Statistical analyses of fluorometry data from chloroform filtrate of lamb feces

Accurately identifying the botanical composition of free-ranging animal die ts remains a challenge. Currently accepted procedures are time consuming, m any requiring painstaking sample preparation while none produce data useful for real-time management, Automated procedures focusing on detection of ch emical and/or physical plant properties using specific molecules called flu orophores offers possibilities for determining the species composition of h erbivore diets. This study was designed to evaluate fluorometry techniques in herbivore diet determinations using fecal samples obtained from 13 lambs fed a basal diet of tobosa hay (Pleuraphis mutica Buckley), and containing 4 different levels (0, 10, 20, and 30%) of tarbush (Flourensia cernua D C, ) leaf material. Chloroform (CHCl3) filtrate obtained from the lamb's feces was exposed to UV light from a xenon are lamp, This caused fluorophore mol ecules in the filtrate to have their outer shell electrons move to a higher energy state as a result of UV light excitation After excitation by UV lig ht at 310, 320, 330, 340, 350, and 355 nm, the fluorophores returned to the ir ground state giving off light (fluorescence). This fluorescence intensit y (counts) varied and when captured using appropriate electronics, produced 1,024 pairs of light intensities (counts) and fluorescent wavelengths betw een 175 and 818 nm in 0.63 nm increments. Previous research indicated diffe rences among diets could be determined using distinct peaks in the red and blue regions of the visible light spectrum and a univariate (1 variable at a time) analysis. This research demonstrates the entire fluorescence data s et can be used to determine differences among diets using multivariate stat istics, Sequences of 5 increasingly complex statistical techniques were use d to distinguish among diets: 2-dimensional plots, polynomial regression mo dels, confidence interval plots, discriminant analysis, and 3-dimensional p lots. Two-dimensional plots indicated 2 spectral fluorescence peaks, 1 in t he blue-green (420-600 nm) and 1 in the red (640-720 nm) region of the visi ble spectrum, Because of the asymmetrical nature of these peaks, fifth-orde r polynomials were developed to differentiate among the 4 diets. Statistica l reliability was high when discriminating between diets containing no tarb ush leaf and the diets containing 30% tarbush leaf; however, it was not pos sible to statistically separate dirts containing intermediate (10 and 20%) amounts of tarbush leaf material From each other or from the 2 extremes (0 and 30% tarbush leaf). These results suggest spectral signatures arising fr om fluorometry data may be useful for differentiating among botanical compo sition diets that differ in plant form, but that a multivariate approach ma y require large sample sizes.