Developing the forest road network to promote tourism using artificial neural network and GIS (Case study: Kheirud Forest)

Tourism in forests  can be effective in developing forest management plans, specially during forest rest periods. Forest areas are important for the protection of ecosystems and natural resources, as well as providing recreational opportunities for people. Therefore, forest recreation, as a form of tourism development, is a significant factor that provides many direct and indirect economic, social, cultural, and environmental benefits.  The aim of this research is to design and complete the forest road network and add walking paths with the goal of tourism development in Kheyrud Forest using the capabilities of artificial neural networks and GIS.  

Initially, layers of slope, aspect, elevation, geology, soil, canopy cover percentage, and existing roads were prepared, and each layer was internally classified and weighted. The internal classification of the layers was done based on the opinion of an expert and researcher in the field of forest engineering and forest tourism, and the weighting of the layers was performed using the Analytic Hierarchy Process (AHP) method. By integrating the various layers and their corresponding weights using the Weighted Linear Combination (WLC) method, a suitability map of the Patam section was prepared as the training section for the neural network to pass the road network. The value of each cell from the shapes, along with coordinates, was extracted using ArcGIS software, and all data were mapped to a range of 1 to 5. In this research, a Multilayer Perceptron (MLP) neural network with 30 neurons in the hidden layer was used for modeling. The data on slope, aspect, elevation, geology, soil, canopy cover percentage, and existing roads were used as inputs, and the suitability data for the Patam section road passage was used as the output for training the network. The neural network estimated the suitability for each of the three active forest sections based on the Patam section.

The overlaying of the existing road network layers, landscape map, and suitability map of the research area showed that the existing road network has made many areas with tourism potential accessible to tourists. Therefore, to complete the access network to areas with tourism potential in the forest, using the PEGGER extension in the ArcView software environment and considering the landscape shape and suitability shape, walking paths of approximately 14 kilometers were designed. During the design, efforts were made to ensure that the road passes through more desirable areas (pixels with lower values). According to the results, the MLP artificial neural network, with a coefficient of determination (R²) of 0.902 and a root mean square error (RMSE) of 0.126, showed greater ability than linear regression in estimating the suitability value for road passage.

The results of this research demonstrate the capability of the intelligent method based on artificial neural networks and GIS for designing and planning the road network. The findings indicate an increase in the learning ability of the MLP network with an increase in iterations up to the 7th iteration.