The land-cover data obtained from the TM remote sensing images of 1990 as well as land-use map of 2005,2011 were used in Turpan to set up a Markov model for predicting changes of land-cover.Given the accuracy of prediction in Markov model is closely related to its simulated step,therefore,firstly,1990 was taken as initial year to predict various areas of land-cover on different classes in 2005,and the optimal simulated step was found out by comparing the minimum difference between the simulate and actual areas of landcover.Secondly,the changes of land-cover in 2011 was predicted by taking the optimal simulated step in Markov model,then the land use degree and land use space change dynamic were analyzed based on this.Results could be summarized as:(1)the areas of transform between unutilized land and grassland are enormous;(2)the cultivated land decreased quickly since 2005,which indicated that the contradiction between supplement and demand was being increased,and also meant that the balance of the cultivated occupation and supplement had not implemented;(3)nearly 6 000 hectares of water disappeared in 21 years which was largely probably associated with global warming;(4)because the change of construction land and garden plot made the land use degree increase year by year,it exhibited that the development of economy was an important factor resulting in the changes of land cover;(5)anthropic factor was a major reason which led to the changes of land-cover,people were taking more and more powerful means to affect the land cover change.Therefore,some appropriate proposals were also offered according to the above results.