Abstract:

A study on distribution of soil nutrient on the field scale is important for improving agricultural management，and for assessing effects of agriculture on environmental quality and degrees of the influence of some random factors.However，soil nutrients are highly heterogeneous whether on a large scale or a small scale，and their heterogeneity results from many processes acting and interacting across a continuum of spatial and temporal scales.In this paper，a new method setting suppositional samples over regions with no measured data is presented.According to the high relationship between soil property and other physical factors，the data of suppositional samples could be simulated.Then a geostatistical analysis and spatial interpolation method could be used with the enough samples data.This method doesn’t completely rely on the relationship of soil property and other physical factors，and could be an easy way for soil property interpolation over regions with sparse measured samples.A case study in Pengshui County using geostatistics of ArcGIS shows that (1)Tridimensional characters of soil property is represented reasonably by the spatial prediction model based on geographical factors；(2)the improved method integrating geographical environmental factors and spatial distributions of regional soil properties could get a better result with the least prediction error and more abundant spatial information than the traditional methods.

Key words: geostatistics/spatial prediction model/ spatial interpolation/sparse samples/ hilly region