Pénot et al. (2025) Combining Landsat optical/thermal and LiDAR High Definition data to estimate turbulent fluxes over Mediterranean forests

Identification

• Journal: Science of Remote Sensing
• Year: 2025
• Date: 2025-11-01
• Authors: Victor Pénot, Olivier Merlin
• DOI: 10.1016/j.srs.2025.100323

Research Groups

• Victor Penot, Olivier Merlin: Université de Toulouse, CNES/CNRS/IRD/INRAe, CESBio, Toulouse, France

Short Summary

This study addresses the challenge of estimating sensible (H) and latent (LE) heat fluxes over semi-arid forests by integrating LiDAR-derived canopy height (hc) into a classical thermal-based contextual method. By normalizing Landsat Land Surface Temperature (LST) for hc effects and constraining the dry edge with an energy balance model, the proposed approach significantly improves the accuracy and consistency of turbulent flux estimates across Mediterranean forest sites.

Objective

• To address the limitation of thermal-based models in estimating turbulent heat fluxes over semi-arid forests by accounting for canopy height (hc) effects on satellite Land Surface Temperature (LST).
• To quantitatively assess the effect of hc on LST over extended areas.
• To develop a normalization method for Landsat LST to account for hc effects.
• To evaluate the Landsat-derived H and LE estimates using the new (LSTnorm-fgv) and classical (LST-fgv) contextual methods.

Study Configuration

• Spatial Scale: Two Mediterranean forest sites in southern France: Puechabon (6400 hectares) and Fontblanche (5164 hectares). Landsat data were aggregated to a 90 m resolution. LiDAR data were rasterized to a 90 m resolution.
• Temporal Scale: A nine-year period, focusing on the hottest months (April to October). Puechabon data covered 2014 to August 2022 (96 overpass dates), and Fontblanche data covered 2013 to 2021 (110 overpass dates).

Methodology and Data

• Models used:
  • Classical contextual method (LST-fgv)
  • New contextual method with LST normalized for hc effect (LSTnorm-fgv(context))
  • New contextual method with LST normalized for hc effect and an Energy Balance (EB) model for dry edge estimation (LSTnorm-fgv(EB))
  • Simple soil energy balance (EB) model (detailed in Appendix A)
• Data sources:
  • Satellite: Landsat-7 Enhanced Thematic Mapper (ETM+), Landsat-8/9 Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) (Collection 2 Level-2 Science Products). Derived products include Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI), fractional green vegetation cover (fgv), and albedo (α).
  • Airborne: High-resolution airborne laser scanning (ALS) LiDAR data (IGN campaign, June 2021) for canopy height (hc).
  • In-situ (Eddy Covariance stations): Continuous measurements of meteorological data (relative humidity (RH), air temperature (Tair), wind speed (WS), solar radiation (Rg)) and surface energy fluxes (latent heat flux (LE), sensible heat flux (H), net radiation (Rn)) at Puechabon and Fontblanche.
  • Reanalysis: SAFRAN reanalysis dataset (hourly meteorological data on an 8-km grid for RH, Tair, and WS).

Main Results

• Canopy Height Effect on LST: Land Surface Temperature (LST) consistently decreases with increasing canopy height (hc). The slope of the linear upper edge in the LST-hc space was estimated at approximately -0.53 ± 0.14 K m⁻¹.
• Threshold Canopy Height (hcmax): The threshold canopy height (hcmax), where LST becomes insensitive to turbulent fluxes (LST ≈ Tair), was estimated as 42 ± 4 m for Puechabon and 35 ± 3 m for Fontblanche. The combined All-Site hcmax was 39 m.
• LST Normalization (LSTnorm-fgv(context)): Normalizing LST for hc effects significantly improved the correlation between remotely sensed and in situ H (from 0.40 to 0.72) and LE (from 0.06 to 0.43) when both sites were considered together. It also reduced inter-site discrepancies in error metrics. However, H remained underestimated (bias = -93 W m⁻², slope = 0.61) and LE overestimated (bias = 62 W m⁻², slope = 0.32).
• EB-modeled Dry Edge (LSTnorm-fgv(EB)): Integrating an Energy Balance (EB) model to constrain the dry edge further reduced biases and increased sensitivity.
  • Bias for H was reduced from -163 W m⁻² (classical method) to -56 W m⁻².
  • Bias for LE was reduced from +132 W m⁻² (classical method) to +25 W m⁻².
  • The slope of the linear regression for H increased from 0.28 to 0.84, and for LE from 0.07 to 0.60, indicating a more realistic representation of flux dynamics.
  • Correlation for H was 0.72 and for LE was 0.46.
• Sensitivity Analysis: The LSTnorm-fgv(EB) approach demonstrated robustness to variations in hcmax and, to a lesser extent, to meteorological inputs from the SAFRAN reanalysis dataset, still yielding improved results compared to classical methods despite some persistent biases.

Contributions

• This is the first study to successfully incorporate LiDAR-derived canopy height (hc) into contextual methods for estimating turbulent heat fluxes, addressing a critical gap in thermal-based modeling over forested ecosystems.
• A novel method was developed to normalize Landsat LST for hc effects, effectively circumventing the need for complex and often non-unique parameterizations of surface roughness in traditional Energy Balance (EB) models.
• The study demonstrates that contextual methods, previously limited to low-vegetation areas, can be effectively applied to forested landscapes with significant canopy height heterogeneity.
• The integration of a simple EB model to constrain the dry edge of the LSTnorm-fgv feature space significantly enhances the accuracy and consistency of flux estimates across sites with contrasting canopy structures.
• The proposed approach offers a robust, low-parameter, and operationally efficient alternative for turbulent flux estimation in complex forest ecosystems, outperforming existing methods in accuracy and consistency at the studied Mediterranean sites.

Funding

Not explicitly stated in the provided text.

Citation

@article{Pénot2025Combining,
  author = {Pénot, Victor and Merlin, Olivier},
  title = {Combining Landsat optical/thermal and LiDAR High Definition data to estimate turbulent fluxes over Mediterranean forests},
  journal = {Science of Remote Sensing},
  year = {2025},
  doi = {10.1016/j.srs.2025.100323},
  url = {https://doi.org/10.1016/j.srs.2025.100323}
}