Green methodology for soil organic matter analysis using a national near infrared spectral library in tandem with learning machine

Abstract

Precision agriculture requires faster and automatic responses for fertility parameters, especially regarding soil organic matter (SOM). In Brazil, the standard methodology for SOM determination is a wet procedure based on the oxidation of the sample by an excess of potassium dichromate based on Walkley-Black method. This methodology has serious drawbacks, since, at a national level, generates approximately 600,000 L/year of toxic acid waste containing Cr³⁺ and possibly Cr⁶⁺, besides time consuming and expensive. Herein, we present a faster green methodology that can eliminate the generation of these hazardous wastes and reduces the costs of analysis by approximately 80%, democratizing the soil fertility information and increasing the productivity. The methodology is based on the use of a national near infrared spectral library with approximately 43,000 samples and learning machine data analysis based on a random forest algorithm. The methodology was validated by submitting the prediction results of 12 blind soil samples to a proficiency assay used for fertility soil laboratories qualification, receiving the maximum quality excellence index, indicating that it is suitable for use in routine analysis.

Keywords: Learning machine; Near infrared spectroscopy; Random forest; Soil; Spectral library.