Improved precipitation and temperature over Western Europe due to convection-permitting scales and 3DVAR data assimilation

In recent years, using convection-permitting scales in regional climate simulations has become more and more frequent. These scales allow the models to include previously parameterized atmospheric processes, which improve the accuracy of a model in simulating precipitation. Data assimilation techniques can also improve the simulation of precipitation and temperature. However, the number of simulations combining these two options is scarce because of the high computational costs. Hence, it is important to evaluate the effect of data assimilation schemes on such convection-permitting simulations and to determine their added values.

In this study, we use the Weather Research and Forecasting model (WRF; version 3.8.1) to dynamically downscale the ERA5 reanalysis over Western Europe for period 2010–2020. We use one year of spin-up to allow the correct initialization of both soil and atmosphere. The spatial resolution of the simulation is set to 3 km and the temporal resolution to 1 hour. 51 vertical levels are employed, up to 20 hPa. Two model configurations are tested: with and without data assimilation (the Da and NoDA experiments, respectively). The former uses the 3DVAR data assimilation (WRFDA), which is run every six hours (00, 06, 12 and 18 UTC – analysis times). Observations are provided by the PREPBUFR dataset (NCEP ADP Global Upper Air and Surface Weather Observations), but only the observations inside a 120-minute window around analysis times are assimilated. Additionally, the DA experiment employs monthly-varying background error covariance matrices. Both experiments share the same parameterization options (e.g., Noah-MP land surface model) and use the daily-varying SST field from the NOAA OI SST v2 dataset instead of SST from ERA5.

The results show that both experiments generate monthly precipitation patterns over Europe which are similar to those from observational datasets such as IMERG and CHIRPS, or the reanalysis ERA5. However, in general, and particularly during summer months, DA generates larger amounts of precipitation than NoDA. These amounts are in line with those from CHIRPS. In terms of temperature, the DA experiment shows warmer temperatures than NoDA over central Europe during winter and colder temperatures over most of the domain in the remaining months. The temperatures of DA are again more in line with those from observational data sets such as CRU or EOBS. The datasets employed in the verification were not assimilated in WRFDA, and independent from each other.

These results highlight the fact that increasing the spatial resolution to reach convection-permitting scales has a positive impact on the simulation, allowing the generation of reliable precipitation and temperature fields. The use of 3DVAR data assimilation can additionally improve the performance of the regional model due to its effect on the positive feedbacks between soil moisture, air temperature, water vapour content and precipitation.