Remotely sensed functional diversity and its association with productivity in a subtropical forest

Functional diversity is a critical component driving ecosystem functioning. Spatially explicit data of plant functional traits and diversity are essential for understanding biodiversity effects on ecosystem functioning. Here we retrieved three morphological traits (95th quantile height, leaf area index, foliage height diversity) and three physiological traits (chlorophyll a + b content, specific leaf area, equivalent water thickness) from airborne laser scanning and multispectral Sentinel-2 data, respectively. We found airborne LiDAR-derived parameters correlated well with in-situ plot-level morphological data (R2 ≥ 0.67). For satellite-derived physiological traits, partial least squares regression (PLSR) obtained higher prediction accuracy (R2 = 0.26–0.43, cross-validation with in-situ community-weighted mean (CWM) leaf physiological trait data) than a vegetation index (VI) approach. The remotely-sensed traits were used as input to estimate multi-trait functional diversity (FD) indices in a species-rich subtropical mountainous forest. Finally, we investigated the influence of single-trait CWMs, multi-trait FD indices and environmental variables on remotely-derived aboveground ecosystem carbon stocks (aboveground biomass, AGB) and primary productivity (kernel normalized difference vegetation index, kNDVI). CWMs of all functional traits were significant predictors of AGB and kNDVI, as suggested by the mass-ratio hypothesis. Morphological FD indices were also important predictors of AGB and kNDVI, indicating effects of complementarity in crown architectures. In best-fit multivariate models, the first principal component CWM of morphological traits and that of physiological traits were the most important predictors of AGB and kNDVI, respectively. The FD index of morphological richness was additionally selected in the best-fit models for AGB and kNDVI at ecosystem and landscape scales. Our work highlights the potential of using remotely-sensed functional traits to asses