Improvement of hail detection and nowcasting by synergistic combination of information from polarimetric radar, model predictions, and in-situ observations

Abstract

Large hail can pose a major threat to people, infrastructure, and property. While polarimetric radar observations enable us to reliably detect hail, discrimination of hail size and strategies to nowcast and predict hail and its size are still subject of research. At S band hail size discrimination algorithms do already exist, while retrievals from C-band radars, which are more widespread in Europe, remain challenging due to resonance scattering effects from large raindrops which tend to interfere with large hail signatures. <br /> Advanced nowcasting tools capable of detecting hail and discriminating its size at C band are required to enable earlier and more precise severe weather warnings being issued by weather services in Europe.

In this thesis 16 severe hail events monitored with the polarimetric C-band radar network of the German national meteorological service (<i>Deutscher Wetterdienst</i>, DWD) are investigated together with hail reports from the European Severe Weather Database (ESWD; https://www.eswd.eu/) for precursory information. This data set includes the largest hailstone (14.1 cm in diameter) reported so far in Germany hitting the ground in Undingen on August 6, 2013. <br /> First, an algorithm for correcting anomalously high attenuation in hail-bearing thunderstorms is evaluated using four overlapping C-band radars. Post-processing of polarimetric moments, like spike-filtering of differential phase, is done to reduce noise and improve attenuation corrections. Results illustrate the capability to mitigate attenuation of up to 15 db km^-1 but only small overcorrection in precipitation without hail. <br /> Second, T-matrix scattering simulations for dry and wet hail are conducted to adjust a fuzzy logic based hail size discrimination algorithm (HSDA) developed for S band for usage at C band. Hereby, dual-layered spheroids are simulated to mimic melting hail. The revised fuzzy logic utilizes an <i>ad-hoc</i> weighting of input variables to improve performance through self-adjustment similar to other unsupervised learning techniques. <br /> Finally, hail precursors and storm dynamics are investigated in an object-based, precipitation system oriented analysis to develop and evaluate nowcasting tools to detect hail growth and predict the hail's terminal diameter. This analysis utilizes a custom-made cell tracker and exploits the results of the attenuation correction algorithm. <br /> Results show that sudden increases of the vertical extent of columns of enhanced differential reflectivity Z_DR, so called Z_DR-columns, precede peaks of strong signal attenuation most probably linked to hail. Also, strong attenuation was often accompanied by specific differential phase K_DP above 10 ° km^-1. It is demonstrated that the intensity of Z_DR-columns, described e.g. by their maximum Z_DR value, and the height of Z_DR-columns above the melting layer, allow for nowcasting hail size at the ground with lead times between 10 and 20 minutes, which can be exploited for warnings.