APPLICATION OF ARTIFICIAL NEURAL NETWORK SYSTEM BASED ON ANFIS USING GIS FOR PREDICTING FOREST ROAD NETWORK SUITABILITY MAPPING

There are different factors which affect the well-planning of forest road networks and originated from both general and regional characteristics. It is necessary to support the decision maker effectively by considering the selected criteria during the process of planning the forest roads in the forest areas which are located in the mountainous regions. So, first of all these factors need to be ranked according to the significance. This study is intended to reveal the forest road network plan alternatives which have a high predictive power in planning by considering the criteria that affect the planning when the forest road planning is made. To this end, the forest management chief, which is the smallest forest management unit within Turkey, has been selected as the sample area. Nine criteria affecting the forest road network planning of this units were evaluated via Geographic Information Systems (GIS) and Adaptive Network Fuzzy Inference System (ANFIS) and six different forest road network models were established. The AUC values of the generated models are between the range of 0,654 and 0,845. The most appropriate ANFIS model was determined as model 4 which comprises elevation, slope, distance village and NDVI criteria. The multi-directional approach was used to reflect the effect of factors on forest network planning via GIS, artificial neural networks and fuzzy logic. Furthermore, the factors employed in the developed models can be diversified at the regional level and utilized by the implementers or decision makers in forest road network planning.