Monitoring desertification in Mongolia based on Landsat images and Google Earth Engine from 1990 to 2020

Abstract

Desertification is one of the most serious ecological and environmental problems in arid regions. Low-cost, wide-ranging, and high-precision methods are essential for the formulation of appropriate strategies for quantitatively monitoring desertification. In this study, based on Google Earth Engine and Landsat images, six machine learning methods were used to monitor desertification dynamics in 1990–2020 in Mongolia. The spatiotemporal distributions and changes in desertification at different stages were analyzed using gravity center change and intensity analysis models. Subsequently, we quantitatively investigated the factors driving desertification in Mongolia. The results indicate that the maximum entropy method can obtain the most accurate assessment of the degree of desertification in comparison with the other five methods, with an accuracy of 96%. In 1990–2005, the area of desertified land increased significantly, afterward, a decreasing trend was observed. Lightly and moderately desertified lands had the highest change intensities and were most sensitive to environmental factors. Although the desertification dynamics are under the influence of both natural and anthropogenic factors, precipitation plays a dominant role in Mongolia. This study provides a comprehensive analysis of the desertification status and trends in Mongolia, and presents desertification maps that can be used to formulate preventive measures and guide desertification prevention and control.

Keywords

• GEE
• Machine learning methods
• Desertification dynamics
• Gravity center change
• Intensity analysis
• Desertification maps