基于混合效应模型的可变密度可变指数削度方程构建研究

发布时间：2018-07-04 13:58

本文选题：杉木 + 削度方程　；参考：《中国林业科学研究院》2017年硕士论文

【摘要】：杉木(Cunninghamia lanceolata)是我国重要乡土针叶用材树种,第八次全国森林资源清查显示杉木人工林面积和蓄积现已达到全国人工林优势树种的首位。因此,在森林资源调查中需要准确的评估和预测杉木人工林的蓄积量。削度方程在精确预测树干直径和材积有着非常重要的作用。目前国内缺乏对基础模型的系统性比较分析,尤其少见针对杉木的可变指数削度方程的构建,可变密度可变指数削度方程的构建。故本文以江西年珠林场密度试验林为研究对象,拟利用混合效应模型建立针对杉木的可变密度可变指数削度方程和材积预测模型。主要研究结论如下:(1)可变指数削度方程拟合优度最好,其次是分段削度方程,最后是简单削度方程。多个可变指数削度方程彼此间的拟合优度没有较为明显区别。从30个削度方程中选择出Zeng(1997)、Bi(2000)、Kozak(2004)、Sharma(2004)4个拟合优度最高的可变指数削度方程作为基础模型。(2)针对选择的基础模型建立基于样地效应、基于样木效应和基于嵌套两水平效应的混合效应模型。混合效应模型可以提高模型拟合精度,考虑样地效应时,混合效应模型的调整决定系数提高0.0016~0.0020,但不能消除观测值之间的自相关性。考虑样木效应和嵌套两水平的混合模型的调整决定系数均比基于样地效应提高0.0104~0.0117,且可以消除大部分观测值之间的自相关性。Kozak(2004)可变指数削度方程在基于样木效应和嵌套两水平混合模型中表现了最高的预估精度,但由于基于样地效应的混合模型结构比嵌套两水平混合模型简单,故选择基于样木效应的Kozak(2004)混合模型作为最优削度方程。(3)在结论(2)的基础上建立了包含密度因子的可变指数削度方程。在削度方程中加入密度因子后,提高了削度方程的拟合精度,调整决定系数达到0.9945。不同密度下的树干表现出不同的削度,树干的削度随着密度的增大而逐渐减少,减少的程度也随密度的增大而降低。(4)利用建立的包含密度的可变指数削度方程对杉木进行材积预测,可变指数削度方程预测法的平均偏差、均方根误差、平均绝对偏差和相对误差均小于二元材积表法和简单削度方程积分求积法。三种方法在高径比和胸高形树较低的范围内预测材积精度较差,但是在高径比和胸高形树较高范围内,三种预测材积方法精度较高。三种方法中,包含密度的可变指数削度方程预测材积精度最高。
[Abstract]:Chinese fir (Cunninghamia lanceolata) is an important native coniferous timber species in China. The eighth national forest resource inventory shows that the area and accumulation of Chinese fir plantation have now reached the top of the national dominant tree species. Therefore, it is necessary to accurately evaluate and predict the volume of Chinese fir plantation in forest resource survey. The truncation equation plays an important role in accurately predicting the diameter and volume of trunk. At present, there is a lack of systematic comparative analysis of the basic model, especially the construction of variable index taper equation and variable density variable index taper equation for Cunninghamia lanceolata (Cunninghamia lanceolata). Therefore, this paper takes the density test forest of Nianzhu Forest Farm in Jiangxi Province as the research object, and uses the mixed effect model to establish the variable density variable index taper equation and volume prediction model for Chinese fir. The main conclusions are as follows: (1) the variable exponential taper equation is the best, followed by the piecewise taper equation, and finally the simple taper equation. There is no obvious difference in the goodness of fit between several variable exponential taper equations. Four variable exponential taper equations of Zeng (1997) Bi (2000) Kozak (2004) Kozak (2004) and Sharma (2004) were selected from 30 taper equations as the basic model. (2) the mixed effect models based on sample effect, sample wood effect and nested two-level effect were established for the selected basic model. The mixed effect model can improve the fitting accuracy of the model. When the sample effect is considered, the adjustment decision coefficient of the mixed effect model is increased by 0.0016 ~ 0.0020, but the autocorrelation between the observed values can not be eliminated. The adjusted decision coefficients of the mixed model considering both the sample effect and the nested level are 0.0104 / 0.0117 higher than those based on the sample effect, and can eliminate the autocorrelation between most of the observed values. Kozak (2004) variable exponential taper equation is based on the sample wood effect. In the nested two-level mixing model, the prediction accuracy is the highest. However, the structure of hybrid model based on sample effect is simpler than that of nested two-level hybrid model. Therefore, the Kozak (2004) mixed model based on sample wood effect is chosen as the optimal taper equation. (3) based on conclusion (2), the variable exponential taper equation with density factor is established. With the addition of density factor into the taper equation, the fitting accuracy of the taper equation is improved, and the adjusting decision coefficient reaches 0.9945. The trunks with different densities showed different taper, and the taper of the trunks decreased with the increase of the density. The degree of reduction also decreases with the increase of density. (4) the volume of Chinese fir is predicted by using the variable exponent taper equation including density, the mean deviation and root mean square error of the prediction method of variable exponential taper equation are obtained. The mean absolute deviation and relative error are smaller than the binary volume table method and the integral quadrature method of the simple taper equation. The accuracy of the three methods in predicting volume is poor in the range of height to diameter ratio and chest height tree, but the accuracy of the three methods is higher in the range of height to diameter ratio and chest height tree. Among the three methods, the variable exponential taper equation including density has the highest prediction accuracy.
【学位授予单位】：中国林业科学研究院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S791.27

【参考文献】