Radar rainfall estimation in the context of post-event analysis of flash-flood events
Résumé
A method to estimate rainfall from radar data for post-event analysis of flash-flood events has been developed within the EC-funded HYDRATE project. It follows a pragmatic approach including careful analysis of the observation conditions for the radar system(s) available for the considered case. Clutter and beam blockage are characterised by dry-weather observations and simulations based on a digital terrain model of the region of interest. The vertical profile of reflectivity (VPR) is either inferred from radar data if volume scanning data are available or simply defined using basic meteorological parameters (idealised VPR). Such information is then used to produce correction factor maps for each elevation angle to correct for range-dependent errors. In a second step, an effective Z-R relationship is optimised to remove the bias over the hit region. Due to limited data availability, the optimisation is carried out with reference to raingauge rain amounts measured at the event time scale. Sensitivity tests performed with two well-documented rain events show that a number of Z = aRb relationships, organised along hyperbolic curves in the (a and b) parameter space, lead to optimum assessment results in terms of the Nash coefficient between the radar and raingauge estimates. A refined analysis of these equifinality patterns shows that the "total additive conditional bias" can be used to discriminate between the Nash coefficient equifinal solutions. We observe that the optimisation results are sensitive to the VPR description and also that the Z-R optimisation procedure can largely compensate for range-dependent errors, although this shifts the optimal coefficients in the parameter space. The time-scale dependency of the equifinality patterns is significant, however near-optimal Z-R relationships can be obtained at all time scales from the event time step optimisation.