Drivers of spatial variability in greendown within an oak-hickory forest landscape

Abstract

Declining near-infrared (NIR) surface reflectance between early and late summer, here termed greendown, is a common, yet poorly understood phenomena in remote sensing time series of temperate deciduous forests. As revealed by phenology analysis of Landsat satellite data, there are strong spatial patterns in the rate of greendown across temperate deciduous forest landscapes, and analyzing these patterns could help advance our understanding of surface reflectance drivers. Within an oak-hickory (Quercus spp. – Carya spp.) forest landscape in western Maryland, USA, we tested how spatial patterns in greendown related to potential drivers at the landscape-, tree crown- and leaf-levels. We found that 50% of the spatial variability in greendown was explained by landscape variables, with greendown particularly higher in locations with higher maximum greenness, more southerly aspects, or locations with greater abundance of white oak (Quercus alba). The importance of species composition as a driver of greendown was supported at the tree crown level, where, relative to three other tree species, white oak exhibited the most consistent trend toward more vertical leaf angles later in the season. At the leaf level, NIR reflectance decreased in productive sites where %N increased, and δ13C decreased, through the season. However, among all sites, there were no consistent seasonal trends in foliar NIR reflectance, and we found no correlation among leaf-level NIR reflectance and satellite-observed greendown. Collectively, these results suggest that the spatial variability of greendown in this oak-hickory forest is most strongly controlled by an interaction of topographic and species compositional drivers operating at the landscape and tree crown levels. We found spatial analysis of greendown to be a useful approach to explore landscape-, tree crown-, and leaf-level controls on surface reflectance, and thereby help translate land surface phenology data into predictions of ecosystem structure and functioning.