Researches On The Estimation Method Of Aboveground Carbon Storage Of The Sinocalamus Affinis McClure Forest Based On BP Neural Networks

With the convening of the Copenhagen conference,estimation of carbon stock has become an important concern by home and abroad. And forest as the mainly of absorbing CO2, whose carbon sequestration function has been more and more attention by the people. In recent years, Forest carbon storage to the quantitative studyed, Along with global suppress greenhouse gas emissions and presents one of the hot spots. According to the detailed analysis of current forest carbon measurement methods, it can be found that different estimation methods get different forest carbon results, sometimes distinguish remarkably. Currently, the traditional estimation method in practical work of the carbon storage measure has some limitations. The neural network has the feature of distributed parallel processing, nonlinear mapping, adaptive learning and fault tolerance, due to it has unique information processing and computer power, it shows strong advantage on the unclear mechanism of high dimensional nonlinear system. This paper attempts to use the essence part of the artificial neural network-BP algorithm, and to establish Aboveground Carbon storage of the Sinocalamus affnis McClure forest model, so as to provide reference to accurate estimate the forest carbon storage.This paper based on much exploration and studying about carbon storage measurement method and theory, it was to deeply study the BP neural network theory and method, and it was quoted to evaluate aboveground Carbon storage of the Sinocalamus affinis McClure forest, it established BP neural network model and finally determined model structure as7-5-1.The number7means that use seven factors, The number5means that neural network includes five hidden layers and the number1means that neural network has one output layer which is Aboveground Carbon storage of the Sinocalamus affinis McClure, training the mean square error is0.002083. Applying this model to estimate Aboveground Carbon storage of the Sinocalamus affinis McClure forest in muchuan, sichuan province.And compared with the traditional regression model of forecasting method results.The results shows:the BP neural network model of training mean absolute error and the average relative error was0.0478and5.38%respectively, to new sample simulation of mean absolute error and the average relative error were0.8897and9.51%respectively. regression model mean absolute error and the average relative error were0.7351and11.16%respectively, the average forecast for new sample absolute error and the average relative error were1.5812and17.08%respectively.The result shown that the estimation results regression model and the BP neural network have significant distinguish; The BP neural network model error and estimation error is less than regression model method, namely carbon storage estimation method based on the BP neural network model has higher precision and smaller error.This paper used BP neural network model, which was established evaluation model about typical vegetation aboveground Carbon storage of the Sinocalamus affinis McClure forest in Muchuan Sichuan province. It has some theoretical meanings to explore and improve carbon storage measurement method. The model itself prediction results were well, which had the significance of increasing precision to evaluate aboveground carbon storage of the Sinocalamus affinis McClure forest, meanwhile. It can be also used as a new method for estimation the forest carbon storage.