Predicting tree diameter from height with regard to the tree age for Scots pine in Poland

With rapid advancements in remote data collection, individual tree height (H) becomes basic parameter obtained over large areas of forests. Hence the need to predict other attributes, especially tree diameter (D), from remotely sensed height calls for a systematic development of appropriate solutions, namely elaboration of predictive H-D models. This study evaluates performance of 25 height-diameter regression equations extracted from the forest biometrics literature.
The estimation properties and predictive performance of analysed models were evaluated based on the raw data from more than 3000 Scots pine (Pinus sylvestris L.) trees measured all over Poland. Assessment was performer in two variants: with and without inclusion of the tree age as an additional factor shaping H-D relationship. Set of various validation metrics, with weighted effect to final ranking, was used (R2, AdjR2, RMSE, MAE, Bias, AIC and BIC) to select the best performing model. Introducing age resulted in increase in overall performance of majority of investigated methods, however it makes direct modelling of D a bit less effective as precise age as a parameter describing given is not very often available.
Given that modern forest inventory increasingly relies on remotely sensed data, it is crucial to further develop effective models that can estimate various tree parameters using easily obtainable independent variables.