Desertification causes not only a reduction of vegetation cover and surface moisture but also the expansion of sandy land. Considering that the modified soil adjusted vegetation index (MSAVI) is linearly related to canopy cover, albedo plays a significant role in the surface energy balance, and the sandy feature index (SFI) can identify sandy land. This paper proposed a desertification monitoring index (DMI) based on MSAVI, albedo and SFI to accurately monitor desertification. In addition, three classification methods (quantile grading, the susceptibility index segmentation method, and Jenks Natural Breaks) were used to divide the DMI into non-desertification, mild desertification, moderate desertification, and severe desertification. The relative exponential entropy model was adopted to select the optimal method for desertification classification. Spatiotemporal distributions of desertification in Duolun County were finally observed from 2005 to 2020. The results show that (1) although MSAVI, albedo and SFI have their own advantages in monitoring desertification, a single indicator is not suitable for various land uses. The DMI can combine the advantages of MSAVI, albedo and SFI in desertification monitoring and is more consistent with the real desertification levels. (2) Compared with quantile grading and susceptibility index segmentation methods, Jenks Natural Breaks has the smallest relative exponential entropy (2.124), with an overall classification accuracy of 88.21%, indicating that Jenks Natural Breaks is an optimal method for desertification monitoring. (3) Although approximately 1708.60 km2 of desertification areas have improved from 2005 to 2020, accounting for 44.23% of the whole area, the remaining areas are still under fluctuating or degraded conditions. This shows that the local government has made some progress in desertification protection, but the areas under fluctuating or degraded conditions still require protection or management.