Advances in Hydrologic Modeling: Data Provenance, Lateral Flow Routing, and Numerical Accuracy
StatisticsView Usage Statistics
Numerical hydrologic models are useful tools for scientific analysis and water resource management. However, major challenges exist regarding their conceptual and computational underpinnings and in the way they are used. This study addresses key challenges in numerical hydrologic modeling for catchment and global scale applications. Specifically a Python framework was developed to address issues of model reproducibility and management of large datasets and model ensembles. The framework is demonstrated with a global moment-independent parameter sensitivity analysis on a commonly used watershed model. Next, an analysis of model uncertainty and bias that can be attributed to numerical truncation error caused by large time steps in Richards equation is presented for global applications of the Community Land Model. Temporal truncation error is linked to seasonal and spatial model biases. An adaptive sub- time-stepping scheme is shown to effectively remove these errors with minimal trade-offs between numerical accuracy and computational expense. Last, lateral subsurface flow in mountain watersheds in the Great Basin is quantified using a spatially distributed hydrologic model. Different levels of spatial representation and connectivity for lateral flow are shown to significantly alter model estimates of subbasin water budgets. Implications are focused on the general lack of representation of hillslope scale lateral redistribution of subsurface water in Earth System Models. To address this, a conceptual framework is presented for evaluation and inter-comparison of different modeling approaches of lateral flow that are relevant for Earth System Models.