Laser-induced breakdown spectroscopy applied to cattle compost for phosphorus quantification

Abstract

Purpose Agronomic and environmental reasons force farmers to know the total P concentration of composted cattle manure. Laser-Induced Breakdown Spectroscopy seems proper to obtain such information. For logistic reasons (carriage, storage, field application, etc.), a dry matter characterization is also needed.
Method Thirty samples of feedlot compost at different stages of stability and maturity were studied. Samples were dried at 50°C for dry matter characterization. As a reference method to determine total P concentration, wet digestion and colorimetry were employed. The area of the P I line emission obtained by laser-induced ablation of the samples was measured to estimate the total P concentration. Randomized calibrations through a modified version of the Kennard-Stone algorithm based on the Mahalanobis distance were performed.
Results Dry matter varied from 40% to 90%, and no pattern was found related to compost origin, maturity, or stability. The total P concentration of the studied compost ranged from 1800 ppm up to 11200 ppm. Almost 80% of the calibration fittings have an R² ≥ 0.895. The mean validation error was less than 22% for about 80% of the calibrations, with a mean prediction error bound to 40%. Discarding outliers, the errors were reduced to 19% and 30%, respectively.
Conclusion Water content must be considered in addition to other characterizations due to logistic implications. Calibrations with a 30 percent of prediction error were achieved, which seems enough as a first approximation to predict the total P content in compost for utilization in farms to recycle nutrients.

Highlights

• Agronomic and environmental reasons force farmers to know the total P concentration of composted cattle manure.
• Laser-Induced Breakdown Spectroscopy seems proper to obtain such information.
• As a reference method to determine total P concentration, wet digestion and colorimetry were employed.
• Randomized calibrations through a modified version of the Kennard-Stone algorithm based on the Mahalanobis distance were performed.
• Discarding outliers, 80% of the calibration fittings have an R² ≥ 0.92 and the mean validation and prediction errors were reduced to 19% and 30%, respectively.