Abstract: Taking Songnen Plain as an example，this paper carried out research on land use pattern space changes. The study area includes Baicheng City，Songyuan City and other 10 cities and counties，with a total area of 46930km2.For half a century，land­use pattern changed dramatically. Lakes' shrinking，wetlands' disappearing，grassland's degradation，land salinization，and desertification developed rapidly，are seriously threatening the regional sustainable development. This paper used multiple remote sensing data source as spatial data. Supported by 3S technologies，the 1km×1km grid unit is taken as the smallest information atlas unit and Geo­information Atlas­spatial Analysis Model was built up. Spatial regression analysis between the land­use patterns and their impact factors had been investigated，and the land use pattern space change trend was predicted. The south West Songnen Plain land­use information maps in 1989 and 2004 were put into Markov model，the land­use transition probability matrix and the transfer area from 1989 to 2004 were calculated. During the geo­information atlas－CA­Markov model construction,2004 was setting as starting，suitability information atlas for changing land use pattern were put into the transfer matrix and transfer area files，using 5×5 filter，and the cycle number was set to 15.Finally，the forecasted atlas distribution atlas of land use pattern in 2019 was obtained.Through statistical analysis，results are as follows: by 2019 the area of land use structure and function will be changed enormously; wetlands，water and grassland area will decrease 61.37%,52.45% and 45.77%，while the saline land area will increase 21.76%. Saline desertification will expand further. Therefore，the appropriate countermeasures and programs must be established as soon as possible.This paper integrated the Geo­information Atlas and spatial analysis model，constituting a Geo­information Atlas－CA­Markov coupling model.The model option uses the raster atlas data and CA graphical data，which showing the complex changes in the ecological landscape features as information atlas. The whole research had been finished in rasterizing，digitization，spatialization with a dynamic fashion.