Forest canopy structure (CS; the quantity and arrangement of vegetation within the canopy) drives ecosystem functions such as light interception, carbon cycling, etc. and may contribute to ecosystem resilience over multiple spatial scales. High resolution measurements of CS have been restricted to small plots, limiting our understanding of broad patterns of CS and its potential influence on resilience of forest processes. However, better estimates of macroscale CS patterns could improve understanding of the spatial patterns of ecosystem functions. We characterized the relationship between CS and ecosystem functions at a range of scales using three parallel approaches:
- We evaluated the potential to estimate CS from Landsat data by comparing Landsat vegetation indices with metrics of CS derived from terrestrial LiDAR at National Ecological Observatory Network (NEON) sites.
- We assessed scales of CS variation using terrestrial LiDAR to quantify CS along 140 – 800 m transects in forested landscapes within the eastern USA. We calculated stability points and autocorrelation lengths for individual CS metrics along each transect.
- We tested linkages between CS and a suite of ecosystem functions at content-wide subset of NEON forest sites to assess the generality of these relationships within and across scales.
Results/Conclusions
Our results indicate that while CS varies significantly across scales from stand to continent, it does so predictably and with consistently strong links to ecosystem functions. We observed that canopy spectral reflectance is influenced by aspects of CS including stand height, leaf area density and variability, and canopy openness. This indicates satellite imagery is sensitive to the physical arrangement of forest canopies, suggesting the potential to estimate CS at macroscales. We also documented that values for a suite of CS metrics consistently achieved stability over a range of ~300 m. This value corresponds well with the spatial extent at which many ecosystem functions are measured and modeled, e.g., NEP from eddy covariance towers. CS metrics were spatially autocorrelated at 40 to ~90 m, suggesting patch-scale dynamics shape CS at these scales, a finding consistent with contemporary approaches to ecosystem modeling. The strength of the relationship between CS and ecosystem functions tended to vary depending on the particular function and the scale of observation but CS was often a better predictor of functioning than more conventional measures (e.g. biodiversity). Taken together, these results underscore that CS is a robust continental predictor of forest ecosystem functions.