Evaluating land surface albedo estimation from Landsat MSS, TM, ETM +, and OLI data based on the unified direct estimation approach

Abstract

Surface albedo is widely used in climate and environment applications as an important parameter for controlling the surface energy budget. There is an increasing need for albedo data to be available for use in applications that require a fine spatial resolution and for validating coarse-resolution datasets; however, such products with long-term global coverage are not available thus far. Existing algorithms for Landsat albedo estimation all require surface reflectance from explicit and reliable atmospheric correction, which may sometimes be unavailable or carry uncertainties due to saturated visible bands or a lack of dense vegetation. In addition, most of the existing algorithms require concurrent clear-sky observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) for bidirectional reflectance distribution function (BRDF) correction, which limited the data availability for Landsat albedo estimation. To overcome these problems, in this study, we adopt the direct estimation approach previously used with coarser resolution data, such as MODIS and Visible Infrared Imaging Radiometer Suite (VIIRS), and apply it to multiple Landsat data obtained by Multispectral Scanner (MSS), Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+), and Operational Land Imager (OLI). By incorporating Landsat spectral response functions and a database of bidirectional reflectance distribution function (BRDF) into radiative transfer simulations, a unified algorithm is developed to estimate surface albedo directly from the Landsat top-of-atmospheric reflectance data obtained by MSS, TM, ETM+, and OLI with few ancillary inputs. To overcome the saturation problems in the visible bands of TM and ETM+ over very bright surfaces, a refined approach is employed by using only non-saturated bands. The validation results against ground measurements over various land cover types and climate regions show that our algorithm is effective for both snow-free and snow-covered surfaces and can achieve root-mean-square errors (RMSEs) of not more than 0.034. In addition, we show the high potential of the earlier MSS data for producing consistent surface albedo estimations based on inter-comparison with TM-based results with RMSEs of 0.011–0.017 and R2 of 0.858–0.963. This long-term, 30-m resolution surface albedo estimation can date back to the early 1980s, which allows for improved understanding of long-term climate change and land cover change effects.