Authors: Jian Shi, and Chang Sun – University of Alberta; Hilary McMillan, and Ryoko Araki – San Diego State University; Martyn Clark, Cyril Thébault, and Nicolás Vásquez – University of Calgary, Hongli Liu – Montana State University
Title: Identifying Dominant Hydrologic Processes Across Large Domains via Sensitivity Analysis
Presentation Type: Poster Presentation
Abstract: Large-domain streamflow forecasting is hindered by strong spatial variability in hydrologic processes, limiting the use of a single universal model structure. This study develops an automatic approach to identify dominant hydrologic processes across diverse landscapes and improve process representation in hydrologic models. Unlike traditional approaches based on expert judgment and field studies, which are resource-intensive and difficult to generalize, we integrate physically based modeling with sensitivity analysis to systematically diagnose process controls on streamflow. The approach is first applied to a new global dataset of 90 small catchments (median area: 48 km²) with well-documented field-based process understanding. The dataset includes hourly meteorological forcing, geospatial attributes, and daily streamflow. Simulations are conducted with the SUMMA–MizuRoute framework, where SUMMA represents a comprehensive process set (snow, evapotranspiration, infiltration, and surface and subsurface flow, etc.) to enable identification of dominant processes. VISCOUS sensitivity analysis is then used to quantify process influence and identify dominant controls, which are evaluated against field-based literature. The approach is subsequently extended to 1,426 catchments from the CAMELS-SPAT dataset for continental-scale analysis across North America. Ongoing work focuses on model calibration, followed by process identification and spatial extrapolation to produce a seamless map of process dominance. This approach enables automatic characterization of hydrologic heterogeneity and supports improved model development, prediction in ungauged basins, and large-domain water resources applications.