Bridging the organism and landscape scales of deciduous forest phenology using an unmanned aerial vehicle, PhenoCam, and remote sensing

Background/Question/Methods

Vegetation phenology arises from organismic processes, but is often observed at the ecosystem level, in measurements such as near-surface and satellite remote sensing. Recent developments in ecological research methods offer the possibility of bridging this gap in scale. This study uses a UAV (Unmanned Aerial Vehicle) to characterize vegetation phenology of the land surface in a 250 meter MODIS (MODerate resolution Imaging Spectroradiometer) pixel of mixed deciduous-evergreen forest at Harvard Forest. We present a synthesis of UAV imagery with species inventories, as well as near-surface and satellite remote sensing, to answer the question: how do species composition and land cover affect ecosystem phenology in a mixed deciduous-evergreen forest?

We flew the UAV with an RGB digital camera throughout the 2013 growing season, from before budburst to after abscission. Digital images were combined into mosaics and aligned to create a georeferenced series. We divided the mosaics into a 10 meter grid and classified grid cells according to land cover type, as well as species composition, using a detailed map of tree species in the study area. We then derived time series of vegetation indices and phenophase transition dates for spring and autumn for each grid cell.

Results/Conclusions

Results indicate that phenology determined from RGB aerial photography matches near-surface PhenoCam observations within an average of three days for key phenophase transition dates in spring and autumn, using the GCC (green chromatic coordinate) metric of canopy activity, and remote sensing observations of EVI (enhanced vegetation index) from the MODIS satellites within an average of two days. Aerial photography reveals gradients of two to four weeks in spring phenophase transitions of the deciduous land cover type, and five weeks in autumn transitions, which are found to correlate with the spatial distribution of dominant canopy species. We conclude that aerial photography with UAVs is a novel method for characterizing ecosystem phenology across scales from organisms to ecosystems. Georeferenced aerial photography can be analyzed using standard digital image processing techniques and readily synthesized with complementary spatial data, such as species maps, to explain patterns in phenology transitions.