Propagation of radar-rainfall uncertainty in runoff predictions

The primary advantage of radar observations of precipitation compared with traditional rain gauge measurements is their high spatial and temporal resolution and large areal coverage. Unfortunately, radar data require vigorous quality control before being converted into precipitation products that can be used as input to hydrologic models. In this study, a physically-based atmospheric model of convective rainfall is coupled with an active microwave radiative transfer model to simulate radar observation of thunderstorms. Radar observations of these storms are generated and used to evaluate the propagation of radar-rainfall errors through distributed hydrologic simulations. This physically-based methodology allows one to directly examine the impact of radar-rainfall estimation errors on land-surface hydrologic predictions and to avoid the limitations imposed by the use of rain gauge data. Results indicate that the geometry of the radar beam and coordinate transformations, due to radar-watershed-storm orientation, have an effect on radar-rainfall estimation and runoff prediction errors. In addition to uncertainty in the radar reflectivity vs. rainfall intensity relationship, there are significant range-dependent and orientation-related radar-rainfall estimation errors that should be quantified in terms of their impact on runoff predictions. Rigorous statistical analysis of the relationship between estimated rainfall errors and characteristics of the predicted hydrograph is conducted for thousands of simulated events. In addition to the influence of radar estimation error, the relationship between event magnitude and the prediction error and its propagation is studied. Furthermore, Bayesian inference is applied for estimating the hydrologic output driven by radar-estimated rainfall based on statistical analysis of radar-rainfall error propagation.