Advancing Mass-Conserving Deep Learning for Distributed Hydrologic Modeling: A Case Study in a Snow-Dominated Watershed
Authors: Yihan Wang, Lujun Zhang, N. Benjamin Erichson, Yang Hong, Tiantian Yang
Organizations: School of Civil Engineering and Environmental Science, University of Oklahoma, Norman, OK, United States · Lawrence Berkeley National Laboratory, Berkeley, CA, United States · International Computer Science Institute, Berkeley, CA, United States · School for Environment and Sustainability, University of Michigan, Ann Arbor, United States
Recent progress in physics-aware deep learning (DL) has embedded mass conservation within Long Short-Term Memory (LSTM) architecture for hydrologic modeling, improving physical consistency and enabling diagnostics of internal states. However, existing mass-conserving neural networks are primarily developed in lumped formulations. Consequently, previous studies provide only fragmented and often indirect evidence of when and why imposed physical constraints may lead to unrealistic hydrologic behavior. To address this limitation, we propose Physics-Enhanced Conservation-Augmented Net (PECAN), a mass-conserving convolutional LSTM-based framework for distributed hydrologic modeling and diagnosing structural limitations of physics-aware DL models. PECAN is implemented in two variants: (1) PECAN-MC with a strict precipitation-based conservation constraint and (2) PECAN-MCR with a relaxed conservation constraint allowing additional water inputs beyond precipitation. Both variants are trained solely on gauge streamflow and evaluated in a snow-dominated watershed in Colorado. We benchmark PECAN against lumped DL baselines and the distributed process-based Variable Infiltration Capacity (VIC) model. Results show that PECAN achieves competitive outlet streamflow simulation accuracy while producing spatially explicit latent states for interpretation. Comparisons with reference soil moisture indicate that PECAN cell states resemble meaningful temporal soil moisture dynamics. However, analyses also reveal that PECAN misrepresents snow processes by prematurely removing winter snow water storage from the system rather than retaining it for spring release. PECAN demonstrates accurate streamflow simulation at gauges and shows diagnostic potential to expose structural limitations in hydrologic process representation, thereby informing future physics-aware neural network development for hydrology.
Figures & tables
Figure 1: Maps of the North Platte Headwaters watershed. The left panel shows the watershed location within the western United States. The right panel shows the watershed elevation, boundary, and river network. Red triangles indicate USGS streamflow stations used in this study.
Figure 2: Overall workflow of the Physics-Enhanced Conservation-Augmented Network (PECAN). Gridded meteorological forcings are used as spatial inputs to PECAN, which is implemented in two variants (i.e., PECAN MC and PECAN MCR ). The model produces spatially distributed runoff, which is routed to gauge locations to generate streamflow predictions for optimization against streamflow observations. PECAN also produces trash cell flux and cell states for hydrologic interpretation.
Figure 3: Schematic illustration of the PECAN runoff module. The fast-flow storage ctfast and slow-flow storage ctslow evolve through conservation-constrained gating mechanisms. Under mass conservation (MC) constraint, the input gate it partitions precipitation xtprcp into fast-flow storage channels, the redistribution operator Re mixes existing fast-flow storage, and the output gate ot controls the release flux from the fast-flow storage. This released flux consists of the trash cell flux qtTC representing water sinks such as ET or unresolved losses, as well as the fast runoff qtfast . Under the mass conservation relaxed (MCR) constraint, an additional bidirectional relaxation gate GtMR is included to allow for additional mass injection or removal within the fast-flow storage. The leakage coefficient ρ governs both recharge from fast-flow storage to slow-flow storage, as well as discharge from the slow-flow storage as slow runoff qtslow . The effective runoff field used for routing excludes the trash cell and sums the fast and slow runoff.
Hydrologic models must reconcile predictive flexibility with physical interpretability, yet the complexity required to reproduce catchment dynamics remains uncertain. Here we compare conceptual hydrologic models and neural networks built from mass-conserving computational units across 513 catchments in the contiguous United States, varying state dimensionality and cross-state information sharing. Under likelihood-based selection, median testing-period Kling-Gupta efficiency skill score (KGEss) increases from 0.83 with one allowable state to 0.89 with two, but reaches only 0.90 with five. KGE-based selection shows similarly rapid saturation. Cross-state information sharing provides an additional source of predictive skill: sharing-enabled candidate spaces restricted to two states outperform non-sharing spaces allowing up to five states, with lower median errors in five of six hydrologic diagnostics. Although likelihood- and KGE-based criteria favor different architectures, both preserve the same structural relationships. Conceptual and neural formulations likewise select different specific architectures but retain the same state dimensionality in 75% of catchments. These findings indicate that effective hydrologic representations are low-dimensional but interaction-dependent, while predictive success alone does not uniquely identify physical storage structures.
Yuan-Heng Wang, Hoshin V. Gupta
Department of Hydrology and Atmospheric Science, University of Arizona, Tucson, AZ
Dooge asked whether hydrology has laws at the catchment scale. We ask whether a low-dimensional system built under physical laws can match the accuracy of neural networks that outperform process-based models on the same inputs. Three laws constrain a lumped catchment model. Conservation makes it a flux-connected storage network, dissipation fixes the sign of every flux and aggregation determines how the unresolved composition enters the fluxes, through effective laws over threshold distributions and a memory of the forcing. These laws admit four kinds of refinement. A parsimonious model under them, CaRT (11 parameters per catchment), is refined in a loop assisted by AI agents. Each refinement addresses an observed deficit through a hypothesis from the literature. The result, CaRT+, has fourteen states, 35 parameters per catchment and no daily neural network. Most refinements follow from aggregation over thresholds or from variables that carry no water but modulate fluxes. On 602 British and 593 US gauged catchments, single CaRT+ models reach median Nash-Sutcliffe efficiencies of 0.863 and 0.765 in held-out years. A stateful LSTM reaches 0.869 and 0.751 and an entity-aware LSTM 0.862 and 0.754, while three-seed LSTM ensembles remain ahead of the CaRT+ ensemble by 0.003-0.012 in median paired NSE. On 505 drier Australian gauged catchments unused in development, single CaRT+ models reach 0.778 against 0.762 and 0.753 for the two LSTMs. The results suggest that catchment-scale laws exist as laws of form, with the constitutive families shared and only their parameters learned.
Xilin Xia
School of Engineering, University of Birmingham, Birmingham, UK
Machine learning models can achieve high predictive accuracy in hydrological applications but often lack physical interpretability. The Mass-Conserving Perceptron (MCP) provides a physics-aware artificial intelligence (AI) framework that enforces conservation principles while allowing hydrological process relationships to be learned from data. In this study, we investigate how progressively embedding physically meaningful representations of hydrological processes within a single MCP storage unit improves predictive skill and interpretability in rainfall-runoff modeling. Starting from a minimal MCP formulation, we sequentially introduce bounded soil storage, state-dependent conductivity, variable porosity, infiltration capacity, surface ponding, vertical drainage, and nonlinear water-table dynamics. The resulting hierarchy of process-aware MCP models is evaluated across 15 catchments spanning five hydroclimatic regions of the continental United States using daily streamflow prediction as the target. Results show that progressively augmenting the internal physical structure of the MCP unit generally improves predictive performance. The influence of these process representations is strongly hydroclimate dependent: vertical drainage substantially improves model skill in arid and snow-dominated basins but reduces performance in rainfall-dominated regions, while surface ponding has comparatively small effects. The best-performing MCP configurations approach the predictive skill of a Long Short-Term Memory benchmark while maintaining explicit physical interpretability. These results demonstrate that embedding hydrological process constraints within AI architectures provides a promising pathway toward interpretable and process-aware rainfall-runoff modeling.
Mohammad A. Farmani, Hoshin V. Gupta, Ali Behrangi +3
Department of Hydrology and Atmospheric Sciences, University of Arizona, Tucson, AZ, USA · Department of Geosciences, University of Arizona, Tucson, AZ, USA