林分生长率非线性混合模型的构建

以浙江省丽水市为研究地，利用连续2期固定样地监测数据为建模样本，以林分平均胸径为自变量，以保留木进界木生长量之和减去枯损木消耗量后的林分蓄积年生长率为因变量，先采用联立方程组方法进行模型拟合和初选，再采用非线性混合模型方法对选定模型进行固定效应和随机效应分析，构建了胸径与生长率混合模型的联立方程组。结果表明：利用联立方程组方法构建的模型及参数，反映了建模总体生长率的平均水平，但由于未考虑起源、龄组、树种组等随机效应，拟合效果欠佳，需要采用混合模型方法进一步分析拟合。联立方程组方法拟合的模型参数，已能很好解释混合模型的固定效应；混合模型的随机效应对生长率具有显著作用，其参数值差异达显著水平。利用混合模型的随机变量各类目随机效应参数之和为零的特点，可在联立方程组模型中加入随机效应参数值建立生长率混合模型，构建胸径模型和生长率混合模型的联立方程组。模型适用性检验表明，其预测值与实测值无系统偏差，方程组具有较好的适用性；利用同期小班数据，对全市林分生长进行模型预测，其生长量和生长率与同期固定样地监测结果相比，准确度分别达到91.5%和98.7%。

Construction of Nonlinear Mixed Model of Stand Growth Rate

The simultaneous equations of diameter at breast height and mixed growth rate were established based on the data of 2 periods of fixed sample plots in Lishui, Zhejiang. The model takes the average diameter at breast height as the independent variable, and the annual growth rate of stand volume as the dependent variable. First, simultaneous equation method was used to model fitting and primary selection, then the nonlinear mixed model method was used to analyze the fixed effect and random effect of the selected model. The results showed that the model and parameters constructed by simultaneous equations reflect the average growth rate of the model. However, without considering the random effects of origin, age group and tree species group, the fitting effect was not good, further analysis was needed using the mixed model method. The model parameters fitted by simultaneous equations method could explained the fixed effect of the mixed model well. The random effect of mixed model had significant effect on the growth rate, and the difference of parameters was significant. Due to the sum of random effect parameters of random variables of mixed model was zero, the random effect parameter was used to construct the simultaneous equations model to establish the growth rate mixed model, and construct the simultaneous equations of DBH model and growth rate mixed model. The applicability test of the test samples shows that the predicted and measured values of the simultaneous equations have no systematic deviation, and the system of equations has strong applicability. Based on the data of the same period, the growth of the whole city was predicted by the model. Compared with the same period of monitoring results of fixed sample plots, forecast accuracy of the number and rate of growth is 91.5%, 98.7% respectively.