多重密度布点对土壤有机质空间特性的解析

论文以张家港市耕层土壤有机质（Soil organic matter, SOM）为目标变量,通过网格布点建立11种样点密度梯度,运用经典统计学和地统计学方法研究其空间特性随布点密度变化的趋势规律。结果表明,各密度下SOM的变异系数稳定在21.70%~23.12%之间,表现出中等程度变异;随着布点密度降低,SOM的块基比和分维数分别由37.52%、1.838波动降至18.70%、1.714,Moran’s I则由0.485起伏升高至0.857,相应地其空间变异中结构变异组分增多、随机变异贡献减少,空间自相关性也逐渐增强;SOM呈现西南-东北递减分布,随着样点密度减小,其空间格局趋于简化。论文以布点密度为视窗,揭示了SOM的多重空间尺度特征,可为解析土壤属性的层级结构提供参考依据。

Spatial Characters of Soil Organic Matter withMulti-sampling Density

The spatial variation of soil properties is a function of spatial scales. The spatial structure of soil properties in special scale can be discovered with single sample density, however, the whole variation could be discovered with multiple sample densities. The study area locates in Zhangjiagang County, which is a part of Yangtze River delta. Soil organic matter (SOM) is one of the important issues in global change research, thus in this paper the top SOM content in fertilized area is studied. Four hundred and thirty one samples were collected as the fundamental data, 11 levels of density or scale were designed with grid centroid sampling method, then classic statistics, semi-variance function, Moran’s I, spatial fractal dimensions and Kriging interpolation were explored to investigate the variation of SOM in spatial structure, randomness and spatial patterns at different sampling densities. The results shows: There is a low influence of sampling density on SOM variation coefficient. The variation coefficients of SOM at 11 sampling densities are among 21.70%-23.12%, which belongs to the moderate variation. With the decrease of the sampling density or the increase of scale unit, the corresponding ratio of nugget/sill falls from 37.52% to 18.70%, fractal dimension falls from 1.838 to 1.714, and the Moran’s I index increases from 0.485 to 0.857, which indicates the components of spatial structure variation is increasing and spatial correlation is gradually enhanced. The content of SOM shows decreasing trend from southwest to northeast. The spatial pattern of SOM content becomes simple with the decrease of sampling density since more and more details in micro-areas are ignored. In general, the model inference accuracy of SOM is reliable and the spatial character is stable when the sampling density is lower than 0.179 sample per km2. In consideration of science and economy, it can be concluded that the density of 0.179 sample per km2 can reflect the spatial randomness and structure of SOM in the study area, and this density can be regarded as the best sampling density. The paper takes sampling densities as cases to reveal the multiple spatial scale characteristics of SOM, and provides a helpful reference for hierarchical analysis of soil properties.