Individual-tree basal area growth models of three main tree species in Larix gmelinii natural forests
[Objective]Developing individual-tree basal area growth models for Larix gmelinii,Betula platyphylla,Quercus spp.based on mixed-effects model method,in order to improve the prediction accuracy of models and provide some reference of forest dynamics of L.gmelinii mixed-species natural forests.[Method]Based on the data from 626 sample plots of L.gmelinii mixed-species natural forests(Chinese National Forest Inventory in 2003,2008,2013 and 2018)in the Daxing'an mountains,Inner Mongolia Autonomous Region,study subjects were divided into three main species groups,i.e.L.gmelinii,B.platyphylla and Quercus spp..The individual-tree basal area growth models were constructed by using the multiple stepped-regression method.Then,the optimal based model was selected,and based on that,the random effects of within-plot were introduced to mixed-effects model for basal area prediction.Finally,the fitting effect was compared with the based model.[Result]The reciprocal transformation of initial diameter at breast height(1/D),the number of trees per hectare(N),the sum of the basal area in trees with DS larger than the subject tree's D(BL),the ratio of BL and D(BL/D),elevation(E)and the combination of slope and aspect(Tc)had significant effects on the growth of individual-tree basal area in the L.gmelinii mixed-species natural forests(P＜0.05).The AIC and BIC of the mixed-effects models were obviously decreased,while LogLik were increased.Mixed-effects models,which were considering heteroscedasticity function and autocorrelation structure,showed an improved prediction accuracy with the Bias and RMSE were lower than that of based model.The R2 of optimal mixed-effects model of L.gmelinii,B.platyphylla,Quercus spp.increased from 0.435 to 0.658,from 0.354 to 0.489,and from 0.307 to 0.379,respectively.[Conclusion]Compared with the based models,the mixed-effects models considering the hierarchical structure show an improved performance,and provide biological significance and statistical reliability,which help to predict the dynamic growth of L.gmelinii mixed-species natural forests in Daxing'an mountain of Inner Mongolia.
万方数据
维普
