GL-ST: A Data-Driven Prediction Model for Sea Surface Temperature in the Coastal Waters of China Based on Interactive Fusion of Global and Local Spatiotemporal Information
The spatiotemporal multimodal variations in sea surface temperature refer to its diverse changes across different temporal and spatial scales. Understanding and predicting these variations are crucial for climate research and marine ecosystem conservation. Data-driven methods for sea surface tempera...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10804073/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841536156234153984 |
|---|---|
| author | Ning Song Jie Nie Qi Wen Yuchen Yuan Xiong Liu Jun Ma Zhiqiang Wei |
| author_facet | Ning Song Jie Nie Qi Wen Yuchen Yuan Xiong Liu Jun Ma Zhiqiang Wei |
| author_sort | Ning Song |
| collection | DOAJ |
| description | The spatiotemporal multimodal variations in sea surface temperature refer to its diverse changes across different temporal and spatial scales. Understanding and predicting these variations are crucial for climate research and marine ecosystem conservation. Data-driven methods for sea surface temperature prediction have made significant advancements in capturing these spatiotemporal multimodal variations. These data-driven techniques often utilize classic convolutional networks (CONV) and long short-term memory networks (LSTM) to extract spatial and temporal features. Leveraging these spatiotemporal multimodal features enhances the analysis of evolving patterns in sea surface temperature over time and space, thereby improving prediction accuracy. However, the separate processing of these two dimensions impedes the effective interaction and fusion of spatiotemporal features. In addition, both CONV and LSTM suffer from inadequate modeling of long-range dependencies, resulting in suboptimal prediction accuracy. To tackle these obstacles, we propose a global–local spatiotemporal information interactive fusion model. This model explicitly extracts temporal and spatial features, facilitating early interaction between temporal and spatial information and resolving the prior difficulty in achieving spatiotemporal consistency. This enables effective modeling of long-range dependencies, ultimately enhancing sea surface temperature prediction. We applied the model to forecast sea surface temperatures in the coastal waters of China (Bohai Sea, East China Sea, South China Sea) over durations of 1, 3, 7, 10, and 14 days. Furthermore, we introduced an iterative optimization scheme for predictions to enhance the model's stability in iterative forecasting. The findings demonstrate that, across various regional and temporal prediction scenarios, the model exhibits superior accuracy and iterative stability compared to existing methods. |
| format | Article |
| id | doaj-art-13242bbb7d2e4f229c7fb4e51bd8b1b6 |
| institution | Kabale University |
| issn | 1939-1404 2151-1535 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj-art-13242bbb7d2e4f229c7fb4e51bd8b1b62025-01-15T00:00:57ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01182959297410.1109/JSTARS.2024.351563810804073GL-ST: A Data-Driven Prediction Model for Sea Surface Temperature in the Coastal Waters of China Based on Interactive Fusion of Global and Local Spatiotemporal InformationNing Song0https://orcid.org/0000-0002-4209-7387Jie Nie1https://orcid.org/0000-0003-4952-7666Qi Wen2https://orcid.org/0009-0005-5395-7724Yuchen Yuan3Xiong Liu4Jun Ma5https://orcid.org/0009-0007-5822-6191Zhiqiang Wei6https://orcid.org/0000-0002-2830-8301College of Information Science and Engineering, Ocean University of China, Qingdao, ChinaCollege of Information Science and Engineering, Ocean University of China, Qingdao, ChinaCollege of Information Science and Engineering, Ocean University of China, Qingdao, ChinaCollege of Information Science and Engineering, Ocean University of China, Qingdao, ChinaCollege of Information Science and Engineering, Ocean University of China, Qingdao, ChinaCollege of Information Science and Engineering, Ocean University of China, Qingdao, ChinaCollege of Information Science and Engineering, Ocean University of China, Qingdao, ChinaThe spatiotemporal multimodal variations in sea surface temperature refer to its diverse changes across different temporal and spatial scales. Understanding and predicting these variations are crucial for climate research and marine ecosystem conservation. Data-driven methods for sea surface temperature prediction have made significant advancements in capturing these spatiotemporal multimodal variations. These data-driven techniques often utilize classic convolutional networks (CONV) and long short-term memory networks (LSTM) to extract spatial and temporal features. Leveraging these spatiotemporal multimodal features enhances the analysis of evolving patterns in sea surface temperature over time and space, thereby improving prediction accuracy. However, the separate processing of these two dimensions impedes the effective interaction and fusion of spatiotemporal features. In addition, both CONV and LSTM suffer from inadequate modeling of long-range dependencies, resulting in suboptimal prediction accuracy. To tackle these obstacles, we propose a global–local spatiotemporal information interactive fusion model. This model explicitly extracts temporal and spatial features, facilitating early interaction between temporal and spatial information and resolving the prior difficulty in achieving spatiotemporal consistency. This enables effective modeling of long-range dependencies, ultimately enhancing sea surface temperature prediction. We applied the model to forecast sea surface temperatures in the coastal waters of China (Bohai Sea, East China Sea, South China Sea) over durations of 1, 3, 7, 10, and 14 days. Furthermore, we introduced an iterative optimization scheme for predictions to enhance the model's stability in iterative forecasting. The findings demonstrate that, across various regional and temporal prediction scenarios, the model exhibits superior accuracy and iterative stability compared to existing methods.https://ieeexplore.ieee.org/document/10804073/Fusionglobal featureinteractionlocal featuresea surface temperature (SST)spatiotemporal information |
| spellingShingle | Ning Song Jie Nie Qi Wen Yuchen Yuan Xiong Liu Jun Ma Zhiqiang Wei GL-ST: A Data-Driven Prediction Model for Sea Surface Temperature in the Coastal Waters of China Based on Interactive Fusion of Global and Local Spatiotemporal Information IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Fusion global feature interaction local feature sea surface temperature (SST) spatiotemporal information |
| title | GL-ST: A Data-Driven Prediction Model for Sea Surface Temperature in the Coastal Waters of China Based on Interactive Fusion of Global and Local Spatiotemporal Information |
| title_full | GL-ST: A Data-Driven Prediction Model for Sea Surface Temperature in the Coastal Waters of China Based on Interactive Fusion of Global and Local Spatiotemporal Information |
| title_fullStr | GL-ST: A Data-Driven Prediction Model for Sea Surface Temperature in the Coastal Waters of China Based on Interactive Fusion of Global and Local Spatiotemporal Information |
| title_full_unstemmed | GL-ST: A Data-Driven Prediction Model for Sea Surface Temperature in the Coastal Waters of China Based on Interactive Fusion of Global and Local Spatiotemporal Information |
| title_short | GL-ST: A Data-Driven Prediction Model for Sea Surface Temperature in the Coastal Waters of China Based on Interactive Fusion of Global and Local Spatiotemporal Information |
| title_sort | gl st a data driven prediction model for sea surface temperature in the coastal waters of china based on interactive fusion of global and local spatiotemporal information |
| topic | Fusion global feature interaction local feature sea surface temperature (SST) spatiotemporal information |
| url | https://ieeexplore.ieee.org/document/10804073/ |
| work_keys_str_mv | AT ningsong glstadatadrivenpredictionmodelforseasurfacetemperatureinthecoastalwatersofchinabasedoninteractivefusionofglobalandlocalspatiotemporalinformation AT jienie glstadatadrivenpredictionmodelforseasurfacetemperatureinthecoastalwatersofchinabasedoninteractivefusionofglobalandlocalspatiotemporalinformation AT qiwen glstadatadrivenpredictionmodelforseasurfacetemperatureinthecoastalwatersofchinabasedoninteractivefusionofglobalandlocalspatiotemporalinformation AT yuchenyuan glstadatadrivenpredictionmodelforseasurfacetemperatureinthecoastalwatersofchinabasedoninteractivefusionofglobalandlocalspatiotemporalinformation AT xiongliu glstadatadrivenpredictionmodelforseasurfacetemperatureinthecoastalwatersofchinabasedoninteractivefusionofglobalandlocalspatiotemporalinformation AT junma glstadatadrivenpredictionmodelforseasurfacetemperatureinthecoastalwatersofchinabasedoninteractivefusionofglobalandlocalspatiotemporalinformation AT zhiqiangwei glstadatadrivenpredictionmodelforseasurfacetemperatureinthecoastalwatersofchinabasedoninteractivefusionofglobalandlocalspatiotemporalinformation |