Development of a Deep Learning Emulator for a Distributed Groundwater–Surface Water Model: ParFlow-ML

Integrated hydrologic models solve coupled mathematical equations that represent natural processes, including groundwater, unsaturated, and overland flow. However, these models are computationally expensive. It has been recently shown that machine leaning (ML) and deep learning (DL) in particular could be used to emulate complex physical processes in the earth system. In this study, we demonstrate how a DL model can emulate transient, three-dimensional integrated hydrologic model simulations at a fraction of the computational expense. This emulator is based on a DL model previously used for modeling video dynamics, PredRNN. The emulator is trained based on physical parameters used in the original model, inputs such as hydraulic conductivity and topography, and produces spatially distributed outputs (e.g., pressure head) from which quantities such as streamflow and water table depth can be calculated. Simulation results from the emulator and ParFlow agree well with average relative biases of 0.070, 0.092, and 0.032 for streamflow, water table depth, and total water storage, respectively. Moreover, the emulator is up to 42 times faster than ParFlow. Given this promising proof of concept, our results open the door to future applications of full hydrologic model emulation, particularly at larger scales.

A national topographic dataset for hydrological modeling over the contiguous United States

Topography is a fundamental input to hydrologic models critical for generating realistic streamflow networks as well as infiltration and groundwater flow. Although there exist several national topographic datasets for the United States, they may not be compatible with gridded models that require hydrologically consistent digital elevation models (DEMs). Here, we present a national topographic dataset developed to support gridded hydrologic simulations at 1 km and 250 m spatial resolution over the contiguous United States. The workflow is described step by step in two parts: (a) DEM processing using a Priority Flood algorithm to ensure hydrologically consistent drainage networks and (b) slope calculation and smoothing to improve drainage performance. The accuracy of the derived stream network is evaluated by comparing the derived drainage area to drainage areas reported by the national stream gage network. The slope smoothing steps are evaluated using the runoff simulations with an integrated hydrologic model. Our DEM product started from the National Water Model DEM to ensure our final datasets will be as consistent as possible with this existing national framework. Our analysis shows that the additional processing we provide improves the consistency of simulated drainage areas and the runoff simulations that simulate gridded overland flow (as opposed to a network routing scheme). The workflow uses an open-source R package, and all output datasets and processing scripts are available and fully documented. All of the output datasets and scripts for processing are published through CyVerse at 250 m and 1 km resolution. The DOI link for the dataset is https://doi.org/10.25739/e1ps-qy48 (Zhang and Condon, 2020).

A national topographic dataset for hydrological modeling over contiguous United States

Topography is a fundamental input to hydrologic models critical for generating realistic streamflow networks as well as infiltration and groundwater flow. Although there exist several national topographic datasets for the United States, they may not be compatible with gridded models that require hydrologically consistent Digital Elevation Models (DEMs). Here, we present a national topographic dataset developed support physically based hydrologic simulations at 1 km and 250 m spatial resolution over contiguous United States. The workflow is described step-by-step in two parts (a) DEM processing using a Priority Flood algorithm to ensure hydrologically consistent drainage networks and (b) slope calculation and smoothing to improve drainage performance. The accuracy of derived stream network is evaluated by comparing the derived drainage area to drainage areas reported by the national stream gage network. The slope smoothing steps are evaluated using the runoff simulations with an integrated hydrologic model. The processed DEM is designed to capture the topographic features and improve the runoff simulations for the models solving partial differential equations. The workflow uses an open-source R package and all output datasets and processing scripts are available and fully documented here. All of the output datasets and scripts for processing are published through Cyverse at 250 m and 1 km resolution. The DOI link for the dataset is https://doi.org/10.25739/e1ps-qy48 (Zhang and Condon, 2020).

Examining cross-scale influences of forcing resolutions in a hillslope-resolving, integrated hydrologic model

Water management in semiarid regions of the western United States requires accurate and timely knowledge of runoff generated by snowmelt. This information is used to plan reservoir releases for downstream users and hydrologic models play an important role in estimating the volume of snow stored in mountain watersheds that serve as source waters for downstream reservoirs. Physically based, integrated hydrologic models are used to develop spatiotemporally dynamic estimates of hydrologic states and fluxes based on understanding of the underlying biophysics of hydrologic response. Yet this class of models are associated with many issues that give rise to significant uncertainties in key hydrologic variables of interest like snow water storage and streamflow. Underlying sources of uncertainty include difficulties in parameterizing processes associated with nonlinearities of some processes, as well as from the large variability in the characteristic spatial and temporal scale of atmospheric forcing and land-surface water and energy balance and groundwater processes. Scale issues, in particular, can introduce systematic biases in integrated atmospheric and hydrologic modeling. Reconciling these discrepancies while maintaining computational tractability remains a fundamental challenge in integrated hydrologic modeling. Here we investigate the hydrologic impact of discrepancies between distributed meteorological forcing data exhibiting a range of spatial scales consistent with a variety of numerical weather prediction models when used to force an integrated hydrologic model associated with a corresponding range of spatial resolutions characteristic of distributed hydrologic modeling. To achieve this, we design and conduct a total of twelve numerical modeling experiments that seek to quantify the impact of applied resolution of atmospheric forcings on simulated hillslope-scale hydrologic state variables. The experiments are arranged in such way to assess the impact of four different atmospheric forcing resolutions (i.e., interpolated 30 m, 1 km, 3 km and 9 km) on two hydrologic variables, snow water equivalent and soil water storage, arranged in three hydrologic spatial resolution (i.e., 30 m, 90 m and 250 m). Results show spatial patterns in snow water equivalent driven by atmospheric forcing in hillslope-scale simulations and patterns mostly driven by topographical characteristics (i.e., slope and aspect) on coarser simulations. Similar patterns are observed in soil water storage however, in addition to that, large errors are encountered primarily in riparian areas of the watershed on coarser simulations. The Weather Research Forecasting (WRF) model is used to develop the environmental forcing variables required as input to the integrated hydrologic model. WRF is an open source, community supported coupled land-atmosphere model capable of capturing spatial scales that permit convection. The integrated hydrologic modeling framework used in this work coincides with the ParFlow open-source surface-subsurface hydrology model. This work has important implications for the use of atmospheric and integrated hydrologic models in remote and ungauged areas. In particular, this work has potential ramifications for the design and development of observing system simulation experiments (OSSEs) in complex and snow-dominated landscapes. OSSEs are critical in constraining the performance characteristics of Earth-observing satellites.

Simulating coupled surface–subsurface flows with ParFlow v3.5.0: capabilities, applications, and ongoing development of an open-source, massively parallel, integrated hydrologic model

Surface flow and subsurface flow constitute a naturally linked hydrologic continuum that has not traditionally been simulated in an integrated fashion. Recognizing the interactions between these systems has encouraged the development of integrated hydrologic models (IHMs) capable of treating surface and subsurface systems as a single integrated resource. IHMs are dynamically evolving with improvements in technology, and the extent of their current capabilities are often only known to the developers and not general users. This article provides an overview of the core functionality, capability, applications, and ongoing development of one open-source IHM, ParFlow. ParFlow is a parallel, integrated, hydrologic model that simulates surface and subsurface flows. ParFlow solves the Richards equation for three-dimensional variably saturated groundwater flow and the two-dimensional kinematic wave approximation of the shallow water equations for overland flow. The model employs a conservative centered finite-difference scheme and a conservative finite-volume method for subsurface flow and transport, respectively. ParFlow uses multigrid-preconditioned Krylov and Newton–Krylov methods to solve the linear and nonlinear systems within each time step of the flow simulations. The code has demonstrated very efficient parallel solution capabilities. ParFlow has been coupled to geochemical reaction, land surface (e.g., the Common Land Model), and atmospheric models to study the interactions among the subsurface, land surface, and atmosphere systems across different spatial scales. This overview focuses on the current capabilities of the code, the core simulation engine, and the primary couplings of the subsurface model to other codes, taking a high-level perspective.