SUMMARY: Groundwater withdrawals have increased significantly in developing parts of the world to support growing populations and increased domestic food production. But in some of these areas, the groundwater is dangerously contaminated with heavy metals like arsenic, raising concerns over the safety of the food and water being consumed by the population. Investigate a site in the Munshiganj district in Bangladesh and analyze impact of drinking water wells and irrigation wells with respect to the groundwater-surface water interactions that drive the arsenic groundwater contamination problem.

Figure 1: Site location in the Munshiganj district, Bangladesh.

Background & Motivation

Like many developing parts of the world, Bangladesh has experienced significant population growth in recent decades. To support the growing population, domestic food production has expanded, especially through the cultivation of Boro, a dry season rice which provides greater yields than the traditional rice grown during the wet season.

A critical dependence on groundwater has emerged because of the need for dry season irrigation and  a growing drinking water demand, but there are concerns over recent findings: much of the region’s groundwater is dangerously contaminated with arsenic. The dissolved arsenic in groundwater appears to be related to the surface. The original source is likely from oxidation of sulfide minerals in near-surface sediments from source regions of the Himalayas.  The right kind of biogeochemical and hydrological conditions can partition arsenic from the solid to aqueous phase and transport arsenic into contaminated aquifers.

Let’s consider an example in the developing area near Sreenagar, Munshiganj district, Bangladesh (see Figure 1). In this area, the sediments in the surface water ponds and Ichimatti River are known to be arsenic “hotspots”. Originally excavated to provide clay/silt material for construction of villages above the monsoon flood levels, the ponds now receive human waste from the villages and are also used for aquiculture. These conditions are suitable for partitioning arsenic from the solid to aqueous phase. The ponds sit in a thin clay layer on top of a relatively homogenous sand aquifer. If a downward hydraulic gradient exists (i.e., the pond water level is higher than the head in the underlying aquifer), dissolved arsenic will be added to the aquifer and move with the groundwater flow.

This is concerning to local officials, given the presence of five drinking wells and two irrigation wells in the vicinity of three of these ponds (see Figure 2).  If pond water is getting into drinking water wells, there is a clear human health risk! And if pond water is being captured by irrigation wells and applied to rice fields, there is food safety concern! Not to mention the possible further contamination of the aquifer because of return-flow (irrigation water that infiltrates down past the root zone)!

Objectives and Deliverables

Perform a 3D modeling analysis groundwater flow at the site near Sreenagar and assess the vulnerability of the drinking water wells and irrigations wells to arsenic contamination from the nearby ponds and/or river. You will simulate groundwater local conditions for the dry-season irrigation activities under different water demands representing different growth periods. You will also evaluate the impact of pumping by investigating flow patterns and fluxes (e.g., between the pond and aquifer) under these different scenarios.

Prepare a 1-2-page report that summarizes your approach and findings. You should discuss your findings with regards to responsibility for the contamination. Include any detailed model results / graphics in support of your conclusions in an appendix. 

Details about the model setup and application are provided in the following sections.

Figure 2: Salient features of the site. Left: plan view with approximate extents of Boro rice fields, surface water ponds, and the Ichmatti River; and approximate locations of the drinking water wells and irrigation wells. Right: Conceptual cross-section.

Model Setup:

Model Domain and Model Layers

Search for and zoom to the following GPS coordinates: 23.534113,90.288401. This is the approximate location of the site. Draw a model domain that includes all of the water wells, rice fields ponds, and the Ichmatti River (see Figure 2). The lateral and bottom boundaries of the domain are assumed to be no-flow boundaries. Lateral boundaries are placed such that assumption of a hypothetical hydrologic barrier does not significantly impact the flow results in the area of significant head variability (i.e., near the wells). The bottom boundary represents the interface of the sand aquifer and a 2^nd layer of low permeability clay. The top boundary should follow the land surface as represented in the Digital Elevation Model (DEM) available on the MAGNET server.

By default, the model consists single layer.  This first layer will represent the thin (5m) clay layer occurring at the surface. Add a 2^nd layer beneath the clay layer to represent the sand aquifer (100m thick). Based on a field study, the following is known about hydraulic properties of each layer:

• Clay layer hydraulic conductivity: 5.2x10^-8 m/s
• Sand aquifer hydraulic conductivity: 9.3x10-5m/s (26 ft/d)
• Effective porosity: 0.2
• Storage compressibility: 0.01

Sources and Sinks of Water

First, add the surface water features as zones in the model domain. The ponds should be represented as (two-way) head-dependent boundary conditions that allow water to enter or leave the aquifer from/to the pond depending on the direction of the head gradient.  Field measurements show that the average pond depths are 4 m. The stage of the ponds can be derived from the aquifer top elevation (the DEM). A leakance (hydraulic conductivity per unit thickness) of 1 d^-1 may be used. Represent the Ichimatti River as a prescribed head boundary condition that follows the aquifer top elevation. Note that the Ichimatti River extends into sand aquifer, so it should exist as zone feature in both layers

Next, add the wells to the model. Their approximate locations are shown in Figure 2. The irrigation wells and drinking water wells are screened at an average depth of 37m and 30m, respectively.  Assume five people are using each drinking water well, and the per capita consumption is about 20 L/d. For now, assign a pumping rate of zero to each irrigation well (we will look at their impact later).

During the dry season (mid-January to late-April), the average precipitation is about 191mm. The potential evapotranspiration is about 417mm. Apply a net recharge to the model domain based on this information (note: recharge can be positive or negative as an input in MAGNET). Later when we add irrigation wells, we will add rice field-specific recharge values to represent return flow of applied irrigation water.

Initial Conditions and Simulation Settings

At the end of the wet season, the land is completely inundated. Therefore, you should assign the initial groundwater levels to be equal to the top aquifer elevation (the DEM).

Assign a time-step of 1 day and a simulation length of 100 days. Use enough grid cells to resolve the head changes at/near the pumping wells (e.g., NX=100).

You will want to resolve the vertical details of the movement of water in the sandy aquifer, so you will need to add ‘computational layers’ to the 2^nd layer of your model. Use the options in the Simulation Settings of the Domain Attributes menu sub-divide the 2^nd layer into five layers of equal thickness.

Model Application

Submit the model for simulation and answer the following questions by investigating your results (your responses and supporting evidence should be included in your written report):

1. Explain the flow patterns in the different layers. Provide some visualizations and plots in support of your answer.
2. On day 100, what is the vertical flux of water from the clay layer into the aquifer? How much of this flux comes from the ponds? 
3. Has the system reached steady-state by the end of the dry season?
4. Are any of the drinking wells drawing water from the pond or river? You will want to add a "tracer" (particles) and see where this tracer goes during the simulation.
   • Add a particle zones where the ponds and/or river are located before simulating
   • You may need to run your simulation for much longer than 100 days to see complete flow-paths. Assume that flow and transport do not occur during the monsoon seasons that interrupt the dry seasons.

Early Growth Period Simulation

Over time, the population grew and the number of people dependent on each drinking well increased to 10, but the daily per capita consumption remained the same.  To overcome the food deficit due to increased population over time, the cultivation of Boro started. At this early stage, irrigation wells pumped at 15 L/s for 12 hours a day for 100 days. Convert this to continuous rate to be applied over the 100-day period.

Note that irrigation water is applied only to the rice fields. You will need to add the rice fields as zone features to represent the return flow of applied irrigation water. Their approximate extents are shown in Figure 2. Assume that return flow (local recharge) is 30% of the irrigation well pumping rates.

Run the updated model and answer the following questions:

1. How have the flow patterns and/or fluxes changed? Explain/interpret.
2. Are any of the irrigation or drinking water wells drawing water from the ponds or river? Again, take advantage of particle tracking techniques in MAGNET)
3. Has the system reached steady-state by the end of the dry season?

Later Growth Period Simulation

The population continues to increase; now 15 people depend on each drinking well! There is even a need to add a new drinking well (put the new well in a sensible location). As the demand for rice kept increasing, the old irrigation pumps were replaced with new pumps that can withdraw water at 24 L/s for 12 hours a day for 100 days.

Run the updated model and answer the following questions:

• How have the flow patterns and/or fluxes changed? Explain/interpret.
• Are any of the irrigation or drinking water wells drawing water from the ponds? 
• Has the system reached steady-state by the end of the dry season?

Sensitivity Analysis

Use your model to determine the following:

• How do order of magnitude changes in the clay layer conductivity change the situation?
• How do order of magnitude changes in the storage compressibility change the situation?

Extra Credit: Scenario Testing

Using the knowledge gained from the above simulations, where would you move the irrigation well in order to reduce the amount of 1) field water, 2) pond water, and 3) river water enter the drinking water wells? Why? Try moving the irrigation well to these locations and for each situation show us the resulting concentrations entering the drinking water wells after 30 years worth of dry seasons (assume some initial concentration of arsenic leaving the pond / entering the aquifer).

For each situation, also determine the residence time of water within the aquifer.

Again, assume that flow and transport do not occur during the monsoon seasons that interrupt the dry seasons.