基于不同空间模型的马尾松林生态系统碳密度研究

以江西省马尾松林生态系统为研究对象,基于样地调查及样品碳含量测定结果计算其碳密度,并选取立地、植被及气象等方面的15个因子,采用多元线性逐步回归方法筛选出对生态系统碳密度影响显著的因子,然后分别利用最小二乘模型(OLS)、空间误差模型(SEM)、空间滞后模型(SLM)和地理加权回归模型(GWR)构建生态系统碳密度与其影响因子之间的关系模型,筛选出最优的拟合模型。结果表明:对马尾松林生态系统碳密度影响显著的因子分别为海拔、坡度、土层厚度、胸径、年均温度和年均降水量。4种模型拟合结果均显示碳密度与坡度呈负相关,与海拔、土层厚度、胸径呈正相关。模型的决定系数(R~2)由大到小分别为GWR(0.8043)SEM(0.6371)SLM(0.6364)OLS(0.6321),模型均方误差(MSE)与赤池信息准则(AIC)最大的均为OLS模型,最小的均为GWR模型;残差检验表明GWR模型能有效降低模型残差的空间自相关性。综合分析得出GWR模型的拟合效果最优,更适用于江西省马尾松林生态系统碳密度的估测。

Study on carbon density in Pinus massoniana forest ecosystem based on different spatial models

This study investigated Pinus massoniana forest in Jiangxi Province. Based on plot investigation and measurement of sample carbon content, the carbon density of Pinus massoniana forest ecosystem was calculated. The influence factors on ecosystem carbon density were screened from fifteen factors, including site, vegetation and meteorological factors by using multiple linear stepwise regression method. The relationship models between ecosystem carbon density and influence factors were established by using ordinary least squares model (OLS), spatial error model (SEM), spatial lag model (SLM) and geographically weighted regression model (GWR), and the best fitting model was selected from them. The results showed that the influence factors of ecosystem carbon density were elevation, slope, soil depth, diameter at breast height (DBH), mean annual temperature and annual precipitation. The fitting results of four models showed that ecosystem carbon density was negatively correlated with slope and positively correlated with elevation, soil depth and DBH. The determination coefficient (R²) of four models were GWR (0.8043) > SEM (0.6371) > SLM (0.6364) > OLS (0.6321). The largest mean square error (MSE) and Akaike information criterion (AIC) of the model was OLS, and the smallest was GWR. The residual tests showed that GWR could effectively reduce the spatial autocorrelation of model residuals. In conclusion, the fitting effect of GWR was the best, which was more suitable for estimating carbon density of Pinus massoniana forest ecosystem in Jiangxi Province.