Spatial variability of soil nutrition in the typical landscape unit of red soil hilly in subtropical region

Based on regionalized variable theory and spatial analysis functions of GIS,spatial heterogeneity of soil properties in the typical landscape ecological units of subtropical region was studied using semivariogram of geostatistics in this paper.The results showed that theoretical semivariogram models of soil organic carbon,and soil total P were spherical one,the best model for semivariograms of soil total N was gaussian model and that of pH was exponential model. Soil(organic) carbon,soil total N,and pH belonged to high and soil total P belonged to medium degree spatial correlation at 60m lag intervals,respectively.However,there were significant differences in ranges of spatial autocorrelation among four soil properties,among which the effective range of soil organic carbon was the highest(261m),followed with soil total N(208m),pH(156m),and soil total P(133m).The differences in their ranges of spatial autocorrelation were dependent on their properties,land use pattern,topography and man-made factors.Kriging method,under the support of GIS,was applied to analyze the spatial distribution of these four soil properties.The results indicated that spatial distribution (patterns) of organic carbon and soil total N were similar appearing an apparent belt-shaped and spot massive gradient change,while that of soil total P was strong random distributed and its patches were more trivial showing a star-studded pattern.Spatial distribution pattern of soil pH did not show a distinctive strip belt-shaped distribution.In general,the fitting results using semivariograms of four soil properties at the scale of whole landscape scales were better than that at the scale of different land use types.