This study intercompares precipitation from gridded surface- (E-OBS, CARPATCLIM) and satellite-based (CMORPH, GPM IMERG, TRMM TMPA, PERSIANN-CDR, CHIRPS) observational datasets, the ERA5 reanalysis, and WRF regional climate simulations with grid spacings of 50 km and 10 km. The evaluation is carried out for the Carpathian region (south-east Central Europe), covering the 2001–2010 and 1985–2010 periods. Area-averaged mean precipitation is compared on the annual, monthly, daily, and hourly scales over a low-altitude, flat subregion within the Carpathian Basin. On the daily scale, ERA5 agrees the most with the ground-based datasets (E-OBS and CARPATCLIM), followed by the satellite-based products (in order: GPM IMERG, TRMM, CMORPH, CHIRPS, and PERSIANN-CDR), and the WRF simulations. E-OBS displays consistently lower precipitation than the reference CARPATCLIM and most other datasets. Precipitation is considerably underestimated by CMORPH in winter and overestimated by PERISANN-CDR throughout the year, questioning their direct applicability in the study region. Conversely, TRMM and CHIRPS are reasonable complements to the ground-based CARPATCLIM in climate monitoring and model verification over the target area. However, the performance of the different observation-based data products is heavily region-dependent. WRF over- and underestimates spring and summer precipitation, respectively, compared to all observation-based datasets. The grid spacing reduction from 50 km to 10 km improves simulated monthly mean precipitation during the period dominated by large-scale weather processes (October to April) and the course of the diurnal precipitation cycle over the Pannonian Plain. Relying on individual observational products in model verification can potentially lead to false bias characterization. The diurnal precipitation maximum occurs 2 and 4–5 h earlier in ERA5 and WRF, respectively, compared to GPM IMERG. However, discrepancies between the CMORPH and GPM IMERG diurnal cycles prevent a rigorous evaluation of sub-daily precipitation from the model-based datasets.
Read full abstract