Development of a Semi-Arid, Site-Specific Flash Flood Forecasting System for the Western Region: Results, Insights and Way Forward

An increasingly drier and more variable climate trend has significantly increased the incidence of intense (extreme) precipitation events during the 20th century. This has led to a continuous growth of extreme-flood- event losses, despite the widespread problem of water scarcity. In semi-arid regions, the extremely localized and intense summertime convective storm systems cause 'short-fused' flash floods, often result in significant risk to life and property. These arid/semi-arid regions have significant coverage, currently spanning approximately one-third of the earth's surface, and possibly spanning more in the future, for example due to current global warming. The short spatial and temporal extents of flash-floods and the highly complex nature of semi-arid hydrology make the subsequent predictions extremely difficult. Current NWS flash-flood predictions and warnings are based either on usually lumped humid region hydrology models, or on areal rain averages translated to experience-based guidelines. To improve the flash-flood predictive capability to be more `site-specific', an established event-based semi-arid rainfall-runoff model KINEROS2 was modified to allow for the continuous simulation of the basin response driven by high-resolution radar precipitation measurements in an uncertainty framework. The model contains process descriptions required to represent semi-arid and arid regions, including a dynamic infiltration algorithm and the ability to account for channel transmission losses. This forecasting system has been made operational at the Tucson NWS office. This talk focuses on the results from the sophisticated multi-objective distributed model calibration framework developed and used on the operational forecast points which are also burn recovery basins, along with the lessons learned, and the possible future directions.