Elsevier

Remote Sensing of Environment

Volume 217, November 2018, Pages 523-536
Remote Sensing of Environment

The Chlorophyll Fluorescence Imaging Spectrometer (CFIS), mapping far red fluorescence from aircraft

https://doi.org/10.1016/j.rse.2018.08.032Get rights and content

Highlights

  • CFIS, a new airborne instrument for retrievals of far red chlorophyll fluorescence was built.

  • A new algorithm for CFIS-based fluorescence retrievals is described and tested on real data.

  • We mapped fluorescence across agricultural fields in Mead at 30 m resolution.

Abstract

The Chlorophyll Fluorescence Imaging Spectrometer (CFIS) is an airborne high resolution imaging spectrometer built at NASA's Jet Propulsion Laboratory (JPL) for evaluating solar-induced fluorescence (SIF) from the Orbiting Carbon Observatory-2 (OCO-2). OCO-2 is a NASA mission designed to measure atmospheric CO2 but one of the novel data products is SIF, retrieved using reductions in the optical depth of Fraunhofer lines in OCO-2’s O2 A-band, covering 757–775 nm at 0.042 nm spectral resolution. CFIS was specifically designed to retrieve SIF within the wavelength range of OCO-2, but extends further down to 737 nm, nearly maintaining the high spectral resolution of the OCO-2 instrument (0.07 vs. 0.042 nm). Here, we provide an overview of the instrument calibration and performance as well as the retrieval strategy based on non-linear weighted least-squares. To illustrate the retrieval performance using actual flight data, we focus on data acquired over agricultural fields in Mead, Nebraska from an unpressurized Twin Otter (DHC-6) aircraft at a flight altitude of 3000 m above ground level (AGL). Spectral residuals are consistent with expected detector noise, which enables us to compute realistic 1-σ precision errors of 0.5–0.7 W/m2/sr/μm for typical SIF retrievals, which can be reduced to <0.2 W/m2/sr/μm when individual data is gridded at 30 m spatial resolution. The 30 m resolution also enabled direct comparison with the Crop Data Layer from the National Agricultural Statistics Service as well as Landsat imagery (NDVI, EVI, Tskin), taken just a day prior to the CFIS overflights. Results show consistently higher vegetation indices and SIF values over soy fields compared to corn, likely due to the respective phenological stage, which might already have affected chlorophyll content and canopy structure (August 15, 2016). While this work is intended to highlight the technical capabilities and performance of CFIS, the comparisons against Landsat and crop types provide insights into how CFIS can be used to study mechanisms related to photosynthesis at fine spatial scales, with the fidelity needed to obtain un-biased SIF retrievals void of atmospheric correction.

Introduction

In recent years, global remote sensing products of solar induced chlorophyll fluorescence (SIF) have become widely available from satellites such as GOSAT, GOME-2, SCIAMACHY and OCO-2 (Joiner et al., 2011, Frankenberg et al., 2011a, Frankenberg et al., 2011, Guanter et al., 2012, Joiner et al., 2012, Frankenberg et al., 2014, Köhler et al., 2015, Sun et al., 2017). A small fraction of light absorbed by chlorophyll is re-radiated at longer wavelengths (660–800 nm), which is termed chlorophyll a fluorescence and has been widely used in photosynthesis research for decades (e.g. Krause and Weis, 1984, Genty et al., 1989, Moya et al., 2004, Corp et al., 2006, Baker, 2008, Campbell and Middleton, 2008. To first order, steady-state SIF, as measured by satellites, places a direct constraint on the amount of excited chlorophyll a, hence on the total absorbed Photosynthetically Active Radiation (APAR) by chlorophyll only. To second order, the yield of SIF (fraction of excited energy undergoing fluorescence) is also related to the amount of non-photochemical quenching (NPQ), which can provide a constraint on environmental stress. Combined, information on SIF and NPQ provides insight into gross primary production (GPP) and thereby helps us gain a more mechanistic understanding of ecosystem carbon exchange (Porcar-Castell et al., 2014). Unlike what we might expect at the leaf scale (Porcar-Castell et al., 2014, Magney et al., 2017), a surprisingly linear relationship between SIF and GPP was observed globally (Frankenberg et al., 2011, Guanter et al., 2012, Yang et al., 2017), which lacks a thorough mechanistic understanding. Vegetation structure and chlorophyll content affect canopy illumination, scattering and absorption of light within the canopy and escape probabilities of SIF photons emitted at different parts of the canopy. At the temporal and spatial scales observed from space, we are currently unable to cleanly attribute variations in SIF to changes in light interception or environmental stress.

Accurate airborne observations of SIF are needed to I) validate satellite-based SIF retrievals, II) disentangle the contributors to changes in SIF at the canopy and landscape scale and III) testing and improving canopy radiative transfer models. The first proposed airborne instrument was the Fraunhofer line discriminator MKII (Plascyk and Gabriel, 1975). Several studies underlined the importance and possibility of observations at fine spatial scale only possible from aircraft (Rascher et al., 2009, Damm et al., 2011, Damm et al., 2015). HyPlant (Rascher et al., 2015), built in support of a the FLEX satellite mission (FLEX: FLuorescence EXplorer), was the first airborne instrument specifically designed to measure SIF from 670 through 780 nm with a spectral resolution of 0.24 nm. Most of the ground-based and airborne SIF retrievals to date rely on changes in the fractional depth of oxygen absorption features in the 760 and 685 nm bands. Here, we will not discuss these methods but refer to Meroni et al. (2009) for an overview of fitting techniques using oxygen absorptions and to Frankenberg et al. (2011a) for a discussion as to why these are more complex to use in space-borne applications.

All currently available satellite retrievals are based on changes in the fractional depth of solar Fraunhofer lines (absorption features in the sun's photosphere) due to the additive SIF signal. These lines are unaffected by elastic scattering within the Earth's atmosphere. None of the aforementioned satellite missions have been specifically designed to retrieve the SIF signal. However, they have still enabled retrievals because the spectral range of the instruments covers Fraunhofer lines and the oxygen A-band at 760 nm that is used in atmospheric remote sensing to account for scattering effects in the atmosphere. Owing to the narrow bandwidth of current satellite instruments, instruments with spectral resolutions <0.1 nm (GOSAT, OCO-2) only measure a small set of Fraunhofer lines. Others, such as GOME-2 or SCIAMACHY provide a larger spectral range with much higher signal-to-noise ratios (SNR) but at much coarser spectral resolution. Both of these trade-offs impact accuracy and precision of SIF retrievals in various ways. Based on experience from prior satellite missions, we designed and built the Chlorophyll Fluorescence Imaging Spectrometer (CFIS), a dedicated airborne instrument with the specific aim of enabling SIF retrievals using Fraunhofer lines, just as current satellites do. CFIS enables SIF retrievals at a much finer spatial resolution than satellites and features a spectral resolution of 0.07 nm and SNR close to OCO-2 but with a much larger spectral range suitable for retrieval. A simulated nadir radiance observed over a typical canopy is shown in Fig. 1, also indicating the spectral ranges of OCO-2 and CFIS, where CFIS is exploiting the extended wavelength range with isolated Fraunhofer lines around 750 nm while keeping almost the same spectral resolution as OCO-2. The primary objective of CFIS was inter-sensor comparison with OCO-2, which has been obtained using OCO-2 underpasses (Sun et al., 2017). In this manuscript, we focus on a detailed technical description of the instrument, the retrieval strategy and retrieval performance using real data. We also underline the potential of CFIS data using overflights near Mead, Nebraska but focus the discussion on more technical and retrieval-related aspects, with an in-depth scientific interpretation being beyond the scope of this work.

Section snippets

Instrument overview

The rationale for CFIS instrument characteristics is largely based on the proven experiences from GOSAT and OCO-2 retrievals using Fraunhofer lines only. The OCO-2 instrument is a 3-channel grating spectrometer launched into a sun-synchronous orbit on July 2, 2014. SIF retrievals are performed using high resolution spectra of the O2 A-band (0.757–0.775μm, FWHM=0.042 nm) recorded at 3 Hz readout rate with 8 independent along-slit focal plane array readouts. The nominal spatial resolution per

SIF retrieval strategy

In the absence of a complete ILS knowledge, we rely on data-driven methods to retrieve SIF (Guanter et al., 2012). If we consider spectral windows without atmospheric absorption features, we can model the observed spectrally resolved radiance L(λ,x) as a function of the state vector elements xi: L(λ,x)=I0(λ)i=0n1xiλi+xnS(λ),where I0 is the basis vector containing the average spectral shape in the absence of any fluorescence signal. Without any additive signal, any changes in this

Discussion and conclusions

We have presented CFIS, the Chlorophyll Fluorescence Imaging Spectrometer, the first high-resolution airborne spectrometer to retrieve far-red fluorescence using only Fraunhofer lines for retrieval. CFIS has been used for satellite inter-comparison (Sun et al., 2017) and here, we provide a technical overview, ranging from instrument design to calibration and retrieval theory and real fit examples. We developed a relatively simple and fast retrieval algorithm, which can be applied to all CFIS

Acknowledgments

The research described in this paper was carried out by the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. ©2018. Additional funding from NASA's Terrestrial Ecology program is acknowledged (16-TE16-0034).

References (37)

  • MoyaI. et al.

    A new instrument for passive remote sensing:: 1. measurements of sunlight-induced chlorophyll fluorescence

    Remote Sens. Environ.

    (2004)
  • RothmanL.S. et al.

    The hitran2012 molecular spectroscopic database

    J. Quant. Spectrosc. Radiat. Transf.

    (2013)
  • BadgleyG. et al.

    Canopy near-infrared reflectance and terrestrial photosynthesis

    Sci. Adv.

    (2017)
  • BakerN.

    Chlorophyll fluorescence: a probe of photosynthesis in vivo

    Annu. Rev. Plant Biol.

    (2008)
  • CiddorP.E.

    Refractive index of air: new equations for the visible and near infrared

    Appl. Opt.

    (1996)
  • CorpL. et al.

    Fluorescence sensing techniques for vegetation assessment

    Appl. Opt.

    (2006)
  • FrankenbergC. et al.

    Disentangling chlorophyll fluorescence from atmospheric scattering effects in O-2 A-band spectra of reflected sun-light

    Geophys. Res. Lett.

    (2011)
  • FrankenbergC. et al.

    New global observations of the terrestrial carbon cycle from GOSAT: patterns of plant fluorescence with gross primary productivity

    Geophys. Res. Lett.

    (2011)
  • Cited by (47)

    • Sensitivity of solar-induced fluorescence to spectral stray light in high resolution imaging spectroscopy

      2023, Remote Sensing of Environment
      Citation Excerpt :

      To test how Fmeasured is affected by stray light when F and r are not constant, we used measurements of vegetation reflectance, solar radiance, and chlorophyll fluorescence emission spectra to simulate reflected vegetation radiance in the presence of F. We then retrieved Fmeasured using the FLD method (Plascyk, 1975; Plascyk and Gabriel, 1975), the 3FLD method (Maier et al., 2003), a spectral fitting method (Cendrero-Mateo et al., 2019), and a singular value decomposition-based method (Frankenberg et al., 2018) under four simulated stray light scenarios that encompass the range of stray light in measurements from typical gratings-based imaging spectrometers. Standard FLD assumes that r and F are constant across the narrow absorption feature used for SIF retrieval.

    View all citing articles on Scopus
    View full text