LAUSR

Abstract There are several methods to extract soil information by spectral sensing. For this reason, the database should be built including standards and protocols both in the lab and in field acquisitions. If we does not align the measurement one to each other, the models will have no merit in term of large-scale application and stay as an academic exercise only. The use of standard samples with known spectra is expected to allow parameterizing mathematically soil spectra collected by different equipment and with distinct geometries. This study was proposed for the reason that to date, the new methodology proposed has not been evaluated in Brazilian tropical soils (Oxisols) in visible, near and short infrared (VIS-NIR-SWIR – 350 to 2500 nm) spectral regions from three spectrometers (FieldSpec 3) in three distinct protocols (Long Light, Near Light and Contact Probe). The current study compared the spectral intensity of each combination between spectrometers and protocols by ANOVA module and the clay prediction capacity by PLSR with cross-validation, before and after the internal soil standard (ISS) method application. The results showed that visual spectral variation is minimized by the ISS method and its correction enables better proxy modeling of clay content especially if the data are mixed from different protocols. The model for clay prediction was improved showing a favorable case to use the ISS technique in any soil spectral measurement in a better way to merge spectral libraries from different sources.