The knowledge of weed spatial distribution allows specific site management that leads to herbicide savings and environmental quality protection. This work aimed to evaluate the spatial distribution of weeds in a clay Oxisol in Campinas, Sao Paulo State, Brazil during the period between 2003 and 2008. The sampled area measured 3.42 ha and featured 302 sampling points, in a 10 x 10 meter regular grid. Based on high and very high values of coefficient of variation and on its asymmetry, it was shown that inferences of mean values did not represent the sampled population of weeds. Also, the semivariograms showed a spatial pattern of data for weeds, even if they were separated according to its leaf morphology. The clustered weed distribution were well designed at the kriged maps and also observed according to distribution index. With the geostatistical approach, it was possible to clear mapping site-specific managing zones with 5 to 10 fold differences in weed count, in order to confirm that this technique can be used as ancillary tool in agricultural practices.
geostatistics; krigging; mapping
