ML-driven cross-calibration of microwave and infrared instruments

Microwave (MW) and infrared (IR) observations are fundamental for weather forecasting, climate monitoring, and atmospheric sounding. Traditionally, inter-calibration efforts have focused on harmonizing measurements between similar instruments or within the same spectral domain. In this study, we explore the use of machine learning (ML) to directly cross-calibrate MW and IR observations, leveraging data from the Cross-track Infrared Sounder (CrIS) and the Advanced Technology Microwave Sounder (ATMS) onboard the NOAA-20 (N20) and Suomi-NPP (NPP) satellites. We demonstrate that an ML-model trained using CrIS observations as input (2211 channels) can predict the ATMS observations (22 channels) with a bias less than 0.1 K and a the coefficients of determination generally greater than 90%. Our results highlight the potential of ML-based cross-calibration approaches for bridging instruments across different spectral regions, provided that the input observations contain all necessary information to reconstruct the target observations. The proposed methodology offers a path forward for the cross-calibration of instruments operating in different spectral domains.