Efficient use of spontaneous vegetation in Mediterranean rangelands is challenged by strong spatiotemporal variability in forage quantity and quality. This heterogeneity, exacerbated by climate change, complicates real-time assessment and the availability of shared, operational knowledge on forage value. This study aimed to develop a functional classification of rangeland resources to support grazing management. In spring and autumn 2024, we monitored the feeding behaviour of sheep and goats on three farms in southern France. We analysed 165 samples of plant parts consumed by animals using nearinfrared spectroscopy (NIRS). Unsupervised classification primarily grouped the samples by plant part (leaf, stem, etc.) rather than botanical species, and captured seasonal shifts in resource quality and availability. The resulting functional classes provided a simplified but informative view of the grazing diet, revealing species-specific feeding behaviours. In parallel, interviews with 18 pastoral farmers explored their knowledge and expectations regarding rangeland use. They assessed forage value based on plant species, animal behaviour, and seasonal changes. By combining grazing behaviour, NIRS analysis, and local knowledge, we developed a typology validated through farmers' feedback and field observations, that will serve a decision-support tool to optimise the use of spontaneous vegetation in Mediterranean pastoral systems.