An improvement of convective precipitation nowcasting through lightning data dynamic nudging in a cloud-resolving scale forecasting system

Abstract

A lightning data dynamic nudging (LDN) method was designed to adjust the dynamic field in convective clouds based on the physical relationship between lightning and vertical velocity. Under this framework, vertical velocity data retrieved from the Guangdong–Hongkong–Macau lightning location system were assimilated by a nudging approach using the Global/Regional Assimilation and Prediction System in mesoscale model (GRAPES-Meso). The effect of the assimilation on short-term (<6 h) precipitation forecast was evaluated by simulating the squall line event on 7 May and a continuous experiment from May to August 2018. Firstly, a series of sensitivity experiments was carried out to adjust the configuration of the assimilation system based on the squall line event. The results show that LDN expanded the spatial distribution of the positive vertical velocity (>1 m s−1) at 700 hPa by about 2%, and thus enlarged the spatial distribution of severe rainfall (>40 mm h−1). It should be noted that LDN did not significantly change the intensity of the updraft (i.e., the maximum vertical velocity) of the squall line relative to the simulation without assimilation. Overall, the equitable threat score (ETS) and fractions skill score (FSS) were increased by 0.04 and 0.13, respectively, and the positive effect lasted for 2–3 h after the data were assimilated. In view of the data assimilation frequency, using two successive nudging procedures with an interval of 12 min had the best effect on the forecast of severe precipitation. For the assimilation of both lightning and radar observations, the asynchronous assimilation of lightning and radar measurements performed slightly better than the synchronous assimilation. LDN improved the accuracy of severe precipitation forecast (>40 mm h−1 and > 20 mm h−1 for the squall line case and the continuous run, respectively), whereas assimilating radar data improved weak precipitation forecast (1–20 mm h−1 in the squall line case and 1–10 mm h−1 in the continuous run).