Assessment and prediction of land use changes in the Anzali wetland Basin, Based on Land Change Modeler (LCM)

Natural and human activities in coastal areas cause dynamic changes in land use and land cover. Rapid population growth in these areas accelerates the process of land use and natural land cover changes, and the transition to residential use and infrastructure development. This research was conducted to investigate and modeling land use changes in Anzali wetland basin between 1975 and 2015 using satellite imagery and predicting possible land use change in 2045 using the LCM model. In order to achieve quantitative and qualitative changes in the study area, the land use maps of the Anzali wetland basin have been produced based on Landsat satellite images for years 1975, 1989, 2000, and 2015. For this purpose, six land use classes including agriculture, rangeland, forest, wetland areas, urban lands, and wetland surface were considered. The accuracy of the land use maps was verified by overall accuracy and kappa coefficients using 323 points based on stratified random sampling and these two parameters were 87% and 0.71, respectively. The LCM model was used to detect and map the changes of different land use categories in the Anzali wetland basin during the periods 1975-1989, 1989-2000, 2000-2015, and predict land use changes in 2045. Analysis of the change detection matrix shows that during the period 1975 to 2015, the total change and transfer of different land uses to each other is 76648.14 hectares. The most changes among different land use during this time are related to the transfer of different land uses to agriculture for 49827.69 hectares, which is equivalent to 65% of the total changes of different land uses. Changing of different land uses to agricultural use is the main change in the uses of this period. forests (64%), rangelands (16%), wetland areas (10%), wetland surface (8%) and residential areas (2 %) have the largest share, respectively. Throughout the study, the expansion of urban land use has always been a positive trend in line with population growth. Based on these changes and by taking 7 independent variable and 8 sub-models, transition potential modeling was done using Artificial Neural Network. The results of modeling in most scenarios showed high accuracy (60.14 to 88.73 percent). To verify modeling accuracy, the standard Kappa coefficient (0.8948) and Null Successes error (77.9%), Hits (3.1%), Misses (15.9%), False Alarms (3.1%) were calculated and accuracy of the position and number of pixels in each class was determined. The ratio of Hits to the total pixels has changed (14.2) indicates that model results are acceptable in predicting land-use changes. Comparison of the results of the changes and conversion of land use classes in the period 2015 to 2045 (predicted) in the region shows that if the land utilization trend continues with current management mode, 10036.26 hectares of forest lands would change to agricultural lands (67.69%), rangeland (32.04 %), urban areas (0.16 %) and wetland surface, and considering the transfer of other uses to forestry, eventually the 9963.36 hectares of forest will be reduced during this period. In general, agriculture, rangeland, and urban areas will increase during this period.