Accuracy assessments of land use change simulation based on Markov-cellular automata model

Abstract

Abstract The potential of Markov-cellular automata model has been recognized and used widely in land use changes simulation and forecast, but previous researchers focused on analysis of future simulation result and land use change. Few thoughts have given to the analysis of modal accuracy and uncertainty and previous accuracy assessments have largely been conducted by simply calculating the overall value of selected accuracy metric. To increase the applied assessment value of Markov- cellular automata model, we explored the accuracy of simulation through the calculation of Kappa index for location and quantity. The case study area was Changping District, which is a rapidly growing area of Beijing. We classified the 1988 and 1995 TM image into six land use types in accordance with Chinese standards. Computed the transition suitable atlas based on different range and measurement units of the physical and socioeconomic factors. And then, we integrated these datasets to simulate the 2000 land use change map by Markov- cellular automata model. Lastly, we analyzed the simulation result and assess the accuracy of different cell size and neighbour size through various kappa index. The results of simulation showed that simulation accuracy of small cell size is better than the big cell size although it takes a substantial amount of time to run. However, the prediction accuracy of the model was quite stable with the neighbour sizes from 3*3 to 13*13. These results allowed us to understand what and how the factor including cell size and neighbour size affected simulation accuracy and how to select the best cell size for simulation.