Abstract:

Objective: The aim of this study is to develop stand-level biomass and carbon stock models for major forest types in Guangdong Province, so as to facilitate the estimation of the biomass and carbon stock in the province, and realize the effective connection of the provincial with municipal- and county-level data. In addition, this study also analyzes the impact of tree species structure and climate on models and its impact mechanism, providing theoretical guidance for more detailed carbon sink monitoring and forest quality improvement. Method: Taking 12 major forest types in Guangdong Province as research object, and using the national forest inventory in 2007, 2012 and 2017 in Guangdong Province, the models of above-ground biomass, below-ground biomass, above-ground carbon stock and below-ground carbon stock compatible with the volume for each forest type were established by using the nonlinear error-in-variable simultaneous equation approach. The influence of tree species structure was introduced into each model in the form of dummy variables, and the climate-sensitive stand biomass and carbon stock models were established by the method of reparameterization. Finally, the model fitting results were evaluated, and the impact of climate variables on stand biomass and carbon stock was also analyzed. Result: The models of volume, above- and below-ground biomass, and the mean carbon content coefficient of above- and below-ground biomass were obtained. 1) The adjusted coefficients of determination ($ R_{\mathrm{a}}^2 $) of basic models based on basal area and mean height were between 0.947?0.997. The total relative errors (TRE) and the mean systematic errors (MSE) were within ±1.54% and ±2.48%, respectively, which were not more than ±3%. The mean prediction errors (MPE) were between 0.30%?3.61%, and only some of models of Quercus spp. forest and Acacia spp. forest were slightly more than 3%. The mean percent standard errors (MPSE) were between 3.30%?13.39%, which were not more than 15%. 2) The $ R_{\mathrm{a}}^2 $ of simplified models based on basal area were between 0.876?0.996. Except for the poor fitting effect of below-ground biomass model of Acacia spp. forest, the TRE and MSE of other models were within ±3.19% and ±2.74%, respectively, with MPE of 0.36%?4.70% and MPSE of 4.18%?15.61%. The $ R_{\mathrm{a}}^2 $ of supplementary models based on mean DBH and stand density were between 0.775?0.977, most of which were above 0.9. Except for the poor fitting effect of some models of Acacia spp. forest, the TRE and MSE of other models were within ±2.28% and ±1.83% respectively, with MPE of 1.12%?6.24% and MPSE of 5.91%?17.44%. 3) The $ R_{\mathrm{a}}^2 $ of stand-level models for distinguishing tree species structure were between 0.960?0.997. TRE and MSE were within ±1.61% and ±2.33% respectively, MPE were between 0.30%?3.41%, and MPSE were between 2.67%?12.92%. In addition, most models were significantly better than the basic models. 4) The climate-sensitive stand-level biomass and carbon stock models for 8 forest types were constructed, which $ R\mathrm{_a^2} $ were between 0.947?0.998. TRE and MSE were within ±1.86% and ±1.96% respectively, MPE were between 0.29%?2.65%, and MPSE were between 3.18%?13.29%. Most models were significantly improved compared with the basic models. In most cases, biomass is negatively correlated with temperature index, and negatively correlated with evapotranspiration or positively correlated with precipitation. Conclusion: The models constructed in this study have good fitting effect and high prediction accuracy. In practical application, appropriate models can be selected according to the availability of data and the estimation range. High temperature, excessive evapotranspiration, or insufficient precipitation are the main factors limiting the growth of forest biomass and carbon stock in Guangdong.

Key words: biomass, carbon stock, stand-level model, nonlinear error-in-variable simultaneous equation, climate, national forest inventory