GSMaP Rainfall Correction through Optimized Power Law Model exploiting the Lightning-Rainfall Relationship

Rain is associated with various natural hazards, such as tropical cyclones, drought, and flooding, which pose significant threat to humans, animals, and agriculture. As a primary source of freshwater, rain is also important to different environmental and anthropogenic processes. Because of these, accurate information on rainfall is a global necessity. Satellites provide rainfall measurements at wide spatial coverage with consistent data availability. However, space-based rainfall estimates may not always be accurate; therefore, data validation and further algorithm improvements are still needed. In this study, we developed a correction scheme to improve the accuracy of satellite-based rainfall measurements by exploiting the inherent relationship between rainfall and lightning.

Severe weather is typically associated with heavy rain and occurrences of lightning. The temporal patterns of rainfall and lightning flashes usually follow similar trends seasonally and diurnally. Considering these, we first established the relation between collocated rainfall and lightning using ground-based data. Lightning flashes were collected from the Lightning Detection Network (LIDEN) operated by the Japan Meteorological Agency (JMA). Ground rainfall data were obtained from the CBAND radar network operated by the Ministry of Land, Infrastructure, transport and Tourism of Japan (MLIT). The radar data were gridded to 0.1° resolution for every hour to match the resolution of the Global Satellite Mapping of Precipitation (GSMaP), the satellite-based product of interest. Only radar and GSMaP pixels with collocated lightning events were considered in the analyses, using hourly data for the whole month of August 2019. Approximately 75% of the radar and GSMaP pixels covering Japan were used to establish the lightning-rainfall relation. The remaining 25% of the pixels, located in northern Japan, were used to validate the model's performance.

Results showed that the relationship between lightning flash rate and radar rainfall volume can be best described empirically by a power law. Based on this analysis, we applied a correction to the GSMaP rainfall by incorporating the established lightning-rainfall power law model. The power law correction model parameters were then optimized to minimize the root mean square error (RMSE) between radar and GSMaP rainfall volumes. The corrected GSMaP data exhibited similar patterns with lightning and radar during the study period. Validation results in northern Japan showed a 71.62% reduction in RMSE between radar data and corrected GSMaP data. A case study during a tropical cyclone from September 5-7, 2020 also demonstrated a 42.7% reduction in RMSE after incorporating lightning data into the GSMaP rainfall. The correlation between radar and GSMaP also improved from 0.89 to 0.97 after the correction. These results demonstrate the potential for enhancing the performance of GSMaP by incorporating lightning data during severe weather events. Future developments in this study include the integration of lightning data into the GSMaP algorithm using satellite-based lightning data from current and future space missions.