基于混合效应的湖南马尾松次生林单木生长模型

【目的】建立湖南省马尾松次生林单木断面积与材积生长模型,为林木的生长预估提供理论依据。【方法】以湖南省2014年一类清查样地中的20块马尾松次生林为研究对象,选取5个具有生物学意义的生长方程,建立马尾松断面积和材积随年龄变化的基础模型,在此基础上,加入以样地为随机效应的随机参数,构建基于混合效应的湖南马尾松次生林单木断面积和材积生长模型。【结果】断面积生长最优基础模型为Logistic方程,其确定系数(R2=0.746)和预测精度(P=98.13%)最大,残差平方和(SSE=0.025)最小;材积生长最优基础模型为Richards方程,其R2为0.703,预测精度为97.20%,SSE为1.034;混合效应模型模拟结果显示,断面积和材积生长模型的随机参数均为μ1、μ2、μ3。混合效应模型的拟合效果较基础模型有显著提升,其中断面积生长模型的R2由0.746提升到0.974,平均误差Bias由0.000 26降低到0.000 01;材积生长模型的R2由0.703提升到0.984,平均误差Bias由0.001 73降低到0.000 13。两个模型的预测精度较对应的基础模型均有所提升。【结论】含样地效应的混合效应模型拟合效果和预测精度均优于基础模型,具有更高的适用性,可为该林分的可持续经营提供科学指导。

Growth model for individual tree of secondary Pinus massoniana forest in Hunan province based on mixed effect

【Objective】Establishing growth model of individual basal area and volume of Pinus massoniana secondary forest in Hunan province, to provide theoretical basis for its forest growth prediction.【Method】20 sample plots of Pinus massoniana in the national forest inventory in Hunan province in 2014 were selected as the research objects. Five growth equations with biological significance were selected to establish the basic model of basal area and volume of Pinus massoniana varying with age. Random parameters with sample plots effect were added to construct the mixed effected model of individual basal area and volume of Pinus massoniana secondary forest in Hunan province.【Result】The optimum basic model for basal area growth was Logistic equation, with the largest determination coefficient(R2=0.746) and prediction accuracy(P=98.13%) and the smallest sum of squares of residuals(SSE=0.025). The best basic model for volume growth was Richards equation with R2 of 0.703, prediction accuracy of 97.20% and SSE of 1.034. Mixed effect model simulation results show that the random parameters of basal area and volume growth model are both μ1, μ2 and μ3. Compared with the basic model, the fitting effect of the mixed effect model is significantly improved. The R2 of the basal area growth model is increased from 0.746 to 0.974, the Bias is reduced from 0.000 26 to 0.000 01, the R2 of the volume growth model is increased from 0.703 to 0.984, and the average error is reduced from 0.001 73 to 0.000 13. The prediction accuracy of the two models are higher than that of the basic model.【Conclusion】The fitting effect and prediction accuracy of mixed effect model with sample plot effect are better than those of the basic model, which has higher applicability and can provide scientific guidance for the sustainable management of this stand.