A Unified and Scalable Framework for the Comprehensive Evaluation of Hydrological Model Behavior and Structural Diagnostics

Carlos Erazo, Yusuf Sermet, Ali Jannesari, Chaoqun Lu, Ibrahim Demir

Robust evaluation of hydrological models requires assessing not only predictive accuracy, but also uncertainty behavior and structural realism. This study introduces the Hydrological Model Evaluation System (HyMES), a unified framework that organizes deterministic, probabilistic, and structural diagnostics, including dynamic hysteresis analysis, into a coherent, station level evaluation workflow. HyMES is implemented as a fully browser based application, enabling interactive, reproducible analysis without local installations. The framework is demonstrated through a comparative evaluation of retrospective streamflow simulations from the National Water Model (NWM v2.1 and v3.0) across stations in the Mississippi River region for 2010–2023. Results show that modest deterministic improvements in newer model versions often arise from compensatory error structures and do not consistently translate to improved uncertainty robustness or structural behavior. Structural and hysteresis diagnostics reveal persistent deficiencies in flow variability, seasonal organization, and catchment memory that are masked by aggregate metrics. HyMES provides a scalable, transparent platform for diagnosing hydrological model behavior beyond single metric performance assessment.