Abstract: The Multi-criteria Evaluation (MCE)-Cellular Automata Markov Chain (CA-Markov) model was adopted to simulate the changes of landscape pattern in the south part of Poyang Lake wetland from 2000 to 2015 derived from four Landsat satellite images. The accuracy of MCE-CA-Markov simulation was tested against multi-temporal remote sensing images and the landscape patterns for 2025 were forecasted in consideration of constraint factors. Research results show that :1) Arable land area decreased continuously from 2000 to 2015, but unused land decreased more dramatically. On the contrary, construction land, forest and grassland showed an increasing trend, and water area was relatively stable. Landscape elements increased while landscape connectivity became weak and landscape fragmentation increased. 2) The dynamic changes of Poyang Lake wetland were affected by natural and human factors, among which socio-economic development and urbanization played a leading role. 3) The simulated landscape patterns for 2010 and 2015 were roughly consistent with the interpreted landscape patterns. The simulation accuracies were high with the Kappa coefficients being 0.927 1 and 0.863 2, respectively for 2010 and 2015. 4) The results of simulated landscape pattern in 2025 showed that the areas of cultivated land and unused land would decrease continually, whereas the areas of construction land and forest and grassland would increase, and the water area would witness no obvious change. In addition, the changes of landscape pattern in the study area were relatively active; but as we would face more ecological environment pressure, measures to protect arable land and utilize rationally unused land need to be implemented.

Key words: Poyang Lake wetland, Landscape Pattern Index, MCE-CA-Markov model, dynamic simulation, Kappa coefficient