PLS Regression Based on ATR-FTIR to Predict Organomineral Fertilizers Properties and Nutrient Pools

Abstract

Enrichment of organic residues with mineral fertilizers is a sustainable route to produce high agronomic value organomineral fertilizers (OMFs). OMFs agronomic value was conditioned by the properties and nutrients pools accessed by chemical analysis. Partial least squares (PLS) regression based on infrared analysis is a fast and alternative technique to assess the properties of OMFs, while replacing laborious, non-environmental-friendly, time-consuming, and high-cost conventional lab analytical procedures. OMFs were produced by composting of mixtures of different proportions of low-grade and soluble P sources with chicken manure and coffee husk for 150 days. After composting, the OMFs were dried and analyzed for: pH in CaCl2, electrical conductivity, total contents of C, P, N, and K, and C soluble in water, as well as for fertilizer-P soluble in water, citric acid, and neutral ammonium citrate. The compost MAP-based OMFs had a greater agronomic value than low-grade rock P-based OMFs. PLS regression models based on the ATR-FTIR spectral signature were a suitable tool to predict all OMFs chemical properties and nutrient pools evaluated through lab conventional analytical procedures. The good performance, robustness, and non-random correlation of PLS regression models were attested by their high coefficient of determination (R2) to calibration (0.92–0.99), cross-validation (0.87–0.99), and prediction capacity (0.89–0.99) combined with the lowest values of the root-mean-square error (RMSE) and reduced values of R2 (0.19–0.44), and to high values of RMSE to y-randomization. PLS based on ATR-FTIR is a rapid and alternative chemometric approach to assess the properties and nutrients pools of OMFs.