Abstract: Yulin is one of the prefecture-level cities bordering Mu Us Sandy Land and Loess Plateau in Shaanxi Province, China. Among them, the Mu Us Sandy Land is one of the most typical ecologically fragile areas. The soil environment in Yulin City has presented the profound impact after remarkable remediation and utilization in recent years. Soil organic matter (SOM) can be an important indicator of soil fertility and productivity. The trend of SOM can also greatly contribute to the decision-making on the stability and security of soil ecosystem under different natural conditions and anthropogenic influence. It is a high demand to monitor the SOM dynamic changes from the large-scale space and long time series. Taking the conspicuous sandy land in Yulin City as the research area, the purpose of this study is to determine the characteristics of land use changes from 1990 to 2020. A systematic investigation was also performed on the variations in the SOM content under different land types that transformed from the sandy land, in order to clarify the effects of different remediation and utilization on the SOM in the sandy land. The dominant sandy land was selected to calculate the dynamic attitude of land use for the transformation characteristics of land use. Three machine learnings (decision tree, random forest, and XGBoost) were used to evaluate the factors related to the natural conditions and land use change characteristics of sandy land. The fitting accuracy was then obtained, according to each waveband of multispectral remote sensing images and related spectral indices. Finally, the XGBoost was selected to invert the SOM content. A systematic analysis was made on the SOM content and spatial distribution characteristics under different land types. After that, a semi-variance function was used to reveal the spatial variability. The average SOM content was calculated to clarify the influence of anthropogenic factors and natural environment on the desert SOM. The results show that more than half of the sandy land was remediated and utilized from 1990 to 2020, indicating the fastest transformation. Specifically, the sandy grassland was the most important land transformation, whereas, the fastest increase was found in the construction land area. The better inversion was also obtained to estimate the SOM content with the inversion error within 13% than before, according to the multispectral remote sensing using XGBoost machine learning. The average SOM of arable land and water area reached nearly 0.8% of land use types. The SOM of all land use types decreased significantly, with the average SOM of 0.51% in 2020. A strong spatial autoregulation of SOM was depended mainly on the natural environmental factors, such as temperature, precipitation, and topography. Initially, the human activities posed a positive impact. But, a negative impact was led to the decline in the SOM content and the land degradation, as the intensity of sand use increased. Some recommendations were given to strengthen the restoration and improvement of degraded forest and grass. The finding can provide the theoretical and practical implications for the sandy land remediation, particularly for the protection of the soil environmental safety of Yulin sandy land.