Conference Agenda

Reliable Sun-Induced Fluorescence (SIF) retrieval at high spatial and temporal resolution is essential for understanding vegetation photosynthetic dynamics and for preparing FLEX L2B/L2C validation strategies. As part of the pre-operational activities supporting the FLEX Cal/Val campaigns planned for 2026–2027, an intensive airborne experiment was conducted in summer 2025 over potato and maize research sites under multiple nitrogen treatments. The campaign benchmarked the SIF measurement capabilities of two complementary systems: the Headwall Nano-Hyperspec VNIR imaging spectrometer and the FluorSpec ground-based SIF instrument. The experiment was not intended to upscale observations to FLEX resolutions, but to quantify the sensitivity, stability, and consistency of these systems under controlled physiological gradients, informing future ESA-approved FLORA Cal/Val workflows. SIF from Nano-Hyperspec imagery was retrieved using the Fraunhofer Line Discrimination method at the O₂-A band and validated against synchronous FluorSpec measurements. The campaign was supported by coordinated leaf-level observations, including gas exchange, chlorophyll fluorescence, pigment and nitrogen content, pigments, and leaf area index. Overall, the experiment establishes a robust testbed for assessing UAV-based SIF retrieval performance and uncertainties in preparation for FLEX Cal/Val activities.

Keywords: Gross Primary Production, Sentinel-2, Gaussian Processes, Cloud Computing, Eddy Covariance, Solar-Induced Fluorescence.

Accurate estimation of Gross Primary Productivity (GPP) at high spatial resolution is essential for understanding ecosystem carbon dynamics and for linking remote sensing observations to terrestrial photosynthesis. Optical satellite data from Sentinel-2 enable detailed monitoring of vegetation structure and phenology, yet their capacity to resolve GPP variability varies across ecosystems. GPP estimation based solely on Sentinel-2 reflectance is limited, particularly in structurally complex and evergreen canopies, highlighting the need for complementary physiological measurements at high spatial and spectral resolution, such as solar-induced fluorescence (SIF), which the upcoming FLEX mission is designed to provide.

Here, we present a plant functional type (PFT)–specific framework for high-resolution GPP estimation based solely on Sentinel-2 surface reflectance data. Ten dedicated GPP models were developed for major vegetation types using a hybrid approach that combines SCOPE radiative transfer simulations with active learning to construct compact, informative training datasets. The resulting Gaussian Process Regression (GPR) models were evaluated against eddy-covariance GPP measurements from 67 ICOS sites across Europe, with independent temporal validation (2021–2024) and additional spatial testing using AmeriFlux sites in North America. Model performance was also inter-compared with MODIS GPP products.

Results demonstrate that PFT-specific GPR modeling substantially improves GPP retrievals relative to generalized approaches, with strong performance in deciduous forests, savannas, and wetlands. In contrast, evergreen forests exhibit persistent limitations, reflecting the reduced sensitivity of optical reflectance to physiological variability in these canopies. The inclusion of meteorological variables from ERA5-Land generally did not improve model performance. Importantly, the multi-model GPR framework provides consistent epistemic uncertainty estimates, offering insight into where Sentinel-2 reflectance sufficiently constrains GPP and where it does not.

These findings, recently published in De Clerck et al. (2026), establish a robust baseline for future FLEX applications. By explicitly delineating the ecosystems and conditions under which optical reflectance alone can—or cannot—constrain GPP, this work highlights where the direct physiological signal from SIF will be critical. The presented PFT-specific framework, implemented on cloud platforms, provides a scalable reference for integrating FLEX SIF with Sentinel-2 and complementary observations to advance carbon-cycle monitoring across heterogeneous landscapes.

Reference: De Clerck, E., Reyes-Muñoz, P., Prikaziuk, E., D.Kovács, D., and Verrelst, J. (2026). High-spatial-resolution gross primary production estimation from Sentinel-2 reflectance using hybrid Gaussian processes modeling. ISPRS Journal of Photogrammetry and Remote Sensing, 232, 172–195. https://doi.org/10.1016/j.isprsjprs.2025.11.033

The FRM4FLUO (Fiducial Reference Measurements for Fluorescence) project represents a comprehensive initiative aimed at establishing best practices and guidelines for field-based measurements supporting the validation of ESA's Fluorescence Explorer (FLEX) satellite mission products. The project addresses the critical need for standardized, high-quality ground reference data that can be confidently used for satellite product validation across multiple product levels.

The primary objective of FRM4FLUO is to provide the scientific community with robust methodological frameworks for conducting field measurements suitable for validating FLEX products, including L2A reflectance products, L2B solar-induced chlorophyll fluorescence (SIF) retrievals, and L2C higher-level biophysical parameters such as leaf area index (LAI), chlorophyll content, and photosynthetic electron transport rate. The project integrates both theoretical and experimental components to ensure that the proposed guidelines are scientifically sound and practically applicable.

The theoretical framework developed within FRM4FLUO includes formalized uncertainty propagation methodologies, enabling traceable quality assessment of field measurements and their applicability to satellite validation activities. This rigorous approach ensures that measurement uncertainties are properly characterized and can be compared against satellite product requirements.

The experimental component of FRM4FLUO involved two extensive field campaigns conducted in Tuscany, Italy, where a multi-scale measurement approach was implemented and tested. This scale-up strategy encompassed ground-based measurements using automated and portable spectrometer systems, unmanned aerial vehicle (UAV) platforms, and airborne observations, creating a comprehensive dataset that bridges the spatial gap between point measurements and satellite pixel scales. This hierarchical approach allows for investigation of spatial representativeness and upscaling methodologies essential for proper satellite validation.

The data acquisition report documents the campaign activities, measurement protocols, and data quality procedures implemented during these field experiments. Preliminary results demonstrate the feasibility of the multi-scale approach and provide initial insights into spatial variability of reflectance and fluorescence signals across different observation scales. The outcomes from these campaigns directly inform the development of best practice guidelines and highlight practical considerations for future validation activities.

FRM4FLUO contributes to establishing a standardized framework for FLEX product validation, supporting the broader goal of ensuring the scientific quality and reliability of fluorescence remote sensing data for advancing understanding of terrestrial photosynthesis and carbon cycle dynamics.

Recent advances in hyperspectral and microwave remote sensing products, particularly solar-induced chlorophyll fluorescence (SIF) and vegetation optical depth (VOD), provide complementary insights into photosynthetic activity and vegetation water status of forests, offering new opportunities for improved stress detection under climate change. To translate these observations into robust water and carbon flux estimates, mechanistic models are needed that explicitly link remote sensing signals to underlying biological processes. The FLORES research project (funded by BELSPO) addresses this objective by developing an innovative mechanistic approach through integration of multiscale SIF and VOD observations, from leaf physiology to satellite monitoring. Anchored within the ICOS network, the project leverages continuous carbon and water flux measurements, top-of-canopy SIF, and canopy water dynamics from GNSS-T–derived VOD as key calibration and validation datasets for model development and evaluation. UAV-based hyperspectral (including SIF), thermal, and LiDAR sensors will be deployed to bridge leaf-scale processes and satellite observations. FLORES will further exploit FLEX and SMAP mission data to deliver robust estimates of forest carbon and water fluxes across temperate forests, strengthening the scientific basis for sustainable forest management under a changing climate.

Sodankylä in northern Finland is dominated by boreal forests and bogs that are sensitive to ongoing climate-driven environmental change. The research site in Sodankylä hosts a set of chlorophyll fluorescence measurements that provide a multi-scale view of vegetation fluorescence. Over recent years, the site has supported drone-based, tower-based, and needle-level observations. Current measurements include active MONIPAM observations at the needle scale, as well as passive canopy-level SIF measurements using Piccolo in Scots pine forest and FloX in bog environments.

We use these datasets to illustrate different aspects of fluorescence measurements. Drone-based measurements support analyses of atmospheric effects on the SIF signal, while initial comparisons between FloX observations and satellite-derived SIF demonstrate both opportunities and challenges for high-latitude validation. Time series from bog environments reveal ecosystem-specific fluorescence dynamics, and active MONIPAM measurements provide complementary physiological information that supports SIF interpretation.

Together, these observations highlight Sodankylä’s ESA SUPERSITE as a valuable platform for high-latitude fluorescence research and satellite validation, supported by extensive complementary measurements, including eddy covariance data linking fluorescence to ecosystem carbon uptake (GPP).

The Fluorescence Explorer (FLEX) mission operations are built upon a robust and highly adaptable ground segment designed to transform complex hyperspectral measurements into actionable scientific data. The data processing orchestration employed for this task evolved from a mature, operational science-mission framework within the ESA Earth Explorer programme. While missions such as EarthCARE, BIOMASS, and SWARM utilize this same underlying framework, the FLEX implementation is uniquely tailored to accommodate the specific scientific requirements and production models developed by the mission’s science clusters. The framework’s methodology emphasizes a modular approach, allowing for the seamless integration of mission-specific algorithms coded into operational processors. Currently, this processing orchestrator is deployed within the FLEX Reference Platform serving as the primary environment for the integration, verification, and validation (IV&V) of operational processors, ground segment components, and external interfaces during the mission preparation phase. This presentation details the technical implementation within a public cloud infrastructure, highlighting how the framework manages data streams and dependencies between processing levels and the tandem Sentinel-3 mission. Initial results from the preparation phase provide insights into the computational performance and resource consumption of individual production steps. By analyzing the current versions of the pre-operational processors, we demonstrate how the framework handles varying computational loads. The stability, scalability, and flexibility inherited from its operational predecessors prove ideal for the FLEX mission, and will ensure the delivery of high-quality scientific products with the reliability required for satellite-based environmental monitoring of fluorescence.

This study, within the PHY2FLEX project supporting the ESA FLEX mission, focuses on using sun-induced chlorophyll-a fluorescence to assess phytoplankton physiology, estimate chlorophyll-a concentration (Chl-a), and develop algorithms for phytoplankton characterization and satellite calibration/validation.

This algorithmic development needs in-situ data to relate radiometric measurements with the biogeochemical. PHY2FLEX has already accomplished two field campaigns to test the FloX spectrometer, developed to mimic the FLORIS sensor and hence supporting the Cal/Val activities. Two field campaigns were conducted in October 2025 with synchronous EnMAP acquisition: Venice Lagoon and near to the Acqua Alta Oceanographic Tower (7-9 October); Lake Garda (28 October). During the campaigns, radiometric measurements were acquired from the FLoX and with other spectroradiometers widely used for water applications: WISP-3/Orca, SE RS-3500, and ROX. Water samples for subsequent laboratory analysis and inherent optical properties were collected.

FLoX-derived spectra resulted to be useful for the identification of the local maximum reflectance at 685 nm, in agreement with those gathered by the other devices, proving to be an efficient instrument for fluorescence retrieval in waters. Finally, FLoX data were fitting the corresponding measurements derived by EnMAP.

This study, developed within the FLEX-ITA project, supports the FLEX mission by highlighting its potential for aquatic environments, where sun-induced chlorophyll-a (Chl-a) fluorescence (SIF) from phytoplankton serves as a proxy for Chl-a concentration and provides insight into trophic status, algal blooms, and phytoplankton physiology.

This study focuses on dataset quality control and signal-to-noise ratio assessment of airborne sensors as a prerequisite for reliable aquatic SIF retrieval. Two field campaigns over Lake Garda (2-9 July 2024 and 10-11 June 2025) combined airborne hyperspectral data from IBIS and HyPlant with synchronous PRISMA, Sentinel-3 OLCI, and extensive in situ measurements (radiometric, atmospheric, and limnological data).

The work then focuses on the evaluation of the HYPLANT-SFM algorithm, which retrieves SIF using polynomial and Voigt functions spectral fitting within oxygen absorption bands. SIF products were validated against in situ data and analyzed in combination with Chl-a maps derived from the BOMBER bio-optical model. The validation of SIF and Chl-a maps showed ~10% mean deviations, confirming robust retrievals. The spatial patterns and temporal variability in retrieved SIF provide complementary information to Chl-a on phytoplankton physiology and dynamics, supporting FLEX mission preparations and future SIF product validation for inland waters.

The detectability of genetically encoded color signals in plants from space remains an open and fundamental question in plant biotechnology and remote sensing. From a biochemical standpoint, the key challenge lies in achieving sufficient chromoprotein concentration to generate a detectable optical contrast against dominant background signals without affecting plant development and physiology. Here, we present a theoretical layout of our TRANSPARENT project. In this project, we investigate whether chromoproteins can be expressed in cotton (Gossypium hirsutum cv. Coker 312) leaves to be detected across increasing observation scales. Our strategy relies on genetically modifying cotton plants via somatic embryogenesis to stably integrate DNA constructs encoding chromoproteins under strong, leaf-specific promoters, enabling uniform pigment accumulation within the canopy while remaining compatible with plant physiology and growth. In the next steps, we will test whether the signals can be detected across increasing observation scales, starting from a laboratory setup, drones equipped with cameras, and finally satellite imagery. The data analysis we will develop AI models for signal deconvolution at any stage of the observation scales, which are validated experimentally by field work.

VICALOPS is the European Space Agency's (ESA) new service dedicated to monitoring the radiometric performance of satellite sensors. It integrates various vicarious calibration methods that leverage reflectance from bright desert surfaces, snow, deep convective clouds (DCC), atmospheric molecules, and sunglint. The system synthesizes results by combining data from these methods to provide a comprehensive overview of sensor calibration and stability.

The VICALOPS service has been set up to monitor the radiometric performance of the FLEX FLORIS sensor using stable desert sites and deep convective clouds. Seven spectral windows, selected to avoid strong gaseous absorption features, are used to ensure robust and consistent radiometric assessments. Time series derived from these windows enable the detection of radiometric drifts and support the long-term monitoring of sensor stability throughout the mission lifetime.

Sensitivity study to the spectral window choices will be presented.

Solar-induced chlorophyll fluorescence (SIF) is widely used to infer canopy photosynthesis, but its capacity to separate physiological from structural signals is uncertain. We tested whether SIF can disentangle the onset and progression of physiological drought stress from concurrent canopy biochemical/physical changes, and evaluated leaf–canopy links. In a mesocosm, we manipulated water (control vs. drought) in two herbaceous canopies with contrasting leaf-angle distributions (planophile vs. erectophile). We measured active and passive fluorescence at leaf and canopy scales, canopy traits, and used SCOPE to upscale and analyze drivers. Drought progressively reduced soil water, depressed stomatal conductance, and increased non‑photochemical quenching, lowering fluorescence yields at leaf and canopy scales. Top-of-canopy (TOC) SIF responded significantly to drought, but neither TOC nor down‑scaled SIF detected stress earlier than NIRv. Modeling and observations showed strong non‑physiological influences on TOC SIF as drought altered canopy biochemistry and structure. Leaf-scale TOC fluorescence yields were higher and weakly correlated with canopy‑scale yields; upscaling with SCOPE improved agreement. We conclude SIF reliably captures drought onset, but offers limited early‑detection advantage over greenness indices when concurrent canopy structural/biochemical changes are substantial.

The Ocean and Land Colour Instrument (OLCI) aboard Sentinel-3 is designed to provide high-accuracy measurements of ocean and land surface reflectance. It acquires data at high spectral resolution in the region of the O₂-A oxygen absorption band. The concept of using these OLCI bands to retrieve information on O₂-A Solar-Induced Fluorescence (SIF) is not new (Taveira et al., 2023) but it has never been thoroughly investigated. This presentation will demonstrate how OLCI Level-1 data can be effectively used to estimate the Relative Band Depth (RBD) of O₂-A atmospheric absorption. Initially, this is done indirectly by comparing RBD values calculated from the high-resolution spectrometer of FLOX Units (JB-Hyperspectral, Germany) with those obtained after convolving FLOX spectra to Sentinel-3 resolution. Subsequently, OLCI-derived RBD was compared with long-term FLOX measurements collected over large fields. In addition, OLCI-derived RBD data were analyzed over extended periods across large pivot-irrigated areas in Saudi Arabia to examine cases of decoupling between RBD and both Near-Infrared reflectance indices (NIRv) and the standard APAR product from OLCI during the growing season. The potential applications of OLCI-derived RBD data and the possibility of using Sentinel-3 to bridge acquisition intervals of FLEX (ESA-Earth Explorer 8) mission are outlined and discussed.

Within the landscape of remote sensing signals, the unique selling point of sun-induced chlorophyll fluorescence (SIF) is its sensitivity to the energy splitting in the photosystems. However, the information on the energy splitting cannot simply be read from a SIF value on a 1:1 basis. Difficulties in establishing this link arise from (i) non-linearities in the link between the quantum yields at the photosystem level, (ii) the large within-tree variation in quenching behaviour, (iii) the upscaling from the leaf to the leaf to the landscape scale, where some leaves have a disproportionate effect on the canopy-scale SIF signal, depending on their orientation relative to the sensor. The cumulative effects of these errors might obscure any photosynthetic imprint withing the SIF signal. In this study, we investigate the conditions under which the photosynthetic imprint is still observable. The link between photosynthesis and fluorescence is evaluated at both leaf and landscape scale, for various sites and at a large range of environmental conditions.

To evaluate the strength of the LUE imprint within the φF signal, canopy-scale φF was modelled with a Random Forest in function of LUE alongside with confounding factors affecting the φF signal. The LUE SHAP values turned out more negative under high PAR conditions, with LUE SHAP approximating zero under low-PAR conditions.

Airborne nocturnal hyperspectral measurements at different altitudes were made on 6^th of July 2024 with IBIS and Hyplant sensors over an illuminated greenhouse operated by the company FRI-EL GreenHouse srl. The greenhouse is located in the vicinity of Ostellato (Ferrara, Italy) at 44.71N, 12.71W. It has a surface of 10ha and at the time of the measurements was cultivated with tomatoes in soil-less farming. During nighttime hours Illumination is provided by a large number of LED lamps having a spectrum with two major peaks in the blue and in the red and that does not extend above 700nm. The light travelling from the surface to the top of the atmosphere is the radiance reflected and emitted by the vegetation, and this was measured by two paired high-res spectrometers (IBIS and Hyplant) at 7 altitudes (2200, 3350, 4500, 5600, 6700, 8900 and 11900 ft).

The fluorescence signal from the vegetation was small, due to the relatively low illumination level (PAR< 400umol m-2 sec-1) and to incomplete soil cover but clearly detectable at all flying altitudes. The analysis of the O2A region of the spectrum showed a consistent altitude-dependent deepening of the O2A absorption band that accurately matched the prediction made with Radiative Transfer model.

This presentation highlights the value of night-time measurement for cross calibration of sensors and measurements’ sensitivity analysis.

The FLuorescence EXplorer (FLEX) mission will deliver a global Sun-Induced Fluorescence (SIF) product at 300 m spatial resolution enabling field-scale analysis. However, its low temporal sampling rate limits its utility for time-critical applications. To bridge this gap, we propose the use of a sensor-agnostic SIF retrieval method to combine the FLEX SIF product with estimates from other spaceborne hyperspectral sensors.

In this contribution, we present our ongoing efforts to extend a deep-learning-based retrieval framework (Buffat et al., 2025) to enable sensor-agnostic retrievals in the O₂-A band. The original method has yielded state-of-the-art accuracy for HyPlant and the first DESIS SIF retrievals by combining a small neural network encoder—trained via self-supervised reconstruction—and hyperspectral radiative transfer emulation (presented by Pato et al.). Though promising, the focus on individual-sensor datasets and small networks limits its generalizability.

To overcome this, we introduce a large multi-sensor dataset (HyPlant FLUO/DUAL, DESIS, EnMAP, EMIT, Aviris-NG) to train a dynamic-one-for-all foundation model (Xiong et al., 2024) as a masked autoencoder (MAE) across multiple sensors. We present preliminary results on fine-tuning this model for SIF retrieval in HyPlant and DESIS data and apply it to FLEX FLORIS simulations, assessing the method’s potential for FLEX.

Water availability strongly influences photosynthetic efficiency and crop productivity, making water stress a key challenge for crop monitoring and irrigation management. Chlorophyll fluorescence is a promising signal for the remote sensing of photosynthesis, as it reflects the partitioning of absorbed light energy among photochemical processes, non-photochemical thermal dissipation, and fluorescence emission. However, fluorescence is often interpreted assuming a linear relationship with photosynthetic efficiency, which may be misleading under stress conditions. This study analyses the non-linear dynamics between chlorophyll fluorescence and photosynthetic efficiency in tomato plants subjected to water deficit. Results from three complementary experiments conducted under controlled conditions were integrated to assess how water stress modulates the photosynthetic response, using active chlorophyll fluorescence imaging, hyperspectral spectroscopy, and infrared thermography. The results show that water deficit alters the relationship between fluorescence and photosynthetic efficiency, with non-linear dynamics emerging under specific stress conditions. These findings highlight the need to consider non-linear behaviour when interpreting fluorescence signals and reinforce the potential of chlorophyll fluorescence for early detection of water stress in remote sensing applications.

AndesFlux is a network of four eddy covariance and forest-plot sites situated along a 1200 km north-south transect in the western Amazon. This latitudinal transect closely follows precipitation and dry season length gradients and aligns with the trajectory of the South American Low-Level Jet (SALLJ), which transports moisture from north to south. Our work aims to understand the carbon and water budgets of these forests, with special emphasis on the southwestern Amazon, the least studied region in Amazonia.

This report concerns the Tambopata site (Ameriflux PE-TNR), located in southeastern Peru. We have measured water and carbon eddy-covariance fluxes on a 53 m high tower operating since 2017 in this primary forest within the Tambopata National Reserve. Since March 2025, a FloX field spectrometer (JB Hyperspectral Devices GmbH) / F760/ F687 system has also been operating on the site. Analyses of Net Ecosystem Productivity (NEP) from 2017–2024 reveal that, during the wet season (January-March), the ecosystem is a very small carbon sink, becoming a larger sink during the El Niño phase of ENSO. In the dry season (June-August), the ecosystem becomes an overall source of carbon to the atmosphere. Collected FloX data for April–December 2025 also show a marked seasonality in SIF: lower values for both far-red (F760) and red (F687) fluorescence have been observed during the dry season when compared to the transitional months between seasons (April-May and October-November), correlating well with calculated NEP and evapotranspiration values. Lower SIF values may be related to changes in canopy leaf area index and seasonal leaf shedding. Alternatively, reduced water availability and higher evaporative demand may cause stomatal limitation, which would be reflected in the SIF values. We are currently evaluating the underlying mechanisms of these observations by including leaf-level photosynthesis measurements and a functional characterization of the dynamic acclimation of the plants.

The Doñana site presents a unique biodiversity gradient where xerophilic (Monte Blanco) and hygrophilic (Monte Negro) habitats are abruptly separated by groundwater depth, which is essential for testing the sensitivity of FLEX products to functional diversity and ecosystem stability. In 2025, during the field campaign carried out within the “Spanish FLEX-S3 Mission Calibration and Validation Plan Implementation (SpaFLEXImp)”, different photoprotection measures (fluorescence and pigments, via hyperspectral signatures) were studied in both areas of Doñana as adaptive mechanisms to stressful environmental conditions and as indicators of photosynthetic efficiency and energy dissipation capacity in Mediterranean shrubland plants: Rosmarinus officinalis, Halimium halimifolium, and Phillyrea angustifolia.

Hyperspectral measurements were performed with a spectroradiometer (FieldSpec 3 Hi-Res, Analytical Spectral Devices (ASD) Inc., Boulder, USA) coupled with the FluoWat leaf clip. A spectral unmixing technique was employed to retrieve the fluorescence quantum efficiency and the flux of photons absorbed by beta-Carotene and xanthophyll pigments, key photoprotection mechanisms under stress conditions.

The results show that the increase in photoprotection mechanisms is associated with dynamic regulation of fluorescence and pigments, allowing the safe redistribution of absorbed energy and reduction of photosystem damage, although not equally across all studied species. These processes contribute significantly to stress tolerance and the physiological plasticity of plants in variable environments. The findings highlight chlorophyll fluorescence and photoprotection pigments as key tools for studying adaptive strategies and assessing the physiological status of plants under environmental stress conditions.

Heat dissipation is a crucial photoprotective mechanism that regulates plant energy balance by dissipating excess excitation energy. This process is commonly quantified through non-photochemical quenching of chlorophyll fluorescence (NPQ), which serves as an indicator for physiological status and stress responses. However, while leaf-level NPQ is well-characterized, comparative studies on estimating NPQ from canopy reflectance remain limited.

In this study, we evaluated three distinct approaches for NPQ estimation using data from a field experiment on maize. The dataset consists of canopy reflectance and intensively sampled NPQ observations at the top of the canopy. The three approaches represent physically based, semi-empirical, and data-driven methods, respectively: (1) Model Inversion (MI), which retrieves the photochemical reflectance parameter (C_x) through numerical inversion of the SCOPE model; (2) Vegetation Indices (VIs), including the photochemical reflectance index (PRI), soil-adjusted canopy PRI (SaPRI), and the corrected PRI form (ΔPRI) designed to reduce the influence of leaf pigment composition; (3) Spectral Fitting Method (SFM), which utilizes Principal Component Analysis (PCA) to extract xanthophyll-related reflectance signature, and derives a spectral score by projecting canopy reflectance onto the principal component to quantify NPQ-induced variation.

Our results indicate that the MI method showed the weakest performance, primarily due to the low sensitivity of canopy reflectance to C_x. Among the vegetation indices, PRI showed stable and accurate in NPQ estimation at the diurnal scale, and SaPRI outperformed PRI under low canopy cover. Over longer temporal scale, ΔPRI demonstrated superior accuracy by accounting for leaf pigment pool. The SFM method was slightly less effective than PRI in capturing diurnal variations and required corrections for the constitutive pigment pool to maintain accuracy over longer temporal scales.

This work provides critical insights into the development of NPQ-related products for the upcoming ESA’s Fluorescence Explorer (FLEX) mission, enhancing the capability for monitoring vegetation physiological stress from space.

Sun-induced chlorophyll-a fluorescence (SICF) and its photosynthetic quantum yield (φ_f) are key indicators of aquatic photosynthetic activity and phytoplankton stress. Optical remote sensing, with synoptic and repetitive coverage, is particularly effective for such monitoring when fine spectral measurements are available. The Fluorescence Explorer (FLEX) mission, in synergy with Sentinel-3 optical data, offers strong potential to assess phytoplankton physiological state beyond biomass-based metrics. Here, we present an elastic and inelastic scattering coupled radiative transfer (RT) model forming the basis of a retrieval framework for simultaneous estimation of SICF properties and φ_f, together with bulk optically active constituents including chlorophyll-a concentration ([Chl-a]), detrital absorption (a_dg), and particulate backscattering (b_bp), across diverse aquatic optical conditions. The forward RT model includes a depth-integrated formulation of SICF accounting for vertical re-absorption of fluorescence radiance, reducing fluorescence peak overestimation by 30–50% in productive waters ([Chl-a] > 10 mg m⁻³). Inversion is performed using the SABER (Semi-analytical Bayesian Estimate Retrieval) algorithm, a probabilistic framework applicable to both optically deep and shallow waters. Performance is evaluated using synthetic spectra from RT simulations spanning oligotrophic to hyper-eutrophic conditions, exploiting the complementarity of Sentinel-3 OLCI broad visible bands, which constrain elastic scattering and inherent optical properties, and FLEX ultra-spectral measurements resolving fine SICF features in the red to near-infrared. The combined Sentinel-3–FLEX configuration yields robust retrievals of [Chl-a] and b_bp (biases of +11.20% and −0.11%). Targeted ultra-spectral sampling of fluorescence peaks (~685 nm and ~730 nm) enables accurate φ_f retrieval (bias: +2.03%). In contrast, a_dg retrieval accuracy remains limited (bias: +52.94%), likely due to the spectral gap between Sentinel-3 blue-green bands and the FLEX operational range. Future work will address these limitations through modeling of atmospheric absorption coupled fluorescence, species-dependent optical properties, multi-temporal constraints from Sentinel-3 revisit capability, and regionally tuned detrital priors. Overall, this study demonstrates the potential of FLEX-Sentinel-3 synergy for joint retrieval of aquatic constituents and fluorescence-based physiological parameters.

Remote sensing (RS) of solar-induced chlorophyll fluorescence (SIF) is increasingly recognized as a key tool for ecosystem research, providing direct and mechanistic insights into photosynthetic activity across spatial and temporal scales. However, several confounding factors complicate the accurate physiological interpretation of retrieved canopy SIF. A precise understanding of these factors is essential and includes: i) absorbed photosynthetically active radiation by chlorophyll (APAR_chl), ii) complementary radiation pathways such as NPQ, iii) scattering and reabsorption of SIF within the leaf and the canopy, iv) scattering and absorption of SIF in the atmosphere, and v) sensor-related influences on the retrieved SIF. In particular, a comprehensive understanding of the processes of re-absorption and scattering of SIF within the canopy is fundamental for reliably comparing fluorescence observations across different scales and for disentangling the physiological component of SIF detected by RS sensors. In this study, we present the results of a comparative assessment of existing methods for deriving the red and far-red SIF escape probability from the leaf to the canopy scale (σ_F-LC) using a diurnal airborne dataset collected over an agricultural area.

Since the experimental determination of is highly labour-intensive, strongly affected by the measurement techniques and used instruments, and limited to specific illumination conditions, viewing angles, and leaf properties, various approaches have recently been developed making use of TOC reflectance data to correct canopy level SIF measurements for these effects. Re-absorption and scattering of SIF are strongly wavelength dependent. While red SIF (F₆₈₇) is more susceptible to re-absorption due to the overlap with chlorophyll absorption, far-red SIF (F₇₆₀) tends to be more strongly scattered. Several approaches have been proposed to derive surrogates for the product of σ_F-LC and the fraction of APAR_chl (fAPAR_chl). Most of these approaches exploit the similarity between the radiative transfer of intercepted incident light and emitted SIF to approximate σ_F-LC. In this study, we applied four methods to estimate σ_F-LC of F₆₈₇ (σ_F687-LC), namely Redv_Liu, Redv_Wie, CNL and Rom₆₈₇ and seven methods to estimate of σ_F-LC of F₇₆₀ (σ_F760-LC), namely FCVI, NIRv, NIRvH1, NIRvH2, NIRvSR, saR2F and Rom₇₆₀, using HyPlant TOC reflectance data. The resulting estimates were compared with σ_F-LC derived using the hybrid method. Specifically, σ_F-LC simulated with the SCOPE model was combined with Gaussian process regression (GPR) to generate maps of σ_F-LC for both F₆₈₇ and F₇₆₀ (σ_F687-LC SCOPE₆₈₇ and σ_F760-LC SCOPE₇₆₀) from HyPlant TOC reflectance data, which served as the reference dataset for the comparative analysis.

The data used in this study were acquired during the 2019 ESA FLEXSense campaign, a large campaign conducted in preparation for the upcoming ESA Earth Explorer 8 satellite mission FLEX. The investigated area is located in the western part of Germany (50.865228° N, 6.450074° W), 40 km west of Cologne and is composed of 102 agricultural fields. The area is characterized by the cultivation of typical regional crops, such as, winter wheat, sugar beet, potato and maize. Top of canopy (TOC) reflectance and SIF data were recorded by the HyPlant airborne imaging spectrometer on 26 June 2019. The first overflight took place in the morning at 10:45 local time, then data acquisition was continued with the second overflight close to solar noon at 13:40 and the final airborne data set was recorded in the afternoon at 16:15. Each dataset consists of a mosaic created from six flight lines, acquired at an altitude of 680 m above ground level, leading to a special resolution of 1 m.

The comparative analysis of the different approaches used to approximate σ_F760-LC indicates that NIRvH2 and saR2F yield the most consistent estimates relative to those derived from SCOPE₇₆₀. The corresponding scatterplots show a close agreement with the 1:1 line, accompanied by high R² values (> 0.81) and low RMSE (< 0.1), indicating strong overall performance for both methods. In addition, the derived maps of σ_F760-LC derived with SCOPE₇₆₀ and NIRvH2, respectively, based on the midday airborne data set exhibit comparable spatial patterns, with the highest values observed over sugar beet, followed by potato and winter wheat, whereas maize represents a special case due to its early growth stage. In contrast, the methods used to approximate σ_F687-LC show substantial deviations from the SCOPE₆₈₇ reference, as indicated by lower R² (< 0.5) and higher RMSE values (> 0.1).

Overall, the results of this study demonstrate that σ_F760-LC can be estimated reliably using multiple approaches, whereas σ_F687-LC estimates derived from the currently available methods should be interpreted with caution, as they do not show consistent agreement with SCOPE-based σ_F687-LC or with each other. Furthermore, the presented results support the validation of the FLEX σ_F product, which is based on a similar SCOPE-based approach, by identifying which of the tested methods performs best and is therefore most suitable for an independent validation.

SpaFLEX project is implementing a comprehensive Calibration and Validation (Cal/Val) plan for the FLEX-S3 mission. The key pillars of this plan are: Cal/Val test sites network, in-situ (ground, UAV and airborne) instruments characterization, fiducial in-situ measurements, plot spatial characterization and sampling protocols, and uncertainty budgets for Level-2 products

The SpaFLEX Cal/Val strategy is based on spatial characterization of Cal/Val test sites, estimating a number of Elementary Sampling Units (ESUs) and sampling points according to the Cal/Val plot heterogeneity. In-situ FLoX (fixed point and To_Go) and Piccolo (UAV and FluoCat) measurements over Cal/Val plot are carried out following spatial sampling determined. The uncertainty propagation of in-situ fluorescence (SIF) and surface reflectance (Ref) for the 300x300 m area representing a FLEX pixel is performed using the Law of Propagation of Uncertainties and Monte Carlo methods, in successives steps. First, systematic and random uncertainty components of radiance and irradiance are propagated using CoMet-Punpy tool (Community Metrology Toolkit - Propagation of UNcertainties in Python). Then, Monte Carlo simulations of SpecFit algorithm propagate the single FLoX and Piccolo SIF and Ref retrievals. Finally, the upscaled FLEX pixel SIF and Ref retrievals are propagated using Monte Carlo simulations of upscaling transfer function applied. The ultimate goal is to achieve a minimum fluorescence of 2 mW m-2 sr-1 nm-1 (10% uncertainty) and a relative reflectance change of <30% (500-650 nm) over a 300 x 300 FLEX pixel to meet ESA uncertainty requirements for Level 2 products.

This work presents the uncertainty estimation of in-situ SIF and Ref measurements for two Cal/Val plots in the recent Cal/Val campaign carried out in July 2025 in SpaFLEX Cal/Val test site of Doñana (Huelva, Spain) . The two Cal/Val plots are xerophytic shrubland (Monte Blanco) and juniper woodland (Monte Negro). In both plots, Piccolo-FluoCat a FLoX-To_Go measurements were performed over corresponding ESUs and sampling points established. SpaFLEX uncertainty propagation procedure were applied for both setups, in order to evaluate the upscaled SIF and Ref FLEX pixel retrieval over these non-homogeneity Cal/Val Plots.

CO₂ uptake in the Southeast Asian tropical forests is an essential component of the global carbon cycle. Since seasonal changes in canopy structure are subtle in evergreen forests, it is challenging to detect weekly to hourly variation in photosynthetic capacity using remote sensing vegetation indices. Solar-induced chlorophyll fluorescence (SIF) is linked with photosynthetic processes and reflects the short-term dynamics of CO₂ uptake. In the present study, we examined SIF to reveal its contributions from the overstory and understory, as well as the effects of vegetation stresses on efficiency of CO₂ uptake and fluorescence. The study was conducted in the natural tropical evergreen broadleaf forest of Pasoh, lowland Peninsular Malaysia, in 2024. SIF was retrieved at 760 nm and 687 nm wavelengths with the Fraunhofer-line approach on fine spectral resolution radiance data. Data were collected from at 52m and 13m, representing overstory and understory signals. CO₂ flux has been measured with the eddy covariance method. Leaf-scale fluorescence was also measured by the pulse amplitude modulation on two overstory tree species Dipterocarpus sublamellatus and Ptychopyxis caputmedusae at 30 m height. Based on the escape probability approach, the proportion of understory SIF and top-of-canopy SIF were approximately 10% and 20% of total fluorescence at 760 nm, suggesting a substantial role of dense midstory and understory within the vertical profile of the forest. We also found that diurnal patterns were different between overstory and understory SIF in the afternoon, influenced by VPD and light conditions. These results may help us gain a deeper understanding of the fluorescence applications to the carbon cycle and environmental stress in tropical forests.

The SpaFLEX project is implementing a comprehensive Calibration and Validation (Cal/Val) plan for the FLEX-S3 mission. This plan will define Cal/Val test sites, ground, UAV and airborne instruments characterization, fiducial measurements (FRM), sampling protocols, and uncertainty budgets for Level-2 products. In order to achieve in-situ FRM measurements, SpaFLEX establishes a protocol for indoor and outdoor intercomparisons of field spectroradiometers. The aforementioned protocol aims to ensure the reliability of in situ measurements (reflectance, radiance, irradiance), serving as a robust reference for subsequent calibration and validation processes. This work presents the intercomparison experiment carried out in conjunction with the Cal/Val campaign in Doñana (Huelva, Spain) in July 2025. At a set-up with tables and tripods, sets of measurements were recorded for various targets (soil, vegetation) and reference panels with different reflectance properties (90%, 20%) under variable illumination conditions, using an ad hoc circular multi-fiber holder to keep the instrument's radiance (L) fibers close and aligned. The irradiance (E) RCR were placed high above and aligned over the measurement table. Four spectroradiometers with distinct characteristics were employed to data acquisition (ASD FieldSpec3 and Flox). At the final step of the intercomparison protocol, in order to evaluate the consistency, stability, and interoperability of the instruments employed during the campaign the L and E measurements were checked radiometrically and spectrally against LibRadtran radiative transfer code L and E simulations. The results indicate that there are significant differences between the ASD Fieldspec 3 and FLOX measurements. For white reference panel, the irradiance measured by the ASD instruments was substantially higher than that recorded by FLOX, with differences ranging approximately 20%. Similarly, radiance measurements over the gray panel (target) exhibited large discrepancies. In contrast, intra-instrument comparisons revealed greater agreement, with the two FLOX systems recording values with minimal differences around 2% for both radiance and irradiance. Likewise, comparisons between the two ASD instruments showed slightly moderate variability. These intercomparisons highlight the need to employ LibRadtran-based radiative transfer simulations as an independent and physically consistent reference for radiometric and spectral checking for field measurements.

The satellite data products of solar induced fluorescence (SIF) of the TROPOMI mission are of high temporal resolution (daily), while still having a relatively high spatial resolution (3x7 km). Two products have been published so far: TROPOSIF (743-748 nm and 735-748 nm), and CalTech SIF (740 nm). These SIF products are complementary to the coming FLEX data, but also to solar reflectance data. In our study we jointly use solar reflectance and SIF to derive vegetation data products. This approach has similarities with the FLEX Level-2 processor, where hyperspectral reflectance is used together with SIF.

In the Vegetation-CCI project, we use OptiSAIL, a system to simultaneously retrieve soil and vegetation properties from surface reflectance by automatic differentiation of a radiative transfer model, to produce vegetation data products including leaf area index (LAI) from multiple satellite missions combined. A cost function comprising a data and prior component is minimized using a smart selection of available measurements in a moving 10-day window interval. The output also includes diagnostic (forward simulated) data products including the fraction of absorbed photosynthetically active radiation (fAPAR).

We present an extension of this approach with TROPOSIF. We developed a forward simulator for SIF in OptiSAIL, using radiative transfer components of the model SCOPE. The OptiSAIL-SIF inversion outputs the additional vegetation data product of the quantum efficiency of fluorescence (FQE) at TROPOMI spatial resolution, and the diagnostic (forward simulated) product of angularly normalized nadir SIF and SIF in TROPOMI observation geometry at the 1 km resolution of the Vegetation-CCI grid that is smoother in time than TROPOSIF, but otherwise consistent with TROPOSIF.

In our presentation, we demonstrate the algorithm, show results for selected locations, and address the challenges. The challenges include noisiness of the SIF data, the coarser resolution of the SIF data, and computational demands of the inversion.

Validating FLEX-S3 mission products requires statistically robust sampling strategies to ensure spatial representativeness of field measurements at 300 × 300 m pixel resolution. The mission's strict uncertainty requirements (<30% overall, ~10% for solar-induced fluorescence, ~1% for reflectance) necessitate rigorous protocols connecting in situ observations to satellite-scale retrievals. This work presents a comprehensive validation methodology implemented across diverse ecosystem types within the Spanish Cal/Val framework.

The study was conducted at four sites: two holm oak forests in Teruel Province (north-east Spain) — Sarrión (low tree density) and Manzanera (high tree density) — and two in Doñana National Park (south-west Spain): Monteblanco (Mediterranean xerophytic shrubland) and Montenegro (humid shrubland). This ecological diversity provided an ideal testing ground for evaluating methodological adaptability across different spatial heterogeneity regimes.

The first component establishes the statistical framework for determining optimal elementary sampling unit (ESU) numbers (20 × 20 m), integrating finite population correction, spatially weighted k-means clustering with semivariogram-based spatial autocorrelation, and Neyman optimal allocation. Variable selection prioritises proxies exhibiting maximum heterogeneity and minimum spatial clustering using Moran's index and standard deviation analysis.

The second component quantifies within-ESU variability using high-resolution hyperspectral imagery (4 cm, Cubert S185, 125 bands, 450–950 nm). Optimal measurements per ESU are determined by applying finite population sampling formulas to PRI and ChIRE independently as reflectance and fluorescence proxies, considering sensor field-of-view constraints and variable-specific accuracy requirements.

The third component provides multi-resolution upscaling via a hierarchical chain weighted by measurement representativeness. Stratified pixel-level estimation accounts for stratum-specific variability and size, while spatial interpolation addresses incomplete coverage, ensuring consistency and preservation of physically based biophysical variable relationships.

Combined with uncertainty propagation analysis, this framework provides a statistically rigorous, operationally feasible approach adaptable to alternative satellite validation missions requiring similar spatial scaling methodologies.

Scheduled for launch in 2026, the FLuorescence EXplorer (FLEX) mission will deliver unprecedented observations of sun-induced chlorophyll fluorescence (SIF) at 300 × 300 m spatial resolution. Robust calibration and validation (cal/val) strategies are crucial to ensure the accuracy of FLEX products. The Spanish SpaFLEXImp project aims to establish a network of permanent validation sites for FLEX Level-2 reflectance and fluorescence products. This study proposes the Doñana Biological Reserve (RBD) as a cal/val supersite and describes the extensive field activities carried out in 2025.

Doñana offers optimal conditions for satellite validation due to its flat topography, which minimizes terrain effects, low cloud cover frequency, and strong radiometric contrast among land cover types. The site is integrated into international environmental monitoring networks such as eLTER, FLUXNET, and LifeWatch, and hosts eddy covariance flux towers across Mediterranean ecosystems, providing continuous carbon and energy flux measurements essential to assess the relationship between fluorescence and photosynthesis.

Field campaigns in July and September 2025 were conducted. July focused on two contrasting vegetation types: xerophytic shrubland (“Monte Blanco”) and juniper woodland (“Monte Negro”), both instrumented with flux towers. A multi-sensor, multi-scale sampling strategy was implemented following established SpaFLEX protocols. Elementary Sampling Units (ESUs) were defined using Sentinel-2 imagery and UAV-based hyperspectral data (Cubert S185) to characterize spatial heterogeneity within FLEX pixels. Ground observations included continuous fluorescence measurements from tower-mounted FLOX systems, spatial sampling with mobile FLOX instruments, and leaf-level measurements using FluoWat devices coupled with ASD FieldSpec spectroradiometers. A cable-cam system equipped with a Piccolo spectrometer enabled transect measurements and instrument intercomparison. In September focused on airborne sensor validation. Ground reference data were collected using ASD spectroradiometers and UAV-mounted sensors to characterize surface reflectance over vegetation, bare soil and calibration tarpaulins supporting the validation of airborne hyperspectral imagery. The campaigns demonstrate Doñana as a permanent FLEX cal/val site, combining long-term standardized monitoring with the capability to conduct intensive, multi-scale measurements across heterogeneous Mediterranean ecosystems.

Physiological interpretation of vegetation spectral signals under biotic stress remains a key challenge in precision agriculture remote sensing. Pest and virus infections alter the functioning of the photosynthetic apparatus, affecting photochemical efficiency and energy dissipation mechanisms. These changes are often not detected by traditional reflectance-based spectral indices, particularly during early stages of infection. This study evaluates the potential of chlorophyll fluorescence and photoprotective pigment-related traits for early detection of virus-induced stress in Ipomoea batatas. A multi-instrumental leaf-level approach was applied, combining high spectral resolution reflectance and passive fluorescence, active fluorescence measurements, and gas exchange. The relationship between fluorescence parameters and photoprotective pigment estimates was examined in relation to photosynthetic efficiency. Infected plants exhibited reduced photochemical efficiency and increased energy dissipation, as indicated by fluorescence signals. However, structural indices such as NDVI were unable to distinguish between healthy and infected plants. The results demonstrate the value of high-resolution reflectance and chlorophyll fluorescence in the early monitoring of biotic stress.

Airborne imaging spectroscopy systems such as HyPlant provide meter-scale measurements of solar-induced chlorophyll fluorescence (SIF), bridging the gap between ground-based and satellite observations and enabling spatially detailed assessments of forest photosynthesis. However, canopy-level SIF is strongly influenced by illumination, scattering, and reabsorption processes and therefore cannot be directly interpreted as a physiological proxy for photosynthetic performance. This study presents a physically based approach that combines airborne imaging spectroscopy, 3D Discrete Anisotropic Radiative Transfer (DART) modelling, and machine learning to downscale HyPlant top-of-canopy (TOC) SIF radiance to leaf and photosystem levels in deciduous, coniferous, and mixed forest stands.

TOC SIF radiance was retrieved from HyPlant FLUO data using the Spectral Fitting Method extended with a neural network (SFMNN). Forest structure was derived from high-resolution panchromatic imagery and LiDAR data, enabling the delineation of individual tree crowns and the estimation of canopy height. Based on these data, a set of 20 × 20 m virtual forest scenes was constructed in DART, spanning variability in canopy cover and leaf area index, and incorporating 3D representations of beech and spruce trees derived from terrestrial laser scanning point clouds. The DART virtual stands were used to generate a look-up table (LUT) linking HyPlant-like reflectance signals to known structural and leaf biochemical properties. This LUT serves to train neural networks that retrieve tree-level biophysical and biochemical traits from HyPlant hyperspectral reflectance image data. The retrieved traits are subsequently incorporated into DART to parameterize a 3D representation of the entire study area. Forward DART SIF simulations, using the exact illumination conditions and sensor geometry of the HyPlant acquisitions, are then conducted to retrieve spatial maps of photosystem fluorescence quantum efficiency (FQE). In addition, DART radiative budget simulations quantify the fraction of absorbed photosynthetically active radiation (fAPAR) and the canopy SIF escape probability (σ_F), enabling SIF downscaling to the leaf level. For comparison, σ_F is also estimated from reflectance-based optical indices, such as the Fluorescence Correction Vegetation Index (FCVI).

Preliminary results show pronounced spatial variability in σ_F derived from the FCVI–fAPAR approach, with higher values in beech stands and lower values in spruce stands, reflecting contrasting canopy architectures and scattering regimes. By accounting for these structural effects, the TOC-SIF₇₆₀ is downscaled to the leaf level, revealing higher leaf-level SIF₇₆₀ values. Ongoing analysis compares index-based and DART-based estimates of σ_F to assess if spectral indices can adequately represent complex forest canopy APAR and SIF radiative transfer processes. In addition, photosystem’s FQEs retrieved from DART at the individual tree level will be compared with index-based leaf SIF efficiencies, providing a systematic evaluation of their performance as photosynthetic performance indicators. Such a systematic comparison is currently unexplored but critical for robust interpretation of canopy-level SIF observed by airborne and satellite sensors.

The FLEX bottom of atmosphere products surface apparent reflectance and at-surface solar irradiance will be validated using autonomous ground reference data supplied by the LANDHYPERNET and INSIF networks. The validation itself is based on a comparison metric, where the difference between the FLEX products and the reference datasets should be smaller than the combined extended uncertainty of both. When computing the difference between satellite and ground reference data, it is important that the measurands are as similar as possible, i.e. they must be harmonised spatially, spectrally, angularly and temporally.

LANDHYPERNET is an automated network of hyperspectral radiometers, providing multi-angular reflectance measurements for satellite validation. The instrument consists of 2 sensors, covering VNIR to SWIR range. The VNIR data is collected at a wavelength range of 380 – 1000 nm with 0.5 nm sampling and 3 nm resolution, while the SWIR sensor measures in between 1000 – 1680 nm with 3 nm sampling and 10 nm resolution. Reflectance measurements are taken with a field of view (FOV) of 5° and irradiance measurements with a 180° FOV.

INSIF (International Network of Sun Induced Chlorophyll fluorescence) is a network providing processed FLOX (fluorescence box) measurements. It provides measurements from 650 – 800 nm with a 0.17 nm sampling and 0.3 nm resolution and from 400 – 950 nm with 0.65 nm sampling and 1.5 nm resolution. FLOX has a dual FOV allowing it to measure upwelling radiance (FOV: 25°) and downwelling irradiance (FOV: 180°) simultaneously.

The harmonisation approach will require the following steps (fully for reflectance, partially for irradiance): In the spectral harmonisation the dataset with the finer resolution will be convolved to the one with the coarser resolution, i.e. FLEX to HYPERNETS or FLOX to FLEX. For spatial harmonisation, Sentinel-2 data will be analysed to find the least variable, hence best suited region of interest (ROI) and quantify the spatial variability. The angular and temporal correction will be done using a BRDF model and linear interpolation. Each of those steps will introduce additional uncertainties, which will propagated together with the input uncertainties of FLEX, FLOX and HYPERNETS.

In this contribution, we will present the strategy and tools under development to validate the afore mentioned bottom of atmosphere FLEX products starting from ground data using FLEX simulated data.

SIFcam is a novel lightweight dual camera system deployable on uncrewed aerial vehicles (UAVs) for sensing solar-induced chlorophyll fluorescence (SIF) at 760 nm. First introduced in Kneer et al. 2023, it has a potential for scaling SIF observations, specifically for characterizing spatial variation. Unlike previously deployed SIF sensors on UAVs, which are point spectrometers with a wide spectral range but no information on the vegetation structure. SIFcam captures vegetation structural information from images with centimeter spatial resolution. SIFcam has two co-mounted, simultaneously triggered cameras equipped with optical interference filters centered at 760.7 nm and 757.9 nm (1 nm FWHM), enabling SIF retrieval at 760 nm using the standard Fraunhofer line discriminator method. Image pre-processing includes correction for dark current and flat-field effects, conversion of raw digital numbers to radiance, estimation of downwelling radiance using Lambertian reference panels visible in the imagery, and atmospheric correction using the empirical line method.

Since the introduction of SIFcam, the data processing workflow has been refined, and further assessments of data accuracy have been performed. A key challenge is generating robust orthomosaics from overlapping images. To address this, an established photogrammetric workflow was applied, alongside the development and testing of an alternative, customizable processing workflow.

For the first workflow (WF1), the small spatial misalignment between image pairs is corrected within the photogrammetric processing after mosaic generation. In contrast, the second workflow (WF2) applies feature-based image matching and planar transformations to align image pairs prior to mosaicking. During mosaic generation, WF1 blends all overlapping pixels, while WF2 chooses a single pixel value, constrained by quality criteria of the available pixel value for each location e.g. avoiding pixel values influenced by oblique viewing geometries. The final WF2 mosaic is therefore constructed from deliberately selected single-image pixels rather than blended values.

To assess the measurement error and uncertainty of SIF, SIFcam observations were directly compared to (Air)FloX and HyPlant sensors and a systematic assessment of sensor uncertainties was performed on simulated data. During the Fiducial Reference Measurements for Fluorescence (FRM4FLUO) campaign SIFcam was compared to ground-based and airborne FloX and HyPlant observations. In addition, SIFcam observations were explored in a winter wheat field experiment to evaluate their added value for phenotyping, alongside UAV-based multispectral vegetation indices, in distinguishing new and old winter wheat cultivars based on structural and pigment differences. The 3D Discrete Anisotropic Radiative Transfer model (DART) was employed to simulate SIFcam images of different spatial resolutions for a variety of structurally contrasting canopies, varying solar zenith angles and for variations in sensor viewing angle. Analyzing these simulations with known radiance and SIF emissions as well as the signal recorded by SIFcam allows us to pinpoint the most crucial drivers of measurement uncertainty as well as the most suitable parameters for SIFcam image acquisition depending on vegetation canopy structure.

Solar-induced chlorophyll fluorescence (SIF) provides a direct, non-invasive proxy of photosynthetic activity and has become a central observable for the interpretation of vegetation functioning across scales. In the context of the upcoming FLEX mission, there is a strong demand for proximal and airborne reference data that bridge the gap between leaf-level measurements, airborne imaging spectroscopy, and satellite observations. Here, we present the SIFcam, a compact snapshot imaging system designed to measure SIF under natural illumination with high spatial and temporal resolution.

SIFcam is based on a dual-camera architecture equipped with ultra-narrow bandpass filters sampling the oxygen absorption band at 760.7 nm and an adjacent reference band at 757.9 nm (FWHM 1 nm, 90% transmissivity). Camera optics feature a 25 mm lens, resulting in a 29.8° field of view in horizontal and vertical direction. Spatial pixel binning enables sensor integration times suitable for Uncrewed Aerial Vehicle (UAV) applications. This configuration enables quantitative retrieval of emitted SIF at 760 nm, using the standard Fraunhofer line descrimination method, while maintaining centimeter-scale spatial resolution. The system is lightweight (ca. 2 kg), UAV-deployable (154x90x195 mm), and suitable for both mobile campaigns and stationary time-series observations. Previous field experiments have demonstrated robust SIF retrievals and a strong agreement with established ground-based and airborne reference instruments, confirming the system’s capability to capture physiologically meaningful signals.

Floronics, a research-driven start-up, is advancing SIFcam from a validated research prototype towards a scalable, flexible platform over the next two years. Planned developments include further hardware miniaturization, improved functionality such as geotagging and WiFi interface, stable radiometric performance, and a consolidated processing chain for automated SIF mapping. These efforts aim to provide standardized, robust, high-resolution SIF data sets that support FLEX calibration and validation activities, process-based model development, and applications in plant phenotyping and precision agriculture.

Detecting physiological drought stress in forests is crucial for anticipating drought-induced declines in productivity and resilience. However, most operational approaches use broadband vegetation indices (VIs) that only indirectly indicate tree physiological function. Here we evaluate whether a multi-sensor combination of SIF, LST and hyperspectral reflectance–based VIs improves drought-stress detection, leveraging their sensitivity to photochemical regulation, photosynthetic efficiency and stomatal closure–driven reductions in transpiration.

The study is conducted in Pfynwald, Switzerland, a drought-prone ~100-year-old Pinus sylvestris forest with a long-term irrigation manipulation. Airborne data were acquired during three campaigns capturing early and high drought stress conditions. We generate 2 m products from DUAL (hyperspectral reflectance), HyPlant FLUO (hyperspectral fluorescence for SIF retrieval) and TASI (thermal infrared for LST) sensors. Ground measurements acquired close to the flight dates, including soil water potential, tree water deficit, leaf water potential and leaf-level gas exchange, quantify drought responses.

We apply a CNN regression to predict each ground-measured drought proxy from single- and multi-sensor feature sets. Reflectance features include narrowband VIs related to photochemical regulation and plant water status alongside broadband VIs. Top-of-canopy SIF is retrieved using the SFMNN approach and FQE is estimated using the saR2F method. We evaluate performance using pooled out-of-fold R², range-normalized RMSE and Spearman’s ρ under spatially independent 10-fold cross-validation. Using the best-performing models, we aggregate target-specific predictions into a dimensionless ensemble drought index.

Model skill is moderate to high for soil water potential (R² ≈ 0.4–0.7) and lower for tree- and leaf-level variables (R² ≈ 0.1–0.6), consistent with scale mismatches and limited sample size. Multi-sensor models that include FQE and LST outperform single-sensor approaches: FQE most improves predictions of leaf-level gas exchange, whereas LST is important for soil water potential. The ensemble index separates seasonal and treatment-driven drought stress, demonstrating the value of integrating SIF, LST and hyperspectral VIs to move beyond greenness-based drought monitoring toward improved detection of forest physiological stress. This points to new opportunities from the upcoming satellite mission FLEX.

The FLEX mission aims to measure Sun-Induced Fluorescence (SIF) in vegetation, a robust indicator of photosynthetic activity and plant stress. The calibration and validation of FLEX require the use of airborne systems capable of providing estimates of the SIF observable. In order to strengthen its capabilities, INTA has implemented the Headwall Chlorophyll Fluorescence Sensor (CFL) into a remote sensing system designed for manned aerial platforms. The general objective of SpaFLEX project is to develop a Spanish Cal/Val plan for FLEX products. This plan includes the selection of representative test sites, measurement protocols, and uncertainty budgets for Level-2 products. Airborne campaigns provide essential medium-scale data over large areas, with high efficiency, to improve algorithms and models estimating fluorescence and reflectance propagation. Three Cal/Val sites were established across diverse Iberian ecosystems: Doñana (Huelva), La Roda (Albacete), and Sarrión (Teruel), offering variability in vegetation cover, geometry, and environmental conditions.

The hyperspectral acquisition system designed for SpaFLEX was specified by INTA’s Remote Sensing Systems Division and integrated into the platform by the company Heligrafics. This system is based on the implementation of INTA’s CFL, a high–spectral-resolution fluorescence sensor (FWHM, 0.1–0.2 nm) featuring 2160 spectral bands in conjunction with a CASI-1500i VNIR sensor, together with their associated positioning systems. They were installed onboard the Partenavia P68 Observer, which provides a suitable airborne platform due to its high-wing configuration, unobstructed nadir field-of-view and adaptable cabin architecture. The flight operations were conducted over two consecutive days (7–8 October 2025) to ensure full coverage of the three target areas. The study area was subdivided into 7 polygons (~ 8 km²), delineated based on the spatial distribution of in situ instrumentation, vegetation cover and eddy covariance flux towers. Data acquisition was performed in coordination with ground-based sensors (FLOX, Piccolo) and instrumented unmanned aerial vehicles, enabling synchronized measurements and cross-platform consistency among satellite, airborne, and in situ observations. The campaign successfully generated a unique, multi-scale and multi-platform dataset for integrated analysis, that will provide a critical foundation for validating FLEX algorithms and improving Level-2 product quality.

Land cover is an essential variable for monitoring land surface and its change. While the advances in spatial resolution of global land cover maps over the past years have been remarkable, the thematic resolution (number of distinguishable classes) has lagged behind. The upcoming FLEX satellite will provide open hyperspectral data globally. This presents an opportunity to improve the thematic detail of land cover products. FLEX will provide enough detail to be able to discern plant traits and health characteristics, which in turn will allow us to track e.g. ecosystem composition at unprecedented thematic detail. The downside of FLEX is its low spatial resolution compared to satellite sensors like Sentinel-2 (300 m vs 10 m), and a lower temporal resolution (revisit time) compared to other 300 m satellite sensors like Sentinel-3.

The crucial step to increase thematic detail in land cover maps is their extraction from 300 m imagery, where most pixels cover a mix of land cover classes. To retain spatial detail and enable tracking of gradual change, I propose to use land cover fractions for this purpose. Fractions represent each land cover class as percentage area covered, instead of only naming the largest class in an entire pixel. This approach allows intuitively expressing change in area cover over time.

I propose to establish an open-source system that will build on existing land cover mapping initiatives, such as SITS and Iota², but will be oriented towards gradual change mapping and big data processing. To deal with the global scale challenge, it needs to run on demand using a cloud computing platform. ESA’s openEO platform allows upscaling by processing data in European data centers with local access to the satellite data archives, and providing user-friendly on-demand processing. I propose integrating the processing chain in openEO, allowing a combination of multiple sensors, and enabling users to generate land cover maps for their area of interest on demand.

The system would fuse information from multiple satellite sensors to: 1) extract thematic detail from high spectral but low spatial resolution of FLEX; 1) detect changes in land cover and fill gaps based on high temporal resolution sensors, like the daily Sentinel-3 imagery, 3) extract spatial details from low spectral but high spatial resolution sensors, like Sentinel-2, by pixel unmixing; and 4) merge these data streams into a single final product.

Accurate retrieval of solar-induced chlorophyll fluorescence (SIF) from satellite measurements is particularly valuable for advancing our understanding of photosynthetic processes of plants and for monitoring ecosystem health at the global scale. Many existing satellite-derived SIF products generally rely on specific narrow spectral regions, such as the Solar Fraunhofer lines or oxygen absorption bands. However, the accurate retrieval of the full SIF spectrum spanning approximately from 650 to 800 nm, remains challenging due to complex radiative interactions between the atmosphere and the surface of the Earth.

A major difficulty in satellite-based SIF retrieval arises from the uncertainties in the atmospheric state parameters, such as aerosol optical depth and water vapor content. Inaccurate assumptions in the atmospheric state can introduce substantial biases in SIF estimates from top-of-atmosphere (TOA) level observations when attempting to retrieve SIF outside the specific narrow absorption bands. Therefore, any retrieval algorithm aiming to retrieve the full SIF emission spectrum is dependent on the accuracy achieved at the atmospheric correction step.

To address these challenges, we developed a novel Bayesian SIF retrieval algorithm known as SIFFI. SIFFI is designed to jointly retrieve the full SIF emission spectrum and surface reflectance without imposing parametric constraints on the spectral shape of the retrieved signal. SIFFI has been applied at top-of-canopy (TOC), tower, and TOA levels, enabling its use across a wide range of measurement platforms, from ground-based instruments to different satellite sensors.

At TOC level, SIFFI is successfully applied to measured data from a Fluorescence Box (FloX) instrument. At TOA-level, SIFFI is able to provide accurate retrievals using simulated datasets both under perfect atmospheric correction and when the atmospheric state is inaccurately characterized during the retrieval process. The latter is achieved by utilizing the Approximation Error (AE) method, which aims to marginalize the poorly known auxiliary atmospheric parameters in the forward modeling.

Overall, SIFFI represents a step toward more robust and comprehensive SIF retrievals from hyperspectral satellite observations and contributes to maximizing the scientific return of fluorescence-focused missions such as FLEX. This presentation introduces the SIFFI framework and demonstrates how the AE method can be used to mitigate the impact caused by atmospheric uncertainties on SIF retrievals.

Solar-Induced Fluorescence (SIF) is a critical signal for monitoring vegetation health and photosynthetic activity. Accurately retrieving the full SIF emission spectrum (640–850 nm) remains a major challenge due to the signal’s weakness compared to reflected radiance, generally confining reliable measurements to atmospheric absorption bands, specifically Oxygen-A (O₂A at 760 nm) and Oxygen-B (O₂B at 687 nm). Anticipating missions like ESA’s upcoming FLuorescence EXplorer (FLEX), which targets high-resolution, full-spectrum SIF retrieval, necessitates the development and validation of robust reconstruction methods.

This study compares two distinct methods for retrieving the full SIF spectrum: the established Spectral Fitting Method (SpecFit) and the novel Principal Component Reconstruction (PCR) approach. SpecFit models the SIF spectrum based on high-resolution hyperspectral radiance measurements, which are fitted to a dual Lorentzian function. In contrast, PCR uses Principal Component Analysis (PCA) applied to a large database of simulated SIF spectra (e.g., from the SCOPE radiative transfer model). This technique establishes a linear relationship between the input SIF values at the O₂A and O₂B absorption bands and the first two PCs, which capture over 99.84% of the spectral variability, allowing for computationally efficient reconstruction. Hereafter, we also developed a PCR model for leaf-level SIF using simulated databases (e.g., Fluspect-B) and evaluated it with real data obtained by leaf-clip measurements.

The methods were compared using Top-of-Canopy (TOC) SIF field data acquired by FloX spectrometers, with SIF values reconstructed by PCR and validated against SpecFit outputs. Results demonstrated high compatibility between the two techniques, particularly for integrated SIF emission (SIF total) over the 670–780 nm range, achieving a high correlation (R²=0.981) and an RMSE of 11.5 mWm⁻²sr⁻¹. Spectrally, performance varied, revealing the inherent differences in their modeling approaches. The Far-Red peak region (F₇₄₀) showed good agreement, with correlation values generally above R=0.96 and RMSE less than 0.2 mWm⁻²sr⁻¹nm⁻¹. However, the PCR method struggled more significantly in the Red peak region (F₆₈₅) and the spectral valley (695–735 nm), where correlation dropped to R<0.92 and the Normalized Root Mean Square Error (NRMSE) increased to 0.2–0.45%. This discrepancy is attributed to SpecFit's constrained dual Lorentzian structure, which yields distinct peaks and a pronounced valley, versus the smoother, physically unconstrained profile derived from the RTM-based PCA. Finally, we validated the leaf-level PCR with leaf-clip measured fluorescence, achieving correlation values generally above R=0.93 and RMSE less than 0.35 mWm⁻²sr⁻¹nm⁻¹ in full spectrum.

The PCR method proves highly promising for remote sensing applications, especially as it only requires the readily available O₂A and O₂B SIF retrievals, demonstrating comparability to methods requiring full hyperspectral data. Future work will focus on comparing the PCR and SpecFit results at the leaf-level with real data obtained by leaf-clip measurements.

The representation of gross primary production (GPP) in land surface models remains highly uncertain, despite GPP being a key driving component of the terrestrial carbon cycle (Gier et al., 2024). These uncertainties mainly arise from both the lack of direct measurements of GPP above the leaf scale and an incomplete representation of plant physiological processes (in terms of both parameter values and equations), in particular the links between carbon assimilation and nutrient availability.

Solar-induced chlorophyll fluorescence (SIF) has therefore emerged as a useful proxy of photosynthetic activity and of GPP by terrestrial ecosystems (Li et al., 2018). To further constrain parameters controlling photosynthetic activity, satellite-based SIF observations can be assimilated (e.g., from the TROPOSIF product, and, in the near future, the FLEX fluorescence product), as SIF provides information on plant physiological traits that regulate photosynthetic activity and GPP.

A fluorescence module previously developed for ORCHIDEE (Bacour et al., 2019) enables the simulation and assimilation of SIF observations. The ORCHIDEE-N land surface model now includes an explicit representation of the nitrogen cycle (Vuichard et al., 2019), allowing a more mechanistic description of photosynthesis through nitrogen limitations on key leaf traits controlling GPP, such as chlorophyll and Rubisco contents.

Integrating the fluorescence module into a model that explicitly represents leaf nitrogen limitation is expected to improve the simulation of both SIF and GPP by providing a more realistic description of chlorophyll content and photosynthetic capacity. In this study, an updated fluorescence module is implemented in ORCHIDEE-N to consistently link nitrogen availability, SIF, and photosynthetic activity.

We present a first intercomparison of these two model versions (with and without the nitrogen cycle) based on the seasonal cycles of GPP and SIF at seven observational sites in Europe. These sites are drawn from the AustroSIF database (Martini et al., in prep.), which integrates in situ measurements of eddy-covariance fluxes (used to estimate GPP), SIF, and pulse-amplitude modulated fluorescence measurements.

So far, neither the fluorescence model parameters nor those of the nitrogen-explicit module have been optimised in this new version. This preliminary study paves the way for assimilating both site-level data and satellite-derived SIF retrievals to further constrain the model.

Remotely sensed solar-induced fluorescence (SIF) is increasingly used as a proxy for vegetation photosynthetic activity at various spatial and temporal scales. However, the top-of-the-canopy (TOC) SIF radiance signal is driven by the incoming photosynthetically active radiation (iPAR) and highly impacted by other non-physiological factors, such as canopy structure as well as leaves and soil background optical properties. Methods scaling the canopy SIF radiance down to photosystems’ efficiency are, therefore, essential for extracting physiologically relevant information from TOC SIF signal. Among the existing downscaling methods, 3D radiative transfer-based approaches are uniquely suited for highly heterogeneous vegetation canopies, such as forest stands. Our recently developed method, based on the DART 3D radiative transfer model, was capable to retrieve the spatial and temporal variations of the photosystems’ fluorescence quantum efficiency (FQE) of an alfalfa crop from airborne (HyPlant) and in-situ (FloX) SIF observations. In our current research, we adapted this approach for airborne SIF images of structurally more complex forest canopies, specifically, beech and oak deciduous stands located close to Rajec in Czech Republic. Airborne LiDAR-derived voxelized 3D scenes of the two forest stands are first used to simulate hyperspectral airborne images labeled with per-pixel canopy biophysical and leaf biochemical properties. A combined radiative transfer and deep learning approach is employed by training a regressive 3D convolutional neural network (CNN) on the DART synthetic labeled images and applying it to real airborne images to retrieve per-pixel canopy properties of interest. The retrieved properties, i.e., contents of leaf chlorophyll a+b, carotenoids, and anthocyanin, leaf and wood area indices, and canopy height, are subsequently imported in DART to forward simulate the TOC SIF image. The per-pixel FQE is estimated through matching the simulated far-red SIF radiance with the observed counterpart. The retrieved FQE maps are free of confounding canopy structural impacts, providing information on light-adapted, steady-state chlorophyll fluorescence emissivity from photosystems I and II.

Solar-induced chlorophyll fluorescence (SIF) is an emission from plant leaves that can be observed from space. Chlorophyll fluorescence is one of the pathways through which absorbed radiation is lost from plant leaves. The other pathways are photosynthesis and non-photochemical quenching. Because of its link to photosynthesis, modelling of SIF with terrestrial biosphere models (TBMs) enables making use of this remotely sensed data to better understand the biogeochemical cycles.

Photosynthesis, also known as gross primary production (GPP), can be estimated using eddy covariance flux data. In this study, we explore how to use the remotely sensed SIF centered at flux tower sites to improve the model in terms of modelling SIF and GPP. For this purpose, we use SIF observations from TROPOMI, which is on board the Sentinel-5 Precursor (S5P) satellite. We are using a TBM called QUINCY, which models fully coupled biogeochemical cycles and includes a description of SIF. The SIF formulation has been tested at evergreen coniferous sites and in this study we extend our analysis to additional ecosystems equipped with eddy covariance towers (deciduous broadleaf forest, agricultural site and tree-grass savanna). Our analysis focuses on the seasonal cycle of these ecosystems and on the impact of droughts during the growing season. This work paves the way for leveraging SIF remote sensing observations with TBMs and demonstrates cases where the higher spatial resolution of FLEX will provide additional benefits.

The experimental station “Majadas de Tiétar” (ICOS/FLUXNET ES-LMa) is a well-established, high-profile observatory for monitoring ecosystem processes in an open evergreen Quercus ilex forest, representative of Iberian dehesa tree–grass systems. With a sparse tree density of ~25 trees ha⁻¹, the site constitutes a challenging testbed for studying spatial heterogeneity in carbon, water and energy fluxes. Continuous eddy covariance (EC) flux measurements, high-quality meteorological observations and extensive soil monitoring have been maintained since 2003. Since 2015, additional subcanopy EC systems, sap flow sensors and lysimeters have been deployed to quantify the relative contributions of tree and grass layers to ecosystem-scale CO₂ and H₂O fluxes.

The flux tower and associated infrastructure deliver continuous measurements of turbulent fluxes of energy, evapotranspiration (ET) and CO₂ (NEE, GPP), together with a comprehensive suite of meteorological (air temperature, relative humidity, four-component radiation, PPFD, precipitation) and soil profile variables (temperature, water content, water potential). Complementary point dendrometers, micro-tensiometers and sap flow measurements provide detailed information on tree water status and transpiration at the tree scale.

Building on this unique long-term dataset, Majadas is being upgraded to host new continuous proximal remote sensing observations in synergy with existing measurements, to support the development and validation of upcoming FLEX mission products. New instrumentation includes: (i) a FLOX system to monitor hyperspectral reflectance and sun-induced chlorophyll fluorescence; (ii) SWIR spectrometers for remote sensing of canopy water content; and (iii) a thermal camera coupled to a multispectral camera to resolve spatial patterns of surface temperature and vegetation properties across trees and grassland. All systems are mounted on a mast with a rotating arm to alternately observe oak crowns and adjacent open grass.

The newly measured variables will be integrated into TSEB/3SEB energy-balance models to better characterize hydraulic and physiological constraints on water and energy fluxes and to enhance model performance. Within the SPAFLEX project framework, Majadas will contribute to FLEX cal/val activities by testing upscaling strategies in heterogeneous dehesa ecosystems. Besides, we will produce accessible, high-quality data to foster the development of innovative FLEX-based products by the user community.

Solar-induced fluorescence (SIF) has emerged as a valuable proxy for photosynthesis and shows promise for advancing our understanding of ecosystem dynamics. To harness the full potential of SIF at the satellite level, it is crucial to quantify the temporal and spatial variability at smaller scales within satellite footprints.The Fluorescence Imaging Spectrometer (FLORIS) onboard FLEX will measure between 500-780 nm at a spectral sampling interval of 0.1 nm within the oxygen absorption bands. While the FLEX spatial resolution of 300 meters is a major improvement from past SIF products, there still exists a need to quantify spatial variability and validate SIF measurements from within the footprint.

Recent advances have significantly improved our ability to observe canopy level SIF via highly specialized hyperspectral imagers such as the Headwall SIF imager (Headwall Photonics, Fitchburg, MA, USA). This instrument is an ultra-high resolution imaging spectrometer that measures radiance between 670-780 nm at a spectral sampling interval of 0.02 nm (0.3 nm FWHM), fully resolving both the O2-A and O2-B absorption bands. We deployed the Headwall imager at a grassland in California, USA over 3 growing seasons (2022-2024), generating more than 270 images of SIF, fluorescence yield ɸF, and normalized difference vegetative index (NDVI). Our site is highly heterogeneous within an area roughly the size of the FLEX footprint. To capture the resulting variation in photosynthesis and phenology, we deployed the imager at two distinct regions within our site: one highly-dynamic and seasonally dry area and one wetter area adjacent to a creek. Each region contains vegetation plots that measure soil moisture throughout the rooting zone. We find that seasonal NDVI captures phenological changes within the two regions, with the drier area having consistently higher NDVI and being more seasonally dynamic due to the seasonal lack of soil moisture within the rooting zone. Despite lower NDVI, both SIF and ɸF are consistently higher in the wetter area where water remains within the rooting zone for longer into the dry season. It is clear from images that the imager captures significant spatial variation within a small area. We see through our study that the imager is capable of linking physiology to phenology and canopy structure. Ground-based field studies like this will be essential for advancing the ability to FLEX SIF for ecosystem productivity quantification and stress detection.

The FLuorescence EXplorer (FLEX) mission aims to globally map the Sun-Induced Fluorescence (SIF) spectrum. Accurate SIF retrieval from reflected radiance is achieved by the FLORIS imaging spectrometer using a dedicated global spectrum-fitting algorithm. The Optimal Estimation iteratively optimizes the state vector by integrating at-surface apparent reflectance observations, prior information, and their associated uncertainty covariance.

This work introduces the Error Budget Tool (EBT), a unified diagnostic framework to quantify SIF retrieval sensitivity to multiple error sources and support Level-2 processor refinement. EBT enables controlled simulations from ideal to more realistic conditions, including anisotropy and instrumental or atmospheric uncertainties. It also enables to track key diagnostics such as changes in the Averaging Kernel Matrix, the state vector, cost function and errors in key SIF metrics.

The impact of several factors was systematically evaluated: (i) surface properties, including Lambertian and anisotropic; (ii) instrument modeling, from noise-free to realistic conditions, accounting for absolute radiometric gain (ARG), relative spatial and spectral radiometric accuracy (RXRA, RSRA), photon noise, and spectral calibration errors; (iii) uncertainties in atmospheric parameters such as Aerosol Optical Thickness, Ångström exponent, and Henyey–Greenstein phase function.

Results from anisotropic simulations show that the Lambertian surface assumption in Level-2 SIF processing is inadequate for surfaces with strong bidirectional reflectance. These effects introduce residual O₂ absorption features in apparent reflectance that the Lambertian forward model cannot reproduce, causing SIF retrieval biases. To mitigate this, a simplified BRDF model was implemented in L2B, reducing systematic errors.

The EBT includes a simplified instrumental noise model accounting for systematic and random sources. Current ARG characterization (3.5%) is significant relative to mission requirements (MR). Spectral calibration errors at MR thresholds (±0.005 nm) induce notable SIF errors, especially for far-red and total SIF.

Sensitivity tests on atmospheric parameters revealed that errors in the HG parameter represent the dominant atmospheric driver of SIF retrieval errors.

This work has enabled the identification of the main error sources in SIF retrieval and the development of targeted improvements in the Level-2 SIF processor.

Monitoring declines in forest vitality caused by biotic damage is essential for implementing effective forest protection measures and for understanding ecosystem dynamics. Leaf area index (LAI) and the fraction of absorbed photosynthetically active radiation (fAPAR) are key functional indicators of forest productivity and structure, directly linking canopy architecture to photosynthetic performance and decline from biotic disturbance. Unlike broad vegetation indices that provide coarse estimates of canopy greenness, LAI and fAPAR are biophysically based variables that are sensitive to changes in light interception and leaf distribution, making them suitable proxies for spatially heterogeneous patterns of forest dynamics and stress. ESA’s upcoming FLEX mission will provide sun-induced chlorophyll fluorescence (SIF) together with LAI and fAPAR as complementary Level-2 vegetation products, enabling joint assessment of structural, functional, and physiological canopy status - albeit at a relatively coarse spatial resolution of 300 m × 300 m. To accurately interpret FLEX signals and relate them to fine-scale canopy dynamics, rigorous validation will be required. Ground-based measurements must be carefully upscaled and cross-compared with FLEX observations and established satellite products to quantify spatial variability, link spectral signals to underlying physiological processes, and assess uncertainties in the retrieval of forest functional traits.

The Hakel Reserve in the northeastern Harz foothills of Saxony-Anhalt, Germany, represents an ideal natural laboratory for studying LAI and fAPAR dynamics. The oak-hornbeam forests, mixed with linden, cherry, and maple, include tree species that have been increasingly affected by biotic stressors such as the oak splendour beetle. The area contains both actively managed stands and strictly protected unmanaged zones, creating strong contrasts in structure and disturbance history.

In this study we will focus on oak-dominated stands. We will implement a multi-sensor, multi-scale monitoring approach to estimate LAI and fAPAR. At the ground level, four 50 m × 50 m plots will be equipped with digital hemispherical photography (DHP) and cosine-corrected PAR sensors beneath the canopy. LAI will be estimated from DHP-derived canopy gap fraction, using a Beer-Lambert-type formulation that accounts for leaf angle distribution and foliage clumping. fAPAR will be directly estimated from measurements of transmitted and incoming PAR, accounting for both direct and diffuse radiation.

Terrestrial measurements will be upscaled using UAV (unmanned aerial vehicle) campaigns conducted during the growing season and timed to coincide with FLEX overflights. RGB imagery supports canopy segmentation and the upscaling of plot-level DHP metrics, while LiDAR provides 3D structure for deriving leaf area distribution and gap fraction. fAPAR will be estimated both empirically from vegetation indices and through approaches that combine multispectral reflectance with LiDAR-derived structural metrics.

These UAV-based products yield spatially explicit LAI and fAPAR maps. These can be aggregated to larger areas and validated against FLEX, Sentinel-2/3, and EnMAP observations to assess consistency and uncertainties across scales. Complementary physiological and soil measurements will be conducted in accordance with FRM4VEG principles, to promote traceability and standardized uncertainty assessment. The methods developed in this project will provide a novel tool for detecting declines in forest productivity and for early detection of forest vitality loss from biotic disturbance.

We developed TECs-SIF, a terrestrial biosphere model (TBM) that explicitly couples canopy radiative transfer and plant photosynthesis to simultaneously simulate solar-induced fluorescence (SIF) and gross primary productivity (GPP). The model integrates a spectral invariant–based radiative transfer scheme across leaf and canopy scales, enabling mechanistic investigation of how the SIF–GPP relationship varies across forest ecosystems and temporal scales.

TECs-SIF was calibrated and evaluated using observations from four AmeriFlux sites spanning evergreen needleleaf and deciduous broadleaf forests (CA-Obs, US-xDJ, US-NR1, and US-UMB). The model reproduced observed SIF and GPP across multiple timescales, with strong performance at both hourly (SIF: R² = 0.48–0.87, RMSE = 0.03–0.12 W m⁻² μm⁻¹ sr⁻¹; GPP: R² = 0.60–0.79, RMSE = 1.82–5.31 μmol m⁻² s⁻¹) and daily resolutions (SIF: R² = 0.64–0.91, RMSE = 0.02–0.09 W m⁻² μm⁻¹ sr⁻¹; GPP: R² = 0.89–0.97, RMSE = 0.51–2.05 μmol m⁻² s⁻¹). The model captured nonlinear SIF–GPP behavior at sub-daily timescales and emergent linear relationships at daily and monthly scales.

Across sites, the simulated SIF–GPP relationship exhibited strong ecosystem dependence, primarily controlled by canopy structural properties (e.g., clumping index) and leaf physiological traits. These results demonstrate the capability of TECs-SIF to mechanistically resolve SIF–GPP coupling and highlight the importance of explicitly representing radiative transfer–physiology interactions in TBMs for improved ecosystem carbon cycle prediction.

Solar-Induced chlorophyll Fluorescence (SIF) could be used as an indicator of photosynthetic status due to the close relationship between SIF and the photosynthetic apparatus. Terrestrial SIF is emitted throughout the red and near-infrared spectrum and is characterized by two peaks centered around 685 nm and 740 nm, respectively. In this study, we present a data-driven approach to reconstruct the terrestrial SIF spectrum from measurements by TROPOspheric Monitoring Instrument (TROPOMI) on board the Sentinel-5 precursor mission. This approach makes use of solar Fraunhofer lines in the combined spectral windows devoid of strong atmospheric absorption features to retrieve SIF signal from the solar radiation reflected by the surface and atmosphere system. A linear forward model is proposed with a proper selection of its parameter settings. The evaluation of the retrieval results is performed by inter-comparison of the SIF peaks with other SIF datasets. The comparisons display similar spatial distributions for the weekly global composites for the first two weeks in June and December of 2024. Especially the comparison of the far-red SIF datasets with another dedicated far-red SIF retrievals, S5P-TROPOMI SIF Data Product (TROPOSIF), demonstrates close agreement, indicating consistency between the two retrieval approaches. The retrieval uncertainty for the weekly global composite is about 10% and 2% of the peak red and far-red SIF values, respectively, which can be considered as satisfactory error thresholds for global composites of SIF observations. Different spectral features for several typical biomes from reconstructed SIF spectra suggest that red and far-red SIF may carry complementary information on photosynthetic function and biophysical properties of the plant. The reconstruction of the SIF spectrum is achieved for spaceborne measurements with the potential to open new applications for better understanding of the ecosystem function.

Global terrestrial ecosystems are showing a troubling decline in carbon sequestration—driven by nutrient shortages and intensifying drought stress—that current land-surface models struggle to reproduce. Satellite observations reveal a widening gap between real-world CO₂ uptake and model projections, highlighting deficiencies in how we represent nutrient dynamics, water limitations, and long-term vegetation acclimation. Closing this gap demands an integrated, physics-based framework that fully leverages Earth Observation—from VNIR, SWIR, TIR, and microwave sensors—to trace water–energy–carbon–nutrient interactions across scales.

To this end, we propose SPACEO (Soil-Plant-Atmosphere Continuum enhanced by Earth Observation), a digital twin framework that couples the STEMMUS-SCOPE soil-plant process model with the radiative-transfer models. By simulating signals from VNIR, SWIR, to TIR end-to-end, SPACEO uniquely links satellite measurements to the underlying ecophysiology of the soil-plant system.

SPACEO’s research is organized into three interlocking science cases:

1. Unified Forward Simulation:With STEMMUS-SCOPE, we build a single simulator that coherently predicts multi-frequency EO signals, establishing the scientific foundations for synergy of multi-satellite multi-frequency EO data.

2. Advanced Retrievals & Reference Dataset: Drawing on extensive field campaigns, we will build a FAIR-compliant Reference Dataset and develop data-driven, hybrid, and physics-informed machine-learning algorithms to estimate key ecosystem metrics—Nitrogen Use Efficiency (NUE), Water Use Efficiency (WUE), canopy and soil temperatures, and stress indicators—from optical, fluorescence, thermal, and microwave data.

3. Digital-Twin Data Assimilation: By assimilating multi-mission observations into our digital twin, we will generate self-consistent soil and plant states and fluxes across diverse biomes. We will quantify how multi-frequency data tighten model constraints versus single sensors and translate these insights into design blueprints for next-generation, multi-sensor monitoring systems.

Accurately quantifying ecohydrological processes across scales remains challenging due to uncertainties in model parameters and environmental forcings. Solar-induced chlorophyll fluorescence (SIF), which is tightly linked to photosynthesis, provides new opportunities to constrain carbon–water fluxes using remote sensing.

Here, we develop a process-based inverse modeling framework (STEMMUS-MLR) by coupling a soil water–heat transfer model with a mechanistic light response scheme, explicitly representing root-zone soil moisture dynamics and vertical heterogeneity. By assimilating ground-based or satellite SIF as an input variable, the framework directly constrains ecosystem physiology to jointly estimate gross primary productivity (GPP) and evapotranspiration (ET).

Evaluations across diverse ecosystems and a continental-scale network of AmeriFlux sites demonstrate that STEMMUS-MLR robustly reproduces daily and seasonal GPP and ET dynamics, showing strong consistency with eddy covariance observations across plant functional types. Compared with empirical linear SIF–GPP and SIF-ET relationships, the mechanistic framework improves physical consistency, scalability, and reduces parameter dependency, while enabling reliable ET partitioning.

This work highlights the critical role of integrating SIF and soil moisture constraints into land surface models to reduce uncertainty in coupled water–carbon simulations. It provides a scalable pathway for quantifying ecosystem fluxes from site to regional scales and advances the application of satellite observations for ecohydrological prediction.

MicroCarb is a UK–French bilateral satellite mission designed to map the sources and sinks of atmospheric CO₂ at the global scale. Launched in July 2026, the mission is currently in its calibration and validation phase. Among its data products, MicroCarb will provide a Level 2 solar-induced chlorophyll fluorescence (SIF) product through the central digital processing unit.

The SIF retrieval algorithm, developed by the National Centre for Earth Observation at the University of Leicester (UK), employs an optimal estimation framework to derive global SIF from MicroCarb observations. SIF is a valuable measurement in its own right: as a proxy for photosynthetic activity, it provides key insights into terrestrial carbon uptake and the functioning of the carbon cycle. In addition, SIF plays an important role as an auxiliary constraint in the retrieval of atmospheric CO₂.

We present an overview of the MicroCarb mission and describe the SIF retrieval algorithm and its processing workflow. We also present results from orbit simulation experiments, demonstrating the potential to deliver robust global SIF estimates from space across a range of scenarios, with suitable auxiliary retrieval characteristics.