Case Study 3 - Municipal Groundwater Extraction in Southwestern British Columbia
This groundwater extraction project (the "Project") is situated in the Fraser lowlands of southwestern British Columbia, serving a population of nearly 150,000 residents (Figure 1). The region is host to both urban and agricultural activities. Presently, the municipality draws water from 19 extraction wells in the Abbotsford Sumas aquifer (the "Aquifer") and from several surface water intakes. Four additional extraction wells were recently added to the well field and operate at a rate such that an Environmental Assessment (EA) was not initially required. However, in response to a growing population and a need for emergency supply capacity, the municipal Proponent has proposed that the new wells may need to be operated at a combined pumping rate of up to 300 L/s. Under this new proposal, the increased rate requires that the project undergo an EA. This increased capacity would be sustained for five years, until such time that surface water supply systems could be commissioned.
Figure C3-1: Municipal project site location map for Case Study 3 showing aquifer extent outlined in black.
The unconfined, glaciofluvial sand and gravel source aquifer, which covers an area of approximately 200 km2 , is well understood from previous development of the water supply system and from previous academic research concerning the groundwater system. A hydrogeological consultant was retained to complete a hydrogeological assessment of potential impacts to the aquifer and to the local surface water bodies.
The objectives of the hydrogeological study and modelling effort were to:
- Assess potential impacts to the aquifer;
- Define the groundwater Zone of Influence (ZOI); and to
- Evaluate impacts to surface water flows and lake water levels.
This would be accomplished with a three-dimensional numerical model for groundwater flow and application of the model to estimate base flow changes.
An understanding of the groundwater system was well established at the time of this modelling. A detailed regional groundwater flow model had been previously developed during an earlier investigation of the aquifer.
The data available for this project included geologic maps of the project area, water level monitoring data, production well drilling records and monitoring data, academic research, and additional data collected during the consultant's one-year field program.
The field program included the establishment of groundwater, creek, and lake monitoring stations across the aquifer. Groundwater levels were monitored at 40 locations across the aquifer over a one-year period to assess the seasonal variability of groundwater levels and flow directions, and to investigate groundwater-surface water relationships. Most wells were monitored on a monthly basis. Monitoring locations were selected to maximize spatial coverage across the aquifer, and to complement existing monitoring well networks operated by Environment Canada and BC MOE. Surface water quality sampling was conducted to establish baseline chemical and physical parameters. Field parameters were measured and the samples were retained for laboratory analysis.
Pumping tests were performed in multiple wells across the aquifer over several years of study and confirmed that the source aquifer is a leaky-artesian type. The production records for existing wells were also examined. Estimates of current groundwater extraction rates from the aquifer were based on records of instantaneous flow measurements and yield estimates from operating pumping wells.
Several short-duration studies related to storm-water management provided estimates of creek flows at eight locations in the area and one-year hydrographs were produced for estimates of groundwater baseflow and recharge.
Overall, the project benefits from a well distributed set of monitoring locations, a multi-season monitoring program, and good documentation of historic and current aquifer usage.
The conceptual model for this groundwater flow modelling exercise is based on the previous understanding of subsurface geology and the hydraulic characteristics of this aquifer.
Analysis of the monitoring program and historic data showed that the water table is sensitive to seasonal variations in the infiltration of precipitation and that water levels are highest in January and lowest in October (which is consistent with regional observations in the B.C. lower mainland). Furthermore, inter-well drawdown caused by pumping of the production wells was evident.
Conceptually, the groundwater flow direction does not change seasonally but the elevation of the water table increases notably during the wet months.
A basic water balance was created, using estimates of precipitation, evapotranspiration, infiltration, run-off, and pumping records. Current extraction from the aquifer is significant and a reasonably robust estimate of current use was determined, albeit uncertain. The water balance showed that only a portion of aquifer recharge was consumed during pumping (33%) and that aquifer storage had not been depleted. Further water balance analysis suggested that the proposed increase in pumping rates (to 38% of recharge) would not negatively impact groundwater storage in the aquifer.
An existing three-dimensional groundwater flow model was used as the basis for this numerical modelling exercise. The model domain was reduced and discretization was refined to focus on the Site's surface water and groundwater systems. The three-dimensional MODFLOW code was used to simulated steady-state and transient flow. The groundwater flow model was calibrated against a spatially and temporally distributed set of observations, including transient calibration to a pumping test. Predictions included an estimate the groundwater zone of influence due to increased withdrawal rates and estimates of baseflow impacts to potentially vulnerable creeks.
Figures 2 and 3 present the distribution of hydrogeologic units assumed for the modelling exercise.
Figure C3-2: Plan view showing distribution of hydrogeologic units in the 3D model (shown here for only Layer 11 as an example). The legend shows anisotropic hydraulic conductivity values for each unit.
Figure C3-3: Model sections showing vertical distribution of hydrogeologic units.
The numerical model for this study was based on an existing regional groundwater flow model. The model domain comprises an extensive aquifer that spans the Canada-US border. The current modelling effort focuses solely on the Canadian aquifer component, targeting approximately 80 km2 of the total aquifer extent. The model domain was further reduced in size based on proximity to the Site, and by removing areas isolated by groundwater flow divides.
The model domain was discretized horizontally into grid cells 200 by 120 metres, consistent with the regional flow model, but was refined in the vicinity of the well field where significant stresses would be applied through pumping. Furthermore, the grid spacing was refined at the location of a potentially vulnerable creek. Vertically, the domain was discretized into 11 layers representing the hydrostratigraphic units identified in the previously developed conceptual model.
A constant-head boundary was simulated along the Canada-US border to simulate outflows to the US portion of the aquifer. Additionally, a constant-head boundary condition was assumed along the southeast model boundary where springs were interpreted to exist. In both cases, the head values were inferred from piezometric contour maps from October and January. A groundwater divide was interpreted at the northeast edge of the model domain; therefore, a no-flow boundary was simulated. The remaining model domain limits were assigned constant-head conditions according to the piezometric contour maps. Figure 4 provides an illustration of these boundary conditions.
Figure C3-4: Boundary conditions for the 3D numerical model, showing constant head (red), river/lake (blue), and drain (light grey along streams) boundaries as well as extent of active cells (white).
Creeks and lakes that were interpreted to be perched above the aquifer for most of the year were assigned river boundary conditions, whereas creeks that intercept the aquifer were assigned drain boundaries in MODFLOW. The nodes were assigned a range of conductance values; however, justification for these values is not provided in the documentation reviewed for this case study. In cases where the surface water feature was interpreted to be in constant hydraulic communication with the aquifer, constant head boundaries were assigned.
Assuming a uniform precipitation distribution and a constant infiltration coefficient representative of the overburden materials, the net recharge was calculated for the model. The value was considered to account for less infiltration in urban settings.
The primary sinks in this model include the municipal groundwater extraction wells, industrial supply wells, and groundwater withdrawal for crop irrigation. These withdrawal rates were assumed from pumping records, instantaneous monitoring, and estimation.
The model was calibrated under steady and transient states. The steady state calibration, performed on the basis of hydraulic head observations, was used as the basis for defining the starting hydraulic head distribution for the transient simulations.
The initial transient calibration included simulation of a pumping test completed in the existing well field to determine the aquifer transmissivity and storage properties. Back-simulation of the pumping test was complicated by the presence of other operating wells in the vicinity, and by a limited observation dataset. The drawdown values simulated were deemed reasonably close to the observed values, although no error statistics were provided due to a lack of data.
An additional transient calibration was performed for the entire aquifer. The model was calibrated to seasonal head observations, incorporating the effects of both time-dependent recharge and variable demands on the system due to crop irrigation.
For the steady-state and transient runs, the normalized root-mean-square error (NRMSE) was calculated between 4% and 6%, which was considered acceptable (Figure 5). The seasonal variations were well simulated by the model.
Figure C3-5: Example of model calibration reporting showing goodness of simulation fit and residual statistics (including geographic locations of maximum and minimum residuals).
The groundwater flow model was further calibrated by comparing simulated groundwater discharge to the creeks against estimates from manual monitoring. Aquifer properties were varied until groundwater inflows were within 15% of the measured values.
Two predictive transient runs were completed for the base case project scenario (pumping for 100 days at 290 L/s) and for a more conservative scenario (pumping for 120 days at 290 L/s). The zone of influence for each scenario was established based on the pre-determined minimum drawdown criteria of 10 centimetres. Figure shows the zone of influence in plan view.
Figure C3-6: Zone of Influence for groundwater extraction at 290 L/s. The blue and green zone borders indicate the predicted maximum extents determined through sensitivity analyses.
The reduction in creek baseflow was calculated based on the difference in zone budget calculations for the baseline transient simulation and the two predictive scenarios. The surface watercourses closest to the well field were simulated to experience the most significant reduction in flows. As two of the creeks were predicted to experience a reduction greater than the agreed Trigger Level of 10%, further investigation was recommended. As the lakes in the area are interpreted to be perched above the groundwater table, no significant impacts from pumping were anticipated.
Sensitivity analyses were conducted to determine the prediction sensitivity to uncertainty in the hydraulic conductivity and aquifer storage parameters. Where a lower and upper bound estimates were supported by field data, the aquifer parameters were adjusted to produced lower and upper bound predictions. The results of the sensitivity analysis showed a change in the hydraulic gradients in the well field; however, there was no significant change in the extent of the groundwater zone of influence. The creeks identified for further investigation in the initial predictive runs were again shown to exceed the Trigger Level for baseflow reduction in both the upper bound and lower bound cases. Again this was a result of the proximity of those creeks to the well field and to the aquifer zones which were manipulated during the sensitivity analysis.
A cumulative effects assessment regarding the loss of aquifer recharge due to future land development was deemed unnecessary by the consultant.
Transporting modelling was not completed for this groundwater extraction project.
This modelling exercise benefited from refining a regional aquifer model to simulate groundwater flow within a smaller domain that was anticipated to potentially be influenced by the proposed extraction wells. In this case, the modeller was successful in achieving the modelling objectives for this project.
This model benefitted greatly from both steady-state and transient calibrations, with stresses in the transient calibration reflective of the anticipated stresses applied to the aquifer during operations (pumping).
Results of the calibration process were only presented for the final model, and only in the form of observed vs. calculated head residuals. Spatial distribution of residuals was not presented, nor calibration sensitivity statistics.
Validation was defined as comparing groundwater discharge to creeks for the calibrated model and observations. During this validation aquifer properties were adjusted. The model was verified against an additional dataset, such as records from active operation of the well field, which commenced in July 2009 at a lower extraction rate (75 L/s) than proposed for the Project (290 L/s). .
A baseline transient model was used to model the pre-project aquifer groundwater system, which was then used to assess the change in groundwater conditions (from baseline) as a result of the project. Furthermore, a sensitivity analysis on predictive models resulted in the adoption of both a project base case and upper bound case as results.
Uncertainty analyses did not include river, lake or stream conductance values.
The modeller has presented several key assumptions for this study, namely:
- Each layer is assumed to be continuous in areal extent, but its geometry is not always congruent with a geologic formation
- The aquifer extends uniformly across the Canada-US border and groundwater outflows occur along the border
- The only source of recharge to the model is direct precipitation, which is uniformly distributed over the model domain
Furthermore, the modeller indicates that the primary sources of uncertainty in the model are identified as:
- Limited reliable information on aquifer properties
- Limited information regarding boundary conditions
- Limited understanding of the distribution of recharge across the aquifer
- Incomplete records of current groundwater usage (particularly crop irrigation)
Due to these assumptions and limitations, the model cannot perfectly represent the actual domain; however, calibration and sensitivity analyses were used to demonstrate that the model was capable of simulating observed flow behaviour in the aquifer and that the level of uncertainty did not jeopardize confidence in the model predictions.