摘要
红锥是南亚热带地区重要的材用和景观树种。运用多个统计分布模型同时拟合树种(特别是包括红锥在内的乡土常绿阔叶树种)多个个体大小指标的研究报导较为少见。本研究仔细调查了广州城市景观林中两年生红锥的冠幅、胸径、基径和树高4个指标,选用12个具有不同函数形式的主要连续型分布,组成一个相对完整的模型集合,运用最大或然法和最小二乘法,同时模拟上述4个长度指标及其衍生的6个面积和体积指标。研究发现:(1)10个指标都有各自最适合的分布模型,也共同遵从几个分布模型,其中以伽玛分布最佳,其次是逻辑斯谛分布和韦伯分布;(2)分布曲线形状因指标的维度而异,从一维直径和高度、二维面积到三维体积,曲线变得越来越低矮、趋于正偏,甚至从单峰形变成凹形;(3)期望概率分布对观察频数分布的拟合优度与指标的尺度有关,通常是对数尺度优于线性尺度,例如,对数柯西分布比柯西分布更适合模拟10个指标中除胸径和胸高面积之外的8个指标;(4)一个整合一致性Akaike信息标准CAIC、Kolmogorov-Smirnov检验统计量KS和回归决定系数R2的统计量CAIC×KS/R2,可作为拟合优度的综合评价标准。本研究可望为林木栽培育种和种群生态学等领域的研究提供借鉴。
Studies have been rare on several competing models of statistical distributions simultaneously confronted with data from several body size indicators of tree species, especially for native evergreen broad-leaved trees such as Castanopsis hystrix A. DC., an important timber and landscaping tree in south subtropics of China. We carefully surveyed 4 body size indicators （canopy diameter, diameter at breast height, ground diameter and tree height） of 2-year C. hystrix in an urban landscape forest of Guangzhou, South China. Here we collected a set of relatively complete statistical distributions consisting of 12 major continuous distributions with rather different function types to model the observed frequency distributions of 4 length indicators and their 6 derived area and volume ones using the methods of both maximum likelihood and minimum squares. We found that： （1） 10 body sizes each have their own best-fitted distribution models and also share several ones, with the gamma distribution best, followed by the logistic and the Weibull; （2） distribution curves in shape vary with body size dimensions, i.e., from 1- （diameter and height）, 2- （area） to 3-dimension （volume）, they become shorter, more positively skew and even change from unimodal to hollow shape; （3） the goodness-of-fit of expected probability distributions to observed frequency distributions is related to indicator scale, and is generally greater on log scale than on linear scale, e.g., the logCauchy distribution has better fit than the Cauchy distribution to data from 8 out of 10 indicators except the diameter and area at breast height; and （4） a new statistic CAICxKS/R2 integrated with the consistent Akaike information criterion （CAIC）, Kolmogorov-Smirnov test statistic （KS） and the determination coefficient of regression （R2） can serve as a comprehensive criterion of goodness-of-fit. Our study can provide insight into some other research areas including tree silviculture and breeding, and population ecology.
出处
《生态环境学报》
CSCD
北大核心
2013年第2期189-198,共10页
Ecology and Environmental Sciences
基金
国家林业公益性行业科研专项(201004042)
国家科技支撑计划课题(2012BAD22B05)
关键词
红锥
个体大小
统计分布模型
伽玛分布
最大或然估计
城市景观林
Key words： Castanopsis hystrix
body sizes
statistical distribution models
gamma distribution
maximum likelihood estimate
urban landscape forest
