| During the course of forest sustainable management of independentwork,especially in the management of natural forests,it’s significant for forest managementto grasp the status of the development and predict the trend of forest in time.The paperregarded the oak forest of Boshan Forest Farm in Zhumadian as the research object,thenbuilt its models of site index,density index and whole-stand growth.In graphic interfaces ofMatlab,the establishment of the oak forest dynamic growth models can reflect the changeprocesses and the trend of each factors in stand growth procedure, which can also monitorthe changes of forest resources.Artificial Neural Network(ANN) was one of the modem intelligent algorithms,which hasmany characteristics of self-learning,self-organizing,self-adaptive and fault-tolerantcapabilities.Within the process of building models,ANN needn’t suppose premiseconditions,consider the interior structure of mathematical model，and confirm factors powerartificially.It can also build models to predict the non-linear or complicated problems.ANNis utilized in this paper as methodology.Based on BP neural network,the correlative modelsof stand growth were built and some systemic researches on the aspects of site index,densityindex and stand growth have been done.(1)By stand age as input variables,average height of dominant trees as outputvariables,the paper constructed the height of dominant trees network model.As thestandard age of ark trees is30years,so the site index models was built by the definition ofsite index.After repeated training of the model,a site index model whose network structure is1:3:1was constructed,and its fitting precision reached98.36%which can betterly predict thesite index simulation of Boshan Forest Farm.(2)By average stand diameter as input variable,stand number density per hectare asoutput variables,number density model was constructed.The standard diameter of ark treeswas set to10cm,and the number density model was built by the definition of densityindex.With the analysis and comparison of models,when the network structure is1:1:1,themodel is the best with its fitting precision reaching95.68%.(3)By stand age,site index and single plant area as input variables,average standdiameter,average height and accumulation of per hectare as output variables,the standgrowth model of oak natural forests was constructed.Logsig was set as the transfer functionof hidden layer, the transfer function of output layer was purelin while its network structureis3:3:3.With the analysis of network performance,we found that the overall fitting precisionof the model was97.26%,the fitting precision of average stand diameter was97.43%,thefitting precision of average height is96.5%,the fitting precision of stand accumulation of perhectare was93.32%.The overall prediction test precision of the model is97.12%. (4)By using the human-computer interaction tools and methods in Matlab,some controlswere added in the interface of building natural forest growth dynamic model system,then thecontrols were set on their attribute and code,finally,combined the established models of siteindex,density index and stand growth.We ran this system,then input the impact factors ofstand growth,the growth variable was gained.We can view the influence processes andvarious trends of various factors on the stand growth in the graphic windows of the system.According to the study results,the dynamic model system of oak natural forests growthcan betterly simulate the growth of oak natural forests of Boshan Forest Farm in Zhumadianwhich basically achieved the goal of this research.The interface of the established systemwas convenient operation and simple. Comparing with the traditional professionalprogramming languages like VB,VC or C++,simple and easy-used Matlab languages whichwere suitable for the non-computer-major foresters were employed to construct this system. |
PDF Full Text Request