Spatio‐Temporal Machine Learning for Regional to Continental Scale Terrestrial Hydrology.

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. Plain Language Summary: Computational models are important for understanding and predicting terrestrial hydrology, but our most physically detailed models can be time‐consuming and expensive to run over large regions. In this study, we trained deep learning models to emulate a complex hydrology model that simulates groundwater flow over the contiguous US. We developed a new model architecture called FSTR that captures spatiotemporal patterns better than standard deep learning models. FSTR is over 1,000 times faster than ParFlow, the original hydrologic model. This enables new possibilities like forecasting groundwater changes and estimating uncertainties. Our results show that specialized deep learning architectures can accurately emulate complex hydrologic models while drastically reducing computation time. Key Points: Deep learning models can emulate a complex integrated hydrology model that simulates groundwater over the United StatesWe developed a new model architecture that is more robust over longer simulation periods than off‐the‐shelf neural networksDeep‐learning based emulators of complex models enable new applications such as real‐time forecasting and estimating uncertainties [ABSTRACT FROM AUTHOR]