Study On Growth And Yield Model Of Secondary Forest Of Quercus In Hunan

Stand growth and yield models are widely used in estimating stand growth dynamics,formulating forest management measures and evaluating forest management effects.The oak secondary forest of the national continuous forest inventory in Hunan Province was taken as the research object,and the retest data of 176 fixed sample plots(6 periods from 1989 to 2014)with complete data,oak as dominant tree species and natural secondary forest types were selected as the research data.By putting forward the method of estimating age of uneven-aged forest based on multi-stage tree diameter,developing a breast diameter status index table,constructing compatible stand growth and yield prediction model,and then adding the mixed effect of sample layers constructing compatible stand growth and yield prediction model based on the mixed effect model is established.It provides a theoretical basis for accurately predict and simulate the future oak forest growth dynamics and establishment of a harvest table that obeys the law of oak growing.(1)This study is based on a multi-period diameter measurement data of a fixed sample plot.By setting different initial ages and selected the Richard equation to simulate multiple diameter-age growth curves.Then the growth curve was accurately located according to the age when the tree height grew to 1.3 m and estimate the age of individual tree,diameter scale and stand.The method was tested by parse wood data.The results show that the average relative error of the method is less than 10%when estimating the average age and the diameter of the forest.The method is effective and can be applied in scientific research and practice.(2)A diameter at breast height(DBH)site index table of natural secondary oak forest in Hunan Province was compiled.First the logistic equation was selected as the optimal guiding curve equation.The datum age was calculated according to the age at which the variation coefficient of the DBH tended to vary little with age.The exponential distance was determined according to the ratio of the absolute variable range of the DBH at the datum age to the exponential number.The datum age was 40a and the exponential distance was 3 cm.Then the DBH site index tables for natural secondary oak forests in Hunan Province were compiled using the standard deviation adjustment method,the coefficient of variation adjustment method,and the relative DBH method.The accuracy of the DBH site index table was determined from the results of a chi-square test of the theoretical values and the actual values for the DBH sample and whether there was an advanced placement of the status index at the same sample site.The test results showed that the site index table compiled by the relative DBH method was the most accurate.(3)Two-step least square method and three-step least square method were used to fit the simultaneous equations of the compatible stand growth and yield model.The evaluation indexes such as the mean error,mean absolute relative error,root mean square error(RMSE),prediction accuracy and decision coefficient(R2)are used to evaluate the accuracy of the fitting results.Finally,the two-step least squares method with the smallest error and the highest precision is used to fit the simultaneous equations.The problem of heteroscedasticity of the model uses the reciprocal of the original function as the weight function,and the simultaneous equations model is modified.The residual distribution map after the model is uniformly distributed,which can better explain the model heteroscedasticity problem.(4)The mixed effect model method was used to simulate the compatible forest growth and yield model.The mixed effect parameter of b3 was selected as the initial stand accumulation model and a2 as the mixed effect parameter of the ending stand area model had the best fitting effect.When considering the model heteroscedasticity problem,the power function is the best choice to simulate the heteroscedasticity structure;When considering the time series auto-correlation,the AR(1)structural matrix model has the best simulation effect as the auto-correlation structure.The average error,mean absolute relative error,root mean square error(RMSE)and prediction accuracy of the mixed effect model were calculated by using the verification data.It was found that the four evaluation indexes of the mixed effect model method were better than the weighted two-step least square method.The residual distribution diagram of the mixed effect model method is uniformly distributed,and the distribution range is greatly reduced.It shows that the fitting effect of the mixed effect model method with heteroscedastic structure and autocorrelation structure is better than that of the weighted two-step least squares method.The mixed effect model can not only describe the change of sample plots by modifying random effect parameters with different difference covariance matrix structures,but also explain the problem of model heteroscedasticity and time series correlation of continuous measurement data through heteroscedasticity structure and autocorrelation structure.Therefore,mixing effect can effectively improve the applicability of the model and the accuracy of prediction.