Estimating the groundwater table threshold for mitigating soil salinization in the Songnen Plain of China
Date
2025-03-24
Authors
Ding, Yiding
Lü, Haishen
Xu, Ligang
Jiang, Mingliang
Zhu, Yonghua
Cheng, Junxiang
Fan, Hongxiang
Su, Jianbin
Major Professor
Advisor
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Elsevier B.V.
Abstract
Study region: The Songnen Plain is a key part of China’s largest plain, situated in northeastern China.
Study focus: Soil salinization has become one of the largest ecological issues in the Songnen Plain, and regulating groundwater levels is a crucial strategy for mitigating it. To address this, a comprehensive framework is developed to estimate the groundwater table threshold for soil salinization by integrating field sampling, remote sensing big data, and machine learning models.
New hydrological insights for the region: The soil salinity inversion model, which utilizes a random forest algorithm, achieves the highest accuracy (R² = 0.75, d = 0.94, RPD = 2.05), outperforming SVM, LightGBM, and XGBoost algorithms. From 2020–2023, areas with mild salinization accounted for 9.3 % of the total area, while moderate salinization accounted for 3.2 %, severe salinization accounted for 4.0 %, and saline soil areas accounted for 0.6 %. A probabilistic model further identifies groundwater depth thresholds for salinization: 2.3 m for sandy soil, 3.1 m for loamy soil, and 1.1 m for silty soil. Based on current groundwater depths, it is anticipated that 15.5 % of the Songnen Plain area will continue to be affected by soil salinization or remain at risk of potential salinization.
Study focus: Soil salinization has become one of the largest ecological issues in the Songnen Plain, and regulating groundwater levels is a crucial strategy for mitigating it. To address this, a comprehensive framework is developed to estimate the groundwater table threshold for soil salinization by integrating field sampling, remote sensing big data, and machine learning models.
New hydrological insights for the region: The soil salinity inversion model, which utilizes a random forest algorithm, achieves the highest accuracy (R² = 0.75, d = 0.94, RPD = 2.05), outperforming SVM, LightGBM, and XGBoost algorithms. From 2020–2023, areas with mild salinization accounted for 9.3 % of the total area, while moderate salinization accounted for 3.2 %, severe salinization accounted for 4.0 %, and saline soil areas accounted for 0.6 %. A probabilistic model further identifies groundwater depth thresholds for salinization: 2.3 m for sandy soil, 3.1 m for loamy soil, and 1.1 m for silty soil. Based on current groundwater depths, it is anticipated that 15.5 % of the Songnen Plain area will continue to be affected by soil salinization or remain at risk of potential salinization.
Series Number
Journal Issue
Is Version Of
Versions
Series
Academic or Administrative Unit
Type
Article
Comments
This aritcle is published as Ding, Yiding, Haishen Lü, Ligang Xu, Robert Horton, Mingliang Jiang, Yonghua Zhu, Junxiang Cheng, Hongxiang Fan, and Jianbin Su. "Estimating the groundwater table threshold for mitigating soil salinization in the Songnen Plain of China." Journal of Hydrology: Regional Studies 59 (2025): 102326. doi:10.1016/j.ejrh.2025.102326
Rights Statement
© 2025 The Authors. This is an open access article under the CC BY-NC license
(http://creativecommons.org/licenses/by-nc/4.0/).
Copyright
Funding
This research was supported by the Strategic Priority Research Program of the Chinese Academy of Sciences, China (XDA28020501); NNSF National Natural Science Foundation of China, China (Grant Nos. 42471019, 42307106 and 42071033).