Modeling runoff and sediment yield at the event scale in semiarid watersheds

Authors: WEI, H. - University Of Arizona; Polyakov, Viktor; Goodrich, David; EFRAT, M. - Hebrew University; GUERTIN, D.P. - University Of Arizona; ASSOULINE, S. - Agricultural Research Organization - Volcani Center; Heilman, Philip; Unkrich, Carl; SHMILOVICH, Y. - Hebrew University Of Jerusalem; MARRA, F. - University Of Bologna, Italy

Submitted to: International Soil and Water Conservation Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/3/2025
Publication Date: 7/4/2025
Citation: Wei, H., Polyakov, V.O., Goodrich, D.C., Efrat, M., Guertin, D., Assouline, S., Heilman, P., Unkrich, C.L., Shmilovich, Y., Marra, F. 2025. Modeling runoff and sediment yield at the event scale in semiarid watersheds. International Soil and Water Conservation Research. 13(4):860-875. https://doi.org/10.1016/j.iswcr.2025.07.001.
DOI: https://doi.org/10.1016/j.iswcr.2025.07.001

Interpretive Summary: This study tested the ability of rangeland watershed model (K2-RHEM) to predict runoff response from several small watersheds in southern Arizona. The model was able to do so on a highly incised watershed without calibration. Other watersheds required calibration of channel transfer rate. The use of additional rain gauges and channel surveys resulted in improved predictions of runoff and sediment yield in comparison to those derived with regional regression equations.

Technical Abstract: K2-RHEM, a recently integrated event-based rangeland watershed model, represents one of the few process-based models available for rangeland applications. However, to gain wider acceptance and demonstrate its reliability, comprehensive evaluation results are essential. In this study, K2-RHEM was evaluated in fivesmall semi-arid watersheds within the USDA-ARS Walnut Gulch Experimental Watershed. Using extensive runoff and sediment data, along with fieldsurveys on channel heads, soil textures, and channel cross-sections, the model showed strong performance in predicting hydrology metrics without calibration: NS ranged from 0.53 to 0.87 and KGE from 0.54 to 0.88 for runoff; NS from 0.59 to 0.85 and KGE from 0.69 to 0.90 for runoff peak; and NS from 0.98 to 0.99 and KGE from 0.94 to 0.98 for time to peak. Sediment yield predictions were particularly accurate in watersheds with significantchannel incisions, with NS of 0.65 and KGE of 0.79. Good sediment yield calibration and validation results were achieved in three watersheds, and reasonable results achieved in the smallest watershed. Sediment yield and runoff peak were found to be sensitive to level of watershed discretization. Improved model performance was seen with additional rain gauges even in small watersheds. These findingsdemonstrate the potential of K2-RHEM as a reliable tool for the prediction of hydrology and erosion for small-scale rangeland watershed management and highlight the importance of both proper watershed discretization and rainfall data resolution in model applications.