中文摘要：

以临安市为例,利用2004年森林资源清查样地数据和同年度Landsat TM影像数据,采用一元二次非线性回归和序列高斯协同模拟方法分别模拟森林地上部分碳密度及其分布,并对模拟结果进行比较分析。结果表明：一元二次非线性回归估计得研究区森林碳储量为2365404.37t,碳密度平均值为9.0000t·hm-2,最大值为73.7144t·hm-2,最小值为0.7156t·hm-2;序列高斯协同模拟得研究区森林碳储量为3291659.83t,碳密度平均值为12.5233t·hm-2,最大值为78.9133t·hm-2,最小值为0.0833t·hm-2;根据2004年森林资源清查样地数据,按随机抽样方法估计研究区森林碳储量为2708897.90t,样地碳密度平均值为10.3065t·hm-2,其最大值为96.9625t·hm-2,最小值为0;序列高斯协同模拟结果更接近地面样地估计结果,而且碳密度分布范围更合理;一元二次非线性回归估计结果与地面样地估计结果之差的累积平方和为9857.4619,而序列高斯协同模拟结果与实测结果之差的累积平方和为8018.4625;序列高斯协同模拟较一元二次非线性回归在估计区域森林碳空间分布上有明显优势。

英文摘要：

Estimation of the forest carbon distribution is an important subject in study of forest carbon. Based on National Forest Inventory （NFI） data and the Landsat TM image data collected in Linan County, Zhejiang in 2004, this research applied two methods, namely unary quadratic nonlinear modeling and Sequential Gaussian co-Simulation to reproduce the distribution of above ground forest carbon, and compared and analyzed the estimation results of above ground forest carbon density. The estimation results with unary quadratic nonlinear regression estimation show that the sum of above ground carbon is 2 365 404.37 t, the mean carbon density is 9.000 0 t·hm-2, with the maximum carbon density of 73.714 4 t·hm-2, and the minimum carbon density of 0.715 6 t·hm-2. With the Sequential Gaussian co-Simulation, the sum of the carbon is 3 291 659.83 t, the mean carbon density is 12.523 3 t·hm-2, with the maximum carbon density of 78.913 3 t·hm-2, and the minimum carbon density of 0.083 3 t·hm-2. According to the NFI data in 2004 , the carbon storage for the study area is estimated with the random sampling method. The total carbon is 2 708 897.90 t, the mean carbon density is 10.306 5 t·hm-2, with the maximum carbon density of 96.962 5 t·hm-2, the minimum carbon density of 0.000 0 t·hm-2. The carbon density from the Sequential Gaussian co-Simulation are closer to that calculated from the NFI data, and the carbon density distribution is more reasonable. The sum of squares of differences between unary quadratic nonlinear regression result and the sample plot data is 9 857.461 9, while that between the results from the Sequential Gauss co-Simulation and the sample plot data is 8 018.462 5. The Sequential Gaussian co-Simulation is relatively better than unary quadratic nonlinear regression on regional forest carbon density estimation.