The ESA globAlbedo project: Algorithm

P. Lewis, L Guanter, G. Lopez Saldana, J-P. Muller, G. Watson, N. Shane, T. Kennedy, J. Fisher, C. Domenech, R. Preusker, P. North, A. Heckel (+6 others)
2012 2012 IEEE International Geoscience and Remote Sensing Symposium  
A land surface broadband albedo map of the entire Earth's land surface (snow and snow-free) is required for use in Global Climate Model initialisation and verification. A group of 10 users have been selected to work with the GlobAlbedo* Implementation team to define requirements and drive the project towards practical applications of the product. These requirements defined the need to generate a final product on 8-daily at spatial resolutions of 1km in sinusoidal projection using the MODIS 10º
more » ... 10º tiling scheme and 0.05º and 0.5º on monthly time-steps. To generate such a global map by temporal compositing requires both sufficient directional looks and the very precise correction of top-of-atmosphere radiances to "at surface" directional reflectances (SDRs). In addition, such a map requires precise radiometric calibration and inter-calibration of different sensors [1] and the computation of radiative transfer coefficients to derive broadband SDRs from different input narrowband SDRs and given sufficient angular sampling from all the directional looks within a given temporal window, derive a suitable BRDF. This BRDF can be integrated to produce DHR (Direct Hemispherical Reflectance known as "black-sky") and BHR (BiHemispherical Reflectance, known as "white-sky") [2]. The final albedo product has been integrated in three spectral broadband ranges, namely the solar spectrum shortwave (400-3000nm), the visible PAR region (400-700nm) and the near-and shortwave-infrared (700-3000nm). In addition, maps of normalized difference vegetation index (NDVI) and the fraction of absorbed photosynthetically active radiation (fAPAR) will be generated consistent with the albedo product to complement the Globalbedo data set for analysis of vegetation-related processes [3] . To achieve the aim of deriving independent estimates using European only assets, GlobAlbedo set out to create a 15 year time series by employing SPOT4-VEGETATION and SPOT5-VEGETATION2 as well as MERIS. Legacy algorithms for deriving SDRs using an optimal estimation approach are outlined [2] as well as a novel system for gap-filling using ten year mean estimates derived from equivalent BRDFs from MODIS [2]. Each and every output pixel albedo value has an estimated uncertainty associated with it and the corresponding BRDF a full uncertainty matrix for each pixel. Separate BRDFs are computed for snow and snow-free pixels and combined together to yield a gap-free dataset. An example of a sample output product browse in Figure 1 shows the BHR and the coefficient of variation derived from the uncertainty divided by the expectation value (loc.cit.)
doi:10.1109/igarss.2012.6352306 dblp:conf/igarss/LewisGSMWSKFDPNHDKZFBS12 fatcat:226li7casrahhdnshvuhswxptu