Comparison between energy dispersive X-ray fluorescence spectral data and elemental data for soil attributes modelling

Abstract

Energy dispersive X-ray fluorescence (EDXRF) has been successfully applied for soil attribute prediction. Although, both spectral and elemental EDXRF data may be used in soil fertility attributes modelling, as far as we are concerned, there are no studies that compare the models performance using the two input possibilities. The objective of this study was to compare the performance of partial least squares regression models for soil fertility attributes prediction using different EDXRF inputs (spectral data and peak intensity data) by the evaluation of some figures of merit such as linearity, trueness, precision and robustness. The results show that the input EDXRF dataset type used in the modelling did not significantly interfere in the fertility attributes prediction, valid for the soils evaluated in the present work. However, the use of peak intensity data presented lack of linearity and robustness in some models. Moreover, the present study report relevant protocols for analytical validation in soil fertility attributes prediction when EDXRF data is applied.