Development of bonitet index models with artificial neural networks: The case of Scots pine (Pinus sylvestris L.) forests in Çankırı region

ABSTRACT Master of Science Thesis DEVELOPMENT OF BONITET INDEX MODELS WITH ARTIFICIAL NEURAL NETWORKS: THE CASE OF SCOTS PINE (PINUS SYLVESTRIS L.) FORESTS IN ÇANKIRI REGION Abidin AYRANLI Çankırı Karatekin University Graduate School of Natural and Applied Sciences Department of Forest Engineering Advisor: Asst. Prof. Dr. Muammer ŞENYURT In this study, the possibility obtaining the predictions about the relationships between height and age that can be used for site index predictions were investigated by using Artificial Neural Network Models (ANN) for scots pine stands located in Çankırı Forets. The 112 sample trees obtained by Ercanlı et. al. (2014) from Scots pine (Pinus sylvestris L.) stands located in Çankırı and Yapraklı Forest Planning Units in Çankırı Forest Enterprise, and Yenice Planning Unit in Ilgaz Forest Enterprise were used as the research material. For this purpose, 60 different ANN models including various calculation algorithms (the feed-forward backprop, the Cascade-forward backprop and Elman backprop), various neuron numbers and transformation functions were trained and compared with various statistical success criteria. ANN model including Cascade Correlation Structure based on 12 neurons and logistic sigmoid transformation function was found to produce the most satisfactory fits with R2 (0,9519), MSE (2,4940), RMSE (1,5793), SSE (1152,2), AIC (4,8278), BIC (440,6307)