2026-10-14T03:22:44Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33969058 2026-09-22T22:11:30Z category_80 portal_316 item_type_3 month_year_09_2026

Data Sheet 1_Predicting runoff in data-sparse drylands: a case study evaluating regionalization success for two parsimonious loss models.pdf Farzaneh Soltani (18568039) Sara Rassa (25092187) Gerhard Schoener (25092190) Hydrology arid curve number flash flood linear-constant infiltration model regionalization Introduction<p>Reliable flash flood forecasting is essential for sustainable storm infrastructure planning and hazard mitigation in sensitive dryland landscapes. Loss or infiltration models are central to flash flood prediction in arid watersheds, where runoff is dominated by infiltration excess. Parsimonious models are critical for use in practice. Accurate prediction remains a challenge because calibration data are typically scarce or unavailable. Regionalization can help fill this data gap: model parameters from a physically similar donor basin with available calibration data are transferred to a target watershed with no or limited data.</p>Methods<p>In this study, we tested whether regionalization success depends on model selection. Two simple loss models available in the widely-used HEC-HMS software – the SCS curve number (CN) and the linear-constant (LC) methods – were calibrated and validated based on long-term observations from the Walnut Gulch Experimental Watershed in Arizona. Optimized parameters derived from this donor basin were then transferred to the target basin, a partially urbanized watershed in New Mexico with a dense network of rain and flow gaging stations. The target basin has a limited observational record and therefore serves as a suitable test case for regionalization under data-limited conditions.</p>Results<p>Validation results from the donor basin demonstrate that the LC model outperformed the CN model for large storm events across four performance metrics. Performance for small events was not significantly different. When parameters were transferred to the target basin and tested for a rare, high-intensity storm event recorded in 2025, the LC model again performed better for most metrics.</p>Discussion<p>While regionalization can be a viable strategy to support stormwater infrastructure design in data-scarce, semi-arid urban watersheds, this study indicates that its success depends on model selection.</p> 2026-09-22T22:11:30Z Dataset Dataset 10.3389/frwa.2026.1959714.s001 https://figshare.com/articles/dataset/Data_Sheet_1_Predicting_runoff_in_data-sparse_drylands_a_case_study_evaluating_regionalization_success_for_two_parsimonious_loss_models_pdf/33969058 CC BY 4.0