Modeling Land Subsidence Susceptibility and Analysis of Influencing Factors in Shabaster Plain using Random Forest

Land subsidence refers to the vertical and downward movement of the earth's surface. This phenomenon may occur naturally or as a result of human activities. It causes damages to communication networks, infrastructure, and structures and agricultural land. In the past few years, land subsidence has occurred as a crisis in many of plains in Iran and has raised significant concerns. In this study, Shabestar Plain, one of the high-risk plains of East Azarbaijan Province, was selected as the study area. In order to identify the land subsidence susceptibility at different locations and also to identify the factors that influence land subsidence, random forest algorithm is used. In this study, a land subsidence map used as the inventory map, prepared by using time series based SAR interferometry technique as training data. The results confirmed that the southern part of the study area has very high susceptibility to land subsidence. In the modeling of land subsidence susceptibility by random forest, 5.63 % of the study area had Very High susceptibility and 6.37% had High land subsidence susceptibility. Also, based on random forest modeling, the highest effect was related to "elevation", "groundwater level", and "land use", and the least effect among the factors involved was related to "distance from fault". Finally the model was evaluated using area under Receiver Operating Characteristic curve, AUC= 96% confirmed very good predictive power of the model. The results of this paper can be used for planning for the development of various constructions and land uses in the future. Also, considering the possibility to identify regions and facilities at risk, timely and appropriate actions can be taken for prevention and minimizing financial and human losses.