Integration of SimWeight and Markov Chain to Predict Land Use of Lavasanat Basin

Production and prediction of land-use/land cover changes (LULCC) map are among the significant issues regarding input of many environmental and hydrological models. Among various introduced methods, similarity-weighted instance-based machine learning algorithm (SimWeight) and Markov-chain with lower complexity and proper performnce are frequently used. The main aim of this study is utilizing SimWeight along with Markov chain to predict land-use map of Lavasanat basin located in north-east of Tehran for the year 2018. In this regrad, eight driver variables and two land-use maps of the sudy area which were created from two Landsat-5 TM image sensor for the years 2000 and 2011 were considered as input. To evaluate the result of SimWeight, Receiver Operating Characteristic was used. The Landuse map of year 2018 was predicted using the proposed method. To evaluate this map, a landuse map of 2018 was produced using classification of a Landsat-8 OLI image. The results of model and value of area under curve (AUC) for transition potential map was about 0.78, which indicated good performance. Furthermore, the comparison of two produced and predicted land-use maps of 2018 shows great similarity. Generally, the results indicated the proper performance of the propsed method to predict LULCC.