Study On The Diameter Distribution Of Natural Birch In Inner Mongolia Region

Natural birch forest is one of the most important timber speciesin the north forestry area. In the recent years, this natural resource is suddenly decreased due to its unreasonable logging and management. So, to strengthen the existing management of natural birch forest is of great significance. Based on the200pieces of natural birch forestsample data which is collected in Daxing’an Mountains, of Inner Mongolia, this paper studied on the law of diameter structure of natural birch forest, and tried to build diameter distribution function and parameters of natural birch forest which can be accurately described. The majority of diameter distributions of this study area are shown in the unimodal mountain shaped distribution curve, a few in the reverse-J type curve. Draw the diameter class distribution histogram of tree numbers on the basis of the known information of stand diameter distribution. As the model foundation, Weibull distribution function is chosen to study the law of diameter distribution. This study used four different methods for parameter estimation-maximum likehood estimate (MLE), percentile-based estimation (PPM), method of moments incorporating skewnedd (MOM1) and traditional method of moments (MOM2), and compared the merits to the four methods. The results showed that the parameter and fitting effect of MOM1are more superior to other methods, followed by fitting effect MOM1> MLE> MOM2> PPM. Relational model between the parameters and stand characteristics factorwas built based on the requested parameters. Stepwise regression technique was used to establish the equation of linear regression: β=-6.1648+1.4765Dg-0.1733Hi;γ=0.4115+0.1654Dg, which is consisted of scale parameter β, shape parameter γ and stand factorsobtained by MOM1. R2in the scale parameter β was0.89, and in the shape parameter γ was0.32. Obviously, the scale parameter forecast accuracy was more higher, but both of these two parameters were passed t Test and F Test. And the multicollinearity was not existed between variables. The standard error of prediction model was very small, the fitting effect was good and the applicability was better.