Abstract The potential of an imaging spectroscopy system with high spatial (0·16–1·45 mm2) and spectral resolution (5–13 nm) was explored for monitoring light interception and biomass of grass swards. Thirty-six Lolium perenne L. mini-swards were studied for a total of eleven consecutive growth periods. Hyperspectral images and light interception (LI) were recorded twice weekly. On two dates ground cover was scored visually (GCv). At harvest, leaf area index (LAI), fresh-matter yield and dry-matter yield (DMY) were determined. Classification of images yielded several estimates of the image ground cover (GCi) and the index of reflection intensity (IRI). The GCi was highly correlated with GCv ( = 0·94), LAI ( = 0·88) and LI ( = 0·95, for dense swards under cloudy skies). However, the relationship between GCi and LI depended on sky conditions and sward structure. Under cloudy skies, LI was linearly related to GCi, whereas under clear skies, this relation was logistic. Regression analysis of GCi and yields showed correlations with of between 0·75 and 0·82. The mean error of DMY estimates was 340 kg. In conclusion, estimates of GCi and IRI can be used to predict DMY, even for high yield levels (up to 3500 kg DM ha−1), allowing accurate, non-destructive monitoring of biomass and light interception of grass swards.