Objective A regional-scale stand biomass growth model was established to provide methodological support for predicting the biomass and carbon storage of natural broadleaved forests in Guangdong Province of southern China in the future.
Method Based on the five forest inventory data of Guangdong Province from 1997 to 2017, 203 natural forest sample plots with six broadleaved tree species such as Quercus spp., Schima superba and other soft broadleaved species as dominant tree species were selected. The site quality difference was reflected by parameter classification, the density effect was expressed by competition index, and the modeling method was distinguished by step-by-step modeling (univariate nonlinear regression method) and joint modeling (nonlinear simultaneous equations method). The DBH growth model, constructed by the theoretical growth equation, was used to estimate the stand age, and then various stand biomass growth models were constructed. The goodness of fit of the model was evaluated by four indexes such as determination coefficient and average prediction error. For the model with high goodness of fit, 183 sample plots by continuously inventory in four periods from 2002 to 2017 were taken as test samples, and the total relative error was used to verify its application effect.
Result To compare the fitting effect and the estimation accuracy at regional scale and sample plot level for exploring the influence of four factors including stand density, different parameter classification, classification method and modeling method on the biomass growth model, it was found that nonlinear simultaneous equation was better than step-by-step modeling; the classification of model parameter b related to growth rate was better than that of model parameter a related to growth potential; considering the stand density and adding competition index to the hierarchical equation had little effect on optimizing model performance. Based on the classification of parameter b, the joint model without competition index in independent variable and the hierarchical equation was the optimal model, i.e. Model 10. The determination coefficient of the biomass growth model was 0.970 1. When Model 10 was used to predict the biomass of four periods, the prediction effect was good. But the estimation error in the later stage was significantly lower than that in the earlier stage. For example, when Model 10 was used to estimate the biomass of Quercus spp. at regional scale from 2002 to 2017, the estimation errors of four periods were 6.22%, 15.27%, 4.80% and −1.84%, respectively.
Conclusion It is a feasible method to establish stand biomass growth model based on the Richards growth equation to estimate regional-scale biomass, which not only provides a basis for evaluating the carbon sink capacity of forest ecosystem at regional scale in a certain period in the future, but also provides a reference for the construction of stand biomass growth model in other regions.