Evaluation of precipitation reanalysis products in space and time for ungauged sites in Slovenia

Precipitation reanalysis products (PRP) are a promising data source for ungauged regions. Since observed time series are often i) too short, ii) their temporal resolution is not sufficient or iii) the network density is too low, they cannot be used as e.g. input for rainfall-runoff (r-r) modelling and derived flood frequency analysis. Reanalysis products as global simulation of the atmosphere over the last decades solve the aforementioned issues.

From the latest PRP three are most promising due to their spatial and temporal resolution for r-r modelling of small to mesoscale catchments: ERA5-Land (raster with approx. 9 km width), REA6 (6 km) and CFSv2 (22 km). These three PRP are able to cope with the dynamics of the r-r process due to their hourly resolution. The PRP are evaluated for Slovenia (Europe) with both, precipitation characteristics in space and time, and runoff characteristics. For areal precipitation, continuous and event-based characteristics are evaluated as well as precipitation extreme values. Simple correction methods for identified biases are suggested and applied. It can be seen that although the PRP clearly differ from each other, there is no clear ‘favourite’ to use as input for the r-r modelling.

To conclude about the suitability of the PRP for r-r modelling, continuous simulations are carried out with GR4H for 20 catchments in Slovenia (55 km²-480 km²). Models are re-calibrated for each PRP input based on KGE. Simulation results of calibration and validation period are evaluated by runoff extreme values, KGE, flow duration curve and intra-annual cycle. Interestingly, first results show that the deviations of some rainfall characteristics do not necessarily transfer to deviations in runoff characteristics, which can be explained by the high nonlinearity of the r-r process. PRP lead to better, at least similar results for runoff characteristics for catchments without rain gauges in their centre.