Three dimensional spatial distribution modeling of soil texture under agricultural systems using a sequence indicator simulation algorithm

Understanding the distribution of alluvial soil textures on a large scale is crucial for agricultural and environmental management. In our study an indicator variogram and a sequence indicator simulation (SIS) algorithm were used to analyze and simulate the spatial distribution of soil textures based on observations of 139 soil profiles in a 15 km2 region in the Huabei alluvial plain in China. The nugget-to-sill ratio value (SH) of the indicator variograms for all textures in a vertical direction (Z) was equal to 1. This suggests that spatial auto-correlation dominates in the direction of sedimentary deposition with 0.05 m sampling intervals. In contrast, SH ratios from 0.48 to 0.81 show that the soil textures have a degree of randomness in the horizontal direction (X, Y) where the sampling distance was about 300 m. Using the indicator variograms in 3 directions (X, Y and Z) as outlined above, a 3D SIS algorithm was used to simulate textures. Finally, the simulation results were evaluated by the reproduction of a histogram, variogram and the mean absolute error (MAE) of prediction. The mean absolute percentage error (MAPE) of the histogram reproduction showed that the main textures (sand, sandy loam and clay) were described well, whereas the less prevalent textures were underestimated. The MAPE of the indicator variograms reproduction were reasonable although some deviation existed as less prevalent textures in the vertical direction. The mean absolute error (MAE) of the SIS prediction was 0.47. This result is considered acceptable for a category variable because of the stochastic nature of soil textures in a horizontal direction, and hence may provide useful data for other agricultural research.