The same atmospheric correction algorithm is applied to both sensors data — Land Surface Reflectance Code (LaSRC), developed by Eric Vermote (NASA/GSFC) (Vermote et al. 2016). LaSRC is based on the 6S radiative transfer model and a heritage from the MODIS MCD09 products (Vermote and Kotchenova 2008) as well as the earlier LEDAPS algorithm implemented for Landsat-5 and Landsat-7 (Masek et al. 2006). A detailed description of the method is given in Vermote et al. (2016), and results of surface reflectance validation for Landsat 8 and Sentinel-2 within CEOS ACIX-I are provided in Doxani et al. (2018). HLSv1.4 uses LaSRC v3.5.5.

HLS provides per-pixel cloud, shadow, snow, and water masks. The Landsat-8 mask of cloud, cloud-shadow, snow and water is a union of the mask derived from the LaSRC, and the mask in the USGS Landsat TOA data (i.e. the BQA file). The Sentinel-2 mask of cloud, cloud-shadow, snow and water mask is a union of LaSRC mask and the mask generated from the Fmask algorithm which has been adapted from (Zhu et al. 2015). Fmask is run on 30 m aggregated TOA reflectance.

 true color composite (bands 4-3-2) of Sentinel-2B image

Figure: A true color composite (bands 4-3-2) of Sentinel-2B image with top-of-atmosphere (TOA) reflectance (left) and surface reflectance after atmospheric correction (right) with Land Surface Reflectance Code (LaSRC). The image is acquired on November 17, 2017 over tile 18SUJ in USA.