Study On Bark Thickness And Bark Factor Of Larch Plantation Based On Mixed Model

This paper takes the larch plantation in Mengjiagang Forest Farm of Jiamusi City,Heilongjiang Province as the research object.Analytical data of 49 analytical wood trunks based on 7 larch plantation standard plots,a total of 1186 disc data were obtained from the annual ring image analysis system WinDENDOR^TMV6.5.The changes of bark thickness(BT)and bark factor(DIB/DOB)with tree height(HT)and DBH were analyzed.Using the PROC MIXED module in SAS 9.4 software,based on the selected optimal basic model,considering the tree effect and plot effect,a linear mixed-effect prediction model of the bark thickness of the larch plantation(bark factor,the bark thickness at arbitrary height)was constructed.The model evaluation indicators used Akaike Information Criterion(AIC),Bayesian Information Criterion(BIC),-2 log likelihood value(-2LL)and likelihood ratio test(LRT).Firstly,the results of the bark factor indicate,the bark factor remains substantially constant along the vertical direction of the trunk with increasing relative height.When the relative height exceeds 80%,the bark factor decreases as the relative height increases.There was no significant correlation between bark factor and DBH.The mixed effect model of any combination of parameters has a better fitting effect than the traditional model.For the bark factor model,based on the tree effect,b1,b2,b4,ie relative height(h/HT)and square of relative height((h/HT)2),the ratio of the diameter of the skin to the diameter of the breast(DBHIB/DBH)is combined as the optimal mixed effect model.After adjustment,the coefficient R2 is determined to be 0.734,the average absolute error MAE is 0.022,and the root mean square error RMSE is 0.032.Based on the plot effect,the combination of b1,b2,ie relative height(h/HT)and square of relative height((h/HT)2),is the optimal mixed effect model.After adjustment,the coefficient R2 is determined to be 0.696,the average absolute error MAE is 0.022,and the root mean square error RMSE is 0.034.All optimal models have the best fitting effect when they have a non-structural(UN)variance-covariance matrix.Secondly,the results of bark thickness at any height indicate,The thickness of the bark varies with the relative height and changes in an S-shaped curve,and the overall thickness is negatively correlated with the relative height.In the interval of 0y20%relative height,the bark thickness decreases rapidly with the increase of height.In the interval of 20%-80%,the bark thickness decreases with the increase of height,and the relative height is 80%to the tip,the rate of decrease in bark thickness is accelerated.There is a significant positive correlation between bark thickness and DBH at any height.The mixed effect model of any combination of parameters has a better fitting effect than the traditional model.For the bark thickness model at any height,the optimal mixing model is based on the combination of b1,b2,the relative height(h/HT)and the skin diameter(dob)at any height.After adjustment,the coefficieit R2 is determined to be 0.878,the average absolute error MAE is 0.053,and the root mean square error RMSE is 0.083.Based on the plot effect,b0,b2,b3,that is,the intercept term,the arbitrary height of the strip diameter(dob),and the aspect ratio(DBH/HT)are the optimal mixing models.After adjustment,the coefficient R2 is determined to be 0.820,the average absolute error MAE is 0.061,and the root mean square error RMSE is 0.093.All optimal models have the best fitting effect when they have a non-structural(UN)variance-covariance matrix.Whether it is a bark factor or a bark thickness model,the effect of the tree effect on the model is greatest.The prediction accuracy of the mixed-effects model is significantly improved compared to the traditional regression model.The construction of the mixed effect model of bark factor and bark thickness can more accurately predict the debark volume of the tree,the bark volume,and provide help and guidance for the actual forest management and wood production.