Spectral heterogeneity from the spaceborne imaging spectrometer EnMAP reveals biodiversity patterns in forest ecosystems

The Spectral Variation Hypothesis (SVH) proposes that spectral heterogeneity (SH), derived from optical data, can serve as a proxy for estimating biodiversity. In this study, we tested the SVH across 42 forest plots in the Italian Alps using imaging spectroscopy data from the EnMAP satellite. We investigated the relationship between SH—quantified using two different metrics, Rao's Q and the coefficient of variation (CV)—and tree species diversity (using Shannon's H index and species richness). We applied three levels of spectral analysis: (1) SH calculated for each individual EnMAP band; (2) SH aggregated across broader spectral ranges (Visible -VIS-, Near Infrared -NIR-, and Shortwave Infrared -SWIR-) and (3) SH derived from vegetation indices (VIs). These analyses were performed under three spatial approaches: (A) a normal approach assigning equal weight to all four EnMAP pixels intersecting a plot; (B) a weighted approach based on the proportional overlap of each pixel with the plot area; and (C) a weighted canopy cover (CC)>70% approach, which included only plots with CC greater than 70% as derived from airborne laser scanning (ALS) LiDAR data. Weak to moderate correlations were observed when SH was derived from single bands, with the strongest relationships in the NIR (R2 approaching 0.4), followed by the VIS and SWIR regions. A similar trend emerged when SH was aggregated across broader spectral ranges, with the highest correlations again found in the NIR (R2 up to 0.35). In contrast, lower R2 values were obtained when SH was computed from specific VIs. The weighted approaches, especially when restricted to plots with CC >70%, consistently yielded higher R2 values than the equal-weight approach in all three the spectral analysis. Results were consistent across both SH metrics (Rao's Q and CV), with stronger correlations when species richness was used as the biodiversity metric. This work highlights how EnMAP hyperspectral data, despite inherent constraints, can provide valuable insights into forest biodiversity monitoring.