Original scientific paper
PREDICTION OF DRY MATTER YIELD FROM SEMINATURAL GRASSLAND BY NIR SPECTROSCOPY
Marina Vranić
; University of Zagreb, Faculty of Agriculture, Department of Field Crops, Forage and Grasses, Grassland Research Centre, Zagreb, Croatia
Krešimir Bošnjak
; University of Zagreb, Faculty of Agriculture, Department of Field Crops, Forage and Grasses, Grassland Research Centre, Zagreb, Croatia
Josip Leto
; University of Zagreb, Faculty of Agriculture, Department of Field Crops, Forage and Grasses, Grassland Research Centre, Zagreb, Croatia
Božica Lukšić
; University of Zagreb, Faculty of Agriculture, Zagreb, Croatia
Siniša Glavanović
; Belupo Inc., Research and Development, Koprivnica, Croatia
Abstract
Near-infrared (NIR) spectroscopy (1100 – 2500 nm) was used to analyze the dry matter (DM) yield from seminatural grassland community Arrhenatheretum medioeuropaeum. Modified partial least square (MPLS), principal component regression (PCR) and partial least square (PLS) techniques were used for spectral data processing. Total of 225 forage samples were used in this investigation. Standard errors of calibration (SEC) were 0.71, 0.71 and 0.77 for MPLS, PLS and PCR, respectively, while standard errors of cross validation (SECV) 0.80, 0.77 and 0.77 respectively. Comparing NIRS and the chemical procedure the standard error of prediction (SEP) for MPLS, PLS and PCR were 0.748, 0.698 and 0.873 respectively. The results show the great potential of NIR spectroscopy for DM yield prediction in samples originated from semi natural grassland. Based on SEC, SEP and RSQ (R2) the PLS method is the most relailable to predict DM yield from seminatural grassland samples followed by MPLS method. The PCR method was at least reliable for DM yield prediction in this research..
Keywords
dry matter yield; semi natural grassland; NIR spectroscopy; PLS; MPLS; PCR
Hrčak ID:
180241
URI
Publication date:
21.4.2017.
Visits: 1.385 *