Study On Standing Volume Model Of Pinus Taiwanensis In Henan

Forestry data table is a table format and map form complied by using mathematical modeling method based on stand tree measurement factor or site factor.It plays an important role in forest cultivation and forest resource management.As a kind of forestry data tables,volume table is the basis of forest resources monitoring and investigation.Pinus taiwanensis is one of the unique main timber tree species in China,which mainly distributed in Huangbai mountain forest farm and small distributed in Shihe District,Luoshan County,Xinxian County and Shangcheng of Xinyang City in Henan Province.The planting area of Pinus taiwanensis accounts for 49.23%of the total area reaching 31526 hectares in Huangbaishan Forest Farm.However,there is no standing wood volume table of Pinus taiwanensis for reference in Henan Province,so other pines volume table was used in practice in the investigation of Pinus taiwanensis resources,thus leading to a larger deviation.In this study,we carried out a experiment of Huangshan pine standing wood volume investigation and its model establishment,and we hope it may provide a support for improving the accuracy of Pinus taiwanensis resource investigation in Henan Province.In this study,all 1102 Pinus taiwanensis samples in the Ta-pieh Mountains area,where the main distribution area of Pinus taiwanensis in Henan Province,were took as the research object.According to the principle that the test sample data shall not be less than one-third of the modeling sample data and the diameter order distribution,the data of 1076 Pinus taiwanensis samples screened by abnormal samples are divided into 776 modeling sample data and 300 inspection sample data,the least square method was used to estimate and analyze the regression of 19 binary standing volume models,6 unitary standing volume models and 7 ground diameter standing volume models of Pinus taiwanensis;the binary,unitary and ground diameter multilayer perceptron neural network models of Pinus taiwanensis were constructed by using multilayer perceptron neural network algorithm,and the sum of deviation square SSR and determination coefficient R² were used,the total relative error RS,relative error absolute value rea,relative error absolute value average reaa and residual distribution were used to optimize the nonlinear binary,unitary,ground diameter standing volume model and multilayer perceptron neural network model of Pinus taiwanensis;three test methods were used to test the applicability of the model:the total relative error RS,the average value of the absolute value of the relative error REAA and F.The model was tested by using the optimal nonlinear binary,unitary,ground diameter standing volume model and multi-layer perceptron neural network model of Pinus taiwanensis;using the determination coefficient R²、root mean square error RMSE,relative root mean square error RMSE%and mean absolute error MAE are four precision indexes,which are divided into modeling samples and test samples.The accuracy of the nonlinear binary,unitary,ground diameter Standing Volume optimal model of Pinus taiwanensis and the optimal model of multi-layer perceptron neural network are compared and analyzed.The main results are as follows:1.After constructing and optimizing the nonlinear binary volume model of 19 Pinus taiwanensis by using the multi model optimization method,it is found that the optimal nonlinear binary volume model of Pinus taiwanensis in Henan Province is the variable parameter Yamamoto volume model:V=0.00013283D^1.72818744H^0.88183454,and the applicability test indexes of the model are RS=0.003,within±3%;REAA=0.0837<10%;F=1.069<F_{0.05（2,298）}=3.026.2.According to the modeling samples,six commonly used nonlinear univariate volume models of Pinus taiwanensis are fitted and optimized.It is determined that the optimal nonlinear univariate volume model of Pinus taiwanensis is Desekku Meyer formula,and the model formula is V=-0.00303885D+0.00071464D².According to the test samples,the applicability of the Diseakoku Meyer formula of the nonlinear model of Pinus taiwanensis was tested.It was found that the total relative error RS was-0.0004,located at ±5%,REAA=0.0972<10%,F=0.917<F_{0.05（2,298）}=3.026.3.Seven widely used ground diameter volume models are adopted.Based on the modeling samples,it is determined that the traditional optimal model of ground diameter nonlinear standing of Pinus taiwanensis is V=0.00001852D₀^3.05834511e^{-0.01785397D0}.Calculating the model applicability test indicators based on the test samples,in which the total relative error S=-0.0021 is within 士 5%,the average value of absolute relative error REAA=0.0998（<10%）,F=0.376<F_{0.05（2,298）}=3.026.4.The binary multilayer perceptron neural network volume model of Pinus taiwanensis is constructed based on the modeling samples.It is found that the optimal structure of Henan Pinus taiwanensis binary multilayer perceptron neural network model is 2:2:1.The volume is predicted according to the test samples and passes the applicability test,in which the total relative error RS=0.00013 is within ±3%;mean absolute value of relative error REAA=0.05336（<10%）;F=0.34<F_{0.05（2,298）}=3.026.5.The volume model of Pinus taiwanensis univariate multilayer perceptron neural network is constructed based on the modeling samples.It is found that the fitting effect of Pinus taiwanensis univariate multilayer perceptron neural network volume model with network structure of 1:2:1 is the best.The calculation of model applicability test index based on the test samples shows that the total relative error is RS=-0.0013,within ± 5%,the average of absolute value of relative error REAA=0.07921<10%,F=0.83<F_{0.05（2,298）}=3.026.6.When the neural network is built with 1:2:1 neural network,the best fitting effect is determined based on the tree volume of 1:2:1 neural network.The applicability of the neural network structure is tested,in which the total relative error RS=-0.00004 is within ± 5%;mean absolute value of relative error REAA=0.98889（<10%）;F=1.01<F_{0.05（2,298）}=3.026.7.The results show that the root mean square error RMSE and relative root mean square error RMSE%of the binary tree volume multilayer perceptron neural network model of Huangshan pine are not better than the nonlinear model,and the determination coefficient R²,the absolute error of MAE model is slightly better than that of traditional model;the neural network model is slightly better than the neural network model in both single-layer and multi-layer.8.When using the Standing Volume Model of Pinus taiwanensis in Henan Province,the DBH range of the binary standing volume model is 5cm～40.9cm,and the tree height range is 3.6m～19.6m;the DBH range of one-dimensional standing volume model is 5cm～40.9cm;the ground diameter range of the ground diameter standing volume model is 6cm～50cmTherefore,,the binary,unitary and ground diameter Standing Volume Tables of Henan Pinus taiwanensis can be obtained by expanding the model respectively within the range above.