Modeling knot geometry from scanned images of Korean pine plantations

Based on 1038 knots from 42 sample trees from 14 standard plots in Korean pine (Pinus koraiensis Sieb. et Zucc.) plantations in Heilongjiang Province, China, we developed a knot profile model system. The concepts of growth inflection point and death inflection point were proposed. Depending on the growth conditions near the growth inflection point of the knot, we divided the knots into two types: curved knots and linear knots. A logistic regression model was constructed to predict the shape of the knots. The area under the curve of the logistic regression model was 0.699 (95% confidence interval 0.661∼0.736), and the prediction accuracy was 0.69. Our research shows that for middle-aged and young forests, the relative radial distances of the growth inflection points at the upper edge and bottom edge of curved knots were larger than those in linear knots. The curved knots did not bend at the death inflection point, but linear knots may bend at the death inflection point. Models were constructed separately for curved knots and linear knots. The results showed that the application of the mixed-effects model significantly improves the model fitting effect.