Search for: All records

Abstract Hydrogeologic models generally require gridded subsurface properties, however these inputs are often difficult to obtain and highly uncertain. Parametrizing computationally expensive models where extensive calibration is computationally infeasible is a long standing challenge in hydrogeology. Here we present a machine learning framework to address this challenge. We train an inversion model to learn the relationship between water table depth and hydraulic conductivity using a small number of physical simulations. For a 31M grid cell model of the US we demonstrate that the inversion model can produce a reliable K field using only 30 simulations for training. Furthermore, we show that the inversion model captures physically realistic relationships between variables, even for relationships that were not directly trained on. While there are still limitations for out of sample parameters, the general framework presented here provides a promising approach for parametrizing expensive models. 

Abstract. The Heihe River basin in northwest China depends heavilyon both anthropogenic and natural storage (e.g., surface reservoirs, rivers andgroundwater) to support economic and environmental functions. The QilianMountain cryosphere in the upper basin is integral to recharging thesestorage supplies. It is well established that climate warming is drivingmajor shifts in high-elevation water storage through loss of glaciers andpermafrost. However, the impacts on groundwater–surface-water interactionsand water supply in corresponding lower reaches are less clear. We built anintegrated hydrologic model of the middle basin, where most water usageoccurs, in order to explore the hydrologic response to the changingcryosphere. We simulate the watershed response to loss of glaciers (glacier scenario),advanced permafrost degradation (permafrost scenario), both of these changes simultaneously (combined scenario) andprojected temperature increases in the middle basin (warming scenario) by alteringstreamflow inputs to the model to represent cryosphere-melting processes, aswell as by increasing the temperature of the climate forcing data. Netlosses to groundwater storage in the glacier scenario and net gains in the permafrost and combined scenarios showthe potential of groundwater exchanges to mediate streamflow shifts. Theresult of the combined scenario also shows that permafrost degradation has more of animpact on the system than glacial loss. Seasonal differences ingroundwater–surface-water partitioning are also evident. The glacier scenario hasthe highest fraction of groundwater in terms of streamflow in early spring. Thepermafrost and combined scenarios meanwhile have the highest fraction of streamflowinfiltration in late spring and summer. The warming scenario raises the temperatureof the combined scenario by 2 ∘C. This results in net groundwater storageloss, a reversal from the combined scenario. Large seasonal changes inevapotranspiration and stream network connectivity relative to the combined scenario show thepotential for warming to overpower changes resulting from streamflow. Ourresults demonstrate the importance of understanding the entire system ofgroundwater–surface-water exchanges to assess water resources underchanging climatic conditions. Ultimately, this analysis can be used toexamine the cascading impact of climate change in the cryosphere on theresilience of water resources in arid basins downstream of mountain rangesglobally. 

Abstract Integrated hydrologic models can simulate coupled surface and subsurface processes but are computationally expensive to run at high resolutions over large domains. Here we develop a novel deep learning model to emulate subsurface flows simulated by the integrated ParFlow‐CLM model across the contiguous US. We compare convolutional neural networks like ResNet and UNet run autoregressively against our novel architecture called the Forced SpatioTemporal RNN (FSTR). The FSTR model incorporates separate encoding of initial conditions, static parameters, and meteorological forcings, which are fused in a recurrent loop to produce spatiotemporal predictions of groundwater. We evaluate the model architectures on their ability to reproduce 4D pressure heads, water table depths, and surface soil moisture over the contiguous US at 1 km resolution and daily time steps over the course of a full water year. The FSTR model shows superior performance to the baseline models, producing stable simulations that capture both seasonal and event‐scale dynamics across a wide array of hydroclimatic regimes. The emulators provide over 1,000× speedup compared to the original physical model, which will enable new capabilities like uncertainty quantification and data assimilation for integrated hydrologic modeling that were not previously possible. Our results demonstrate the promise of using specialized deep learning architectures like FSTR for emulating complex process‐based models without sacrificing fidelity.