Improving ATMS Imagery Visualization Using Limb Correction and AI Resolution Enhancement

Abstract

The advanced technology microwave sounder (ATMS) is an important satellite instrument that provides vital data on atmosphere temperature and moisture for weather forecasting and climate research, and helps us plan for extreme weather. However, its coarse resolution and angular dependence have long been a challenge for improving image visualization. This article proposes a method to enhance the imagery visualization for ATMS, combining limb correction (LC) with artificial intelligence (AI) resolution enhancement (RE). Measurement data from the ATMS onboard NOAA-20 were utilized to train the LC method, which were then validated using newly acquired NOAA-21 ATMS data. The AI RE was performed using enhanced super-resolution generative adversarial networks, which increased the pixel resolution by a factor of four. The high-resolution (HR) Advanced Microwave Scanning Radiometer 2 data served as a reference to initially and quantitatively evaluate the RE method. The combined method of LC and AI RE produced an angular-dependence-free and HR ATMS image, resulting in a significant improvement in image visualization, including surface and atmosphere information, and allows for clear identification of severe weather events. For the swift identification and analysis of tropical cyclones in the upcoming season, as of this writing, this proposed method has been routinely employed to produce high-quality global ATMS images, and these images are showcased and tested in the NOAA internal HR imagery visualization system—JSTAR Mapper. Moreover, concentrated efforts are being made to further enhance these images in preparation for an official release.

Keywords

• Advanced technology microwave sounder (ATMS)
• convolutional neural network (CNN)
• enhanced super-resolution generative adversarial networks (ESRGAN)
• generative adversarial network (GAN)
• image visualization
• limb correction (LC)