Estimation of macronutrients content in mixed swards by Near Infrared Reflectance Spectroscopy

Abstract Text:

Near Infrared Reflectance Spectroscopy (NIRS) is a rapid, non-destructive and inexpensive technique for providing chemical analysis of forage and feedstuffs. The aim of this study was to develop NIRS calibration equations to determine macronutrients (P, Ca, Mg and K) content in mixed swards. Two hundred twenty samples were taken from mixed ryegrass (Lolium perenne L. and Lolium multiflorum Lam.) and clover (Trifolium repens L. and Trifolium pratense L.) swards located in a grassland area from Galicia (NW Spain). Reference data that were used for developing calibration equations were analyzed using standard laboratory methods widely applied for determination of macronutrients content in swards samples (P was measured colorimetrically as molybdovanado-phosphoric acid and Ca, Mg and K were measured by atomic absorption spectrophotometry). All the samples were scanned with monochromatic radiation from 1100 to 2500 nm using NIRSystems 6500 (FOSS Analytical AS, Denmark). Predictive equations were developed using modified partial least squares (MPLS) regression with internal cross-validation and scatter correction using standard normal variate (SNV) and detrend. Cross validation was used to avoid overfitting of the equations. The best equations of MPLS regression for the four macronutrients were obtained by using the second derivative of spectra than using the first derivate. The coefficient of determination in calibration (R²_c), cross validation (R²_cv) and prediction (R²_p) and the standard error of calibration (SE_C), cross validation (SE_CV) and prediction (SE_P) found for each macronutrient were: P (R²_c= 0.72 and SE_C= 0.43; R²_cv= 0.61 and SE_CV= 0.51; R²_p= 0.70 and SE_P= 0.46), Ca (R²_c= 0.90 and SE_C= 1.44; R²_cv= 0.69 and SE_CV= 2.50; R²_p= 0.84 and SE_P= 1.35), Mg (R²_c= 0.76 and SE_C= 0.19; R²_cv= 0.58 and SE_CV= 0.26; R²_p= 0.60 and SE_P= 0.20) and K (R²_c= 0.82 and SE_C= 3.46; R²_cv= 0.71 and SE_CV= 4.37; R²_p= 0.74 and SE_P= 2.98). Taking into account RPD values for validation, calculated as standard deviation of reference data (SD) divided by SE_P, a good quantitative prediction of Ca (RPD= 2.27) and approximately of P, Mg and K (RPD= 1.79, 1.66 and 1.95, respectively) were achieved. Nevertheless, these results might be improved by adding samples to the calibration set.

Keywords: minerals, NIRS, ryegrass-clover pastures.