Species-specific and generalized allometric biomass models for eight Fagaceae species in the understory of evergreen broadleaved forests in subtropical China

Quantifying the biomass of saplings in the regeneration component is critical for understanding biogeochemical processes of forest ecosystems.However,accurate allometric equations have yet to be developed in sufficient detail.To develop species-specific and generalized allometric equations,154 saplings of eight Fagaceae tree species in subtropical China’s evergreen broadleaved forests were collected.Three dendrometric variables,root collar diameter(d),height(h),and crown area(ca)were applied in the model by the weighted nonlinear seemingly unrelated regression method.Using only d as an input variable,the species-specific and generalized allometric equations estimated the aboveground biomass reasonably,with R _(adj)^(2) values generally>0.85.Adding h and/or ca improved the fitting of some biomass components to a certain extent.Generalized equations showed a relatively large coefficient of variation but comparable bias to species-specific equations.Only in the absence of species-specific equations at a given location are generalized equations for mixed species recommended.The developed regression equations can be used to accurately calculate the aboveground biomass of understory Fagaceae regeneration trees in China’s subtropical evergreen broadleaved forests.

Uncertainties in above ground tree biomass estimation

Models of above-ground tree biomass have been widely used to estimate forest biomass using national forest inventory data.However,many sources of uncertainty affect above-ground biomass estimation and are challenging to assess.In this study,the uncertainties associated with the measurement error in independent variables(diameter at breast height,tree height),residual variability,variances of the parameter estimates,and the sampling variability of national inventory data are estimated for five above-ground biomass models.The results show sampling variability is the most significant source of uncertainty.The measurement error and residual variability have negligible effects on forests above-ground biomass estimations.Thus,a reduction in the uncertainty of the sampling variability has the greatest potential to decrease the overall uncertainty.The power model containing only the diameter at breast height has the smallest uncertainty.The findings of this study provide suggestions to achieve a trade-off between accuracy and cost for above-ground biomass estimation using field work.

Long-term response of living forest biomass to extensive logging in subtropical China

Forest disturbance and recovery are critical ecosystem processes,but the temporal patterns of disturbance have not been studied in subtropical China.Using a tree-ring analysis approach,we studied post-logging above-ground(ABG)biomass recovery dynamics over a 26-year period in four plots with different degrees of logging disturbance.Before logging,the ABG biomass ranged from 291 to 309 t ha-1.Soon after logging,the plots in primary forest,secondary forest,mixed forest and singlespecies forest had lost 33,91,90 and 100%of their initial ABG biomass,respectively.Twenty-six years after logging,the plots had regained 147,62,80 and 92%of their original ABG biomass,respectively.Over the 26 years following logging,the mean CAI(Current annual increment)were 10.1,5.5,6.4 and 10.8 t ha^-1 a^-1 and the average MAI(Mean annual increment)8.7,2.5,5.6 and 7.8 t ha^-1 a^-1 for the four forest types,respectively.The results indicate that subtropical forests subjected to moderate logging or disturbances do not require intensive management and single-species plantings can rapidly restore the above-ground biomass to levels prior to heavy logging.

DEPENDENCE ANALYSIS OF REGRESSION MODELS IN TIME SERIES

In this paper, the relative dependence of a linear regression model is studied. In particular, the dependence of autoregressive models in time series are investigated. It is shown that for the first-order non-stationary autoregressive model and the random walk with trend and drift model, the dependence between two states decreases with lag. Some numerical examples are presented as well.