A Machine Learning-Trained ICVS VIIRS Clear-Sky Mask Algorithm Applicable for Multiple Satellites

The monitoring of the Visible Infrared Imaging Radiometer Suite (VIIRS) radiometric Observation against the Community Radiative Transfer Model (CRTM) simulation (O-M) biases for the thermal emissive bands (TEBs) has been implemented in the NOAA Integrated Calibration/Validation System (ICVS) Long-Term Monitoring (LTM) ( www.star.nesdis.noaa.gov/icvs/) for VIIRS calibration and validation. The VIIRS clear-sky mask (CSM) is a critical parameter employed in O-M calculation to improve O-M accuracy. Commonly, the CSM is sensor independent and provided by the environment data records (EDR) team, which is not available during the sensor data records (SDR) calibration and validation period right after the sensor launch. Although a proxy CSM with similar sensor can be used to initially validate O-M biases, the O-M statistics are not good as expected. Thus, developing a fast but accurate VIIRS clear-sky flag become critical in O-M evaluation for the sensor calibration and validation. In this study, the machine learning (ML) method is employed to train and generate CSM in global ocean domain for VIIRS, and the NOAA advanced Clear-Sky Processor over Ocean (ACSPO) data is used as reference. Also, the ICVS O-M tool was further used to validate the retrieved ML CSM by comparing with ACSPO. The preliminary result showed that the O-M mean biases are comparable with ACSPO. The standard deviations are slightly smaller for all bands and the number of the clear-sky pixels even increase. Furthermore, using the ML-trained model with SNPP/VIIRS data to generate NOAA-20 CSM showed that the O-M statistics are also comparable with ACSPO for NOAA-20, which indicates that the ML model trained by SNPP/VIIRS has capability use for NOAA-20 and late JPSS satellites. The future work is to further validate the ML CSM algorithm for land conditions. It is expected that the well validated ML CSM algorithm could advance the accuracy of VIIRS O-M biases for NOAA-20 and future JPSS VIIRSs.