Young et al. (2025) Tracking vegetation phenology across diverse biomes using Version 3.0 of the PhenoCam Dataset

Identification

• Journal: Earth system science data
• Year: 2025
• Date: 2025-11-26
• Authors: Adam M. Young, Thomas Milliman, Koen Hufkens, Keith L. Ballou, Christopher Coffey, Kai Begay, Michael Fell, Mostafa Javadian, Alison K. Post, Christina Schädel, Zakary Vladich, Oscar Zimmerman, Dawn M. Browning, Christopher R. Florian, Minkyu Moon, Michael D. SanClements, Bijan Seyednasrollah, Mark A. Friedl, Andrew D. Richardson
• DOI: 10.5194/essd-17-6531-2025

Research Groups

• National Ecological Observatory Network, Battelle, Boulder, CO, USA
• Center for Ecosystem Science and Society, Northern Arizona University, Flagstaff, AZ, USA
• Earth Systems Research Center, University of New Hampshire, Durham, NH, USA
• BlueGreen Labs (BV), Melsele, Belgium
• Information Technology Services, Northern Arizona University, Flagstaff, AZ, USA
• School of Informatics, Computing, and Cyber Systems, Northern Arizona University, Flagstaff, AZ, USA
• Cooperative Institute for Research in Environmental Sciences, Earth Lab, University of Colorado Boulder, Boulder, CO, USA
• Woodwell Climate Research Center, Falmouth, MA, USA
• USDA ARS, Jornada Experimental Range, Las Cruces, NM, USA
• Department of Environmental Science, Kangwon National University, Chuncheon, South Korea
• Department of Earth and Environment, Boston University, Boston, MA, USA

Short Summary

This paper introduces PhenoCam Dataset Version 3.0, a significantly expanded and updated dataset for tracking vegetation phenology across diverse biomes, which now includes a camera-derived Normalized Difference Vegetation Index (cameraNDVI) and simplified data products. It demonstrates that while cameraNDVI offers a complementary measure of canopy structure, the Green Chromatic Coordinate (GCC) generally provides a less noisy signal for phenological tracking.

Objective

• To present PhenoCam Dataset Version 3.0, detailing its significant updates in spatial and temporal coverage, and evaluating its representation across ecoregions and biomes.
• To introduce and evaluate a new PhenoCam-based estimate of the Normalized Difference Vegetation Index (cameraNDVI) and a reduced set of simplified data products.

Study Configuration

• Spatial Scale: 738 unique sites across North America (Level II Ecoregions) and globally (Whittaker Biome Classification), covering diverse biomes including evergreen broadleaf forests, understory vegetation, grasslands, wetlands, and agricultural systems.
• Temporal Scale: 4805.5 site-years of data, spanning from 2000 to 2023, with some individual time series exceeding 16 years. Images are acquired every 15-30 minutes, summarized to 1-day and 3-day time steps.

Methodology and Data

• Models used:
  • Green Chromatic Coordinate (GCC) for canopy greenness.
  • Camera Normalized Difference Vegetation Index (cameraNDVI) calculated from exposure-adjusted pixel intensities of back-to-back visible (RGB) and visible+near-infrared (RGB+NIR) images.
  • Broadband Normalized Difference Vegetation Index (broadbandNDVI) derived from radiometric sensors measuring photosynthetically active radiation and global radiation.
  • Smoothing spline approach for signal characterization and phenological transition date estimation, with optimal span determined by minimizing the Bayesian Information Criterion (BIC).
• Data sources:
  • PhenoCam Network digital cameras (StarDot NetCam SC and StarDot NetCam Live 2) capturing RGB and RGB+NIR images.
  • National Ecological Observatory Network (NEON) flux tower data (Hukseflux NR01 net radiometer, Kipp & Zonen PAR Quantum Sensor) for broadbandNDVI comparison.
  • WorldClim 30-year average temperature and precipitation data for biome classification.
  • USA Environmental Protection Agency Level II Ecoregion classification of North America.
  • AmeriFlux data portal for NEON PAR and radiation data.
  • Oak Ridge National Lab Distributed Active Archive Center (ORNL DAAC) for archived PhenoCam imagery and derived data.

Main Results

• PhenoCam V3.0 includes 738 unique sites and 4805.5 site-years, representing a 170% increase from V2.0, with notable expansion in evergreen broadleaf forests, understory vegetation, grasslands, wetlands, and agricultural systems.
• The network's ecoclimatic representation is strong in boreal forest, temperate forest, temperate grassland desert, temperate rain forest, tropical forest savanna, and woodland/shrubland biomes, but under-represented in high Arctic, most of Mexico, subtropical desert, tundra, and tropical rain forest biomes.
• CameraNDVI, a new product, is now included for approximately 75% of cameras, showing similar seasonal patterns to GCC across diverse ecosystems.
• GCC time series generally exhibit less variability and fewer outliers, providing a smoother signal of canopy greenness and phenology compared to cameraNDVI (GCC Signal-to-Noise Ratio was higher than cameraNDVI in over 80% of cases).
• CameraNDVI lacks the distinct early spring "spike" seen in GCC for deciduous broadleaf sites and shows a delayed fall senescence, potentially better representing seasonal Leaf Area Index (LAI) dynamics.
• CameraNDVI is less sensitive to snow cover and large outliers compared to tower-derived broadbandNDVI, showing strong seasonal correspondence and often a cleaner signal.
• Two new simplified data products (daily mean GCC and smoothed GCC, and simplified transition dates) are introduced for easier user access.
• Consistency between V3.0 and V2.0 transition dates is high, with r² values greater than 98% and mean absolute errors (MAE) less than 2.0 days.

Contributions

• Significant expansion of the PhenoCam dataset (V3.0) in terms of total site-years (170% increase) and coverage of previously under-represented plant functional types (e.g., understory, evergreen broadleaf forests, grasslands, wetlands, agriculture).
• Introduction of cameraNDVI, a novel PhenoCam-derived vegetation index, offering a complementary measure of canopy structure and a more direct comparison to satellite/flux-tower NDVI than GCC.
• Detailed evaluation and comparison of cameraNDVI with GCC and tower-based broadbandNDVI, highlighting their respective strengths and weaknesses for phenological tracking.
• Release of simplified data products to enhance accessibility and usability for educational and general scientific applications.
• Continued provision of a rigorously quality-controlled, multi-year, multi-biome dataset for evaluating satellite products, calibrating phenological models, and understanding ecosystem processes.

Funding

• National Science Foundation
• Long-Term Agroecosystem Research (LTAR) network (supported by the United States Department of Agriculture)
• U.S. Department of Energy
• U.S. Geological Survey
• Northeastern States Research Cooperative
• USA National Phenology Network
• NEON Program (National Science Foundation)
• AmeriFlux data portal (U.S. Department of Energy Office of Science)

Citation

@article{Young2025Tracking,
  author = {Young, Adam M. and Milliman, Thomas and Hufkens, Koen and Ballou, Keith L. and Coffey, Christopher and Begay, Kai and Fell, Michael and Javadian, Mostafa and Post, Alison K. and Schädel, Christina and Vladich, Zakary and Zimmerman, Oscar and Browning, Dawn M. and Florian, Christopher R. and Moon, Minkyu and SanClements, Michael D. and Seyednasrollah, Bijan and Friedl, Mark A. and Richardson, Andrew D.},
  title = {Tracking vegetation phenology across diverse biomes using Version 3.0 of the PhenoCam Dataset},
  journal = {Earth system science data},
  year = {2025},
  doi = {10.5194/essd-17-6531-2025},
  url = {https://doi.org/10.5194/essd-17-6531-2025}
}