Calibrating vegetation phenology from Sentinel-2 using eddy covariance, PhenoCam, and PEP725 networks across Europe

Vegetation phenology obtained from time series of remote sensing data is relevant for a range ecological applications. The freely available Sentinel-2 imagery at 10 m spatial resolution with ~ 5-day repeat cycle provides an opportunity to map vegetation unprecedented fine scale. To facilitate the production Europe-wide Copernicus Land Monitoring based dataset, we design and evaluate framework on comprehensive set ground observations, including eddy covariance gross primary (GPP), PhenoCam green chromatic coordinate (GCC), phases Pan-European Phenological database (PEP725). We test three indices (VI) — normalized difference index (NDVI), two-band enhanced (EVI2), plant (PPI) regarding their capability track seasonal trajectories GPP GCC performance in reflecting variabilities corresponding phenometrics, i.e., start season (SOS) end (EOS). find that phenology, PPI performs best, particular evergreen coniferous forest areas where variations leaf area are small snow prevalent during wintertime. Results inconclusive which no consistently better than others. When comparing PEP725 phases, EVI2 perform NDVI correlation consistency (i.e., lower standard deviation). also link VI phenometrics various amplitude thresholds phenophases SOS 25% EOS 15% provide best matches ground-observed phenological stages. Finally, demonstrate applying bidirectional reflectance distribution function correction step can be excluded mapping Europe.