Xin Zheng , Sha Zhang , Shanshan Yang , Jiaojiao Huang , Xianye Meng , Jiahua Zhang , Yun Bai
{"title":"基于遥感和深度学习预测未来蒸散量","authors":"Xin Zheng , Sha Zhang , Shanshan Yang , Jiaojiao Huang , Xianye Meng , Jiahua Zhang , Yun Bai","doi":"10.1016/j.ejrh.2024.102023","DOIUrl":null,"url":null,"abstract":"<div><h3>Study region</h3><div>The watersheds of the four flux sites in the United States were selected as the study areas for this research.</div></div><div><h3>Study focus</h3><div>This study validates the efficiency of Convolutional Long Short-Term Memory Network (ConvLSTM) models for site-scale <span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span> prediction. We enhanced the ConvLSTM model by adding a Spatial Pyramid Pooling module (SPPM) and a Multi-head Self-Attention Module (MSA-Module), creating the Multi-head Self-Attention ConvLSTM (MSA-ConvLSTM) model, which we applied to predicting regional-scale actual evapotranspiration (<span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span>). This study aims to investigate whether the MSA-ConvLSTM model can enhance the accuracy of predicting regional-scale <span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span>, considering multiple feature variables. Furthermore, we evaluated different performance indicators, discussed possible reasons for errors in regional <span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span> prediction, and conducted sensitivity analysis of the model characteristics.</div></div><div><h3>New hydrological insights for the region</h3><div>The MSA-ConvLSTM model accurately predicts the future state of <span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span>. The average <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> was 0.81, which is 11.6 % and 5.5 % higher than those of the ConvLSTM and Self-Attention ConvLSTM (SA-ConvLSTM) models, respectively. The average RMSE is 11.94 mm/m, which is 21.5 % and 13.7 % lower than ConvLSTM and SA-ConvLSTM, respectively. The average MAE is 9.46 mm/m, which is 21.3 % and 13 % lower than ConvLSTM and SA-ConvLSTM, respectively. Incorporating of a multi-head self-attention module enhances the model’s capacity for comprehensive understanding of input data features. This improvement allows the model to better adapt to feature relationships at varying scales and angles, enhancing its representational capacity and enabling effective adaptation to complex environmental changes.</div></div>","PeriodicalId":48620,"journal":{"name":"Journal of Hydrology-Regional Studies","volume":"56 ","pages":"Article 102023"},"PeriodicalIF":4.7000,"publicationDate":"2024-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Predicting future evapotranspiration based on remote sensing and deep learning\",\"authors\":\"Xin Zheng , Sha Zhang , Shanshan Yang , Jiaojiao Huang , Xianye Meng , Jiahua Zhang , Yun Bai\",\"doi\":\"10.1016/j.ejrh.2024.102023\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><h3>Study region</h3><div>The watersheds of the four flux sites in the United States were selected as the study areas for this research.</div></div><div><h3>Study focus</h3><div>This study validates the efficiency of Convolutional Long Short-Term Memory Network (ConvLSTM) models for site-scale <span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span> prediction. We enhanced the ConvLSTM model by adding a Spatial Pyramid Pooling module (SPPM) and a Multi-head Self-Attention Module (MSA-Module), creating the Multi-head Self-Attention ConvLSTM (MSA-ConvLSTM) model, which we applied to predicting regional-scale actual evapotranspiration (<span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span>). This study aims to investigate whether the MSA-ConvLSTM model can enhance the accuracy of predicting regional-scale <span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span>, considering multiple feature variables. Furthermore, we evaluated different performance indicators, discussed possible reasons for errors in regional <span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span> prediction, and conducted sensitivity analysis of the model characteristics.</div></div><div><h3>New hydrological insights for the region</h3><div>The MSA-ConvLSTM model accurately predicts the future state of <span><math><msub><mrow><mi>ET</mi></mrow><mrow><mi>a</mi></mrow></msub></math></span>. The average <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> was 0.81, which is 11.6 % and 5.5 % higher than those of the ConvLSTM and Self-Attention ConvLSTM (SA-ConvLSTM) models, respectively. The average RMSE is 11.94 mm/m, which is 21.5 % and 13.7 % lower than ConvLSTM and SA-ConvLSTM, respectively. The average MAE is 9.46 mm/m, which is 21.3 % and 13 % lower than ConvLSTM and SA-ConvLSTM, respectively. Incorporating of a multi-head self-attention module enhances the model’s capacity for comprehensive understanding of input data features. This improvement allows the model to better adapt to feature relationships at varying scales and angles, enhancing its representational capacity and enabling effective adaptation to complex environmental changes.</div></div>\",\"PeriodicalId\":48620,\"journal\":{\"name\":\"Journal of Hydrology-Regional Studies\",\"volume\":\"56 \",\"pages\":\"Article 102023\"},\"PeriodicalIF\":4.7000,\"publicationDate\":\"2024-10-23\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Hydrology-Regional Studies\",\"FirstCategoryId\":\"89\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S2214581824003720\",\"RegionNum\":2,\"RegionCategory\":\"地球科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"WATER RESOURCES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Hydrology-Regional Studies","FirstCategoryId":"89","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2214581824003720","RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"WATER RESOURCES","Score":null,"Total":0}
Predicting future evapotranspiration based on remote sensing and deep learning
Study region
The watersheds of the four flux sites in the United States were selected as the study areas for this research.
Study focus
This study validates the efficiency of Convolutional Long Short-Term Memory Network (ConvLSTM) models for site-scale prediction. We enhanced the ConvLSTM model by adding a Spatial Pyramid Pooling module (SPPM) and a Multi-head Self-Attention Module (MSA-Module), creating the Multi-head Self-Attention ConvLSTM (MSA-ConvLSTM) model, which we applied to predicting regional-scale actual evapotranspiration (). This study aims to investigate whether the MSA-ConvLSTM model can enhance the accuracy of predicting regional-scale , considering multiple feature variables. Furthermore, we evaluated different performance indicators, discussed possible reasons for errors in regional prediction, and conducted sensitivity analysis of the model characteristics.
New hydrological insights for the region
The MSA-ConvLSTM model accurately predicts the future state of . The average was 0.81, which is 11.6 % and 5.5 % higher than those of the ConvLSTM and Self-Attention ConvLSTM (SA-ConvLSTM) models, respectively. The average RMSE is 11.94 mm/m, which is 21.5 % and 13.7 % lower than ConvLSTM and SA-ConvLSTM, respectively. The average MAE is 9.46 mm/m, which is 21.3 % and 13 % lower than ConvLSTM and SA-ConvLSTM, respectively. Incorporating of a multi-head self-attention module enhances the model’s capacity for comprehensive understanding of input data features. This improvement allows the model to better adapt to feature relationships at varying scales and angles, enhancing its representational capacity and enabling effective adaptation to complex environmental changes.
期刊介绍:
Journal of Hydrology: Regional Studies publishes original research papers enhancing the science of hydrology and aiming at region-specific problems, past and future conditions, analysis, review and solutions. The journal particularly welcomes research papers that deliver new insights into region-specific hydrological processes and responses to changing conditions, as well as contributions that incorporate interdisciplinarity and translational science.