FLEX-based inferring of terrestrial photosynthesis dynamics for quantifying European vegetation productivity

Información del proyecto

FLEXINEL

Identificador del acuerdo de subvención: 101086622

DOI

10.3030/101086622

Fecha de la firma de la CE 4 Abril 2023

Fecha de inicio 1 Septiembre 2023

Fecha de finalización 31 Agosto 2028

Financiado con arreglo a

• European Research Council (ERC)

Coste total

€ 1 999 943,00

Aportación de la UE

€ 1 999 943,00

Coordinado por

UNIVERSITAT DE VALENCIA
Spain

Periodic Reporting for period 1 - FLEXINEL (FLEX-based inferring of terrestrial photosynthesis dynamics for quantifying European vegetation productivity)

Período documentado: 2023-09-01 hasta 2026-02-28

Resumen del contexto y de los objetivos generales del proyecto

Vegetation photosynthesis drives the uptake of atmospheric CO2 and regulates land–atmosphere feedbacks, yet current satellite systems still struggle to capture its rapid dynamics and response to stress. Reflectance-based indicators inform on structure and pigments, but they do not directly quantify photosynthetic function or clearly separate measurement noise from true physiological change.
FLEXINEL (2023–2028) addresses this gap in the context of ESA’s FLuorescence EXplorer (FLEX) mission—Europe’s first satellite dedicated to measuring solar-induced chlorophyll fluorescence (SIF), a direct proxy of photosynthetic activity. FLEX, flying in tandem with Sentinel-3 (S3) as the FLEX–S3 mission, will enable global monitoring of the “breathing” of terrestrial vegetation with unprecedented spectral fidelity.
The overall objective of FLEXINEL is to develop the scientific foundations and computational infrastructure to transform FLEX and FLEX–S3 data into robust, quantitative estimates of photosynthetic function. The project combines plant physiology, canopy radiative-transfer modelling, Bayesian machine learning (ML) and multi-sensor satellite data to deliver a new generation of vegetation products that are physically consistent, uncertainty-aware and scalable to continental and global domains.
FLEXINEL addresses five key needs:
1. By deriving essential vegetation traits (pigments, structural properties, canopy biochemistry) and linking them to photosynthetic efficiency, FLEXINEL aims to quantify how plants optimize photosynthesis under stress conditions.
2. The project develops probabilistic hybrid retrievals that merge radiative-transfer models (RTMs) with Gaussian Process Regression (GPR) and dimensionality reduction, enabling per-pixel vegetation traits with explicit aleatoric and epistemic uncertainty. A flagship result is a PCA-based reconstruction of full-spectrum SIF, allowing FLEX-like SIF synthesis and airborne-to-satellite upscaling for pre-launch validation and calibration.
3. FLEXINEL builds harmonised frameworks that integrate FLEX–S3 with other optical missions, generating continuous L3 time series of vegetation traits. Cloud-native implementations in openEO and GEE scale these workflows to continental domains and prepare them for Copernicus/ESA operations.
4. FLEX-relevant products are fused into ecosystem and carbon-cycle models to better constrain gross primary productivity (GPP), water-use efficiency and related functional traits, strengthening the quantitative link between optical signals and carbon fixation processes.
5. By developing open-source tools (ARTMO, PyEOGPR), standardised workflows and cloud-ready pipelines, FLEXINEL positions Europe to rapidly transform FLEX fluorescence measurements into operational vegetation and carbon products, consolidating European leadership in global vegetation monitoring.

Trabajo realizado desde el comienzo del proyecto hasta el final del período abarcado por el informe y los principales resultados hasta la fecha

During the first reporting period, FLEXINEL established the core scientific, methodological and computational foundations for uncertainty-aware retrievals of vegetation traits and photosynthesis in preparation for the ESA FLEX mission. Work progressed along four axes: (1) retrieval theory and emulation, (2) scalable vegetation-trait mapping, (3) integration with photosynthesis and carbon-flux modelling, and (4) cloud-enabled deployment.
FLEXINEL developed probabilistic hybrid models that couple radiative-transfer simulations with GPR to retrieve pigments, structural and biochemical traits directly from TOA hyperspectral data, with explicit aleatoric and epistemic uncertainty per pixel. A flagship result is a PCA-based full-spectrum SIF reconstruction framework, enabling FLEX-like SIF synthesis, airborne-to-satellite upscaling and pre-launch algorithm benchmarking. Emulator optimisation and analytical uncertainty propagation provide a robust theoretical backbone for FLEXINEL’s retrievals.
Hybrid GPR–RTM approaches were extended across Sentinel-2, S3, EnMAP and UAV data and across diverse biomes, and were implemented in cloud-native pipelines (openEO, Google Earth Engine) to generate continental-scale maps of LAI, Cab and CNC with pixel-wise Bayesian uncertainty. To link EO products with ecosystem functioning, FLEXINEL developed GPR-hybrid GPP upscaling, the D&B v1.0 vegetation model for assimilating SIF and trait indicators, and causal-inference tools to identify environmental drivers and domain shifts.
Mission readiness was ensured through interoperable retrieval libraries (e.g. PyEOGPR, ARTMO integration) and tools for propagation and visualisation of multi-dimensional uncertainty, including per-band radiometric variance, in line with FLEX and CHIME product requirements.
Main achievements:
1) PCA-based full-spectrum SIF reconstruction for FLEX-like pre-launch products.
2) Probabilistic hybrid GPR retrievals with explicit uncertainty decomposition.
3)Continental-scale cloud pipelines for vegetation traits and uncertainty mapping.
4) Integration of optical traits with carbon-flux and ecosystem models.

Avances que van más allá del estado de la técnica e impacto potencial esperado (incluida la repercusión socioeconómica y las implicaciones sociales más amplias del proyecto hasta la fecha)

FLEXINEL established the core scientific, methodological and computational foundations for uncertainty-aware retrievals of vegetation traits and photosynthesis in preparation for the ESA FLEX mission. Work advanced along four axes: (1) retrieval theory and emulation, (2) scalable vegetation-trait mapping, (3) integration with photosynthesis and carbon-flux modelling, and (4) cloud-enabled deployment.
FLEXINEL developed a new generation of probabilistic hybrid models that combine RTM simulations with GPR, providing physically consistent estimates of pigments, structural and biochemical traits from BOA/TOA hyperspectral data with explicit aleatoric and epistemic uncertainty per pixel. A flagship achievement is the PCA-based full-spectrum SIF reconstruction framework, which reconstructs full-spectrum SIF from a small number of components, enabling FLEX-like SIF synthesis, efficient airborne-to-satellite upscaling (HyPlant–FLEX), and pre-launch calibration/validation testbeds.
Hybrid retrievals were extended across sensors (S2, S3, EnMAP, UAV) and biomes (croplands, forests, mangroves, semi-arid and high-mountain ecosystems), and implemented in cloud-native pipelines (openEO, GEE) to produce continental-scale maps of LAI, Cab, CNC and SIF-related indicators with pixel-wise uncertainty. In parallel, FLEXINEL linked these optical products to ecosystem functioning by: (i) upscaling GPP using S3 + Sentinel-5P synergies and GPR hybrids, (ii) developing the D&B v1.0 vegetation model to assimilate SIF and trait indicators into photosynthesis and water-use calculations, and (iii) applying causal inference to identify environmental drivers and domain shifts in vegetation dynamics.
To ensure mission readiness, FLEXINEL implemented interoperable, cloud-ready retrieval libraries (e.g. PyEOGPR, ARTMO integration) and tools for propagation and visualisation of multi-dimensional uncertainty, including per-band radiometric variance and covariance-aware structures, aligned with FLEX and S3product requirements.
FLEXINEL is expected to deliver FLEX-ready retrieval algorithms for SIF and vegetation traits (with uncertainties), harmonised multi-mission L3 time series of photosynthesis indicators, L4 functional products (GPP, NPP), operational cloud pipelines for Copernicus Data Space, open-source UQ libraries (ARTMO, PyEOGPR), validated HyPlant–FLEX–S3 upscaling methods, and a consolidated framework uniting RT physics, ML and ecosystem modelling.