Modelling tree height-diameter allometry of Chinese fir in relation to stand and climate variables through Bayesian model averaging approach
Modelling tree height-diameter allometry of Chinese fir in relation to stand and climate variables through Bayesian model averaging approach
复制标题
DOI：
10.14214/sf.10415
复制
发表时间：
2021
期刊：
影响因子：
1.8
通讯作者：
Lele Lu;S. Chhin;Xiongqing Zhang;Jianguo Zhang
中科院分区：
文献类型：
--
作者：
Lele Lu;S. Chhin;Xiongqing Zhang;Jianguo Zhang
树高直径异速生长反映了特定树种对地上和地下资源分配格局的响应。然而，传统的方法（如逐步回归（SR））可能会忽略模型的不确定性在变量选择过程中。本研究对450株杉木（Cunninghamia lanceolata（Lamb.）Hook.）以五个间距生长。通过贝叶斯模型平均（Bayesian model averaging，BMA）分析了林分和气候变量对树高-胸径异速生长的影响，并与SR方法进行了比较。结果表明，SR模型与BMA模型中后验概率第三高的模型相当。虽然SR方法的参数估计值与BMA相似，但BMA产生的估计值的95%区间略窄。高随密度、优势高和年平均气温的增加而增加，随林分断面积和夏季平均最高气温的增加而降低。结果表明，温度是影响杉木人工林高径异速生长的主导气候变量。而SR模型包含了最冷月平均气温和冬季平均最低气温，这些变量被排除在BMA，这表明冗余变量可以通过BMA删除。
Tree height-diameter allometry reflects the response of specific species to above and belowground resource allocation patterns. However, traditional methods (e.g. stepwise regression (SR)) may ignore model uncertainty during the variable selection process. In this study, 450 trees of Chinese fir (Cunninghamia lanceolata (Lamb.) Hook.) grown at five spacings were used. We explored the height-diameter allometry in relation to stand and climate variables through Bayesian model averaging (BMA) and identifying the contributions of these variables to the allometry, as well as comparing with the SR method. Results showed the SR model was equal to the model with the third highest posterior probability of the BMA models. Although parameter estimates from the SR method were similar to BMA, BMA produced estimates with slightly narrower 95% intervals. Heights increased with increasing planting density, dominant height, and mean annual temperature, but decreased with increasing stand basal area and summer mean maximum temperature. The results indicated that temperature was the dominant climate variable shaping the height-diameter allometry for Chinese fir plantations. While the SR model included the mean coldest month temperature and winter mean minimum temperature, these variables were excluded in BMA, which indicated that redundant variables can be removed through BMA.