Quantitative Analysis of Agricultural Compost Indicator Factors Based on Different Nir Feature Variable Selection Methods

Composting is an effective method for managing livestock manure and crop straw. An exhaustive monitoring of the composting process and of the final compost characteristics is necessary to gain well-composted product suitable for application to agricultural lands. The near infrared reflectance spectroscopy (NIRS) provided non-destructive, fast and relatively cheap methodologies for qualitative and quantitative analysis of different compounds in agricultural composting. In this study, the near-infrared spectroscopy analysis technology was used to determine compost properties (organic matter (OM) content, total nitrogen (TN) content and carbon-nitrogen(C/N) ratio) based on more than 100 samples from 2 different composting procedures and the models built by using partial least squares regression (PLSR). The multiplicative scatter correction (MSC), stepwise linear regression (SR), Synergy Interval Partial Least Squares(siPLS), successive projections algorithm (SPA), synergy interval partial least squares (siPLS) and linear baseline correction (LBC) were employed as spectra pretreatment. The performance of LBC-siPLS-PLSR for OM content , MSC-SPA-PLSR for TN content and R-SPA-PLSR for C/N ratio was evaluated respectively according to root mean square error (RMSE) of prediction (RMSEP, OM was 4.061, TN was 0.205 and C/N 1.11), the coefficient of determination for prediction (R p 2 of OM 0.746, TN 0.65 and C/N 0.706 ) and residual predictive deviation(RPD) being obtained for this latter values of 2.02, 1.71 and 2.07 for OM, TN and C/N, respectively. These results showed that the NIRS technique needs to be fitted to each element and property, using specific spectrum pretreatment, in order to achieve an acceptable accuracy in the prediction.