Spatial simulation and prediction of land use and land cover using adaptive stochastic rules and landscape pattern characteristics

Land use and land cover has become an indicator of regional sustainable development. This paper describes a spatially explicit stochastic model based on Markov chains and maximum likelihood rule, which includes the landscape pattern expressed by class concurrences probability matrix and adaptive local estimation. The experimental results in Beijing mountainous areas show the overall simulation accuracy can be increased by 2.4% and 0.045 in Kappa coefficient if considering the landscape pattern. With increasing local adaptive estimations of model parameters, the simulation accuracy can be increased to above 90%. The model only needs two different time land use and land cover maps to run without the need to describe the complex relationships between biophysical, economic and human factors that affect actual land use and land cover changes.