Modelling land use change by an integrated Cellular Automata and Markov Chain Model (case study; Azadshahr County)

Abstract

In this research, land-use changes in Azadshahr County were investigated from 1998 to 2009, using the imageries from Landsat 5 satellite and an integration of the Markov chain and Cellular Automata methods. Using the object-based support-vector-machine image classification method, land-use maps were classified into three major categories, namely agriculture fields, forest lands and built-up areas for the years of 1987, 1998 and 2009; their overall accuracies have been obtained 91.0%(1987), 91.0% (1998) and 88.8% (2009), with the respective Kappa values of 86.5%(1987), 86.5% (1998) and 83.2%(2009). The built-up areas had the greatest changes by increasing 2.02% and 2.17% for the periods of 1987-1998 (as first period) and 1998-2009 (as second period), respectively. During the first period, forest area has shrunk by approximately -1.80%. However, as a result of the afforestation project during 1998-2009, forest area has increased 1.59%, while over the 22-year period the total area of forest has merely reduced by -0.21%. Agricultural areas on the one hand has shrunk in favor of the built-up areas, and on the other hand, increased by the conversion of the forest lands, making a total reduction of -0.22 and -3.75% for the first and second periods, respectively. The land-use pattern of 2020 was simulated using the MULOSCE extension of the QGIS software based on the integrated cellular automata and Markov chain technique. It is expected for this period to encounter a 0.62% increase in built-up areas, with 0.48% and 0.15% reduction in agriculture fields and forest lands, respectively.

Keywords

• landsat
• object-oriented classification
• support-vector machine
• markov chain
• azadshahr