LIBERTY—Modeling the Effects of Leaf Biochemical Concentration on Reflectance Spectra Terence P. Dawson,* Paul J. Curran,* and Stephen E. Plummer† T he conifer leaf model LIBERTY (Leaf Incorporating Biochemistry Exhibiting Ref lectance and Transmittance Yields) is an adaptation of radiative transfer theory for determining the optical properties of powders. LIBERTY provides a simulation, at a fine spectral resolution, of quasiinfinite leaf ref lectance (as represented by stacked leaves) and single leaf reflectance. Single leaf ref lectance and transmittance are important input variables to vegetation canopy reflectance models. A prototype parameterization of LIBERTY was based upon measurements of pine needles and known absorption coefficients of pure component leaf biochemicals. The estimated infinite-reflectance output was compared with the spectra of both dried and fresh pine needles with root mean square errors (RMSE) of 2.87% and 1.73%, respectively. The comparisons between measured and estimated ref lectance and transmittance values for single needles were also very accurate with RSME of 1.84% and 1.12%, respectively. Initial inversion studies have demonstrated that significant improvements can be made to LIBERTY by utilizing in vivo absorption coefficients which have been determined by the inversion process. These results demonstrate the capability of LIBERTY to model accurately the spectral response of pine needles. Elsevier Science Inc., 1998 INTRODUCTION Understanding the interaction of radiation with vegetation canopies requires a knowledge of the spectral prop* Department of Geography, University of Southampton, Highfield, Southampton, United Kingdom † Section for Earth Observation, Institute of Terrestrial Ecology, Monks Wood, Abbots Ripton, Cambridgeshire, United Kingdom Address correspondence to Terence P. Dawson, Univ. of Oxford, Environmental Change Unit, 1a Mansfield Rd., Oxford OX1 3TB, UK. E-mail: [email protected] Received 22 October 1996; revised 13 December 1997. REMOTE SENS. ENVIRON. 65:50–60 (1998) Elsevier Science Inc., 1998 655 Avenue of the Americas, New York, NY 10010 erties of individual leaves (Aber, 1994). Leaves are an important component of forest canopies, and it is the concentration of their biochemical constituents, namely, pigments, water, nitrogen, cellulose, and lignin, together with canopy structure that shapes the absorption features of remotely sensed ref lectance spectra. Absorption features in the near infrared region of the spectrum (1000– 2500 nm) are a function of the bending and stretching vibrations of biochemical bonds (between, for example, hydrogen–carbon and nitrogen–oxygen atoms) together with their harmonics and overtones (Kemp, 1987). In the visible region, chlorophyll and carotenoid pigments have strong absorption due to electron energy transitions (Lichtenhaler, 1987; Mackinney, 1941). Fine spectral resolution remotely sensed data can be statistically analyzed to estimate the concentration of biochemicals in forest canopies. Such information has been used to drive ecosystem simulation models and estimate photosynthetic efficiency, the rate of nutrient cycling, and the degree of vegetation stress (Curran, 1994). Strong correlations between leaf biochemical concentrations and specific wavebands of measured spectra have now been reported by Wessman et al. (1988), Martin and Aber (1990), Curran and Kupiec (1995), and Zagolski et al. (1996). However, wavebands selected by multiple linear regression using biochemical assay data are often not consistent with the absorption features of the biochemicals within the leaves. Even when known causal wavebands are selected in the regression equation, spectral correlation with biochemical concentrations are not always strong (Dawson et al., 1995a). To investigate this further, a physical modeling approach is required to describe and quantify the ref lectance, scattering, and absorption of canopy leaves as a function of their biochemical and physical properties. Some of the earliest research on leaf reflectance mod0034-4257/98/$19.00 PII S0034-4257(98)00007-8 LIBERTY—Conifer Leaf Model eling was based upon the Kubelka–Munk (K–M) radiative transfer theory (Allen and Richardson, 1968), and this theory has since been used by others (Yamada and Fujimura, 1988; Conel et al., 1993). The K–M approach is a two-stream approximation to the radiative-transfer equation. Later Allen et al. (1969) proposed a single layer “plate” model. By extending the plate model to several layers with air spaces between plates Allen et al. (1970) was able to simulate multiple scattering which occurs as a result of the air space/cell wall discontinuities within the mesophyll structure of a leaf (Gausmann, 1974). The PROSPECT model (Jacquemoud and Baret, 1990) develops Allen’s multiple layer plate model and adopts a solid angle of incident radiation instead of an isotropic direction. This modification accommodates shadowing caused by the undulating shape of the leaf surface at microscopic scales (Grant, 1987). Designed to illustrate the effects of chlorophyll and water content of leaves on the spectra of those leaves, PROSPECT has now been upgraded (Jacquemoud et al., 1996) to incorporate the biochemicals of cellulose, lignin, and protein. The K–M and plate theories assume the absorption and scattering of the radiation is distributed homogeneously throughout the leaf when in fact it is characterized by the absorbing biochemicals being localized into hydrated cells which are separated by intercellular air spaces causing multiple refractive index discontinuities. Other leaf models have utilized ray tracing (Allen et al., 1973) and stochastic methods (Tucker and Garratt, 1977), but these models are difficult to invert and require computer-intensive processing. In addition, existing leaf models have usually been based upon broadleaf vegetation where the leaf structure is considered to be an infinitely extending plane having distinct layers. Pine needles present their own particular problem. They do not contain a well-defined pallisade layer, and so a transverse section of a pine needle will display roughly spherical cells. There are also problems of how to characterize and model the pine foliage elements. Because of the complex spatial distribution of pine canopies, it may be more appropriate to treat shoots (clumped needles) as the independently located foliage element instead of individual needles (Stenberg et al., 1994). Finally, their shape and size make it difficult to accurately measure the spectral properties of individual needles even in the laboratory (Daughtry et al., 1989). LIBERTY (Leaf Incorporating Biochemistry Exhibiting Reflectance and Transmittance Yields) was developed specifically to address these problems. The objective was to generically characterize and model conifer (particularly pine) needles at the cellular level by adapting Melamed’s theory of light interaction with suspended powders (Melamed, 1963) and infer leaf optical parameters from stacked needles in the laboratory. An enhancement of LIBERTY to accommodate a thickness parameter, thereby providing a transmittance 51 output, facilitates the modeling of individual conifer needle elements. Therefore, based upon known parameters derived in the laboratory, LIBERTY can be coupled with a suitable canopy model for investigations of forest canopy ref lectance. Inversion of LIBERTY, providing in vivo absorption coefficients, indicate significant improvements in model accuracy and hint at the potential to estimate biochemical concentrations and biophysical variables directly from high spectral resolution remote sensing (Dawson et al., 1997). MODEL THEORY Infinite Ref lectance Function The total reflectance R of a medium, such as a leaf or stacked leaves, will be the sum of the scattered radiation f luxes emerging in an “upwards” direction out of the leaf, assuming a horizontal leaf surface. R is derived by first obtaining an approximate expression for the total ref lectance of an individual leaf cell as a function of the optical absorption coefficient k of the cell medium and taking into account both internal and external scattering. The parameters and variables of R can then be related to the bulk optical characteristics for the diffusion of radiation over the whole leaf. The Interaction of Radiation with a Leaf Cell Let us assume that the internal cells of a leaf are spherical particles whose surface will scatter radiation according to Lambert’s cosine law and whose average reflectance coefficients for internally and externally incident radiation are defined as mi and me, respectively. Leaves actually consist of cells having different shapes and orientation with respect to the surface, but we can approximate the optical properties of the internal structure of a leaf by assuming that the average surface-scattering effects and transmission s of an aggregation of leaf cells with the same mean diameter can be characterized by a spherical model. In a perfect sphere, the angle at which a refracted ray is incident on the inside of the sphere is equal to that at which it enters the sphere. However, as emphasized by Melamed (1963), for an ensemble of irregularly shaped cells the two angles are uncorrelated with the refracted ray having virtually an equal probability of encountering interior surfaces at all orientations. The ref lectance coefficient of a surface is defined as the ratio of the ref lectance of the surface to that of a perfectly diffuse surface under the same conditions of illumination and measurement. Therefore, the reflection coefficient m(h) for a direction h is determined by the Fresnel equations for the specular ref lection of unpolarized radiation. The evaluation of the average ref lection coefficients me and mi assume the cell surfaces are ideally diffuse and obey Lambert’s cosine law whereby the reflected radiance in a direction h with respect to the nor- 52 Dawson et al. mal will be proportional to cos h. The average value of the external ref lectance coefficient me, for light moving from a medium having a low refractive index to one having a high one will be Eq. (1): p/2 me5# m(h) sin h cos h dh 0 (1) and for radiation incident mi for when light passes from a high to low medium Eq. (2): hc mi5(12sin2 hc)12# m(h) sin h cos h dh, 0 (2) where hc is the critical angle. The LIBERTY model assumes that the cell spheres are separated by an air gap; therefore, me will be the reflectance coefficient for light passing from the air to the cell. For unit flux radiating into the sphere from the surface, the intensity, defined as flux per unit solid angle, in a direction h (with h taken along the diameter d of the sphere) will be (1/p)cos h. Let the medium of individual leaf cells absorb light with absorption coefficient k. Then radiation reaching the cell wall with an angle dh, in the direction h will be Eq. (3): (1/p)e2kd cos h cos h 2p sin h dh. (3) After one pass, therefore, the total radiation reaching the surface M will be Eqs. (4,5): p/2 M52# e2kd cos h cos h 2p sin h dh 0 52[12(kd11)e2kd]/(kd)2. (4) x5xu1(xas)xu1(xas)2xu1···, (5) which is a converging series where the solution at the asymptotic limit will be The proportion of the initial radiation emerging from the cell will be miM since mi is the average reflectance coefficient of the cell surface for internally incident radiation. Assuming that the cell has identical scattering properties at every point on its surface, the total radiation after infinite interreflections is the sum of the geometric series in miM. Therefore, the total transmitted component s will be Eq. (6): s5(12mi)M/(12miM), (6) and the component radiation that is absorbed will be Eq. (7): 12s5(12M)/(12miM). rounded by leaf material having a ref lectance R. However, the cells on the surface of the leaf are exposed on the side of the material boundary and radiation must enter or leave by way of this boundary. If we assume that the leaf surface is horizontal, then we can define a set of parameters xu, xa, and xd, which represent the fraction of the radiation emerging from the interior of a cell in upward, adjacent, and downward directions, respectively, and which is then intercepted by the overlying, adjacent, and underlying cells. Because the radiance emerging from any point of an inner surface of the cell is assumed to be isotropic, these parameters approximate the solid angle of the radiation, expressed as fractions of 4p steradians. For the special case at the medium boundary, the fraction of the emitted radiation xu for cells in the surface layer is scattered through an angle of approximately 2p steradians, which is the angle of free space seen by the emitted radiation. For simplification, the lateral radiance xa can be resolved into its upward and downward components, thereby facilitating a one-dimensional solution to the diffusion of radiation in a leaf. We define a probability coefficient x as the total fraction of the radiation which emerges from the interior of a cell and which is scattered upwards towards a cell layer one cell distance closer to the leaf surface. Then we can define x as (7) Now consider an assembly of spherical cells of uniform size each having a transmission s. By striking an individual cell, therefore, an incident light ray will be partially reflected from the surface with a value me and the remainder (12me) will enter the cell. This component of the radiation will reemerge from the cell with a value (12me)s. The Interaction of Radiation with a Layer of Cells By assembling many cells in close proximity, a quasiinfinite thickness is achieved. Assuming that the cells are in layers, all of the cells, save the surface cells, will be sur- x5xu/(12xas). (8) At low absorption coefficients where kd,1, xd>xu, and xu1xd1xa>1, we have Eq. (9): x5xu/(12(122xu)s). (9) The parameter x is important as it can be treated as a function of the intercellular air gap or scattering efficiency. By increasing x, we will have a higher ref lectance due to increased radiation scattering from lower to upper layers. Initially, however, the ref lected component me of an external incident ray is scattered over an angle 2p steradians; therefore, xd is 0 and xu is twice the value of radiation which emerges from within a cell. The initial reflected component of radiation contributing to R for unit incident radiation is 2xme with Eq. (8) expressing the effect of shadowing by the adjacent cells at the surface. The remainder (122xme) enters the cell. Of this fraction, the amount of radiation which is reemitted from the leaf thereby contributing to the total ref lectance R is x(12 2xme)s. An amount (12x)(122xme)s is therefore transferred to the lower leaf layers, of ref lectance R. In a similar fashion, (12x)(122xme)sR is ref lected back by the lower layers, returns to the surface cell, and the amount, x(12x)(122xme)2s2R is transmitted upwards. The next in- LIBERTY—Conifer Leaf Model 53 is the fraction of radiation that emerges after an infinite number of interreflections for unit intensity striking the lower layers, one cell depth below the surface. In addition to the radiation which has emerged after all interref lections of the initial radiation component, we add the radiation which, after the second, third, and subsequent ref lections, entered the surface cells and was scattered back into the interior. The sum of these components will be the residue of the right-hand component of Eq. (10): (12x)2(122xme)s[(12me)sR/(12meR)]. (12) This term represents the initial intensity so by applying Eq. (11), the fraction of Eq. (12) which emerges after an infinite number of interref lections will be Eq. (13): x(12x)2(122xme)s[(12me)sR/(12meR)]2, (13) and the fraction which returns to the interior of the leaf will be Eq. (14): (12x)3(122xme)s[(12me)sR/(12meR)]2. (14) By continuing this process, the equation for the sum of all the radiative components which contribute to the total radiation emerging from the leaf surface Eq. (15): R52xme1x(122xme)s 1x(12x)(122xme)s[(12me)sR/(12meR)] Figure 1. A schematic representation of LIBERTY showing the scattered radiation paths in a leaf. Each sphere can represent the same or a different cell. In an aggregate of cells, each surface cell is surrounded on all but one side by the bulk of the leaf having reflectance R. The coefficient x represents the fraction of the radiation that has emerged from the interior of the cell and is scattered upwards, me is the reflectance coefficient for radiation passing from the intercellular air space to the cell, s is the transmitted component of radiation of an individual cell, and R is the total ref lection of the leaf bulk. 1x(12x)2(122xme)s[(12me)sR/(12meR)]2 1x(12x)3(122xme)s[(12me)sR/(12meR)]31···, (15) which can be rewritten: R5 2xme1x(122xme)s1x(12x)(122xme)2s2R 1x(12x)(122xme)2s2meR21x(12x)(122xme)2s2me2R3 1x(12x)(122xme)2s2me3R41··· 52xme1x(122xme)s (10) A schematic representation of the mechanism is shown in Figure 1. From an examination of Figure 1, the term x[(12me)sR/(12meR)] (16) The form of Eq. (16) is quadratic in nature; therefore, an approximation of the root for R can be resolved easily using the Newton–Raphson iterative methodology (Jeffrey, 1986). terreflection of the initial ray will contribute the component x(12x)(122xme)2s2meR2 to the total ref lectance R. After an infinite number of interactions, the radiation being emitted by the leaf and contributing to the total radiation of the leaf R will be 1x(12x)(122xme)s[(12me)sR/(12meR)]. 2xme1x(122xme)s(12meR) . (12meR)2(12x)(12me)sR (11) The Transmission of a Needle Having Finite Layers An adaptation of LIBERTY is required to provide the transmission of a single needle element. Laboratory measurements of the spectral properties of conifer needles (Daughtry et al., 1989) have shown that, in the nearinfrared wavelengths (800–1300 nm), about 40% of the total incident radiation is transmitted while a slightly higher percentage is reflected. The Melamed (1963) model was therefore modified to determine the radiation components as a function of leaf thickness. This was achieved by resolving the average reflectance and transmittance values of a single layer of leaf cells and, through adaptation of a procedure demonstrated by Benford (1946), providing a measure of total reflectance and transmittance as a continuous function of thickness. In considering a single layer of leaf cells, examina- 54 Dawson et al. tion of Eq. (16) shows that, where there is no underlying leaf material whose backscatter will contribute to the total reflective radiation, the total reflection R will be equal to R52xme1x(122xme)s. (17) In a series of techniques used by Benford (1946), the number of individual layers of leaf cells, each having identical ref lectance and transmittance functions, can be increased providing a total ref lection and transmission value as a function of thickness. When R reaches a maximum (i.e., infinite ref lectance, R∞), it can be shown that R∞5R1 T 2R , (12R∞R) (18) where R and T are the values of ref lectance and transmittance for the single layer of leaf cells. Because Eq. (18) relates the general terms T, R, and R∞ and the values of R∞ and R are known from Eqs. (16) and (17), respectively, we can determine T by solving Eq. (19): R) 3(R 2R)(12R 4 R T5 ∞ ∞ ∞ 1/2 . (19) The equations for Tn and Rn, where n is the number of layers, have been derived and incorporated into LIBERTY based upon a continuous function of thickness, where n is not limited to an integral thickness. This enables us to quantify total leaf transmittance and ref lection as a function of the average leaf thickness. LEAF OPTICAL PROPERTIES USED TO PARAMETERIZE LIBERTY To parameterize LIBERTY, we used data from the laboratory analysis of jack pine (Pinus banksiana) needles. The laboratory reflectance measurements made on stacked freeze-dried and ground needles and pure and in vitro leaf component biochemicals were made using a Perkin Elmer Lambda 19 double-beam spectrophotometer equipped with an integrating sphere. The instrument was continually purged with gaseous nitrogen to remove atmospheric moisture. The calibration of the instrument was performed using a barium sulfate (BaSO4) reference plate and internal wavelength calibrations. The absorption was measured between wavelengths of 400 nm and 2500 nm at 1 nm intervals. To measure and to remove leaf moisture, the samples were freeze-dried in an Edwards Modulyo freeze drier. Deep frozen (2308C) needles were placed in a vacuum oven and atmospheric pressure was reduced to less than 1.431024 Torr. The temperature was then increased slowly to 1308C for 48–60 h. The freeze-drying process is well suited to the spectrophotometric studies of leaves as the low drying temperatures do not denaturize the component biochemicals. All measured and adopted absorption coefficients were standardized to zero absorption at the minima (occurring in the 800–900 nm wavelength region) and a Figure 2. Pure combined chlorophyll-a and chlorophyll-b, in vitro extracted pigments and pure lignin absorption in visible wavelengths. wavelength-independent linear absorption coefficient was incorporated into the model. Jacquemoud and Baret (1990) fixed the value of this elementary absorption using the estimated absorption of dry albino maize leaves having the most compact structure. However, we allowed this absorption to vary in order to i) analyze empirical variation in spectra maxima as a function of biochemical concentration and/or leaf internal structure and ii) quantify, through inversion studies, the causal effects of the structure on the spectra as a result of water stress and/ or senescence (Peñuelas et al., 1996; Aldakheel and Danson, 1996). Absorption in the Visible Wavelengths Initial investigations of the absorption coefficient of freeze-dried leaves in visible wavelengths showed that a negative natural logarithmic absorption continuum in the visible wavelengths was convolved with absorption attributed to pigments. This was verified by bleaching chopped pine needles in 100% acetone. The albino leaves were then air-dried and measured with the spectrophotometer. The observed 2ln absorption coefficent was incorporated into the model as it represented absorption due to leaf structure and biochemicals, in particular lignin. This absorption has been validated by others (Elvidge, 1990; Schubert, 1965) and is near-identical to the absorption coefficient of “pure” lignin (Fig. 2). Lignin has a strong absorption in ultraviolet wavelengths peaking at a wavelength of 280 nm and extending through the visible and near-infrared wavelengths. The absorption spectra of pure biochemicals were examined and combined linearly to determine the fit of the total absorption effects of the leaf. Chlorophyll a and b pigments were dissolved in 80% acetone and their respective absorption coefficients were measured and then combined to a chlorophyll a/b ratio of 2.5;1. This ratio has been shown to be reasonably constant for slash pine needles grown in diverse environmental conditions (Kupiec, 1995). The resultant formulation was then compared to absorption coefficients of total in vitro leaf pigments LIBERTY—Conifer Leaf Model 55 extracted using the techniques of Mackinney (1941) (Fig. 2). A difference spectrum exhibits a close approximation to published carotenoid absorption coefficients (Lichtenthaler, 1987) in the 440–500 nm wavelength region (Fig. 3). The large absorption maximum at a wavelength of 410 nm in the difference spectra is probably attributable to extracted chlorophyll a and b components that have a broader and higher blue region absorbency as a result of the increased water content of solvent extracted pigments. Acetone causes a shift of the pigment absorption peaks to shorter wavelengths. The chlorophyll a maximum absorption peak in the red region had an in vivo maxima at a wavelength of 672 nm and an in vitro (90% aqueous acetone) maxima at a wavelength of 664 nm, a shift of 8 nm. Because of the increased water content and greater interaction of the combined pigments, there was a further shift of about 2 nm between the absorption maxima of the extracted whole pigments and the pure chlorophyll a and b mixture in the same aqueous acetone solution. Initially, the wavelength corrected in vitro absorption data were incorporated into the LIBERTY model. cellulose concentrations in the slash pine needles used here (Kupiec, 1995) showed that their mean concentrations were 23% and 35% of dry weight, respectively. Their pure absorption coefficients were measured from purchased laboratory powders (The Aldrich Chemical Company Inc.) and, because of their reasonably positive cocorrelation, combined linearly and adjusted to attain the lowest root mean square error (RMSE) between the empirical freeze-dried data set and LIBERTY-predicted output. This resulted in a lignin to cellulose ratio of 1.13;1 which suggests that, although there is a larger proportion of cellulose in conifer needles, lignin has a stronger absorption weighting. This effect was probably due to the encrusting nature of lignin (Schubert, 1965) causing a masking of the cellulose and a reduction of its interaction with radiation. To incorporate the effects of nitrogen concentration, an absorption spectrum of protein (which is dominated spectrally by nitrogen) was adapted from the revised PROSPECT (Jacquemoud et al., 1996) model. Using the same method as that for lignin and cellulose absorption, nitrogen concentration were set in the model to produce the lowest RMSE between the measured spectra and model spectra. Although the accuracy of the model spectra was increased, our studies did not produce a strong correlation between nitrogen concentrations and spectral variation as a result of variability in protein absorption. Because leaf ref lectance and transmittance properties are more appropriately related to biochemical content (mass per unit leaf area) than concentrations (mass per unit leaf dry mass), the absorption coefficients were scaled using the specific leaf area (SLA) for jack pine which has been estimated to be 80 g m22. Using biochemical content as the input variable enables the model to be used to estimate the spectra of other conifer species with different SLAs. Absorption in the Near-Infrared The water absorption coefficient for LIBERTY was based upon the published absorption coefficients of pure distilled water over the spectral range 750–2500 nm (Curcio and Petty, 1951), which has been shown to be in agreement with published absorption coefficients of in vivo leaf water (Jacquemoud and Baret, 1990) estimated from PROSPECT inversion methods. Cellulose and lignin are the most abundant of the compounds found in the cell walls of conifer leaves. The cellulose molecules are organized into highly hydrated microfibrils which form the basic structural unit of the cell wall. Extensive lignin deposition within the interfibrillar spaces confers considerable strength to the leaf tissues. The absorption coefficients of these biochemicals have similar features and their combined in vivo absorption peaks have proven difficult to separate due to their intimate mixture. A biochemical analysis of lignin and Characterization of LIBERTY Structural Parameters To determine the refractive index of the air-cell interfaces, leaves can be infiltrated with oil mixtures having variable refractive indexes (Woolley, 1971; Gausman, 1974). The ref lectance of leaves is minimal when the refractive index of the liquid matches the refractive index of the cell wall. The effective refractive index of a typical leaf in the visible wavelengths is not inconsistent with the published refractive index, n51.47, of epicuticular wax, but Allen et al. (1969) demonstrated that it decreases almost linearly with wavelength to a value of about 1.2 at a wavelength of 2500 nm. Refractive indexes incorporated into LIBERTY were adopted from a regression equation of calculated refractive indexes (Jacquemoud and Baret, 1990) against wavelength as a best approximation to the average refractive indices of wet mesophyll cells for the 400–2500 nm wavelength region, respectively. Examination of a transverse section through red spruce Figure 3. A comparison of estimated and measured carotenoid absorption, determined by spectrophotometer measurements of b-carotene extracted from carrot roots. 56 Dawson et al. needles showed the diameter of internal leaf cells were in the region of 30–60 lm (Rock et al., 1986). The scattering efficiency of the Melamed (1963) model as a function of the absorption coefficient in visible and near-infrared wavelengths has been tested (Hapke and Wells, 1981) for different sphere sizes against those of crushed glass powders. The measured values of the internal scattering curve were indistinguishable from the Melamed (1963) model for an average sphere diameter d535–50 lm. VALIDATION DATA To compare LIBERTY output with the ref lectance spectra of dried needles, spectrometer measurements were made on 160 needle samples collected from sample sites in Nipawin National Reserve, Saskatchewan, Canada as part of the Boreal Ecosystem Atmosphere Study (BOREAS) in 1994. This site, comprising jack pine (Pinus banksiana), provided 10 sample plots with 16 samples collected from each plot; eight of these samples were first year needles and eight were second year needles. The reflectance of dried and finely-chopped samples were measured to produce a mean ref lectance spectrum for each plot and all ages of needle. The concentration of chlorophyll, carotenoids, water content, nitrogen, cellulose, and lignin were measured using wet chemical assay and stepwise linear regression techniques at the Universities of Southampton and New Hampshire. Comparison of LIBERTY output with the reflectance spectra of fresh, stacked whole needles were made using slash pine (Pinus elliottii) foliage samples collected from Gainesville, Florida in 1992 (Curran and Kupiec, 1995). Laboratory ref lectance measurements were made for 80 samples using an IRIS Mk IV spectroradiometer. Both dried and fresh samples had been stacked to ensure a sufficient optical thickness for infinite ref lectance, and the spectra were averaged and compared to the LIBERTY-predicted outputs. Figure 4. A comparison between measured and LIBERTY-predicted ref lectance and transmittance values for several quasiinfinite and single needle experiments. Validation of the single needle reflectance and transmittance model outputs were made using microspectrometer measurements of jack pine samples. The equipment used was a Zeiss UMSP 50 (universal microspectrophotometer) microscope. Reflectance and transmittance measurements were taken at 1 nm steps over a wavelength range of 400–800 nm, individual values being based on a mean of 10 measurements taken along the length of each needle. RESULTS AND DISCUSSION Based upon the model variables and biochemical data outlined in Table 1, the laboratory spectrometer data and the LIBERTY predictions were highly correlated (r50.99) across all wavelengths (Fig. 4). For both the dried and fresh needle samples, the RMSE between estimated and measured spectra for the quasiinfinite setup were 2.87% Table 1. Leaf Biochemical Variation and LIBERTY Variables Used for Stimulating the Spectra of Jack Pine (Pinus banksiana) Needles Biochemical Mean Maximum Minimum g21) 2.48 103.3 1.15 25.1 38.3 4.00 165 1.51 28.2 45.4 1.28 55.6 0.8 22.0 31.3 Structural Variable Dried Needles Fresh Needles Chlorophyll (mg Water (% dry mass) Nitrogen (% dry mass) Lignin (% dry mass) Cellulose (% dry mass) Average internal cell diameter (lm) Intercellular air space determinant, xu Leaf thickness Baseline absorption Specific Leaf Area (SLA) g m22 40 0.03 0.000446 80 45 0.028 1.62 0.00058 80 LIBERTY—Conifer Leaf Model Figure 5a. LIBERTY-predicted and dried needle ref lectance spectra. for the predicted freeze-dried needles (Fig. 5a) and 1.73% for fresh needles (Fig. 6a). A sensitivity analysis of LIBERTY indicated that it was necessary to reduce the average cell diameter for dry needles, which is to be expected as water loss would likely cause a contraction of the cells. In addition, the scattering parameter x was increased for dry needles where the model exhibits a higher reflectance due to many more air/cell interfaces as a result of cell distortion. All of the above observations agree with our underlying physical understanding of leaves. However, an increase in the linear absorption coefficient was also required to reduce the ref lectance of fresh whole needles. This suggested that there were changes to the sample density that had been unaccounted for in the model. The dried needles had been finely chopped and stacked dry samples were more compact than their fresh counterparts. This difference in sample density affected both the depth of radiation penetration and the degree of multiple scattering from adjacent needles. The pigment absorption features in visible wavelengths were narrower in the predicted spectra as a result of adapting the absorption coefficient of in vitro pigments in acetone. Examination of different extracted pigment Figure 5b. Revised LIBERTY-predicted and dried needle ref lectance spectra using in vivo absorption coefficients derived from an inversion of LIBERTY. 57 Figure 6a. LIBERTY-predicted and fresh slash pine needle ref lectance spectra. samples in aqueous acetone also highlighted large differences in the carotenoid absorption region between samples where chlorophyll concentrations are similar. This suggested that separate carotenoid and chlorophyll absorption coefficients are necessary in the model. The model output in particular shows an excessive absorption region between wavelengths of 400–470 nm when compared to empirical spectra in both fresh and dried leaves. Although LIBERTY exhibits the distinct side “lobes” observed on each side of the green reflectance peak of measured conifer needles, centered around a wavelength of 520 nm, the depth of these effects do not mirror those in the empirical data very well. Numerical inversion studies of LIBERTY, based upon the Newton–Raphson iterative methodology, show that the intermixing of the biochemicals affecting absorption in the visible wavelengths produces a distinct in vivo absorption coefficient (Fig. 7) that is different from the absorption due to in vitro pigments. The increased absorption can be accounted for by the addition of in vivo lignin absorption effects and the fact that, in the chloroplast membrane of leaf cells, chlorophyll does not exist in free solution, but as pigment-protein complexes. Therefore, the absorption spectrum of chlorophyll a, for exam- Figure 6b. Revised LIBERTY-predicted and fresh jack pine needle ref lectance spectra. 58 Dawson et al. Figure 6c. Revised LIBERTY-predicted and fresh slash pine needle ref lectance spectra. ple, which has interacted with protein, gives rise to several chlorophyll a classes (French et al., 1972), each of which contribute linearly to the total absorption of visible light. The only other significant deviation between LIBERTY-predicted and measured reflectance for both fresh and dry leaves is in the 1000–1600 nm wavelength region where LIBERTY predicts alternatively higher and lower reflectances in regions of low and high absorption peaks, respectively. This may be the consequence of the logarithmic nature of the absorption of radiation. LIBERTY is very sensitive to very subtle absorption effects in the highest reflected regions of the wavelength; hence measurement errors of the absorption coefficients will be magnified where the total absorption coefficient is low. This explanation also accounts for the high spectrometer instrument noise in absorption coefficient measurements when reflection was high. Incorporating the predicted in vivo absorption coefficient for combined chlorophyll and carotenoids improves model performance significantly in the visible region, as is demonstrated by the revised quasiinfinite predictions for dry and fresh jack pine needles (Figs. 5b and 6b, respectively) and slash pine needles (Fig. 6c). A comparison of the LIBERTY-predicted single needle ref lectance and transmittance spectra with measured jack pine needles using the in-vivo-estimated pigment Figure 7. Comparison of in vitro and LIBERTY-predicted in vivo absorption in visible wavelengths. Figure 8. Comparison of the measured and LIBERTY-predicted single needle ref lectance and transmittance for jack pine (leaf thickness parameter51.26). absorption demonstrated a high accuracy for the model prediction (RMSE for ref lectance and transmittance values were 1.84% and 1.12%, respectively) (Fig. 8). The leaf thickness parameter was set to 1.62 which corresponded to a ref lectance and transmittance maxima of 53% and 32%, respectively, at a wavelength of 800 nm. Wavelength mismatches between fresh needle and predicted spectra in the near-infrared (Fig. 6a) and red edge (Fig. 8) may be associated with calibration errors between the different spectrometers. Sensitivity studies on the effects of varying lignin/cellulose content with LIBERTY-predicted spectra on both dried (Fig. 9) and fresh leaves (Fig. 10) have been conducted in which the average value of the cellulose/lignin content was varied between 615% of mean (42.6–58.9 g m22). This deviation was chosen as it ref lected normal variations in measured leaf content (Kupiec, 1995; Martin and Aber, 1990). The model output demonstrates that, in fresh leaves, the resultant variation in predicted Figure 9. Sensitivity of the LIBERTY-predicted ref lectance for dry needles to variation in combined cellulose and lignin concentrations. LIBERTY—Conifer Leaf Model Figure 10. Sensitivity of the LIBERTY-predicted ref lectance for fresh needles to variation in combined cellulose and lignin concentrations. 59 The research was funded by the Natural Environment Research Council (Research grant GR3/7647 to P. J. C. and studentship GT4/94/407/L to T. P. D.) and the University of New Hampshire (Accelerated Canopy Chemistry Program grant to P. J. C.). The authors are indebted to many individuals who made this work possible, notably John Kupiec (Scottish Natural Heritage), Geoff Smith (Institute of Terrestrial Ecology, Monks Wood), John Marshall (University of Southampton), and John Aber and Mary Martin (University of New Hampshire). We are particularly grateful to Stephane Jacquemoud (University of Paris), who allowed us to investigate and anatomize his PROSPECT model and utilize some of the model absorption data sets. REFERENCES spectra was a change in overall ref lection in the 1500– 2300 nm wavelengths. No individual absorption peak due to the cellulose and lignin content could be identified, but spectral differences were in agreement with empirical observations (Gao and Goetz, 1994; Dawson et al., 1995b). The problem of trying to observe this same effect in measured leaves and, in particular, to associate it with a particular biochemical change in content, was complicated by strong correlations between the component biochemicals. Further research in this arena, in association with a robust inversion of LIBERTY, is directed at verification of the statistical relationships between leaf biochemical concentration and spectral properties identified by others. CONCLUSION A new model, LIBERTY, has been presented which predicts, quite accurately, the reflectance spectra of both dried and fresh conifer needles. Overall root mean square differences between predicted and measured quasiinfinite reflectance outputs were 2.87% for the freeze-dried needles and 1.73% for fresh needles. For single needle reflectance and transmittance estimates, the RMS errors were 1.84% and 1.12%, respectively. For the freeze-dried needle predictions, LIBERTY mirrors the absorption peaks due to lignin, cellulose, and protein quite accurately in the 1600–2500 nm wavelength region. Leaf water content masks most of the near-infrared absorption features in fresh leaves that are due to biochemical constituents in the 1400–1600 nm wavelengths region. However, a variation of the ref lectance can be observed when content of these biochemicals were altered. Initial inversion studies demonstrated significant increases in accuracy by providing in vivo absorption coefficients. This suggested that it may be possible to estimate biochemical cocentrations using minimization techniques, such as linear programming, on the predicted aggregate absorption coefficient. By providing single leaf ref lectance and transmittance outputs, LIBERTY can be coupled with forest canopy reflectance models and further research is currently being conducted to verify this. Aber, J. D., Ed. (1994), Accelerated Canopy Chemistry Program Final Report to NASA-EOS-IWG, National Aeronautics and Space Administration, Washington, DC. Aldakheel, Y. Y., and Danson, F. M. (1996), Spectral ref lectance of dehydrating leaves: Measurements and modelling. In RSS96: Remote Sensing Science and Industry, Remote Sensing Society, Nottingham, UK, pp. 71–78. Allen, W. A., and Richardson, A. J. (1968), Interaction of light with a plant canopy. J. Opt. Soc. Am. 58:1023–1028. Allen, W. A., Gausman, H. W., Richardson, A. J., and Thomas, J. R. (1969), Interaction of isotropic light with a compact plant leaf. J. Opt. Soc. Am. 59:1376–1379. Allen, W. A., Gausman, H. W., and Richardson, A. J. (1970), Mean effective optical constants of cotton leaves. J. Opt. Soc. Am. 60:542–547. Allen, W. A., Gausman, H. W., and Richardson, A. J. (1973), Willstätter–Stoll theory of leaf reflectance evaluated by ray tracing. Appl. Opt. 12:2448–2543. Benford, F. (1946), Radiation in a diffusing medium. J. Opt. Soc. Am. 36:524–537. Conel, J. E., Van den Bosch, J., and Grove, C. I. (1993), Application of a two-stream radiative transfer model for leaf lignin and cellulose concentrations from spectral ref lectance measurements (Parts 1 and 2). In Proceedings, 4th Annual JPL Airborne Geoscience Workshop. Vol. 1. AVIRIS (R. O. Green, Ed.), Jet Propulsion Laboratory Publication No. 9326, Pasadena, CA, pp. 39–51. Curcio, J. A., and Petty, C. C. (1951), The near infrared absorption spectrum of liquid water. J. Opt. Soc. Am. 41:302–304. Curran, P. J. (1994), Attempts to drive ecosystem simulation models at local to regional scales. In Environmental Remote Sensing from Regional to Global Scales (G. M. Foody and P. J. Curran, Eds.), Wiley, Chichester, pp. 149–166. Curran, P. J., and Kupiec, J. A. (1995), Imaging spectrometry: a new tool for ecology. In Advances in Environmental Remote Sensing (F. M. Danson and S. E. Plummer, Eds.), Wiley, Chichester, pp. 71–88. Daughtry, C. S. T., Biehl, L. L., and Ranson, K. J. (1989), A new technique to measure the spectral properties of conifer needles. Remote Sens. Environ. 27:81–91. Dawson, T. P., Curran, P. J., and Kupiec, J. A. (1995a), Causal correlation of foliar biochemical concentrations with AVIRIS spectra using forced entry linear regression. In Proceedings, 5th Annual JPL Airborne Geoscience Workshop. Vol. 1. 60 Dawson et al. AVIRIS (R. O. Green, Ed.), Jet Propulsion Laboratory Publication No. 95-1, Pasadena, CA, pp. 39–51. Dawson, T. P., Curran, P. J., and Plummer, S. E. (1995b), Modelling the spectral response of coniferous leaf structures for the estimation of biochemical concentrations. In Remote Sensing in Action, Remote Sensing Society, Nottingham, pp. 587–594. Dawson, T. P., Curran, P. J., and Plummer, S. E. (1997), The potential for understanding the biochemical signal in the spectra of forest canopies using a coupled leaf and canopy model. In Physical Measurements and Signatures in Remote Sensing (A. Guyot and T. Phulpin, Eds.) Balkema, Rotterdam, pp. 463–470. Elvidge, C. D. (1990), Visible and near infrared ref lectance characteristics of dry plant materials. Int. J. Remote Sens. 11:1775–1795. French, C. S., Brown, J. S., and Lawrence, M. C. (1972), Four universal forms of chlorophyll-a. Plant Physiol. 49:421–429. Gao, B.-C., and Goetz, A. F. H. (1994), Extraction of dry leaf spectral features from reflectance spectra of green vegetation. Remote Sens. Environ. 47:369–374. Gausmann, H. W. (1974), Leaf ref lectance of near-infrared. Photogramm. Eng. 40:183–191. Grant, L. (1987), Diffuse and specular characteristics of leaf reflectance. Remote Sens. Environ. 22:309–322. Hapke, B. and Wells, E. (1981), Bidirectional ref lectance spectroscopy. 2. Experiments and observations. J. Geophys. Res. 86:3055–3060. Jacquemoud, S., and Baret, F. (1990), PROSPECT: a model of leaf optical properties spectra. Remote Sens. Environ. 34: 75–91. Jacquemoud, S., Ustin, S. L., Verdebout, J., Schmuck, G., Andreoli, G. and Hosgood B. (1996), Estimating leaf biochemistry using the PROSPECT leaf optical properties model. Remote Sens. Environ. 56:194–202. Jeffrey, A. (1986), Mathematics for Engineers and Scientists, Van Nostrand Reinhold, Wokingham. Kemp, W., Ed. (1987), Organic Spectroscopy, Macmillan, London. Kupiec, J. A. (1995), The remote sensing of foliar chemistry, Ph.D. thesis, University College of Swansea, Swansea (unpublished). Lichtenthaler, H. K. (1987), Chlorophylls and carotenoids: pigments of photosynthetic biomembranes. Methods Enzymol. 148:350–382. Martin, M. E., and Aber, J. D. (1990), Effects of moisture content and chemical composition on the near infrared spectra of forest foliage. In Imaging Spectroscopy of the Terrestrial Environment (G. Vane, Ed.), Proceedings SPIE 1298, pp. 171–177. Mackinney, G. (1941), Absorption of light by chlorophyll solutions. J. Biol. Chem. 140:315–322. Melamed, M. T. (1963), Optical properties of powders. Part I. Optical absorption coefficients and the absolute value of the diffuse ref lectance. J. Appl. Phys. 34:560–570. Peñuelas, J., Filella, I., Serrano, L., and Savé, R. (1996), Cell wall elasticity and Water Index (R970nm/R900nm) in wheat under different nitrogen availabilities. Int. J. Remote Sens. 17:373–382. Rock, B. N., Vogelmann, J. E., Williams, D. L., Vogelmann, A. F., and Hoshizaki, T. (1986), Remote detection of forest damage. Bioscience 36:439–445. Schubert, W. J. (1965), Lignin Biochemistry, Academic, New York. Stenberg, P., Kuuluvainen, T., Kellomaki, S., Grace, J. C., Jokela, E. J., and Gholz, H. L. (1994), Crown structure, light interception and productivity of pine trees and stands. Ecol. Bull. 43:20–34. Tucker, C. J., and Garratt, M. W. (1977), Leaf optical system modelled as a stochastic process. Appl. Opt. 16:635–642. Wessman, C. A., Aber, J. D., Peterson, D. L., and Melillo, J. M. (1988), Remote sensing of canopy chemistry and nitrogen cycling in temperate forest ecosystems. Nature 335(8): 154–156. Woolley, J. T. (1971), Ref lectance and transmittance of light by leaves. Plant Physiol. 47:656–662. Yamada, N. and Fujimura, S. (1988), A mathematical model of reflectance and transmittance of plant leaves as a function of chlorophyll pigment content. In Proceedings, IGARSS’88: International Geosciences and Remote Sensing Symposium, IEEE, Piscataway, NJ, pp. 833–834. Zagolski, F., Pinel, V., Romier, J., et al. (1996), Forest canopy chemistry with high spectral resolution remote sensing. Int. J. Remote Sens. 17:1107–1128.
© Copyright 2025 Paperzz