A high-resolution rainfall re-analysis based on radar-raingauge merging in the Cévennes-Vivarais region, France

- A 8-year radar-raingauge rainfall reanalysis in a region prone to heavy rainfall and flash-floods.
- KED radar-raingauge merging always better than observation systems separately.
- Clear added value of radar versus raingauges for short space-time scales.

SummaryThis work aims at providing quantitative precipitation estimates (QPEs) for the Cévennes-Vivarais region, France, over temporal (1-6 h) and spatial (1-300 km2) scales relevant for flash-flood prediction in that region. A systematic implementation of three estimation methods (radar QPE, hourly raingauge Ordinary Kriging - OK - and merging of radar and raingauge data through Kriging with External Drift - KED) proves the KED method to systematically outperform the concurrent approaches for the 131 main rain events selected during the period 2007-2014. Error models, assuming the standard deviation of the QPE error to be a bi-linear function of the rain rate and the kriging normalized estimation standard deviation, are parameterized for the KED and OK QPEs for the considered temporal and spatial scales. The error models are shown to depend on the type of rain event (Cévennes rain events, localized convection, widespread rainfall) and physical parameters such as the 0 °C isotherm altitude and the rain intermittency. The added-value of the radar network in terms of QPE with respect to the hourly raingauge network is larger for localized convection rain events as well as for the smallest space-time scales which are those of interest for flash-flood prediction in the region.