High-resolution rainfall variability simulated by the WRF RCM: application to eastern France

The Weather Research and Forecasting (WRF) model, driven laterally by ERA-Interim reanalyses, is used here to downscale rainfall, at relatively high resolution (similar to 8 km) over Burgundy (eastern France), during the period 1989-2009. Regional simulations are compared to the M,t,o-France Station Network (MFSN; 127 daily rain-gauge records), at various temporal scales, including interannual variability, the annual cycle, and weather types. Results show that the spatial distribution of WRF-simulated rainfall climatology is consistent with MFSN observation data, but WRF tends to overestimate annual rainfall by similar to+15 %. At the interannual scale, WRF also performs very well (r similar to 0.8), despite almost constant, systematic overestimation. Only the average annual rainfall cycle is not accurately reproduced by WRF (r similar to 0.5), with rainfall overestimation in spring and summer, when convective rainfall prevails. During the winter season (October-March), when stratiform rainfall is prevalent, WRF performs better. Despite the biases for summertime convective events, these results suggest that high-resolution WRF simulations could successfully be used to document present and future climate variability at a regional scale. Nevertheless, because of overestimated convective rainfall, WRF-simulated rainfall should probably not be used directly to feed impact models, especially during the vegetative summer period.