You could take data from global cover maps and evaluate cover transitions. One source of yearly Land Cover is MODIS MCD12C1. You also have MODIS GlobalCover from 2005 and 2009.

A good source of global gridded data is Worldgrids website.

Transition matrix can be obtained from tools like SPC plugin for QGIS or lulcc R package, just to cite some open source options.

Links bellow.

Hope it helps

https://lpdaac.usgs.gov/dataset_discovery/modis/modis_products_table/mcd12c1

http://due.esrin.esa.int/page_globcover.php

http://worldgrids.org/doku.php?id=wiki:land_cover_and_land_use

https://plugins.qgis.org/plugins/SemiAutomaticClassificationPlugin/

https://cran.r-project.org/web/packages/lulcc/lulcc.pdf

Assuming you refer to independent Mongolia, rather than Chinese Mongolia, some of my ground observations for your interpretation of imagery. Large collective farms with (subsidized) dry-land cereal cultivation (kolkhozy) in the steppes have been abandoned after the end of communist rule. These will show up as bare rectangular patches. After privatization and individualization of livestock herds, pastoralism has has been intensified near  roads and extensified in the periphery. This has not lead to major cover changes. It may be detectable by textural differences. Further, some coniferous forest areas are heavily infected by pests and many, if not most trees are dead. Forest fire scarce may be found as well. Finally, open cast mining has been on the increase and should appear in change detection. 

You can check on the images provided by Google Earth Engine for detecting the changes over a temporal frame. Based on the knowledge of remote sensing imagery it can be easily understood.

Dear Changliang ,

You can easily use landsat images with google earth imageries for different time periods to understand the land changes in your specific area, then after that you can compare your work with theoretical statistics if you have.

Kind Regards.