Abstract:

Objective: Based on the accurate measuring of the stem volume of moso bamboo(Phyllostachys edulis), the research on bamboo stem volume models was conducted to accurately estimate the stem volume and to provide a theoretical basis for the management of bamboo forest. Method: Taking moso bamboo as the research object, based on the measured data of 216 samples of bamboo from Lin'an, Qingyuan, Wuyi, Changshan, Ninghai, Anji, Zhuji, Yuyao, Huangyan and Taishun county of Zhejiang Province, drainage method that could measure the volume of irregular shaped object was used to measure the bamboo stem volume. Firstly, based on the allometric equation and the equation proposed by Terasaki Watari, taking the diameter at breast height(D)and the internode length of bamboo at breast height(L)as the independent variables, one unary model(M₁)and four binary models(M₂-M₅)were selected to establish the three bamboo stem volume models with all the samples data. Secondly, model error structures were analyzed with the likelihood analysis with which whether the log-transformed linear models or nonlinear weighted models were chosen for model fitting. The parameter estimations were obtained by fitting regression of the models using ForStat2.2 software. Finally, the bamboo stem volume model that fit for Zhejiang Province well and had stable parameters and higher prediction accuracy was selected according to analysis towards the results of t test for the parameter estimates and 4 validation statistics, namely the adjusted coefficient of determination(R_a²), the standard error of the estimate(SEE), the mean errors(ME)and the mean systematic errors(MSE). Result: The Akaike's information criterion(AICc)of the 5 basic models were calculated with the likelihood analysis of model error structure. The results showed that the value of ΔAICc was larger than 2, which indicated the multiplicative error and the use of log-transformed model for model fitting. The parameters of the 5 bamboo stem volume log-models established were stable for the absolute t-value was larger than 2. All of the R_a² were above 0.95, and the SEE and MSE were around 0, respectively, which reflected the well-fitting, and the model M₂ fit the best. The statistics of model fitting and prediction across diameter classes were investigated as well. The results illustrated that the fitting and prediction performance across diameter classes of the 5 models was quite preferable, yet better when assuming the medium diameter classes and a bit worse as assuming the class of 4.0-5.9 cm and that of 14.0-15.9 cm. Finally, the prediction log-models with correction did not perform a better fitting effects and prediction precision. Conclusion: Drainage method is a better method for accurate measurement of bamboo stem volume. Likelihood analysis is suitable to analyze the error structure and provide with a reliable justification of choosing the fitting models. The evaluation accuracy of binary models is higher than that of unary model as the variable H and L are introduced. Considering the easier and more accurate measurement of internodes length of bamboo at breast height(L)in practice, and the stem volume model M₅ performs better than model M₂, the suggested suitable model for predicting the stem volume is M₅ that is based on variable D-L, i.e.V=0.191 2D^{2.114 9}e^{-6.841 1/L}.

Key words: Phyllostachys edulis, bamboo stem volume, internode length of bamboo at breast height, volume model, drainage method