Towards rapid analysis with XRF sensor for assessing soil fertility attributes: Effects of dwell time reduction

Abstract

The analysis time used for the diagnosis of soil fertility attributes using portable X-ray fluorescence spectroscopy (XRF) sensors (between 30 and 90 s) is too long for in situ applications. The present study aimed at evaluating the trade-off between dwell time and XRF performance for assessing soil fertility attributes. A total of 102 soil samples acquired in two Brazilian agricultural fields were used, whose spectra were obtained using dwell times of 90, 60, 30, 15, 10, 7, 4, and 2 s to build and validate calibration models for clay, cation exchange capacity, and extractable K and Ca. Results revealed that it is possible to make drastic reductions in the XRF dwell time (from 90 to 2 s), while keeping excellent prediction performance [ratio of performance to interquartile distance (RPIQ) between 3.52 and 8.32] for all the studied attributes. A dwell time of only 2 s performed satisfactorily and is an analysis time suitable for rapid in situ applications. In addition, it was shown that data from spectral databases previously collected that used long dwell times (e.g., 30, 60, 90 s) can be extrapolated to fast applications with shorter dwell times (e.g., 2 and 4 s), once standardization by the detector's live time has been performed. Anyhow, calibrations using a dwell time similar to the one of the validation set tended to show superior results and are therefore recommended. This study addresses the need and provides guidelines for optimizing XRF sensor analysis time for in situ applications in the context of precision agriculture.