土壤有机质含量是衡量土壤肥力和评价耕地质量的重要指标。在县域尺度下,利用合理数量的样点反映其空间分异特征具有重要意义。论文选取位于江汉平原腹地的湖北省公安县为研究区,布设了4 045个样点并测定其有机质含量,采用半方差分析得到土壤有机质空间变异特征,基于Moran’s I揭示空间变异较大地区,并通过二分法的思想逼近最佳合理采样数值。结果表明：研究区土壤有机质分布呈中等空间相关性,有6个局部区域的空间变异明显高于邻近地区,采样点数量与插值精度呈正相关,合理采样数为598个。研究结果可以为江汉平原县域耕地质量监测和管控提供指导。

Soil organic matter (SOM) is a key index to estimate soil fertility and quality of cultivated land. Finding an appropriate number of samples on county scale is very important in saving costs while accurately expressing the spatial variability of SOM. In this research, Gong’an County, with an area of 2 258 km², locating in the hinterland of Jianghan Plain (111°25′-111°48′E, 29°37′-30°19′N), was chosen as a typical case. A total of 4 045 soil samples were collected for analysis of SOM content. After the rejection of outliers, 3 950 sample sites were retained for further analysis. First, semi-variance function was used to explore the whole spatial variability. Then, a gradient template was defined to calculate the gradient vector, and the area was devided into seven parts based on the density of the pixels which had high gradient values. Moran’s I was used to describe the spatial variability of each part. In order to determine the appropriate number of samples, 100 points were randomly sampled and independent verification was used to validate the precision of Kriging interpolation result under different number of sample sites, i.e., 3 850, 2 695, 1 886, 1 320, 924, 647, 452 and 317. In order to improve the accuracy of sampling, a simple dichotomy method was employed to find the most appropriate number of samples. Geostatistical analysis suggested that the spatial variability of SOM distribution was moderate, and structural factors showed that SOM was affected by human activities in the research region. Six high spatially variated areas were found by gradient calculation on the basis of Kriging interpolation, and the local Moran’s I showed great anisotropy of SOM distribution and significant spatial variability in these regions, which suggested that more sampling sites are needed in these regions to obtain accurate spatial interpolation. Independent verification showed that the number of sampling points was positively correlated with the modeling accuracy and the reasonable number wasl was between 452 and 657 in this area. Finally, the appropriate sampling number was determined as 598 by dichotomy method. The results can provide guidance for monitoring and controling farmland quality in Jianghan Plain.