Premium
Understanding uncertainty in distributed flash flood forecasting for semiarid regions
Author(s) -
Yatheendradas Soni,
Wagener Thorsten,
Gupta Hoshin,
Unkrich Carl,
Goodrich David,
Schaffner Mike,
Stewart Anne
Publication year - 2008
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/2007wr005940
Subject(s) - flash flood , watershed , environmental science , surface runoff , radar , flood myth , variance (accounting) , hydrology (agriculture) , calibration , warning system , hydrological modelling , meteorology , climatology , computer science , statistics , geography , mathematics , geology , machine learning , geotechnical engineering , archaeology , ecology , telecommunications , accounting , business , biology
Semiarid flash floods pose a significant danger for life and property in many dry regions around the world. One effective way to mitigate flood risk lies in implementing a real‐time forecast and warning system based on a rainfall‐runoff model. This study used a semiarid, physics‐based, and spatially distributed watershed model driven by high‐resolution radar rainfall input to evaluate such a system. The predictive utility of the model and dominant sources of uncertainty were investigated for several runoff events within the U.S. Department of Agriculture Agricultural Research Service Walnut Gulch Experimental Watershed located in the southwestern United States. Sources of uncertainty considered were rainfall estimates, watershed model parameters, and initial soil moisture conditions. Results derived through a variance‐based comprehensive global sensitivity analysis indicated that the high predictive uncertainty in the modeled response was heavily dominated by biases in the radar rainfall depth estimates. Key model parameters and initial model states were identified, and we generally found that modeled hillslope characteristics are more influential than channel characteristics in small semiarid basins. We also observed an inconsistency in the parameter sets identified as behavioral for different events, which suggests that model calibration to historical data is unlikely to consistently improve predictive performance for different events and that real‐time parameter updating may be preferable.