广东西樵山国家森林公园森林碳储量空间分布研究

Study on the Spatial Distribution of Forest Carbon Storage in Xiqiao Mountain National Forest Park,Guangdong Province

为阐明不同模型对森林碳储量空间分布估测精度的影响,依据2020年广东西樵山国家森林公园森林资源管理“一张图”数据,采用地理加权回归模型(Geographic weighted regression model,GWR)、空间误差模型(Spatial error model,SEM)和最小二乘模型(Ordinary least squares,OLS)对研究区森林碳储量的空间分布进行研究,并对3种模型的拟合效果和影响森林碳储量的因子进行分析,利用全局和局域Moran's I分别对模型残差的全局空间自相关性和空间分布状况进行描述,对空间异质性作用下各模型之间的差异进行说明,采用GWR模型绘制研究区森林碳储量的空间分布。结果表明,处于不同位置下的GWR模型,各参数估计值对森林碳储量的影响大小不断变化;GWR模型在数据拟合方面要明显优于SEM模型和OLS模型,OLS模型的拟合效果最差;GWR模型产生了较大范围的参数估计值,且GWR模型各参数估计值的范围均包括了OLS模型和SEM模型的参数估计值,得到了较好的模型残差局域化空间分布效果,模型稳定性较高;研究区中部森林碳储量较多,GWR模型的拟合偏差为1.36 t/hm 2,在所有模型中最小。

To clarify the influence of different models on the accuracy of spatial distribution of forest carbon storage estimation,this study used geographically weighted regression model(GWR),spatial error model(SEM)and ordinary least squares model(OLS)to analyze the fitting effect of the three models and the factors affecting the forest carbon storage in the study area based on the forest resource management"one map"in Guangdong Xiqiao Mountain National Forest Park in 2020.Global Moran's I and local Moran's I were used to respectively describe the global spatial auto-correlation and the spatial distribution of the model residuals,to illustrate the differences between the models in the action of spatial heterogeneity.GWR was used to map the spatial distribution of forest carbon storage in the study area.The results showed that the influence size of various parameter estimates on forest carbon storage was constantly changing for GWR at different positions.GWR significantly outperformed SEM and OLS in terms of the data fitting,while OLS had the worst fit.GWR yielded a wide range of parameter estimates.Moreover,the range of all parameter estimates of GWR included the parameter estimates of OLS and SEM,which obtained a good localized spatial distribution effect of model residues and high model stability.The central forest of the study area had more forest carbon storage,and the fitting deviation of GWR was 1.36 t/hm 2,which was minimal in all models.