Consistent crown phenology, not functional traits, predicts whether tropical trees can be mapped from imaging spectroscopy

- Airborne imaging spectroscopy can identify tree species in hyperdiverse tropical forests, but accuracy varies greatly among species. We asked which ecological and evolutionary attributes make a tropical tree species spectrally separable. - In moist tropical forest in French Guiana, we fitted the learning curve of accuracy against the number of labelled crowns for 169 species, then tested six hypothesised determinants on 45 species rarefied to ten crowns, at species, species-pair and crown scales. - Dispersed sampling and larger crowns increased accuracy, and liana infestation more than halved the odds of correct identification. Phenological regularity, how closely conspecific crowns share one leaf cycle, in step or not, was the strongest species-level correlate and survived every multivariate analysis. It did not act by making conspecifics uniform on the acquisition date: regular species were spectrally more distinct from other species, and within species the crowns that best followed the species' leaf cycle were classified correctly. Congeners reduced accuracy; leaf and bark traits did not predict separability. - Separability depends on the consistency of a species' signal, how faithfully its crowns express a shared leaf cycle, how widely it is sampled and how free its crowns are from contamination, rather than on its measured functional traits.