Research On Growth And Profit Model Of Taxus Chinensis Var. Mairei Plantation

Key words:Taxus chinensis var. mairei, Plantation, Site Index, Density Index, Growth ModelThe research object of this paper is Taxus chinensis var. maire. More than 2,500 single tree plants in 370 sampling plots and of 9 counties and cities were researched, in Fujian, Zhejiang and Jiangsu province. 50 years growing variables including site index, business density index and single tree positioning of Taxus chinensis var. maire were investigated and collected. Based on BP modeling neural network programming combined with stand growth theory, changes of different growth model under different site index and density index were'studied systematically and comprehensively.First, there are detailed research on origin name, differences between ancient and modern, traditional medical function of Taxus, which made up the blank area of history, culture and functional applications of Taxus in our country. Second, research Materials, software and the key technology and methods were introduced. In the main part, according to the rank of site index and stand density index, site index model and site index tables, single tree growth model, stand model, all stand growth model and dynamic income forecasting model were discussed and constructed in sub-chapters. After organized the survey data, they were separated into training samples and test samples, and 3/5 are training samples and 2/5 are test samples. According to the principle of BP neural network model, a simple and effective hidden layer determination method was found. At the same time, the fitted models from artificial neural networks (MATLAB software) were selected combining with tree measurement science and tree growth theory, and then got accurate modeling analysis results and effective security.Tree age is the input variable and stand average superiority height is purpose output variable and then got 1:4:1 stand average superiority height BP neural network model. And then, drew up site index table of Taxus chinensis var. maire according to standard deviation adjustment method. Based on full application on site index table, effects of average high stand, stand average DBH, stand volume growth were researched comprehensively and systematically, and income of Taxus chinensis var. maire were predicted and evaluated using these models, combining site index and stand density. And, the overall training accuracy of single tree growth model of Taxus chinensis var. maire was up to 93%, which was higher than general modeling methods. The test accuracy of stand model cumulative frequency of tree height of Taxus chinensis var. maire was 97% on average, frequency test accuracy was 86% on average, and diameter cumulative frequency test accuracy o was 87.34%. The stand height DBH is a approximate two-dimensional distribution, and the number of Middle height and Middle diameter is Maximum, and tree height and diameter were less in the two poles. Two-dimensional distribution model of tree height diameter was 2:S:1, which was a one input and two outputs neural network model. It had high fitting accuracy and testing accuracy, and can up to relative satisfactory results.Innovation points of this paper:(1) Simplified modeling procedure of artificial neural networks plantation law, and built neural network models of famous species stands, individual tree growth, and wrote related program. The usage of model structure and the matrix determinant could reflect the relationship between input and output of neural network and input and output of neurons intuitively. The written programs could be used and modeled in MATLAB software, c language, c++ language and simulation models directly, and relations and the structure were simplified.(2) Selected fitting models of artificial neural networks (MATLAB software), combining several relevant theoretical knowledge, such as tree measurement science and tree growth theory. This qualitative and quantitative analysis combined method can not only determine numbers of hidden layer neurons simplify, but could avoid major defects which could deduct the model generalization ability, like prone single fitting data in training when only using MATLAB software. So, this is an effective protection for model and analysis accuracy.(3) Calculated theoretical value of variation coefficient of tree height in all age groups, according to working requirements of site index table, combining stand advantages height growth oriented curves. Built 2: s:1 neural network model made by site index table and built site index table model whose output variable was the site index, using age and average stand dominant height as variables. at the meantime, as a respect of individual tree growth, stand structure and stand growth of taxus chinensis var. maire, there was new progress, compared with previous studies. As a respect of data collection, model construction and model performance analysis, there was a special insight.Stand' structure, stand growth and individual tree growth system all have nonlinearity, diversity, complexity and other characteristics. Model construction technology which is closer to the natural growth pattern is needed to simulate forest dynamic growth model. But, artificial neural network has a significant advantage in this respect, which is a new technology of stand growth model. According to this model construction thought, a successful systematic study on BP neural network construction modeling technology was made to apply in stand structure and all stand growth model, combining theoretical knowledge of stand growth, putting using Taxus chinensis var. maire as the research object. This method can be extended to evaluation and analysis in the macro level and other research of Taxus chinensis var. maire, and even to growth amount assessment and growth model of other famous tree.