Explaining discrepancies between spectral and in-situ plant diversity in multispectral satellite earth observation

In light of the ongoing global biodiversity crisis, urge to monitor and map terrestrial plant at large spatial extents has spurred research on adequate quantitative methods. The use spectral diversity metrics from different remote sensing platforms emerged as a promising tool for such assessments. Satellite presents next frontier implementation these methods assess with temporal continuity truly regional or scales. However, question what exactly is monitored by relatively coarse multi-spectral satellite observations remained largely unanswered. this research, we examined which components contribute remotely sensed diversity. We assessed relationships between in-situ taxonomic trait diversity, evaluated role confounding factors, vegetation cover, landscape morphology (slope elevation), in shaping relationships. Hereto, used Sentinel-2 imagery field species count data collected Montesinho-Nogueira Natura 2000 site (Portugal) together radiative transfer models quantify theoretical link simulated Through linear mixed-effect models, our results highlight that variation cover dominates Sentinel-2's signal (contributing 53–84% R2marginal). component encompasses variability canopy architecture traits well fraction bare soil litter spectra. These elements strongly impact overall signal, shown both simulations empirical comparisons. Next found significant predictor covariate while leaf appeared insignificant multispectral dataset. Variation dominated study it not necessarily correlated We, therefore, recommend future applications consider including morphological traits, - reflectance better understand ambiguous performance proxy This will result higher robustness, consistency, scalability predicting across scales, sensors, ecosystems