Improving Radar Reflectivity Reconstruction with Himawari-9 and UNet++ for Off-Shore Weather Monitoring

• Published in: Remote sensing (Basel, Switzerland) Vol. 16; no. 1; p. 56
• Main Authors: Wan, Bingcheng, Gao, Chloe Yuchao
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.01.2024
• Subjects: Brightness temperature; Comparative analysis; Datasets; Deep learning; Diurnal; Diurnal variations; Himawari-9; Meteorological satellites; Meteorological stations, Radar; Precipitation; Radar; radar composite reflectivity; Radar data; Radar systems; Rain; Reflectance; Regions; Remote monitoring; Root-mean-square errors; Severe weather; UNet; Weather; Weather forecasting; China
• ISSN: 2072-4292; 2072-4292
• DOI: 10.3390/rs16010056

Weather radars play a crucial role in the monitoring of severe convective weather. However, due to their limited detection range, they cannot conduct an effective monitoring in remote offshore areas. Therefore, this paper utilized UNet++ to establish a model for retrieving radar composite reflectivity based on Himawari-9 satellite datasets. In the process of comparative analysis, we found that both satellite and radar data exhibited significant diurnal cycles, but there were notable differences in their variation characteristics. To address this, we established four comparative models to test the influence of latitude and diurnal cycles on the inversion results. The results showed that adding the distribution map of the minimum brightness temperature at the corresponding time in the model could effectively improve the model’s performance in both spatial and temporal dimensions, reduce the root-mean-square error (RMSE) of the model, and enhance the accuracy of severe convective weather monitoring.