Conference Agenda

The European Space Agency’s (ESA) Fluorescence Explorer (FLEX) mission is designed to operate in tandem with the Sentinel-3 mission, utilising the combined capabilities of both platforms’ instruments. On board Sentinel-3, the OLCI multispectral instrument and the SLSTR dual-view conical scanning broadband sensor are utilised to characterise key atmospheric components including aerosols and water vapour and to facilitate cloud detection and screening. FLEX, on the other hand, is equipped with the FLORIS high-resolution spectrometer covering wavelengths from 500 to 780 nm with a spectral resolution ranging from 0.3 to 3 nm and a spectral sampling interval between 0.1–2 nm. Such a high spectral resolution allows the disentangling of the solar-induced chlorophyll fluorescence (SIF) signal emitted by vegetation alongside the photosynthetic activity. The accurate estimation of the SIF signal within the 650-780 nm range is a primary FLEX mission objective. This estimation is challenging as the SIF signal – an in vivo proxy for photosynthesis – comprises only a small fraction (∼4%) of the top-of-atmosphere radiance in the red and far-red regions. Therefore, for an accurate spectrally-resolved SIF retrieval, the atmospheric correction process plays a crucial role as inaccuracies in this step can propagate into errors in the final SIF estimates.

This presentation outlines the design and implementation of FLEX’s Level-2 atmospheric correction processor developed as part of the Level 2 Prototype Processor (L2PP) for the ESA FLEX-DISC project (2024-2030). The FLEX-DISC initiative (reference 4000144004/24/I-DT) contributes to the FLEX mission ground segment preparations activities, focusing after launch on supporting FLEX’s commissioning and product validation phases. In the event that Sentinel-3 is unavailable, the processor switches to a contingency processing branch which is described.

To assess the current performance, the results of the L2PP atmospheric correction are presented. These evaluated results include the retrieved atmospheric parameters such as aerosols and water vapour and the apparent surface reflectance which is further used for the retrieval of the SIF. The atmospheric correction is presented for selected scenes from the FLEX End-To-End Mission Performance Simulator (FLEX-E) project. These scenes are suitable for testing the correction as they are created using the FLEX observation model and depict various atmospheric conditions including aerosols and clouds over a realistic surface in the Catalonia region of Spain with an optional elevation model.

[1] European Space Agency. Flex - fluorescence explorer mission, 2024. URL https://www.esa.int/Applications/Observing_the_Earth/FutureEO/FLEX. Accessed: 2024-10-29.

[2] European Space Agency. Flex: Fluorescence explorer mission – report for mission selection. Esa sp-1330/2, European Space Agency (ESA), 2008. URL https://esamultimedia.esa.int/docs/EarthObservation/SP1330-2_FLEX.pdf. Accessed: 2024-10-29.

[3] Neus Sabater, Pekka Kolmonen, Shari Van Wittenberghe, Antti Arola, and José Moreno. Challenges in the atmospheric characterization for the retrieval of spectrally resolved fluorescence and pri region dynamics from space. Remote Sensing of Environment, 254:112226, 2021.

This contribution presents the design and implementation of the Level-2 Solar-Induced Fluorescence (SIF) retrieval processor developed for the FLuORescence Imaging Spectrometer (FLORIS) onboard ESA’s 8^th Earth Explorer FLEX mission. SIF retrieval is a core component of the Level-2 Prototype Processor (L2PP), developed within the ESA FLEX Data Innovation and Science Cluster (DISC) project. The L2PP consists of four sequential modules: (i) L1C, performing geometric, radiometric, and spectral co-registration of FLORIS and Sentinel-3 observations; (ii) L2A, providing atmospheric correction and retrieval of surface apparent reflectance (R*) and surface solar irradiance; (iii) L2B, dedicated to SIF retrieval; and (iv) L2C, which estimates biophysical and photosynthesis related parameters. This work focuses on the L2B module, which retrieves spectrally resolved SIF in the 670–780 nm range and derived products, including SIF in the O₂ absorption bands, red and far-red peak intensity and position, and spectrally integrated SIF.

The L2B module disentangles the SIF from the actual surface reflectance (R) by exploiting the outputs provided by the L2A module, namely the R* spectrum and its associated uncertainty. The inversion is performed within an Optimal Estimation framework, where the inverse problem is solved using Bayes’ theorem considering probability densities and assuming Gaussian distributions for the uncertainties. The forward model simulates R*, and the state vector is iteratively optimized by minimizing a cost function that accounts for both measurement residuals and a-priori constraints. Measurement and a-priori covariance matrices are used to weight the spectral information and to regularize the solution, respectively. Uncertainties propagated from upstream modules are ingested as inputs and consistently propagated through the L2B processing, ensuring coherent uncertainty characterization along the full processing chain.

Recent developments of the SIF retrieval module include the implementation of the Error Consistency Method (ECM) as an iterative regularization strategy. The inversion now consists of two phases: an initial non-regularized phase, followed by a regularized phase guided by ECM. In addition, a new Bidirectional Reflectance Distribution Function (BRDF) model has been introduced to improve the representation of surface reflectance anisotropy. This improvement is particularly important within the O₂ absorption bands, where accurate radiative transfer modelling is critical for reliable SIF retrieval.

The performance of the L2B processor has been evaluated using Test Data Sets generated by the Scene Generation Module and mission End-to-End simulations. Retrieval accuracy was assessed across different canopy types, atmospheric conditions, and instrumental configurations. Results show robust performance under ideal conditions, with low dispersion and high correlation between retrieved and reference SIF. When instrumental noise is introduced, an increase in dispersion is observed, leading to reduced precision. Simulations including radiometric gain uncertainties exhibit additional positive biases, highlighting the sensitivity of SIF retrieval to calibration errors.

Overall, this work demonstrates the feasibility of operational SIF full spectrum retrieval within the FLEX mission framework. Ongoing investigations focus on disentangling the impact of different instrumental noise components and improving the numerical robustness of the inversion, particularly during the initial non-regularized iterations, to mitigate noise sensitivity while preserving retrieval accuracy.

A unique feature of the Fluorescence Explorer (FLEX) mission is the combination of a wide band VNIR hyperspectral spectrometer of relatively low spectral resolution (LR) with a narrow band sub-nanometre spectrometer (HR) for fluorescence retrieval. The combination of HR and LR enables the retrieval of data products related to photosynthesis. The main goal is to differentiate the energy dissipation pathways of photochemistry and heat. In the Level-2 algorithm this is achieved by estimating the fluorescence emission efficiency (FQE) and the non-photochemical quenching (NPQ).

We present the theoretical background of the algorithm for these quantities, and review its performance. First, the fluorescence escape probability and the absorbed photosynthetically active radiation (aPAR) are retrieved with machine learning algorithms trained with a globally representative dataset of spectra generated with the model SCOPE, along with the leaf area index (LAI), leaf chlorophyll content (LCC), and leaf carotenoid content (LCCAR). Second, the fluorescence quantum efficiency (FQE) is computed by normalizing the fluorescence by the escape probability and the absorbed PAR. Third, the NPQ is retrieved by decomposing a soil-corrected reflectance spectrum with principle component analysis (PCA) into fast and slow varying components, and attributing the fast components to NPQ. Finally, the NPQ and FQE are combined using Bayasian statistics to estimate the photochemical and non-photochemical yields and electron transport rate (ETR). A weak prior relationship between NPQ and FQE stabilizes the results. The uncertainty of the data products is estimated by propagation of the uncertainty of reflectance and SIF, and inclusion of uncertainties in prior coefficients. The algorithms have been validated to synthetic FLEX data and field data collected with the FLoX instrument in several campaigns.

The FLuorescence EXplorer (FLEX) mission, developed by the European Space Agency (ESA) as its 8th Earth Explorer, will provide global maps of vegetation fluorescence as an indicator of photosynthetic activity together with the necessary parameters for deriving the amount of carbon assimilated by plants. FLEX is currently planned to be launched in 2026. The FLEX on-board instrument, the FLuORescence Imaging Spectrometer (FLORIS) sensor, will acquire data in the 500 - 780 nm spectral range with a spectral resolution between 0.3 nm (High Resolution, HR) and 1.8 nm (Low Resolution, LR). The spectral sampling interval will be from 0.1 nm in the oxygen absorption bands (748-769 nm and 686–697 nm) up to 2 nm outside the atmospheric absorption bands. The FLEX satellite will deliver data at a spatial resolution of 300 meters, with observations scheduled around 10:00 local time. It will have a swath width of 150 km and a repeat cycle of 27 days. FLEX will operate in tandem with Sentinel-3, working synchronously with the Sentinel-3 Camera 4 (nadir-looking), which has a 14-degree field of view. These characteristics make the validation of FLEX products complex and challenging, especially for those with high dynamism and significant spatial variability. In this context, this contribution provides an overview of the FLEX strategy for the calibration and validation (cal/val) of the satellite’s operational Level 1C and Level 2 science data products.

Dedicated tools and resources are being developed to support the validation and quality assessment processes for L1C and L2. An interactive portal (i.e. the FLEX Collaborative Platform, CP), is under development and it is expected to provide important validation support functionalities. The FLEX CP will provide analysis tools to enable assessment of quality indicators from specific products and address any specialized data processing requirements.

Since FLEX will provide global products, validation activities will be conducted in a wide range of global climate and vegetation conditions. Product accuracy will be assessed over a widely distributed set of locations and time periods via several ground-truth and validation efforts. The input data for validation origins from different ground networks, as well as field campaigns. The Radiometric Calibration Network (RadCalNet)and VICALOPS sites will be exploited for L1C top of atmosphere radiances products, while AERONET (AErosol RObotic NETwork), HYPERNETS and the upcoming International Network of Sun Induced Chlorophyll Fluorescence (INSIF) will be exploited for L2 reflectance and fluorescence products. Sentinel 2 will be used for geolocation validation of the L2 products. For vegetation biophysical products and surface temperature, the LAND VALidation (LANDVAL), the Ground-Based Observations for Validation (GBOV), the Surface Radiation Budget Network (SURFRAD) Atmospheric Radiation Measurement (ARM) and the Advanced Surface Temperature Radiometer Network (ASTeRN) and the United States Climate Reference Network (USCRN) will be exploited

Different validation categories will be exploited, encompassing direct, indirect approaches and inter-comparison with space products, with the aim to fully validate FLEX core products according to the Committee on Earth Observation Satellites (CEOS) guidelines. A comprehensive strategy for the validation of FLEX operational products is planned. This strategy ensures the traceability to the mission requirements, and guarantee that all parameters relevant for the operational products have been adequately validated. The mission requirements are highly challenging, and they cover the end-to-end Earth observation system including high-level requirements, mission operations, data product development and processing, data distribution and data archiving. The validation of FLEX products will consider that both the FLORIS products and the ground truth measurements have inherent uncertainties and variances due to several factors. Risks associated with the validation of the FLEX products have been identified, possible mitigation scenarios have been outlined and back-up solutions proposed. Finally, the validation timeline during pre-launch preparation, commissioning, and routine operations will be detailed in the presentation.

Automated measurement networks such as INSIF and LANDHYPERNET will provide the main source of reference data for the bottom-of-atmosphere reflectance validation of FLEX. One of the main challenges for this kind of validation activities is due to the mismatch between the field of view of the in-situ instruments (between 1 and 10 m) and the FLEX pixel size (300m). In order to make the in-situ measurements representative of the entire FLEX pixel, an upscaling needs to be performed. This can be done by using a third, auxiliary, reflectance dataset covering the whole FLEX pixel with a higher spatial resolution e.g. Sentinel-2 (S2; 10 – 60 m) L2A. Here the auxiliary reflectance data at the location of the in-situ measurement is compared to the mean auxiliary reflectance over the region of interest (ROI) used for the FLEX validation (one or multiple FLEX pixels).

There are significant uncertainties in the upscaling process related to the spatial variability of the validation site being used, which can vary both in time and space due to changes to the surface within a pixel. This spatial variability can be quantified by calculating the standard deviation of the auxiliary pixels within the validation ROI. Since spatial variability varies with spatial scale, a scaling relationship can be established for each validation site by using different spatial resolutions. Which then can be used to scale the spatial variability to the two scales relevant for the upscaling (i.e. the FOV of the in-situ instrument and the pixel size of FLEX), as well as to determine the uncertainties on the upscaling factor. These uncertainties will be spectrally interpolated to the FLEX wavelengths. Results from the recent ESA-funded FRM4FLUO campaign and from commercial high-resolution satellite data have been used to validate this approach.

In addition to the spatial uncertainties, we will also discuss the uncertainties due to temporal variability (related to the small temporal mismatch between the field measurements and FLEX), as well as the main uncertainty components on the reference and FLEX reflectance data. All these are combined in a metrologically robust validation metric to determine whether or not the FLEX reflectances are consistent with expectations and mission requirements.

The ESA FLuorescence EXplorer (FLEX) mission will offer an unprecedented global view of terrestrial photosynthesis through high-resolution observations of sun-induced chlorophyll fluorescence (SIF). Realizing the full scientific potential of FLEX, however, critically depends on the availability of robust and traceable calibration and validation (Cal/Val) strategies for Level-2B (L2B) SIF products. Validating satellite-scale SIF remains inherently challenging, particularly in heterogeneous landscapes where point-based ground measurements may fail to represent the satellite footprint, and conventional airborne imaging approaches are costly and temporally constrained.

To address these limitations, the ESA DISC project explores a multi-scale FLEX Cal/Val framework that bridges ground, airborne, and satellite observations. This contribution presents the FLEX L2B validation concept and the initial findings of testing complementary validation approaches. The methodologies proposed were evaluated using both simulated datasets and real measurements acquired within the ESA FRM4FLUO project.

The FLEX validation strategy primarily relies on in situ–based approaches, which are supported by image-based methods and inter-comparisons with other satellite missions. Two complementary ground-based validation scenarios are considered: single-point validation, based on continuous tower-based spectrometer measurements at well-characterized sites, and multi-point validation, which uses spatial sampling from unmanned aerial systems (UAS) or mobile platforms to explicitly address sub-pixel heterogeneity. Scale-bridging between ground and satellite observations is achieved through transfer functions derived from high-resolution satellite data, while UAS acquisitions are synchronized with FLEX overpasses to minimise temporal mismatches.

The experimental dataset was collected during two intensive field campaigns in agricultural areas in Tuscany (Italy) in May and June 2025. Multi-scale SIF measurements were acquired using ground-based FloX systems and airborne platforms, including the lightweight UAS-mounted AirFloX system, as well as a helicopter-mounted configuration (HELiPOD). Dedicated optimisation methods supported flight planning and sampling design, maximising spatial representativeness while minimising measurement effort.

The results demonstrate that the proposed FLEX Cal/Val approaches are robust under realistic observational conditions. Tower-based measurements provide accurate local validation when supported by appropriate transfer functions, while UAS-based multi-point observations effectively capture spatial variability within the FLEX footprint, significantly improving representativeness. Overall, this study emphasises the key role of UAS-based SIF observations in bridging the gap between ground and space-based measurements, providing a practical and scalable pathway toward reliable validation of FLEX L2B SIF products.

The FLuorescence EXplorer (FLEX) mission, developed by the European Space Agency (ESA), will deliver spatially explicit maps of key biophysical and photosynthesis-related parameters, contributing to an improved understanding of the global carbon and water cycles. The FLEX Level-2 (L2) biophysical products can be grouped into two main categories: (i) traditional biophysical products, including leaf area index (LAI), leaf chlorophyll content (LCC), and the fraction of absorbed photosynthetically active radiation (fAPAR); and (ii) advanced biophysical (photosynthesis-related) products, including absorbed photosynthetically active radiation by chlorophyll a and b (APARchl), leaf carotenoid content (LCARC), fluorescence quantum efficiency (FQE), reversible energy dissipation (RED), electron transport rate (ETR), and fluorescence escape probability (fesc).

FLEX L2 biophysical products are fundamental to the future development of higher-level (L3 and L4) products and therefore require rigorous quality assessment. A comprehensive validation strategy has been developed to quantify their accuracy, consistency, and uncertainty. This strategy integrates direct and indirect validation approaches. Direct validation is supported by in situ observations acquired during dedicated field campaigns as well as by data provided by established observation networks, such as the Ground-Based Observations for Validation (GBOV) network and the forthcoming International Network of Sun-Induced Chlorophyll Fluorescence (INSIF). These datasets include measurements of multiple biophysical parameters, surface reflectance, and chlorophyll fluorescence. Indirect validation strategies include the comparison of FLEX L2 biophysical products with proxy data sources or with variables estimated using independent models or algorithms. In particular, FLEX L2 biophysical products will be evaluated against values retrieved from radiometric measurements through radiative transfer model inversion. The Soil Canopy Observation of Photosynthesis and Energy fluxes (SCOPE) model will serve as the primary framework for this purpose. Another indirect validation strategy involves the systematic intercomparison of the traditional FLEX L2 biophysical products LAI and fAPAR with comparable products derived from other satellite data (e.g. Sentinel-3, Sentinel-2). This intercomparison is intended to assess the consistency, robustness, and overall performance of the FLEX L2 traditional biophysical products across different sensors, spatial resolutions, and retrieval methodologies, thereby strengthening confidence in their accuracy and long-term applicability.

This contribution will give an overview of the different validation strategies currently being developed to assess the performance of the FLEX L2 biophysical products. It further provides insights into the evaluation and refinement of the validation approaches based on data acquired during field campaigns conducted in spring and summer 2025 as part of the ESA Fiducial Reference Measurements for Fluorescence (FRM4FLUO) project.