Assimilation of Rainfall and Total Lightning Data for Nowcasting Torrential Rainfall During Summer Thunderstorms in Japan

Recently, detailed spatio-temporal analysis, using X-band multi-parameter radar-derived 3D volume scan and total lightning data in Japan, have revealed the peak in-cloud (IC) lightning occurs ~10 mins before maximum ground precipitation for individual cells of a summer thunderstorm (TS) producing torrential rain. This study investigates the potential of utilizing the total lightning data for monitoring and short-term prediction of torrential rain during three summer TS events causing heavy rainfall over Japan: an isolated TS, a TS possessing a merging of two cells, and a splitting TS cell. We construct simple linear regression models using (1) only ground precipitation volume (PV) and (2) a combination of ground PV and IC pulse rate. These models are continuously updated with the latest observations of IC and ground PV values to predict the one-step and multi-step ahead values of ground rainfall. We demonstrate a promising approach for short-term prediction of ground rainfall, by simultaneous application of the current and historical data of IC pulse rate and PV, which showed high accuracy (cross-correlation coefficient between observed and predicted PV was 0.84~0.94).