Journal of near infrared spectroscopy vol:13 issue:2 pages:99-107
In this study, the reflectance and transflectance sample presentation mode were compared for the analysis of the nutrient content of hog (Sus domesticus) manure using visible and near infrared (vis-NIR) spectroscopy. A total of 194 hog manures, which were collected in the spring of 2004 from farms in the northern part of Belgium, were assayed by conventional wet chemical analysis and spectroscopy for the following constituents: dry matter content (DM), organic matter content (OM), pH, total Kjeldahl nitrogen (Ntot), ammonium nitrogen (NH4-N), phosphorus (P), potash (K), calcium (Ca), sodium (Na) and magnesium (Mg). Samples were scanned with a Foss NIRSystems Model 6500 scanning monochromator in reflectance and transflectance mode, respectively. A ceramic reference was measured in between the two modes. The monochromator was equipped with a DCFA sample presentation unit and ranges from 400 to 2498 nm. Partial least squares regression was employed to relate the spectral information to the nutrient content. The PLS models were calibrated for both sample presentation modes using leave-one-out cross-validation. The results of this study showed that the transflectance mode performed better than the reflectance mode. From the transflectance measurements, very good quantitative predictions for total N, good quantitative predictions for K, DM and OM, approximate predictions for NH4-N, P and Mg, very approximate predictions for Ca and a discrimination between high and low values for Na were obtained. pH was not predictable. The reflectance measurements were able to provide good quantitative predictions for total N and K, approximate quantitative predictions for NH4-N, very approximate predictions for DM, OM, P and Mg and discrimination between high and low values for Ca. Na was even less predictable and pH might be unpredictable.