科研之友
微信
新浪微博
Facebook
分享链接
摘要
Tree height is a key variable in forest monitoring studies and for forest management. However, tree height measurement is time consuming, and the recommended procedure is to use estimates from tree height (H) - diameter at breast height (DBH) models. Increasingly, H-DBH models are being developed for urban forests, providing tools to forest management and ecosystem services estimation. Here, we compared model forms and approaches for predicting H as a function of DBH and additional stand level covariates variables. Four model forms were evaluated: (i) basic models (which only used DBH as predictor variable); (ii) generalized models (which used DBH and other predictor variables based on the best basic model); (iii) a mixed-effects model based on the best basic model; and (iv) a mixed-effects model based on the generalized model. Several sampling designs aimed at minimizing height measurement were tested in terms of accuracy and applicability. Taking predicted accuracy and investigation cost into account, we recommend generalized non-linear mixed-effects model (NLME) when there were two or less tree height measurements taken in a given stand. The basic NLME model could be calibrated when there were 3 or more tree height measurements, depending on the required level of accuracy and investigation cost. Additionally, we first reported that soil pH as a covariate variable in H-DBH model and our generalized NLME model implied that the difference in the H-DBH relationship caused by pH varies among different stands. This finding may be attributable to differing biological properties of the similar alkaline tolerance species.
