Spectral analysis of meat combined with chemometric analysis has been identified as a promising tool for authenticating livestock-animal diets. The objectives of this study were (i) to determine whether the visible-NIR spectrum of perirenal adipose tissue (PAT) and caudal adipose tissue (CAT) can reliably discriminate lambs pasture-finished for different durations before slaughter, and (ii) to analyze the kinetics of appearance and stabilization of the visible-NIR spectrum-based pasture signature in PAT and CAT. Four groups of 50–55 lambs were used over three years: lambs finished on lucerne pasture for 0 (L0, concentrate-fed in stall), 21 (L21), 42 (L42) and 63 (L63) days before slaughter. Partial least squares discriminant analysis was applied on PAT or CAT visible-NIR spectra to discriminate the groups. No one adipose tissue reliably discriminated the four groups, with less than 62 % lambs correctly classified. However, visible-NIR spectroscopy was able to discriminate stall-fed (L0) from pasture-finished (L21 + L42 + L63) lambs, with an accuracy of 93.8 % and 87.5 % lambs correctly classified based on PAT and CAT spectra, respectively. The lucerne pasture fingerprint (or signature) on visible-NIR spectrum appeared between 0 and 42 days in more than 95 % of lambs. It stabilized between 42 and 63 days in CAT, but had not stabilized within the range of grazing durations pre-slaughter explored in PAT. Further research into shorter and longer pasture-finishing durations could help determine more precisely the time required for the pasture signature to appear and stabilize in animal tissues.