Research on Typhoon Prediction by Integrating Numerical Simulation and Deep Learning Methods
Typhoons rank among the most destructive natural disasters, significantly affecting human activities and daily life. Atmospheric numerical model wind fields, which are widely utilized, often underestimate typhoon intensity. This study proposes a model for predicting typhoon maximum wind speeds using...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Atmosphere |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2073-4433/16/1/111 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832589123864494080 |
|---|---|
| author | Tianyi Lv Huaming Yu Liangshi Lin Yijun Tao Xin Qi |
| author_facet | Tianyi Lv Huaming Yu Liangshi Lin Yijun Tao Xin Qi |
| author_sort | Tianyi Lv |
| collection | DOAJ |
| description | Typhoons rank among the most destructive natural disasters, significantly affecting human activities and daily life. Atmospheric numerical model wind fields, which are widely utilized, often underestimate typhoon intensity. This study proposes a model for predicting typhoon maximum wind speeds using the Long Short-Term Memory (LSTM) neural network. The model predicts maximum wind speeds based on existing atmospheric numerical forecasts, constructs a parametric wind field model from these predictions, and integrates it with the numerical model wind fields to generate an LSTM-optimized wind field. The results show that the LSTM model accurately predicts typhoon maximum wind speeds, with the predicted extreme values closely aligning with actual observations and capturing the trends of maximum wind speed variations. Compared with the ERA5 typhoon maximum wind speed, the C of the LSTM model for predicting the typhoon maximum wind speed is improved from 0.801 to 0.859, and the RMSE and MAE are reduced by 58% and 64%, respectively. In the simulation of Typhoon DELTA (2020), the LSTM-optimized wind field exhibits substantially higher wind speed intensities in the central region of the typhoon compared to the ERA5 wind field, providing a more accurate representation of the intensity and structure of the typhoon. |
| format | Article |
| id | doaj-art-7630f808966a4d468a66ef7cfd906596 |
| institution | Kabale University |
| issn | 2073-4433 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Atmosphere |
| spelling | doaj-art-7630f808966a4d468a66ef7cfd9065962025-01-24T13:22:04ZengMDPI AGAtmosphere2073-44332025-01-0116111110.3390/atmos16010111Research on Typhoon Prediction by Integrating Numerical Simulation and Deep Learning MethodsTianyi Lv0Huaming Yu1Liangshi Lin2Yijun Tao3Xin Qi4College of Oceanic and Atmospheric Sciences, Ocean University of China, Qingdao 266100, ChinaCollege of Oceanic and Atmospheric Sciences, Ocean University of China, Qingdao 266100, ChinaWenzhou Marine Center, Ministry of Natural Resources, Wenzhou 325011, ChinaNational Marine Data and Information Service, Ministry of Natural Resources, Tianjin 300171, ChinaManagement College, Ocean University of China, Qingdao 266100, ChinaTyphoons rank among the most destructive natural disasters, significantly affecting human activities and daily life. Atmospheric numerical model wind fields, which are widely utilized, often underestimate typhoon intensity. This study proposes a model for predicting typhoon maximum wind speeds using the Long Short-Term Memory (LSTM) neural network. The model predicts maximum wind speeds based on existing atmospheric numerical forecasts, constructs a parametric wind field model from these predictions, and integrates it with the numerical model wind fields to generate an LSTM-optimized wind field. The results show that the LSTM model accurately predicts typhoon maximum wind speeds, with the predicted extreme values closely aligning with actual observations and capturing the trends of maximum wind speed variations. Compared with the ERA5 typhoon maximum wind speed, the C of the LSTM model for predicting the typhoon maximum wind speed is improved from 0.801 to 0.859, and the RMSE and MAE are reduced by 58% and 64%, respectively. In the simulation of Typhoon DELTA (2020), the LSTM-optimized wind field exhibits substantially higher wind speed intensities in the central region of the typhoon compared to the ERA5 wind field, providing a more accurate representation of the intensity and structure of the typhoon.https://www.mdpi.com/2073-4433/16/1/111typhoontyphoon maximum wind speedslong short-term memory neural networkparametric model wind fields |
| spellingShingle | Tianyi Lv Huaming Yu Liangshi Lin Yijun Tao Xin Qi Research on Typhoon Prediction by Integrating Numerical Simulation and Deep Learning Methods Atmosphere typhoon typhoon maximum wind speeds long short-term memory neural network parametric model wind fields |
| title | Research on Typhoon Prediction by Integrating Numerical Simulation and Deep Learning Methods |
| title_full | Research on Typhoon Prediction by Integrating Numerical Simulation and Deep Learning Methods |
| title_fullStr | Research on Typhoon Prediction by Integrating Numerical Simulation and Deep Learning Methods |
| title_full_unstemmed | Research on Typhoon Prediction by Integrating Numerical Simulation and Deep Learning Methods |
| title_short | Research on Typhoon Prediction by Integrating Numerical Simulation and Deep Learning Methods |
| title_sort | research on typhoon prediction by integrating numerical simulation and deep learning methods |
| topic | typhoon typhoon maximum wind speeds long short-term memory neural network parametric model wind fields |
| url | https://www.mdpi.com/2073-4433/16/1/111 |
| work_keys_str_mv | AT tianyilv researchontyphoonpredictionbyintegratingnumericalsimulationanddeeplearningmethods AT huamingyu researchontyphoonpredictionbyintegratingnumericalsimulationanddeeplearningmethods AT liangshilin researchontyphoonpredictionbyintegratingnumericalsimulationanddeeplearningmethods AT yijuntao researchontyphoonpredictionbyintegratingnumericalsimulationanddeeplearningmethods AT xinqi researchontyphoonpredictionbyintegratingnumericalsimulationanddeeplearningmethods |