Soil Erosion Of Alpine Grassland Ecosystem On Tibetan Plateau Sanjiangyuan Region

Due to its particular geographical location, the Sanjiangyuan region, located in the hinterland of the Qinghai-Tibet Plateau, has its unique alpine grassland ecosystem. This region is China’s major water supply source, thus having extreme ecological importance in terms of water supply conservation. It also has a rich variety of grassland animals and plants. However, constant land erosion in this region has given rise to a series of problems, such as worsening condition of the grassland, desertification, and “black soil patch”. Thus, it is of enormous significance to study the methods to monitor and assess land erosion in this region. In this study, we took GIS as our technological platform, used the RUSLE Model for reference, and adopted several machine learning methods. Based on the 137 Cs erosion modulu, we built a self-defined erosion model, and carried out a comparative analysis of their accuracy. We acquired relatively accurate estimated volume of erosion in the research area. Meanwhile, we analyzed the respective methods for parameter selection and modeling and the result of the RUSLE Model. At last, based on the erosion results, we came up with strategies to prevent and control erosion. Major findings of this study include:1) Adopting several machine learning-based parameter selection and modeling methods for the first time, we acquired relatively accurate volume of erosion and distribution map of the erosion modulu in the Sanjiangyuan region. The order of accuracy for each algorithm model is:Generalized Linear Model > Cubist Model > Random Forest Model > Boosting Tree Model. The order of accuracy for each method of parameter selection is: Genetic Algorithm(GA) > Recursive Features Elimination(RFE) > Univariate Filters(UF) > Simulated Annealing(SA)> Boruta All-relevant Selection(BORUTA). Although the Genetic Algorithm and the Generalized Linear Model are the most consistent methods for parameter selection and modeling, the Simulated Annealing and the Cubist Model are the most accurate self-defined models.2) The respective estimated volume of erosion in the Sanjiangyuan region for the RUSLE and SA-Cubist Model is 3.1*109 t*a-1 和 2.3*109 t*a-1, and the accuracy of the SA-Cubist Modelis better than that of the RUSLE Model, which indicates that self-defined model is better than the RUSLE Model in terms of accuracy.3) Through a gravity transfer analysis of the grade of erosion, we found that results for both model showed a gradual increase in the grade of erosion from southeast to northwest. A comparison of the grade of erosion between counties showed that three areas with the highest grade of erosion are Tanglha Town of Golmud, Zhiduo County, and Xinghai County.4) In the Sanjiangyuan region, the grade of erosion for land with different vegetation is:farmland > sparse vegetation/swamp/desert > grassland > forest > meadow > boscage. Except for that the grade of erosion for grassland is dramatically high, our findings are generally in line with other comparative research findings. Considering human factors, grassland vegetation is featured with pasture grass, and overgrazing should be the major cause for the severe erosion in the research area, which points out the key area for the prevention and control of erosion in the future.5) A comparison of the importance of each parameter in the process of parameter selection shows that wind erosion and freeze-thaw erosion are two of the most significant dynamics for alpine grassland erosion and the major features different from those of other types of underlying surface erosion. And in our study, the self-defined model is better than the RUSLE Model in that the RUSLE Model fails to account for average wind speed and land freeze-thaw data.6) The grade of erosion is generally high in the Sanjiangyuan region. Based on the volume of erosion and the market value method, we have estimated the economic value of for annual loss of organics in the research area. The results for self-defined model and the RUSLE Model are 29.9and 39.9 billion RMB respectively, which is a considerable amount of economic loss. At present,the major problem in the research area is how to, based on the fact that several environmental factors varies along each side of the Xinghai-Zaduo line, formulate reasonable measures for the prevention and control of land erosion in the Sanjiangyuan region.