Soil salinization prediction through feature selection and machine learning at the irrigation district scale
IntroductionSoil salinization is a critical environmental issue affecting agricultural productivity worldwide, particularly in arid and semi-arid regions. This study focuses on the Xinjiang region of China, specifically the Xiao Haizi and Sha Jingzi irrigation areas, to explore the use of remote sen...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2025-01-01
|
| Series: | Frontiers in Earth Science |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/feart.2024.1488504/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841558652678307840 |
|---|---|
| author | Junbo Xie Junbo Xie Cong Shi Yang Liu Yang Liu Qi Wang Qi Wang Zhibo Zhong Zhibo Zhong Zhibo Zhong Shuai He Shuai He Shuai He Xingpeng Wang |
| author_facet | Junbo Xie Junbo Xie Cong Shi Yang Liu Yang Liu Qi Wang Qi Wang Zhibo Zhong Zhibo Zhong Zhibo Zhong Shuai He Shuai He Shuai He Xingpeng Wang |
| author_sort | Junbo Xie |
| collection | DOAJ |
| description | IntroductionSoil salinization is a critical environmental issue affecting agricultural productivity worldwide, particularly in arid and semi-arid regions. This study focuses on the Xinjiang region of China, specifically the Xiao Haizi and Sha Jingzi irrigation areas, to explore the use of remote sensing technology for surface soil salinity estimation.MethodsExhaustive and filter-based feature selection methods were employed by integrating soil salinity data measured on the ground with 32 spectral features derived from Landsat 8 OLI remote sensing images. A 5-fold cross-validation method was used to identify feature combinations that resulted in higher R2 values. Moreover, the inversion accuracy of soil salinization monitoring models built using different feature combinations was compared across five machine learning algorithms: Support Vector Machine (SVM), XGBoost, Decision Tree (DT), Random Forest (RF), and AdaBoost.ResultsThe results revealed that: (1) The AdaBoost and DT algorithms demonstrated high efficacy and precision in the prediction of soil salinity, with AdaBoost outperforming other algorithms in the validation set (R2 value of 0.892, MAE of 1.558, RMSE of 2.043), and DT showing the best performance in the training set (R2 value of 0.917, MAE of 0.838, RMSE of 1.182). (2) Feature combination 3, consisting of Salinity Index 5, Salinity Index 1, and Salinity Index 8, not only effectively extracted soil salinity information but also significantly improved the accuracy and efficiency of model estimations, effectively reflecting the actual situation of soil salinization in the irrigation area.DiscussionThis research provides robust methodological support for using remote sensing technology for soil salinity monitoring and management. |
| format | Article |
| id | doaj-art-af22f44cee9647e68f6326c5655ede25 |
| institution | Kabale University |
| issn | 2296-6463 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Earth Science |
| spelling | doaj-art-af22f44cee9647e68f6326c5655ede252025-01-06T06:59:42ZengFrontiers Media S.A.Frontiers in Earth Science2296-64632025-01-011210.3389/feart.2024.14885041488504Soil salinization prediction through feature selection and machine learning at the irrigation district scaleJunbo Xie0Junbo Xie1Cong Shi2Yang Liu3Yang Liu4Qi Wang5Qi Wang6Zhibo Zhong7Zhibo Zhong8Zhibo Zhong9Shuai He10Shuai He11Shuai He12Xingpeng Wang13Institute of Farmland Water Conservancy and Soil-Fertilizer, Xinjiang Academy of Agricultural and Reclamation Science, Shihezi, Xinjiang, ChinaCollege of Water Hydraulic and Architectural Engineering, Tarim University, Alar, Xinjiang, ChinaWestern Agricultural Research Center, Chinese Academy of Agricultural Sciences, Changji, Xinjiang, ChinaInstitute of Farmland Water Conservancy and Soil-Fertilizer, Xinjiang Academy of Agricultural and Reclamation Science, Shihezi, Xinjiang, ChinaCollege of Water Hydraulic and Architectural Engineering, Tarim University, Alar, Xinjiang, ChinaInstitute of Farmland Water Conservancy and Soil-Fertilizer, Xinjiang Academy of Agricultural and Reclamation Science, Shihezi, Xinjiang, ChinaCollege of Water Hydraulic and Architectural Engineering, Tarim University, Alar, Xinjiang, ChinaCollege of Water Hydraulic and Architectural Engineering, Tarim University, Alar, Xinjiang, ChinaKey Laboratory of Northwest Oasis Water-Saving Agriculture, Ministry of Agriculture and Rural Affairs, Shihezi, Xinjiang, ChinaXinjiang Production & Construction Corps Key Laboratory of Efficient Utilization of Water and Fertilizer, Shihezi, Xinjiang, ChinaCollege of Water Hydraulic and Architectural Engineering, Tarim University, Alar, Xinjiang, ChinaKey Laboratory of Northwest Oasis Water-Saving Agriculture, Ministry of Agriculture and Rural Affairs, Shihezi, Xinjiang, ChinaXinjiang Production & Construction Corps Key Laboratory of Efficient Utilization of Water and Fertilizer, Shihezi, Xinjiang, ChinaCollege of Water Hydraulic and Architectural Engineering, Tarim University, Alar, Xinjiang, ChinaIntroductionSoil salinization is a critical environmental issue affecting agricultural productivity worldwide, particularly in arid and semi-arid regions. This study focuses on the Xinjiang region of China, specifically the Xiao Haizi and Sha Jingzi irrigation areas, to explore the use of remote sensing technology for surface soil salinity estimation.MethodsExhaustive and filter-based feature selection methods were employed by integrating soil salinity data measured on the ground with 32 spectral features derived from Landsat 8 OLI remote sensing images. A 5-fold cross-validation method was used to identify feature combinations that resulted in higher R2 values. Moreover, the inversion accuracy of soil salinization monitoring models built using different feature combinations was compared across five machine learning algorithms: Support Vector Machine (SVM), XGBoost, Decision Tree (DT), Random Forest (RF), and AdaBoost.ResultsThe results revealed that: (1) The AdaBoost and DT algorithms demonstrated high efficacy and precision in the prediction of soil salinity, with AdaBoost outperforming other algorithms in the validation set (R2 value of 0.892, MAE of 1.558, RMSE of 2.043), and DT showing the best performance in the training set (R2 value of 0.917, MAE of 0.838, RMSE of 1.182). (2) Feature combination 3, consisting of Salinity Index 5, Salinity Index 1, and Salinity Index 8, not only effectively extracted soil salinity information but also significantly improved the accuracy and efficiency of model estimations, effectively reflecting the actual situation of soil salinization in the irrigation area.DiscussionThis research provides robust methodological support for using remote sensing technology for soil salinity monitoring and management.https://www.frontiersin.org/articles/10.3389/feart.2024.1488504/fullremote sensingLandsat 8agricultural sustainabilitysoil salinitymachine learningfeature selection |
| spellingShingle | Junbo Xie Junbo Xie Cong Shi Yang Liu Yang Liu Qi Wang Qi Wang Zhibo Zhong Zhibo Zhong Zhibo Zhong Shuai He Shuai He Shuai He Xingpeng Wang Soil salinization prediction through feature selection and machine learning at the irrigation district scale Frontiers in Earth Science remote sensing Landsat 8 agricultural sustainability soil salinity machine learning feature selection |
| title | Soil salinization prediction through feature selection and machine learning at the irrigation district scale |
| title_full | Soil salinization prediction through feature selection and machine learning at the irrigation district scale |
| title_fullStr | Soil salinization prediction through feature selection and machine learning at the irrigation district scale |
| title_full_unstemmed | Soil salinization prediction through feature selection and machine learning at the irrigation district scale |
| title_short | Soil salinization prediction through feature selection and machine learning at the irrigation district scale |
| title_sort | soil salinization prediction through feature selection and machine learning at the irrigation district scale |
| topic | remote sensing Landsat 8 agricultural sustainability soil salinity machine learning feature selection |
| url | https://www.frontiersin.org/articles/10.3389/feart.2024.1488504/full |
| work_keys_str_mv | AT junboxie soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT junboxie soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT congshi soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT yangliu soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT yangliu soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT qiwang soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT qiwang soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT zhibozhong soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT zhibozhong soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT zhibozhong soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT shuaihe soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT shuaihe soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT shuaihe soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale AT xingpengwang soilsalinizationpredictionthroughfeatureselectionandmachinelearningattheirrigationdistrictscale |