Development of height estimation models for pure beech, beech and oak mixed stands of Çatalca Binkılıç

In this study, height-diameter models were developed separately for pure beech stand and beech oak tree species and on a general basis in mixed beech-oak stands in the Çatalca Binkılıç region. The data used for developing the height-diameter models were obtained from sample plots located in pure beech and mixed beech-oak stands, representing various developmental stages and canopy closure levels across different sites. A total of 38 sample plots were taken from pure beech stands, with 1861 beech trees sampled. Additionally, 62 sample plots were taken from mixed beech-oak stands, where a total of 3956 trees were sampled, including 2441 beech and 1515 oak trees. Height-diameter models, using DBH as the sole explanatory variable, were fitted through nonlinear regression analysis (ONLS) and support vector machine learning techniques. Ten height-diameter models, previously proven effective in various studies, were fitted using nonlinear regression analysis (ONLS). Among these, the model with the highest predictive performance based on goodness-of-fit statistics was selected as the best-performing model. The height-diameter models developed for pure beech stands, as well as for general, beech, and oak tree species in mixed stands, explained 69.7% to 82.7% of the variation in height, with prediction error rates ranging from 5% to 9%. The height estimation models developed using support vector machine learning achieved a prediction error rate of 6% in pure beech stands. In mixed stands, the models predicted with error rates of 14% on a general basis, 12% for beech, and 16% for oak tree species, respectively. For pure beech stands, the height-diameter models developed using both methods demonstrated superior prediction performance compared to the other models.