The potential of the measure of the Bidirectional Reflectance Distribution Function (BRDF) of terrestrial surfaces has been demonstrated in recent years for several objectives. These include the correction of directional effects in time series of vegetation indices and reflectances from coarse resolution sensors (Leroy and Roujean, 1994; Wu et al., 1995), and the direct use of directional effects for the estimation of leaf area index and other biophysical parameters by inversion of radiative transfer models (Privette et al., 1994; Knyazikhin et al., 1998; Bicheron and Leroy, 1999), albedo retrieval (Wanner et al, 1997; Capderou, 1998; Strugnell and Lucht, 2001), land cover classifications (Abuelgasim et al., 1996; Bicheron et al., 1997; Hyman and Barnsley, 1997), and radiance to flux conversion factors for Earth Radiation Budgets studies (Manolo-Smith et al., 1998; Liang et al., 2000).
The development of these various applications at large scale (regional or global) requires that the intrinsic spatial and temporal variations of the BRDF be documented, for different types of biomes and at different seasons. The BRDF has been measured in the field (Kimes, 1983; Kuusk, 1991; Deering et al., 1992; Sandmeier and Strahler, 2000) or from airborne instruments (Irons et al., 1991; Leroy and Bréon, 1996), with most often an adequate sampling of directional space, but with a poor spatial coverage. Directional effects on land surfaces have been seen from space with AVHRR (Gutman, 1987; Roujean et al., 1992a; Burgess and Pairman, 1997), or more recently with the along track scanning radiometer (ATSR-2) launched on ERS-2 (Godsalve, 1995 ) and MODIS launched on TERRA and AQUA platforms. The space coverage is potentially adequate, but the sampling of the BRDF is limited to the across-track (AVHRR) or along track (ATSR-2) plane of acquisition.
The spaceborne POLDER instrument (Deschamps et al., 1994) offers new opportunities to sample the BRDF of every point on Earth for viewing angles up to 60-70°, and for the full azimuth range at a spatial resolution of about 6 km, when the atmospheric conditions are favorable (Hautecoeur and Leroy, 1998). A first version of the instrument has delivered 8 months of global data from November 1996 to June 1997 onboard the Japanese ADEOS platform. A second version of the instrument is foreseen to be launched on the ADEOS-2 platform in November 2000.
The objective of this work is to document with POLDER / ADEOS 1 data the BRDF of every biome on Earth, and also at several seasons whenever possible, so as to serve for the development and prototyping of science applications of the BRDF measure. The biomes are chosen on the basis of the 17 land cover classes of the IGBP 1-km land cover classification, called DISCover land cover data set (Loveland and Belward, 1997). 395 BRDF data sets have been collected in the blue, red and near infrared bands of POLDER. These data are available to the science community (http://smsc.cnes.fr/POLDER/indexbr.htm). A more global POLDER/ADEOS 1 database of almost 23000 BRDF over 8 months will be soon produced and available through the POLDER Web site.
2. Instrument and data processing
2.1 The POLDER measurements
The POLDER instrument is a radiometer designed to measure the directionality and polarization of the sunlight scattered by the ground atmosphere system. The instrument is made of a bidimensional CCD matrix, a rotating wheel that carries filters and polarizers, and a wide field of view lens (114°). The field of view seen by the CCD matrix is ± 43° along track and ± 51° across track. The view zenith angles seen at surface level are larger due to Earth curvature, ± 50 ° along track and ± 61 ° across track (± 70° in the matrix diagonal). The pixel size on the ground is about 6 km for an ADEOS altitude of 800 km. The rotating wheel carries filters which allow spectral measurements at 8 wavelengths (443, 490, 565, 670, 763, 765, 865, and 910 nm). Three of the channels (443, 670 and 865 nm) measure the polarization of the incident light. Images of the same band are acquired every 19.6 s, which permits a large overlap between successive images. During the satellite overpass, a surface target is viewed up to 14 times with each time a different viewing angle. The directional configuration changes each day due to orbital shift between successive days. Therefore, after a few days, assuming favorable atmospheric conditions, the slices of measurements provide a sampling of the BRDF in the limits of the instrument field of view.
Several science products, in the fields of land surfaces, clouds, aerosols and ocean color are derived from the POLDER measurements (Deschamps et al., 1994; Leroy and Lifermann, 1999). For land surfaces, the global set of POLDER data is processed so as to obtain so-called level 2 data, that is, reflectances geocoded, calibrated, cloud screened and atmospherically corrected for each orbit (Leroy et al., 1997; Bréon and Colzy, 1999). To declare if a pixel is cloudy or not, cloud screening is made of three tests: derivation of the pressure of the main reflectors (surface or clouds) based on the ratio of the channels at 763 and 765 nm, polarized reflectance exhibiting large value for clouds at scattering angles close to 140 °, and a threshold at 443 nm where the cloud reflectance is brighter than the soil or vegetation reflectance. If the pixel is recognized as clear, a correction for stratospheric aerosols and absorbing gases (O3, O2, and H20) is applied. The ozone correction is made with TOMS/ADEOS data. A correction accounting for molecular scattering correction is applied. At this stage, no correction for tropospheric aerosols is yet taken into account.
Global maps at 10-day frequency of BRDF model parameters, hemispherical reflectances, anisotropy corrected vegetation indices, and quality indices such as the number of cloud-free observations are also produced (so-called level 3 data) (Leroy et al., 1997; Leroy and Hautecoeur, 1999). They result from the adjustment, for each spectral band and at a full space resolution, of a 30 day time series of directional reflectances and the 3-parameter semi-empirical BRDF model of Roujean et al. (1992).
2.2 The global land cover dataset
The DISCover data set, developed in the framework of the IGBP core project Data Information System (DIS), has been processed through an unsupervised classification using one year of monthly NDVI 1-km AVHRR data acquired from March 1992 through February 1993. A post classification is then applied with the addition of digital elevation model, ecoregions data, and a collection of other land cover/vegetation reference data (Brown et al., 1993). The DISCover has not been yet completely validated, but represents the first opportunity of a global landcover at this resolution. Several degrees of refinement in the number of classes of the classification exist, ranging from the general 17 IGBP classes to the more precise nomenclature of the 84 classes of Olson (1994). The present work is based on the 17 IGBP classes, from which the water body class has been excluded. The classes under study are therefore: 1) Evergreen needleleaf forest, 2) Evergreen broadleaf forest, 3) Deciduous needleleaf forest, 4) Deciduous broadleaf forest, 5) Mixed forest, 6) Closed shrublands, 7) Open shrubland, 8) Woody savannas, 9) Savannas, 10) Grasslands, 11) Permanent wetlands, 12) Croplands, 13) Urban and built-up, 14) Cropland/natural vegetation mosaic, 15) Snow and Ice, 16) Barren or sparsely vegetated.
2.3 Methodology for the construction of the BRDF dataset
The basic input for the construction of the BRDF data set is level 2 surface reflectance data over two periods of two months (November and December 1996, May and June 1997). The choice of two periods allows to observe the temporal changes of the BRDF, due to the seasonal behaviour of vegetation and to the variations of sun elevation. The location of the pixels for the BRDF extraction results from the cross-consideration of level 3 maps of numbers of cloud-free observations, and of the DISCover land cover map. For each month, the level 3 map is used to extract large areas (500 - 1000 km) with a sufficient number of clear days (typically larger than 6) and thus a good sampling of directional space. Within these large areas, the DISCover land cover map is used to select homogeneous zones (50 - 100 km) of 1 km pixels pertaining to the same class. The BRDF is then acquired for the central pixel of selected zones at a full space resolution during one month. The choice of zones contains some part of arbitrary. It is such that it satisfies an approximate balance between classes, and that a geographical coverage as even as possible is obtained. At this point 980 BRDF data are selected.
Among all the BRDFs, we have only chosen those presenting the smallest possible noise, and with a sufficiently even sampling of directional space so as to be able to extract consistent principal and perpendicular planes. The principal sources of noise are expected to originate from partial cloudiness, undetected by cloud screening tests, and aerosol events. Practically, the data are processed as follows. Sections of the BRDF in specific planes of observation are retrieved from the BRDF data. Are said to belong to the principal (resp. perpendicular) plane the data points which belong to 16°-wide bands of directional space centered around the principal (resp. perpendicular) planes. As in Hautecoeur and Leroy (1998), fitted polynomial curves are superimposed to the data points to outline the general shape of the directional signatures. They are designed to be symmetric about the origin in the perpendicular plane, and, in the principal plane, to fit the data located on each side of the data point closest to the sun direction. Noisy data, contaminated by partial cloudiness or aerosol events, are expected to have a substancially higher level at 443 nm than uncontaminated data. The data selection has been such that the data for which the surface reflectance in the perpendicular plane of any orbit exceeds the value of the fitted polynomial by more than 0.025 have been discarded. The whole process is visually controlled so as to ensure that the directional sampling permits adequate reconstruction of the principal and perpendicular planes. As a result of this selection procedure, we obtain 395 BRDF signatures for the 16 IGBP land cover classes. The data are available to the science community at the Internet address mentioned in the Introduction. An individual BRDF data set comprises level 2 surface reflectance data for a given 6 km resolution pixel, at 443 nm, 670 nm, and 865 nm, with associated sun and view zenith angles, and relative azimuth between view and sun directions.
3. Data format
Each BRDF file is characterized with its month of synthesis and its pixel location in the POLDER grid. Each file contains 7 columns. This is a sample of one of the BRDF acquired from POLDER for the IGBP classe 1 over Eurasia.
File name : brdf_may_520_3429
4. Requested form for aknowledgement
Cite the following publication: P. Bicheron and M. Leroy, " BRDF signatures of major biomes observed from space ", published in J. of Geo. Res. 116, D21, 26669-26681, 2000.
Abuelgasim A., S. Gopal, J.R. Irons, and A.H. Strahler, " Classification of ASAS multiangle and multispectral measurements using artificial neural networks ", Rem. Sen. Env., 57, 79-87, 1996.
Bicheron P., and M. Leroy, " A method of biophysical parameter retrieval at global scale by inversion of a vegetation reflectance model ", Rem. Sen. Env., 67, 251-266, 1999.
Bicheron P., M. Leroy, O. Hautecoeur, and F.M. Bréon, " Enhanced discrimination of boreal forest covers with directional reflectances from the airborne polarization and directionality of Earth reflectances (POLDER) instrument ", J. Geophys. Res., 102, D-24, 29517-29528, 1997.
Brown J.F., T.R. Loveland, J.W. Merchant, B.C. Reed, and D.O. Ohlen, " Using multisource data in global land cover characterization : concepts, requirements and methods ", Photogramm. Eng. Rem. Sen., 59, 977-987, 1993.
Capderou M., " Determination of the shortwave anisotropic function for clear sky desert scenes from ScaRab data : comparison with models issued from other satellite data ", J. Applied Met., 37, 1398-1911, 1998.
Chen J., " Canopy architecture and remote sensing of the fraction of photosynthetically active radiation absorbed by boreal conifer forests ", IEEE Trans. Geosci. Rem. Sen., 34, 6, 1996.
Chen J., and J. Cihlar, " A hot-spot function in a simple bidirectionnal reflectance model for satellite applications ", J. Geophys. Res., 102, D 22, 25907-25913, 1997.
Cosnefroy H., M. Leroy, X. Briottet, " Selection and characterization of sahara and arabia desert sites for the calibration of satellites sensors ", Rem. Sen. Env., 58, 1, 101-114, 1996.
Deschamps P.Y., F.M. Bréon, M. Leroy, A. Podaire, A. Bricaud, J.C. Buriez, G. Sèze, " The POLDER mission : instrument characteristics and scientific objectives ", IEEE Trans. Geosci. Rem. Sen., 32, 598-615, 1994.
Engelsen O., B. Pinty, M. Verstraete, and J. Martonchik, " Parametric bidirectional reflectance factors model : evaluation, improvements, and applications ", Technical Report EUR, 16426 EN, EC Joint Research Centre.
Goldsave C., " Bidirectional reflectance sampling by ATSR-2 : a combined orbit and scan model ", Int. J. of Rem. Sen., 16, 269-300, 1995.
Gutman G.G., " The derivation of vegetation indices from AVHRR data ", Int. J. of Rem. Sen., 10, 107-132, 1989.
Hautecoeur O., and M. Leroy, " Surface bidirectional reflectance distribution function observed at blobal scale by POLDER/ADEOS ", Geophy. Res. Let., 25, n°22, 4197-4200, 1998.
Hyman A.H., and M.J. Barnsley, " On the potential for land cover mapping from multiple view angle (MVA) remotely sensed information ", Int. J. Rem. Sen., 18, 2471-2475, 1997.
Irons J.R., K.J. Ranson, D.L. Williams, R.R. Irish, and F.G. Huegel, " An off-nadir pointing imaging spectroradiometer for terrestrial ecosystem studies ", IEEE Trans. Geosci. Rem. Sen., 29, 66-74, 1991.
Justice C.O., J.G.R. Townshend, B.N. Holben, and C. J. Tucker, " Analysis of the phenology of global vegetation using meteorological data ", Int. J. of Rem. Sen., 6, 8, 1271-1318, 1985.
Kimes D.S., " Dynamics of directional reflectance factor distributions for vegetation canopies ", App. Opt., 22, 1364-1373, 1983.
Knyazikhin Y., J.V. Martonchik, D.J. Diner, R.B. Myneni, M.M. Verstraete, B. Pinty, and N. Gobron, " Estimation of vegetation canopy leaf are index and fraction of absorbed photosynthetically active radiation from atmosphere-corrected MISR data ", J. Geophys. Res., 103, D 24, 32239-32256, 1998a.
Kuusk A., " Determination of vegetation canopy parameters from optical measurements ", Rem. Sen. Env., 37, 207-218, 1991.
Leroy M., and F.M. Bréon, " Surface reflectance angular signatures from airborne POLDER data ", Rem. Sen. Env., 57, 97-107, 1996.
Leroy M., J.L. Deuzé, F.M. Bréon, O. Hautecoeur, M. Herman, J.C. Buriez, D. Tanré, S. Bouffiès, P. Chazette, and J.L. Roujean, " Retrieval of atmospheric properties and surface bidirectional reflectances over land from POLDER/ADEOS ", J. of Geophys. Res., 102, 17023-17037, 1997.
Leroy M. and O. Hautecoeur, 'Anisotropy-corrected vegetation indices from POLDER/ADEOS', IEEE Transactions on Geoscience and Remote Sensing, 37, 1698-1708, 1999.
Leroy M., and A. Lifermann, 'The POLDER instrument onboard ADEOS : scientific expectations and first results', Proceedings of the COSPAR 98 conference of Nagoya, to be published in Advance of Space Research, 1999.
Liang S., J.C. Stroeve, I.F. Grant, A.H. Strahler and J.P. Duvel, 'Angular corrections to satellite data for estimating Earth Radiation Budget', Rem. Sen. Rev., 18, 103-136, 2000
Loveland T.R., and A.S. Belward, " The IGBP-DIS global 1-km land cover data set DISCover : first results ", Int. J. of Rem. Sen., 18, 3289-3295, 1997.
Olson J.S., " Global ecosystem framework definitions ", USGS EROS data center internal report, Sioux Falls, SD, 1994.
Privette J.L., " Inversion of a vegetation reflectance model with NOAA/AVHRR data ", Rem. Sen. Env., 58, 187-200, 1996.
Privette J.L., R.B. Myneni, C.J. Tucker, and W.J. Emery, " Invertibility of an 1-D discrete ordinates canopy reflectance model ", Rem. Sen. Env., 48, 89-105, 1994.
Rahman H, B. Pinty, and M. Verstraete, " Coupled surface atmosphere reflectance (CSAR) model. 2 Semi-empirical surface model usable with NOAA advanced very high resolution radiometer data ", J. Geophys. Res., 98, 20791-20801, 1993.
Roujean J.L., M. Leroy, A. Podaire, and P.Y. Deschamps, " Evidence of surface reflectance bidirectional effects from a NOAA/AVHRR multitemporal dataset ", Int. J. Rem. Sen., 13, 685-698, 1992.
Roujean J.L., M. Leroy, and P.Y. Deschamps, " A bidirectional reflectance model of the Earth's surface for the correction of remote sensing data ", J. Geophys. Res., 97, 20455-20468, 1992.
Sandmeier S.R., and A. H. Strahler, 'BRDF Laboratory measurements', Rem. Sen. Rev., 18, 481-502, 2000.
Strugnell N., and W. Lucht 'Continental scale albedo inferred from AVHRR data, land cover class and field observations of typical BRDFs', J. Climate, 14, 1360-1376, 2001.
Wanner W, A.H. Strahler, B. Hu, et al., " Global retrieval of bidirectional reflectance and albedo over land fro EOS/MODIS and MISR data ", J. Geophys. Res., 102, 17143-17162, 1997.
Wanner W., X. Li, and A.H. Strahler, " On the derivation of kernels-driven models of bidirectional reflectance ", J. Geophys. Res., 100, D 10, 21077-21089, 1995.
Wu A., Z. Li, and J. Cihlar, " Effects of land cover type and greeness on advanced very high resolution radiometer bidirectional reflectances : analysis and removal ", J. Geophys. Res., 100, 9179-9192, 1995.