Complementing ERA5 and E-OBS with high-resolution river discharge over Europe


The 0.5° resolution of many global observational or quasi-observational datasets is not sufficient for the evaluation of current state-of-the-art regional climate models or the forcing of ocean model simulations over Europe. While higher resolved products are available for meteorological data, e.g. ERA5 reanalysis and the E-OBS vs 22 (EOBS22) datasets, they lack crucial information at the land-ocean boundary. ERA5 is frequently used to force regional climate models (RCMs) or ocean models and both datasets are commonly used as reference datasets for the evaluation of RCMs. Therefore, we extended both datasets with high-resolution river discharge for the period 1979–2018. On the one hand, our discharge data close the water cycle at the land-ocean interface so that the discharges can be used as lateral freshwater input for ocean models applied in the European region. On the other hand, the data can be used to identify trends in discharge that are induced by recent climate change as ERA5 and EOBS22 are rather independent datasets. The experimental setup to generate the discharges was chosen in a way that it could be easily adapted in a climate or Earth system modelling framework. Consequently, the recently developed 5 Min. horizontal resolution version of the hydrological discharge (HD) model was used to simulate discharge. It has already been applied in multiple climate modelling studies and is coupled within several global and regional Earth system models. As the HD model currently does not regard direct human impacts of the river runoff, it is well suited to investigate climate change-related discharge trends. In order to calculate the necessary gridded input fields for the HD model from ERA5 and EOBS22 data, we used the HydroPy global hydrological model. For both experiments, we found that the general behavior of discharge is captured well for many European rivers, which is consistent to earlier results. For the EOBS22 based discharges, a widespread low bias in simulated discharge occurs, which is likely caused by the missing undercatch correction in the underlying precipitation data. The analysis of trends over Southeastern Europe was hampered by missing data in EOBS22 after 2004. Using both experiments, we identified consistent trend patterns in various discharge statistics, with increases in low flow characteristics over Northern Europe and general drying trends over Central and Southern Europe. In summary, we introduced an experimental setup that is useful to generate high-resolution river runoff data consistent with the meteorological forcing for historical periods and future scenarios from any climate model data instead of having to rely on observed time series.
QR Code: Link to publication