FLEX

Variations in photosynthesis still cause substantial uncertainties in predicting photosynthetic CO₂ uptake rates and monitoring plant stress, which are difficult to measure by reflectance based optical remote sensing techniques. Sun-induced fluorescence in contrast is directly emitted from the core of the photosynthetic apparatus and is a direct indicator for plant health and the effciency of photosynthettic energy conversion.

Here we present several validated maps of sun-induced fluorescence, employing the novel airborne imaging spectrometer HyPlant. HyPlant has an unprecedented spectral resolution, which allows for the first time quantifying sun-induced fluorescence emission in physical units according to the Fraunhofer Line Depth Principle that exploits solar and atmospheric absorption bands. HyPlant serves as both an airborne demonstrator for the FLEX satellite mission, and it also is valuable for strategically focused activities in the validation and interpretation of space-based fluorescence signals at the field scale. Maps of sun-induced fluorescence show a large spatial variability between different vegetation types, which complement classical remote sensing approaches. Different crop types largely differ in emitting fluorescence that additionally changes within the seasonal cycle and are related to the seasonal activation and deactivation of the photosynthetic machinery. Additionally, we show examples how fluorescence can track acute environmental stresses and can be used to improve our forward modelling of actual photosynthesis.

The FLuorescence EXplorer (FLEX) satellite mission is explicitly optimized for detecting the sun-induced fluorescence (SIF) emitted by plants. It allows consistent measurements with a spectral resolution of 0.3 nm at the O₂-B (677 to 697 nm) and O₂-A (740 to 780 nm) bands, related to the red and far-red fluorescence emission peaks respectively. The FLuORescence Imaging Spectrometer (FLORIS) sensor, aboard of FLEX, will provide continuous medium and high-resolution radiance spectra over the entire red and near-infrared spectrum from 500 to 780 nm. The high resolution part of the spectrum is exploited for retrieving the full SIF emission spectrum in the 677 to 780 nm spectral range by using the Spectrum Fitting algorithm (SpecFit). This new way for looking at plant’s fluorescence provides better information about SIF compared to the values retrieved from single narrow channels at the oxygen absorption bands only. The total fluorescence spectrum is composed of the fluorescence emitted by Photosystem I (PS_I) and Photosystem II (PS_II), the first contributes only in the far-red while the second has one peak in the red and one in the far-red. The spectra are subjected to re-absorption and scattering effects which occur within the canopy before SIF is detected by the sensor. Consequently, the top of canopy fluorescence spectrum is strongly modified and loses part of its information related to plant’s activity. These effects can be potentially quantified and accounted for by studying the spectral signature of the top of canopy SIF spectrum. In this context, this work shows the current version of the fluorescence spectrum retrieval algorithm for FLEX and the first attempts in decoupling the PSI and PSII radiance spectra.

The algorithm is tested and evaluated by employing atmosphere-surface radiative transfer simulations obtained by coupling SCOPE and MODTRAN5 codes. The simulation dataset produced in the framework of the ESA PARCS and BRIDGE projects, considers more realistic conditions because it includes directional effects, and the top-of-atmosphere radiance spectra are resampled to the current specifications of the FLORIS spectrometer. The accuracy in deriving the total fluorescence spectrum, PSI and PSII spectra is evaluated by comparing the reference simulated spectra against the retrieved one at different wavelengths.

The capability of deriving the PSI and PSII spectra by using the novel retrieval algorithm is also evaluated exploiting real measurements. The leaf level fluorescence emission spectrum detected by using the FluoWat leaf clip device is used to provide an initial evaluation on the real data. Afterwards, top of canopy radiance measurements collected by ground-based high-resolution spectrometers and by the airborne imaging spectrometer HyPlant. The results show the possibility of deriving total fluorescence and the PSI and PSII spectra from the radiative transfer simulations. The results obtained using real spectral measurements are encouraging. The capability of the novel SpecFit retrieval algorithm will open new perspectives for studying and using SIF for understanding plant functioning.

In this paper, we outline and prototype our design for an Earth Observation Land Data Assimilation System (EO-LDAS) to produce improved estimates of land surface biophysical properties.In this, multiple physically-based operators (based on radiative transfer codes) are used to map from state space to the observations in a variational system that provides updates of state conditioned on the observations (and associated uncertainty). Any form of EO data can be incorporated, provided we have a suitable (and consistent) observation operator. This means that we can make optimal use of all available data (and other constraints) to help constrain what is otherwise a ill-posed problem. The system also incorporates regularisation (smoothing) and other model constraints.

The heart of the system is a (3D - space/time) grid of locations and defined support over which the biophysical properties are represented. The core resolution is matched to a particular monitoring task, but is typically that of the highest spatial resolution sensors (e.g. 10s of metres for Landsat/Sentinel-2 type observations). A full set of biophysical parameters X is defined, and subsets of these used as apprriate for each sensor sensitivity (e.g. using optical data, leaf area, orientation, leaf pigments and other biochemical properties, and soil related terms). The system solves for updated biophysical parameters by forward modelling the sensor response given X within the variational framework.

In this development of the EO-LDAS concept, we consider the processing bottlenecks and present solutions. In particular:

- spatial scaling is explicitly incorporated as we synthesise observations from the core representation. This requires knowledge of the spatial dependencies of X, but is then straightforward.

- whilst running original computer codes for radiative transfer models can be slow and a significant bottlenect, we are only really interested in the mapping between inputs (X) and outputs (satellite observations), and so can apply methods that mimic this relationship at much decreased computational costs. We are using Gaussian process emulation for this task.

- the state representation can be generalised by defining local regions with masks, within which we can treat the biophysical parameters as constant (e.g. over a field) or with a local linearisation

- Careful consideration of the regularisation constraint can lead to efficient linearised processing algorithms making use of e.g. Discrete Cosine Transformations.

- Given the vast amounts of coarse spatial resolution data available, we consider a variant of the system where we use a generic representation to seperately treat and describe these data (based on the ESA GlobAlbedo approach), and then combine this with the higher resolution datasets.

The paper will present examples of these issues and their solutions from various case studies we have been investigating.

A compact dual-field-of-view (DFOV) dual spectrometer system (Piccolo Doppio) has been developed and described previously in ESA workshop proceedings (MacArthur et al 2014). This field spectrometer can measure both down-welling and up-welling solar radiant flux near-simultaneously and is designed for logging applications from fixed locations or measurements from rotary-wing unmanned aerial vehicles. By measuring these fluxes, Earth surface reflectance can then be computed and used to infer surface state variables, e.g. the photosynthetic pigment content of vegetation, or for the validation of airborne or space-based observations. This Piccolo system incorporated up to two USB controlled and fibre-optic input based original equipment manufacturers’ (OEM) optical benches (individual spectrometers). This enabled the selection of spectrometers with different spectral resolutions and sampling intervals for a diverse range of applications. For example, spectrometers to measure across the visible near infra-red range (VNIR) and the short wave infra-red range (SWIR) or the O2-A and O2-B spectral regions could be incorporated. This system has now been successfully field trialled during 2015 and some preliminary results and application descriptions presented in Mac Arthur et al (2015). However, the system described by Mac Arthur et al (2014) is only capable of incorporating two optical benches. Therefore, one of the three possible spectral regions (VNIR, SWIR, or O2-A and O2-B) could not be included. This omission could possibly restrict the use of the system for atmospheric correction of images acquired from airborne or satellite platforms as water vapour and aerosol optical thickness (from the SWIR) or oxygen absorption (from the O2-A and O2-B region) parameters would be omitted. Canopy reflectance investigations are also be more limited by the omission of one of these spectral regions.
To address this omission, the Piccolo system has now been further developed first as a DFOV field spectrometer covering three spectral regions (VNIR, SWIR, or the O2-A and O2-B) (Piccolo Trio) but also as a hyperspectral sun tracking sunphotometer (Piccolo Solaro) covering the same three spectral regions. These Piccolo systems have, as before, two fore optics and input fibre optic bundles but each of these now contains three fibre optic cores. The cores are now distributed to three USB controlled fibre optic input spectrometers. Each input fibre optic bundle has, as before, a high-speed miniature shutter incorporated into the fore optic housing. These shutters are controlled by the same electronic system as the Piccolo Doppio. This enables each to be opened and closed sequentially (within less than 500 ms of each other) and, therefore, allows the systems to measure down-welling and up-welling fluxes and these to be recorded by each optical bench. The Piccolo Trio system is therefore a DFOV measuring simultaneously across the VNIR, SWIR and O2-A and O2-B spectral region and at the high resolutions necessary for validation of fluorescence observations from air- or -space borne imaging sensors, if appropriate optical benches are selected. Furthermore, through the inclusion of a commercial mechanised sun tracker in the system, the Piccolo Trio can be used as a logging sunphotometer. As this system has two input channels (previously described for up-welling and down-welling measurements), it can be configured to measure direct solar irradiance and diffuse solar irradiance. Direct solar irradiance can be measured by incorporating a very narrow field-of-view fore optic onto the end of the up-welling fibre bundle and aligning this with the solar disc and down-welling irradiance by masking the solar disc and shielding the fore optic from direct irradiance. These fore optics, when mounted on a commercially available sun tracker which follows Sun's path automatically, enable and direct and diffuse solar irradiance measurements made. The measurements from these systems can then be used to inform and validate the atmospheric correction and for the validation of Earth surface radiance for Sentinel-2 and Sentinel-3 optical imaging sensors and for airborne and space-based fluorescence observations, if the Fluorescence Explorer (FLEX) mission is selected as the 8th ESA Earth Explorer.
These two Piccolo systems will be more fully described and instrument laboratory characterisation and performance testing (spectral and radiometric accuracies, sampling intervals and spectral band widths) presented and discussed. In addition, preliminary field measurement results will be presented and their utility for atmospheric correction and Earth surface reflectance applications assessed.
References
MacArthur, A. and Robinson, I. (2015). A critique of field spectroscopy and the challenges and opportunities it presents for remote sensing for agriculture, ecosystems, and hydrology. In proceedings SPIE Remote Sensing for Agriculture, Ecosystems, and Hydrology XVI, Toulouse, France. 21st to 25th September, 2014.
MacArthur, A., Robinson, I., Rossini, M., Davis, N., MacDonald, K. (2014) A Dual-Field-of-View Spectrometer System for Reflectance and Fluorescence Measurements. 5th International Workshop on Remote Sensing of Vegetation Fluorescence. 22 - 24 April 2014, Paris, France

Mapping at a global scale the actual photosynthesis of terrestrial vegetation is of particular interest for the improvement in the predictive capability of global models through new parameterizations for canopy photosynthesis and the corresponding exchange processes of energy, water and carbon between the surface and the atmosphere.  Current remote sensing techniques can only provide an estimate of the “potential” photosynthesis, rather than “actual”, but sun-induced chlorophyll fluorescence is a sensitive indicator of the actual photosynthesis in both healthy and physiologically stressed vegetation, which can be used as a powerful non-invasive marker to track the status, resilience, and recovery of photochemical processes. The variations in amplitude and shape of the fluorescence emission spectrum reflect the efficiency of photosynthesis. The integral of the overall fluorescence emission provides information about actual photosynthetic light conversion. The shape of the emission spectrum provides additional information about the vegetation health status.

When measured from space, vegetation fluorescence contributes only a tiny fraction of the signal coming on top of the reflected radiance by the surface, so that specific retrieval algorithms are needed to disentangle fluorescence from reflectance. Also, retrievals of fluorescence from space have to make corrections for atmospheric effects, such as surface pressure, atmospheric temperature profile, vertical distribution of aerosols concentration, and water vapour content. On the other side, proper interpretation of fluorescence levels requires information about vegetation status (leaf chlorophyll content, leaf area index, fractional cover, canopy temperature, etc.) so that a dedicated fluorescence mission must also include such additional measurements of vegetation status.

The ESA’s Earth Explorer FLEX (Fluorescence EXplorer) mission is the first space mission focused on the estimation of fluorescence emission by terrestrial vegetation on a global scale with high spatial resolution (300 m) and resolving the spectral shape of fluorescence emission.  The FLEX mission not only includes the measurement of the full spectrum of fluorescence emission, but also includes explicit measurement of photochemical changes in reflectance (i.e., PRI), canopy temperature measurements and all the relevant variables (chlorophyll content, Leaf Area Index, etc.) needed to asses the actual physiological status of vegetation and to provide quantitative estimates of photosynthetic rates and gross primary production.

The FLEX mission concept consists in a single platform that carries the Fluorescence Imaging Spectrometer (FLORIS), which has been designed and optimised for discrimination of the fluorescence signal in terrestrial vegetation. FLEX will fly in formation with Copernicus Sentinel-3 inorder to further enhance the spectral coverage from measurements made by the Sentinel-3 instruments OLCI and SLSTR, exploiting the synergy between their data and helping in the proper characterization of the atmospheric state and cloud screening, essential for a reliable retrieval of fluorescence emission.

In this paper, we provide the relevant scientific background and an overview of the FLEX mission concept, measurement methods and scientific challenges, describing current status and perspectives.  Ongoing developments in instrumentation, atmospheric correction procedures, signal extraction techniques, and utilization of the fluorescence signal in models and applications will be presented.