23AI The Retrieval of Surface Pressure and Wind Speed over Ocean using NOAA-20 ATMS Measurements through a U-Net Machine Learning Algorithm

In this presentation, surface pressure and wind speed were retrieved over ocean for tropical cyclones and in the neighboring area of each tropical cyclone using the satellite microwave measurements. For tropical cyclone location, International Best Track Archive for Climate Stewardship (IBTrACS) data have been used over North Atlantic Ocean and West Pacific Ocean between 2018 and 2021. We also used the National Oceanic and Atmospheric Administration (NOAA)-20 Advanced Technology Microwave Sounder (ATMS) reprocessed Sensor Data Record (SDR) brightness temperatures and geolocation information, and the hourly European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) surface pressure and wind speed in quarter degree interval of latitude and longitude. For each tropical cyclone case in IBTrACS, ATMS granules were combined to have 96 FOVs * 96 Scanlines (referred to as an image) for each. More than 2000 ATMS images have been generated for the time periods and regions studied, but only 266 images were selected that contain pure ocean cases. ERA5 surface pressure and wind speed values were collocated to the ATMS geolocations. Therefore, an image includes 22 channel ATMS brightness temperatures, latitude, longitude, viewing zenith angle and the collocated ERA5 surface pressure and wind speed. A U-Net machine learning algorithm was developed and applied to the 266 ATMS images to predict surface pressure and wind speed. Of the total number, 239 images were used for the training and 27 images were used for independent validation.

Due to inherent changes in along scan spatial resolution and mixed H and V polarizations, it is a challenge for cross-track scanning satellite microwave instruments to retrieve surface pressure and wind speed with an accuracy comparable to conical scanning satellite microwave instruments. Nevertheless, preliminary results indicate that near-TC estimates do exhibit some skill and that estimates for more general over ocean conditions are quite skillful. Further work is underway and will be presented at the meeting.