Abstract: Aeolian desertification is one of the most serious ecological environment problems, for which remote sensing and spatial-temporal changes analysis provide an effective technical support. Combining these inversion indicators and the Decision Tree Classification method based on MODIS data, the information of aeolian desertification was extracted, a new index(Modified Aeolian Desertification Index, MADI) was given to analyze dynamic changes of aeolian desertified lands in the middle-west part of Inner Mongolia from 2000 to 2014. Adopting the spatial-temporal changes analysis method, changes of aeolian desertified lands at three different temporal scales were explored, and changes of desertification types and deterioration levels also were analyzed in this paper. The results show that: (1) The combination of indicators and the Decision Tree Classification can efficiently extract aeolian desertification information. (2) The types and area of aeolian desertification have a general decline trend in the middle-west part of Inner Mongolia during 2000-2014, the annual average aeolian desertified land area during 2011-2014 is about 1.72×10⁵ km² which is smaller than that during 2000-2010 by 2×10⁴ km². (3) Dynamic change of aeolian desertified lands have multi-temporal scale characteristics. Deterioration or rehabilitation trends of aeolian desertified lands at shorter temporal scale is different from those at longer temporal scale. With the increase of temprocal scale, the dynamaic degrees of slight desertification, moderate desertification, severe desertification and very severe desertification land show a decreasing trend. According to our results limited to the shorter research period, it is more reasonable to detect dynamic trend of aeolian desertification and its driving mechanism at 10-year or longer time scales, while it is helpful only to detect those hot spots such as higher desertification risk areas or restoration areas at 5-year or shorter time scales. (4) Different desertication reversion proportions present at different temproal scales, the lager of net desertication reversion proportion is 9.08% from 2005 to 2010 and 13.95% from 2005 to 2014. (5) Aeolian desertification in the whole Mu US Sandy Land continually reverse during 2000-2014, and desertification in the eastern Hobq Desert and the agro-pastrue zone of northern foot of Yinshan Mountains also gradually reverse during 2005-2014. On the contrary, aeolian desertification in the middle-west part of the Otindag Sandy Land present an instability, and tend to deteriorate during 2000-2014.

Key words: MODIS, remote sensing monitoring, aeolian desertification, space-temporal change analysis