Study On Whole-Stand Growth Modeling With Artificial Neural Network For Chinese Fir

During the course of forest management,it is important to grasp the status in quo in time and predict the development trend of forest resources for maco-management decision. The paper regarded Chinese Fir plantation as the research object, according to the practical condition in Shunchang forest farm,under MATLAB environment,the paper,using back propagation neural network modeling technology,carried on systematic research to one kind of growth model of the Chinese fir plantation.The model can apply theoretical basis for mastering the growth dynamics of Chinese fir plantation and scientific management,It's of great significance to the research and application in the whole-stand growth model.The full text introduced the development of the whole stand model and the general neural network learning and training strategy,analysed the strengths and weaknesses of BP neural network,made an comprehensive and intensive study on theory and modeling of BP neural network,and explained the perfomance analysis on BP neural network the paper used at length.Basing on these,the paper proposed an new application method in setting up whole-stand growth model by BP neural network.Then by the age,the site index and the stand denisty as input matrix and by the stand average breast-height diameter,the stand average height and the stock per hectare as output matrix,the paper constructed a whole-stand growth neural network model with multi-input and multi-output ,and selected suitable model structure. This is a breakthrough comparing with the the previous single-output whole-stand growth model. In order to improve indentification accuracy and generalization ability,it's necessary to simplify the original imput data by normalizing method before modeling,and anti-normalize the output data when the mode simulation has been completed.The results show that the network works more stable and get small errors with fast convergence afte normalizing.After a great deal of trainings and tests with 260 temporary plots data as training samples,determine the ideal BP neural model structure of 3:13:3, in this case, the mean square error mse is 0.0096, the overall fitting accuracy is over 95 percent, three outputs of the fitting accuracy are more than 93 percent.In order to test the model's predictive power,the paper used 92 temporary plots data as inspecting samples,and verified the proposed model .The overall forecast accuracy is 94 pecent, the forecast accuracy of three outputs are more than 93 pecent, these data showed that the model can work well on predicting the entire Chinese fir plantation forest trends. Visible qualitative analysis with the simulation curved surface of three-dimensional and quantitative analysis by calculation of precision and regress showed that the model was not only fit the growth law of forest but also has high precision and strong generalization ability,so the model can constitude the basis of prediction in whole-stand growth of Chinese fir plantation. It also shows that BP neural network model has superiority of predicting the whole-stand model, and illustrate a good prospect of application and extension.