Multi-Observation Calibration of a Regional Scale Groundwater Flow Model

English Abstract:

Since 2009, a regional groundwater characterization program is ongoing in the province of Québec (Canada). Among the numerous results from these projects, regional distributed groundwater recharge rates, groundwater ages from isotopic tracers and hydraulic properties were obtained. Integrating these data into a regional groundwater flow model has not yet been attempted in the St. Lawrence Lowlands. The objective of this work was to use multiple source data to calibrate a regional scale groundwater flow model. The study area is located in the central part of the St. Lawrence Lowlands and in the Appalachian mountains on the south shore of the St. Lawrence River (SLR) (7750 km²). Between 2009 and 2014, groundwater was sampled for ³He, ³H and ¹⁴C analyses (42 samples). A distributed water budget calculation based on the Runoff Curve Number was used to calculate recharge for the entire study area. Groundwater outflow to peatlands was estimated based on a local Darcy approach. The steady-state 3D groundwater model was developed using Modflow, based on 100 m x 100 m cells and six vertical layers. The first layer represents the Quaternary sediments and the five others represent the fractured bedrock aquifer. The model was calibrated to reproduce 11 300 measured water levels, 22 measured river baseflows, estimated groundwater seepage to peatlands and groundwater ages obtained from ³H/³He and ¹⁴C. The calibrated model shows that the majority of the recharge water is discharged at the scale of local watersheds, that regional groundwater flow is limited, and that almost no deep groundwater flows to the SLR. This confirms data from the regional water budget. Groundwater outflow to peatlands represent the range of flows estimated from local hydraulic gradients and hydraulic conductivities. The average and maximum particles travel times at the sampled wells compare favorably to the results from isotopic tracers with ages ranging from 2 to 22 000 years. Using different types of calibration data provided a model that simulate adequately a wide range of processes.