2007 12:00 Low Frequency Data Collection (2006-2007) Appendix B – Piezometric Monitoring 228    Table B.1 (Continued):  Summary of in-situ water level and physico-chemical time-series data collection from Gotra observation wells Well Instrument Data Sampling Frequency Start End GSI-06-03 Solinst ®  Levelogger Water Level  15 Minutes 2/16/2008 15:00 4/3/2009 0:00 M10/F30 Temperature Low Frequency Data Collection (2008-2009) Well Instrument Data Sampling Frequency Start End GSI-06-03 Solinst ®  Levelogger Water Level  1 Minute 2/17/2007 12:11 2/20/2007 11:21  M10/F30 Temperature 1 Minute 2/20/2007 15:37 2/25/2007 10:13 2 Minute 2/25/2007 12:00 3/2/2007 13:24 Solinst ®  Levelogger Water Level M10/F30 Temperature GSI-06-05 Solinst ®  Levelogger Water Level  1 Minute 2/17/2007 12:32 2/23/2007 10:13  M5/F15  Temperature 1 Minute 2/24/2007 10:34 2/25/2007 10:05 2 Minute 2/25/2007 10:42 2/28/2007 12:40 2 Minute 2/28/2007 14:06 3/1/2007 10:38 GSI-06-06 Solinst ®  Levelogger Water Level  1 Minute 2/21/2007 15:44 2/22/2007 9:46 M10/F30 Temperature 1 Minute 2/22/2007 13:50 2/24/2007 14:35 1 Minute 2/24/2007 15:14 2/25/2007 10:13 2 Minute 2/25/2007 12:10 3/2/2007 8:38 GSI-06-07 Solinst ®  Levelogger Water Level  1 Minute 2/17/2007 12:22 2/18/2007 16:43 M5/F15 Temperature 1 Minute 2/18/2007 16:49 2/22/2007 13:40 1 Minute 2/22/2007 16:02 2/25/2007 10:13 2 Minute 2/25/2007 11:16 3/2/2007 9:58 2 Minute 3/2/2007 10:54 3/2/2007 12:32 GSI-06-08 Solinst ®  Levelogger Water Level  1 Minute 2/17/2007 13:04 2/23/2007 12:33 M10/F30 Temperature 1 Minute 2/23/2007 15:50 2/25/2007 10:13 Solinst ®  Levelogger Water Level M5/F15 Temperature GSI-06-09 2 Minute 3/1/2007 15:24 3/2/2007 13:24 GSI-06-04 2 Minute 3/1/2007 16:06 3/2/2007 12:54 High Frequency Data Collection (2007) Well Instrument Data Sampling Frequency Start End Solinst ®  Levelogger Water Level M10/F30 Temperature Solinst ®  Levelogger Water Level M10/F30 Temperature Solinst ®  Levelogger Water Level M10F30 Temperature Solinst ®  Levelogger Water Level 20 Minute 2/10/2008 14:40 2/10/2008 15:20 M5/F15 Temperature  1 Minute 2/10/2008 15:25 2/16/2008 14:23 Solinst ®  Levelogger Water Level M5/F15 Temperature GSI-06-09 GSI-07-13 1 Minute 2/11/2008 15:40 2/16/2008 14:18 GSI-06-04 1 Minute 2/10/2008 15:25 2/16/2008 14:18 GSI-06-07 1 Minute 2/10/2008 14:25 2/16/2008 13:13 GSI-06-03 1 Minute 2/10/2008 15:25 2/16/2008 14:18 High Frequency Data Collection (2008) Appendix C – Estimation of Hydrogeologic Parameters 229 APPENDIX C – ESTIMATION OF HYDROGEOLOGIC PARAMETERS The goal of this section is to characterize the hydraulic parameters of the main hydrostratigraphic units defined in the geologic model.  These include: point-bar aquifer sands; channel-fill clayey-silts; and levee/crevasse-splay silts.  Parameter estimates were made based on the analysis of piezometer test data, pump tests and analytical superposition modeling, and grain-size based methods, all described in the following text. C.1.0 Piezometer Tests A piezometer test allows the measurement of hydraulic conductivity (and possibly storage, with lesser accuracy) in-situ by monitoring the recovery water level in a piezometer over time after the initiation of an instantaneous change in head.  The change in water level is initiated by the sudden introduction or removal of a known volume of water, or solid cylinder into the well. When volume is added, these tests are known specifically as slug tests, and when it is removed, they are known as bail tests.  In this section, slug tests will often be referred to as falling head tests, and bail tests as rising head tests.  This terminology describes the respective water level responses observed in each. Several mathematical solutions for interpreting the water level recovery with time are commonly used to calculate aquifer parameters, and all involve specific assumptions about the piezometer configuration, aquifer confinement and groundwater flow in time.  The analysis methods employed in this study include solutions developed by Cooper, Bredehoeft and Papadopulos [1967], Hvorslev [1951] and Bouwer and Rice [1976]. These models include both transient and quasi steady-state assumptions about groundwater flow in response to an induced instantaneous disturbance.  The fully transient Cooper- Bredehoeft-Papadopulos [1967] solution (CBP) accounts for specific storage in the aquifer, and was only used on the data from the wells screened in the sand.  This is because the model assumes that the medium is confined by lower conductivity units, which is not the case for the channel-fill silt.  The models of Bouwer-Rice [1976] (BR) and Hvorslev [1951] (Hv) assume quasi-steady-state flow through the formation, where hydraulic head changes with time, but elastic storage is ignored.  Head changes at the test well are therefore assumed to be transmitted immediately through the medium in these analyses. Appendix C – Estimation of Hydrogeologic Parameters 230 CBP analyses were conducted using graphical techniques (Section C.1.1), whereas the software package AQTESOLV 4.01 was used to calculate hydraulic conductivity from the field data for both Hvorslev and Bouwer-Rice methods (Sections C.1.2 and C.1.3). The general observation well geometry shown in Figure C.1 was used in order to perform analyses on piezometer test data using all methods.  The figure depicts an example of the water level‟s response during a falling head test, shortly after the slug is released into the well. During a slug test, water level responses may be over-damped (monotonically returning to the static water level), or under-damped (oscillating as it returns to the static water level).  Under- damped responses sometimes occur in high hydraulic conductivity aquifers due to inertial effects in the well (e.g. Cooper et al., [1967]).  Figure C.2 shows an example of the data collected from tests performed at GSI0605.  As the familiar over-damped response is observed in rising head slug tests (Figure C.2(bottom)), the oscillations exhibited in the falling head data (Figure C.2 (top)) are likely not due to well inertial effects.  Scatter in the data observed in falling head tests in this study likely is the result of additional energy being added to the system due to dropping of the slug into the well.  Raw data from the piezometer bail tests are plotted in Figure C.3.  Semi-log plots of normalized displacement (Ht/H0) versus time (seconds) were also produced using the test data for both aquifer and channel-fill silt and are shown in Figures C.4 and C.5. A geometric average of individual hydraulic conductivity values (Kgeo) is used to summarize the calculated results.  This is value is deemed to best represent the overall aquifer conductivity, as the probability distribution function for hydraulic conductivity generally is log-normal.  This is used as opposed to an arithmetic, or harmonic average, which are biased towards the highest and lowest possible conductivity values, respectively. C.1.1 Cooper-Bredehoeft-Papadopulos (CBP) Analysis CBP analysis assumes that the aquifer is perfectly confined, homogeneous, isotropic, infinite in extent, that the piezometer where a slug test is being performed is fully penetrating, and that flow to the well is horizontal [Cooper et al., 1967]. The CBP solution identifies normalized displacement, Ht/H0, is a function (F) of the parameters η and α: Appendix C – Estimation of Hydrogeologic Parameters 231 Eqn. C1 ),( F H H o t   where: Eqn. C2 2 cr Tt   Eqn. C3 S r r c s 2 2   and: Ht = water level at time, t [L]; Ho = initial displacement [L]; T = aquifer transmissivity [L2/T]; t = elapsed time [T]; rc = radius of the interior of the piezometer [L]; rs = radius of screen + gravel pack [L]; S = aquifer storage coefficient [-] At the Gotra site, aquifer conditions are leaky-confined (see Section 4.3.2).  The aquifer itself is saturated, whereas the confining layer may not be fully saturated.  In addition, all of the observation wells are partially penetrating, thus the real system does not comply with all the assumptions inherent to the CBP method.  However, the analysis was still performed on the data to obtain a rough estimate aquifer transmissivity and storage coefficient.  Semi-log plots of normalized displacement versus time were prepared using slug test data at the same scale as the CBP type curves [Papadopulos et al., 1973].  Slug test data were visually matched to the type curves which had the most similar curvature by superimposing the plots and keeping the Appendix C – Estimation of Hydrogeologic Parameters 232 arithmetic axes coincident.  A match point, t, was selected from the data plot that overlaid the vertical axis η = 1.0 on the type curve plot, so that transmissivity could be calculated from: Eqn. C4 t r T c 0.1 2   Values of hydraulic conductivity, K [L/T], were then calculated using screen length, Le [L], which is the thickness of the saturated region of flow.  This is because all wells are partially penetrating, and T values from this analysis are only representative of the formation and flow conditions immediately in the vicinity of the test hole. The geometry of the Gotra wells in Figure C.1 show that rc = rs in all cases, so it was possible to estimate storativity from the relationship: Eqn. C5 S In most cases, the curvature of test data could be matched to multiple type curves, hence α (S values) from this analysis are highly non-unique.  This is a common result in permeable materials, where the shape of the response, is only weakly affected by α, and consequently α is not easily identifiable from the data.  Examples of semi-log plots of Ht/H0 versus time and scaled CPB type curves are shown in Figures C.6 and C.7.  Aquifer parameters calculated from this analysis are summarized in Table C.1. C.1.2 Hvorslev Analysis The Hv method does not require that observation wells be fully penetrating in order to determine the hydraulic conductivity of the formation.  The method assumes quasi-steady-state flow throughout the test, where head changes at the test well are transmitted instantaneously through the formation.  As a consequence, specific storage is neglected in parameter estimation, in contrast to the CBP method, which as previously mentioned, is fully transient and accounts for aquifer storage.  This Hv method also assumes that the aquifer has infinite aerial extent, is homogeneous and is of uniform thickness and is confined. The following relationship describes the water-level response due to the instantaneous injection or withdrawal of a known volume from a well [Hvorslev, 1951]: Appendix C – Estimation of Hydrogeologic Parameters 233 Eqn. C6                      2 2 0 2 1 2 ln 2 ln we e we e e t r L r L r tKL H H  and Eqn. C7 xZwwe KKrr /  where Ht  = water level at time t [L]; H0  = initial displacement [L]; K, Kx = hydraulic conductivity [L/T]; t = time [T]; Kz  = vertical hydraulic conductivity [L/T]; Le  = screen length [L]; r =  well casing radius [L]; rw  =  well radius [L]; rwe  =  equivalent well radius [L]; rw represents the radius of the gravel envelope surrounding the well and was set equal to r in all calculations for wells screened within the sands.  The additional area of the well annulus was incorporated into analysis of the recovery data for piezometers screened in the clay unit. Significant anisotropy in the aquifers of the Bengal Delta Plain (BDP) has been noted by several authors and has been estimated within a range of 0.1 [Ashfaque, 2007] to 0.0001 [Michael and Appendix C – Estimation of Hydrogeologic Parameters 234 Voss, 2009a].  Estimates made by the former author are for a site of similar scale to this study, whereas those made by the latter authors represent regional values of deeper materials, which incorporate layering effects of numerous aquitards.  In the absence of any modeled anisotropy value for our site, it was assumed that Kh >> Kv, and an anisotropy ratio (Kv/Kh) of 0.01 was assigned for the purpose of slug test data analysis.  Conductivity results obtained using this method are summarized in Table C.2. C.1.3 Bouwer-Rice Analysis The BR unconfined analysis method was also performed on slug test data for comparison with the results obtained by other analyses.  As with the Hv method, BR analysis employs a quasi- steady state model to determine hydraulic conductivity, and so ignores elastic aquifer storage. The method is used on over-damped responses where the well may be fully or partially penetrating.  It may also be applied to approximate conditions in confined aquifers if the top of the well screen is positioned significantly far from the water-table boundary so that it does not affect slug test response, as is the case here. The recovery of water-levels as a result of instantaneous displacement in this method is described by Eqn. C8 )/ln( 2 )/ln( 20 wee e t rrr tKL HH   re is the effective radial distance over which head is dissipated, or the distance away from the well over which the average conductivity value is being measured.  This is an empirical quantity which accounts for well geometry [Bouwer and Rice, 1976], and is estimated by AQTESOLV 4.01 when well specifications are entered into the input.  Once again, an anisotropy ratio of 0.01 was used for horizontal to vertical hydraulic conductivity.  Results obtained using the BR method are documented in Table C.3. C.1.4 Summary of Slug Results Cross-plots of calculated hydraulic conductivity between Hv and CBP, and Hv and BR are shown in Figure C.8.  The plots show that the Hv and BR methods produce comparable results, whereas there is considerable scatter among the values calculated using the CBP method.  In Appendix C – Estimation of Hydrogeologic Parameters 235 addition, data from rising head tests appear to give the most consistent, reproducible results in all cases, although values calculated using the CBP method are consistently lower. The observed scatter in results from falling head versus rising head tests, and from the CBP versus other methods reflects the greater difficulty in fitting a slope to the test data.  This is expected given the scatter in the falling head data, as well uncertainty when applying visual techniques in the CBP method.  Figure C.9 shows a cross-plot between Hv and BR methods used for determination of conductivity in the channel-fill silt unit.  Again, both of these methods appear to produce consistent, reproducible results from the collected data, and are therefore considered to be reliable calibration values. The calculated values of aquifer hydraulic conductivity are summarized graphically in Figure C.10 (a) and (b), and upscaled (i.e. geometric averages) values (+/- 1SD) are shown in Figure C.11.  Conductivity estimates for the aquifer made through Hv and BR analyses appear to be well constrained around 6.5 x 10-4 m/s (Tables C.3 and C.4), which is a reasonable value for a clean sand [Freeze and Cherry, 1979].  Results calculated for the silt unit (Figure C.10(c)) are also reasonable estimates for a silt, or silty sand (~2 x 10-7 m/s, [Freeze and Cherry, 1979]. There are no apparent spatial trends of hydraulic conductivity in the aquifer or the channel fill unit (Figure C.12) likely because there are insufficient data to characterize such trends with so few wells.  Storage estimates are highly variable (Figure C.10 (d)), reflecting the uncertainty with fitting CBP type curves to the data, and therefore are only accepted with caution. C.2.0 Time-Drawdown Analysis and Superposition Modeling To make use of time-drawdown data resulting from ongoing irrigation pumping in the area, periods of high-frequency head data collected during 2007 and 2008 were selected for analysis using Theis [1935] analytical models.  These analyses were meant to provide a first cut estimation of parameters for numerical modeling based on known pumping schedules, in spite of the numerous simplifying assumptions about the aquifer associated with the solutions.  Major uncertainties with these methods will be reflected in estimates of elastic storage, which will be over predicted if the aquifer exhibits leaky or unconfined behaviour.  However, the early time portion of a leaky or unconfined response often coincides with the drawdown predicted by the Theis solution, and may therefore provide an estimate of elastic storage [Narasimhan and Zhu, 1993]. Appendix C – Estimation of Hydrogeologic Parameters 236 An automated parameter estimation method to minimize residual error was performed using MATLAB (Section C.2.1), and estimates are compared to values computed using standard semi-log regression methods (Section C.2.2). C.2.1 Least Squares Fitting to Theis Solution – Superposition Modeling The premise of this method is to minimize the total squared misfit between modeled and observed drawdowns using the Theis model in order to deduce aquifer parameters.  Aquifer transmissivity and storage were estimated by fitting the water level data to the Theis solution using 3 nearby pumping wells (Table C.5).  Drawdown was estimated assuming that the static water level, h0, in each observation well was equal to the last measured head value prior to when pumping began.  Parameters were calculated using a least squares method in MATLAB in order to minimize the difference between the observed and drawdown predicted using the Theis solution: Eqn. C9      2 2 ,;,,  obsn obsjitheis tsSTtrQs  where Eqn. C10 )( 4 uW T Q stheis    Eqn. C11 Tt Sr u 4 2   and: nobs = number of drawdown data points [-]; stheis = Theis modeled drawdown at time, t, for aquifer parameters Ti and Sj [L]; Q = pumping rate [L3/T]; Appendix C – Estimation of Hydrogeologic Parameters 237 r = distance of the observation well from the pumping well [L]; t = elapsed time since pumping began [T]; Ti  =  range in transmissivity for a 30m thick aquifer [L 2/T]; Sj = range in storage [-]; sobs = observed drawdown at time t [L] W(u) = the exponential integral of u, or the well function [-] The values of Χ2 were calculated for transmissivity values within the range 1.5x10-1 to 12 m2/min (10-5 m/s to 10-3 m/s assuming an aquifer with a thickness of 30m), and the storage coefficient was varied from 10-4 to 0.2.  Trial calculations of T, S and X2 values helped to narrow the range of values selected for parameter estimation. Some limitations of the above method include: 1) The aquifer is confined.  This means that the drawdown response follows the shape of the well function, W(u), as time progresses.  This is likely not the case, but because the duration of each pumping test is relatively short (<4 hours), it is possible that the drawdown response in these tests will resemble confined conditions; 2) The aquifer is uniformly 30m thick.  The interpreted stratigraphy at the site indicates that the Holocene sands extend to a maximum depth of ~30m below the surface.  Below the inhabited area (i.e. near the channel fill silt) however, the sand may be considerably thinner; 3) Drawdown is only the result of specified pumping wells.  Although pumping schemes from wells 4, 33 and 50 are available for these test periods, it cannot be guaranteed that other, distant irrigation wells are not contributing to the observed drawdown; 4) The model requires that flow to the pumping well is horizontal.  However, well 4 is screened below the inferred paleohorizon (Section 4.1.4).  As a result, there must be vertical components of flow towards this well. Appendix C – Estimation of Hydrogeologic Parameters 238 5) Theis assumes an initial steady condition of the water table.  At the beginning of these “tests”, it is certain that the water table is not at steady state, because the aquifer is pumped erratically at the onset of irrigation season (~November).  Therefore, transient effects are superimposed on the drawdown pattern and cannot be separated from observation data. Semi-confined or unconfined analytical models (e.g. Hantush and Jacob, [1955], Boulton, [1954], and Neuman, [1972]) may in fact be more appropriate for this analysis, as they account for inflow from above units, or vertical flow due to aquifer drainage.  The mathematical solutions of all of these are analogous to the Theis model, except for modifications made to the exponential integral to incorporate additional parameters (e.g. aquitard conductivities and thicknesses, specific yield and vertical aquifer conductivity).  In our case, implementing these models would involve the introduction of additional uncertainty (e.g. aquitard parameters) as well as added tedium with extra calculations involving specific yield in MATLAB.  Given the uncertainties in the data, particularly with pumping (i.e. the potential of numerous wells pumping), as well as the short duration of tests, the added complexity of using these models is unwarranted for the purpose of the exercise. Because the Theis method incorporates numerous simplifying assumptions in this case, it is only intended as a first cut at estimating hydraulic conductivity and storage, prior to numerical modeling.  Moreover, calculations of storage using this analytical method may provide insight into confinement conditions of the aquifer, that is, high storativity values >10-4 ft-1or 3.28 x 10-4 m-1 [Fetter, 2001] calculated using the Theis method may be an indication of semi-confined conditions where specific yield plays a role in the drawdown response [Boulton, 1954]. Tests 1 and 2 involved pumping from single wells, (W4 and W50 respectively), thus calculations of X2 for these events followed Equation C9 directly.  Total drawdown observed during test 3 however was the result of pumping from W4, W50 and W33.  This required the theoretical drawdown in Equation C9 to be modified in order to include the effects of pumping at multiple wells, Q, with variable radial distances, r, from the observation points.  This was accomplished through linear superposition of the Theis solution: Eqn. C12   wellsn jinnntheis STtrQss ,;,  Appendix C – Estimation of Hydrogeologic Parameters 239 where nwells represents the individual pumping wells.  As discussed in Section 4.4.1, pumping rates at individual wells were estimated in the field by measuring the time required to fill a bucket with the outflow.  These rates were applied to the respective pumping tests in the modeling exercise.  Radial distances from observation points to the pumping wells were calculated from GPS data (Table C.6), and also used in the analytical model. Time-drawdown data as well as fits to the Theis solution are plotted in Figure C.13 (a)-(c).  The plots show that at early times, observed drawdown appears to be systematically higher than predicted values, implying that the aquifer in Gotra does not conform to the inherent assumptions of the Theis solution.  Data begin to conform to Theissian behaviour at the observation wells after about 5-15 minutes of pumping in the W4 and Superposition pumping tests, but not so much in the W50 pumping test.  Discrepancies in the fit to the Theis solution are likely the result of initial transient effects (i.e. well bore storage) at early times, and possibly interference pumping from irrigation wells in the W50 test. Contours of Log(X2) were plotted (Figure C14) by interpolating the results in Log(K)-Log(S) space (Kriging), to investigate whether the estimated parameters were correlated.  The plots show that the S and K values which provide the best fit to the data are associated with unique X2 minima in most of the wells, with one exception being GSI0713 in Pumping Test 2.  Here, multiple combinations of S and K may provide a reasonable fit to the data.  Test results at the majority of the wells however indicate that a unique head distribution will be produced at the site for linear changes in parameter values when a numerical model is constructed.  Parameter sets associated with minimum X2 values are shown in Table C.7. C.2.2 Semi-Log Methods The purpose of reanalyzing the data using semi-log fitting methods was to provide a check to the values estimated using the Theis solution in the previous section.  The Cooper and Jacob [1946] method involves a simplification of the Theis equation by approximation of the well function for large times, so that aquifer transmissivity can be calculated according to: Eqn. C13` s tQ T    4 log3.2  Where: Appendix C – Estimation of Hydrogeologic Parameters 240 T = transmissivity, [L2/T], Q = pumping rate [L3/T], ∆s = difference in drawdown measured at elapsed time values t2 and t1 [L]; ∆logt = logarithmic difference in time of measurements at t2 and t1 [T]. Since the calculation of transmissivity only requires differences in drawdown, ∆s, the equation becomes: Eqn. C14  214 log3.2 hh tQ T      where h2 and h1 represent water level values at given times after pumping began. A similar calculation using time recovery data was used to estimate transmissivity also using a modification of the Theis [1935] recovery calculation: Eqn. C15  124 'log3.2 hh tQ T      Where: t‟ = time since pumping ceased. Equations C14 and C15 are independent of static water levels, which is convenient since pumping appears to be influencing the piezometric surface at all times, and a reliable measure of static head value is not possible.  Although calculation of aquifer transmissivity remains valid, it is not possible to estimate the storage coefficient using head instead of drawdown values with this method. Because these methods are simplifications of the Theis solution described in the previous section, the same assumptions in the previous analysis apply, in addition to the following condition: Appendix C – Estimation of Hydrogeologic Parameters 241 Eqn. C16 05.0 4 2  u Tt Sr  Where: r = distance of observation well from pumping well, [L]; S = storage coefficient, [-]; T = aquifer transmissivity, [L2/T]; t = elapsed time since pumping began, [T]. Values of u < 0.05 are necessary to truncate the exponential integral of u; that is, this condition ensures that values beyond the second term in the series become negligible [Cooper and Jacob, 1946].  T and S values from Theissian analyses were used to estimate the pumping time required from each testing period to satisfy this condition (Table C.8).  As the lengths of Pumping Tests 1 and 2 are 56 and 177 minutes respectively, it is only possible to perform this analysis on GSI0603 for Pump Test 1, and GSI0603 and GSI0609 for Test 2.  For recovery calculations, data from GSI0603 may be used from Recovery Test 1, and only GSI0603 and GSI0609 may be used for Recovery Test 2 (Table C.8). Hydraulic conductivity values were computed assuming a 30m thick aquifer, and are summarized in Table C-9.  Linear regression of head versus log-time data is shown in Figures C.15 (a) to (d). C.2.3 Summary and Discussion of Time-Drawdown Analysis Variations in hydraulic conductivity and storage obtained from least squares fitting of drawdown data to the Theis solution as well as with straight line methods are plotted in Figure C.16. Geometric means of aquifer hydraulic conductivities estimated using Theis and superposition modeling vary from 5.15x10-4 to 8.19x10-4 m/s (Table C.7), which is a reasonable range of values for a sandy aquifer [Freeze and Cherry, 1979].  Values calculated using the log-linear approximation are similar, and range from 5.52x10-4 to 2.24x10-3 m/s (Table C.9).  Figure C.16 also shows that the parameters estimated from Pumping Test 1 (W4) generally are the most Appendix C – Estimation of Hydrogeologic Parameters 242 tightly constrained, whereas estimates made from Pumping Test 2 (W50) appear to vary the most. Storage values tend to vary over orders of magnitude (Table C.7).  The fact that some storage estimates are quite high (up to ~10-2 (storativity) or 3x10-3 m-1 for a 30m thick aquifer) indicates that the Gotra aquifer may experience conditions close to semi or unconfined, and therefore there is inconsistency in using a confined model for such an aquifer.  Storage estimates in these tests more likely represent a combination of elastic storage and specific yield, however, this is difficult to confirm since the inflection point that occurs in time-drawdown data from an unconfined aquifer test is not apparent in any of the data sets.  This is likely because sufficient time has not elapsed in any of the pumping periods for the aquifer to undergo leaky conditions, or to approach a time when delayed yield ceases to effect drawdown.  Consequently, it is not possible to extract a reasonable estimate of specific yield from the current data; however, there is an obvious need to consider the effects of leaky or unconfined flow.  To perform a pumping test of adequate length would require a power source independent from current the village lines, which are subject to daily load shedding, and cannot be relied upon for several days of continuous pumping. Other complications with these analyses include initial transient conditions of the piezometric surface, the possible effect of additional pumping from irrigation wells and the likelihood that pumping wells are partially penetrating.  The fact that the piezometric surface has not reached a state of equilibrium at the beginning of each pump test means that the true drawdown as a result of pumping cannot be calculated, or even that desaturation of the aquifer has occurred and nature of drawdown cannot be properly matched by the well function.  Failing to account for the possibility of interference from distant irrigation wells underestimates the impact they may have on observed drawdown, resulting in erroneous estimates of transmissivity and storage. Partially penetrating wells induce vertical components of flow around the screen, invalidating the general assumption that the well only receives horizontal flow.  This effectively increases flow to the screen and can result in extra head loss near the well, rendering predictive solutions inaccurate. Rather than pursue additional analytical models to remedy the above discrepancies due to leaky-unconfined conditions or partially penetrating wells (e.g. Boulton [1954], Neuman [1974]; Hantush –Jacob [1955]), it is more efficient to construct a numerical model that will also allow the incorporation of aquifer geometry and initial transient conditions of the piezometric surface. Appendix C – Estimation of Hydrogeologic Parameters 243 Parameter estimates made in this section from analytical solutions of time-drawdown data will however help to define calibration ranges for the numerical model. C.3.0 Grain Size Methods Particle size separation was performed on fine-grained sediment samples collected from the observation boreholes, and the data were used to estimate hydraulic conductivities of fine- grained materials.  As discussed in Section 3, only intact “clay” samples were kept for laboratory analysis because representative coarse grained materials could not be retained due to washing of fine components from the bulk sample.  Values of d10 and d60 (grain sizes for which 10% and 60% of the total grains are finer, respectively) were estimated from grain size curves plotted from the data (Appendix A, Figure A.2), and coefficients of uniformity, U, were computed as d60/d10 to help determine an appropriate method for calculating conductivity (Table C.10). The Hazen Method [Hazen, 1911; Kresic, 1997], is valid for well sorted samples with U<5 and 0.1mm <d10 < 3mm, and the Breyer Method [Kresic, 1997] is appropriate for heterogeneous samples where 1<U<20 and 0.06 mm <d10 < 0.6 mm.  Based on the available data, neither method is appropriate for conductivity calculations because sample grain sizes are simply too small.  However, the Breyer method was still used to calculate hydraulic conductivity to provide an approximate estimate.  This method is preferable to the Hazen method because it is for poorly sorted samples, and values of U in the dataset are typically greater than 5. Hydraulic conductivity using the Breyer method is computed by: Eqn. C17 2 10dC g K b    Where: K = hydraulic conductivity, [L/T]; g = gravitational constant, [L/T2] μ = kinematic viscosity; [L2/T]; Appendix C – Estimation of Hydrogeologic Parameters 244 Cb = empirical constant = 6x10-4log(500/U), [-] [Kresic, 1997]; d10 = the grain sizes for which 10% of the total grains are finer, [L]. Results from this calculation are summarized in Table C.10. Hydraulic conductivity values were also calculated using the USDA computer program ROSETTA [Schaap et al., 2001].  The program uses five hierarchical pedotransfer functions (PTFs) in order to estimate the saturated hydraulic conductivity of sediment from textural class data deduced from grain size analyses [Schaap et al., 2001].  Saturated conductivity estimates are made based on the pore-size model of Mualem [1976], and the comparison of textural class data to measured parameters surmised in a database of 1306 soil samples, collected from temperate to subtropical zones within North America and Europe [Schaap et al., 2001].  The program also provides uncertainty estimates of conductivity using artificial neural network and bootstrap models [Effron and Tibshirani, 1993].  Conductivity estimates made using the ROSETTA method are also tabulated in Table C.10.  A discussion of parameter values obtained from grain-size estimates is found in Section 4.2.2.3. Appendix C – Estimation of Hydrogeologic Parameters 245   Figure C.1: Schematic of a falling-head slug test.  R represents well screen radius (with or without gravel pack, (Hv)); r, rw = well casing radius (Hv); rs = screen radius + gravel pack (CBP); rc = radius of piezometer interior (CBP).  The radius of the gravel pack and the annulus of the well were considered negligible in the piezometers screened in the sand because the well diameter ~ the diameter of the drill pipe.  Thus, rw=rc=rs in the sand.  In the piezometers screened in the silt, the gravel pack radius was considered significant because the piezometer diameter (1 –inch, 2.54cm) was small compared to the drill pipe (2-inch, 5.08cm).  Thus, rs > rw in these piezometers.   Appendix C – Estimation of Hydrogeologic Parameters 246   Figure C.2: Examples of raw data from falling (top) and rising (bottom) head slug tests at GSI0605. Oscillations in the falling head data are likely the result of greater energy being added to the system due to dropping the slug. Thus, over-damped analyses may be applied to all data. GSI 06 05 Falling Head Tests -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 0 5 10 15 20 25 30 Elapsed Time (sec) D is p la c e m e n t (m ) GSI 06 05 F01 GSI 06 05 F02 GSI 06 05 F03 GSI 06 05 Rising Head Tests -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0 5 10 15 20 25 30 Elapsed Time (sec) D is p la c e m e n t (m ) GSI 06 05 R01 GSI 06 05 R02 GSI 06 05 R03 Appendix C – Estimation of Hydrogeologic Parameters 247  Figure C.3: Recovery of water table after purging piezometers near GSI0605. Symbols represent data collected from individual piezometers at 3 distinct depths: 20ft (green), 37ft (red) and 56ft (white).  Y-axis measurements are depths to water from piezometer casing tops.  0 1 2 3 4 5 6 7 8 0 5000 10000 15000 20000 25000 D e p th  t o  W a te r (m ) Elapsed Time (sec) Aquitard Bail Tests 20 ft 37 ft 56 ft Appendix C – Estimation of Hydrogeologic Parameters 248   Figure C.4: Semi-log plots of normalized displacement versus elapsed time observed in falling (top) and rising (bottom) head slug tests.  Falling Head Tests -3 -2.5 -2 -1.5 -1 -0.5 0 0 5 10 15 20 25 30 t (seconds) L o g [H (t )/ H 0 ] GSI0604 GSI0605 GSI0606 GSI0607 GSI0608 GSI0609 GSI0610 Rising Head Tests -3 -2.5 -2 -1.5 -1 -0.5 0 0 5 10 15 20 25 t (seconds) L o g [H (t )/ H 0 ]             GSI0604 GSI0605 GSI0606 GSI0607 GSI0608 GSI0609 GSI0610 Appendix C – Estimation of Hydrogeologic Parameters 249  Figure C.5: Semi-log plot of normalized displacement versus elapsed time observed in piezometers installed adjacent to GSI0605 as a result of well purging.  Note that piezometers are screened within the channel-fill unit thus recovery times are significantly longer than in regular slug tests.  -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0 5000 10000 15000 20000 25000 L o g [H (t )/ H 0 ] t (seconds) Channel-Fill Bail Tests 20 ft 37 ft 56 ft Appendix C – Estimation of Hydrogeologic Parameters 250   Figure C.6: Examples of semi-logarithmic plots of normalized displacement of the water table during rising head slug tests (top).  The CBP type curves used to match with this data are also shown (bottom).  GSI 06 04 Rising Head CBP Analysis -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 Log(t) H (t )/ H 0 Rising Head 01 Rising Head 02 Rising Head 03 CBP Type Curves - GSI 06 04 Rising Head -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 Log(Tt/r 2 ) H (t )/ H 0 1.00E-01 1.00E-02 1.00E-03 1.00E-04 1.00E-05 1.00E-06 1.00E-07 Alpha Appendix C – Estimation of Hydrogeologic Parameters 251   Figure C.7: Examples of semi-logarithmic plots of normalized displacement of the water table during falling head slug tests (top).  The CBP type curves used to match with this data are also shown (bottom).  GSI 06 04 Falling Head CBP Analysis -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 Log(t) H (t )/ H 0 Falling Head 01 Falling Head 02 Falling Head 03 CBP Type Curves - GSI 06 04 Falling Head -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 Log(Tt/r 2 ) H (t )/ H 0 1.00E-02 1.00E-03 1.00E-04 1.00E-05 1.00E-06 1.00E-07 Alpha Appendix C – Estimation of Hydrogeologic Parameters 252 (a)  (b)  (c)  (d)  Figure C.8:  Comparison of aquifer conductivity results from falling head, and rising head data for different analysis methods.  Falling Head Results 1.00E-04 1.00E-03 1.00E-02 1.00E-04 1.00E-03 1.00E-02 K (m/s) CBP K  ( m /s ) H v o rs le v GSI0604 GSI0605 GSI0606 GSI0607 GSI0608 GSI0609 GSI0610 y=x Rising Head Results 1.00E-04 1.00E-03 1.00E-02 1.00E-04 1.00E-03 1.00E-02 K (m/s) CBP K  ( m /s ) H v o rs le v GSI0604 GSI0605 GSI0606 GSI0607 GSI0608 GSI0609 GSI0610 y=x Falling Head Results 1.00E-04 1.00E-03 1.00E-02 1.00E-04 1.00E-03 1.00E-02 K (m/s) Bouwer-Rice K  ( m /s ) H v o rs le v GSI0604 GSI0605 GSI0606 GSI0607 GSI0608 GSI0609 GSI0610 y=x Rising Head Results 1.00E-04 1.00E-03 1.00E-02 1.00E-04 1.00E-03 1.00E-02 K (m/s) Bouwer-Rice K  ( m /s ) H v o rs le v GSI0604 GSI0605 GSI0606 GSI0607 GSI0608 GSI0609 GSI0610 y=x Appendix C – Estimation of Hydrogeologic Parameters 253   Figure C.9:  Comparison of channel-fill conductivity results data for different analysis methods.   Channel-Fill Bail Test Results 1.00E-08 1.00E-07 1.00E-06 1.00E-08 1.00E-07 1.00E-06 K (m/s) Bouwer-Rice K  ( m /s ) H v o rs le v 20 ft 37 ft 56 ft y=x Appendix C – Estimation of Hydrogeologic Parameters 254 (a) Falling Head Slug Tests  (b) Rising Head Slug Tests    (c) Channel Fill Bail Tests   (d) Estimates of Storage from CPB analysis of Slug Tests   Figure C.10: Ranges of hydraulic conductivity values, K, and storage, S, obtained from slug-test data using different solution methods.  Conductivity values for the sand unit from falling and rising slug test data are shown in (a) and (b) respectively (N=18), and results from the bail tests in the silt unit are shown in (c).  Values for the aquifer obtained from falling head data are more scattered than those obtained from rising head tests, which is the result of excess energy produced as the slug was dispensed into the wells.  This is especially evident in the CBP plot in (a) that has the added handicap of visual interpretation.  Storage coefficient values (d) range over orders of magnitude, reflecting the uncertainty of the CBP curve fitting method.   1.00E-04 1.00E-03 1.00E-02 CBP Hv Bouwer-Rice K  ( m /s ) 1.00E-04 1.00E-03 1.00E-02 CBP Hv Bouwer-Rice K  ( m /s ) 1.00E-08 1.00E-07 1.00E-06 K  ( m /s ) Hv Bouwer-Rice 1.00E-08 1.00E-07 1.00E-06 1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 Falling Head Rising Head S Appendix C – Estimation of Hydrogeologic Parameters 255  Figure C.11: Geometric means of hydraulic conductivities, (+/- 1 standard deviation). Conductivity values obtained through the Hvorslev (Hv) and Bouwer-Rice (BR) analysis methods of the rising head data are considered the most reliable values for the aquifer.  These values range from 3.41 x 10-4 to 1.00 x 10-3 m/s.  Conductivity values for the silt unit range from 4.63 x 10-8 to 3.28 x 10-7 m/s.  (a)  (b)   Figure C.12: Variation of calculated hydraulic conductivity from slug tests (a) along a NE transect through the village, and (b) with depth.  No spatial trends are apparent in the calculated values.  1.00E-08 1.00E-07 1.00E-06 1.00E-05 1.00E-04 1.00E-03 1.00E-02 CBP Falling Hv Falling BR Falling CBP Rising Hv Rising BR Rising Hv Silt BR Silt K  ( m /s ) 0.00E+00 2.00E-04 4.00E-04 6.00E-04 8.00E-04 1.00E-03 1.20E-03 0 500 1000 1500 Distance (m) K  ( m /s ) GSI0604 GSI0605 GSI0606 GSI0607 GSI0608 GSI0609 GSI0610 0 5 10 15 20 25 30 1.00E-08 1.00E-06 1.00E-04 1.00E-02 1.00E+00 K (m/s) D e p th  ( m ) Aquifer Channel Fill Appendix C – Estimation of Hydrogeologic Parameters 256  (a)   (b)  (c)  Figure C.13:  Calculated drawdown from time-series data plotted with drawdown predicted from the Theis solution for well GSI0603.  Tests shown in (a) and (c) show a good visual fit to the Theis solution, whereas the response in (b) is not well matched by the model.  The response in (b) may be the result of interference of pumping from irrigation wells, or from transient conditions that existed prior to pumping of W50. GSI0603 Test 1 Fit to Theis Solution (Pumping at W4) 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0 10 20 30 40 50 60 Elapsed Time (min) D ra w d o w n Observed Theis GSI0603 Test 2 Fit to Theis Solution (Pumping at W50) 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0 50 100 150 200 Elapsed Time (min) D ra w d o w n Observed Theis GSI0603 Test 3 Fit to Theis Solution (Superposition of W4, 50 and 33 Pumping) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0 50 100 150 200 250 Elapsed Time (min) D ra w d o w n Observed Theis Appendix C – Estimation of Hydrogeologic Parameters 257    Figure C.14 (a): Log(X2) contours in Log(K)-Log(S) fields produced from least squares fitting of drawdown data to the Theis solution due to pumping in Well 4. -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0603 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0605 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0606 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) GSI0607 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0608 -4 -3 -2 -1 0 L o g  ( S ) Appendix C – Estimation of Hydrogeologic Parameters 258       Figure C.14 (b): Log(X2) contours in Log(K)-Log(S) fields produced from least squares fitting of drawdown data to the Theis solution due to pumping in Well 50.  -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0603 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0607 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0609 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0713 -4 -3 -2 -1 0 L o g  ( S ) Appendix C – Estimation of Hydrogeologic Parameters 259    Figure C.14 (c): Log(X2) contours in Log(K)-Log(S) fields produced from least squares fitting of drawdown data to the Theis solution due to pumping in Wells 4, 33 and 50.   -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0603 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0607 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0609 -4 -3 -2 -1 0 L o g  ( S ) -4 -3.5 -3 -2.5 Log (K) (m/s) GSI0713 -4 -3 -2 -1 0 L o g  ( S ) Appendix C – Estimation of Hydrogeologic Parameters 260 (a)  (b)  (c)  (d)  Figure C.15: Semi-log plots of head versus elapsed time used to estimate hydraulic conductivity with the Cooper-Jacob method. Pumping tests shown in (a) and (c) plot head versus time since pumping began (t), and recovery tests shown in (b) and (d) plot head versus time since pumping ceased.  Regression fits and coefficients of determination are also shown.   Pumping Test 1 (W4) y = -0.094x - 1.2687 R2 = 0.9919 -1.45 -1.4 -1.35 -1.3 -1.25 -1.2 0.00 0.50 1.00 1.50 2.00 Log (t) (min) H e a d  ( m ) Recovery Test 1 (W4) y = 0.115x - 1.4481 R2 = 0.9841 -1.4 -1.35 -1.3 -1.25 -1.2 -1.15 0.50 1.00 1.50 2.00 Log (t') (min) H e a d  ( m ) Pumping Test 2 (W50) y = -0.047x + 0.3382 R2 = 0.9584 y = -0.0687x + 0.0113 R2 = 0.9427 -0.2 -0.1 0 0.1 0.2 0.3 0.4 1.00 1.50 2.00 Log (t) (min) H e a d  ( m ) GSI0603 GSI0609 Recovery Test 2 (W50) y = 0.0275x + 0.2374 R2 = 0.8354 y = 0.0862x - 0.2032 R2 = 0.9948 -0.1 0 0.1 0.2 0.3 0.4 1.50 2.00 2.50 Log (t') (min) H e a d  ( m ) GSI0603 GSI0609 Appendix C – Estimation of Hydrogeologic Parameters 261 (a)  (b)  Figure C.16: Aquifer parameter estimations made by fitting data from pump tests to the Theis solution using a least squares method. The plots show variations in hydraulic conductivity (a) and storage (b) calculated using drawdown data obtained from various pump tests.   1.00E-04 1.00E-03 1.00E-02 W4  Theis W50 Theis W4, W33 and W50 Theis W4 CJ W50 CJ K  ( m /s ) ` 1.00E-04 1.00E-03 1.00E-02 1.00E-01 W4 W50 W4, W33 and W50 S Appendix C – Estimation of Hydrogeologic Parameters 262   Figure C.17: Comparison of conductivity values computed using different grain size methods. The ROSETTA method generally predicts systematically higher conductivities.  1.00E-08 1.00E-07 1.00E-06 1.00E-05 1.00E-08 1.00E-07 1.00E-06 1.00E-05 K (m/s) Breyer Method K  ( m /s ) R o s e tt a  M e th o d y=x Appendix C – Estimation of Hydrogeologic Parameters 263  Figure C.18: Hydraulic conductivity calculated for fine grained sediments using grain size data. Data are colour coded to reflect the interpreted geological unit that the sample was taken from. Diamonds represent values computed using the Breyer Method, and squares represent those calculated using the ROSETTA method.  0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 1.00E-08 1.00E-07 1.00E-06 1.00E-05 1.00E-04 K (m/s) D e p th  ( m ) Appendix C – Estimation of Hydrogeologic Parameters 264  Figure C.19: Geometric mean hydraulic conductivity values from grain size methods. The values shown for Lens deposits are arithmetic averages because they are not continuous hydrostratigraphic units. 1.00E-08 1.00E-07 1.00E-06 1.00E-05 1.00E-08 1.00E-07 1.00E-06 1.00E-05 Breyer Kgeo (m/s) R o s e tt a  K g e o  ( m /s ) Channel Fill Lens Older Alluvium Overbank y=x Appendix C – Estimation of Hydrogeologic Parameters 265  Table C.1: Summary of calculated aquifer parameters calculated using the CBP analysis method  Well Name Test No. α (Storativity) T (m 2 /s) K (m/s) Arithmetic Average (m/s) Test No. α (Storativity) T (m 2 /s) K (m/s) Arithmetic Average (m/s) 1 1.00E-02 6.67E-03 2.58E-04 1.41E-04 1.72E-04 1 1.00E-07 3.67E-04 4.56E-04 2.49E-04 1.95E-04 2 1.00E-07 3.74E-04 2.05E-04 2 1.00E-04 3.28E-04 1.79E-04 3 1.00E-02 3.14E-04 1.71E-04 3 1.00E-03 2.88E-04 1.57E-04 1 1.00E-02 3.37E-03 7.23E-04 4.31E-04 3.83E-04 1 1.00E-06 4.00E-07 1.00E-03 5.97E-04 5.14E-04 2 1.00E-07 1.80E-04 1.08E-04 2 1.00E-07 6.85E-04 4.09E-04 3 1.00E-04 1.02E-03 6.11E-04 3 1.00E-07 9.00E-04 5.37E-04 1 1.00E-02 1.00E-02 1.10E-03 9.99E-04 1.29E-03 1 1.00E-04 5.01E-05 1.56E-04 1.42E-04 1.66E-04 2 1.00E-02 1.74E-03 1.58E-03 2 1.00E-07 2.07E-04 1.89E-04 1 1.00E-06 1.00E-06 5.29E-04 2.48E-04 2.85E-04 1 1.00E-03 3.33E-04 3.28E-04 1.54E-04 2.07E-04 2 1.00E-06 6.85E-04 3.21E-04 2 1.00E-07 4.75E-04 2.23E-04 3 1.00E-07 5.19E-04 2.43E-04 GSI0608 1 1.00E-05 5.05E-04 1.03E-03 5.62E-04 3.88E-04 1 1.00E-07 1.00E-07 7.88E-04 4.31E-04 3.93E-04 2 1.00E-03 3.91E-04 2.14E-04 2 1.00E-07 6.48E-04 3.54E-04 1 1.00E-05 5.05E-06 9.16E-04 5.01E-04 5.55E-04 1 1.00E-04 5.01E-05 6.90E-04 3.77E-04 4.24E-04 2 1.00E-07 1.11E-03 6.08E-04 2 1.00E-07 8.61E-04 4.71E-04 1 1.00E-02 5.50E-03 3.28E-04 1.79E-04 2.14E-04 1 1.00E-03 7.00E-03 4.75E-04 2.60E-04 2.85E-04 2 1.00E-03 4.55E-04 2.49E-04 2 1.00E-02 5.43E-04 2.97E-04 3 1.00E-02 5.43E-04 2.97E-04 Sgeo 4.39E-04 Kgeo 3.79E-04 Sgeo 3.65E-05 Kgeo 2.88E-04 +/- 1 SD 3.86E-03 +/- 1 SD 3.83E-04 +/- 1 SD 2.60E-03 +/- 1 SD 1.34E-04 Geometric Average GSI0604 GSI0607 GSI0609 GSI0610 Rising Head TestsFalling Head Tests GSI0605 GSI0606 Appendix C – Estimation of Hydrogeologic Parameters 266  Table C.2: Summary of calculated aquifer parameters from Hvorslev Analysis  Well Name Falling Head Tests Rising Head Tests Test No. K (m/s) Arithmetic Average (m/s) Test No. K (m/s) Arithmetic Average (m/s) 1 3.57E-04 3.54E-04 1 4.45E-04 4.49E-04 2 3.48E-04 2 4.75E-04 3 3.58E-04 3 4.28E-04 1 7.93E-04 5.96E-04 1 1.07E-03 1.09E-03 2 7.01E-04 2 8.88E-04 3 2.93E-04 3 1.31E-03 1 3.39E-04 3.32E-04 1 3.16E-04 3.24E-04 2 3.26E-04 2 3.33E-04 1 6.71E-04 5.48E-04 1 5.99E-04 6.08E-04 2 4.25E-04 2 6.32E-04 3 5.92E-04 GSI0608 1 5.47E-04 4.91E-04 1 9.95E-04 9.53E-04 2 4.35E-04 2 9.12E-04 1 1.07E-03 9.91E-04 1 1.44E-03 1.14E-03 2 9.14E-04 2 8.36E-04 1 4.54E-04 4.73E-04 1 6.89E-04 6.82E-04 2 4.93E-04 2 6.58E-04 3 6.99E-04 Kgeo 5.09E-04 Kgeo 6.86E-04 +/- 1 SD 2.20E-04 +/- 1 SD 3.17E-04 GSI0610 Geometric Average GSI0604 GSI0605 GSI0606 GSI0607 GSI0609 Appendix C – Estimation of Hydrogeologic Parameters 267  Table C.3: Summary of calculated aquifer parameters from Bouwer-Rice Analysis  Well Name Falling Head Tests Rising Head Tests Test No. K (m/s) Arithmetic Average (m/s) Test No. K (m/s) Arithmetic Average (m/s) 1 3.04E-04 3.03E-04 1 3.82E-04 3.82E-04 2 3.00E-04 2 3.96E-04 3 3.04E-04 3 3.68E-04 1 8.89E-04 6.24E-04 1 1.00E-03 1.04E-03 2 7.07E-04 2 8.73E-04 3 2.76E-04 3 1.25E-03 1 3.23E-04 3.16E-04 1 3.05E-04 3.15E-04 2 3.09E-04 2 3.24E-04 1 6.16E-04 4.99E-04 1 5.87E-04 5.96E-04 2 3.81E-04 2 6.26E-04 3 5.75E-04 GSI0608 1 5.26E-04 4.54E-04 1 9.89E-04 9.32E-04 2 3.83E-04 2 8.75E-04 1 9.47E-04 7.73E-04 1 1.28E-03 1.04E-03 2 6.00E-04 2 8.04E-04 1 4.65E-04 5.09E-04 1 7.13E-04 6.71E-04 2 5.54E-04 2 6.30E-04 3 6.69E-04 Kgeo 4.73E-04 Kgeo 6.49E-04 +/- 1 SD 1.66E-04 +/- 1 SD 3.02E-04 GSI0606 GSI0607 GSI0609 GSI0610 Geometric Average GSI0604 GSI0605 Appendix C – Estimation of Hydrogeologic Parameters 268  Table C.4: Summary of calculated conductivity for the channel-fill silt.   Table C.5:  Time spans and pumping wells for which water level data was used in pump test analysis.   Table C.6: Distances (metres) from observation points in pumping tests to pumping wells obtained using GPS data.  Test No. K (m/s) Arithmetic Average (m/s) Test No. K (m/s) Arithmetic Average (m/s) G1 7.11E-08 2.05E-07 G1 9.22E-08 2.20E-07 G2 2.05E-07 G2 2.26E-07 G3 3.39E-07 G3 3.42E-07 R1 3.65E-07 3.68E-07 R1 3.58E-07 3.63E-07 R2 3.58E-07 R2 3.55E-07 R3 3.82E-07 R3 3.74E-07 W3 8.91E-08 8.72E-08 W3 8.92E-08 8.57E-08 W2 8.52E-08 W2 8.22E-08 Kgeo 1.87E-07 Kgeo 1.90E-07 +/- 1 SD 1.41E-07 +/- 1 SD 1.38E-07 Geometric Mean GSI0607-20ft GSI0607-37 ft GSI0607-56 ft Hvorslev AnalysisBouwer Rice Analysis Piezometer Name Pumping Test No. Well(s) Pumping Start Time End Time 1 4 2/23/2007 9:08 2/23/2007 10:04 2 50 2/15/2008 10:27 2/15/2008 13:24 3 4, 50, 33 2/13/2008 8:20 2/13/2008 12:09 Pumping Well GSI0603, 04 GSI0605 GSI0606 GSI0607 GSI0608 GSI0609 GSI0713 W4 74.90 131.70 89.01 41.65 93.99 283.33 204.83 W50 33.26 213.54 127.70 140.66 124.37 190.04 299.81 W33 31.48 162.30 143.35 104.36 67.86 244.74 252.51 Appendix C – Estimation of Hydrogeologic Parameters 269  Table C.7: Hydraulic conductivity and storage values providing the best fit of drawdown to the Theis solution.  Well Name K (m/s) S (-) X 2 GSI0603 6.67E-04 3.60E-04 5.22E-04 GSI0605 5.28E-04 3.10E-03 2.00E-04 Kgeo (m/s) 5.36E-04 GSI0606 4.17E-04 6.60E-03 1.55E-04 Sgeo (-) 3.29E-03 GSI0607 4.72E-04 1.80E-02 2.71E-04 GSI0608 6.39E-04 2.90E-03 9.49E-05 Well Name K (m/s) S (-) X 2 GSI0603 8.89E-04 2.40E-02 4.71E-03 GSI0607 2.78E-04 9.60E-02 1.09E-03 Kgeo (m/s) 5.15E-04 GSI0609 1.14E-03 1.40E-03 8.81E-04 Sgeo (-) 1.23E-02 GSI0713 2.50E-04 7.20E-03 4.04E-03 Well Name K (m/s) S (-) X 2 GSI0603 8.89E-04 1.10E-02 3.94E-03 GSI0607 5.83E-04 1.70E-02 5.29E-04 Kgeo (m/s) 8.18E-04 GSI0609 1.83E-03 3.80E-03 4.78E-03 Sgeo (-) 7.21E-03 GSI0713 4.72E-04 3.80E-03 1.54E-03 Test 2: Well 50 Pumping Test 3: Wells 4, 33 and 50 Pumping Appendix C – Estimation of Hydrogeologic Parameters 270  Table C.8: Calculation of testing time required for Cooper-Jacob analysis of pumping test data.  A minimum time requirement must be met in each case (according to Equation C-13) for accurate estimates of aquifer parameters using Cooper-Jacob analysis.  Drawdown and recovery times in most wells do not meet this requirement.   Table C.9: K (m/s) calculated using high frequency water level data and straight line methods  Well r (m) T(m 2 /s) S (-) Minimum t required for analysis (min) Pump (56 min) Recovery (161 min) GSI0603 74.9 2.00E-02 3.60E-04 8 Y Y GSI0605 131.7 1.58E-02 3.10E-03 283 N N GSI0606 89.01 1.25E-02 6.60E-03 349 N N GSI0607 41.65 1.42E-02 1.80E-02 184 N N GSI0608 93.99 1.92E-02 2.90E-03 111 N Y Well r (m) T(m 2 /s) S (-) Minimum t required for analysis (min) Pump (177 min) Recovery (784 min) GSI0603 33.26 2.67E-02 2.40E-02 83 Y Y GSI0607 140.66 8.33E-03 9.60E-02 18994 N N GSI0609 190.04 3.42E-02 1.40E-03 123 Y Y GSI0713 299.81 7.50E-03 7.20E-03 7191 N N Pumping Test 1 (W4 pumping) Pumping Test 2 (W50 Pumping) Well Pump Well 4 Recovery Well 4 Pump Well 50 Recovery Well 50 Arithmetic Average GSI0603 6.75E-04 5.52E-04 8.97E-04 7.15E-04 7.10E-04 GSI0609 N/A N/A 1.31E-03 2.24E-03 1.78E-03 Kgeo 6.75E-04 5.52E-04 1.08E-03 1.27E-03 1.12E-03 Appendix C – Estimation of Hydrogeologic Parameters 271  Table C.10: Grain-size parameters and calculated conductivities for fine grained samples collected from Gotra boreholes. Borehole Log Description Interpreted Unit Depth (m) d10 (mm) d60 (mm) U K 1  (m/s) K 2  (m/s) K 3  (m/s) K 4  (m/s) GSI-07-13  10-15 ft Brown Soft Clay Older Alluvium 3.81 0.0055 0.03 5.5 na na 3.01E-07 6.01E-06 GSI-07-13  15-20 ft Grey Silt Older Alluvium 5.334 0.0088 0.055 6.2 na na 7.65E-07 1.19E-05 GSI-07-13  20-30 ft Grey Silty Clay Older Alluvium 7.62 0.0044 0.026 5.9 na na 1.94E-07 3.91E-06 GSI-07-13  30-35 ft Grey Silty Clay Older Alluvium 9.906 0.0041 0.02 4.9 na na 1.74E-07 2.48E-06 GSI-07-13  35-38 ft Grey Hard Clay Older Alluvium 11.1252 0.0026 0.016 6.2 na na 6.58E-08 1.37E-06 GSI-07-13  38-43 ft Grey Soft Clay Older Alluvium 12.3444 0.002 0.007 3.4 na na 4.69E-08 9.95E-07 GSI-07-13  46-71 ft Grey Silty Clay Older Alluvium 17.8308 0.0021 0.0075 3.6 na na 4.85E-08 1.01E-06 GSI-07-13  85-93 ft Grey Soft Clay Older Alluvium 27.1272 0.0012 0.0034 2.8 na na 1.71E-08 2.81E-06 GSI-06-05 16-38 ft Grey Soft Clay Channel Fill 8.2 0.0029 0.01 3.5 na na 9.16E-08 1.32E-06 GSI-06-05 38-69 ft Grey Soft Clay Channel Fill 16.3 0.0034 0.03 8.8 na na 1.06E-07 1.74E-06 GSI-06-05  56 ft Grey Soft Clay Channel Fill 17.1 0.0022 0.01 4.5 na na 5.30E-08 1.02E-06 GSI-06-05  >56 ft Grey Soft Clay Channel Fill 17.1 0.0034 0.017 5 na na 1.20E-07 2.00E-06 GSI-06-07 20-55 ft  Grey Soft Clay Channel Fill 11.4 0.0042 0.02 4.8 na na 1.85E-07 2.25E-06 GSI-07-14  7-10 ft Brown Soft Clay Overbank 2.6 0.0057 0.029 5.1 na na 3.30E-07 5.53E-06 GSI-07-14  15-50 ft Grey Soft Clay Channel Fill 9.9 0.0028 0.012 4.3 na na 8.14E-08 1.21E-06 GSI-07-14A  20-68 ft Grey Soft Clay Channel Fill 13.4 0.0031 0.015 4.8 na na 1.02E-07 1.50E-06 GSI-07-14A  95-105 ft Grey Sticky Clay Older Alluvium 30.5 0.0012 0.017 14.1 na na 1.17E-08 1.31E-06 GSI-07-14A 105-115 ft Grey Hard Clay Older Alluvium 33.5 0.0021 0.017 8.2 na na 3.94E-08 1.43E-06 GSI-06-06 12-15.5 ft Grey Soft Clay Channel Fill 4.2 0.0035 0.019 5.4 na na 1.24E-07 2.39E-06 GSI-06-06  20-27 ft Grey Soft Clay Channel Fill 7.2 0.0026 0.015 5.8 na na 6.76E-08 1.46E-06 GSI-06-06 10-30 ft Grey Soft Clay Channel Fill 6.1 0.0055 0.025 4.6 na na 3.16E-07 3.63E-06 GSI-06-08 15-31 ft Grey Soft Clay Channel Fill 7 0.0027 0.012 4.4 na na 7.88E-08 1.30E-06 GSI-06-03 7-17 ft Grey Soft Clay Channel Fill 3.7 0.0025 0.017 6.7 na na 6.23E-08 1.25E-06 GSI-07-15  61-62 ft Grey Soft Clay Lens 18.7 0.004 0.025 6.2 na na 1.61E-07 3.51E-06 GSI-06-09 3-15 ft Brown Sticky Clay Overbank 2.7 0.0021 0.015 7.2 na na 4.15E-08 1.25E-06 GSI-06-10 85-86 ft Grey Soft Clay Lens 26.0604 0.0048 0.027 5.6 na na 2.00E-07 4.90E-06 1  method of Hazen (Kresic, 1997), applicable for: U < 5 and (0.1 mm < d10 < 3 mm) 2  method of Breyer (Kresic, 1997), applicable for: (1 < U < 20) and (0.06 mm < d10 < 0.6 mm) 3  approximate method of Breyer 4  ROSETTA pedotransfer function method Appendix D – Climate Data and Estimation of Hydrologic Parameters 272 APPENDIX D – CLIMATE DATA AND ESTIMATION OF HYDROLOGIC PARAMETERS This section summarizes the data used and the procedures followed for the estimation of reference monthly evapotranspiration (ETo), crop evapotranspiration (ETc i), potential evaporation from ponds (Ep), and surface runoff values for the study area.  Hydrometeorological data were obtained from various government weather stations, evapotranspiration estimates were made using the FAO Penman-Monteith Method [R G Allen et al., 1998] and pond evaporation was computed using a method described by Maidment, [1993].  Surface runoff was estimated using a modification of the empirical USDA Soil Conservation Service (SCS) Runoff Curve Number (CN) method, for Indian conditions [SCD, 1972].  All precipitation and temperature data are summarized in Table D.1. D.1.0 Evapotranspiration Agricultural fields that surround the Gotra site are planted to numerous types of crops, and as such, will have differing evapotranspiration rates throughout the year depending on which are in season.  The goal of this section is to estimate approximate monthly evapotranspiration rates for the various land zones, which include “High Lands”, “Low Lands”, ponds and village areas. Subdividing the area into these zones allows variable evapotranspiration associated with land use (i.e. crop types, vegetative cover, free water surfaces) to be accounted for [R G Allen et al., 1998].  This procedure consists of first determining a reference monthly evapotranspiration, ETo (Section D.2.1), then multiplying these values by monthly crop coefficients Kci (Section D.2.2) to obtain unique values for evapotranspiration of an area depending on land use [R G Allen et al., 1998].  A separate method is used for estimating evaporation from ponds Ep, which is also known as potential evaporation from a free surface [Maidment, 1993], and is discussed in Section D.2.3. D.1.1 Calculation of Reference Evapotranspiration, ETo Reference crop evapotranspiration (ETo) was calculated using monthly climate Normals for Kolkata (http://www.imd.gov.in/section/climate/kolkataweb.htm) and the FAO Penman-Monteith equation [R G Allen et al., 1998], which determines evapotranspiration from the hypothetical grass reference surface.  ETo was calculated according to: Appendix D – Climate Data and Estimation of Hydrologic Parameters 273 Eqn. D1   )34.01( 273 900 )(408.0 2 2 u eeu T GR ET asn O         where: ETo = reference evapotranspiration [mm day -1] Rn = net radiation at the crop surface [MJ m -2 day-1] G = soil heat flux density [MJ m-2 day-1] T = mean monthly air temperature at 2 m above the ground surface [°C] u2 = wind speed at 2 m above the ground surface [m s -1], assumed to be 2 m s-1 es = saturation vapour pressure [kPa] ea = actual vapour pressure [kPa] es - ea = saturation vapour pressure deficit [kPa] Δ = slope vapour pressure curve [kPa °C-1] γ = psychrometric constant [kPa °C-1] Calculation of the net crop surface radiation, Rn, involved the following series of relationships: Eqn. D2 nlnsn RRR   Eqn. D3   sns RR  1 Eqn. D4 asss R N n baR         Appendix D – Climate Data and Estimation of Hydrologic Parameters 274 Eqn. D5 30 1 r n N n   Eqn. D6                    35.035.114.034.0 2 4 min, 4 max, so s a KK nl R R e TT R   Eqn. D7   aso RzR 510275.0  Eqn. D8  ssrsca dGR   sincossinsin )60(24   Eqn. D9        Jdr 365 2 cos033.01   Eqn. D10   tantanarccos s and Eqn. D11        39.1 365 2 sin409.0 J    where: Rns = net solar or shortwave radiation [MJ m -2 day-1] α = albedo or canopy reflection coefficient, which = 0.23 for the hypothetical green grass reference crop [-] Rs = incoming solar radiation [MJ m -2 day-1] as+bs = fraction of extraterrestrial radiation reaching the earth on clear-sky days (n = N).  Values of as  =0.25 and bs =0.5 were used in lieu missing radiation data, as recommended by Allen et al., [1998]. Appendix D – Climate Data and Estimation of Hydrologic Parameters 275 n = number of daylight hours [hour] N = maximum possible duration of daylight hours [hour] n/N = relative sunshine duration [-] nr = number of rainy days [day] obtained from the India Meteorological Service website Rnl = net outgoing longwave radiation [MJ m -2 day-1] ζ = Stefan-Boltzmann constant [4.903 x 10-9 MJ K-4 m-2 day-1] Tmax, K = maximum absolute temperature during the 24-hour period [K = °C + 273.16] Tmin, K = minimum absolute temperature during the 24-hour period [K = °C + 273.16] Rso = clear-sky radiation [MJ m-2 day-1] z = station elevation above sea level, [m], assumed 10m for our case Ra = extraterrestrial radiation [MJ m-2 day-1] Gsc = solar constant = 0.0820 MJ m-2 min-1 dr = inverse relative distance Earth-Sun ωs = sunset hour angle [rad] ψ = latitude [rad] = 23 N = 0.401 radians at our site δ =  solar declination [rad] and Appendix D – Climate Data and Estimation of Hydrologic Parameters 276 J = number of the day in the year between 1 (1 January) and 365 or 366 (31 December), known as “Julian Day”.  Values for J were taken as the 15th of each month. The monthly soil heat flux density, G, was calculated using the relationship: Eqn. D12  1,1,, 07.0   imonthimonthimonth TTG where Tmonth, i+1 and Tmonth, i-1 represent the average monthly temperatures prior to, and following the month in question. The psychrometric constant is given as: Eqn. D13   PcP  where: P = atmospheric pressure [kPa] λ = latent heat of vaporization [MJ kg-1] =2.45 MJ kg-1 for normal temperature ranges [R G Allen et al., 1998] ε = ratio or water MW of water vapour to dry air =0.622 and cP = specific heat capacity at constant pressure, [MJ kg -1 °C-1]=1.013 10-3 MJ kg-1 °C-1 for average atmospheric conditions [R G Allen et al., 1998]. Atmospheric pressure was calculated as a function of station elevation (z) according to a simplification of the ideal gas law, assuming 20°C for a standard atmosphere [R G Allen et al., 1998]: Eqn. D14 26.5 293 0065.0293 3.101         z P  Appendix D – Climate Data and Estimation of Hydrologic Parameters 277 Saturation vapour pressures, es, and actual vapour pressure, ea, were calculated using: Eqn. D15     2 minmax TeTee oo s    Eqn. D16           3.237 27.17 exp6108.0 T T Teo  and Eqn. D17           3.237 27.17 exp6108.0 dew dew dew o a T T Tee  where Tmax and Tmin were determined from average daily minimum and maximum temperatures for each month from Kolkata climate normals.  Tdew was assumed to equal Tmin. [Maidment, 1993]. The slope of the vapour pressure curve, Δ, was calculated according to: Eqn. D18   2 3.237 3.237 27.17 exp6108.04098                T T T  Finally the wind speed at 2m above the ground surface, u2, was assumed to be 2 m/s, as no wind data were available from weather stations.  This value represents an average of over 2000 weather stations across the planet, measured at a height of 2m, as reported by Allen et al., [1998]. D.2.2 Calculation of Crop Evapotranspiration, ETc Crop evapotranspiration, ETc, accounts the evaporation and transpiration characteristics of cropped surfaces compared to the reference surface (Section D.2.1), by multiplying ETo by a crop coefficient, Kc  [R G Allen et al., 1998]: Appendix D – Climate Data and Estimation of Hydrologic Parameters 278 Eqn. D19 occ ETKET   Since various types of crops are cultivated in the fields surrounding the research site, it is necessary to apply specific Kc values to regions where crop types are known to accurately estimate ETc.  This applies for both agricultural lands, as well as the village area, thus, Kc values must be established for these areas, and will contribute to different recharge zones. There are two types of cultivated lands in the research vicinity: High Lands and Low Lands, [Pal, 2009].  In the High Lands, various crops are cultivated throughout the field in rotation. From February to May, jute is grown in the High Lands, and from June to October, Aman rice is grown.  From October to December, the High Land fields are divided to grow potatoes, mustard seed, cauliflower, cabbage and tomato, and from January to February, they are left fallow.  The Low Land crops consist of two types: Aman and Boro rice, which are grown from June to October and from December to April, respectively.  The Low Land fields are left fallow for the months of May and November.  A summary of crop types and months of cultivation are presented in Table D.2. Allen et al., [1998] present tabulated Kc values for various crops and their associated development stages.  The Kc values relevant to the Gotra site are shown in Table D.3.  Because no local information about lengths of crop development stages for the research area are available, the standard crop development durations provided in Allen et al., [1998] were scaled over the lengths of time that the respective crops were farmed in Gotra (Table D.4).  These scaled growth periods were used to calculate daily crop coefficients associated with the Gotra crops in their respective cultivation periods, and are plotted in Figure D.1. The curves show in Figure D.1 were calculated as explained by Allen et al., [1998], where Kc values from the end of the initial to the beginning of the mid stage, and from the end of the mid to the end of the late stage are just linear interpolations between the two points.  Monthly averages of these values were used in the following calculations of ETc, since it is monthly evapotranspiration of interest. To give consideration to the various coefficients associated with crops grown simultaneously in the High Lands (i.e. from October to December) the monthly crop coefficient was calculated according to: Appendix D – Climate Data and Estimation of Hydrologic Parameters 279 Eqn. D20    cropsn j jcjc KfK 1 where j represents a given crop (see Table D.3) and fj is an assumed fraction of land associated with a given crop, and Kcj the given crop coefficient (e.g. potatoes). According to Allen et al., [1998] Kc for bare soil can be approximated as Kc = Kc ini for months where the crop fields are left fallow.  For both the High and Low Lands, these were calculated as: Eqn. D21    cropsn j jinijfallowc KfK 1 and fj is an assumed fraction of land associated with a given crop (Table D.1). The monthly crop coefficients computed according to the above procedure are shown in Table D.5. The crop coefficient for the village area must consider evapotranspiration from the tree cover throughout the year, and so a constant coefficient, Kc village, can be used for this region. Although palm trees have been observed in the area, the overall distribution of tree types is unknown.  Kc village was therefore set to 0.95, which is the scaling factor used by Harvey et al., [2006] for village evapotranspiration in Munshiganj, Bangladesh, which is similarly covered by tropical trees. D.1.3 Calculation of Potential Evaporation, Ep Calculation of potential evaporation from a free water surface under atmospheric conditions, EP, was calculated using relationships defined in Chapter 3 of the Handbook of Hydrology [Maidment, 1993].   The main equation used was: Eqn. D22         as hnP eeu ARE       2 536.0143.6  Appendix D – Climate Data and Estimation of Hydrologic Parameters 280 where: Ah = energy advected to the water body [mm day -1] and is calculated according to: Eqn. D23  pooiih PTTqTqA  31019.4 with qo = rate of outflow per unit area from the water body [mm day -1] qi = rate of inflow per unit area from the water body [mm day -1] Ti = inflow water temperature [ºC] To = outflow water temperature [ºC] TP = precipitation water temperature = average daily temperature [ºC] and P = daily precipitation rate [mm day-1] Equation D22 was formulated to describe advection energy transferred to a lake, [Maidment, 1993], which experiences significant inflow and outflow.  At the Gotra site, it was assumed that inflow and outflow were negligible compared to precipitation, since open surface water bodies consist of ponds, with no surface inflows or outflows to advect energy. D.2.0 Surface Runoff A modification of empirical USDA Soil Conservation Service (SCS) Runoff Curve Number (CN) method to account for Indian conditions  [SCD, 1972] was used to estimate surface runoff in the study area.  The modified SCS runoff is described by the relationships: Appendix D – Climate Data and Estimation of Hydrologic Parameters 281 Eqn. D24 )7.0( )3.0( 2 SP SP R     and: Eqn. D25 S CN   254 25400  where: R = actual direct runoff, (in cm) [L]; P  = total storm rainfall, (in cm) [L]; S  = potential maximum retention (in cm) [L] CN  = runoff curve number of hydrologic soil cover complex [-]. CN is a function of soil type, land cover and antecedent moisture condition (AMC), and is traditionally established for specific watersheds using field surveys and aerial photo interpretation.  As CN data were not available for Nadia District, runoff curve numbers for the nearby Kaliaghai River Basin, (Midnapore District, West Bengal) were used for this estimate. These data were established by Kumar et al., [1997] using IRS digital satellite data for various land use covers.  Categories that are relevant to the Gotra area as defined in Kumar et al., [1997] include cultivated land (good hydrologic condition), rice paddies and open forested areas.  These correspond to High Lands, Low Lands and Village areas, as defined in Section 4.4.1.  Kumar et al., [1997] also quote curve number values for fallow areas.  Accordingly, these values were applied to the Gotra High Lands and Low Lands for the months of January, and the months of May and November, respectively, as fields are left fallow during these months (Table D.2).  CN values are summarized in Table D.6. The CN numbers for each land use cover in Table D.6 were assigned to one of four hydrologic soil groups defined by the US SCS [2006].  These groups and their respective characteristics are also summarized in Table D.6.  A combination of types C and D were considered to approximate the soil conditions at the site given that 1- these materials are classified as fine Appendix D – Climate Data and Estimation of Hydrologic Parameters 282 grained silts and clays (Appendix A) and 2- during monsoon season (i.e. times for potential runoff) the water table is close to the ground surface.  CNs for High Lands, Low Lands, Village and Fallow areas were computed as averages of the tabulated values, and are also shown in Table D.6.  Appendix D – Climate Data and Estimation of Hydrologic Parameters 283   Figure D.1: Monthly crop coefficients used to calculate various values for evapotranspiration in Gotra.  The length periods in Gotra were scaled to fit the standard durations quoted by Allen et al., 1998, and Kc values were linearly interpolated between Kc ini, Kc mid and Kc end.  High Land Crop Coefficients 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 Feb Mar Apr May Jun Jul Jul Aug Sep Oct Nov Dec K c Jute Aman Rice Potato Oil Crops (Mustard Seed) Cauliflower Cabbage Tomato Low Land Crop Coefficients 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr K c Aman Rice Boro Rice Appendix D – Climate Data and Estimation of Hydrologic Parameters 284  Table D.1: Meteorological data used for estimation of hydrologic parameters.  Average Historical PPT (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Haringhata Weather Station 10.84 17.26 21.65 58.83 97.57 208.85 230.17 262.09 177.94 104.78 17.51 0.94 Ranaghat Weather Station 13.15 25.82 29.32 64.10 130.90 230.74 247.66 264.16 207.38 90.65 13.42 3.65 Hooghly Weather Station 14.64 30.25 32.35 62.79 136.76 235.97 286.18 291.29 236.62 99.90 21.00 3.62 Bongaon Weather Station 12.09 26.18 46.41 76.65 156.33 279.27 311.15 322.93 247.78 114.40 16.92 4.19 Kolkata / Dum Dum Weather Station 12.40 21.60 27.37 56.06 128.53 275.94 330.94 313.73 282.26 125.68 25.53 6.00 Kolkata / Alipore Weather Station 12.54 24.74 32.57 53.24 130.42 292.34 337.05 337.85 267.23 137.89 22.62 6.47 Kolkata 1 16.80 22.90 32.80 47.70 101.70 259.90 331.80 328.80 295.90 151.30 17.20 7.40 Total Nadia PPT (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 2004 0.00 0.00 1.00 38.30 87.40 208.70 146.50 215.60 414.50 141.30 0.00 9.00 2005 20.00 6.00 104.20 104.20 59.90 145.10 274.20 143.80 139.60 351.40 0.00 7.20 2006 0.00 0.00 0.00 6.00 144.10 95.90 353.40 218.60 427.10 51.50 0.50 0.00 2007 0.00 80.80 35.90 19.70 71.80 205.30 471.40 208.70 454.30 107.00 38.00 0.00 2008 70.60 35.00 10.20 92.00 96.40 259.80 301.20 219.50 314.00 81.50 0.00 0.00 Average Historical Temperature (C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Kolkata / Dum Dum Weather Station 19.44 22.36 27.13 30.21 30.73 30.01 29.01 28.90 28.93 27.61 23.83 19.88 Kolkata / Alipore Weather Station 19.75 22.71 27.49 30.22 30.71 29.99 29.02 28.86 28.85 27.65 23.84 20.11 Daily Minimum Temperature 1 13.90 16.90 21.70 25.10 26.40 26.50 26.10 26.10 25.80 24.00 18.90 14.30 Daily Maximum Temperature 1 26.60 29.70 34.00 36.30 36.00 34.10 32.20 32.00 32.20 31.90 29.80 27.00 1India Meteorological Service Website climate normals Appendix D – Climate Data and Estimation of Hydrologic Parameters 285  Table D.2:  Crop types, fraction of land and months of cultivation of agriculture in the research area.   Table D.3:  Crop coefficients obtained from Allen et al., 1998.  Subscripts ini, mid and end represent initial, middle and final development stages of the crops in question.  Kc end for potato and sugar cane were reported as 0.7-0.9 and 0.6-0.9 respectively in Allen et al., 1998, whereas Kc end values presented in this chart are averages  High Lands Low Lands Crop, j Jute Aman Rice Potato Mustard Seed Cauli- flower Cabbage Tomato Crop, j Aman Rice Boro Rice Contribution fraction in Fallow Periods, f 0.33 0.33 0.07 0.07 0.07 0.07 0.07 Contribution fraction in Fallow Periods, f 0.5 0.5 Months Cultivated Months Cultivated Jan Fallow Fallow Fallow Fallow Fallow Fallow Fallow Jan x Feb x Feb x Mar x Mar x Apr x Apr x May x May Fallow Fallow Jun x Jun x Jul x Jul x Aug x Aug x Sep x Sep x Oct x x x x x Oct x Nov x x x x x Nov Fallow Fallow Dec x x x x x Dec x Crop Name, j Kc ini Kc mid Kc end Rice 1.05 1.2 0.75 Potato 0.5 1.15 0.75 Oil Crops (Mustard Seed) 0.35 1.15 0.35 Cauliflower 0.7 1.05 0.95 Cabbage 0.7 1.05 0.95 Tomato 0.6 1.15 0.8 Appendix D – Climate Data and Estimation of Hydrologic Parameters 286  Table D.4: Scaled growing periods for the various crops grown in the fields that surround Gotra.   Table D.5: Calculated monthly crop coefficients for the various vegetables grown in the fields surrounding Gotra.  Crop Cultivation Period (days) Fibre Crops (Jute) Aman Rice Potato Oil Crops (Mustard Seed) Cauliflower Cabbage Tomato Aman Rice Boro Rice Init. (Lini) 15 22 20 16 23 22 19 28 27 Dev. (Ldev) 24 22 20 25 23 33 24 28 27 Mid (Lmid) 28 52 34 33 23 28 32 65 64 Late (Lend) 22 26 17 20 23 8 17 32 32 Total Days of Cultivation in Gotra 89 122 92 92 92 92 92 153 151 High Lands Low Lands Kc Kc High Lands Low Lands Jan 0.66 1.15 Feb 0.43 1.2 Mar 0.9 1.2 Apr 0.76 0.95 May 0.47 1.05 Jun 1.06 1.05 Jul 1.18 1.14 Aug 1.2 1.2 Sep 1 1.2 Oct 0.62 0.96 Nov 1.05 1.05 Dec 1.01 1.05 Appendix D – Climate Data and Estimation of Hydrologic Parameters 287  Table D.6: Determination of CN and S for Gotra recharge zones.  As CN data were not available for Nadia District, runoff curve numbers for the nearby Kaliaghai River Basin, (Midnapore District, West Bengal) were used for this estimate.  These data were established by Kumar et al., 1991 using IRS digital satellite data for various land use covers.  Gotra CN Land Use Cover Condition Gotra Domain A B C D Average of Types C and D Cultivated Land Poor Hydrologic Condition N/A 68 76 82 84 N/A N/A Good Hydrologic Condition High Lands 62 72 78 82 80 572 Paddy Low Lands 95 95 95 95 95 521 Forest Dense N/A 26 40 58 61 N/A N/A Open Village 28 44 60 64 62 664 Fallow 68 79 86 89 88 544 C = Low minimum infiltration capacity when thoroughly wetted.  Moderately fine- to fine-grained soils, or soils with an impeding layer (fragipan) D = High overland flow potential; very low minimum infiltration capacity when thoroughly wetted.  Clay soils with high swelling potential, soils with permanent high water table, soils with a clay layer nesar the surface, shallow soils over impervious rock Gotra S Fallow (Includes Bare Ground) SCS 1964 CN and Soil Type1 1SCS 1964 Soil Groups: A = Low overland flow potential; high minimum infiltration capacity even when thoroughly wetted B = Moderate minimum infiltration capacity when thoroughly wetted. Moderately deep to deep, moderately to well drained, moderately fine- to moderately coarse- grained (e.g. sandy loam) Appendix E - Geochemistry 288 APPENDIX E – GEOCHEMISTRY Appendix E - Geochemistry 289  Table E.1: Field and analytical geochemistry of groundwater in Gotra. Well ID Well Type Sampling Date UTM Easting (m) UTM Northing (m) Depth (m bgl) Sampling Elevation (masl) As (ug/l) pH EC (uS/cm) Alkalinity (eq / L) Temp (C) Dissolved Oxygen (mg/l) NH4 (mg/l) Na (mg/l) K (mg/l) Ca (mg/l) Fe (mg/l) Mg (mg/l) Mn (ug/l) Cl (mg/l) SO4 (mg/l) S (mg/l) HCO3 (mg/l) NO3 (mg/l) P (mg/l) d18O (ppth) 2H (ppth) TOC (mg/l) Fe(II) Field (mg/l) PhreeqC Charge Balance Error (%) GSI0601 Observation Well 26-Feb-07 661680.0 2547721.0 21.9 -14.2 208.1 6.75 1299.0 13.5 27.1 . 8.30 62.64 3.29 127.82 5.412 38.075 13.5 6.62 . 0.20 790.66 0.13 . . . 5 1.29 . GSI0602 Observation Well 26-Feb-07 661599.5 2547791.0 45.7 -38.5 4.0 6.94 1400.0 8.0 27.1 . 0.01 33.07 1.45 92.16 0.210 24.104 8.0 4.42 0.03 . 471.94 . 0.03 . . . 0.13 -2.61 GSI0603 Observation Well 21-May-06 662724.8 2546738.0 22.6 -15.0 13.1 7.09 720.0 6.1 28.6 0.15 0.41 13.99 5.05 98.22 5.523 19.480 6.1 21.15 6.97 2.56 354.88 1.57 6.97 -2.10 -20.70 2 0.57 -0.08 GSI0603 Observation Well 20-Feb-07 662724.8 2546738.0 22.6 -15.0 14.1 7.00 688.0 6.4 26.5 . 0.34 12.53 4.08 98.43 6.123 17.443 6.4 18.75 23.46 8.30 379.22 . 23.46 . . . 0.74 4.18 GSI0603 Observation Well 14-Feb-08 662724.8 2546738.0 22.6 -15.0 13.2 7.11 639.0 6.1 26.5 . 0.20 13.65 4.16 92.87 4.889 16.402 6.1 22.40 23.83 7.84 361.97 11.57 23.83 -2.77 -19.19 . 0.83 -0.06 GSI0603 Observation Well 14-Feb-08 662724.8 2546738.0 22.6 -15.0 13.3 7.11 639.0 6.1 26.5 . 0.20 13.67 4.14 93.99 4.924 16.570 6.1 . . 7.74 361.52 . . . . . 0.85 -6.97 GSI0604 Observation Well 21-May-06 662724.8 2546738.0 12.2 -4.7 35.0 7.14 1094.0 8.6 27.0 0.25 0.70 34.93 30.62 122.90 6.141 28.597 8.6 60.87 0.85 0.76 501.67 . 0.85 -0.50 -11.80 4 0.83 4.43 GSI0604 Observation Well 19-Feb-07 662724.8 2546738.0 12.2 -4.7 39.9 6.84 1445.0 11.6 27.1 . 2.80 58.34 41.78 168.46 8.572 37.624 11.6 125.79 38.81 13.62 674.12 . 38.81 . . 4 0.86 3.06 GSI0604 Observation Well 14-Feb-08 662724.8 2546738.0 12.2 -4.7 30.5 6.93 1957.0 14.4 26.7 . 0.26 103.94 122.07 168.01 7.507 55.875 14.4 212.83 59.28 20.19 837.03 9.38 59.28 -3.20 -21.61 . 1.13 0.25 GSI0605 Observation Well . 662563.4 2546657.0 22.5 -14.3 488.4 6.79 1525.0 16.2 27.1 . 14.40 50.72 5.30 176.44 14.495 47.415 16.2 25.39 . 0.18 939.60 0.05 . . . 9 0.77 -1.71 GSI0605 Observation Well 14-Feb-08 662563.4 2546657.0 22.5 -14.3 467.9 6.68 . 12.3 26.5 . 7.81 51.10 5.38 170.61 15.258 48.074 12.3 31.49 0.35 0.19 713.23 11.83 0.35 -5.05 -34.20 . 0.76 -1.95 GSI0605 Observation Well 23-Feb-07 662563.4 2546657.0 22.5 -14.3 495.7 6.79 1525.0 16.2 27.1 . 14.40 50.54 5.29 177.38 14.387 47.103 16.2 24.39 . 0.21 939.54 0.39 . . . 9 0.70 5.56 GSI0606 Observation Well 21-May-06 662775.5 2546631.0 24.4 -16.7 2.9 7.01 867.0 7.1 26.5 0.3 0.38 17.69 2.50 117.23 7.627 25.057 7.1 31.31 40.51 13.67 414.81 . 40.51 -2.70 -21.50 . 0.65 -1.94 GSI0606 Observation Well 22-Feb-07 662775.5 2546631.0 24.4 -16.7 2.5 7.05 811.0 7.1 26.4 . 0.50 17.80 2.01 105.54 7.813 24.132 7.1 27.58 34.40 12.10 414.26 0.1 34.4 . . 1 0.64 4.01 GSI0606 Observation Well 14-Feb-08 662775.5 2546631.0 24.4 -16.7 3.0 7.06 698.0 6.5 26.6 . 0.60 15.82 1.89 92.26 6.618 19.401 6.5 20.10 25.34 8.29 382.17 0.21 25.34 -2.67 -20.31 . 2.09 0.91 GSI0607 Observation Well 21-May-06 662655.7 2546654.0 24.4 -16.4 334.6 7.05 1390.0 12.4 27.0 0.2 6.80 41.23 15.06 157.22 13.758 42.194 12.4 51.63 3.58 1.55 718.60 0.05 3.58 -4.06 -25.40 6 0.94 -1.15 GSI0607 Observation Well 22-Feb-07 662655.7 2546654.0 24.4 -16.4 268.7 6.87 1701.0 14.3 26.9 . 7.60 47.69 59.82 170.73 10.717 48.292 14.3 108.30 13.60 4.79 827.88 1.34 13.6 . . 5 0.85 1.09 GSI0607 Observation Well 14-Feb-08 662655.7 2546654.0 24.4 -16.4 247.4 6.87 1639.0 14.8 26.8 . 1.15 49.27 67.87 177.39 11.077 51.458 14.8 122.74 12.56 4.28 859.55 0.07 12.56 -4.20 -29.09 . 1.07 -2.08 GSI0608 Observation Well 22-May-06 662629.1 2546737.0 25.9 -18.1 33.0 7.13 1316.0 8.8 27.8 . 1.04 47.58 71.68 124.34 1.020 33.097 8.8 119.17 31.68 10.55 513.26 . 31.68 -1.40 -19.00 4 0.82 -3.35 GSI0608 Observation Well 23-Feb-07 662629.1 2546737.0 25.9 -18.1 47.1 7.07 1338.0 9.7 25.6 . 1.90 46.16 66.60 131.77 7.517 31.346 9.7 125.36 20.14 6.90 568.79 1.15 20.14 . . 2 0.72 2.60 GSI0608 Observation Well 14-Feb-08 662629.1 2546737.0 25.9 -18.1 28.6 7.01 1350.0 10.2 26.7 . 0.24 49.88 85.04 133.80 8.313 32.527 10.2 130.22 20.44 6.61 594.72 4.34 20.44 -1.20 -12.09 . 0.72 -1.35 GSI0608-40 Shallow Piezometer 14-Feb-08 662629.1 2546737.0 12.2 -4.4 42.0 6.90 2980.0 19.2 25.7 . 0.63 102.10 203.00 271.90 8.280 79.870 19.2 393.75 112.34 36.80 1102.40 0.25 112.3 -3.60 -21.22 . 0.50 -1.54 GSI0608-40 Shallow Piezometer 20-Feb-07 662629.1 2546737.0 12.2 -4.4 39.9 6.82 3080.0 19.2 25.7 . 3.00 100.19 198.65 299.82 7.708 91.177 19.2 416.23 113.66 40.66 1098.68 0.17 113.7 . . 7 0.43 -1.88 GSI0608-55 Shallow Piezometer 14-Feb-08 662629.1 2546737.0 16.8 -9.0 29.9 7.14 1122.0 9.4 25.4 . 0.24 40.77 85.93 105.20 6.280 24.300 9.4 90.71 0.11 0.08 554.17 0.58 0.11 -1.98 -14.27 . 0.99 0.79 GSI0608-55 Shallow Piezometer 22-May-06 662629.1 2546737.0 16.8 -9.0 23.9 7.39 1132.0 6.1 27.5 . 0.70 48.89 60.04 94.79 0.534 31.512 6.1 115.34 67.00 22.15 355.52 . 67 2.00 3.00 9 1.81 -3.03 GSI0608-55 Shallow Piezometer 20-Feb-07 662629.1 2546737.0 16.8 -9.0 28.6 7.12 1246.0 9.2 25.6 . 1.20 41.30 83.98 109.05 5.654 28.458 9.2 98.99 2.89 0.89 541.69 . 2.89 . . 2 0.83 6.36 GSI0609 Observation Well 22-May-06 662927.1 2546832.0 20.7 -14.4 9.2 7.24 511.0 4.9 26.8 . 0.34 10.01 2.03 72.66 2.593 14.285 4.9 6.12 0.24 0.21 290.41 . 0.24 -2.70 -24.30 . 0.39 -0.57 GSI0609 Observation Well 20-Feb-07 662927.1 2546832.0 20.7 -14.4 10.9 7.21 498.0 5.2 26.4 . 0.40 9.21 1.82 70.09 2.230 13.594 5.2 5.63 . 0.10 309.63 0.56 . . . . 0.41 2.75 GSI0609 Observation Well 14-Feb-08 662927.1 2546832.0 20.7 -14.4 10.2 7.24 475.0 5.2 26.6 . 0.21 9.85 1.78 67.43 2.367 12.594 5.2 5.54 . . 306.22 . . -2.85 -20.79 . 0.46 -2.82 GSI0609-30 Observation Well 14-Feb-08 662927.1 2546832.0 9.1 -2.8 6.4 7.16 548.0 5.8 25.7 . 0.20 11.94 1.57 78.44 1.956 14.979 5.8 7.40 7.50 2.42 344.35 0.09 7.5 -2.53 -21.24 . 0.45 -3.88 GSI0609-30 Observation Well 22-Feb-07 662927.1 2546832.0 9.1 -2.8 6.8 7.27 599.0 5.7 26.0 . 0.39 11.17 1.49 81.01 2.088 16.155 5.7 6.98 7.60 2.93 336.50 3.18 7.6 . . . 0.44 -2.67 GSI0609-40 Observation Well 14-Feb-08 662927.1 2546832.0 12.2 -5.8 27.8 7.05 478.0 4.6 26.0 . 0.44 7.97 1.31 69.58 4.557 12.033 4.6 5.41 0.03 . 269.78 . 0.03 -2.71 -18.24 . 0.98 -1.78 GSI0609-40 Observation Well 22-Feb-07 662927.1 2546832.0 12.2 -5.8 29.1 7.16 508.0 5.4 26.1 . 0.00 7.70 1.50 74.83 5.123 13.772 5.4 5.76 0.02 0.08 316.03 0.34 0.02 . . 1 0.92 2.28 GSI0609-60 Observation Well 14-Feb-08 662927.1 2546832.0 18.3 -11.9 16.1 7.16 451.0 5.3 26.0 . 0.27 9.09 1.75 71.47 3.590 13.559 5.3 4.89 0.02 . 313.95 . 0.02 -2.83 -14.76 . 0.45 -2.35 GSI0609-60 Observation Well 20-Feb-07 662927.1 2546832.0 18.3 -11.9 15.4 7.17 533.0 5.6 25.9 . 0.53 8.42 1.82 74.45 3.929 14.879 5.6 5.42 0.02 0.08 330.24 . 0.02 . . . 0.43 -2.29 GSI0610 Observation Well 22-May-06 663246.1 2546871.0 20.7 -14.0 7.7 7.34 587.0 5.6 26.8 . 0.27 16.33 3.93 82.68 1.297 16.873 5.6 8.76 5.49 2.05 330.30 . 5.49 -3.10 -20.50 . 0.57 -2.56 GSI0610-20 Observation Well 14-Feb-08 663246.1 2546871.0 6.1 0.6 2.4 7.11 560.0 5.9 26.1 . 0.14 14.72 3.25 79.66 3.082 14.841 5.9 8.28 6.51 2.07 350.87 3.33 6.51 -3.42 -22.89 . 0.24 4.03 GSI0610-20 Observation Well 14-Feb-08 663246.1 2546871.0 6.1 0.6 2.6 7.11 560.0 5.9 26.1 . 0.14 14.59 3.22 79.54 3.022 14.876 5.9 . . 2.08 351.64 . . . . . 0.21 -3.67 GSI0610-20 Observation Well 22-Feb-07 663246.1 2546871.0 6.1 0.6 3.1 7.22 814.0 6.1 25.4 . 0.06 14.45 3.42 83.29 3.035 15.944 6.1 8.78 4.67 1.88 359.59 0.28 4.67 . . . 0.24 0.15 GSI0610-40 Observation Well 14-Feb-08 663246.1 2546871.0 12.2 -5.5 1.7 7.10 644.0 6.8 25.6 . 0.11 16.17 3.27 94.78 3.118 16.984 6.8 12.19 9.94 3.41 399.24 0.31 9.94 -4.04 -28.04 . 0.21 -1.19 GSI0610-40 Observation Well 21-Feb-07 663246.1 2546871.0 12.2 -5.5 3.5 7.15 636.0 6.5 25.2 . 0.05 14.04 3.12 86.66 3.024 17.303 6.5 8.45 5.21 1.97 383.35 5.93 5.21 . . . 0.25 -1.51 GSI0610-60 Observation Well 14-Feb-08 663246.1 2546871.0 18.3 -11.6 2.3 7.13 568.0 6.1 25.9 . 0.13 12.91 3.15 81.19 3.359 15.546 6.1 11.63 7.13 2.53 362.58 2.04 7.13 -2.99 -22.25 . 0.23 -5.07 GSI0610-60 Observation Well 21-Feb-07 663246.1 2546871.0 18.3 -11.6 2.9 7.17 611.0 6.1 25.7 . 0.04 15.73 3.09 81.34 3.602 15.597 6.1 8.48 5.64 2.14 362.12 0.13 5.64 . . . 0.34 -4.94 GSI0610-IRRW Irrigation Well 14-Feb-08 663246.1 2546871.0 54.9 -48.1 34.7 7.08 560.0 6.1 26.9 . 0.45 13.69 2.40 78.81 2.832 15.810 6.1 8.36 4.27 1.58 363.67 . 4.27 -3.40 -27.93 . 0.89 -2.00 GSI0611 Observation Well 26-Feb-07 660596.3 2548001.0 . . 135.2 6.80 302.0 11.6 27.3 . 5.50 56.97 3.35 112.06 5.284 30.867 11.6 4.22 . 0.15 685.41 0.22 . . . 4 1.26 -3.64 GSI0605-20 Shallow Piezometer 14-Feb-08 662563.4 2546657.0 6.1 2.1 406.3 6.86 1295.0 12.7 . . 12.10 29.88 12.54 162.99 11.016 47.036 12.7 106.13 0.09 0.12 739.47 0.03 0.09 -2.41 -19.13 . 1.50 -2.72 GSI0605-37 Shallow Piezometer 14-Feb-08 662563.4 2546657.0 11.3 -3.1 496.3 6.85 1512.0 16.4 26.0 . 16.39 51.06 8.16 188.99 17.079 50.386 16.4 69.11 0.09 0.19 943.81 1.8 0.09 -4.95 -31.34 . 1.34 -3.43 GSI0605-56 Shallow Piezometer 14-Feb-08 662563.4 2546657.0 17.1 -8.8 417.5 6.81 1396.0 13.2 26.2 . 11.92 33.82 8.29 166.46 15.289 48.113 13.2 91.58 0.02 0.12 762.71 0.41 0.02 -3.43 -24.86 . 1.20 -2.62 GSI0606-35 Shallow Piezometer 14-Feb-08 662775.5 2546631.0 10.7 -3.0 24.0 7.04 2090.0 11.9 26.1 . 1.03 59.10 5.80 276.60 4.660 58.200 11.9 306.70 87.40 28.50 677.47 0.09 87.4 -4.01 -28.29 . 0.90 -2.40 GSI0606-35 Shallow Piezometer 14-Feb-08 662775.5 2546631.0 10.7 -3.0 23.0 7.04 2090.0 11.9 26.1 . 1.03 57.70 5.90 276.30 4.750 57.250 11.9 301.81 85.34 27.60 677.60 0.37 85.34 -4.01 -28.29 . 0.80 0.91 GSI0712 Observation Well 26-Feb-07 661874.7 2547939.0 . . 5.2 7.10 1171.0 6.5 26.7 . 0.40 19.51 2.29 73.80 0.438 21.587 6.5 10.97 0.03 . 386.42 . 0.03 . . 1 0.38 0.88 GSI0713 Observation Well 14-Feb-08 662511.3 2546579.0 23.2 -14.9 133.0 6.68 1696.0 16.6 26.9 . 15.79 61.79 7.07 184.61 18.633 50.280 16.6 108.53 0.05 0.21 960.04 0.49 0.05 -4.00 -26.03 . 0.81 -3.13 GSI0713-44 Shallow Piezometer 14-Feb-08 662511.3 2546579.0 13.4 -5.2 117.1 7.01 1772.0 18.2 28.9 . 11.89 65.97 6.67 211.33 20.829 53.029 18.2 113.45 0.04 0.35 1040.16 0.12 0.04 -3.48 -23.23 . 0.49 -5.37 GSI0714 Observation Well 2-Mar-07 662605.6 2546698.0 25.3 -17.3 424.1 6.88 1002.0 8.8 26.8 . . 30.48 6.75 105.55 7.508 28.806 8.8 40.67 0.14 0.19 520.02 . 0.14 . . 4 1.00 -5.70 GSI0714 Observation Well 14-Feb-08 662605.6 2546698.0 25.3 -17.3 436.0 6.82 967.0 9.6 27.1 . . 33.24 5.36 105.78 9.893 29.820 9.6 44.04 0.05 0.14 560.91 0.21 0.05 -3.65 -26.44 . 1.05 -3.98 Appendix E - Geochemistry 290  Table E.1 (continued): Field and analytical geochemistry of groundwater in Gotra. Well ID Well Type Sampling Date UTM Easting (m) UTM Northing (m) Depth (m bgl) Sampling Elevation (masl) As (ug/l) pH EC (uS/cm) Alkalinity (eq / L) Temp (C) Dissolved Oxygen (mg/l) NH4 (mg/l) Na (mg/l) K (mg/l) Ca (mg/l) Fe (mg/l) Mg (mg/l) Mn (ug/l) Cl (mg/l) SO4 (mg/l) S (mg/l) HCO3 (mg/l) NO3 (mg/l) P (mg/l) d18O (ppth) 2H (ppth) TOC (mg/l) Fe(II) Field (mg/l) PhreeqC Charge Balance Error (%) GSI0715 Observation Well 14-Feb-08 662706.4 2546825.0 29.9 -22.5 2.7 7.42 551.0 4.9 27.1 . 0.15 12.03 1.84 78.57 1.117 14.587 4.9 30.85 8.43 2.90 290.45 0.65 8.43 -2.42 -21.70 . 0.73 -6.71 GSI0715 Shallow Piezometer . 662706.4 2546825.0 29.9 -22.5 . 7.42 551.0 4.9 . . . 12.03 1.84 78.57 1.117 14.587 4.9 30.85 8.43 2.90 292.00 0.65 8.43 . . . . -1.50 GSI0715-55 Shallow Piezometer 14-Feb-08 662703.3 2546830.0 16.8 -9.4 3.8 7.10 809.0 6.5 26.1 . 0.09 30.98 4.40 99.44 1.328 20.882 6.5 57.52 35.62 12.01 381.88 0.92 35.62 -3.06 -23.16 . 0.17 -1.77 GSI0715-65 shallow Piezometer 14-Feb-08 662703.3 2546830.0 19.8 -12.4 2.9 7.13 791.0 6.8 26.0 . 0.16 28.16 2.93 101.07 1.499 21.112 6.8 54.35 22.94 7.89 398.91 0.08 22.94 -3.49 -19.70 . 0.25 -1.31 GSI0716 Observation Well 14-Feb-08 662222.2 2546820.0 . . 1.8 7.08 578.0 6.5 27.0 . . 31.62 1.23 70.68 0.194 17.843 6.5 10.79 0.08 0.06 389.71 0.13 0.08 -3.24 -25.11 . 0.18 -2.00 GSI0803-22 Shallow Piezometer 14-Feb-08 662724.8 2546738.0 6.7 0.8 21.6 . . . . . . 16.36 34.91 42.34 3.850 13.172 . . . 17.03 0.00 . . . . . 0.29 -3.55 GSI0803-15 Shallow Piezometer 14-Feb-08 662724.8 2546738.0 4.6 2.9 14.3 . . . . . . 16.91 34.49 41.02 1.515 12.403 . . . 23.15 0.00 . . . . . 0.17 . GSI0803-27 Shallow Piezometer 14-Feb-08 662724.8 2546738.0 8.2 -0.7 26.5 . . . . . . 18.58 33.97 54.68 5.777 15.133 . . . 15.55 0.00 . . . . . 0.59 . GSI0805-10 Shallow Piezometer 14-Feb-08 662563.4 2546657.0 3.0 5.2 29.1 7.80 531.0 . . . . 24.29 50.56 36.07 0.049 12.319 . 32.63 4.06 1.92 0.00 . 4.06 -0.59 -16.95 . 1.89 . GSI0817 Observation Well 14-Feb-08 661966.0 2546817.0 . . 2.3 7.13 528.0 5.9 26.8 . . 23.61 1.83 64.84 0.081 17.005 5.9 7.94 0.20 0.11 353.81 0.26 0.2 -2.43 -15.77 . 0.21 . GSI Blank N/A 14-Feb-08 . . . . 0.4 . . . . . . 0.07 . 0.15 0.013 0.039 . . . . . . . . . . . -3.97 IRRW0601 Irrigation Well . 662752.3 2546756.7 200.0 -192.6 47.8 . . . . . . 13.40 2.27 81.70 4.749 16.771 . 8.33 4.06 1.41 0.00 4.02 4.06 -3.32 -24.10 . 0.79 . North Pond N/A . 662719.0 2546755.0 0.0 7.5 . . . . . . . . . . . . . . . . . . . 1.00 0.40 . . . South Pond N/A 14-Feb-08 662563.4 2546657.0 0.0 8.2 18.8 8.90 517.0 . . . . 22.85 26.28 44.25 0.078 17.111 . 34.08 6.64 2.35 0.00 8.29 6.64 0.21 -9.70 . 0.48 . TW01 Domestic Tube Well 12-May-06 662713.6 2546671.0 115.8 -108.1 55.4 7.13 713.0 7.7 27.1 0.7 0.12 15.21 3.73 93.15 0.948 27.017 7.7 2.64 . . 450.32 0.08 . -3.80 -32.30 2 0.88 . TW01 Domestic Tube Well 17-Feb-07 662713.6 2546671.0 115.8 -108.1 61.6 7.28 1201.0 7.8 26.7 . 0.60 15.12 3.48 88.58 0.941 27.125 7.8 2.52 . . 458.57 0.09 . . . 1 0.89 -0.04 TW02 Domestic Tube Well 12-May-06 662708.1 2546702.0 21.9 -14.3 9.3 7.08 961.0 7.5 26.6 0.5 0.08 47.86 10.14 110.60 6.017 25.554 7.5 44.65 26.68 9.26 439.65 0.04 26.68 -2.30 -28.00 2 0.66 -2.37 TW02 Domestic Tube Well 17-Feb-07 662708.1 2546702.0 21.9 -14.3 8.4 7.18 1217.0 6.9 26.0 . 0.40 20.94 15.35 90.08 4.994 22.745 6.9 25.16 13.09 4.75 408.08 0.02 13.09 . . 1 0.83 6.10 TW03 Domestic Tube Well 12-May-06 662756.8 2546629.0 21.3 -13.6 3.4 7.06 788.0 6.5 26.6 0.15 1.00 16.66 2.36 104.97 7.608 25.358 6.5 25.75 32.26 10.97 382.35 . 32.26 -2.33 -24.50 2 0.69 0.21 TW03 Domestic Tube Well 17-Feb-07 662756.8 2546629.0 21.3 -13.6 3.6 7.10 1208.0 6.8 25.3 . 0.50 16.67 2.34 97.98 6.623 24.137 6.8 24.84 28.17 9.84 396.72 . 28.17 . . . 0.72 5.45 TW04 Domestic Tube Well 12-May-06 662694.8 2546669.0 61.0 -53.1 58.3 . 536.0 5.2 27.3 1 1.00 11.28 2.80 76.56 2.689 15.004 5.2 5.90 1.63 0.73 309.89 0.2 1.63 -3.10 -25.90 2 0.93 1.01 TW04 Domestic Tube Well 17-Feb-07 662694.8 2546669.0 61.0 -53.1 61.4 7.15 894.0 5.5 27.0 . 0.55 10.64 2.33 75.12 2.740 15.060 5.5 5.70 1.53 0.68 324.79 . 1.53 . . . 0.93 2.90 TW04 Domestic Tube Well 14-Feb-08 662694.8 2546669.0 61.0 -53.1 46.7 7.17 505.0 5.3 26.2 . 1.00 11.19 2.27 73.11 2.657 13.622 5.3 7.41 2.67 0.89 315.47 2.68 2.67 -3.38 -13.68 . 0.95 -0.58 TW05 Domestic Tube Well 12-May-06 662760.7 2546581.0 21.9 -14.1 14.7 7.10 645.0 5.1 26.2 0.7 1.41 11.28 2.51 85.49 4.922 21.666 5.1 18.99 23.84 8.22 302.80 . 23.84 -2.70 -26.00 2 1.13 -2.88 TW05 Domestic Tube Well 17-Feb-07 662760.7 2546581.0 21.9 -14.1 15.9 7.17 1070.0 4.5 25.3 . 1.10 11.15 2.32 77.94 4.706 21.695 4.5 17.67 22.47 7.84 263.72 0.06 22.47 . . . 1.14 7.51 TW06 Domestic Tube Well 12-May-06 662690.6 2546582.0 21.9 -13.7 222.6 7.13 837.0 6.8 26.5 0.15 7.80 21.37 7.85 94.78 10.657 22.774 6.8 38.34 . . 394.56 0.02 . -3.50 -28.70 5 0.88 10.84 TW06 Domestic Tube Well 17-Feb-07 662690.6 2546582.0 21.9 -13.7 222.6 7.02 1207.0 7.1 25.8 . 5.90 19.20 7.65 83.71 8.482 21.769 7.1 27.91 . 0.12 416.86 . . . . 2 0.92 3.82 TW07 Domestic Tube Well 12-May-06 662659.0 2546661.0 21.3 -13.4 422.2 7.14 1408.0 13.7 26.6 0 8.90 42.78 19.29 162.47 16.740 42.197 13.7 37.19 . . 788.52 . . -4.40 -29.80 8 0.74 -2.21 TW07 Domestic Tube Well 18-Feb-07 662659.0 2546661.0 21.3 -13.4 436.9 6.97 1327.0 14.3 26.6 . 8.60 43.58 18.15 159.94 16.011 42.683 14.3 25.59 . 0.17 825.44 0.78 . . . 6 0.71 0.45 TW08 Domestic Tube Well 12-May-06 662624.5 2546694.0 67.1 -59.1 38.8 7.08 765.0 7.6 27.3 0.15 0.36 27.06 3.30 93.45 0.875 27.904 7.6 4.48 . . 450.63 0.14 . -3.86 -26.30 1 1.02 -1.04 TW08 Domestic Tube Well 18-Feb-07 662624.5 2546694.0 67.1 -59.1 45.3 7.24 765.0 8.3 27.1 . 0.46 25.93 3.33 91.82 1.104 28.556 8.3 4.52 . 0.07 489.67 0.48 . . . . 0.99 3.28 TW09 Domestic Tube Well 12-May-06 662571.6 2546683.0 27.4 -19.3 514.3 7.14 1180.0 11.6 26.4 0.15 6.10 43.10 4.75 127.65 12.059 37.054 11.6 26.50 . 0.06 673.87 0.03 . -4.20 -26.80 6 1.00 -1.59 TW09 Domestic Tube Well 18-Feb-07 662571.6 2546683.0 27.4 -19.3 492.1 6.96 1073.0 12.2 25.4 . 9.30 42.32 4.54 129.66 11.320 37.171 12.2 26.57 . 0.19 711.46 0.2 . . . 5 1.01 -1.38 TW10 Domestic Tube Well 12-May-06 662566.4 2546697.0 47.2 -39.1 8.5 7.07 595.0 5.0 26.4 0.15 0.44 11