Study On Growth Models With Artificial Neural Network For Betula Platyphlla Plantation

The paper reagarded Betula platyphlla plantation as the reasearch object, and reagarded the Betula platyphlla plantation survey data standards of Northeast Forestry University in Maoershan experimental forest as the training samples and testing samples. Under MATLAB environment, the paper, using BP neural network modeling technology, carried on systematic research to the whole-stand growth model and the individual tree growth model of Betula platyphlla plantation.On the whole-stand growth model study, by the age, the site index and the stand density index as input matrix and by the stand volume per hectare as output matrix, the paper constructed a structure of 3: S: 1 BP neural network model with multi-input and single-output. The model was trained by 300 Betula plantation in standards so that a suitable network structure was 3:3:1.Model training results show that the mean square error function mse = 0.0011203. The accuracy of the results of the analysis model were that the whole-stand growth model fits the overall accuracy of 94.17% and the model that the theoretical value of the fitting are in good agreement with the actual value. Using 196 samples tested of the model to the test, the overall accuracy of 93.97 %, and rendering of the actual value of stand volume and theoretical value of the control chart in three-dimensional,we could see that the whole-stand growth model could predict the theoretical value which was close to the measured values.The predictive power of model was strong.The individual tree model of BP ANN was created,in which the model structure was 4:S:1, by diameter of individual tree, stand density index,site index and age of stand as input variables, and by diameter increase of individual tree as output variable,using log-sigmoid function (logsig) and linearity function(purelin)of MATLAB as the neural functions. The neural network model was trained by 200 growth data of individual trees in thinning stands so that the best network structure is4 : 3 : 1, the MSE is 0.00160179, the matching precision is 96.86%.The whole-stand growth model and the individual tree growth model of Betula platyphlla plantation fully fitted the sample of data while maintained the regularity of tree growth equation .In the stand management, the model can be used for analysis, computing,simulation, prediction and so on.