Aesthetic quality evaluation modeling of forest landscape using artificial neural network

Considering decreasing trend in wood harvesting potential of Hyrcanian forests, we need to plan for utilization of other ecosystem services, such as ecotourism, more than past. For ecotourism planning, comprehensive information of ecological beauty and natural landscape structure should be summarized. On the other hand, accurate evaluation of different landscapes in a region requires comprehensive information of affective criteria, and its impact on user perception of landscape quality. Locating lookouts, which have high quality in landscape structure, is known as the first step to promote aesthetic quality of landscape and protection of natural ecosystems. This research aims to evaluate aesthetic quality of forest landscape using quantitative comprehensive approach and artificial neural network modeling for determination of the most effective landscape visual parameters in subjective aesthetic quality promotion of landscape.
The study area is three districts (with high diversity in landscape quality) of Khyrud research educational forest managed by Natural Resources College of University of Tehran which are named Patom, Namkhaneh and Gorazbon. In study forest, totally 200 landscapes, with different structure of tree cover and view composition, were selected to record landscapes characteristics. Landscape quality which is in the eyes of beholder, was evaluated in 200 studied landscapes.
In this study, in order to model the aesthetic quality evaluation of forest landscape, structural features and landscape parameters were recorded and aesthetic quality of landscape was classified in three classes of weak(1), desirable(2) and extremely desirable(3). Multilayer Perceptron network was used to data processing with artificial neural network.
Considering network coefficients of determination (Test samples) which is 0.88, 0.896, 0.969 in 1 to 3 classes, the accuracy of artificial neural network in aesthetic quality evaluation of landscape is assessed in extremely desirable level. Sensitivity analysis prioritizes landscape composition, tree diversity and thick trees view respectively to achieve class 1 and 2 in quality of forest landscape. On the other hand, tree diversity, landscape composition and view point respectively play a significant role in class 3 in quality of forest landscape.
results of the most effective variables on aesthetic quality of forest landscape, proved that landscape composition with higher diversity in its scenes, diversity in tree views with higher tree species and also thick and old trees in landscape should be a priority for forest landscape planning and management to achieve lookouts with higher quality of landscape in the eyes of beholders. This research prepared a new method for aesthetic quality evaluation of forest landscape and the introduced model is known as an environmental decision support system in forest ecosystems with an application in similar forests. Also practical criteria in aesthetic quality evaluation of forest landscape were introduced.