MARKOV CHAINS BASED LAND COVER ESTIMATION MODEL DEVELOPMENT: THE CASE OF ANKARA PROVINCE

Abstract

Land cover changes due to many factors ranging initially from the numerical size and density of the population to socio-economic characteristics and technology, from land policies to natural factors. Most of the land cover change is human-induced and involves many economic and ecological problems. Using the land by protecting it is of great importance to solve these problems before they worsen and to take precautions. Temporal monitoring of land cover changes and improving Geographical Information Systems and Remote Sensing technologies provide decision makers with many opportunities for the future of this change. The speed, direction and type of land cover change are identified with the developed forecast models, thus forming a basis for planning for sustainable land use. Although many probability methods are used to make land cover estimations, one common practice is Markov chain model. In Markov chains, a special class of random processes, two or more results that emerge based on repeated observations can be determined by means of probability laws. The main objective of this study is to estimate the 2018 land cover of Ankara province with Markov chains technique. In this context, CORINE 1990, 2000, 2006 and 2012 data sets were evaluated according to first level land cover classes. In the model for 2012 land cover estimation, 92% accuracy was achieved based on the first three data sets. With the addition of 2012 CORINE data to the model, evaluation for 2018 was made and the accuracy of the model was found to be 90% for 2018 based on the added data set. According to these latest estimation data, the change of land cover in the province will continue to develop in favor of artificial areas and against agricultural areas.