Résumé
Aims: Tropical forests face significant threats from climate change, land-use change, and anthropogenic pressures. Effective conservation requires repeated, rapid, and accurate biodiversity assessments over large areas to identify priority sites and evaluate the efficiency of protection efforts. High-resolution hyperspectral imaging has been proposed for estimating biodiversity by linking spectral variance to species diversity. However, inconsistencies in the relationship between spectral variance and taxonomic diversity raise concerns about the reliability of this approach. This study addresses two critical questions: (1) How does intraspecific spectral variance impact the relationship between total spectral variance and taxonomic diversity? (2) How do uneven spectral distances between species influence the accuracy of biodiversity estimates derived from spectral data? We aim at emphasizing that while hyperspectral data offer valuable insights, its limitations must also be acknowledged. Location: The research was conducted in a hyper-diverse tropical forest located at the Paracou experimental site in French Guiana, using spectral data from airborne hyperspectral imagery. Methods: A simulation-based approach was used to generate artificial communities with controlled taxonomic diversity by drawing pixels from a spectral database derived from a forest inventory of trees with a diameter at breast height (DBH) > 10 cm. Intraspecific spectral variance was manipulated to evaluate its effect on total spectral variance and taxonomic diversity. The Rao index, integrating species abundance and spectral dissimilarity, was employed to assess the impact of uneven interspecies spectral distances. Results: Intraspecific spectral variance was a major contributor to total spectral variance, often exceeding interspecific variance contributions, even in taxonomically diverse assemblages. Nonuniform spectral distances between species weakened the correlation between spectral variance and taxonomic diversity. This suggests that spectral variance is not a reliable metric for biodiversity estimation in complex ecosystems. Conclusions: This study highlights challenges in using hyperspectral data for biodiversity estimation in tropical forests. Advanced methods may reduce environmental noise but the high intraspecific diversity and variable interspecies spectral distances pose substantial barriers. Future research should consider alternative diversity metrics, such as functional diversity, which may offer more consistent correlations with spectral data