Fibbi et al. (2025) Assessment of three remote sensing methods for estimating actual evapotranspiration in a Mediterranean region

Identification

• Journal: International Journal of Applied Earth Observation and Geoinformation
• Year: 2025
• Date: 2025-12-04
• Authors: Luca Fibbi, Marta Chiesi, Maurizio Pieri, G. Bartolini, Daniele Grifoni, Bernardo Gozzini, Fabio Maselli
• DOI: 10.1016/j.jag.2025.105003

Research Groups

• CNR IBE (Institute of BioEconomy, National Research Council), Sesto Fiorentino, Italy
• LaMMA Consortium, Sesto Fiorentino, Italy

Short Summary

This study assesses three remote sensing methods (NDVI-Cws, MODIS, LSA SAF) for estimating actual evapotranspiration (ETa) in Tuscany, Italy, over 20 years (2005-2024) using a triple collocation approach. It finds that LSA SAF and NDVI-Cws estimates show strong spatial and temporal concordance and indicate widespread increasing ETa trends, while MODIS estimates are less concordant, especially for forests, and poorly reflect these trends.

Objective

• To inter-compare the regional spatial and temporal actual evapotranspiration (ETa) variations estimated by three remote sensing methods (MODIS, LSA SAF, NDVI-Cws).
• To identify the advantages and drawbacks of the three methods to select the optimum one for maximum spatial and temporal resolution applicability.
• To characterize the spatial and temporal regional ETa trends in relation to existing environmental conditions.

Study Configuration

• Spatial Scale: Tuscany region, Central Italy (approximately 42-45° N latitude, 8°-13° E longitude). Analysis conducted over 12 representative vegetated areas and at a per-pixel level (250 m resolution for NDVI-Cws, 500 m for MODIS, 5 km for LSA SAF).
• Temporal Scale: 20-year period (2005–2024). Daily time step for NDVI-Cws and LSA SAF products, 8-day for MODIS products, with annual cumulations for analysis.

Methodology and Data

• Models used:
  • NDVI-Cws: Semi-empirical method combining conventional weather data and satellite NDVI images to simulate water stress effects on ETa. Modified versions for wetlands and irrigated croplands were also applied.
  • MOD16A2GF: Simplified Penman-Monteith approach.
  • LSA SAF (DMETv3 LSA-312.3 and METREF LSA-303): Physically based and simplified Soil-Vegetation-Atmosphere-Transfer algorithms.
  • DAYMET algorithm: For interpolating ground meteorological observations to 250 m spatial resolution.
  • ERAD algorithm: For estimating daily solar radiation from MSG satellite data.
• Data sources:
  • Meteorological and ancillary data: Daily minimum/maximum air temperature and rainfall (2005–2024) from the LaMMA database (Regional Hydrologic Service of Tuscany, National Weather Service). Daily solar radiation (2005–2024) from ERAD algorithm. Land cover map from CORINE Land Cover project (regrouped into five biome types). Wetlands distribution from CORINE and Ramsar datasets.
  • MODIS products: MODIS NDVI product (MOD13Q1 v061) at 250 m spatial resolution (16-day maximum value composite, linearly interpolated to daily). MODIS ET product (MOD16A2GF) at 500 m spatial resolution and 8-day temporal resolution.
  • LSA SAF products: Reference ET (METREF LSA-303) and actual ET (DMETv3 LSA-312.3) products at 5 km spatial resolution and daily temporal resolution, derived from the SEVIRI sensor onboard the MSG geostationary satellite. Inputs include MSG-SEVIRI derived LAI, albedo, FVC, surface fluxes, snow cover, ECOCLIMAP land cover, and ERA-Interim reanalysis weather data.
  • Statistical methods: Triple collocation strategy, Reduced Major Axis (RMA) regression, Ordinary Least Square (OLS) regression (for time trend analysis), Pearson’s coefficient of correlation (r), Root Mean Square Difference (RMSD).

Main Results

• Spatial and Temporal ETa Variations: Clear spatial and temporal ETa variations were observed across Tuscany, but the three products represented these variations differently.
• MODIS ETa Performance:
  • MODIS ETa estimates were significantly higher for forest areas and lower for non-forest areas (croplands, wetlands) compared to LSA SAF and NDVI-Cws.
  • ETa trends estimated by MODIS were poorly concordant with those of the other two products, particularly for forests.
  • MODIS ETa estimates only marginally reflected the widespread ETa increases evidenced by LSA SAF and NDVI-Cws.
  • Per-pixel trends from MODIS were moderate (forest: 1.40 mm/year, non-forest: 2.10 mm/year), significantly lower than the other products.
• LSA SAF and NDVI-Cws ETa Performance:
  • LSA SAF and NDVI-Cws ETa estimates showed high spatial (r = 0.916) and temporal (r = 0.852) concordance.
  • Both methods indicated widespread increasing ETa trends across the study region, with mean increases close to 4 mm/year (forest: 3.56-4.55 mm/year, non-forest: 4.27-5.07 mm/year), in response to rising air temperature, solar radiation, and precipitation.
  • Exceptions to increasing trends were noted in some southern coastal areas, where ETa decreased due to local rainfall drops and disturbances.
• Spatial Resolution: NDVI-Cws provided a more spatially detailed prediction of local ETa variability (250 m) compared to MODIS (500 m) and LSA SAF (5 km), which tended to smooth variability in heterogeneous areas.
• Meteorological Trends (2005-2024): Mean air temperature significantly increased by 0.069 °C/year. Solar radiation showed a non-significant positive trend of 0.037 MJ/(m²·year). Precipitation showed a non-significant positive trend of 10.81 mm/year (sample areas) or 5.8 mm/year (pixel level).

Contributions

• Provides a spatially distributed assessment of three remote sensing ETa products (MODIS, LSA SAF, NDVI-Cws) over a specific, environmentally heterogeneous Mediterranean region (Tuscany) for a 20-year period, addressing the limitations of sparse ground observations.
• Successfully applies the triple collocation approach to evaluate ETa products with diverse conceptual bases, input datasets, and spatial resolutions.
• Highlights the effectiveness of the semi-empirical NDVI-Cws method for detailed ETa monitoring in Mediterranean, water-limited regions, demonstrating its capability to capture local variability at high spatial resolution.
• Offers evidence of widespread increasing ETa trends in Tuscany over the last two decades, linking them to climate change drivers and identifying areas where MODIS performance is suboptimal, particularly in water-limited forest ecosystems.
• Demonstrates the potential for generating high spatial and temporal resolution ETa datasets using methods like NDVI-Cws, thereby expanding their utility for various operational applications.

Funding

• Funding acquisition by Bernardo Gozzini. No specific project names or reference codes were listed.

Citation

@article{Fibbi2025Assessment,
  author = {Fibbi, Luca and Chiesi, Marta and Pieri, Maurizio and Bartolini, G. and Grifoni, Daniele and Gozzini, Bernardo and Maselli, Fabio},
  title = {Assessment of three remote sensing methods for estimating actual evapotranspiration in a Mediterranean region},
  journal = {International Journal of Applied Earth Observation and Geoinformation},
  year = {2025},
  doi = {10.1016/j.jag.2025.105003},
  url = {https://doi.org/10.1016/j.jag.2025.105003}
}