Jean Verdebout

Critique of stepwise multiple linear regression for the extraction of leaf biochemistry information from leaf reflectance data

Abstract This study examined the use of stepwise multiple linear regression to quantify leaf carbon, nitrogen, lignin, cellulose, dry weight, and water compositions from leaf level reflectance ( R ). Two fresh leaf and one dry leaf datasets containing a broad range of native and cultivated plant species were examined using unconstrained stepwise multiple linear regression and constrained regression with wavelengths reported from other leaf level studies and wavelengths derived from chemical spectroscopy. Although stepwise multiple linear regression explained large amounts of the variation in the chemical data, the bands selected were not related to known absorption bands, varied among datasets and expression bases for the chemical [concentration (g g −1 ) or content (g m −2 )], did not correspond to bands selected in other studies, and were sensitive to the samples entered into the regression. Stepwise multiple regression using artificially constructed datasets that randomized the association between nitrogen concentration and reflectance spectra produced coefficients of determination ( R 2 s) between 0.41 and 0.82 for first and second derivative log(1/ R ) spectra. The R 2 s for correctly-paired nitrogen data and first and second derivative log (1/ R ) only exceeded the average randomized R 2 s by 0.02–0.42. Replication of this randomization experiment on a larger dry ground leaf data set from the Harvard Forest showed the same trends but lower R 2 s. All of these results suggest caution in the use of stepwise multiple linear regression on fresh leaf reflectance spectra. Band selection does not appear to be based upon the absorption characteristics of the chemical being examined.

Investigation of leaf biochemistry by statistics

Abstract The biochemical concentration (total protein, cellulose, lignin, and starch) of 73 plant leaves has been related to their optical properties through statistical relationships. Both fresh and dry plant material, leaves and needles, were used in this study. Stepwise multiple regression analyses have been performed on reflectance, transmittance, and absorptance values (individual leaves) as well as on reflectance values of optically thick samples (stacked leaves + needles), on measured values and on transformations of them such as the first derivative or the logarithm of the reciprocal of the reflectance. They underscored good prediction performances for protein, cellulose, and lignin with high squared multiple correlation coefficients (r2) values. Starch, whose concentration in the leaf was smaller compared to the other components, was estimated with less accuracy. As expected, dry material and optically thick samples provided respectively stronger correlations than fresh material and individual leaves.

A new approach to quantifying abundances of materials in multispectral images

The authors tested a new technique for mapping abundances of materials using multispectral images which they call Foreground-Background Analysis (FBA). The method maximizes the contrast between sets of foreground and background spectra while simultaneously minimizing the variability within these sets. Spectral variability introduces errors in the fractions (abundance) of endmembers when simple mixture models (2-5 endmembers) are applied to complex natural surfaces. FBA was tested using CCD-camera measurements of known materials in the laboratory and included variations in illumination geometry. Lower abundance uncertainties were obtained using FBA than with simple mixture models.<<ETX>>

Optical Properties of Leaves: Modelling and Experimental Studies

This paper deals with the interpretation of leaves spectra following an approach based on modelling and laboratory studies. First, the leaves structure and principal constituents are described together with the way they interact with light. The effects of growth, senescence and environmental factors on the leaf optical properties are summarised. A laboratory study conducted on drought stress of maize (Zea Mays) plants is reported as an example. A succinct review of the existing models is then made: ray tracing, Kubelka-Munk and developments, plate models, and the stochastic model. The use of these models to determine leaf constituents and structure by inversion on reflectance spectra is then discussed with an emphasis on the research of good specific absorption coefficients for the constituents. The validation of the PROSPECT model (generalised plate model) on the basis of leaves spectra is presented. The problems linked with the application of these procedures to remote sensing data is evoked, and an example of inversion on experimental spectra of sugar beet (Beta vulgaris L) fields is briefly reported.

The European optical goniometric facility: technical description and first experiments on spectral unmixing

Presents the new European goniometric facility installed at the Joint Research Center of the Commission of the European Union. Basically, this system allows independent positioning of a light source and a detector anywhere on a 2 m radius hemisphere centered on the target and thereby to perform bidirectional reflectance measurements. The control software allows full programming of an experiment, the graphical visualization of the measurement geometry and the interface to a dedicated data base for the experimental results. A first experiment is described in which multispectral CCD camera images were acquired on various targets (soils, canopies and synthetic leaves) under different illumination angles. The purpose was to provide a data set allowing the comparison of the projected area of spectral endmembers, as measured from the CCD images, with the result of linear spectral mixture inversion algorithms.<<ETX>>

Long-term UV irradiance changes over Moscow and comparisons with UV estimates from TOMS and METEOSAT

We analyzed long-term variations of UV irradiance 300-380 nm over Moscow 55.7N, 37.5E since 1968 using a complex dataset that includes ground-based UV measurements, UV retrievals from two satellites, and the results of a previously developed empirical model. Long-term interannual changes of UV irradiance, 300-380nm, during 1968-2003 show the absence of any linear trends although an increase is detected in the late 90-s due to cloud amount and aerosol content decrease. The ground-based data are compared with UV satellite retrievals from two independent methods as well as with the results of an empirical model that accounts for the physical dependence of UV on cloud parameters (amount and optical thickness), surface albedo, total ozone, and aerosol properties of the atmosphere. UV datasets over Moscow obtained from different satellite instruments: from the Total Ozone Mapping Spectrometer (TOMS) data (version 8) since 1979 and from METEOSAT/MVIRI since 1984. The original METEOSAT processor, using visibility observations at a nearby meteorological station to quantify the aerosol load, leads to a significant underestimation of the UV daily doses (-23% in warm period and -31% in cold period). Substituting the visibility observations by in situ monthly mean aerosol optical depth improves significantly the agreement in both warm and cold periods (respectively, -9% and -10%) but the bias still remains. The difference between TOMS UV retrievals and ground-based data has different signs in warm (+6%) and cold (-15%) periods. Applying off-line absorbing aerosol correction in TOMS UV retrievals eliminates the positive bias in warm period. The negative bias during the cold period can be due to the application of minimum Lambertian effective reflectivity (MLER) approach to determine the surface albedo especially in conditions with non stable snow cover (end of February- March, and November-December). Model reconstruction of UV variability demonstrates high correlation with aerosol corrected satellite UV retrievals (0.83-0.94) as well as with ground data (0.82) during warm period. During cold months the correlation between satellite UV retrievals and ground-based measurements is much worse.

Laser-induced time-resolved fluorescence of vegetation

Time-resolved measurements of the laser-induced chlorophyll fluorescence emission of vegetation detected by two different techniques are described. Fluorescence decay time measurements using single photon counting and picosecond laser pulse excitation have been used to analyze the fluorescence heterogeneity of plant leaves. The fluorescence is described by lifetimes of 10-40, 80-150, 400-500, and 700-1000 ps. By closing the reaction centers via application of the herbicide DCMU, the lifetimes of the two slowest components increase by a factor of about three. Another possible method to monitor the fluorescence after picosecond excitation could be a streak camera detection system. Measurements performed on the slow decay component of stressed and unstressed plants are presented. >

Analysis of imaging spectrometer data to evaluate the biochemical content of vegetation, based on the results of a laboratory experiment

The reflectance spectrum of green vegetation is mainly determined by the leaf content in chlorophyll and water, the respective spectral signatures being modulated by the structural characteristics of the leaf and the canopy. However, correlation between the reflectance spectra and the leaf content in such constituents as lignin, cellulose, nitrogen, starch, etc. has been demonstrated. To further study this question, we have constituted a data set associating high quality spectra (of single leaves, optically thick stacks, needles, stalks, on fresh and dried material) with a number of physical and chemical measurements (leaf thickness, water content, chlorophyll, carotenoids, cellulose, lignin, proteins, nitrogen, starch). This data set has been used to investigate the link between the optical properties and the composition, focusing on the biochemical components. Two approaches have been followed: the classical regression analysis and modeling based on the Kubelka-Munk formula. In the latter case, empirical specific absorption coefficients have been determined and then used to decompose the infrared spectra in water and other components contributions. This method is successful in retrieving the relative water content but does not yet allow us to estimate the biochemical content. It was applied both on laboratory and AVIRIS spectra.

Multitemporal AVIRIS-images of forested and agricultural units in southern Germany

An analysis, based on the inversion of a simple non-linear model of the ground reflectance, was conducted on several AVIRIS scenes. The scenes were acquired during the MAC EUROPE 91 campaign on the 5th and 22nd of July, over two test sites (Black Forest and Freiburg). The model consists in a linear mixing of the soil reflectance and a green vegetation reflectance described with a Kubelka-Munk formula containing the chlorophyll and water specific absorption coefficients. Its inversion provides a Green Vegetation Fraction of the pixel and two parameters related respectively to chlorophyll and water. The model can then be used to evaluate the magnitude of the 1.7 micrometers absorption feature which is thought to be a signature of the vegetation biochemical components. The spatial and temporal variability of this feature over the scenes is commented.17

Surface solar irradiance maps from seviri MSG data

A scheme for solar irradiance retrieval for the Europe from Meteosat Second Generation (MSG) data is here proposed. The surface irradiance is obtained by interpolation of multidimensional LUT which is generated using the radiative transfer code libRadtran. Satellite model assessment is based on set of 25 ground stations with measurements of global solar irradiance from high quality measurements networks.