Peak flow analysis using a two-dimensional watershed model with radar precipitation data

Abstract

The computational speed of computers and availability of spatial hydrologic data make distributed watershed models a viable approach for many applications including peak flow and storm motion analysis. A study is presented that couples distributed watershed modeling with radar rainfall estimates to analyze peak flow with the following objectives: 1) demonstrate the coupling of radar data with a distributed hydrologic model; 2) analyze conditions of storm size, velocity and intensity that produce peak discharge exceeding a specified threshold; and 3) examine the potential for increasing forecast lead-time using radar data and distributed modeling. The CASC2D watershed model is applied on two watersheds in central Arizona, the Cave Creek and Hassyampa River basins. WSR-88D radar data for four rainfall events is analyzed, and uncalibrated radar data are shown to underestimate precipitation intensities. Calibration of the radar Z-R relationship based on ground observations is used, and these calibrated radar data compare well with rain gauge data and are used with the CASC2D model to successfully reproduce two runoff events on the Hassyampa River. Analysis of moving storm characteristics based on radar observations is also presented. A parametric analysis is conducted to analyze the relationship between peak runoff and storm size, velocity, and intensity. A test matrix of moving storms with various velocities, sizes and intensities is applied to the Cave Creek and Hassyampa River models. Results show that for a given peak discharge there exists a nearly linear relationship between the storm length (Ls) and the storm velocity divided by the storm excess intensity (Vs / ie). Changes in this linear relationship are demonstrated for different values of peak flow, different watersheds, and different soil moisture conditions. An application of radar rainfall data as input to the CASC2D model is presented in which precipitation forecasts are generated by extrapolation of precipitation patterns from radar images. Forecast lead-time was increased for two events by 6 hours and 3 hours, respectively, through the inclusion of these precipitation forecasts. Finally, application of the parametric analysis results is demonstrated where peak flow estimates are determined based solely on radar observations of storm size, velocity and intensity.