Advances in the development of multi-sensor heavy rainfall statistics for Germany (KOSTRA-DWD-Hybrid): evaluation of radar-supported interpolation schemes

As heavy rainfall events pose increasing challenges to society, ranging from localised flash floods to widespread flooding that disrupts critical infrastructure and services, accurate and comprehensive analyses of their patterns play a pivotal role in enhancing our understanding of rainfall dynamics, assessing associated risks, and implementing effective mitigation strategies. Among the most practical outcomes of these analyses are depth-duration-frequency (DDF) curves, which serve as a key reference point for the planning and design of water management systems and facilities, such as dams, dikes, flood retention basins, and urban drainage networks. For Germany, DDF curves are provided by KOSTRA-DWD (German: Koordinierte Starkniederschlagsregionalisierung und -auswertung des Deutschen Wetterdienstes), a product that has been developed since the 1980s by the Department of Hydrometeorology at the Deutscher Wetterdienst (DWD).

In recent years, KOSTRA-DWD has undergone substantial revisions, incorporating additional rainfall data along with enhanced methods from extreme value statistics and geostatistics. The latest release, KOSTRA-DWD-2020^[1], provides DDF curves for 22 durations ranging from 5 minutes to 7 days and 9 return periods from 1 to 100 years, with a spatial resolution of 5 km. While KOSTRA-DWD-2020 relies on annual maximum series derived from 1,900 rain-gauge stations, this study focuses on the development of new multi-sensor heavy rainfall statistics that integrate rain-gauge measurements with high-resolution radar estimates.

To this end, we have evaluated a set of radar-supported interpolation schemes aimed at regionalising parameters of the Generalized Extreme Value distribution (μ, σ, ξ), as well as scaling parameters (θ, η) introduced by Koutsoyiannis et al. ^[2]. All interpolation schemes are based on the method of kriging with external drift, using rain-gauge measurements from 233 stations with long observational records (≥ 50 years) as the primary variable guiding spatial interpolation. Rain-gauge measurements from 1,667 stations with shorter records (≥ 10 years), along with radar estimates, in turn serve as external drift variables that refine the spatial structure of the interpolation. Two radar estimates produced by DWD with high temporal (5 min) and spatial (1 km) resolution have been examined: the first product, RY, consists of quality-controlled radar estimates corrected for beam shielding and processed with refined reflectivity-rainfall relationships, while the second product, YW, is derived from RY through additional quasi-calibration against rain-gauge measurements to reduce systematic bias. Our preliminary results indicate that integrating radar estimates improves the spatial representation of DDF curves, preserving the higher rainfall levels derived from rain-gauge measurements while enhancing spatial variability through radar estimates. However, discrepancies between the different interpolation schemes remain, underscoring the pressing need for a more detailed analysis to better understand their respective strengths and limitations.

^[1] https://www.dwd.de/DE/leistungen/kostra_dwd_rasterwerte/kostra_dwd_rasterwerte.html

^[2] Koutsoyiannis et al., 1998, A mathematical framework for studying rainfall intensity‑duration‑frequency relationships. J. Hydrol. (206), 118-135, https://doi.org/10.1016/S0022-1694(98)00097-3