In the framework of studies on the nutritional value of rangeland plants, it can sometimes be useful to have an indication of the lipid content of samples, although it is not a major component of plant vegetative parts. Heterogeneous databases gathering very diverse plants or plant parts require a very large number (several hundreds) of chemical analyses for calibration of chemical composition, so that lipids are generally not considered in such databases. The present study intends to seek for a proxy of lipid content as a single wavelength in the NIR spectrum. The objective is not to predict lipid content, but to be able to rank the plants according to their lipid content. The study is based on samples from a study on Mediterranean shrubby rangelands (“garrigue”) with mixed vegetation: grasses, shrubs, trees (Silué et al., 2016). About 250 samples from 60 species were collected in Corconne (southern France). Samples were dried mildly (55°C) and ground (1mm sieve). Spectra were collected on a FOSS NIRSYSTEM 5000 spectrometer with a wavelength range 1100-2500nm (2mm step). Then 30 samples were selected to represent the botanical diversity and the expected range of lipids. Lipid content was assessed by crude fat analysis by extraction with petroleum ether on Soxhlet. The approach was to correlate lipid content with absorption at individual wavelength, in order to identify wavelengths better representing lipids. This analysis was performed on raw spectra as well as on spectra pretreated with different derivation orders and smoothing options. The correlogram for 2nd derivative is shown as an illustration on Figure 1. The highest correlations obtained with the different pretreatements were R²=0.04 (at 1402nm) with raw spectra, 0.29 (at 1760nm) with 1st derivative, 0.63 (at 1772nm) with 2nd derivative, 0.64 (at 1782nm) with 3rd derivative. The correlation with wavelengths in the 1760-1780nm is high enough to provide a useful information on lipid content. When applied to t