Regional-scale modelling of the spatial distribution of surface and subsurface textural classes in alluvial soils using Markov chain geostatistics

Soil texture is directly associated with other soil physical and chemical properties and can affect crop yield, erodibility and water and pollutant movement. Thus, maps of soil textural class are valuable for agricultural management. Conventional spatial statistical methods do not capture the complex large-scale spatial patterns of multi-class variables. Markov chain geostatistics (MCG) was recently proposed as a new approach for the conditional simulation of categorical variables. In this study, we apply an MCG algorithm to simulate the spatial distribution of textural classes of alluvial soils at five different depths in a 15-km² area on the North China Plain. Soil texture was divided into five classes – sand, sandy loam, light loam, medium loam and clay. Optimal prediction maps, simulated maps and occurrence probability maps for each depth were generated from sample data. Simulated results delineated the distribution of the five soil textural classes at the five depths and quantified related spatial uncertainties caused by limited sample size (total of 139 points). These results are not only useful for understanding the spatial distribution of soil texture in alluvial soils, but also provide valuable quantitative information for precision agriculture, soil management and studies on environmental processes affected by surface and subsurface soil textures.