Improving GSMaP Precipitation Estimates Using a Modified Kalman Filter With Lightning Flash Rate

Precipitation drives various Earth’s processes and is linked to hazards like flash floods and drought, making timely and accurate monitoring of precipitation crucial for disaster mitigation. The Global Satellite Mapping of Precipitation (GSMaP) provides high-resolution rainfall data by combining microwave radiometer (MWR) and infrared (IR) retrievals. Since MWR retrievals are limited and only available every three hours, GSMaP’s near-real-time (NRT) product propagates MWR measurements forward in time and refines rainfall rates based on IR data through a Kalman filter. GSMaP estimates rain at high spatiotemporal resolutions, but is still susceptible to errors due to IR only seeing the cloud top instead of detecting actual precipitation. This study enhances the precipitation estimates of GSMaP NRT by integrating lightning as additional information in the Kalman filter, generating a new product called GSMaP NRTL. The approach is to use specific measurement models in the Kalman filter based on the relationship of rainfall with IR brightness temperature and lightning, classified into two thunderstorm types. Type I mostly incorporates mature stage thunderstorms, while Type II includes initiating and dissipating thunderstorms. Validation results show that NRTL improves on the performance of NRT, particularly reducing errors and biases while increasing the correlation and rainfall detection. However, underestimation persists in moderate to heavy rain (> 10 mmh-1). Kling-Gupta Efficiency (KGE) scores show that NRTL achieves better performance at finer resolutions than NRT usually does, even after aggregation. KGE also shows that NRTL performs best in Type I thunderstorms, especially over land, while improvements in Type II indicate better representation of usually underestimated rainfall in this type.