Application logistic regression and Markov Chain in land cover change prediction in east of Mazandaran province

Document Type : Research Paper

Authors

1 Department of Environment, Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, Noor, Mazandaran, P.O. Box 46414-356, Noor, Iran.

2 Department of Environment, Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, Noor, Mazandaran, P.O. Box 46414-356, Noor, Iran

3 3Department of Forestry, Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, Noor, Mazandaran, P.O. Box 46414-356, Noor, Iran.

4 4Department of Environment, Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, Noor, Mazandaran, P.O. Box 46414-356, Noor, Iran.

Abstract

This study was performed with objective land cover change prediction (forest, agriculture, residential and orchard) using logistic regression and Markov Chain in the GIS environment in east of Mazandaran province. Land cover change detected using satellite imageries belonging to the years 1987 and 2001. Transition potential modeling was conducted using a logistic regression. Seven variables (DEM, distance from residential, distance from agriculture, distance from forest, distance from river, distance from road, and qualitative variable) and 7 sub-models (forest to agriculture, forest to residential, forest to orchard, agriculture to residential, agriculture to orchard, orchard to residential, orchard to agriculture) were employed. Land cover change prediction conducted using Markov Chain and hard prediction for 2006. The accuracy assessment was determined using predicted map compared with actual map 2006. Finally, landcover change prediction done for 2013. Results showed that during the years 1987 to 2001, the large amount of forest and orchard have been reduced and, in contrast, agriculture and residential have been added. Null successes, hits, misses and false alarms were 89.8%, 0.1%, 9.8% and 0.3% respectively. Total error prediction model was 10.12% which is indicative of acceptable model. Furthermore, the prediction results showed that forest and agriculture will be reduced and residential and orchard will be increased in 2013 compared with 2006.

Keywords