A Field Platform for Continuous Measurement of Canopy Fluorescence

(1)

A Field Platform for Continuous Measurement of Canopy Fluorescence

Fabrice Daumard, Sébastien Champagne, Antoine Fournier, Yves Goulas, Abderrahmane Ounis, Jean-François Hanocq, and Ismaël Moya

Abstract—This paper presents a field platform for continuous measurement of fluorescence at the canopy level. It consists of a 21-m-high crane equipped for fluorescence measurements. The crane is installed in the middle of the fields dedicated to agricul- tural research. Thanks to a jib of 24 m and a railway of 100 m distance, fluorescence measurements can be performed at nadir viewing over various field crops. The platform is dedicated to the development and test of future passive or active airborne and space-borne vegetation sensors. A new fully automatic instrument, called TriFLEX, has been installed at the end of the jib. TriFLEX is designed for passive measurement of fluorescence in the oxygen A and B absorption bands. It is based on three spectrometers and allows for continuous measurements with a repetition rate of about 1 Hz. The data products are the radiances of the target, the fluorescence flux at 687 and 760 nm, and several vegetation indexes, including the photochemical reflectance index and the normalized difference vegetation index. A new algorithm for fluo- rescence retrieval from spectral bands measurement is described.

It improves upon the well-known Fraunhofer line discriminator method applied to passive fluorescence measurement by taking into account the spectral shape of fluorescence and the reflectance of vegetation. A measurement campaign of 38 days has been carried out in summer 2008 over a sorghum field. The evolution of the signals showed that the crop was suffering from stress due to lack of water. After several rainy days, a reversion of the water stress was observed.

Index Terms—Oxygen absorption bands, sun-induced fluores- cence (SIF), vegetation remote sensing.

I. INTRODUCTION

P

HOTOSYNTHESIS is one of the most important biochem- ical processes on Earth. Plants absorb light and CO₂from the atmosphere and H₂O from the soil to produce carbohydrate and release O2. This process leads to biomass production and gas exchange with the atmosphere. The assessments of gas exchange and biomass production are two important challenges for remote sensing. Most of the indexes used in remote sensing of vegetation are based on reflectance measurements. The de- rived information is mostly related to the amount of vegetation rather than to its status.

Manuscript received December 4, 2009; revised January 26, 2010. Date of publication June 3, 2010; date of current version August 25, 2010. This work was supported by the Programme National de Télédétection Spatiale through the “Plateforme de test pour capteurs de fluorescence satellitaires ou avionnées”

project. The work of S. Champagne was supported by the Centre National d’Études Spatiales through a Terre, Océan, Surfaces Continentales, Atmosphère program.

The authors are with the Laboratoire de Météorologie Dynamique, Cen- tre National de la Recherche Scientifique, Ecole Polytechnique, 91128 Palaiseau Cedex, France and also with the Institut National de la Recherche Agronomique (INRA), Avignon, France (e-mail: fabrice.daumard@

lmd.polytechnique.fr).

Digital Object Identifier 10.1109/TGRS.2010.2046420

A topic remaining to be covered is the quantitative assessment of actual photosynthesis. In this regard, the measurement of chlorophyll fluorescence emission (ChlF), which is closely related to the photosynthetic activity of plants, offers a chal- lenging perspective [1]. Although the light energy absorbed by plants is efficiently used to drive the photosynthetic mechanisms, a small amount is lost through fluorescence emission.

The ChlF is the result of a deexcitation mechanism in competition with photochemical conversion. Therefore, the ChlF emissionin vivois variable. This variability is widely used to evaluate photosynthetic activity of green leaves in the laboratory [1]. Under natural sunlight, the amount of ChlF emitted by the vegetation represents less than 1% of the radiance of the vegetation and, thus, is a very weak signal. However, at certain wavelengths, where the solar spectrum is attenuated (absorption bands of solar or the Earth’s atmosphere), the sun-induced fluorescence (SIF) signal becomes non negligible and can be quantified by the measurement of the filling-in of the absorption bands. This is the Fraunhofer line discriminator (FLD) method originally proposed by Link [2] to highlight the luminescent component of moon radiance. It was used later by Plascyk and Gabriel [3], who built the first airborne instrument to measure fluorescence emission (i.e., the MKII FLD). This method was successfully adapted to the atmospheric oxygen absorption bands [4] to quantify ChlF from vegetation. In short, SIF is quantified by comparing the depth of the atmospheric oxygen absorption band in the solar irradiance spectrum to the depth of the band in the radiance spectrum of the plants. In the last ten years, several passive instruments using this method have been used to measure the fluorescence of vegetation under different conditions from leaf to canopy level. The effectiveness of the method has been checked using a passive fluorescence detector in the O2-A absorption band (760 nm) based on narrowband filters [4], [5], in parallel with a laser-based fluorimeter [6].

Both instruments measured simultaneously, at the leaf level, the fluorescence transients during the induction period with an excellent correlation.

In addition, based on a rotating wheel of filters, a more com- plex sensor allowed to measure almost simultaneously ChlF at 760 and 687 nm [7], from which the F687/F760 fluorescence ratio was calculated [8]. This sensor was also equipped with other filters to obtain the photochemical reflectance index (PRI) (see [9]). This instrument was later used to monitor SIF at 687 and 760 nm and PRI during spring recovery of a boreal forest [10]. Simultaneous CO2flux measurements with an eddy correlation tower showed good correlation between the PRI and the net CO2assimilation, which evidenced the link between the Scots pine photosynthetic activity recovering and the relaxation of nonphotochemical quenching.

(2)

A further development within the interference filter-based instruments was Airflex, an airborne sensor that measures fluorescence simultaneously at 687 and 760 nm [11], [12]. Airflex is basically a six-channel radiometer aimed at measuring the filling-in of the atmospheric O2bands. Several flights with Air- flex were carried out over cultivated fields, during June and July 2005, in the frame of the Sent2flex campaign organized by the European Space Agency (ESA) in Barrax, Spain. The track flew over a succession of cultivated fields and showed good repro- ducibility of the data (≈2%) when repeating the flights. Airflex was also flown in 2007 in the frame of the ESA Cefles2 campaign (preliminary results are presented in [13]). These studies highlight the strong variability of the fluorescence signal associ- ated with different fields, as well as its strong dependence with the vegetation 3-D structure [11], [12]. Indeed, the measured fluorescence signal contains not only information on the physiological state of vegetation (stress, growth, etc.) but also information on its structure and on the spectral signature of the soil.

It is worth noting the possibility of obtaining images of chlorophyll fluorescence using a narrowband multispectral camera as reported in [14]. Interestingly, in this work, an unmanned aerial vehicle (helicopter) [15] was used, which offers a great perspective for fluorescence remote sensing.

Instead of using dedicated instruments based on narrowband filters, other works have been conducted using general-purpose radiometers or spectrometers. The effect of water stress over a canopy of olive trees was studied [16], showing that water stress is detectable using the filling-in of the O₂-A band. The FLD method applied with a 3-nm-resolution spectrometer was used to monitor the nitrogen fertilization in corn [17], [18]. The effect of inhibition of photosystem II on fluorescence signal was studied with a higher resolution spectrometer (0.06 nm), finding a fluorescence signal on the herbicide-treated plant four times greater than on the control plant [19]. They also studied passive fluorescence as an indicator of ozone-induced stress [20], [21].

However, most of the radiometer-based experiments cited above refer only to O2-A band measurements and were often conducted at the leaf level or at a very short distance. Long- term measurements have been reported by Louis et al. [10], who measured the fluorescence of Scots pines with a spot of about 9 m²at a distance of 50 m. However, the inclined sighting view (120^◦ of zenithal angle) complicates the interpretation of these data.

To better compare ground measurements with flight conditions, we describe a field platform dedicated to continuous fluorescence measurement that allows continuous nadir measurement on the same target and under controlled conditions.

Furthermore, we describe a new instrument, called TriFLEX, which is designed to measure passive fluorescence at 687 and 760 nm in a fully automatic manner.

II. DESCRIPTION OF THEPLATFORM

The 21-m-high crane is surrounded by fields dedicated to agricultural research and is located at the Institut National de Recherche Agronomique (INRA) site in Montfavet near Avignon, France(43^◦553.24N,4^◦5246.60E)[Fig. 1(a) and (b)]. The jib is 24 m long and can rotate over 360^◦in less than 1 min [Fig. 1(c)]. A hundred-meter railway allows the crane to move on a north–south axis between a west plot of 100 m×

Fig. 1. (a) INRA’s site with the two main fields (from Google Earth). West field is about 100 m×200 m; east field is about 60 m×150 m. The railway is between both fields. (b) INRA’s crane with TriFLEX at the end of the jib.

(c) Reference observed from one of the webcams. (D) Closer view of TriFLEX.

200 m and an east plot of 60 m × 150 m; thus, different crops can be studied during the same measurement campaign.

The crane can easily be controlled from the ground by only one operator. However, the maximum permissible wind speed to operate is less than 14 m·s⁻¹, thus, windy days, which are frequent in Avignon, are a real concern. This crane has already been used for microwave remote sensing studies [22].

The site contributes to the CARBOEUROPE-IP project for the assessment of the terrestrial European carbon balance by means of continuous monitoring of surface energy budget, evapotran- sipration, CO2exchange with the atmosphere, ground moisture, surface temperature, and meteorological measurements.

III. TRIFLEX

Fig. 2 shows a schematic diagram of the TriFLEX instrument that is mounted at the top of the crane. As other instruments measuring SIF, TriFLEX compares the depth of the atmospheric oxygen absorption bands in the solar irradiance spectrum to the depth of these bands in the radiance spectrum of the vegetation. Hence, two depth measurements are needed: one on a non-fluorescent reference panel gives the depth of the band without fluorescence, and the second on the vegetation gives the band depth changes induced by fluorescence. It is well known that as the sun elevation changes during the day, the band depth also varies, resulting in a diurnal cycle of depths [4], [5]. The depth of the O2bands also depends on atmospheric conditions such aerosol content and clouds screening, which induce fast changes in a range of seconds to hours.

Pseudoparallel recordings of band depths can be achieved with one spectrometer by making alternative measurements on the vegetation and on a reference board [13], [19]. However, the time delay between irradiance and sample measurements limits the use of this method during cloudy days in which rapid changes of illumination may lead to inaccurate filling-in determination.

(3)

Fig. 2. TriFLEX instrument. Sp1 is a broad band spectrometer (300–900 nm, FWHM∼2 nm). Sp2 and Sp3 are two identical spectrometers (630–815 nm, FWHM∼0.5nm). Sp2 acquires the radiance from the vegetation, and Sp3 acquires the irradiance by continuously measuring the radiance from a reference board (see Sp3 FOV on the top left caption). A laptop controls the three spectrometers. It also controls the optical shutters, the reference board rotation, and the acquisition of several environment parameters through the data acquisition unit (Agilent 34970A). The reference board can be rotated from default position to the calibration position (see top left caption) for cross-calibration purposes.

With the TriFLEX instrument, the measurements of the O2 band depth are made in parallel for both reference and vegetation, rather than sequentially. TriFLEX uses two identical spectrometers (HR2000+, Ocean Optics, Dudenin, FL) to measure simultaneously irradiance and vegetation radiance spectra (see Fig. 2). These spectrometers cover the spectral range of 630–815 nm with a resolution of 0.5 nm [full-width at half-maximum (FWHM)] and an encoding resolution of 0.09 nm/pixel. The main advantage from simultaneous measurement on the reference and on the sample is an improved time resolution by a factor of two to three, which allows for fast fluorescence changes induced by clouds and sun spells to be monitored. A third spectrometer (HR2000+, Ocean Optics) measures vegetation radiance on the spectral range 300–900 nm (50μm entrance slit, FWHM∼2 nm).

The optical chain is the same for the three spectrometers: it consists of a collimating lens (74-UV, Ocean Optics), which allows us to adjust the field of view (FOV), fitted to a 5-m-long optic fiber of a 0.22 numerical aperture and a core diameter of 940 μm (SEDI, Courcouronnes, France). In line with the fiber, an electromechanical optical shutter (Micropack, DE) is used to measure the dark level of the charge-coupled device (CCD). The lenses and optical fiber are fixed on a 3-m-long pole oriented downward in a nadir-viewing direction (Fig. 2).

The pole and optical fiber assembly allows the optical head to be positioned far from the instrument box, reducing reflections that could alter the local lighting conditions. For the same reason, the various elements of the head are painted black. At the working height (21 m), the target spot on the ground has a diameter of 2 m. A webcam installed near the optical head records pictures of the target and its surroundings every 5 min during measurements.

To prevent wavelength drifts due to thermal changes, the spectrometers are enclosed in a thermally insulated box and maintained at a constant temperature of 22±0.02 ^◦C. The thermal regulation is achieved by a thermoelectric cooler, controlled by a proportional–integral–differential regulator (MPT5000, Wavelength Electronics). A data acquisition unit (Agilent 34970A) acquires several environmental parameters including photosynthetically active radiation (PAR) measured with a quantum-meter (JYP 1000, SDEC France), temperature of the spectrometers, and the ambient temperature. The whole acquisition system is enclosed in a waterproof aluminum box (Zarges, France), as shown in Fig. 1(d).

All measuring devices including spectrometers and the data acquisition unit are controlled by a Dell Latitude D630 laptop computer. A software written in VB.NET performs automatic data acquisition in the scheduled time window. An Internet connection allows monitoring and control of the instrument, as well as automatic uploading of the data (about 250 Mb/day).

The spectrometers were wavelength calibrated with a spectral lamp (Cal-2000-Bulb, Micropack), also used to determine the spectral instrumental response functions. Radiometric calibration was done using a black body (LI-COR 1800-02, Li-Cor Inc., Lincoln, NE).

Nevertheless, measurements used to retrieve the fluorescence signal come from two different spectrometers, and peri- odic cross-calibration of the spectrometers is necessary. Every 20 min, the reference board is switched from the default position to the calibration position by the means of a rotary solenoid (GDAX035X20E06, Magnet-Schultz). In the calibration position, the reference intercepts the FOV of all spectrometers (Fig. 2). At the end of the day, a linear relationship is deduced from the measurements of vegetation and reference

(4)

TABLE I

CHANNELWAVELENGTH ANDWIDTH OF THETriFLEX INSTRUMENT (FWHM).λ1, . . . , λ4AREUSED TORETRIEVEFLUORESCENCE IN THE

O2-B ABSORPTIONBAND;λ5, . . . , λ7AREUSED TORETRIEVE THE FLUORESCENCE IN THEO2-A ABSORPTIONBAND

Fig. 3. Radiance spectrum acquired on vegetation (sorghum field) at 8 h local time. This spectrum uses the full range of the spectrometer. Vertical lines represent the wavelength positions of the channels used by TriFLEX to compute fluorescence.

spectrometers (Sp2 and Sp3, respectively) and further used for fluorescence computation. Although the square of the correlation coefficient is about 0.999, this cross-calibration is com- puted each day to prevent any possible drift.

To obtain measurements that are comparable with those made by AirFLEX, TriFLEX uses several channels to retrieve the fluorescence. The central wavelengths and widths of these channels are presented in Table I. They have been chosen to optimize the signal-to-noise ratio (SNR) and defined according to the AirFlex instrument. One channel has been added (683 nm) to improve the retrieval of the fluorescence at 687 nm, as discussed later in this paper.

The products extracted from these data are fluorescence in the O2-A and O2-B bands, the normalized difference vegetation index (NDVI), and the PRI.

Dynamic SNR Optimization: Fig. 3 shows the spectrum of the reflected light measured over a sorghum field as seen by the

Fig. 4. Improvement of SNR on vegetation for each channel. SNR is given without optimization (light gray) and with optimization algorithm activated (dark gray).

spectrometer (i.e., not corrected by the instrumental function) using the full range of the spectrometer. It clearly shows that light reflected by vegetation has a high dynamic range, driven by the depth of the O₂-A absorption band, which ranges from 5 at 12:00A.M. to 17 at 9:00A.M.

The noise of the HR2000+ spectrometers (Sp2 and Sp3) has been found constant and independent of the radiance flux. It limits the accuracy of the depth determination, especially in the morning or in the evening, when the in-band absorption is important. As a result, the SNR depends linearly on the measured flux and can be improved by increasing the integration time up to the maximum acceptable by the spectrometer (14 000 counts in practice). The optimum SNR would be obtained when each channel is acquired independently with an integration time corresponding to the full scale of the spectrometer. However, this method is not optimal regarding the repetition rate and the time correlation between channels. An acceptable compromise is to group channels into a small number of amplitude classes and acquire simultaneously all the channels of the same class. Therefore, we developed an algorithm included in the main program to automatically manage the integration time and the number of classes. The optimization algorithm starts by acquiring a first spectrum with no saturated pixels and divides the channels into classes according to their radiance level. Then, for each class, a spectrum is acquired with an integration time that maximizes the radiance of the class’s channel. Saturation of the CCD is allowed for channel in the other classes but is limited to twice the saturation level of the CCD. By repeating this process from the first step, the program adapts automatically for any radiance change. Fig. 4 illustrates the SNR gain obtained with this method. The improvement is about a factor of 4 for O2-B and 12.5 for O2-A. As a result of the optimization algorithm, the frequency of measurement is

>1 Hz when PAR>1000 μmol·m⁻²·s⁻¹ (from 8 to 15 h solar time, during summer days in Avignon).

IV. MEASUREMENTCAMPAIGN

The first measurements using the fluorescence platform were done during the summer 2008. The target was a sorghum field

(5)

[grain sorghum,Sorghum bicolor(L.) Moench cv. Solarius] of about 20 m×20 m, planted on May 16, 2008 (day of year (doy) 137) in one of the eastern fields (Fig. 1). The rows were oriented in a north–south direction, with an interrow distance of 0.45 m and a density of 22.5 plants·m⁻².

Measurements were made on fully developed plants (1.2 m height) from July 31, 2008 (doy 213) to September 6, 2008 (doy 250). To generate physiological changes, the last day of watering was on July 18, 2008 (doy 200). A water stress was expected to develop during the measuring period as a result of the dry summer.

Additional Measurements: SIF emission spectra at the canopy level are extremely difficult to measure in the field because of the presence of sun excitation at the emission wavelengths. As will be discussed further, the knowledge of the shape of the fluorescence emission spectrum is needed to retrieve the SIF level. We used full sun light leaf level emission spectra as a proxy for the canopy level emission. The spectra were acquired with a fluorimeter, called leaf spectrofluorime- ter, already described [11] in which the sun is used as the source of excitation. It is based on a portable spectrometer (HR 2000+, Ocean Optics, Idil, France) equipped with a high- pass red filter (RG665, Schott, France) to select wavelengths only corresponding to Chl fluorescence. In the illumination system, the solar radiation was filtered by a low-pass filter (Corning 4.96, 5 mm) blocking excitation atλ >600 nm. A plano-convex lens focuses the sun on the leaf in such a way that it compensates the light attenuation introduced by the filter and the optics. As a result, measurements were done at full sun light excitation. After fitting the leaf in the fluorimeter, measurements were performed after a light adaptation period of 10–15 min when a stationary state was reached. Raw emission spectra were corrected for the instrumental response function of the spectrometer and for the transmission of the red filter.

Chlorophyll content on sampled leaves of the more outstand- ing plants of the target were measured using a SPAD 502 (Minolta, Ramsey, NJ).

V. RESULTS

Fig. 5 shows an example of the diurnal variation of the radiance fluxes after correction by the integration time and calibration during a day with alternations between clouds and sun spells. To extract fluorescence from the radiance signals by analysis of the in-filling of the atmospheric O2 absorption bands, several parameters should be taken into account including fluorescence and reflectance spectral variations at the vicinity of the oxygen absorption bands and atmospheric corrections due to the air mass between the target on the ground and the reference at 21 m above. Let us introduce first the fluorescence extraction algorithm.

A. Fluorescence Retrieval

In the following, all the targets (i.e., vegetation and reference panel) are considered as Lambertian targets. At the ground level, the vegetation radiance can be described as

L_i= ρ_i×I_i

π +f_i (1)

Fig. 5. Radiance fluxes acquired in the out-band and in-band channel over grain sorghum. (a) O2-A band. (b) O2-B band. (August 1, 2008, doy 214).

where the indexistands for a given wavelength in the vicinity of an oxygen absorption band,ρ_iis the target reflectance,I_iis the sun irradiance, andfiis the SIF.

In the same way, the reference radiance can be written as L_i= ρ_i×Ii

π . (2)

To retrieve I_i, the reference board reflectance ρ_i has been measured in the laboratory. By combining (1) and (2), one can write for each channel the following equation:

Li=ρi×L_i

ρ_i +fi (3)

whereL_iandL_iare measured, andρ_iandf_iare unknowns.

Let us consider first the case of the O₂-A band.

Case of the O2-A Band: It is worth noting that considering (3) for several values of i generates a system ofn equations with2nunknowns, with nbeing the number of channels and which cannot be solved in the general case. We need additional information on the fluorescence and reflectance variation along the absorption band.

Fluorescence model: Fluorescence emission spectra at the canopy level under solar excitation were not available. We used a fluorescence emission spectrum recorded at the leaf level using the setup previously described. Fig. 6 presents two fluorescence emission spectra acquired during two different days. It shows that the slope of the emission spectrum is almost constant in the vicinity of the O2-A band despite the fact that the fluorescence emission spectrum depends on several parameters, including chlorophyll concentration, light intensity, and relative photosystem I/photosystem II contribution. This was previously reported for fluorescence emission spectra from different species in [11].

Several fluorescence emission spectra acquired on representative leaves of the sorghum field were averaged to generate a mean spectrum. Taking the fluorescence value corresponding to

(6)

Fig. 6. Fluorescence emission spectra of adaxial side of a grain sorghum leaf (arbitrary units) and apparent reflectance spectra at the canopy level.

The continuous curve is a fluorescence spectrum acquired on August 30, 2008 (doy 243) at 18 h and 30 min local time (PAR: 764μmol·m⁻²·s⁻¹; chlorophyll content: 54.7 μg·cm⁻²). The dashed curve is a fluorescence spectrum acquired on August 27, 2008 (doy 240) at 17 h local time (PAR:

1244μmol·m₋2·s⁻¹; chlorophyll content: 49.6μg·cm⁻²). Vertical lines represent the channel used by TriFLEX.

the in-band wavelength (760.51 nm), we expressed, by means of a multiplicative coefficient(K_i), the fluorescence measured in an out-band channel(f_i)as a function of the fluorescence in-band(fIN), i.e.,

f_i=K_i×f_IN (4)

where fIN is the fluorescence in the absorption band to be retrieved.

This fluorescence model bringsn−1 equations with zero additional unknowns. Therefore, we obtained the following equation system:

L_i=ρ_i×^L_ρⁱ

i +f_i, i∈ {1, . . . , n}

fi=Ki×fIN, i∈ {1, . . . , n, i=IN}. (5) This equation system is still underdetermined as we have 2n−1equations and2nunknowns.

Reflectance model: The reflectance spectrum of vegetation is continuously generated by the TriFLEX sensor. We observed that at the canopy level, the reflectance spectrum in the vicinity of the O₂-A band is rather linear (Fig. 6). This led us to introduce a linear variation of reflectance and write

ρ(λ) =a+b×λ. (6)

This bringsnequations and two unknowns, namely, aandb.

The equation system becomes

⎧⎨

⎩

Li=ρi×^L_ρⁱ

i +fi, i∈ {1, . . . , n}

f_i=K_i×f_IN, i∈ {1, . . . , n, i=IN} ρ_i(λ_i) =a+b×λ_i, i∈ {1, . . . , n}.

(7)

This latter system has 3n−1 equations and 2n+ 2 unknowns and can be solved ifn= 3. Therefore, we need three channels (one in-band and two out-band) to calculate the flu-

orescence in the O₂-A band, whose wavelengths are given in Table I. The final explicit expression of fluorescence is shown in the Appendix.

Case of the O2-B Band: The method used for fluorescence retrieval in the O2-B is similar to the one employed for O2-A. Although the variability of fluorescence emission spectrum is greater in the red band, we used the same scheme for fluorescence model and defined multiplicative coefficientsK_i based on the mean of the leaf fluorescence emission spectrum measured under sun light illumination. Indeed, in the vicinity of the O2-B band, the measurement channels are spectrally very close. Thus, it seems reasonable to consider that in this range of wavelengths, the shape of the fluorescence spectrum can be well characterized by constant multiplicative coefficientsKi.

The only change is for the reflectance model. The insert in Fig. 6 clearly shows that a linear interpolation fails to fit the shape of the reflectance across the O₂-B absorption band.

Therefore, we consider a parabolic model for the reflectance in the vicinity of the O2-B band, i.e.,

ρ(λ) =a+b×λ+c×λ². (8) This model brings three unknowns andnequations. There- fore, the system for the O₂-B band can be solved ifn= 4(see the expression of the solution in the Appendix).Therefore, four channels are needed to retrieve the fluorescence in the O2-B band. The wavelength positions used are listed in Table I.

B. Atmospheric Correction

Atmospheric corrections are performed to retrieve fluorescence at the ground level. The signal acquired on vegetation is corrected to compensate for the reabsorption of the atmosphere along the path vegetation-sensor, and the signal acquired on the reference board is corrected for the lack of absorption of incident light from the reference altitude to the ground.

The two measured signals were corrected from these effects using MODTRAN 4, which computes the two transmission factors [23]. The variation of fluorescence flux induced by this atmospheric correction is about 5% for the O2-A absorption band and 1.3% in the O2-B band.

C. Measurements

Fig. 7 shows an example of the diurnal variation of fluorescence fluxes over the sorghum field during a day with alternations between clouds and sun spells. For both bands, one may observe the increase of fluorescence flux when light increased, as expected. However, most of the literature related to in vivo ChlF refers to the fluorescence yield (i.e., ratio of the number of photons emitted by fluorescence to the number of absorbed photons). The fluorescence yield is considered as being tightly related to the photosynthetic electron transport rate (see [1] for a review). For a deeper analysis of the fluorescence parameter in passive chlorophyll fluorescence measurements at the canopylevel, it is useful to introduce the fluorescence yield concept, which can be defined as

Φf = Fs

APAR (9)

(7)

Fig. 7. Diurnal cycle of PAR and fluorescence fluxes in both bands (F s687 andF s760) on grain sorghum [August 1 (doy 214)].

Fig. 8. Diurnal cycle of PAR and fluorescence fraction in both bands (F F687 andF F760) on grain sorghum [August 1 (doy 214)].

where APAR stands for the absorbed PAR and is defined as

APAR=fAPAR×PAR (10)

where fAPAR stands for the amount of incident radiation intercepted by green leaves. Although PAR is continuously measured in our experiment, APAR is not accurately known, as it depends on the sun elevation and the 3-D structure of the vegetation. To better account for changes in light interception, we used the radiance flux at 685 nm (border of the O2-B absorption band) as a proxy for APAR. This wavelength is strongly absorbed by vegetation. Consequently, for equivalent illumination, the variation of fluorescence fluxes divided by radiance flux at 685 nm should be mainly linked to the variation of the fluorescence yield.

The fluorescence fraction is defined as the ratio of the fluorescence flux to the target radiance at 685 nm:F F687 = F s(687)/L(685), which represents the amount of fluorescence in the radiance of the vegetation at 685 nm. For the sake of com- parisonF F760 =F s(760)/L(685), but in this case, it is not the amount of fluorescence in the radiance of the vegetation at 760 nm. FF also has the advantages that it uses two signals coming from the same instrument and the same target; as a result, it presents a higher SNR compared to raw fluorescence signals.

Fig. 8 shows a diurnal cycle for both fluorescence fractions measured on August 1, 2008 (doy 214). This day was chosen because of the huge variations of the solar illumination observed due to the presence of clouds. We observed a good correlation betweenF F687andF F760during light changes.

Fig. 9. Time series of parameters acquired on grain sorghum (solar noon, see text). Fluorescence fraction in both bands (F s687andF s760), NDVI, the PRI, the chlorohpyll content ([Chl]), rainfall, and mean PAR.

The two signals were of the same order of magnitude, although F F760> F F687 for the entire day. We also observed that most of the fluorescence fractions variations were negatively correlated with the PAR variations.

To summarize the evolution of the signals during the campaign, fluorescence levels representative of full sunlight conditions have been extracted from diurnal cycles to produce time series of fluorescence. For each measurement day, we selected in the time window going from 11:20 to 12:05, solar time, a subtime window of 1 min in duration that has a mean PAR value of 1900±100 μmol·m⁻²·s⁻¹. In addition to fluorescence fractions, we calculated two vegetation indexes from the measured data, i.e.,

NDVI=ρ(758)−ρ(685)

ρ(758) +ρ(685) (11) PRI=ρ(531)−ρ(570)

ρ(531) +ρ(570). (12) Fig. 9 presents the time series of the fluorescence fractions, PRI, and NDVI together with chlorophyll content, PAR, and rainfall. During the whole campaign, F F687 and F F760 stayed well correlated and tended to decrease with time, except at the end of the measuring period, where both signals increased. Interestingly, PRI was also well correlated with fluorescence fractions, showing also a decrease followed by a marked increase at the end of the measuring period. The NDVI index maintained an almost constant value of 0.8 for

(8)

Fig. 10. Fluorescence flux(F s)versus PAR for three days: 214 no water stress, 243 maximal water stress effect, 248 after rainy days, and reversion of water stress.

the whole campaign, indicating the absence of any major change in the vegetation structure. This observation is supported by the chlorophyll content, which remains at a high value (48 μg·cm⁻²) during the experiment. Rainfall events were rare, except at the end of the measuring period.

As time series showed a static view of the fluorescence properties of the sorghum field, we choose three days that represented the most extreme variation of the fluorescence and PRI parameters. Doy 214 represented the beginning, doy 243 the minimum of both PRI and fluorescence, and doy 248 the recovery at the end of the campaign.

Fig. 10 presents the variations of fluorescence flux (F s) against PAR for these three days. We plotted only the data acquired from 11 to 15 h, solar time. In all cases and for the two bands, we observed an increasing relationship between fluorescence fluxes and PAR, although that for PAR values

>1000μmol·m⁻²·s⁻¹ data are scattered. Doy 214 showed the maximum slope, doy 243 showed the minimal value, and doy 248 showed an incomplete recovery for both bands. A saturation effect for PAR values>1300 is also present for doy 243.

VI. DISCUSSION

The so-called PRI is based on reflectance changes in the green part of the spectrum: PRI= ((ρ(531)− ρ(570)/ρ(531) +ρ(570))[9], where ρ(531)andρ(570)represent the reflectance at 531 and 570 nm, respectively, with the latter wavelength being used as a reference. The physiological bases for changes in PRI began to be investigated in the 1990s. Under laboratory conditions, absorbance and reflectance changes centered near 531–535 nm have been related toΔpH- mediated chloroplast shrinkage and to changes in the aggre- gation state of antenna pigment–protein complexes mediated by an accumulation of deepoxidized forms of the xanthophyll cycle molecules [24], [25]. At the canopy level, rapid vegetation reflectance changes around 531 nm (PRI changes) due to sudden changes in incident light could be sensed remotely and passively using dedicated instruments [10], [26], [27] or a portable radiometer. The resulting changes of PRI were suggested to relate to chloroplast conformational changes asso- ciated withΔpH and the deepoxidation state of the xanthophyll

cycle pigments [28]. Thus, absorbance and reflectance changes around 531 nm have common origins. Evainet al.[26] showed that PRI correlates better with nonphotochemical than with photochemical quenching and was a good indicator of stomata closure upon water shortage.

The TriFLEX sensor was specifically designed to measure, on the same target, the PRI together with other fluorescence parameters. It can be observed in Fig. 9 that PRI decreases about 46% from the first day of measurement to day 243. This decrease is interpreted as an increase of the energy dissipation mechanisms as a result of stomata closure induced by a water stress. The reversion of this decrease after the rainy episode of day 247 strongly supports this view.

Time series of the fluorescence fractions (see Fig. 9) are also in line with the water stress hypothesis. Previous works at the leaf level [6], [29] reported a decrease of the stationary fluorescence at noon when water stress developed after withholding watering. This decrease parallels the decrease of PRI and was also attributed to the increase of nonphotochemical quenching as a result of stomata closure [26]. As the water stress increases, there is a progressive decline of both the maximum fluorescence value achieved and the PAR value at which this maximum is reached. The water stress was also suggested by a moderate leaf rolling detectable by visual inspection (not shown). As for PRI, a reversion of this decrease after the rainy episode of day 247 was also observed. Changes of the relationship between fluorescence flux and PAR shown in Fig. 10 are similar to those reported in [6] and [30]. They also indicate the increase of nonphotochemical quenching for doy 243, compared to the end or the beginning of the campaign.

It is worth noting that neither the chlorophyll content of leaves (almost constant with a value of49.3±2.4 μg·cm⁻²) nor the NVDI indicates a noticeable variation during the measuring campaign. This emphasizes the greater sensitivity of PRI and fluorescence parameters to physiological changes in the sorghum field than chlorophyll or NDVI. This raises the question of the redundancy between PRI and fluorescence.

PRI is strongly correlated with the deepoxidation state of the xanthophyll cycle. Because xanthophyll-related thermal dissipation is often negatively related to leaf photochemistry, PRI has been generally shown in the literature to correlate

(9)

with several fluorescence parameters likeΔF/F m, NPQ, photosynthesis, or stomatal conductance (for a review, see [31]).

However, the PRI does not unequivocally reflect photosynthesis. This is illustrated by the extremely different PRI variations found in the literature (between −0.04 and +0.2) for similar variations in the deepoxidation state of the xanthophylls. It is yet unclear what causes this variation, although it has been shown that PRI depends on the 3-D structure of vegetation [32].

Despite these potential pitfalls, there is a growing interest in PRI as the reflectance parameter that most dynamically tracks photosynthesis. The determination of PRI simply requires reflectance measurements in bands of 5–10 nm wide, which are easy to achieve.

Passive methods to retrieveF srequire reflectance measurements at very high spectral resolution (< 1 nm) and high encoding accuracy. This makesF smore difficult to determine than PRI. With fluorescence being a deactivation process directly in competition with the photochemical conversion, it may appear at first sight more tightly linked with photosynthesis than PRI. This is certainly the case forvariablefluorescence.

However,F sis an extensive parameter that depends on both the fluorescence quantum yield and the absorbed PAR, which is not accurately known at the canopy level. Fluorescence has, however, some specific advantages. First,F s760is not absorbed by the vegetation and can be used to estimate the absorbed PAR as proposed in [5]. Second, by measuring simultaneousF s687 andF s760, one may calculate theF s687/F s760fluorescence ratio, which contains a useful information on the chlorophyll content [8], although as for the PRI, the 3-D canopy structure also affects the fluorescence ratio, as shown in [11].

In our opinion, it is important to measure both PRI (together with the vegetation reflectance spectrum) and passive fluorescence at the same time as these parameters reflects complementary aspect of the vegetation optical response. That is exactly what the new sensor TriFLEX does.

These results have been obtained thanks to a new algorithm to retrieve Chl fluorescence from measurements in some specific channels of the whole radiance spectra. Their number, width, and location were chosen for their pertinence in or out of the oxygen absorption band feature. The algorithm relies on two spectra models, one for fluorescence and the second one for reflectance. It allows one to obtain an explicit solution for the fluorescence without any iteration and does not require explicit computation of reflectance (see the preceding section).

We compared the results obtained with our new algorithm with the predictions using the so-called Plascyk model or FLD method [33]. This work assumed the constancy of both fluorescence and reflectance spectra of the target in the vicinity of the hydrogen absorption bands (HαorHβ). This was valid because of the narrow band(H_α)used. However, making this assumption in the O₂bands leads to inaccurate fluorescence es- timations because the reflectance of vegetation is not constant.

Accounting for the actual shape of the reflectance spectra [34]

improved the Plascyk’s FLD method to retrieve fluorescence in the O2-A band. They introduced a correction coefficient that links in-band reflectance to off-band reflectance using spline interpolation. Although they improved the fluorescence retrieval, this method requires a continuous spectral sampling to interpolate the reflectance along the absorption feature. In this paper, with real data at the canopy level, a linear model of

reflectance was judged satisfactory in the O₂-A band, whereas a parabolic model was suitable for the O₂-B band. Using these fluorescence and reflectance models rather than the usual Plascyk formulas gives a retrieved fluorescence about 45% and 8% lower in the O2-B band and in the O2-A band, respectively.

VII. CONCLUSION

In this paper, we have presented a new crane-based field platform for continuous passive fluorescence measurement at the canopy level. Thanks to the mobile crane, fields having different vegetation types or under different phenologic states can be compared during extended periods of time and under repro- ducible conditions. Furthermore, one of these fields is equipped with an eddy covariance CO₂flux sensor that would allow to compare direct CO2measurements with fluorescence flux.

The crane is equipped with a new passive sensor, namely, TriFLEX, based on three commercial spectrometers, which acquires automatically several vegetation parameters including the reflectance spectrum in the 400–900 nm wavelength range and the SIF flux at 687 and 760 nm. The sensor automatically compensates the wide dynamics range of the vegetation radiance, allowing to optimize the SNR of the fluorescence signal during illumination changes. The nadir-viewing line of sight was chosen to better compare TriFLEX products with airborne fluorescence measurements. It is also worth noting the possibility to install an active fluorescence system (Lidar) in parallel with TriFLEX, on the same frame and measuring on the same target.

During the first test campaign, the ability of both PRI and fluorescence signals to early detect the occurrence of a re- versible water stress of the sorghum field was demonstrated.

Further studies will focus on the role of 3-D structure on the fluorescence flux, together with the study of bidirectional effects on both sun illumination and direction of observation.

APPENDIX

A. Fluorescence Fluxes Expression

O₂-B Band: λ₃is in the absorption band,λ₁andλ₂are off bands (left),λ4is an off band (right) (see Table I). We have

F687 = N

D (13)

whereN andDare defined in (14) and (15), shown at the top of the next page.

O2-A Band: λ6, is in the absorption band,λ5is an off band (left), λ7 is an off band (right) (see Table I). We have (16), shown at the top of the next page.

B. Fluorescence Fraction

The fluorescence fraction is defined as F F = F

L(685). (17)

Thus,F F is the part of fluorescence in the radiance signal coming from the vegetation, including the fluorescence signal itself. It is possible to choose as reference a wavelength for which there is no fluorescence emission; this will give us a parameter that represents the fluorescence fraction in the

(10)

N =L3L₁L₂L₄ρ₃(λ1−λ2)(λ1−λ4)(λ2−λ4) +L₃(−λ4L₁L₂ρ₄(λ1−λ2)(λ1−λ3)(λ2−λ3) +L₄(λ3−λ4)

×(L₁L₂ρ₁(λ₂−λ₃)(λ₂−λ₄) +L₂L₁ρ₂(λ₁−λ₃)(λ₄−λ₁))) (14) D=L₁(L₂(λ1−λ2) (K4L₃ρ₄(λ1−λ3)(λ3−λ2) +L₄ρ₃(λ1−λ4)(λ2−λ4))

−K2L₃L₄ρ₂(λ1−λ3)(λ1−λ4)(λ3−λ4)) +K1L₂L₃L₄ρ₁(λ2−λ3)(λ2−λ4)(λ3−λ4) (15)

F760 = L₆(L7L₅ρ₇(λ5−λ6)) +L5L₇ρ₅(λ6−λ7) +L6L₅L₇ρ₆(λ7−λ5)

K₅L₆L₇ρ₅(λ₆−λ₇) +L₅(K₇L₆ρ₇(λ₅−λ₆) +L₇ρ₆(λ₇−λ₅)) (16)

reflected signal (F F R). However, it is worth noting that this parameter can easily be deduced fromF F: the ratio between fluorescence and reflected signal is directly linked to F F through the following relationship:

F F R(λ) = F(λ)

I(685)×ρ(685) = F F(λ)

1−F F687. (18) In practice,F F RandF F differs by about 10% at maximum.

ACKNOWLEDGMENT

The authors would like to thank the Unité Expérimentale Environnement et Agronomie d’Avignon for providing the sorghum field and G. Parlant and D. Parlant for their helpful comments and grammar corrections.

REFERENCES

[1] G. C. Papageorgiou and Govindjee,Chlorophyll a Fluorescence: A Sig- nature of Photosynthesis. Dordrecht, The Netherlands: Springer-Verlag, 2004.

[2] F. Link, “Variations lumineuses de la Lune,”Bull. Central Astron. Inst.

Czechoslovakia, vol. 2, pp. 131–133, 1951.

[3] J. A. Plascyk and F. C. Gabriel, “The Fraunhofer line discriminator MKII—An airborne instrument for precise and standardized ecological luminescence measurement,”IEEE Trans. Instrum. Meas., vol. IM-24, no. 4, pp. 306–313, Dec. 1975.

[4] I. Moya, L. Camenen, G. Latouche, C. Mauxion, S. Evain, and Z. G. Cerovic, “An instrument for the measurement of sunlight excited plant fluorescence,” inProc. XIth Int. Congr. Photosynth., G. Garab, Ed., 1999, pp. 4265–4370.

[5] I. Moya, L. Camenen, S. Evain, Y. Goulas, Z. G. Cerovic, G. Latouche, J. Flexas, and A. Ounis, “A new instrument for passive remote sensing—Part 1: Measurements of sunlight-induced chlorophyll fluorescence,”Remote Sens. Environ., vol. 91, no. 2, pp. 186–197, May 2004.

[6] J. Flexas, J. M. Briantais, Z. Cerovic, H. Medrano, and I. Moya, “Steady- state and maximum chlorophyll fluorescence responses to water stress in grapevine leaves: A new remote sensing system,”Remote Sens. Environ., vol. 73, no. 3, pp. 283–297, Sep. 2000.

[7] S. Evain, L. Camenen, and I. Moya, “Three channels detector for remote sensing of chlorophyll fluorescence and reflectance from vegetation,” in Proc. 8th Int. Symp.: Phys. Meas. Signatures Remote Sens., M. Leroy, Ed., Aussois, France, 2001, pp. 395–400.

[8] H. K. Lichtenthaler, R. Hak, and U. Rinderle, “The chlorophyll fluorescence ratio F690/F730 in leaves of different chlorophyll content,”Photo- synth. Res., vol. 25, no. 3, pp. 295–298, Sep. 1990.

[9] J. A. Gamon, J. Penuelas, and C. B. Field, “A narrow-waveband spectral index that tracks diurnal changes in photosynthetic efficiency,”Remote Sens. Environ., vol. 41, no. 1, pp. 35–44, Jul. 1992.

[10] J. Louis, A. Ounis, J. M. Ducruet, S. Evain, T. Laurila, T. Thum, M. Aurela, G. Wingsle, L. Alonso, R. Pedros, and I. Moya, “Remote sensing of sunlight-induced chlorophyll fluorescence and reflectance of

Scots pine in the boreal forest during spring recovery,” Remote Sens.

Environ., vol. 96, no. 1, pp. 37–48, May 2005.

[11] I. Moya, F. Daumard, N. Moïse, A. Ounis, and Y. Goulas, “First airborne multiwavelength passive chlorophyll fluorescence measurements over La Mancha (Spain) fields,” inProc. 2nd Int. Symp. Recent Adv. Quan- titative Remote Sens., Torrent, Spain, 2006, pp. 820–825.

[12] I. Moya, F. Daumard, N. Moise, A. Ounis, and Y. Goulas, “Sun-induced chlorophyll fluorescence at 687 nm and 760 nm: Comparison of airborne and ground measurements,”Remote Sens. Environ., submitted for publication.

[13] U. Rascher, G. Agati, L. Alonso, G. Cecchi, S. Champagne, R. Colombo, A. Damm, F. Daumard, E. de Miguel, G. Fernandez, B. Franch, J. Franke, C. Gerbig, B. Gioli, J. A. Gómez, Y. Goulas, L. Guanter, Ó. Gutiérrez-de-la-Cámara, K. Hamdi, P. Hostert, M. Jiménez, M. Kosvancova, D. Lognoli, M. Meroni, F. Miglietta, A. Moersch, J. Moreno, I. Moya, B. Neininger, A. Okujeni, A. Ounis, L. Palombi, V. Raimondi, A. Schickling, J. A. Sobrino, M. Stellmes, G. Toci, P. Toscano, T. Udelhoven, S. van der Linden, and A. Zaldei, “CEFLES2:

The remote sensing component to quantify photosynthetic efficiency from the leaf to the region by measuring sun-induced fluorescence in the oxygen absorption bands,”Biogeosci. Discuss., vol. 6, no. 1, pp. 2217–2266, 2009.

[14] P. Zarco-Tejada, L. Suarez, J. Berni, G. Sepulcre-Canto, F. Morales, and J. Miller, “Imaging chlorophyll fluorescence with an airborne narrowband multispectral camera for vegetation stress detection,”Remote Sens.

Environ., vol. 113, no. 6, pp. 1262–1275, Jun. 2009.

[15] J. Berni, P. Zarco-Tejada, L. Suarez, and E. Fereres, “Thermal and narrowband multispectral remote sensing for vegetation monitoring from an unmanned aerial vehicle,”IEEE Trans. Geosci. Remote Sens., vol. 47, no. 3, pp. 722–738, Mar. 2009.

[16] O. Perez-Priego, P. J. Zarco-Tejada, J. R. Miller, G. Sepulcre-Canto, and E. Fereres, “Detection of water stress in orchard trees with a high- resolution spectrometer through chlorophyll fluorescence in-filling of the O2-A band,”IEEE Trans. Geosci. Remote Sens., vol. 43, no. 12, pp. 2860–

2869, Dec. 2005.

[17] E. Middleton, L. Corp, C. Daughtry, and P. Campbell, “Chlorophyll fluorescence emissions of vegetation canopies from high resolution field reflectance spectra,” inProc. IGARSS, 2006, pp. 4064–4067.

[18] L. A. Corp, E. M. Middleton, J. E. McMurtrey, P. K. Campbell, and L. M. Butcher, “Fluorescence sensing techniques for vegetation assessment,”Appl. Opt., vol. 45, no. 5, pp. 1023–1033, Feb. 2006.

[19] M. Meroni and R. Colombo, “Leaf level detection of solar induced chlorophyll fluorescence by means of a subnanometer resolution spec- troradiometer,” Remote Sens. Environ., vol. 103, no. 4, pp. 438–448, Aug. 2006.

[20] M. Meroni, M. Rossini, V. Picchi, C. Panigada, S. Cogliati, C. Nali, and R. Colombo, “Assessing steady-state fluorescence and PRI from hyper- spectral proximal sensing as early indicators of plant stress: The case of ozone exposure,”Sensors, vol. 8, no. 3, pp. 1740–1754, 2008.

[21] M. Meroni, C. Panigada, M. Rossini, V. Picchi, S. Cogliati, and R. Colombo, “Using optical remote sensing techniques to track the development of ozone-induced stress,” Environ. Pollut., vol. 157, no. 5, pp. 1413–1420, May 2009.

[22] L. Prévot, I. Champion, and G. Guyot, “Estimating surface soil moisture and leaf area index of a wheat canopy using a dual-frequency (C and X bands) scatterometer,”Remote Sens. Environ., vol. 46, no. 3, pp. 331–

339, Dec. 1993.

(11)

[23] F. Daumard, Y. Goulas, A. Ounis, R. Pedros, and I. Moya, “Atmospheric correction of airborne passive measurements of fluorescence,” inProc.

ISPMSRS, Davos, Switzerland, 2007.

[24] F. Morales, A. Abadia, and J. Abadia, “Characterization of the xanthophyll cycle and other photosynthetic pigment changes induced by iron deficiency in sugar beet (Beta vulgarisL.),”Plant Physiol., vol. 94, no. 2, pp. 607–613, Oct. 1990.

[25] A. V. Ruban, A. J. Young, and P. Horton, “Induction of nonphotochemical energy dissipation and absorbance changes in leaves (Evidence for changes in the state of the light-harvesting system of photosystem II in vivo),”Plant Physiol., vol. 102, no. 3, pp. 741–750, Jul. 1993.

[26] S. Evain, J. Flexas, and I. Moya, “A new instrument for passive remote sensing—Part 2: Measurement of leaf and canopy reflectance changes at 531 nm and their relationship with photosynthesis and chlorophyll fluorescence,” Remote Sens. Environ., vol. 91, no. 2, pp. 175–185, May 2004.

[27] J. Peguero-Pina, F. Morales, J. Flexas, E. Gil-Pelegrin, and I. Moya,

“Photochemistry, remotely sensed physiological reflectance index (PRI) and de-epoxidation state of the xanthophyll cycle inQuercus coccifera under intense drought,”Oecologia, vol. 156, no. 1, pp. 1–11, May 2008.

[28] J. A. Gamon, C. B. Field, W. Bilger, O. Bjorkman, A. L. Fredeen, and J. Penuelas, “Remote sensing of the xanthophyll cycle and chlorophyll fluorescence in sunflower leaves and canopies,”Oecologia, vol. 85, no. 1, pp. 1–7, Nov. 1990.

[29] J. Flexas, J. M. Escalona, S. Evain, J. Gulias, I. Moya, C. B. Osmond, and H. Medrano, “Steady-state chlorophyll fluorescence (Fs) measurements as a tool to follow variations of net CO2assimilation and stomatal conductance during water-stress in C-3 plants,”Physiologia Plantarum, vol. 114, no. 2, pp. 231–240, Feb. 2002.

[30] Z. G. Cerovic, Y. Goulas, M. Gorbunov, J. M. Briantais, L. Camenen, and I. Moya, “Fluorosensing of water stress in plants: Diurnal changes of the mean lifetime and yield of chlorophyll fluorescence, measured simultaneously and at distance with aτ-LIDAR and a modified PAM- fluorimeter, in maize, sugar beet, and kalanchoe,”Remote Sens. Environ., vol. 58, no. 3, pp. 311–321, Dec. 1996.

[31] I. Moya and J. Flexas, “Terrestrial photosynthesis in a changing environment. The molecular, physiological and ecological bases of photosynthesis driving its response to the environmental changes,” inRemote Sensing of Photosynthesis, J. Flexas, F. Loreto, and H. Medrano, Eds.

Cambridge, U.K.: Cambridge Univ. Press, 2010.

[32] L. Suarez, P. Zarco-Tejada, G. Sepulcre-Canto, O. Pérez-Priego, J. Miller, J. Jiménez-Muñoz, and J. Sobrino, “Assessing canopy PRI for water stress detection with diurnal airborne imagery,”Remote Sens. Environ., vol. 112, no. 2, pp. 560–575, Feb. 2008.

[33] J. A. Plascyk, “The MK II Fraunhofer line discriminator (FLD-II) for airborne and orbital remote sensing of solar-stimulated luminescence,”

Opt. Eng., vol. 14, pp. 339–346, 1975.

[34] L. Alonso, L. Gomez-Chova, J. Vila-Frances, J. Amoros-Lopez, L. Guanter, J. Calpe, and J. Moreno, “Improved Fraunhofer line discrim- ination method for vegetation fluorescence quantification,”IEEE Geosci.

Remote Sens. Lett., vol. 5, no. 4, pp. 620–624, Oct. 2008.

Fabrice Daumardreceived the M.Sc. degree in remote sensing from the University of Paris VI, Paris, France, in 2006. He is currently working toward the Ph.D. degree at the Laboratoire de Météorologie Dynamique, Centre National de la Recherche Scien- tifique, Ecole Polytechnique, Palaiseau, France.

His main research interests include the development of passive remote sensing techniques for chlorophyll fluorescence measurements at the ground level and with airborne sensors. He also works on atmospherics correction needed to retrieve fluorescence from airborne measurements.

Sébastien Champagne received the Eng. degree in optronics from Ecole Nationale Supérieure des Sciences Appliquées et de Technologie, Lannion, France, in 2006.

He is currently an Engineer with the Labora- toire de Météorologie Dynamique, Centre National de la Recherche Scientifique, Ecole Polytechnique, Palaiseau, France. His research interests include optical sensor and remote sensing applications.

Antoine Fournier received the Eng. degree from the Ecole Nationale Supérieure des Sciences Ap- pliquées et de Technologie, Lannion, France, in 2007 and the M.Sc. degree in photonics and telecommu- nications optics from the University of Rennes1, Rennes, France, in 2007. Since 2008, he has been working toward the Ph.D. degree at the Labora- toire de Météorologie Dynamique, Centre National de la Recherche Scientifique, Ecole Polytechnique, Palaiseau, France, working on remote sensing of chlorophyll fluorescence.

Yves Goulas received the Eng. degree from the Ecole Centrale de Paris, Châtenay Malabry, France, in 1984 and the Ph.D. degree in applied physics from the University of Paris XI, Orsay, France, in 1992.

He is currently with the Laboratoire de Météorolo- gie Dynamique, Centre National de la Recherche Scientifique, Palaiseau, France. He has developed several instruments in the domain and is also in- volved in modeling fluorescence properties of leaves and canopies. His main research interests include the utilization of optical signals from vegetation, particularly chlorophyll fluorescence, to monitor plant physiology.

Abderrahmane Ounisreceived the Eng. degree in electronics from the Ecole Nationale Polytechnique, Algiers, Algeria, in 1989. He received the M.Sc.

degree in laser physics and applications from the University of Paris XIII, France, in 1995, and the Ph.D. degree in optics and photonics from the Uni- versity of Paris XI, Orsay, France, in 2001.

Since 1999, he has been working as an Engineer at the Centre National de la Recherche Scientifique (CNRS), France. He has a permanent position at the Laboratoire de Météorologie Dynamique, CNRS, Ecole Polytechnique, Palaiseau, France. He has developed several dedicated instrumentation in this field, particularly for airborne measurements. His main research interests include research and development in passive and lidar remote sensing of vegetation fluorescence and reflectance.

Jean-François Hanocqreceived the Ph.D. degree from the Faculté des Sciences et des Techniques de St Jerôme, Marseille, France, 1991.

He is currently with the Institut National de Recherche Agronomique, Avignon, France. He is in charge of radiometric measurements at ground level and airborne.

Ismaël Moyareceived the M.Sc. degree in mathe- matics and the Ph.D. degree in biophysics from the University of Paris XI, Orsay, France, in 1968 and 1979, respectively.

In 1972, he joined the Centre National de la Recherche Scientifique, Ecole Polytechnique, Palaiseau, France, where he is currently with the Laboratoire de Météorologie Dynamique. His research interests include fluorescence remote sensing using both active (lidars) and passive techniques.

He has developed several ground-based sensors and airborne sensors based on the filling-in of the atmospheric oxygen absorption bands.