三个主要树种单木生物量及其器官分配模型

目前已有不同方法构建生物量相容性模型,但基于非线性似乎不相关回归估计法实现不同树种生物量模型相容性及各器官生物量分配模型的研究较少。因此,论文以塞罕坝华北落叶松、油松、白桦3种林分为对象,基于非线性似乎不相关回归估计法和广义多项Logit模型,建立了包含哑变量的非线性可加生物量模型及各器官生物量分配模型。结果表明：不同树种树干生物量模型确定系数均大于0.90,树枝、树叶、树根生物量模型确定系数在0.77~0.93范围内,各器官生物量均方根误差和绝对误差分别在2.68~17.19 kg/株和0.83~1.39 kg/株范围内,经过检验不同树种各器官生物量模型均能满足精度需求。不同树种广义多项Logit分配模型,通过似然比检验、比分检验和Wald检验均达到显著水平（P<0.001）,各器官参数均表现为显著水平（P<0.05）。不同树种树干、树枝、树叶及树根生物量比例分别在0.76~0.87、0.07~0.11、0.02~0.07、0.04~0.07范围内。包含哑变量的非线性似乎不相关生物量模型及广义多项Logit各器官分配模型,实现了生物量模型在不同树种间的通用性,并对森林生物量器官分配格局研究提供了科学参考。

Models for Estimating Biomass and Its Distribution in Organs of Three Main Tree Species

At present, there are different methods to construct compatible models of biomass. However, there are few reports on the compatibility of models for different tree species and the allocation model of component-specific biomass based on the nonlinear seemingly unrelated regression method. In this paper, the research objects were Larix principis-rupprechtii, Pinus tabuliformis and Betula platyphylla forests in Saihanba National Forest Farm, Hebei Province, China. Based on the nonlinear seemingly unrelated method and the generalized multinomial Logit model, the nonlinear additive models of biomass with dummy variables and allocation models of biomass in organs were established. The results showed that the discriminant coefficients (R²) of stem biomass models for different tree species were all higher than 0.90, and the root mean square errors (RMSE) and absolute errors were in the range of 16.68-17.19 and 0.84-1.07; in the biomass models of branch, leaf and root, the discriminant coefficients (R²) were in the range of 0.77-0.93, and the root mean square errors (RMSE) and absolute errors were in the range of 2.58-12.18 and 0.83-1.39. After inspection, the biomass models of different tree species can meet the precision requirement. The generalized multinomial Logit allocation models of different species all reached a significant level (P< 0.001) after Likelihood ratio test, Score test and Wald test. The parameters of each organ were significant (P < 0.05). The proportions of biomass in trunks, branches, leaves and roots of different tree species were in the range of 0.76-0.87, 0.07-0.11, 0.02-0.07 and 0.04-0.07, respectively. The models of nonlinear seemingly unrelated regression with dummy variables and the generalized multinomial Logit model of biomass allocation in organs can be used universally among different tree species and provide scientific basis for the study of forest biomass allocation pattern.