Paramanik et al. (2025) Automated and continuous estimation of FAPAR from distributed wireless PAR sensor networks

Identification

• Journal: Agricultural and Forest Meteorology
• Year: 2025
• Date: 2025-11-07
• Authors: Somnath Paramanik, Harry Morris, Rémi Grousset, Gabriele Bai, Christophe Lerebourg, Ernesto López-Baeza, Ana Pérez-Hoyos, David García-Rodríguez, Darius Culvenor, Alexander Knohl, Anne Klosterhalfen, Frank Tiedemann, Christian Lanconelli, M. Clerici, Nadine Gobron, Luke A. Brown, Finn James, Stefan Maier, Fabrizio Niro, Jadunandan Dash
• DOI: 10.1016/j.agrformet.2025.110904

Research Groups

• School of Geography and Environmental Science, University of Southampton, United Kingdom
• Climate and Earth Observation Group, National Physical Laboratory, United Kingdom
• ACRI-ST, Sophia-Antipolis, France
• Albavalor, Science Park Universitat de Valencia, Spain
• University Research Institute on Robotics and Information and Communication Technologies (IRTIC), Universitat de Valencia, Spain
• Environmental Sensing Systems, Bentleigh East, Victoria, Australia
• University of Göttingen, Bioclimatology, Germany
• UniSystems, Milan, Italy
• European Commission, Joint Research Centre (JRC), Ispra, Italy
• School of Science, Engineering & Environment, University of Salford, United Kingdom
• maitec, Darwin, Australia
• Serco for European Space Agency (ESA), Frascati, Italy

Short Summary

This study evaluates the performance of two-flux (2f) and four-flux (4f) FAPAR measurement systems and digital hemispherical photography (DHP) across multiple vegetation types and temporal scales using automated wireless PAR sensor networks. It reveals strong agreement between 2f- and 4f-FAPAR (R² > 0.99, RMSE ≤ 0.04), suggesting that 2f systems are a reliable and cost-effective alternative, and underscores the importance of daily integrated FAPAR for long-term ecosystem monitoring.

Objective

• To evaluate the consistency between 2f- and 4f-FAPAR measurements across three distinct ecosystems (vineyard, deciduous forest, savanna woodland).
• To assess the impact of quality control (QC) procedures on the recorded PAR dataset.
• To compare daily integrated (sunrise to sunset) and instantaneous (satellite overpass) FAPAR estimates.
• To intercompare FAPAR measured by wireless sensor networks (WSNs) with DHP-derived FAPAR.

Study Configuration

• Spatial Scale: Elementary Sampling Units (ESU) of approximately 60 m × 60 m (Valencia Anchor Station - VAS) and 50 m × 50 m (Hainich National Park - HNP, Litchfield Savanna SuperSite - LSS). PAR sensors were deployed at various heights (e.g., 0.2 m, 1.8 m, 31 m, 43 m) and DHP systems at 1.3 m above ground.
• Temporal Scale: PAR data measured every 5 minutes. FAPAR estimated as instantaneous (10:00 local solar time ± 15 min) and daily integrated (sunrise to sunset). Data collected from February 2021 to November 2024 (VAS), April 2023 to November 2024 (HNP), and December 2021 to June 2024 (LSS).

Methodology and Data

• Models used:
  • Two-flux FAPAR (2f-FAPAR)
  • Four-flux FAPAR (4f-FAPAR)
  • Digital Hemispherical Photography (DHP) for FAPAR estimation
  • HemiPy Python module for DHP image processing and FAPAR derivation.
• Data sources:
  • Distributed wireless PAR sensor networks using Apogee SQ-110-SS quantum sensors.
  • Digital Hemispherical Photography (DHP) using Canon EOS 80D/1300D cameras with Sigma 4.5 mm F2.8 EX DC fisheye lenses (manual campaigns at VAS, automated system at HNP).
  • Data obtained from the Ground-Based Observations for Validation (GBOV) initiative.

Main Results

• 2f- vs. 4f-FAPAR: Strong agreement was observed across all three study sites (VAS, HNP, LSS) for both instantaneous and daily integrated FAPAR, with R² values consistently greater than 0.99 and RMSE values ≤ 0.04. A minimal positive bias (≤ 0.04) indicated slight overestimation by 2f-FAPAR, suggesting it can reliably substitute the more complex 4f setup.
• Instantaneous vs. Daily Integrated FAPAR: Daily integrated FAPAR consistently exhibited greater stability and lower uncertainty compared to instantaneous FAPAR across all sites. Agreement varied by site, with HNP showing the strongest correlation (R² = 0.97, RMSE = 0.03) and LSS the weakest (R² = 0.34, RMSE = 0.11). Instantaneous FAPAR remains crucial for satellite product validation.
• WSNs-FAPAR vs. DHP-derived FAPAR: Moderate to good agreement was found. At VAS, instantaneous 2f-FAPAR showed R² = 0.39 (RMSE = 0.34), improving to R² = 0.56 (RMSE = 0.15) for daily integrated. At HNP, agreement was better (instantaneous R² = 0.76, RMSE = 0.14; daily integrated R² = 0.74, RMSE = 0.14), with a consistent positive bias (0.12) for 2f-FAPAR.
• Quality Control (QC): Rigorous QC steps (e.g., removal of non-physical values, energy balance, clearness index filtering, solar zenith angle restriction, percentile outlier removal) significantly filtered the raw dataset, ensuring data integrity (e.g., 14.30% data loss in VAS due to energy balance QC, 18.70% in HNP due to cloud filtering).
• GCOS Requirements: The percentage of data meeting GCOS requirements (absolute difference of 0.005 units and relative difference of 10% of reference 4f-FAPAR) varied significantly by site: VAS (47.14%), HNP (99.78%), LSS (92.00%).

Contributions

• Provides a comprehensive evaluation of 2f- and 4f-FAPAR estimation using automated wireless PAR sensor networks across diverse vegetation types (vineyard, deciduous forest, savanna) and temporal scales.
• Demonstrates that the simpler and more cost-effective 2f-FAPAR system can provide comparable accuracy to the more complex 4f-FAPAR system for most ecological and remote sensing applications.
• Highlights the distinct utility of instantaneous FAPAR for satellite product validation and daily integrated FAPAR for long-term ecosystem monitoring and modeling.
• Establishes practical guidelines for selecting appropriate FAPAR estimation schemes based on specific field conditions and research objectives.
• Contributes to the development of robust, automated, and continuous ground-based FAPAR measurement protocols, addressing existing challenges in data consistency and spatial representativeness for satellite product validation.

Funding

• GBOV (Ground Based Observations for Validation) funded by the European Commission Joint Research Centre FWC 932059, part of the Global Component of the European Union’s Copernicus Land Monitoring Service.
• Living Planet Fellowship, a programme of and funded by the European Space Agency.
• zukunft.niedersachsen as part of the project FoResLab (ZN4445).
• Instrument Data Evaluation and Analysis Service (IDEAS+) Quality Assurance for Earth Observation (QA4EO) service, funded through European Space Agency contract number QA4EO/SER/SUB/39.

Citation

@article{Paramanik2025Automated,
  author = {Paramanik, Somnath and Morris, Harry and Grousset, Rémi and Bai, Gabriele and Lerebourg, Christophe and López-Baeza, Ernesto and Pérez-Hoyos, Ana and García-Rodríguez, David and Culvenor, Darius and Knohl, Alexander and Klosterhalfen, Anne and Tiedemann, Frank and Lanconelli, Christian and Clerici, M. and Gobron, Nadine and Brown, Luke A. and James, Finn and Maier, Stefan and Niro, Fabrizio and Dash, Jadunandan},
  title = {Automated and continuous estimation of FAPAR from distributed wireless PAR sensor networks},
  journal = {Agricultural and Forest Meteorology},
  year = {2025},
  doi = {10.1016/j.agrformet.2025.110904},
  url = {https://doi.org/10.1016/j.agrformet.2025.110904}
}