Fourier transform infrared spectroscopy and partial least square regression for the prediction of substrate maturity indexes

Traditional methods to evaluate the stability and maturity of organic wastes and composting matrices are laborious, time-consuming and generate laboratory chemical wastes. This study focused on the development of partial least square (PLS) regression models for the prediction of the stability and maturity of compost-based substrates based on Fourier transform infrared (FTIR) spectroscopy. The following parameters, selected as conventional maturity indexes, were modeled and used as dataset: dissolved organic carbon (DOC), C/N and NH4+/NO3− ratios, cation exchange capacity (CEC), degree of polymerization (DP), percentage of humic acid (PHA), humification index (HI) and humification ratio (HR). Models were obtained by using data from a wide range of compost based growing media of diverse origin and composition, including 4 commercially available substrates and 11 substrates prepared in our facilities with varying proportions of different organic wastes. The PLS models presented correlation coefficient of calibration (R2cal) close to 0.90 and correlation coefficient (R2) of cross validation (R2cv) presented acceptable values (> 0.6), ranging from 0.67 (HR) to 0.92 (C/N). The good performance of the method was also confirmed by the low correlation obtained from the Y-randomization test. R2 for test samples (R2pred) ranged from 0.66 (C/N) to 0.97 (HI) confirming the good correlation between measured and PLS predicted maturity indexes. FTIR spectroscopy combined with PLS regression represents, after modeling process, a fast and alternative method to assess substrate maturity and stability with reduction of time, lower generation of laboratory chemical wastes residues and lower cost per sample than conventional chemical methods. All models adjusted for maturity indexes are predictive, robust and did not present chance correlation.