Tree growth is an important rate for understanding the dynamics of forested ecosystems. Commonly, information about growth rates comes from times series of inventoried plots. While useful, this information is available for limited areas and may not represent the full spectrum of forest landscapes. Due to its relatively large spatial coverage and high spectral resolution, hyperspectral remote sensing data has potential to generate landscape scale estimates of tree growth because it captures information about canopy traits, such as foliar nitrogen content, which are correlated with tree growth. In previous work, we developed an empirical relationship between plot-level tree growth and canopy hyperspectral reflectance for even-aged reforestation plots in a tropical dry forest of Panama. This work suggested a suite of hyperspectral canopy reflectance metrics that capture tree growth across multiple species. While these spectral regions been uncovered, the mechanism for this relationship has not yet been directly tested. Our objective is to understand the trait-based mechanisms underlying the relationship between tree growth and canopy reflectance. To address this question, we calculated indices from leaf spectra shown to relate to important leaf traits, as well as direct chemical measurements of some leaf traits, and assessed their correlation with plot-level tree growth.
Results/Conclusions
Foliar traits, as measured by trait indices from hyperspectral data, were correlated with tree growth, but relationships were not consistent across all species. All species showed negative relationships between a chlorophyll index and tree growth (R-squared from -0.31 to -0.69). Additionally, three out of five species have a negative correlation between growth and foliar water content (R-squared -0.7 to -0.54). While seemingly counterintuitive that plots composed of individuals with higher growth rates have lower chlorophyll and higher water, this pattern may reflect that individuals with higher growth senescence earlier in the dry season rather than maintain photosynthetic tissue under conditions of high water limitation. While correlations were found between spectral indices and growth, for most species our results suggest that leaf traits alone may not be the dominant driver of tree growth. Crown-level properties, such as the number of leaves and crown structure, in addition to leaf structure, such as specific leaf area, may also be important traits in explaining the relationship between hyperspectral canopy reflectance and tree growth.