Improvement of Maximum Air Temperature Forecasts Using a Stacking Ensemble Technique
Due to the influence of complex factors such as atmospheric dynamic processes, physical processes and local topography and geomorphology, the prediction of near-surface meteorological elements in the numerical weather model often has deviation. The deep learning neural networks are more flexible but...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-03-01
|
| Series: | Atmosphere |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2073-4433/14/3/600 |
| _version_ | 1797613539732488192 |
|---|---|
| author | Linna Zhao Shu Lu Dan Qi |
| author_facet | Linna Zhao Shu Lu Dan Qi |
| author_sort | Linna Zhao |
| collection | DOAJ |
| description | Due to the influence of complex factors such as atmospheric dynamic processes, physical processes and local topography and geomorphology, the prediction of near-surface meteorological elements in the numerical weather model often has deviation. The deep learning neural networks are more flexible but with high variance. Here, we proposed a stacking ensemble model named FLT, which consists of a fully connected neural network with embedded layers (ED-FCNN), a long short-term memory (LSTM) network and a temporal convolutional network (TCN) to overcome the high variance of a single neural network and to improve prediction of maximum air temperature. The case study of daily maximum temperature forecast evaluated with observation of almost 2400 weather stations shows substantial improvement over that of single neural network model, ECMWF-IFS and statistical post-processing model. The FLT model can more effectively improve the forecast bias of the ECMWF-IFS model than that of any of the above single neural network model, with the RMSE reduced by 52.36% and the accuracy of temperature forecast increased by 43.12% compared with the ECMWF-IFS model. The average RMSEs of the FLT model decreases by 8.39%, 1.50%, 2.96% and 16.03%, respectively, compared with ED-FCNN, LSTM, TCN and the decaying average method. |
| first_indexed | 2024-03-11T06:56:11Z |
| format | Article |
| id | doaj.art-1a5560b0fcf84b4c915829a32b751593 |
| institution | Directory Open Access Journal |
| issn | 2073-4433 |
| language | English |
| last_indexed | 2024-03-11T06:56:11Z |
| publishDate | 2023-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Atmosphere |
| spelling | doaj.art-1a5560b0fcf84b4c915829a32b7515932023-11-17T09:34:01ZengMDPI AGAtmosphere2073-44332023-03-0114360010.3390/atmos14030600Improvement of Maximum Air Temperature Forecasts Using a Stacking Ensemble TechniqueLinna Zhao0Shu Lu1Dan Qi2State Key Laboratory of Severe Weather and Institute of Artificial Intelligence for Meteorology, Chinese Academy of Meteorological Sciences, Beijing 100081, ChinaHunan Meteorological Observatory, Changsha 410118, ChinaNational Meteorological Center, Beijing 100081, ChinaDue to the influence of complex factors such as atmospheric dynamic processes, physical processes and local topography and geomorphology, the prediction of near-surface meteorological elements in the numerical weather model often has deviation. The deep learning neural networks are more flexible but with high variance. Here, we proposed a stacking ensemble model named FLT, which consists of a fully connected neural network with embedded layers (ED-FCNN), a long short-term memory (LSTM) network and a temporal convolutional network (TCN) to overcome the high variance of a single neural network and to improve prediction of maximum air temperature. The case study of daily maximum temperature forecast evaluated with observation of almost 2400 weather stations shows substantial improvement over that of single neural network model, ECMWF-IFS and statistical post-processing model. The FLT model can more effectively improve the forecast bias of the ECMWF-IFS model than that of any of the above single neural network model, with the RMSE reduced by 52.36% and the accuracy of temperature forecast increased by 43.12% compared with the ECMWF-IFS model. The average RMSEs of the FLT model decreases by 8.39%, 1.50%, 2.96% and 16.03%, respectively, compared with ED-FCNN, LSTM, TCN and the decaying average method.https://www.mdpi.com/2073-4433/14/3/600deep learningstacking ensemblepost-processingbias correction |
| spellingShingle | Linna Zhao Shu Lu Dan Qi Improvement of Maximum Air Temperature Forecasts Using a Stacking Ensemble Technique Atmosphere deep learning stacking ensemble post-processing bias correction |
| title | Improvement of Maximum Air Temperature Forecasts Using a Stacking Ensemble Technique |
| title_full | Improvement of Maximum Air Temperature Forecasts Using a Stacking Ensemble Technique |
| title_fullStr | Improvement of Maximum Air Temperature Forecasts Using a Stacking Ensemble Technique |
| title_full_unstemmed | Improvement of Maximum Air Temperature Forecasts Using a Stacking Ensemble Technique |
| title_short | Improvement of Maximum Air Temperature Forecasts Using a Stacking Ensemble Technique |
| title_sort | improvement of maximum air temperature forecasts using a stacking ensemble technique |
| topic | deep learning stacking ensemble post-processing bias correction |
| url | https://www.mdpi.com/2073-4433/14/3/600 |
| work_keys_str_mv | AT linnazhao improvementofmaximumairtemperatureforecastsusingastackingensembletechnique AT shulu improvementofmaximumairtemperatureforecastsusingastackingensembletechnique AT danqi improvementofmaximumairtemperatureforecastsusingastackingensembletechnique |