Evaluation of VEGETATION and PROBA-V phenology using phenocam and eddy covariance data

High-quality retrieval of land surface phenology (LSP) is increasingly important for understanding the effects of climate change on ecosystem function and biosphere-atmosphere interactions. We analyzed four state-of-the-art phenology methods: threshold, logistic-function, moving-average and first derivative based approaches, and retrieved LSP in the North Hemisphere for the period 1999-2017 from Copernicus Global Land Service (CGLS) SPOT-VEGETATION and PROBA-V leaf area index (LAI) 1 km V2.0 time series. We validated the LSP estimates with near-surface PhenoCam and eddy covariance FLUXNET data over 80 sites of deciduous forests. Results showed a strong correlation (R² > 0.7) between the satellite LSP and ground-based observations from both PhenoCam and FLUXNET for the timing of the start (SoS) and R² > 0.5 for the end of season (EoS). The threshold-based method performed the best with a root mean square error of ~9 d with PhenoCam and ~7 d with FLUXNET for the timing of SoS (30th percentile of the annual amplitude), and ~12 d and ~10 d, respectively, for the timing of EoS (40th percentile).