Radar-based extreme rainfall analysis for small areas

Warnings of extreme precipitation averaged over a small area and during a short duration are needed by people and the public services to ensure their safety. For these localized extreme events, accurate statistics can be derived from radar-based precipitation estimation using a regional analysis. This method has already been applied for rainfall averaged over a square of 1 km by Goudenhoofdt, Delobbe and Willems (2017). The extreme statistics can then be combined in realtime with precipitation estimation or nowcasting to provide relevant warnings. For this purpose web and smartphone applications can be developed to let the user define areas, durations and return periods of interest. Additional parameters could be added in the framework of a probabilistic approach.

In the first step of this project, we need to perform the extreme value analysis. The quality of a 10-year radar rainfall product for extreme events will be analyzed by comparison with independent rain gauge measurements. The principles of the regional frequency analysis will be presented together with the underlying hypothesis on the rainfall process. The statistical method will be tested using rectangles with different length, width and orientation. This will allow to analyze the impact of these parameters on the extreme precipitation statistics. Eventually the possibility to extent the method to any given small area will be discussed.