A Pilot Study of 10-Day Composites From the NASA Harmonized Landsat and Sentinel-2 Images for Terrestrial Monitoring

To meet the demand for more frequent medium resolution land observation imagery, the United States National Aeronautics and Space Administration’s Harmonized Landsat and Sentinel-2 (HLS) project creates comparable 30-m surface reflectance data products from Landsat 8/9 and Sentinel-2A/B Top-of-Atmosphere measurements, by applying a series of harmonization algorithms that include atmospheric correction, cloud masking, view angle effect normalization, bandpass adjustment on Sentinel-2, and gridding in the same spatial reference system. Typically, over a hundred HLS images are available in a year for any location when all four sensors are in operation. Although cloud and cloud shadow pixels are masked in the HLS, they are retained in the distributed data, resulting in unnecessarily large data volume, and there are also a small proportion of cloud and cloud shadow pixels that were undetected. The high data volume and cloud mask omission problem combine to impose a challenge for many applications. In this pilot study we produce 10-day image composites from individual HLS observations. To eliminate the residual cloud and cloud shadow pixels, our composite algorithm takes advantage of the aerosol information derived during the atmospheric correction, the distance to the detected cloud and cloud shadow, and a few spectral screening rules. Following the screening of the observations, a maximal value Enhanced Vegetation Index 2 (EVI2) rule is used to select the best observation for each pixel location. The 10-day HLS image composites were created globally for a few growing season months, and at selected location for a few years. They are in general free of cloud and cloud shadow contamination and spatially complete, with a reduced data volume and increased overall data quality. The composited HLS images are made freely available in NASA's Visualization, Exploration, and Data Analysis (VEDA) environment for users’ evaluation and feedback on the composite window length and data quality.