NIR spectroscopy is a useful tool for many applica-tions in diagnostic and biochemical estimation onplant tissues [1]. Today, for various application do-mains such as precision agriculture, global-scaleecology [2], and the validation of satellite remote sen-sing products [3], it is very important to assess theleaf biochemical composition: water content (Cw, alsocalled equivalent water thickness) [4], dry mattercontent (Cm, also called leaf mass per area), and totalchlorophyll content (Cab). Estimations of these bio-chemical compounds are usually performed by labo-ratory procedures, and most of them are lengthy,complex, and destructive. For these reasons thereis a necessity for new procedures that make leaf bio-chemical measurements in situ possible.
Several techniques have been explored for thispurpose, based on the interaction of electromagneticradiation in the internal components of a plant leaf[5,6]. These physical interaction models take into ac-count transmittance, reflectance, and fluorescentmeasurements on the vegetation tissue. This ideasets a practical method to estimate leaf biochemicallevels in a nondestructive way [7]. For example, theestimation of total chlorophyll on a plant leaf, whichnormally involves analytical chemical techniques,can be measured by instruments that exist today:the SPAD-502 (Konica Minolta Sensing Inc, Osaka,Japan) [8,9] and the CCM-200 (ADC BioScientificLtd, Hertford, UK) [9]. Those instruments use se-lected wavelengths from the visible domain to com-pute relative measurements of chlorophyll. Othersinstruments such the Dualex FLAV and the DualexHCA, which have been developed by FORCE-A (Or-say, France) [10] and are based on the fluorescenceemission by chlorophyll after exposure to ultraviolet
(UV) radiation, give relative measurements ofchlorophyll and analytical measurements of otherbiochemical compounds. Our main purpose is to pre-sent an instrument that can perform the analyticalestimation of leaf water content via biophotonic mea-surements in situ, using a nondestructive techniquethat can be easily applied.To achieve the goal of the estimation of water con-
tent, an inversion procedure of the radiative transfermodel PROSPECT is performed on the RAMIS mea-surements.This work involves two research activities: instru-
mentation and model inversion. The instrumenta-tion part describes the RAMIS prototype, includinga quick description of the source system, sensor sys-tem, and the signal acquisition procedure. Next, theinversion procedure of the PROSPECT model is de-scribed and how it is adapted to the output data fromthe RAMIS system is described. Last, the result ofwater contents measured from several leaf samplesare compared with the computed water contentsfrom the inversion procedure.
2. Physical Principles
The interaction of electromagnetic radiation withplant leaves can be computed from knowledge ofthe spectral variation of the complex refractive index[7,11]. This phenomenon is directly related to differ-ent processes such as the transmission, reflectionand absorption of light, and electronic orbital inter-actions with the leaf structure in the visible and in-frared domain. The absorption phenomenon allowsthe definition of a spectrometric method to performanalytical measurements of the principal leaf bio-
chemical compounds. In order to accomplish thisobjective, it is essential to know the spectral absorp-tion characteristics of each compound: water, chloro-phyll, and dry matter. Figure 1 shows the specificabsorption coefficients of these three important bio-chemical components.
The spectral transmittance and reflectance of aplant leaf can be predicted from the spectral res-ponse of its biochemical constituents (Fig. 1) andthe behavior of light in the internal leaf structure.This relationship between the spectral leaf transmit-tance–reflectance and the concentration of biochem-ical compounds presents a useful tool to relateexternal optical measurements to its internal compo-nents suchaswater, drymatter, and chlorophyll (prin-cipally A and B). Therefore, a spectrometric methodcould be implemented in a portable device in orderto perform in situ transmittancemeasurements usinga selected set of wavelengths. More specifically, thewavelengths must be selected as a function of the ap-propriate biochemical leaf compounds consideringtheir sensitivity response at their respective concen-tration levels.
All these characteristics have been implementedin the development process of the RAMIS prototype.In addition, the relationship between optical mea-surements and chemical species concentrations hasbeen established by comparison with regular labora-tory measurement methods and by inversion of thePROSPECT model.
A. Radiative Transfer Model PROSPECT
The radiative transfer model PROSPECT [12] isan optical model of a plant leaf that is capable of
Fig. 1. (Color online) Specific absorption coefficients of water, chlorophyll, and dry matter determined after a chemical extraction [14].
computing its spectral transmittance and reflectancefrom 400nm to 2500nm. PROSPECT is based on adielectric model of a plant leaf that could be com-posed of one or several layers and is able to accountfor multiple internal reflections within them [13].Furthermore, the attenuation of electromagnetic ra-diation due to spectral absorption by biochemicalcompounds is considered for each internal reflection.For this reason, it is important to know the spectralrefractive index of leaf material and the specific ab-sorption coefficients for each biochemical compound[14]: chlorophyll, water, and dry matter (Fig. 1).Finally, the formalism could be extended to a contin-uous layer [15], with the possibility of describing dif-ferent structures of plant leaf mesophyll, namely,monocotyledon leaves (unique compact layer) anddicotyledon leaves (several layers).As a result, the radiative transfer model PRO-
SPECT incorporates three biochemical parameters:chlorophyll content Cab, water content Cw, dry mat-ter content Cm, plus an additional parameter N, tosimulate spectral transmittance and reflectance.The parameterN, called also parameter of structure,describes the mesophyll structure of a plant leaf andit could be used to differentiate one specie fromanother [16,17].For this reason, PROSPECT can be used as a for-
ward model that allows us to apply inverse theorymodeling [18]. It can be performed in order to esti-mate all parameters or a selected biochemical para-meter from the transmittance and the reflectancemeasured by RAMIS.
3. Description of RAMIS
The prototype RAMIS [19] consists principally of twofunctional parts: a source system of photonic radia-tion to illuminate the adaxial faces of the leaf anda photonic sensor system with the function of per-forming transmittance measurements.
A. Source System
In order to justify the spectral band position of theRAMIS source system, several transmittance spectrawere computed using PROSPECT, varying each timeone of the biochemical parameters (Fig. 2). This in-dicates where the spectral sensitive zones are locatedin relation with each biochemical leaf compound.Consequently, we are able to determine the optimalwavelength positions on the spectral transmittance.However, some of the ideal wavelengths were shiftedto other values as a result of limitations on the nom-inal values of LED wavelengths available in the mar-ket. The bands chosen for the source system arecentered on 656, 721, 843, 937, and 1550nm.
However, in our application of determination ofwater and dry matter contents, it is just necessaryto use 843, 937, and 1550nm wavebands. The cap-ability of RAMIS to determine leaf chlorophyll con-tent, using the wavelengths of 656nm and 721nm,will not be discussed in the present paper.
Light emission diodes (LEDs) (Roithner Lasertech-nik, Vienna, Austria) were used with a characteristicband-centered spectrum of nominal values (Fig. 2).Their respective spectral emissions were measuredin order to verify and use them in the adaptation
Fig. 2. (Color online) Leaf spectral transmittance predicted using the PROSPECTmodel in superposition with the source spectrum fromthe RAMIS prototype. These simulated variations of the spectral transmittance are observed as a function of the principal biochemicalcompound: (a) water (Cw), (b) dry matter (Cm), (c) chlorophyll AB (Cab), and (b) the leaf parameter structure (N). The arrow shows theparameter increment variation.
of the PROSPECT model. Finally, they are arrangedin a compact formation (Fig. 3) to obtain an approx-imation of an incident light source that illuminatesthe plant leaf sample.
B. Detector System
The RAMIS detector system is formed by a Si/Gedouble layer photodiode (Judson Technologies, Phila-delphia, Pennsylvania, USA) designed with a respon-sivity range of 500 to 2000nm. It is composed of twospectral zones, situated between 500 and 1200nm,corresponding to the photodiode layer of Si that itis sensitive to the spectral emission of 843 and937nm photodiodes. The second zone is situated be-tween 1000 and 2000nm, corresponding to the photo-diode layer of Ge, and is sensitive at 1550nm (Fig. 4).The mechanical mount of the Si/Ge double layer
photodiode sensor is a cylindrical compartment withthe sensor situated at the center, aimed toward the
source system on the opposite jaw of the RAMISprototype (Fig. 5). The prototype is assembled in apincer configuration, and the sample of plant leafis clamped between the photonic source and sensor.
The electronic systems that drive the source andthe sensor are placed in each respective jaw. Thesource controller circuit is composed of a current dri-ver circuit with the functionality to activate eachphotodiode independently and to control the currentstability of each one. This circuit is controlled by adata acquisition universal serial bus (USB) moduleusing its digital input/output (I/O) parallel port.The activation procedure of each photodiode is con-trolled by the state of the digital I/O port bits; in thiscase the source is controlled by five bits, one bit foreach photodiode.
For the sensor, a current-to-voltage circuit andan amplifier stage were implemented with thefunctionality to produce an output voltage level V ,
Fig. 3. (Color online) Source system of the RAMIS prototype con-formed of five LEDs: (a) transversal view, (b) top view of the closecompartment to avoid external light interference in the measure-ment procedure, and (c) mechanical support for the source.
Fig. 5. (Color online) Detector system of the RAMIS prototypecomposed of Si/Ge double layer photodiode sensor: (a) transversalview, (b) top view of mechanical configuration, and (c) mechanicalsupport for the photodiode.
proportional to intensity I of the radiation incidenton the sensor. Two circuits were implemented, onefor each channel of the Si/Ge photodiode. These twooutput voltage lines are connected to the analog-to-digital converter (ADC) port of the USB data acquisi-tion module. The retrieved values are recorded onthe computer that controls the data acquisitionUSB module (Fig. 6).The procedure to perform the measurement con-
sists of activating each photodiode for an activationtime ton. This time ton should be small enough toavoid an increase of the temperature of each photo-diode; consequentially it could affects their radiatedpower while the measurement procedure is takingplace. At the same time the ADC port acquires thesignal for a sampling time ts. The sampling time tsis shorter than the activation time ton and centeredto avoid the rising and the falling edges of the activa-tion signal (Fig. 7).
4. Signal Treatment
In order to achieve transmittance measurementswith the RAMIS prototype, the transmittance mea-surement intensity I0 must be defined as a measureperformed by the source and detector system withoutany leaf sample inside. This intensity I0 is trans-
duced as a proportional output voltage V0 from theelectronic circuit of the sensor. If the air attenuationof the source signal is neglected, then the intensity I0serves as a reference signal that is related to a trans-mittance equal to unity. However, because such anintensity is a function of the photodiode of the sourcesystem (843, 937, and 1550nm), it is important todefine the reference intensity I0ðλC;iÞ, where thewavebands are centered at each specific λC;i. Conse-quently, we must define a sequential activation foreach diode of the source system.
After I0ðλC;iÞ has been determined, the system isready to measure the signal intensity of incident ra-diation from the source, with a plant leaf sampleplaced between the source and sensor system. Thesame sequential procedure of LED activations mustbe used at this point also, due to the monochannelcharacteristic of the Si/Ge sensor in the wavelengthrange of 500 to 1200nm, where all spectral LEDemissions are situated. The intensities IðλC;iÞ arethen defined for each specific λC;i from the source thatgoes through the plant leaf sample (Fig. 8). In thesame way as I0ðλC;iÞ, the intensities IðλC;iÞ are trans-duced as proportional output voltageV for each diodewavelength λC;i.
At this point, with the measurements of I0ðλC;iÞand IðλC;iÞ, the transmittance measured by theRAMIS system can be defined for each photodiodewaveband as follows:
TRðλC;iÞ ¼VðλC;iÞV0ðλC;iÞ
¼ IðλC;iÞI0ðλC;iÞ
: ð1Þ
It is important to note here that the Si/Ge sensorgives a response signal that is proportional to the in-cident intensity I, integrated along the wavebandcentered at a specific λC;i. This aspect is consideredin computing the plant leaf transmittance in theradiative spectral model. Furthermore, the spectralresponsivity of the Si/Ge detector is not constant
Fig. 6. (Color online) Schematic configuration of the RAMIS system.
Fig. 7. Synchronization of the activation time ton and samplingtime ts for the source and the sensor system.
through the interval of 500 to 2000nm and thespectral emissions of each LED have a characteristicshape (Fig. 9). Last, the transmittance measuredby RAMIS is not a hemispherical quantity. Forthese reasons, two important procedures must becarried out: spectral treatment and hemisphericalcorrection.
A. Spectral Treatment
Based on the characteristics of the photonic compo-nents used in the RAMIS prototype, and considering
the theoretical method that describes the phenom-ena of leaf light transmittance, it is necessary to con-sider the spectral response of the sensor and thespectral emission of the source. For this reason,the spectral emission of each light-emitting diodeis defined as Eðλ; λC;iÞ, where its characteristic spec-trum is centered at a wavelength λC;i (Fig. 9). We alsodefine SðλÞ as the spectral responsivity of the Si/Gedouble layer photodiode (Fig. 4).
Next, the hemispherical transmittance spectrumof a plant leaf is defined as THðλÞ, defined for the
Fig. 8. (Color online) Schematic representation of the reference I0ðλC;iÞ and the sample IðλC;iÞ measurements performed by the RAMISprototype.
Fig. 9. (Color online) Measured spectral emission of 1550nm LED used in the source prototype RAMIS system.
500 to 2000nm wavelength interval. It is then possi-ble to express the intensity measured by the RAMISprototype by only regarding the spectral behavior ofthe photonic components [20]. The intensity mea-sured by RAMIS can be written as
IHðλC;iÞ ¼ kZ
∞
0Eðλ; λC;iÞTHðλÞSðλÞdλ; ð2Þ
where the intensity IHðλC;iÞ is a function of the sourcespectral band centered at λC;i (Fig. 9) and k is a factorof proportionality. The case where there is no mate-rial between the source and sensor must be consid-ered. Here, the spectral transmittance THðλÞ shouldbe equal to unity for all λ. The reference intensity canthen be defined as
I0;HðλC;iÞ ¼ kZ
∞
0Eðλ; λC;iÞSðλÞdλ; ð3Þ
assuming provisionally that the totality of the trans-mitted light enters into the detector. Finally, usingEqs. (2) and (3), the hemispherical transmittanceis defined as
TRHðλC;iÞ ¼
IHðλC;iÞI0;HðλC;iÞ
¼R∞
0 Eðλ; λC;iÞTHðλÞSðλÞdλR∞
0 Eðλ; λC;iÞSðλÞdλ: ð4Þ
This expression represents the hemispherical trans-mittance for each photodiode wavelength λC;i imple-mented into the RAMIS prototype.
B. Hemispherical Correction
Equation (4) only describes the spectral characteris-ticsofan idealhemispherical sensorwithsimilarspec-tral characteristics to those of the RAMIS Si/Gesensor. The idea implicit in this formulation is to havea relationship between RAMIS measurements and aphysical model of transmittance. In this case, we areinterested in the implementation of the radiativetransfer model PROSPECT. However, PROSPECTcomputes hemispherical transmittances, which aregeometrically different from the transmittancesmea-suredbyRAMIS.For this reason, it isnecessary toper-form a geometrical correction, specifically, to find arelationship between the RAMIS transmittance andthe computed spectral hemispherical transmittance.For this reason, reference measurements of trans-
mittancewere performed, especially of hemisphericalleaf transmittance. For this purpose, a spectrometer(Analytical Spectral Device ASD from FieldSpec-FR)with an integrating sphere (LICOR) was used to per-form the reference measurements of hemisphericaltransmittance.In this case, the hemispherical correction can be
described, and it is possible to find a relationshipbetween the transmittance measurement of the RA-MIS prototype and the hemispherical measure-ments. Using Eq. (1) as a definition of RAMIStransmittance and regarding the Lambertian char-acteristic of the sources, we define
TRHðλC;iÞ ¼ ξλC;iTRðλC;iÞ; ð5Þ
where TRðλC;iÞ is the transmittance measured by RA-MIS, TR
HðλC;iÞ is the hemispherical transmittancemeasured by the ASD, and ξλC;i is the correction coef-ficient associated with the geometrical differencesbetween RAMIS and the ASD system.
It is observed that the correction parameter ξλC;idoes not depend strongly on the sampled leaf species;they have thereafter been assumed independentof these.
5. Material and Methods
Two electronic systems are incorporated into RA-MIS. The first drives each LED sequentially, whilethe second makes a current-to-voltage conversionof the output signal from the detector. Both electronicsystems work in synchrony during the measurementprocedure.
In order to have a portable instrument, an auto-matic data acquisition module was incorporated intothe electronic system of the prototype. This consistsof a NIUSB-6009module fromNational Instrumentswith the following characteristics: a 14 bit ADC inputand I/O digital ports. Thus, we can control the LEDactivation and the data acquisition of the prototypeelectronic system. Consequently, this module enablesthe connection of RAMIS to a laptop and uses any Gprogramming language such as LabView to controlthe acquisition and processing data.
The measurement procedure on a plant leaf sam-ple consists in performing two transmittance mea-surements. The first one, a measurement withouta plant leaf sample, determines the reference inten-sity I0ðλC;iÞ. The second determines the intensity thatpasses through the plant leaf sample IðλC;iÞ. Thetransmittance measured by RAMIS can then be de-fined by Eq. (1). The distance between the sampleand the sensor must be constant for all measuredsamples, as well as the distance between the sampleand the source.
The objective of the experiment was to verify theperformance of the RAMIS prototype for estimatingwater and dry matter contents. In order to validateits performance, a comparison was made betweenRAMIS estimations and the values of water anddry matter contents measured by a direct procedureof the drying process. A variety of plant leaves wasused Table 1.
The experimental water thickness (Cw) of eachplant leaf sample can be defined as the water contentper unit leaf surface, denoted as
Cw ¼ M −Mdry
S; ð6Þ
where M is the mass of the sample leaf when thetransmittance measurement is performed, Mdry isthe dry mass of the leaf at the end of the drying pro-cess, and S is the area of the leaf (in cm2). Using thedensity of water, it is possible to express Cw as a
water thickness that it is related to an equivalent op-tical thickness of water.At the same time the mass M is measured, the
measurements of IðλC;iÞ and I0ðλC;iÞ were performedby RAMIS, and finally the transmittance TRðλC;iÞwas computed using Eq. (1) for each wavelength.
6. Results and Discussion
The respective values of measured transmittancesTRðλC;iÞ are denoted as the input parameters forthe inversion procedure.Into the inversion procedure, the source spectral
emission Eðλ; λC;iÞ and the sensor spectral responsiv-
ity SðλÞ are included for PROSPECT model adapta-tion. Then, the output parameters of the inversionprocedure are the water thickness Cw, the dry mattercontent Cm, and the internal structure parameterN (Fig. 10).
The three values of measured transmittanceTRðλC;iÞ could be used for inverting the set of bio-chemical parameters at the same time (N, Cw, andCm); this implies the use of three wavelengths imple-mented into the RAMIS prototype. Furthermore, it ispossible to invert each parameters individually ifthey are not strongly correlated; this is the case ofthe dry matter Cm at 843nm wavelength transmit-tance. However, the leaf structure parameterN mustbe inverted at the same time with all or selected bio-chemical parameters; then for inverting one param-eter we must use two wavelength at least.
For the inversion procedure of water estimations,we must deal with the correlation of the dry matterat the spectral zone where the photodiodes wave-bands were placed. For this reason, the leaf structureparameter N must be inverted together with thewater Cw and the dry matter Cm content, so threewavelengths must be used in this case: 843, 937,and 1550nm.
The result of water estimation content, from differ-ent leaves samples, shows a fair agreement with thewater content measured by the drying process, as isshown in Fig. 11.
For dry matter estimation Cm, only two wavebandswere used: 843 and 937nm. The values of water
Fig. 10. (Color online) Inversion procedure of the transmittances TRðλC;iÞ, measured by RAMIS in order to compute the biochemical leafparameters N, Cw, and Cm using the adaptation of the radiative transfer model PROSPECT.
Table 1. Leaves of Different Plants Species Measured byRAMIS for Water and Dry Matter Estimations
Latin Name Common Name
Prunus avium Sweet CherryHedera helix Common IvyDavidia involucrata Dove TreeQuercus agrifolia Coast Live OakFicus benjamina Weeping FigGinkgo biloba GinkgoFagus sylvatica purpurea Purple BeechIlex aquifolium European HollyLaurus nobilis Laurel TreeSyringa vulgaris Common LilacPlatanus hispanica London PlaneRosa eglanteria Eglantine RoseTilia platyphyllos Large-Leaved Linden
contents Cw previously computed were used as apriori information as a result of its small correlationat 937nm (Fig. 2(b)) for inverting the dry matter con-tent Cm. Also, the structure leaf parameter N mustbe inverted as well. The results obtained do not agreewith those obtained by measuring the dry mass di-rectly, as is shown in Fig. 12.
The dispersion of the estimated values could bedue to different factors such as uncertainty of the di-gital acquisition system and the geometrical varia-tion of source–sensor configuration.
Also, the reasons for the discrepancies are linkedto the small dependence of the transmittance withCm, as can be seen in Fig. 2(b). In fact, a change of
Fig. 11. (Color online) Comparison between the measured and the estimated values of water thickness (Cw) from different plant leaves(Table 1) using the wavelengths of 843, 937, and 1550nm in the inversion procedure.
Fig. 12. (Color online) Comparison between themeasured and the estimated values of drymatter content (Cm) from different plant leaves(Table 1) using the wavelengths of 843 and 937nm in the inversion procedure.
3% of the measured transmittances values producesa variation by more than 100% on the computed drymatter estimation Cm.
7. Conclusions
Our aim in developing RAMIS was to present an in-strument capable of performingmeasurements of themain biochemical compounds of plant leaves, withthe principal characteristic of being a portable in-strument able to perform measurements in situ.The capability of RAMIS to determine water contentCw was demonstrated by computing the inversion oftransmittance measurements. The choice of wave-length used in the developing process of the proto-type is based on the results of the spectralresponse of plant leaf transmittance, verifying thespectral sensitivity zone at each biochemical com-pound and avoiding spectral zones with high correla-tion between them and the LED wavelengthavailability.The estimation of dry matter, regarding its small
effect on spectral variations (Fig. 2), is affected bythe limited accuracy of transmittance measure-ments; consequently, dry matter estimation wasnot accomplished.The implementation of the radiative transfer mod-
el PROSPECT into the RAMIS prototype shows agood result regarding water estimation.The capability of the radiative transfer model
PROSPECT to compute also the plant leaf re-flectance as a function of its biochemical contentscould be used to improve the accuracy of RAMISby setting a reflectance sensor in an adequateposition [21].These possible upgrades have been considered
based on the conception of a portable instrumentcapable of performing multi-biochemical mea-surements using different wavelengths [22] andusing an optical leaf model. This device could per-form biochemical leaf estimations without a pre-vious calibration procedure for any internalleaf structure.
We thank the Centre Régional d’Innovation et deTransfert Technologique—Conception Circuits Spé-ciaux et Télématique (CRITT-CCST), Force A Com-pany, and Diderot Valorisation Bureau of the ParisDiderot University. We thank also Natacha Vendolafor her contribution on this work. This projectreceived the 2006 Environmental Innovative Tech-nology Award of ADEME (French Agency for Envir-onment and Energy Management), Contribution No.IPGP 2636.
References1. W. F. McClure, A. Hamid, F. G. Giesbrecht, and W. W. Weeks,
2. V. Demarez, J. P. Gastellu-Etchegorry, E. Mougin, G. Marty, C.Proisy, E. Dufrne, and V. LeDantec, “Seasonal variation of leafchlorophyll content of a temperate forest, inversion of thePROSPECT model,” Intl. J. Remote Sensing 20, 879–894(1999).
3. G. Le Maire, C. Francois, and E. Dufrene, “Towards universalbroad leaf chlorophyll indices using PROSPECT simulateddatabase and hyperspectral reflectance measurements,” Re-mote Sensing Environ. 89, 1–28 (2003).
4. G. H. Downing, G. Carter, W. Holladayk, and W. G. Cibula,“The radiative-equivalent water thickness of leaves,” RemoteSensing Environ. 46, 103–107 (1993).
5. H. W. Gausman, “Plant Leaf Optical Properties in Visible andNear Infrared Light, Graduate Studies Series (Texas TechUniversity, 1985), Vol. 29.
6. D. Combes, L. Bousquet, S. Jacquemoud, H. Sinoquet, C. Var-let-Grancher, and I. Moya, “A new spectrophotogoniometer tomeasure leaf spectral and directional optical properties,”Remote Sensing Environ. 109, 107–117 (2007).
7. T. Fourty, F. Baret, S. Jacquemoud, S. Schmuck, andJ. Verdebout, “Leaf optical properties with explicit descrip-tion of its biochemical composition: direct and inverse pro-blems,” Remote Sensing Environ. 56, 104–117 (1996).
8. S. Chung, J. Sung, K. A. Sudduth, S. T. Drummond, and B.Hyun, “Spatial variability of yield, chlorophyll content, andsoil properties in a Korean rice paddy field,” Proceedings ofthe 5th International Conference on Precision Agriculture,P. C. Robert, ed. (ASA, 2001).
9. A. D. Richardson, S. P. Duigan, and G. P. Berlyn, “An evalua-tion of noninvasive methods to estimate foliar chlorophyll con-tent,” New Phytologist 153, 185–194 (2002).
10. Y. Goulas, Z. G. Cerovic, A. Cartelat, and I. Moya, “Dualex: anew instrument for field measurements of epidermal ultravio-let absorbance by chlorophyll fluorescence,” Appl. Opt. 43,4488–4496 (2004).
11. R. B. Myneni, Photon–Vegetation Interactions: Applications inOptical Remote Sensing and Plant Ecology (Springer-Verlag, 1991).
12. S. Jacquemoud and F. Baret, “PROSPECT: a model of leafoptical properties spectra,” Remote Sensing Environ. 34,75–91 (1990).
13. W. Allen, “Transmission of isotropic light across a dielectricsurface in two and three dimensions,” J. Opt. Soc. Am. 63,664–666 (1973).
14. J. B. Feret, C. Franois, G. P. Asner, A. A. Gitelson, R. E.Martin,P. R. Bidel, S. L. Ustin, G. Le Maire, and S. Jacquemoud,“PROSPECT-4 and 5: advances in the leaf optical propertiesmodel separating photosynthetic pigments,” Remote SensingEnviron. 112, 3030–3043 (2008).
15. G. G. Stokes, “On the intensity of the light reflected from ortransmitted through a pile of plates,” Proc. R. Soc. London11, 545–556 (1860).
16. S. Jacquemoud, S. L. Ustin, J. Verdebout, G. Schmuck, G. An-dreoli, and B. Hosgood, “Estimating leaf biochemistry usingthe PROSPECT leaf optical properties model,” Remote Sen-sing Environ. 56, 194–202 (1996).
17. Y. Zuhu, S. Runhe, and Z. Ershun, “Calculation of mesophyllstructure parameter and its effect on leaf spectral reflectance,”inProceedings of IEEE International Conference on Geoscienceand Remote Sensing Symposium, 2005 (IEEE, 2005),pp 1299–1301.
18. A. Tarantola, Inverse Problem Theory and Methods for ModelParameter Estimation (Society for Industrial and AppliedMathematics, 2005).
19. J. P. Frangi, S. Jacquemoud, G. De Rosny, B. Equer,I. Roca, P. Cabarrocas, and R. Vanderhagen, “Radiometricdevice and method for determining in situ the biochemicalcontent of leaves, and portable apparatus comprisingsame,” World Intellectual Property Organization WO/2003/006960, patent PCT/FR2002/002494 (23 January 2003).
20. P. Bousquet, Spectroscopie Instrumentale (Dunod Univer-sité, 1969).
21. P. Ceccato, S. Flasse, S. Tarantola, S. Jacquemoud, andJ. M. Grégoire, “Detecting vegetation water content usingreflectance in the optical domain,” Remote Sensing Environ.77, 22–33 (2001).
22. D. A. Sims and J. A. Gamon, “Relationships between leaf pig-ment content and spectral reflectance across a wide range ofspecies, leaf structures and developmental stages,” RemoteSensing Environ. 81, 337–354 (2002).
