Estimating The Total Height Of Individual Trees Based On Log Data From Cut-to-length Harvesters In Australian Pinus Radiata Plantations

As the primary source of big data for forest management,cut-to-length(CTL)harvesters have been increasingly adopted and widely utilised in both natural and plantation forests worldwide.Usually,harvester data contains the length and end diameters of individual logs,while the stump height and length of the top section of a tree are unable to get.According to the variable got in harvester data,the total height of individual trees cannot be calculated.Based on the situation,this study includes(1)harvester data screening and interactive exploratory data analysis,(2)converting more than 3000 detailed taper measurements trees to harvester data through improved cut-to-length(CTL)simulation and(3)building the new CTL model of estimating total tree height with the input variable of the DBH0B,the total log length and SEDOB TL of individual tree.To compare the best estimation form,this study transformed the model into linear and nonlinear,and the estimation of the different forms considered the autocorrelations and heteroscedasticity in the data.In order to validate,the study used cross-validation leaving one out,and compared the New CTL model with Varjo’s model by the benchmarking statistics of the weighted mean squared error of prediction(WMSEP).As a result,New CTL model in nonlinear form generally shows better prediction accuracy than linear form and Varjo’s model in both linear and nonlinear form,with the 1.2%of WMSEP in New CTL model linear form,the 11.2%of WMSEP in Varjo’s model nonlinear form and 20.4%of WMSEP in Varjo’s model linear form.Estimating the total height of each harvested tree is the necessary first step for the most effective use of harvester data.The development of New CTL model contains the essential elements of model such as model formulation,variable selection,parameter estimation,model testing and statistical validation,and the New CTL model is an improved and more efficient approach in the prediction of total tree height of individual tree stems.