Evaluating the Use of Environmental Tracers to Reduce Conceptual Model Uncertainty of Hydrogeologic Models
Abstract/Artist Statement
Uncertainty in groundwater studies is inherently due to unknown complexities of hydrologic processes and subsurface physical structure. We investigate methods to reduce uncertainty in these components by testing a suite of alternate conceptual hydrogeologic flow and transport models at a site contaminated with mining tailings near Riverton, Wyoming. We test the hypothesis that using environmental tracer data will result in a larger decrease in uncertainty over testing with water level data alone.
Multiple conceptual models are developed using hydrologic and geologic data to address uncertainty and behavior of the system. Two process conceptualizations differ in how groundwater recharge is applied across the top boundary: a spatially uniform flux versus dependence on vegetation density. Geologic structure is also conceptualized in several ways: a surficial gravel layer and a permeable sandstone layer, each with homogenous properties, are separated by a semi-permeable clay boundary; there are five homogeneous layers with a middle sandstone/clay pair pinching out; and a heterogeneous system populated by highly permeable channel-floodplain features in the uppermost layer. These representations are conditioned to wellbore data to create stochastic grids used in process-based groundwater models. MODFLOW 2005 and MT3DMS codes are used to simulate groundwater flow and chemical transport of environmental tracers: chlorofluorocarbons (CFCs), sulfur-hexafluoride (SF6) and tritium (3H). These tracers are chosen because they have a known temporal signal to global groundwater systems and their movement in the subsurface is sensitive to those complexities controlling flow and transport. We compare the resulting simulation concentrations and water levels against field data gathered from 2018 to 2020 to test the ensemble of conceptual models. Preliminary results show that differences between simulated concentrations for each alternate conceptual model is greater than the analytical detection limit, meaning differences between simulated concentrations is caused by differences between each model. Further testing will use least squares residuals and Bayesian likelihood calculations to investigate each data type’s ability to reduce model uncertainty and select an optimal hydrogeologic conceptual model for the Riverton site.
Previous site characterization yielded a single conceptual model that was unable to account for increased contaminant concentrations after flooding events in 2010. This work uses a different, multi-model approach to identify a conceptual model to use in future contaminant transport modeling at the site. On a broader scale, this research will further explore the utility of testing hydrogeologic conceptual models using environmental tracer data and the possibility for greater confidence in model selection over testing with hydraulic data alone. This will allow hydrogeologists and forecast analysts to better assess uncertainty of alternate hydrogeologic conceptualizations and make informed groundwater resource decisions.
Evaluating the Use of Environmental Tracers to Reduce Conceptual Model Uncertainty of Hydrogeologic Models
UC 331
Uncertainty in groundwater studies is inherently due to unknown complexities of hydrologic processes and subsurface physical structure. We investigate methods to reduce uncertainty in these components by testing a suite of alternate conceptual hydrogeologic flow and transport models at a site contaminated with mining tailings near Riverton, Wyoming. We test the hypothesis that using environmental tracer data will result in a larger decrease in uncertainty over testing with water level data alone.
Multiple conceptual models are developed using hydrologic and geologic data to address uncertainty and behavior of the system. Two process conceptualizations differ in how groundwater recharge is applied across the top boundary: a spatially uniform flux versus dependence on vegetation density. Geologic structure is also conceptualized in several ways: a surficial gravel layer and a permeable sandstone layer, each with homogenous properties, are separated by a semi-permeable clay boundary; there are five homogeneous layers with a middle sandstone/clay pair pinching out; and a heterogeneous system populated by highly permeable channel-floodplain features in the uppermost layer. These representations are conditioned to wellbore data to create stochastic grids used in process-based groundwater models. MODFLOW 2005 and MT3DMS codes are used to simulate groundwater flow and chemical transport of environmental tracers: chlorofluorocarbons (CFCs), sulfur-hexafluoride (SF6) and tritium (3H). These tracers are chosen because they have a known temporal signal to global groundwater systems and their movement in the subsurface is sensitive to those complexities controlling flow and transport. We compare the resulting simulation concentrations and water levels against field data gathered from 2018 to 2020 to test the ensemble of conceptual models. Preliminary results show that differences between simulated concentrations for each alternate conceptual model is greater than the analytical detection limit, meaning differences between simulated concentrations is caused by differences between each model. Further testing will use least squares residuals and Bayesian likelihood calculations to investigate each data type’s ability to reduce model uncertainty and select an optimal hydrogeologic conceptual model for the Riverton site.
Previous site characterization yielded a single conceptual model that was unable to account for increased contaminant concentrations after flooding events in 2010. This work uses a different, multi-model approach to identify a conceptual model to use in future contaminant transport modeling at the site. On a broader scale, this research will further explore the utility of testing hydrogeologic conceptual models using environmental tracer data and the possibility for greater confidence in model selection over testing with hydraulic data alone. This will allow hydrogeologists and forecast analysts to better assess uncertainty of alternate hydrogeologic conceptualizations and make informed groundwater resource decisions.