An improved approach to design precipitation estimation using radar data and rain-gauge measurements

Design precipitation estimates are crucial for hydrological modeling, flood risk assessment, and infrastructure planning. Traditional methods often rely on relatively sparse rain-gauge measurements or on precipitation derived from weather radar data with varying degrees of accuracy.

The presented study introduces a methodological approach to improving the accuracy and spatial consistency of design precipitation estimates from weather radar data by incorporating rain-gauge measurements, surpassing standard adjustment methods. This approach is based on a geostatistical merging of two design precipitation fields: one derived from high-resolution precipitation intensity data calculated from radar reflectivity and adjusted using direct precipitation measurements, and the other from long-term ombrographic measurements. It preserves relatively reliable estimates at gauge locations while maintaining spatial gradients derived from radar data and mitigates the relatively high uncertainty in the computation of precipitation intensity from radar reflectivity as well as the limited spatial representation of rain-gauge data. It is thus particularly relevant when long station records are available and cover the entire area of interest.

The approach is validated using data from a Central European country of Czechia. Before merging, the robustness of design precipitation estimates is enhanced by applying the L-moment-based index storm procedure and the region-of-influence method. Final design precipitation fields are derived by interpolating the ratios between rain-gauge and adjusted radar design precipitation totals with Empirical Bayesian Kriging. The identified spatial variability of design precipitation adequately reflects heavy precipitation formation mechanisms, causal circulation patterns, and orographic effects. This paves the way for future research focused on local adjustments of extreme value distribution parameters and interpolation settings, which may further mitigate the uncertainty of the estimates in certain areas.