IMPROVING A HYDROGEOLOGICALLY COMPLEX AQUIFER MODEL USING TRANSIENT GROUNDWATER LEVELS, EAST FLATHEAD VALLEY, MONTANA

The Montana Bureau of Mines and Geology is investigating ground water resources within the East Flathead Valley (EFV) due to increasing groundwater development. The groundwater system consists of a complex package of unconsolidated glacial, lacustrine, and fluvial Quaternary sediments underlain by poorly cemented Tertiary sediments and bedrock.

Wells in the valley are completed within sand and gravel aquifers, including a shallow layer (0-200 ft. below surface), an intermediate layer (250-400 ft. below surface), and a deep layer (500-1,500 ft. below surface). Silt and clay rich lacustrine sediments and glacial till form aquitards between the aquifers over most (but not all) of the study area. Water levels were measured monthly in approximately 100 domestic and irrigation wells located throughout the EFV, from August 2019 – September 2021, with groundwater levels varying from 3101 ft. MSL along the Swan mountain front to 2887 ft. MSL near the Flathead River. Change in groundwater levels across the EFV captures the dynamic behavior of the hydrologic system in response to the seasonality of recharge and pumping.

Head values were useful for calibrating the steady-state model. Due to inherent residual errors from calibrating to heads, calibration storativity (Ss) in the transient version of the model to head targets compounded errors for wells with more observations and constant error residual. To avoid this issue, we used drawdown values (relative to steady-state) as the transient model targets. This eliminated the penalization of wells with constant head residuals, while also emphasizing calibration to timing and magnitude of ground water levels. The Nash-Sutcliffe Efficiency (NSE) statistic was used to score each observation well, with a score >0.5 considered optimal. NSE values after calibration varied from -1.76 to 0.85, with an average of 0.13 and a median of 0.25. Poor NSE values were common in wells with little variation in water levels, because these produced erroneously high error calculations due to noise. NSE was useful at wells with dynamic and large magnitude water level changes throughout their observation record.