Modeling and quantification of available potassium spatial uncertainty in the soil by stochastic simulations

The objective of this work was to evaluate the performance of the sequential Gaussian simulation (SGS) and the sequential indicator simulation (SIS) for modeling the uncertainty of available K predictions in a sugarcane area, and to compare both simulations to the already established method of ordinary kriging (OK). A sampling grid with 626 points was installed in an area of 200 ha, in the municipality of Tabapua, in the state of Sao Paulo, Brazil. The simulations reproduced the variability in the available K sample data, whereas OK overestimated the low K levels and underestimated the high ones. The standard deviation map obtained from OK showed less variation along the studied area when compared to the maps obtained from the simulations. SIS achieved an accuracy 22% higher than that obtained by SGS for modeling the conditional distribution function of K. The simulations have higher efficiency than OK for modeling the uncertainty in the spatial distribution of K. SIS has better performance than SGS for estimating the levels of available K in sugarcane area.