AbstractAimsThe link between spectral diversity and in‐situ plant biodiversity is one promising approach to using remote sensing for biodiversity assessment. Nevertheless, there is little evidence as to whether this link is maintained at fine scales, as well as to how it is influenced by vegetation's vertical complexity. Here we test, at the community level in grasslands, the link between diversity of the spectral signal (S_Div) and taxonomic diversity (T_Div), and the influence of vertical complexity.MethodsWe used 196 1.5 m × 1.5 m experimental communities with different biodiversity levels. To measure vertical complexity, we quantified height diversity (H_Div) of the most abundant species in the community. T_Div was calculated using the Shannon index based on species cover. Canopy spectral information was gathered using an unmanned aerial vehicle (UAV) mounted with a multi‐spectral sensor providing spectral information via six 10‐nm bands covering the visible and near‐infrared region at a spatial resolution of 3 cm. We measured S_Div in a core area of 1 m ×1 m within the communities as mean Euclidean distance of all pixels in a feature space spanned between the two first components of a PCA calculated for the complete raster stack. We modelled S_Div through mixed‐effect linear models, using T_Div, H_Div, and their interaction as fixed‐effect predictors.ResultsContrary to our expectations, T_Div was negatively linked to S_Div. The diversity in plant height was positively related to S_Div. More importantly, diversity in plant height and T_Div had a significant negative interaction, meaning the more complex the vegetation was in terms of height, the more the S_Div–T_Div relationship became negative.ConclusionsOur results suggest that in order to exploit the S_Div–T_Div link for monitoring purposes, it needs to be contextualized. Moreover, the results highlight that communities’ functional characteristics (i.e. plant height) mediate such a link, calling for new insights into the relation between S_Div and functional diversity.