Weather Radar Echo Extrapolation with Dynamic Weight Loss
Precipitation nowcasting is an important tool for economic and social services, especially for forecasting severe weather. The crucial and challenging part of radar echo image prediction is the focus of radar-based precipitation nowcasting. Recently, a number of deep learning models have been design...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-06-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/15/12/3138 |
| _version_ | 1797592808130871296 |
|---|---|
| author | Yonghong Zhang Sutong Geng Wei Tian Guangyi Ma Huajun Zhao Donglin Xie Huanyu Lu Kenny Thiam Choy Lim Kam Sian |
| author_facet | Yonghong Zhang Sutong Geng Wei Tian Guangyi Ma Huajun Zhao Donglin Xie Huanyu Lu Kenny Thiam Choy Lim Kam Sian |
| author_sort | Yonghong Zhang |
| collection | DOAJ |
| description | Precipitation nowcasting is an important tool for economic and social services, especially for forecasting severe weather. The crucial and challenging part of radar echo image prediction is the focus of radar-based precipitation nowcasting. Recently, a number of deep learning models have been designed to solve the problem of extrapolating radar images. Although these methods can generate better results than traditional extrapolation methods, the issue of error accumulation in precipitation forecasting is exacerbated by using only the mean square error (MSE) and mean absolute error (MAE) as loss functions. In this paper, we approach the problem from the perspective of the loss function and propose dynamic weight loss (DWL), a simple but effective loss function for radar echo extrapolation. The method adds model self-adjusted dynamic weights to the weighted loss function and structural similarity index measures. Radar echo extrapolation experiments are performed on four models, ConvLSTM, ConvGRU, PredRNN, and PredRNN++. Radar reflectivity is predicted using Nanjing University C-band Polarimetric (NJU-CPOL) weather radar data. The quantitative statistics show that using the DWL method reduces the MAE of the four models by up to 10.61%, 5.31%, 14.8%, and 13.63%, respectively, over a 1 h prediction period. The results show that the DWL approach is effective in reducing the accumulation of errors over time, improving the predictive performance of currently popular deep learning models. |
| first_indexed | 2024-03-11T01:58:28Z |
| format | Article |
| id | doaj.art-e56a033fd6964faa88ca2207534a24ab |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-11T01:58:28Z |
| publishDate | 2023-06-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-e56a033fd6964faa88ca2207534a24ab2023-11-18T12:27:00ZengMDPI AGRemote Sensing2072-42922023-06-011512313810.3390/rs15123138Weather Radar Echo Extrapolation with Dynamic Weight LossYonghong Zhang0Sutong Geng1Wei Tian2Guangyi Ma3Huajun Zhao4Donglin Xie5Huanyu Lu6Kenny Thiam Choy Lim Kam Sian7School of Automation, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaSchool of Automation, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaSchool of Computer Science, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaSchool of Electronics and Information Engineering, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaSchool of Automation, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaSchool of Automation, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaSchool of Automation, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaSchool of Atmospheric Science and Remote Sensing, Wuxi University, Wuxi 214105, ChinaPrecipitation nowcasting is an important tool for economic and social services, especially for forecasting severe weather. The crucial and challenging part of radar echo image prediction is the focus of radar-based precipitation nowcasting. Recently, a number of deep learning models have been designed to solve the problem of extrapolating radar images. Although these methods can generate better results than traditional extrapolation methods, the issue of error accumulation in precipitation forecasting is exacerbated by using only the mean square error (MSE) and mean absolute error (MAE) as loss functions. In this paper, we approach the problem from the perspective of the loss function and propose dynamic weight loss (DWL), a simple but effective loss function for radar echo extrapolation. The method adds model self-adjusted dynamic weights to the weighted loss function and structural similarity index measures. Radar echo extrapolation experiments are performed on four models, ConvLSTM, ConvGRU, PredRNN, and PredRNN++. Radar reflectivity is predicted using Nanjing University C-band Polarimetric (NJU-CPOL) weather radar data. The quantitative statistics show that using the DWL method reduces the MAE of the four models by up to 10.61%, 5.31%, 14.8%, and 13.63%, respectively, over a 1 h prediction period. The results show that the DWL approach is effective in reducing the accumulation of errors over time, improving the predictive performance of currently popular deep learning models.https://www.mdpi.com/2072-4292/15/12/3138precipitation nowcastingradar echo imagedynamic weight lossdeep learning |
| spellingShingle | Yonghong Zhang Sutong Geng Wei Tian Guangyi Ma Huajun Zhao Donglin Xie Huanyu Lu Kenny Thiam Choy Lim Kam Sian Weather Radar Echo Extrapolation with Dynamic Weight Loss Remote Sensing precipitation nowcasting radar echo image dynamic weight loss deep learning |
| title | Weather Radar Echo Extrapolation with Dynamic Weight Loss |
| title_full | Weather Radar Echo Extrapolation with Dynamic Weight Loss |
| title_fullStr | Weather Radar Echo Extrapolation with Dynamic Weight Loss |
| title_full_unstemmed | Weather Radar Echo Extrapolation with Dynamic Weight Loss |
| title_short | Weather Radar Echo Extrapolation with Dynamic Weight Loss |
| title_sort | weather radar echo extrapolation with dynamic weight loss |
| topic | precipitation nowcasting radar echo image dynamic weight loss deep learning |
| url | https://www.mdpi.com/2072-4292/15/12/3138 |
| work_keys_str_mv | AT yonghongzhang weatherradarechoextrapolationwithdynamicweightloss AT sutonggeng weatherradarechoextrapolationwithdynamicweightloss AT weitian weatherradarechoextrapolationwithdynamicweightloss AT guangyima weatherradarechoextrapolationwithdynamicweightloss AT huajunzhao weatherradarechoextrapolationwithdynamicweightloss AT donglinxie weatherradarechoextrapolationwithdynamicweightloss AT huanyulu weatherradarechoextrapolationwithdynamicweightloss AT kennythiamchoylimkamsian weatherradarechoextrapolationwithdynamicweightloss |