Influence of geostatistical rigor on the quality of maps used in precision agriculture

Abstract

Maps used by precision agriculture to prescribe local application are based on sample grid and interpolation. Interpolation by kriging presents some difficulties related to its usage since it's a complex method and considers several parameters for calculation (i.e. anisotropy, trend, nugget effect, partial sill, range) and when they are not used it causes loss of quality in the analyses, reducing reliability of maps created by this method. Accordingly, this study aims to evaluate if a rigor in geostatistical modelling leads to expressive gain of quality of the maps used by precision agriculture. To test this, it will be studied the gain on predictive quality by treating anisotropy and tendency found in data, considering two sample grids (1 sample/ ha and 1 sample/4ha), besides of calculating semivariogram by method of moments and method of restricted maximum likelihood. These parameters will be tested in a simulated virtual field produced by a process of Gaussian Unconditional Simulation calculated by gstat package on software R. In addition, the parameters will be also tested in two experimental fields that present conditions of trend and anisotropy. The results obtained by the analyses will be evaluated through comparison of predicted values by interpolation and observed values obtained in the fields, in addition will be carry out a sensibility analyses and error propagation to identify the impact of each parameter on the estimation error's obtained by interpolation of maps using geostatistics. (AU)