Análisis de componentes principales con datos georreferenciados: Una aplicación en agricultura de precisión

new precision agriculture technologies allow collecting information from several variables at many georeferenced locations within crop fields. the spatial covariation of soil properties and crop yield data can be evaluated by principal component analysis (pca). nevertheless, pca has not been explicitly developed for spatial data as other multivariate descriptive methods. other multivariate techniques that include spatial autocorrelation among data of neighborhood sites have been recently developed. in this paper, we apply and compare two multivariate analyses, pca and spatially constrained multivariate analysis methods (multispati-pca). the latter incorporates the spatial information into multivariate analysis calculating moran's index between the data at one location and the mean values of its neighbors. the results showed that multispati-pca detected relations in the data that were not detected with pca. the mapping of spatial variability from the first principal component was similar between pca and multispati-pca, but maps from the second component were different due to the variance correction by spatial autocorrelation. multispatipca method represents a crucial tool to map spatial variability within a field, and to identify homogeneous zones in a multivariate sense.