A genetic algorithm for resizing and sampling reduction of non-stationary soil chemical attributes optimizing spatial prediction

Aim of study: To evaluate the influence parameters geostatistical model and initial sample configuration used in optimization process; to propose resizing a configuration, reducing its size, for simulated data study spatial variability soil chemical attributes under non-stationary with drift process from commercial soybean cultivation area.Area Cascavel, BrazilMaterial methods: For both, attributes, Genetic Algorithm was resizing, maximizing overall accuracy measure.Main results: The results obtained showed that practical range did not relevant way process. Moreover, local variations, such as variance or sampling errors (nugget effect), had direct relationship reduction mainly smaller nugget effect. efficient since it generated configurations 30 35 points, corresponding 29.41% 34.31% respectively. In addition, comparing optimized configurations, similarities were regarding dependence structure characterization area.Research highlights: is possible reduce allowing lesser financial investments collection laboratory analysis samples future experiments.