GIS is an essential tool for large scale land use planning.The spatio temporal dynamic changes of landscape patterns are an important issue for researches in landscape ecology.The transitions among various land use types over time, using the Markov chain model and combined with economic factors,were conducted for the management and planning of land use system.These models,however,only depend on the transition matrix,whose elements were constant, to describe the dynamic change of spatial patterns. In addition, the anthropogenic and natural disturbances were not considered in these models, which made them somehow ineffective in their applications. This paper, based on GIS and taken account of anthropogenic and natural disturbances, proposes a multi dimension landscape patterns dynamic optimizing decision models(t+1)=s(t)[Q(t)+w t(t)r(t)]for the dynamic programming of land use system, where the vector s(t) is the distribution of areas of land use, and develops a preliminary computer decision support system for land use planning and management. Aimed at certain given objectives, a series of decision variables were generated and the land use system can be made to reach their optimum composition as a whole. The algorithm for the model was provided and proved to be convergent. Linked with GIS, the dynamic simulation using this model for the spatial patterns of land use system can be quantitatively and visually illustrated. As a case study, this model had been applied in land use system of Hangzhou City, China, which was a part of work for the overall planning of Hangzhou City. The results showed that the model operates well.