AUTHOR=Zhou Xiao , Zhou Yang , Zhang Xuan , Sharma Ram P. , Guan Fengying , Fan Shaohui , Liu Guanglu 

TITLE=Two-level mixed-effects height to crown base model for moso bamboo (Phyllostachys edulis) in Eastern China

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 14 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2023.1095126

DOI=10.3389/fpls.2023.1095126

ISSN=1664-462X

ABSTRACT=Height to crown base (HCB), as one of the important prediction variables of forest growth and yield model, is of great significance for carbon storage and bamboo stem utilization of bamboo forests. However, the existing HCB models, which are only a few, have been built on the hierarchically structured data without applying correct modeling approach. As bamboo forests, which play an important role in the ecosystem functioning and mitigation of climate warming, development of HCB model for bamboo provides useful guidance for its effective forest management. Based on the fitting of data acquired from 38 temporary sample plots of Phyllostachys edulis forests in Yixing, Jiangsu Province, we selected the best HCB model among the six basic models. We extended the best basic model (Logistic model) through integration of the number of predictor variables, which involved evaluation of the impact of 13 variables on HCB. We introduced both the block- and sample plot-level random effects to the extended model for accounting for nested data structure through mixed-effects modeling. Results showed that bamboo height (H), diameter at breast height (DBH), total basal area of all the bamboo individuals with diameter larger than that of the subject bamboo (BAL)) and canopy density (CD) provided significantly more contribution to the variations of HCB than many others did. DBH had larger impact on HCB followed by BAL and CD. Compared with the basic and extended model forms, introducing two-level random effects provided significantly larger improvement on the model accuracy. Different sampling strategies were evaluated in the response calibration (model localization) and identified the optimal strategy. With increase of the number of sampled bamboo in calibration, the prediction accuracy of HCB model was substantially improved. However, using many sample bamboos per sample plot in calibration could increase the measurement cost with a very little accuracy gain. Thus, calibration should use only four randomly selected bamboo individuals per sample plot, which would provide a desirable compromise among the measurement cost, model use efficiency, and prediction accuracy.