Petropoulos et al. (2026) Expanding the Ts–VI feature space for retrieving new parameters characterising the water and carbon cycle: proof-of-concept of a new methodological framework and its validation at selected FLUXNET sites

Identification

• Journal: The Science of The Total Environment
• Year: 2026
• Date: 2026-01-23
• Authors: George P. Petropoulos, Spyridon E. Detsikas, Christina Lekka
• DOI: 10.1016/j.scitotenv.2026.181377

Research Groups

• Department of Geography, Harokopio University of Athens, Athens, Greece

Short Summary

This study introduces and validates a novel methodological framework that extends the "analytical triangle" method, integrating Landsat-8 Earth Observation data with the SimSphere SVAT model, to retrieve instantaneous maps of energy, water, carbon, and ozone cycle components. The proof-of-concept was evaluated at three FLUXNET sites, demonstrating satisfactory accuracy for parameters like Gross Primary Productivity (GPP) and Water Use Efficiency (WUE), with the most accurate predictions observed at cropland sites.

Objective

• To introduce and provide the first validation of a novel methodological framework that extends the "analytical triangle" method, using Landsat-8 data and the latest SimSphere SVAT model, for deriving instantaneous maps of mass, energy, carbon, and ozone flux parameters directly from the Land Surface Temperature (Ts) – Vegetation Index (VI) feature space.
• To retrieve new variables characterizing land surface interactions, specifically Gross Primary Productivity (GPP), ozone flux (O₃ flux), water use efficiency (WUE), carbon dioxide concentration ([CO₂]), and ozone concentration ([O₃]), in addition to established variables like latent heat flux (LE), sensible heat flux (H), and surface soil moisture (SSM).

Study Configuration

• Spatial Scale: Three FLUXNET sites in Europe (Ispra ABC-IS, Castelporzianob2 in Italy; Grignon in France), representing deciduous broadleaf forest, evergreen broadleaved forest, and cropland biomes. Earth Observation data were processed at a 100 meter spatial resolution.
• Temporal Scale: A total of 30 selected days across the years 2004–2017, focusing on spring and summer months. In-situ FLUXNET data were recorded at half-hourly intervals, while satellite data provided instantaneous measurements at overpass times.

Methodology and Data

• Models used:
  • "Analytical triangle" method (a Ts-VI feature space interpretation framework).
  • SimSphere Soil Vegetation Atmosphere Transfer (SVAT) model (1-dimensional).
• Data sources:
  • Satellite: Landsat 5 and Landsat 8 (Level 2) Earth Observation datasets, accessed via Google Earth Engine (GEE), filtered for less than 20% cloud cover.
  • Observation (in-situ):
    • FLUXNET network (including Carbo Europe and ICOS regional networks) Eddy Covariance (EC) tower measurements for LE, H, SSM, GPP, O₃ flux, and [O₃].
    • Meteorological variables (air temperature, precipitation rates, soil water content at different depths) from FLUXNET.
    • Radiosonde observations from the University of Wyoming's Atmospheric Science Weather Balloon Data Archive for atmospheric forcing (wind direction/speed, temperature, dew point temperature, pressure).

Main Results

• The proposed methodological framework satisfactorily captured the spatial variability of predicted variables, showing reasonable and consistent patterns related to topography and land cover.
• Overall, the analysis yielded satisfactory Root Mean Square Deviations (RMSDs) and correlation coefficients (R) across the 30 validation days:
  • Gross Primary Productivity (GPP): RMSD of 6.268 μmol CO₂ m⁻² s⁻¹ (average R = 0.93). Best performance was observed at cropland sites.
  • Water Use Efficiency (WUE): RMSD of 8.116 kg m⁻³ (average R = 0.51). Lowest RMSD (1.75 kg m⁻³) at the cropland site (FR-Gri).
  • Ozone concentration ([O₃]): RMSD of 8.270 ppb (average R = 0.82). Consistently underestimated across sites.
  • Ozone flux (O₃ flux): RMSD of 5.429 μmol mol⁻¹ (average R = 0.69). Consistently underestimated across sites.
  • Carbon dioxide concentration ([CO₂]): RMSD of 59.910 μmol mol⁻¹ (average R = 0.62). Excluding post-harvest periods significantly reduced RMSD to 2.23 μmol mol⁻¹.
  • Latent Heat Flux (LE): RMSD of 48.902 W m⁻² (average R = 0.81).
  • Sensible Heat Flux (H): RMSD of 66.565 W m⁻² (average R = 0.82).
  • Surface Soil Moisture (SSM): RMSD of 9.662 m³/m³ (average R = 0.41).
• Cropland sites (FR-Gri) generally showed the most accurate predictions and lowest average errors for most variables, compared to forested ecosystems.

Contributions

• Provides the first validation of the "analytical triangle" method using high-resolution Landsat-8 data and the latest version of the SimSphere SVAT model.
• Introduces a novel methodological framework that significantly extends the application of the "analytical triangle" to retrieve a broader range of land-atmosphere interaction parameters, including GPP, [CO₂], WUE, [O₃], and O₃ flux, directly from the Ts-VI feature space.
• Demonstrates the potential for regional-scale GPP modeling without requiring extensive model calibration.
• Highlights the advantage of integrating Earth Observation's spectral and spatial resolution with the vertical and temporal continuity provided by SVAT models for more accurate parameter estimation.
• Offers a rapid and low-cost method for monitoring key land-atmosphere interaction parameters.

Funding

• LISTEN-EO project (H.F.R.I. Project Number: 15898), funded by the European Union–Next Generation EU under the National Recovery and Resilience Plan “Greece 2.0” (H.F.R.I call “Basic research Financing (Horizontal support of all Sciences)”).

Citation

@article{Petropoulos2026Expanding,
  author = {Petropoulos, George P. and Detsikas, Spyridon E. and Lekka, Christina},
  title = {Expanding the Ts–VI feature space for retrieving new parameters characterising the water and carbon cycle: proof-of-concept of a new methodological framework and its validation at selected FLUXNET sites},
  journal = {The Science of The Total Environment},
  year = {2026},
  doi = {10.1016/j.scitotenv.2026.181377},
  url = {https://doi.org/10.1016/j.scitotenv.2026.181377}
}