Abstrakt
Thunderstorms and especially induced lightning discharges have still not been fully understood, although they are known to cause many casualties yearly worldwide. This study aims at filling the gap of knowledge by investigating the potential of phase and power of the co- and cross-channels of a vertical cloud radar to indicate lightning close to the radar site.
We performed statistical and correlation analyses of vertical profiles of phase and power spectra in the co- and the cross-channel for 38 days of thunderstorms producing lightning up to 20 km from the radar in 2018-2019. Specifically, we divided the dataset into "near" and "far" data according to the observed distance of lightning to the radar and analyzed it separately.
Although the results are quite initial given the limited number of "near" data, they clearly showed different structures of "near" and "far" data, thus confirming the potential of radar data to indicate lightning. Moreover, for the first time in this study the predictability of lightning using cloud radar quantities was evaluated.
We applied a Regression Tree Model to diagnose lightning and verified it using Receiver Operating Characteristic (ROC) and Critical Success Index (CSI). ROC provided surprisingly good results, while CSI was not that good but considering the very rare nature of lightning its values are high as well.