华北落叶松天然林单木冠幅模型研究
发布时间：2018-09-10 12:24
【摘要】：华北落叶松(Larixprincipis-rupprechtii)是我国华北地区山地的主要造林树种。其优点众多,生长速度快、木材材质好、用途广、腐朽耐力强,是营造经济生态林的良好树种,同时也是杰出的防护林树种。树冠是树木进行光合作用和积累能量的重要场所,它反映了树木的生长活力和竞争力,是树木一个重要的健康指标。冠幅是描述树冠生长的重要指标也是森林生长收获预估模型的重要预测变量。因此准确预估冠幅对森林可持续经营和森林生态研究极其重要。本文以多种树木林分因子为协变量构建了一般冠幅-胸径预测模型,并且分析了地域效应和区组效应以及嵌套在区组里的样地效应对华北落叶松冠幅影响,以此为基础使用了非线性混合效应模型构建了相应的冠幅模型,最后还考虑了冠幅与东西南北四个方向的冠径的相关性,并使用了四种可加性模型方法构建了相应的模型系统,四种模型系统都解决了冠幅与东西南北四个方向的冠径的相容性;最后通过综合比较确定了最优的可加性冠幅模型系统。主要研究结果如下:(1)三参数的逻辑斯蒂模型:CW=a/1+exp[b+cln(D+1)]能较好反映华北落叶松天然次生林冠幅和直径之间的非线性关系,与其它候选模型相比,该模型有较高的拟合精度,并且模型各参数都具有一定的生物学意义;对象木冠长(CL)、对象木树高(H)和每公顷株数(M)对冠幅影响较大,当这些因子作为预测变量时能明显改进模型的预测精度;得到的改进后的冠幅模型表达式为:CW = 1.4875-0.031 1H+exp[(-3.6416-0.0002M)+(2.4346+ 0.0045CL)ln(D+1)]通过大量实验数据验证,该模型具有较高的预测精度。(2)研究发现区组效应和嵌套在区组里面的样地效应对华北落叶松的随机影响较大,当模型考虑这些随机影响时模型预测精度能进一步显著提高;,指数方差函数且预测变量为胸高直径能有效剔除模型的异方差,表达式为:var(εijk= σ2exp(2γxijk);利用所构建的嵌套两水平非线性混合效应模型预测冠幅时,利用随机抽取的4株样地计算随机效应参数效果较好。当地域效应作用在固定效应参数β1和β5上时,模型对应的AIC(5425)最小而LogLik(-2697)最大最终构建了华北落叶天然林非线性回归效应单木冠幅模型。(3)以模型(4-3)为基础模型,使用非线性联立方程组(NSE)、非线性似然不相关回归(NSUR)、比例平差法(AP)和最小二乘法独立回归(OLSSR)方法构建冠幅可加性模型系统,这几种方法都能有效的考虑总冠幅和各树冠半径之间的相关性。通过综合对比这几种可加性模型系统,对于总冠幅,分级联合控制平差法构建的冠幅模型系统对应的指标δ,RMSE和TRE均要低于NSUR、AP和OLSSR模型系统;所以在可加性模型构建冠幅模型方法中,分级联合控制平差法构建的冠幅可加性模型系统拟合效果最好。
[Abstract]:Larch of North China (Larixprincipis-rupprechtii) is the main afforestation tree species in the mountainous area of North China. It has many advantages, such as fast growth rate, good wood material, wide use and strong decadent endurance. It is a good tree species for economic and ecological forest, and is also an outstanding shelterbelt tree species. Tree crown is an important place for photosynthesis and energy accumulation of trees. It reflects the growth vitality and competitiveness of trees and is an important health index of trees. Canopy width is an important index to describe tree crown growth and an important predictor of forest growth and harvest prediction model. Therefore, it is very important for forest sustainable management and forest ecology to estimate the crown accurately. In this paper, a general prediction model of crown width and DBH was constructed using a variety of tree stand factors as covariables, and the effects of region effect, block effect and sample plot effect nested in block on crown width of Larix gmelini were analyzed. On this basis, the corresponding crown size model is constructed by using the nonlinear mixed effect model. Finally, the correlation between the crown width and the crown diameter in the four directions of the east, west, north and south is considered, and the corresponding model system is constructed by using four additive model methods. All the four model systems have solved the compatibility between the crown width and the crown diameter in the four directions from east to west, and the optimal additive crown model system has been determined by comprehensive comparison. The main results are as follows: (1) the three-parameter logical Stirt model: CWSA / 1 exp [b cln (D 1] can better reflect the nonlinear relationship between crown width and diameter of natural secondary forest of Larix gmelini. Compared with other candidate models, this model has higher fitting accuracy. All the parameters of the model have certain biological significance, the tree height (H) and the number of trees per hectare (M) have a great influence on the crown width, and the prediction accuracy of the model can be improved obviously when these factors are used as the prediction variables. The expression of the improved crown model is: CW = 1.4875-0.031 1H exp [(-3.6416-0.0002M) (2.4346 0.0045CL) ln (D 1)]. The model has a high prediction accuracy. (2) the block effect and the sample effect nested in the block have a great influence on the random effect of Larix gmelini. When these random effects are taken into account, the prediction accuracy of the model can be further improved significantly, and the exponential variance function and the prediction variable are sternum height diameter, which can effectively eliminate the heteroscedasticity of the model. The expression is: var (蔚 ijk= 蟽 2exp) (2 纬 xijk);). When the constructed nested two-level nonlinear mixed effect model is used to predict the crown amplitude, it is better to calculate the random effect parameters by using the random sampling plots. When the regional effect is on the fixed effect parameters 尾 _ 1 and 尾 _ 5, the corresponding AIC (5425) is the smallest and the LogLik (-2697) maximum is the largest. Finally, the model of the nonlinear regression effect of the North China deciduous natural forest is constructed. (3) the model (4-3) is taken as the basic model. Using (NSE), nonlinear likelihood uncorrelated regression (NSUR), proportional adjustment method (AP) and least square independent regression (OLSSR) method to construct the crown additive model system. These methods can effectively consider the correlation between total crown size and crown radius. Through the comprehensive comparison of these additive model systems, for the total crown amplitude, the corresponding indexes of the crown model system constructed by the hierarchical combined control adjustment method are lower than those of the NSUR,AP and OLSSR model systems. Therefore, in the method of building crown amplitude model with additive model, the fitting effect of the model system based on hierarchical combined control adjustment method is the best.
【学位授予单位】：中南林业科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S758
，
本文编号：2234436
[Abstract]:Larch of North China (Larixprincipis-rupprechtii) is the main afforestation tree species in the mountainous area of North China. It has many advantages, such as fast growth rate, good wood material, wide use and strong decadent endurance. It is a good tree species for economic and ecological forest, and is also an outstanding shelterbelt tree species. Tree crown is an important place for photosynthesis and energy accumulation of trees. It reflects the growth vitality and competitiveness of trees and is an important health index of trees. Canopy width is an important index to describe tree crown growth and an important predictor of forest growth and harvest prediction model. Therefore, it is very important for forest sustainable management and forest ecology to estimate the crown accurately. In this paper, a general prediction model of crown width and DBH was constructed using a variety of tree stand factors as covariables, and the effects of region effect, block effect and sample plot effect nested in block on crown width of Larix gmelini were analyzed. On this basis, the corresponding crown size model is constructed by using the nonlinear mixed effect model. Finally, the correlation between the crown width and the crown diameter in the four directions of the east, west, north and south is considered, and the corresponding model system is constructed by using four additive model methods. All the four model systems have solved the compatibility between the crown width and the crown diameter in the four directions from east to west, and the optimal additive crown model system has been determined by comprehensive comparison. The main results are as follows: (1) the three-parameter logical Stirt model: CWSA / 1 exp [b cln (D 1] can better reflect the nonlinear relationship between crown width and diameter of natural secondary forest of Larix gmelini. Compared with other candidate models, this model has higher fitting accuracy. All the parameters of the model have certain biological significance, the tree height (H) and the number of trees per hectare (M) have a great influence on the crown width, and the prediction accuracy of the model can be improved obviously when these factors are used as the prediction variables. The expression of the improved crown model is: CW = 1.4875-0.031 1H exp [(-3.6416-0.0002M) (2.4346 0.0045CL) ln (D 1)]. The model has a high prediction accuracy. (2) the block effect and the sample effect nested in the block have a great influence on the random effect of Larix gmelini. When these random effects are taken into account, the prediction accuracy of the model can be further improved significantly, and the exponential variance function and the prediction variable are sternum height diameter, which can effectively eliminate the heteroscedasticity of the model. The expression is: var (蔚 ijk= 蟽 2exp) (2 纬 xijk);). When the constructed nested two-level nonlinear mixed effect model is used to predict the crown amplitude, it is better to calculate the random effect parameters by using the random sampling plots. When the regional effect is on the fixed effect parameters 尾 _ 1 and 尾 _ 5, the corresponding AIC (5425) is the smallest and the LogLik (-2697) maximum is the largest. Finally, the model of the nonlinear regression effect of the North China deciduous natural forest is constructed. (3) the model (4-3) is taken as the basic model. Using (NSE), nonlinear likelihood uncorrelated regression (NSUR), proportional adjustment method (AP) and least square independent regression (OLSSR) method to construct the crown additive model system. These methods can effectively consider the correlation between total crown size and crown radius. Through the comprehensive comparison of these additive model systems, for the total crown amplitude, the corresponding indexes of the crown model system constructed by the hierarchical combined control adjustment method are lower than those of the NSUR,AP and OLSSR model systems. Therefore, in the method of building crown amplitude model with additive model, the fitting effect of the model system based on hierarchical combined control adjustment method is the best.
【学位授予单位】：中南林业科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S758
，
本文编号：2234436
本文链接：https://www.wllwen.com/shoufeilunwen/zaizhiyanjiusheng/2234436.html
