Legume-Grass-Mixtures provide valuable forage for livestock, while at the same time they are of high importance for fertilization through nitrogen-fixing, especially in organic agriculture. Through the relationship of fixed nitrogen with N-content in the aboveground biomass of legumes, the percentage of legumes in the total dry matter yield (PL) can be estimated with remote sensing methods. The present study utilized multispectral data captured with a UAV of alfalfa-grass-mixtures (Medicago sativa) from two fields in Hesse, Germany, to model PL. Using PLSR, PL could be modelled with high (RMSEval 12.2 %, R²val 0.88) and medium accuracies (RMSEval 11.01 %, R²val 0.52). It could be shown that UAV-borne remote sensing is well suited to model legume content with a high spatial resolution and coverage. However, the results revealed that modelling accuracies decreased when incorporating data from a region within a field with low variability in PL. Thus, we suggest focusing future research on alfalfa-grass fields with a high variability in PL, which is exactly where additional information for farming practices is needed most.