Statistics analyses of spatio-temporal land use data, such as historical review, flow analysis，change index analysis and trend analysis, are important land management operations, and attract more and more attention from management and planning department. To overcome the difficulty in statistics of annual land use type in any query region and the time consuming problem in long time series change flow analysis, the statistical optimization method based on spatio-temporal variation model was proposed. For the former difficulty, the feature entities in the statistical region and boundary were classified with the proposed method based on the principles of connectivity of graphics, and then the statistical optimization algorithm of sequential snapshots was used to realize the statistics of time point status in any query area. For the latter problem, the spatio-temporal network approximation judging was carried out with the method based on multi-commodity flow principle, to reduce time consuming and improve the efficiency of long time series change flow analysis through reducing the number of spatial overlay analysis. Finally, the effectiveness and feasibility of the proposed method were verified through case study using land use data of Qionghai City, Hainan Province from 2009 to 2012.