Ensemble radar precipitation estimation for hydrology in the Alps

A novel promising solution to express the complex error structure of radar precipitation estimates in a mountainous region is to generate an ensemble of radar precipitation fields. A prototype ensemble generator is currently being implemented and tested in a quasi-operational context in MAP D-PHASE, the Mesoscale Alpine Programme hydro-meteorological forecast demonstration project.

Each ensemble member is a possible realisation of the unknown true precipitation field given the radar reflectivity measurements and the detailed knowledge on the radar error structure. If we knew the radar error for a specific time and location we could simply correct the radar estimate and would not need an ensemble. The fact is we do not know the exact error at a specific time and location, but we know roughly the expected range of errors and how errors correlate in space and time. The idea of the ensemble is to express the uncertainty in a radar precipitation field in a manner that is coherent with our knowledge of the space-time variances and covariances of the radar errors. The original (deterministic) radar precipitation field is perturbed with a stochastic component, which has the correct space-time mean and covariance structure.

The ensemble members can be directly input into a hydrological model. Instead of running the model only once we run it several times. We thus get an ensemble of possible hydrological forecasts, the spread of which represents the sensitivity of the hydrologic system to the uncertainty in the radar precipitation field on input.