思茅松天然林林分生物量混合效应模型构建

本研究以云南省普洱市的思茅松天然林为对象,调查了3个位点45块样地的林分地上、根系和总生物量。以幂函数模型为基础构建林分生物量的基本模型;采用混合效应模型技术,考虑区域效应随机效应,选择基本混合效应模型,并分析模型的方差和协方差结构,分别构建3个维量的区域效应随机效应的混合效应模型;考虑林分因子、地形因子和气象因子固定效应,构建含环境因子固定效应和区域效应随机效应的林分生物量混合效应模型。所有模型均采用拟合指标和独立检验指标进行评价。结果表明:1) 从模型拟合情况看,考虑区域效应的随机效应模型均能显著提高一般回归模型的精度;在3类含环境因子固定效应模型中,含地形因子固定效应的区域混合效应模型均具有最低的AIC和BIC值,表现最好;2) 就模型独立性检验看,除地形因子固定效应的林分根系混合效应模型外,其余模型均优于一般回归模型;考虑环境因子固定效应的混合效应模型与普通区域效应混合模型相比,各个维量模型的独立性检验指标表现不一,但总体上差异不大;3) 综合考虑模型拟合和独立性检验结果,除林分根系生物量选择普通区域效应混合模型外,另2个维量均选择含地形因子固定效应和区域效应随机效应的混合效应模型。 

Building mixed effect models of stand biomass for Simao pine (Pinus kesiya var. langbianensis) natural forest

In this paper we took natural Simao pine ( Pinus kesiya var. langbianensis) forest as the research object, and investigated the aboveground, root and total biomass of 45 plots of at three typical sites ( Tongguan town of Mojiang County, Yunxian town of Simao District, and Nuofu town of Lancang County) in Pu'er City, Yunnan Province. Firstly, we chose the best power function to the basic model. Secondly, considering random effect of the regional effect we constructed the mixed effects models of the biomass components of stand using technology of mixed effects models, and analyzed the variance and covariance structures of the models. Finally, based on the basic mixed effects models of the components, We constructed the mixed effects models including fixed effects from three types of environmental factors (including stand, topographic and climate factors) respectively. The models were evaluated by fitting and independence test indices. The fitting indices include logLik, Akaike information criterion ( AIC ) and Bayesian information criterion ( BIC) , and the test indices include sum relative error ( SRE) , mean relative error ( MRE) , absolute mean relative error ( AMRE) and prediction precision ( p) . The results showed:(1) For the models fitting, the mixed model considering random effect of regional effect were significantly better than the ordinary models, and the mixed models including the fixed effect of environmental factors were better than the ordinary mixed models. Among the models including the fixed effects form three types of environmental factors, the models including topographic factors were the best models because of the lowest values for AIC and BIC. (2) For the independence test of models, except for the mixed models of stand root biomass including the fixed effects of topographic factors, the other mixed models were better than the ordinary models. Compared the mixed models including the fixed effect of environmental factors and the ordinary mixed models, the performance were different for three components, but for each component the differences among the models were small. (3) The best model for root biomass of stand was the mixed effects models only considering the random effect of regional effect, but for the other components ( including aboveground biomass and the total biomass of stand) , the best models were both the mixed effects models including the fixed effects of topographic factors and random effect of regional effect.