Abstract
An indirect method for estimating biodiversity from Earth observations is the Spectral Variation Hypothesis (SVH). SVH states that the higher the spatial variability of the spectral response of an optical remotely sensed image, the higher the number of available ecological niches and hence, the higher the diversity of tree species in the considered area. Here for the first time we apply the concept of the SVH to Light Detection and Ranging (LiDAR) data to understand the relationship between the height heterogeneity (HH) of a forest and its tree species diversity, a concept we have named the 'Height Variation Hypothesis' (HVH). We tested HVH in two different European forest types: a coniferous mountain forest in the eastern Italian Alps and a mixed temperate forest in southern Germany. We used the heterogeneity index Rao's Q to estimate HH using a Canopy Height Model (CHM) at different resolutions derived from LiDAR data, and linear regression models and relation analysis to assess the relationships between HH and three species diversity indices derived from in situ collected data: Shannon's H, Simpson's S and species richness. The relationships were calculated for all plots in both study areas, and separately for plots with a defined Canopy Closure (CC 70%, CC > 80%, CC > 90%) to un-derstand the effect of forest density on the relationship between HH and tree species diversity. Our results showed that HH is related to the tree species diversity of the forest ecosystems reaching (in the case of Shannon's H) values of R2 = 0.63 for the coniferous mountain forest and R2 = 0.56 for the mixed temperate forest, par-ticularly when calculated with a CHM resolution of 2.5 m. The associations also increased with increasing ca-nopy closure suggesting that HVH is scale and forest density dependent. Our results also underlined that the abundance-based diversity measures are more highly correlated with HH than with species richness. Finally, our findings suggest that the HVH is a valuable tool for assessing tree species diversity in forest ecosystems, and could also be useful for overall biodiversity estimates.