Full-Text Search:
Home|About CNKI|User Service|中文
Add to Favorite Get Latest Update

Establishment of predicting models for Pinus tabulaeformis stands volume based on mixed models

WANG Shaojie;DENG Huafeng;XIANG Wei;HUANG Guosheng;WANG Xuejun;College of Forestry,Beijing Forestry University;Academy of Forest Inventory and Planning,State Forestry Administration;  
【Objective】This study established high precision stand volume models to provide personalized model equations for different density stands in Beijing,which would provide theoretical basis for forest management and harvesting.【Method】Using the periodically inventory data of 76 Pinus tabulaeformis plots in Beijing,P.tabulaeformis stands were divided into five levels by different initial densities(ID)including Ⅰ(ID400 strain/hm~2),Ⅱ(400≤ID800 strain/hm~2),Ⅲ(800≤ID1 200 strain/hm~2,Ⅳ(1 200≤ID100 strain/hm~2),and Ⅴ(ID≥1 600 strain/hm~2).Choosing basal area and average height of dominant trees as independent variables,the basic linear model of P.tabulaeformis volume was constructed.Based on the basic model,P.tabulaeformis mixed effects models were constructed considering densitylevel effect and plot effect and using R language NLME.Then absolute mean error(||),root mean square errors(RMSE),and coefficient of determination(R~2)were used to evaluate the models.【Result】The R~2 of two-level mixed model was 0.998 2,higher than that of single-level mixed models of density level effect and plot effect,while both||and RMSEof the two-level mixed model were smaller than the two singlelevel mixed models.The||and RMSEof the two-level mixed model were 0.069 8 and 0.100 6,which were 78.86% and 82.39% lower than the basic model,respectively.When heteroscedasticity of exponential function and[ARMA(1,1)]time series correlation structure were added to the mixed model,the fitting precision was improved.【Conclusion】The precision of both single level mixed models and two-level mixed model were better than that of basic model.The two-level mixed model was better than basic model and single level mixed models.Exponential function variance structures could effectively remove the heteroscedasticity in the data,and[ARMA(1,1)]structure can express the error correlation between plots.
Download(CAJ format) Download(PDF format)
CAJViewer7.0 supports all the CNKI file formats; AdobeReader only supports the PDF format.
©CNKI All Rights Reserved