Forecasting deforestation and forest recovery using Land Transformation Model ‎‎(LTM) in Iranian Zagros forests

Land use changes and its patterns in spatial and temporal scales occur in a non-linear way. Therefore ‎to predict the potential and negative effects of these changes on forest ecosystem services in feature, ‎nonlinear tools such as Artificial Neural Networks (ANNs) are needed. In this study for forecasting ‎deforestation and recovery of Sardasht forests for 10, 20 and 30 years later, Land Transformation ‎Model (LTM) based on ANNs and GIS was used. For this purpose three different scenarios including ‎time periods of 1997-2007, 1997-2017 and 2007-2017 were used, and deforestation and forest ‎recovery of Sardasht using 14 variables for 2027, 2037 and 2047 were predicted. Results showed that ‎over 20-year studied time period (1997 to 2017) despite 2372.57 ha recovery of Sardasht forests, ‎‎10314.63 ha deforestation occurred. Deforestation and forest recovery modeling by all three ‎scenarios with good Receiver Operating Characteristic curve (or ROC curve) (more than 0.8) for all ‎scenarios, show a definite and increasing deforestation process in Sardasht over the next three ‎decades, so based on the 1997-2007 scenario, it is anticipated that 22296.24 ha of forests in the ‎region will be destroyed over the next 30 years. Also, spatial overlapping of models for forecasting ‎deforestation during the same period showed that 7.47% of Sardasht forests in identical locations, ‎have a high potential for deforestation and conversion to other land uses. The results of this research ‎can be used for proper conservation planning and increasing regulatory programs in areas with high ‎degradation potential.‎