Geostatistical analysis of soil reflectance spectra for field-scale digital soil mapping. A case study.

Knowledge of field-scale soil variability is essential for sustainable soil management. Traditional techniques, based on soil analysis, are costly and timeconsuming. An alternative method would be the use of visible-infrared reflectance spectroscopy coupled with multivariate analysis, specifically principal component analysis (PCA) and geostatistics. In this study, after brief reviews regarding reflectance spectroscopy, PCA, and geostatistics, we presented a methodological approach for digital soil mapping in a study area of Southern Italy. Reflectance spectra of 240 surface soil samples collected at geo-referenced sites, were decomposed by PCA. The first three components (PC1, PC2, PC3) explained most (98%) of the total variance of the initial data set, therefore, they were considered for the assessment of soil spatial variability by variography and kriging (geostatistics). The resulting PC1, PC2 and PC3 kriging maps were interpreted in the light of the information contents on reflectance spectra and compared with the results of a previous, conventional soil survey. The presented strategy seems to be efficient and reliable for mapping soil spatial variability.