Trunk Density(Wood Density) Influencing Factors And Biomass Growth Model Of Three Native Broadleaf Tree Species In Guangdong Province

Forest is the main body of terrestrial ecosystem on which human beings live,and plays a key role in the process of global climate change.In this context,as an important indicator of forest quality assessment,forest biomass measurement has become a research hotspot.For the estimation of forest biomass in large areas,biomass models are relatively effective and accurate means and methods.The estimation of biomass in large area and stand depends on individual tree biomass model,which is the basis of estimating and predicting forest biomass.Cinnamomum camphora,Schima superba and Liquidambar formosana are three important native broad-leaved tree species in Guangdong Province.They have huge carbon sink potential and high timber and economic value.Based on the measured data of 90 fallen trees and 40 analytic trees of the three tree species,the effects of different factors on tree trunk density and wood density were explored by using multivariate variance analysis of main effects and general linear model.The factors related to tree trunk density and wood density were screened out from 30 factors of 5 categories.Then,the effect of different factors on trunk density and wood density of three tree species was calculated by using the enhanced regression tree(BRT)method.Then,based on six commonly used tree growth equations,Richard equation,Mitscherlich equation,Gompertz equation,Korf equation,logistic equation and Schumacher equation,the above-ground biomass growth process of tree species was taken as dependent variable,and tree growth age as independent variable.The basic model of biomass growth was established,and the optimal basic model of biomass growth was selected.On this basis,dumb variables are modeled by dumb variables based on age group and origin,principal component analysis and correspondence analysis.Four evaluation indices,including determinant coefficient R~2,standard error SEE,average prediction error MPE and total relative error TRE,and rationality of model parameters,were used to evaluate the effects of each model.Through this study,the following conclusions are drawn:(1)The main factors affecting the trunk density of Cinnamomum camphora are vegetation type,height under branches,DBH,total vegetation coverage and crown width from east to west;the main factors affecting the trunk density of Schima superba are city and vegetation type;the main factors affecting the trunk density of Liquidambar formosana are aspect,altitude and average height.From a single factor point of view,the main influencing factors of trunk density of three tree species are different,and there are no common main influencing factors.(2)The main influencing factors of wood density of three tree species are also different,but the common main influencing factors are high under branches,and their relative contribution rates are similar,all of them are about 10%.The main factors of wood density of each tree species from big to small are Cinnamomum camphora(height under branch,vegetation type,altitude,vegetation coverage,average height,shrub coverage,age,DBH,forest category,soil thickness),Schima superba(age,herb coverage,height under branch,average DBH,soil thickness and vegetation type),and Liquidambar formosana(slope aspect,altitude,average height,height under branch).(3)Stand factor and single tree factor are the main factors affecting trunk density and wood density of Cinnamomum camphora,Schima superba and Liquidambar formosana,but their effects on different tree species are different.Stand factor and single tree factor are the dominant factors affecting the trunk density and wood density of Cinnamomum camphora,and their relative contribution rates are 87.04%and 76.92%respectively.Stand factor,single tree factor and region factor are the main factors affecting the trunk density of Schima superba,and the total relative contribution rate is 79.96%.The main factors affecting the wood density of Schima superba are stand factor,single tree factor and soil factor,and the total relative contribution rate is 83.04%.Terrain factor,stand factors and single tree factors are the main factors affecting the trunk density and wood density of Liquidambar formosana.Their relative contribution rates are 83.98%and 92.70%respectively.(4)The basic biomass growth model of Cinnamomum camphora was the best for Richard equation,with R~2of 0.443 and SEE of 99.73 kg.Based on Richard equation,the dummy variable model of age group and origin of camphor tree only increased slightly,about 0.5.The model R~2based on dimension reduction method is significantly improved.The dumb variable model R~2based on principal component analysis is more than 0.65,and the corresponding analysis model R~2is 0.56.The dumb variables of age group and principal component analysis are introduced into Richard equation parameter a and B.The model achieves the best result with R~2reaching 0.72,which is significantly lower than the optimal basic model SEE,which is71.29 kg.(5)Korf equation was the best basic biomass growth model for Schima superba,R~2was0.557,SEE was 79.09kg.Schima superba based on age group dumb variable,the effect was significant,R~2increased to 0.70,SEE decreased to 65 kg.The dumb variable model based on the factor introduced by correspondence analysis method is better than the model based on the factor introduced by principal component analysis.After many comparisons,it is concluded that the model is the best by introducing dumb variables of classification results after corresponding analysis on parameter B of Korf equation and dumb variables representing age group on C.Compared with the optimal basic model,R~2increased by 0.21 to 0.77 and SEE decreased by 22.03 kg to 57.05 kg.(6)Gompertz function was the optimal basic biomss growth function for Liquidam bar formosana with R~2of 0.6595 and SEE of 76.11kg.It got bad performance when t he age class and origin as the dummy variable were introduced to the basic function w ith the increment of R~2less than 0.01.The dummy variable based on correspondence a nalysis performed better than that based on principal component analysis and the differe nce of R~2was 0.018 between them.Through the comparison of different models,the model that the classified results of correspondence analysis and age class respectively a s the dummy variable introduced to parameter b and parameter c performed best.Com pared with the optimal basic function R~2increased by 0.15 and SEE dropped by 19.479kg to 56.63kg.