Harmonizing Landsat 8 and Sentinel-2: A time-series-based reflectance adjustment approach

Abstract

We developed a Time-series-based Reflectance Adjustment (TRA) approach for reducing the reflectance differences between Landsat 8 and Sentinel-2 observations. This TRA approach used the time series of matched Landsat 8 and Sentinel-2 observations to build linear regression models to adjust reflectance differences between the two sensors for each individual pixel and each spectral band. We evaluated this approach for the NASA harmonized Landsat and Sentinel-2 (HLS) surface reflectance product (V1.4; https://hls.gsfc.nasa.gov/data/v1.4/) and top-of-atmosphere (TOA) reflectance with approximately 4 years of temporal coverage at five Military Grid Reference System (MGRS) tiles. Using this approach, the surface reflectance difference between Landsat 8 and Sentinel-2 in the HLS product reduced 45% for the blue band, 42% for the green band, 38% for the red band, 30% for the Near Infrared (NIR) band, 37% for the Shortwave Infrared (SWIR) 1 band, and 32% for the SWIR2 band. The TRA approach also reduced TOA reflectance difference between Landsat 8 and Sentinel-2 substantially, in which the blue band reduced 46%, the green and NIR bands reduced 42%, the red band reduced 48%, and the SWIR1 and SWIR2 bands reduced 44%. If the high aerosol observations were screened, the reflectance differences in the HLS product could be further reduced by 2–4% and the TOA reflectance differences could be further reduced by 3–6% for the six spectral bands. The TRA approach has also shown good results in reserving the spatial patterns and the heterogeneity of land surface. The transformation parameters estimated from the TRA approach can be directly used for future Landsat 8 and Sentinel-2 reflectance adjustment, with slightly lower (5%) reduction of reflectance difference.