In preparation of the FLEX satellite mission we have conducted a series of field campaigns, during which we aimed to (i) correct real world data on SIF to (i) develop a complete FLEX-like reference data set of field data that can be used to develop and test FLEX satellite data processing concepts and data products, (ii) quantitatively understand the dynamics within the SIF signal and their quantitative link to structural and functional vegetation traits to support the development of vegetation stress indicators, and (iii) to evaluate and test the components required for a FLEX Cal/Val concept.

We have conducted two large FLEX campaign activities, namely the AtmoFLEX and FlexSense campaigns, in the years 2017, 2018 and 2019, which aimed to collect complete optical and SIF data from various ecosystems across five European countries. Additionally, we have finalized the PhotoProxy activity (funded by EO4Science), during which we included further field data from the US cornbelt (cooperation with colleagues from the University of Nebraska, USA) aiming to improve our knowledge on the option to include SIF in GPP predictions. Finally, we have completed the AtmoFLEX campaigns, which focussed on the acquisition of atmospheric data and the development and testing of ground-based SIF references systems (FloX system). These ‘FLEX driven’ campaign activities were integrated with other synergistic campaign activities, such as SARSense, SurfSense, or Chime (Vila-Guerau de Arellan et al. 2020, Mengen et al. 2021). With this presentation we will give an overview on the main outcomes and the conclusions, which we could draw from these challenging campaign activities.

For a correct retrieval of the relatively weak solar-induced fluorescence signal from a satellite platform, a stringent atmospheric correction is essential. This challenge was already identified during the selection of the FLEX satellite mission, and the tandem concept between FLEX and Sentinel-3 is a direct answer to this challenge. During above-outlined campaigns, we could close a crucial gap by collecting a complete data set on synchronously recorded detailed atmospheric data, ground based and airborne radiation and SIF retrievals, as well as top-of-atmosphere satellite data, acquired from a tandem constellation between Sentinel-3A and Sentinel-3B during the commissioning phase. At five times slots, we were successfully underflying the Sentinel tandem constellation with the airborne FLEX-like sensor HyPlant, while flying over dedicated atmospheric measurement stations. This dataset is complemented by 14-month long time series of FloX measurements, more than 700 flight lines from the high-performance imaging spectrometer HyPlant, and various associated measurements of bio-physical plant traits, carbon and water flux measurements and dedicated functional measurements for monitoring the effects of environmental stresses on plant health. Using this large reference dataset, which is currently used for the development and testing of the FLEX satellite and FLEX data processing scheme and which delivered quantitative sensitivity parameters on the impact of atmospheric characterization for SIF retrieval, we could:
- show that solar-induced fluorescence is sensitive to early signs of vegetation drought and shows significant changes of its far-red peak already 3 days after the onset of drought, while reflectance indication was only sensitive after 7 days, once drought effects already left visible marks (Damm et al., submitted),
- detect the effect of summer heat waves, which significantly reduced photosynthetic carbon uptake, as clear changes in the solar-induced fluorescence signal predicted by previous model assumptions (Martini et al. 2021),
- establish a data downscaling approach, which can be used to bring top-of-canopy SIF measurements closer to leaf-level SIF, which is the relevant input parameter for many mechanistic vegetation flux models (Siegmann et al. 2021),
- deliver an overview of relative uncertainty and bias estimates for FloX time series and derive requirements towards a calibration and validation network for SIF satellite missions (e.g. FLEX, Buman et al., submitted).

Thus, executing this ambitious and integrated campaign concept, we could (i) lay the basis for the development of the FLEX Cal/Val scheme, (ii) confirm some hypotheses on SIF being a sensitive drought and heat stress indicator, and (iii) greatly extend the data basis on the natural variability of the SIF signal across different ecosystems, the diurnal and seasonal cycle, and as a reaction to environmental extremes. The campaign data has been delivered to ESA and is currently being made freely available via the ESA campaign webpage and data portals.



Selected publications that emerged from these campaign activities:

Siegmann B., Cendrero-MateoM.P., Cogliati S., Damm A., Gamon J., Herrera D., Jedmowski C., Junker-Frohn L.V., Kraska T., Muller O., Rademske P., van der Tol C., Quiros-Vargas J., Yang P. & Rascher U. (2021) Downscaling of far-red solar-induced chlorophyll fluorescence of different crops from canopy to leaf level using a diurnal data set acquired by the airborne imaging spectrometer HyPlant. Remote Sensing of Environment, 264, article no. 112609, doi: 10.1016/j.rse.2021.112609.

Martini D., Sakowska K., Wohlfahrt G., Pacheco-Labrador J., van der Tol C., Porcar-Castell A., Magney T.S., Carrara A., Colombo R., El-Madanay T., González-Cascón R., Martin M.P., Julitta T., Moreno G., Rascher U., Reichstein M., Rossini M. & Migliavacca M. Heat-wave breaks down the linearity between sun-induced fluorescence and gross primary production. New Phytologist, accepted.

Vila-Guerau de Arellan J., Ney P., Hartogensis O., de Boer H, van Diepen K., Emin D., de Groot G., Klosterhalfen A., Langensiepen M., Matveeva M., Miranda G., Moene A., Rascher U., Röckmann T., Adnew G., Brüggemann N., Rothfuss Y. & Graf A. (2020) CloudRoots: integration of advanced instrumental techniques and process modelling of sub-hourly and sub-kilometre land-atmosphere interactions. Biogeosciences, 17, 4375-4404, doi: 10.5194/bg-17-4375-2020.

Mengen D., Montzka C., Jagdhuber T., Fluhrer A., Brogi C., Baum S., Schüttemeyer D., Bayat B., Bogena H., Coccia A., Masalias G., Trinkel V., Jakobi J., Jonard F., Ma Y., Mattia F., Palmisano D., Rascher U., Satalino G., Schumacher M., Koyama C., Schmidt M., Vereeken H. (2021) The SARSense campaign: air- and space-borne C- and L-band SAR for the analysis of soil and plant parameters in agriculture. Remote Sensing, 13, article no. 825, doi: 10.3390/rs13040825.

The FLuorescence EXplorer (FLEX) will be the first mission designed to monitor the photosynthetic activity of the terrestrial vegetation through the retrieval of the solar-induced chlorophyll fluorescence and the true vegetation reflectance (Level-2 products). To provide reliable estimates on the photosynthesis efficiency on large spatial and temporal scales, the intermediate Level-2 products demand specific uncertainty determination. The validation of the Level-2 products is based on the comparison between the retrieved FLEX products and the ground truth measurements. However, similar to the satellite products, ground truth measurements have also associated uncertainties and variances mainly related to 1) instrument performance and calibration, 2) retrieval error, 3) and site dependent spatial/temporal variability, which need to be characterized to perform a fair comparison.

In this context in July 2020 a ground-based and airborne campaign was carried out to evaluate the calibration/validation (Cal/Val) protocols developed between the Spanish National Institute of Aerospace Technology (INTA) and the University of Valencia for the future FLEX mission in the experimental agricultural site of Las Tiesas, Barrax, Spain. During this field campaign various surface types were investigated including heterogeneous crop fields such as melon, pepper, alfalfa, onion, and corn as well as homogenous surface types such as festuca and bare soil fields. In each surfaces type five elementary sampling units (ESUs) were assigned, and during the airborne overpass simultaneously top of canopy and leaf level measurements were performed at each ESUs. The airborne system was equipped with an Airborne Hyperspectral Scanner (AHS, Sensytech), Compact Airborne Spectrographic Imager (CASI 1500i, ITRES), and a high-resolution Chlorophyll Fluorescence sensor (CFL, Headwall). Top of canopy measurements were performed with ASD FieldSpec3 (Malvern Panalytical) and at leaf scale, the FluoWat leaf clip was used, coupled with an ASD FieldSpec3, to characterize the optical properties and fluorescence emission on the different ESU. Furthermore, with the objective of characterizing the real reflectance and solar-induced fluorescence spatial and temporal variability in a continuous and not supervised mode, a CableCam system was tested for continuous measurements over the festuca and melon field. The CableCam is a custom system consisting in a semi-autonomous trolley traveling along a 70 meters zip line, equipped with two high resolution spectroradiometers (one covering the VIS/NIR spectrum range and the other one covering the spectral range of the O2 A and O2 B atmospheric absorption bands – Piccolo system), a multispectral camera (MAIA-S2, SAL Engineering), and a thermal camera (ThermalCapture Fusion, TEAX). The CableCam was set 4 meters above the canopy. Moreover, continuous measurements in the festuca field were collected by a FloX system (Hyperspectral Devices) placed in a two-meter tower. Finally, a sun tracker attached to a second Piccolo system was used to monitor the direct and global solar irradiance during the campaign.

In this study, first the uncertainty of each measurements system was estimated by comparing the measured irradiance with the modelled irradiance generated using the radiative transfer model libRadtran. Secondly, the instrument uncertainty was propagated to estimate the true reflectance and fluorescence retrieval error. Additionally, the retrieved products were compared with the FluoWat leaf level measurements. Thirdly, considering the absolute uncertainty (instrumental and retrieval error) the spatial and temporal variability of the cited Level-2 products was estimated for each studied site. Finally, the airborne imagery absolute uncertainty and the spectral variability among pixels was used to estimate the number of sampling units needed to capture the spatial heterogeneity of a 300x300 meters FLEX pixel. The strategies tested in this study aim to set the roadmap for the Cal/Val protocols of the future FLuorescence EXplorer mission.

The validation of SIF products that will be provided by the FLEX mission is a challenging issue and has been and will be the focus of both past and future ESA campaigns and initiatives. In brief, a good validation strategy implies i) that a reliable mean fluorescence signal (full spectrum) is provided for an area of at least 3x3 pixels of FLEX and ii) the uncertainty associated to the SIF estimation does not exceed the error requirement recommended for FLEX measurements.

Past experiences acquired over a series of field campaigns and modelling studies, has indicated that the first requisite can be potentially satisfied either by using different solutions. For example by exploiting mobile platforms flying at relatively low elevation and that can rapidly cover (or sampling) the entire area required for FLEX validation, or by integrating a sufficient number of ground-based fixed or mobile measurements over that same area. Here, we will show comparison of measurements at different scales, errors, caveats and potential improvements. The second requisite poses an additional challenge, considering the rather incompressible uncertainty associated with the use of airborne or ground-based spectrometers and the obvious dependency of such uncertainty on measurement height and optical features of the near-surface atmosphere.

We present different validation approaches, sites and instruments requirements and the overall sampling approach for performing the validation of the fluorescence metrics throughout the mission lifetime of FLEX. This presentation will summarize all those aspects, by bringing together theoretical considerations as well as a large amount of data which have been acquired over the last 5 years using ground/tower based, drone-based and airborne-based fluorescence sensors. However, some points still need to be better investigated and our contribution will also highlight new directions and additional activities to be undertaken.

In the last decade, intensifying efforts on field campaigns to measure solar-induced fluorescence (SIF) at different scales have been made by distinct Universities, Research Institutions, and within the framework of space mission activities like the FLuorescence EXplorer (FLEX), 8th Earth Explorer from the European Space Agency (ESA). In this context, the SIF research community has gathered increasing interest in canopy-SIF monitoring at the tower and Unmanned Aerial System (UAS) scale both to continue advancing vegetation-orientated scientific studies and to provide accurate validation reference datasets for current and future fluorescence-related satellite-derived products.

Focusing on satellite validation purposes, the estimation of reliable canopy-level SIF spectra to be used as 'ground truth' is not straightforward. Enlarging the footprint size is usually achieved by placing the measurement instrument at a certain distance from the canopy level, e.g., by using sensors mounted on towers or tilting the observation angle. However, the larger the optical path between the target and the measurement instrument, the stronger the atmospheric impact, being oxygen absorption effects the most critical ones to be corrected.

Oxygen absorption features, particularly the oxygen-A band around 760 nm, have been traditionally used in passive remote sensing experiments at the canopy level to measure the chlorophyll fluorescence signal emitted by plants. Absorption features are advantageous spectral regions to measure SIF since the difference between canopy-reflected solar irradiance and emitted fluorescence is reduced; therefore, increasing the sensitivity to detect SIF. However, monitoring SIF at discrete wavelengths could not be sufficient in the context of the FLEX mission where satellite-derived spectrally-resolved SIF is planned to be estimated.

In this work, we propose a 3-step physically-based processing chain mainly designed for fluorescence retrievals at the tower level. The proposed processing accounts for (1) the oxygen absorption estimation through the own-developed toolbox, the O2_TRANS, given the environmental and observation conditions, (2) the oxygen correction and calibration procedure of the acquired spectra, and (3) the application of a SIF retrieval strategy.

In the proposed approach, only main parameters regarding the illumination and acquisition geometry are required to estimate oxygen absorption effects at a specific tower setup, i.e., solar zenith angle, downward-looking observation zenith angle, and target-sensor distance. Also, meteorological parameters of air temperature and pressure (T,p) can be accounted for in the developed O2_TRANS software package for a more detailed oxygen transmittance modeling. The O2_TRANS toolbox is based on the HITRAN database, https://hitran.org/, producing line-by-line oxygen transmittance modeling for the O16O16, O16O17, O16O18 isotopologues. The developed toolbox (publicly available through a git repository) can be easily configured for a specific tower setup. The O2_TRANS provides the set of oxygen transmittance spectra needed to correct the pairs of quasi-simultaneous downwelling and upwelling radiance spectra typically measured to estimate SIF.

In a second step, the set of acquired upward- and downward-looking radiance spectra are oxygen-corrected and calibrated. Here, the word calibration refers to the application of a calibration spectral factor required to compensate for errors associated with the algebra manipulation at the sensor resolution during the oxygen absorption correction. This calibration spectral factor has been empirically formulated to be merely dependent on the O2_TRANS outputs to avoid any more advanced approximation error approach involving model training dependencies.

Finally, in a third step, the preferred SIF retrieval strategy can be applied, i.e., either a differential absorption technique or(and) a spectral fitting method.

This work illustrates with simulated and real datasets a physically-based approach that can be used to correct measurements for oxygen atmospheric effects at the proximal sensing level and is flexible enough to be coupled with any SIF retrieval technique.

Earth observations via satellite are becoming more and more important for all environmental studies and to monitor political restrictions. Most satellite products are based on level 1 top of atmosphere radiance measurements. However, the quality of those data is difficult to estimate. Instruments can be characterized and calibrated on the ground but the changes during the launch and ageing must be characterized in space. Uncertainty estimates can be done comparing with ground truth stations. To apply that comparison the atmosphere must be characterized completely as level 1 data are dependent on interaction within the atmosphere.
A different method for the uncertainty estimate is the comparison of two satellites with similar instrumentation. If the satellites have similar viewing geometries of the same geographic targets within a short time frame, no atmospheric correction is needed. Different instrument characteristics such as the central wavelength and the width of a channel must be considered .
The Sentinel 3 tandem mission was a unique opportunity to compare the same instruments on two different satellite as Sentinel 3A and 3B were flying in the same orbit with a time shift of only 30 s [Clerc et al. 2020, Lamquin et al. 2020, Hunt et al. 2020]. During the tandem phase, OLCI-B was reprogrammed to mimic the future Earth Explorer 8 FLEX for a number of scenes over Europe. The reprogrammed OLCI (OLCI-FLEX) measured in 45 bands within the visible spectral range with high resolution in the oxygen absorption bands.
This data set is used to establish and test a validation strategy for the future FLEX campaign. One step of this validation strategy is the comparison of OLCI and FLEX level 1 radiances. It is based on a transfer function that modifies radiance measurements at OLCI-FLEX bands towards OLCI-A bands to make the measurements comparable. The transfer function requires possible characterization of the atmosphere and the surface. It is done applying an optimal estimation algorithm on the high resolution OLCI-FLEX measurements. The second step of the transfer function is a simulation of OLCI-A measurements based on the found environmental characteristics using a forward model. The optimal estimation algorithm and the forward model are based on look up tables calculated with the radiative transfer model MOMO [Hollstein et al. 2012].
The application of the transfer function for the OLCI-FLEX data showed a difference between OLCI-FLEX and OLCI-A of about 2 % which agrees well with the findings of Lamquin et al. 2020. Furthermore, possible processing issues could be identified for large wavelengths. The results show that the method of the transfer function is sensitive for both instrument differences as well as processing uncertainties. Thus, a good quality control is possible.


Clerc, Sébastien, Craig Donlon, Franck Borde, Nicolas Lamquin, Samuel E. Hunt, Dave Smith, Malcolm McMillan, Jonathan Mittaz, Emma Woolliams, and Matthew Hammond. ‘Benefits and Lessons Learned from the Sentinel-3 Tandem Phase’. Remote Sensing 12, no. 17 (2020): 2668.

Lamquin, Nicolas, Sébastien Clerc, Ludovic Bourg, and Craig Donlon. ‘OLCI A/B Tandem Phase Analysis, Part 1: Level 1 Homogenisation and Harmonisation’. Remote Sensing 12, no. 11 (3 June 2020): 1804. https://doi.org/10.3390/rs12111804.

Hunt, Samuel E., Jonathan PD Mittaz, David Smith, Edward Polehampton, Rose Yemelyanova, Emma R. Woolliams, and Craig Donlon. ‘Comparison of the Sentinel-3A and B SLSTR Tandem Phase Data Using Metrological Principles’. Remote Sensing 12, no. 18 (2020): 2893.

Hollstein, André, and Jürgen Fischer. ‘Radiative Transfer Solutions for Coupled Atmosphere Ocean Systems Using the Matrix Operator Technique’. Journal of Quantitative Spectroscopy and Radiative Transfer 113, no. 7 (1 May 2012): 536–48. https://doi.org/10.1016/j.jqsrt.2012.01.010.