Precipitation identification and quantification using ABI and GLM observations aboard the GOES-R series: A machine learning perspective

Compared to the legacy GOES Imager, the Advanced Baseline Imager (ABI) aboard the GOES-R series provides measurements at much higher resolution in both spatial and temporal dimensions, which benefits many applications such as quantitative precipitation estimation. However, it is still a challenge to quantify the water content in the cloud and near-surface precipitation intensity using parametric relations based on the Infrared (IR) radiation and converted brightness temperature. Comparison between ground-based rainfall measurements and the operational GOES-16 (formerly known as GOES-R) rain rate products indicates that the uncertainty in GOES-16 rain retrievals is significant, and the uncertainty varies in different geophysical locations. This study designs a two-stage deep learning framework for precipitation identification and quantification using measurements from the ABI and Geostationary Lightning Mapper (GLM). In particular, this framework consists of a classification module for precipitation detection and a regression module for precipitation quantification (if precipitation is detected). Cloud-top brightness temperature observed by ABI channels 8/10/11/14/15 as well as lighting information from the GLM are used as input to this deep learning framework. The rainfall estimates from a ground radar network are used as target labels in the training phase. Independent verification demonstrates that this two-stage learning technique can estimate precipitation time, intensity, and amount very well, especially in heavy rain regions. In addition, the inclusion of GLM data can not only better predict the location of heavy precipitation in the detection module, but also improve the accuracy of precipitation intensity estimation in the regression module.