College of Forestry,Northeast Forestry University,Harbin 150040,China
Show less
文章历史+
Received
Published
2024-12-24
2025-09-15
Issue Date
2025-10-30
PDF (1128K)
摘要
在林业研究中,胸径-树高二元联合分布多由相同边缘分布构造,而林分的胸径与树高的实际分布状况可能有所差异。为降低这种差异带来的影响,依据佳木斯市孟家岗林场的115块长白落叶松人工林数据,选择适用条件低、适应范围广的Copula函数方法拟合落叶松胸径-树高二元联合分布模型。首先选择威布尔(Weibull)、广义威布尔(G-Weibull)、逻辑斯蒂(Logistic)、轻量逻辑斯蒂(Logit-Logistic)、伽马(Gamma)、对数正态(Log-Normal)6个分布函数作为备选基础模型,根据K-S(kolmogorov smirnov test)检验与半参数估计结果筛选并构建Copula胸径-树高二元联合分布模型,再通过负对数似然(negative log-likelihood,NLL)、Sn拟合优度统计量和似然比检验(likelihood ratio test,LRT)与二元对数logistic分布函数和二元Weibull分布函数进行比较,最后使用雷诺误差指数(error index of Reynolds,EI)对模型预测能力进行评估。结果表明,基于Copula函数的二元分拟合结果与模型(EI=0.318 4)预估能力皆优于二元Weibull分布(EI=0.638 1)和二元对数Logistic分布(EI=0.949 0),说明此方法构建胸径-树高二元联合Copula分布模型能够很好地描述落叶松人工林胸径树高联合分布,以Copula方法构建树高-胸径联合分布是可行的。
Abstract
In forestry research, the bivariate distribution of diameter at breast height (DBH) and tree height is often constructed from identical marginal distributions, yet the actual distributions of DBH and tree height within a forest stand may vary. To mitigate the impact of these differences, based on the data from 115 Larix olgensis stands in Mengjia Gang Forest Farm, Jiamusi City, this study employs the Copula function method, which has low application conditions and a broad range of adaptability, to model the bivariate distribution of DBH and tree height in Larix olgensis plantations. Firstly, we selected six distribution functions-Weibull, G-Weibull, Logistic, Logit-Logistic, Gamma, and Log-Normal- as potential base models. These were then screened and used to construct a Copula-based bivariate distribution model for DBH and tree height, based on the results of the Kolmogorov-Smirnov (K-S) test and semiparametric estimation. The model's fit was further evaluated using the negative log-likelihood (NLL), the Sn goodness-of-fit statistic, and the likelihood ratio test (LRT), comparing it with the bivariate logistic distribution function and the bivariate Weibull distribution function. Finally, the predictive capacity of the model was assessed using the Reynolds error index (EI). The findings indicated that the Copula-based bivariate fitting results and predictive capabilities (EI=0.318 4) outperformed those of the bivariate Weibull distribution (EI=0.638 1) and the bivariate logistic distribution (EI=0.949 0). This suggests that the Copula method for constructing a bivariate joint distribution model of DBH and tree height can effectively describe the joint distribution in Larix olgensis plantations, making it a viable approach for modeling tree height and DBH distributions.
使用R4.4.1软件对数据进行统计分析,使用fitdist函数进行胸径、树高一元分布拟合,筛选出最优的胸径、树高边缘分布;使用MATLAB2022b进行Copula函数的半参数分析,筛选出最适合数据的Copula函数;再使用R4.4.1软件中的fitdistrplus、copula、copBasic等R包构建Copula胸径-树高二元联合分布,定义函数计算二元分布的拟合指标与雷诺误差指数(error index of Reynolds,EI),利用ggplot2包作图。
通过上述比较,最终选取Plackett Copula函数作为本联合分布的依赖结构,选取Gamma分布作为胸径的边缘分布,选取对数正态分布作为树高的边缘分布。查找文献选择二元Weibull分布与二元对数Logistic分布作为对照分布,比较二元Copula胸径树高联合分布与传统二元胸径树高预测精度。表6通过负对数似然(negative log-likelihood,NLL)、Sn拟合优度统计量和似然比检验(likelihood ratio test,LRT)对3个二元联合分布模型进行了比较。二元Copula联合分布模型在NLL和Sn统计量上表现最优,而LRT进一步验证了其复杂性带来的显著改善。可以认为二元Copula联合分布在对落叶松人工林的胸径树高分布有优异的描述效果。
ZHOUY P, LEIZ Y, ZHAOG J,et al.Comparing different height-diameter models of Pinus sylvestris var.mongolica in sandy land[J].Journal of South China Agricultural University,2019,40(3):75-81.
FANW, XUC H, CUIJ,et al.Comparisons of height-diameter models of Chinese fir based on mixed effect in Dabie Mountain area,China[J].Chinese Journal of Applied Ecology,2017,28(9):2831-2839.
HUD H, ZHENGH Q, WEIR P,et al.Analysis of the impact of density control and fertilization on the transformation of near-mature forest into large-diameter Chinese fir trees[J].Journal of Mountain Agriculture and Biology,2024,43(3):9-14,32.
WUC Z, HONGW.A study on Weibull D.B.H.distribution of Chinese fir plantation and its optimal fitting[J].Acta Agriculturae Universitatis Jiangxiensis,1998,20(1):86-90.
HONGL X, DUG J, ZHANGQ R,et al.Weibull distributions and their dynamic predic-tions in the natural uneven-aged evergreen broad-leaf juvenile forest[J].Chinese Journal of Plant Ecology,1995,19(1):29-42.
MENGX Y.A study of the relation between D and H- distributions by using the Weibull function[J].Journal of Beijing Forestry University,1988(1):40-48.
[17]
JOHNSONN L.Bivariate distributions based on simple translation systems[J].Biometrika,1949,36(3/4):297-304.
[18]
SCHREUDERH T, HAFLEYW L.A useful bivariate distribution for describing stand structure of tree heights and diameters[J].Biometrics,1977,33(3):471-478.
[19]
LIF, ZHANGL J, DAVISC J.Modeling the joint distribution of tree diameters and heights by bivariate generalized beta distribution[J].Forest Science,2002,48(1):47-58.
[20]
SKLARA.Random variables,joint distribution functions and copulas[J].Kybernetika,1973,9:449-460.
[21]
NELSENR B.An introduction to copulas[M].New York:Springer-Verlag,2006.
[22]
MINGLIANGW, KEITHR.Bivariate distribution modeling with tree diameter and height data[J].Forest Science,2007,53(1):16-24.
[23]
BAILEYR L, DELLT R.Quantifying diameter distributions with the Weibull function[J].Forest Science,1973,19(2):97-104.
[24]
FELSENSTEINJ.Confidence limits on phylogenies:An approach using the bootstrap[J].Evolution,1985,39(4):783-791.
[25]
REYNOLDSM, BURKT E, HUANGW-C.Goodness-of-fit tests and model selection procedures for diameter distribution models[J].Forest Science,1988,34(2):373-399.
[26]
DIVISIOND A.S-Plus 6:Guide to Statistics[M].Amsterdam:Elsevier,2000.
[27]
NEWTONP F, LEIY, ZHANGS Y.Stand-level diameter distribution yield model for black spruce plantations[J].Forest Ecology and Management,2005,209(3):181-192.
[28]
OGANAF N, GORGOSO-VARELAJ J, OSHOJ S A.Modelling joint distribution of tree diameter and height using Frank and Plackett copulas[J].Journal of Forestry Research,2020,31(5):1681-1690.
[29]
GORGOSO VARELAJ, GARCÍA VILLABRILLEJ D, ROJO ALBORECAA,et al.Comparing Johnson's SBB,Weibull and Logit-Logistic bivariate distributions for modeling tree diameters and heights using copulas[J].Forest Systems,2016,25(1):1-5.