National albedo production on THEIA (opendata)
- Creators
- Theia
Description
The Sentinel-2 (S2) high-resolution albedo (HR) calculation method has already been used as part of the Copernicus Global Land Service program (http://land.copernicus.eu/global/), considering PROBA-V data (Roujean et al., RSE, 2018, https://doi.org/10.1016/j.rse.2018.05.015). Here, it uses MAJA L2A data as input and comprises two steps. Firstly, a spectral dataset is built up by accumulating S2 scenes over time to provide directional measurements for inverting a model of the Bidirectional Reflectance Distribution Function (BRDF). Due to the reduced field of view of S2, directionality is mainly due to the variability of solar geometry, and thus the quality of the HR albedo product may vary according to the length of the chosen time period. Secondly, an BRDF model is inverted on each multi-temporal spectral directional dataset and used to simulate reflectance in all directions. Integration of the BRDF model over the space of viewing directions provides the directional hemispheric reflectance (DHR), and integration of the DHR albedo over the space of illumination directions provides the bidirectional hemispheric reflectance (BHR). This is a DHR albedo product corresponding to local solar noon. It is supplied with an uncertainty that reflects the cumulative error in its realization. The final step is to convert the spectral albedos of S2 into a solar albedo. The conversion law is obtained from simulations of a radiative transfer model including thousands of plant cover types. This work was carried out as part of the CNES SCO (Space Climate Observatory) program, through the sat4BDNB project run in collaboration between CESBIO and CSTB (Centre Scientifique et Technique du Bâtiment) and based on the French National Building Database (BDNB). The chosen period is from June 1 to September 30, 2022. An average RH albedo is generated over this period for the whole of mainland France, using a tiling and geographical projection based on the MAJA L2A product. The product has been verified by inter-comparison with measurements made from albedometers positioned on 6 French grasslands equally distributed across the country. The statistical results give a bias of -0.01 and a root mean square error of 0.02 in reflectance units. The albedo value was multiplied by a factor of 10000 for archiving purposes.