An approach for the long-term 30-m land surface snow-free albedo retrieval from historic Landsat surface reflectance and MODIS-based a priori anisotropy knowledge

Authors: SHUAI, YANMIN - National Aeronautics And Space Administration (NASA); MASEK, JEFFREY - National Aeronautics And Space Administration (NASA); Gao, Feng; SCHAAF, CRYSTAL - University Of Massachusetts; TAO, HE - University Of Maryland

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/17/2014
Publication Date: 8/3/2014
Publication URL: http://handle.nal.usda.gov/10113/59890
Citation: Shuai, Y., Masek, J., Gao, F.N., Schaaf, C., Tao, H. 2014. An approach for the long-term 30-m land surface snow-free albedo retrieval from historic Landsat surface reflectance and MODIS-based a priori anisotropy knowledge. Remote Sensing of Environment. 152:467-479.

Interpretive Summary: Land surface albedo is a key parameter driving Earth’s climate. Although the Moderate Resolution Imaging Spectroradiometer (MODIS) provides albedo products at 500-m spatial resolution globally, they are too coarse to characterize albedo evolution of terrestrial biomes happening at a fine scale. This paper presents a “pre-MODIS era” approach to generate albedo at 30-m spatial resolutions for earlier Landsat data (pre-2000). The approach extends the previously developed “MODIS-concurrent” approach (post-2000) and completes the series of approaches to retrieve long-term Landsat albedo since the 1980s from historic Landsat archives as well as the on-going Landsat-8 mission. Understanding the albedo evolution of land cover and land use change at field scales is critical for the USDA to evaluate the impacts and to develop mitigation strategies for sustainable agricultural production.

Technical Abstract: Land surface albedo has been recognized by the Global Terrestrial Observing System (GTOS) as an essential climate variable crucial for accurate modeling and monitoring of the Earth’s radiative budget. While global climate studies can leverage albedo datasets from MODIS, VIIRS, and other coarse-resolution sensors, many applications in heterogeneous environments can benefit from higher-resolution albedo products derived from Landsat. We previously developed a “MODIS-concurrent” approach for the 30-meter albedo estimation which relied on combining post-2000 Landsat data with MODIS Bidirectional Reflectance Distribution Function (BRDF) information. Here we present a “pre-MODIS era” approach to extend 30-m surface albedo generation in time back to 1980s through an a priori anisotropy Look-Up Table (LUT) built up from the high quality MCD43A BRDF estimates over representative homogenous regions. Each entry in the LUT reflects a unique combination of land cover type, terrain information, time of year, disturbance age and type, and Landsat optical spectral bands. An initial conceptual LUT created for the Pacific Northwest (PNW) of the United States provides BRDF shapes estimated from MODIS observations for undisturbed and disturbed surface types, including recovery trajectories of burned areas and non-fire disturbances. With the assumption of invariant BRDF shape for consistent land covers, the spectral white-sky and black-sky albedos are derived through albedo-to-reflectance ratios as a bridge between Landsat and MODIS scale. A further narrow-to-broadband conversion based on radiative transfer simulation is adopted to produce the broadband albedos at visible, near infrared, and shortwave regimes. We evaluate the accuracy of produced Landsat albedo using available field measurements at forested AmeriFlux stations in the PNW region and examine the consistency of surface albedo generated by this approach respectively with that from “concurrent” approach and the coincident MODIS operational surface albedo products. Using the tower measurements as reference, the Landsat 30-m snow-free shortwave broadband albedo yields an absolute accuracy of 0.02 with a root mean square error less than 0.016 and a bias no more than 0.007. A further cross-comparison over individual maps shows that the retrieved white sky shortwave albedo from the “pre-MODIS era” LUT approach is highly consistent (R2=0.988, the scene-averaged low RMSE=0.009 and bias=-0.005) with that generated by the earlier “concurrent” approach, and exhibits more detailed landscape texture and wider dynamic range of albedo value than the coincident 500-m MODIS operational products (MCD43A3), especially in the heterogeneous regions. Collectively, the “pre-MODIS” LUT and “concurrent” approaches provide a practical way to retrieve long-term Landsat albedo since 1980s from historic Landsat archives as well as the on-going Landsat-8 mission, and support investigation on albedo evolution of terrestrial biomes at fine resolution.