Quantifying the spatiotemporal evolution and driving factors of soil erosion in the source region of the Yellow River using 137Cs and machine learning models
Jinxi Su, Zhenying Zhou, Juncheng Li, Mengyao Long, Huilong Lin
{"title":"Quantifying the spatiotemporal evolution and driving factors of soil erosion in the source region of the Yellow River using 137Cs and machine learning models","authors":"Jinxi Su, Zhenying Zhou, Juncheng Li, Mengyao Long, Huilong Lin","doi":"10.1016/j.catena.2025.109296","DOIUrl":null,"url":null,"abstract":"<div><div>The source region of the Yellow River (SRYR), known as the “Water Tower of China”, is not only crucial for the sustainability of alpine grassland and wetland ecosystems, but has significant implications for water quality and sediment deposition in the river’s middle and lower reaches. However, complex topography, climate, and diverse soil erosion types limit the applicability of traditional estimation methods. This study developed a tailored compound erosion prediction model for the SRYR to quantify erosion rates and identify their driving mechanisms. A dataset of 537 soil erosion rates from <sup>137</sup>Cs and 33 conditioning variables were used to construct an optimal model. Two variable selection methods—genetic algorithm (GA) and least absolute shrinkage and selection operator (LASSO)—were applied alongside three machine learning algorithms: categorical boosting (CatBoost), random forest (RF), and k-nearest neighbors (KNN). Geographical detectors were used to identify erosion drivers. Key findings include: (1) The GA-CatBoost model outperformed others (R<sub>test</sub><sup>2</sup> = 0.51). Based on optimal model, the estimated annual soil erosion rate (2001–2022) in the SRYR was 20.21 t·ha<sup>−1</sup>·a<sup>-1</sup>, with total annual erosion of 238.80 × 10<sup>6</sup> t·a<sup>-1</sup>. (2) Spatial analysis revealed high erosion in the northwest, low in the southeast, and mixed patterns in the central region, with 78.28 % of the SRYR exhibiting improvement over 2001–2022. (3) Precipitation and NDVI were identified as the dominant driving factors mitigating soil erosion in the SRYR. These findings demonstrate the effectiveness of ecological restoration efforts in the SRYR, providing empirical evidence for targeted soil and water conservation strategies. Future studies could enhance model accuracy by diversifying sampling, measuring erosion rates during shorter time windows and using higher resolution input data.</div></div>","PeriodicalId":9801,"journal":{"name":"Catena","volume":"259 ","pages":"Article 109296"},"PeriodicalIF":5.7000,"publicationDate":"2025-07-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Catena","FirstCategoryId":"97","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0341816225005983","RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"GEOSCIENCES, MULTIDISCIPLINARY","Score":null,"Total":0}
引用次数: 0
Abstract
The source region of the Yellow River (SRYR), known as the “Water Tower of China”, is not only crucial for the sustainability of alpine grassland and wetland ecosystems, but has significant implications for water quality and sediment deposition in the river’s middle and lower reaches. However, complex topography, climate, and diverse soil erosion types limit the applicability of traditional estimation methods. This study developed a tailored compound erosion prediction model for the SRYR to quantify erosion rates and identify their driving mechanisms. A dataset of 537 soil erosion rates from 137Cs and 33 conditioning variables were used to construct an optimal model. Two variable selection methods—genetic algorithm (GA) and least absolute shrinkage and selection operator (LASSO)—were applied alongside three machine learning algorithms: categorical boosting (CatBoost), random forest (RF), and k-nearest neighbors (KNN). Geographical detectors were used to identify erosion drivers. Key findings include: (1) The GA-CatBoost model outperformed others (Rtest2 = 0.51). Based on optimal model, the estimated annual soil erosion rate (2001–2022) in the SRYR was 20.21 t·ha−1·a-1, with total annual erosion of 238.80 × 106 t·a-1. (2) Spatial analysis revealed high erosion in the northwest, low in the southeast, and mixed patterns in the central region, with 78.28 % of the SRYR exhibiting improvement over 2001–2022. (3) Precipitation and NDVI were identified as the dominant driving factors mitigating soil erosion in the SRYR. These findings demonstrate the effectiveness of ecological restoration efforts in the SRYR, providing empirical evidence for targeted soil and water conservation strategies. Future studies could enhance model accuracy by diversifying sampling, measuring erosion rates during shorter time windows and using higher resolution input data.
期刊介绍:
Catena publishes papers describing original field and laboratory investigations and reviews on geoecology and landscape evolution with emphasis on interdisciplinary aspects of soil science, hydrology and geomorphology. It aims to disseminate new knowledge and foster better understanding of the physical environment, of evolutionary sequences that have resulted in past and current landscapes, and of the natural processes that are likely to determine the fate of our terrestrial environment.
Papers within any one of the above topics are welcome provided they are of sufficiently wide interest and relevance.