Integration of Multinomial-Logistic and Markov-Chain Models to Derive Land-Use Change Dynamics

Land-use change reflects the relationship between human activities and environmental processes over time and space. Modeling land-use change dynamics is of particular environmental, social, and economic importance at regional scales. This paper develops a new approach to model land-use changes by making use of multiple categories that incorporate socioeconomic and environmental factors with multinomial logistic models and Markov chains to quantify the impact of these variables on land-use change dynamics for the years 1990, 2000, and 2010. Spatial autocorrelation and colinearity tests were utilised to screen the most suitable independent variables before modeling. The multinomial logistic model was evaluated by means of a likelihood ratio test and pseudo R-2. A Markov transition matrix was then designed for integration with the multinomial logistic model to illustrate the temporal land-use change dynamics from 1990 to 2010 and to visualise the predicted land-use change map. The approach was calibrated for 2010 by cross comparing actual and simulated land-use maps. The predicted map of 2010 showed accuracies of 78.54, 70.38, and 57.25% for urban land, arable land, and grassland, respectively; the total accuracy for the three simulated land-use types was 72.37%. A future land-use map for 2020 was predicted by means of the hybrid approach demonstrated in this paper, and this approach could be used in similar research. (C) 2014 American Society of Civil Engineers.