摘要
Segmented taper equation was selected to model stem profile of Dahurian larch （Larix gmelinii Rupr.）. The data were based on stem analysis of 74 trees from Dailing Forest Bureau in Heilongjiang Province, Northeastern China. Two taper equations with crown ratio and stand basal area were derived from the Max and Burkhart’s （1976） taper equation. Three taper equations were evaluated： （1） the original equation, （2） the original equation with crown ratio, and （3） the original equation with basal area. SAS NLIN and SYSNLIN procedures were used to fit taper equations. Fit statistics and cross-validation were used to evaluate the accuracy and precision of these models. Parameter estimates showed that the original equation with inclusion of crown ratio and basal area variables provided significantly different parameter estimates with lower standard errors. Overall fit statistics indicated that the root mean square error （RMSE） for diameter outside and inside bark decreased respectively by 10% and 7% in the original model with crown ratio and by 12% and 7.2% in the original model with basal area. Cross-validation further confirmed that the original equation with inclusion of crown ratio and basal area variables provided more accurate predictions at the lower section （relative heights, 10%） and upper section （relative heights, 50%） for both outside and inside bark diameters.
Segmented taper equation was selected to model stem profile of Dahurian larch （Larix gmelinii Rupr.）. The data were based on stem analysis of 74 trees from Dailing Forest Bureau in Heilongjiang Province, Northeastern China. Two taper equations with crown ratio and stand basal area were derived from the Max and Burkhart’s （1976） taper equation. Three taper equations were evaluated： （1） the original equation, （2） the original equation with crown ratio, and （3） the original equation with basal area. SAS NLIN and SYSNLIN procedures were used to fit taper equations. Fit statistics and cross-validation were used to evaluate the accuracy and precision of these models. Parameter estimates showed that the original equation with inclusion of crown ratio and basal area variables provided significantly different parameter estimates with lower standard errors. Overall fit statistics indicated that the root mean square error （RMSE） for diameter outside and inside bark decreased respectively by 10% and 7% in the original model with crown ratio and by 12% and 7.2% in the original model with basal area. Cross-validation further confirmed that the original equation with inclusion of crown ratio and basal area variables provided more accurate predictions at the lower section （relative heights, 10%） and upper section （relative heights, 50%） for both outside and inside bark diameters.
基金
This study was supported by the National Natural Science Foundation of China(30972363)
Special Fund for For-estry-Scientific Research in the Public Interest(201004026)
China Postdoctoral Science Foundation(200902362,20100471014)
the Fun-damental Research Funds for the Central Universities(DL10CA06)
SRF for ROCS,SEM.
