First global Landsat surface reflectance products validated using near simultaneous MODIS observations
Abstract
Surface reflectance (SR) adjusted for atmospheric effects is a primary input for land cover change detection and for developing many higher-level surface geophysical parameters. With the launch of the first Landsat in 1972, a series of Landsat satellites have produced large quantities of images useful for land cover and change studies and other earth science applications. However, standard Landsat imagery products have been provided in DN or higher level products, not SR. This is due in part to many challenges to performing atmospheric correction operationally on Landsat images. With the mass processing capability provided by a 6S based atmospheric correction algorithm implemented in the Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS), we produced global Landsat surface reflectance products using the Global Land Survey (GLS) images acquired around 2000 and 2005. Near simultaneous MODIS observations were used to provide the most comprehensive evaluation of these products, that is, every Landsat surface reflectance image was evaluated as long as a near simultaneous MODIS acquisition was available. For Landsat 7 images, daily MODIS data acquired on the same day as the Landsat images were used in the evaluation. For Landsat 5 images, temporally closest MODIS NBAR data were used. The Landsat surface reflectance products were found highly consistent with the MODIS data, with root mean square difference (RMSD) values between them ranging between 1.2 and 2.3 percent absolute reflectance for Landsat-7 images and between 1.5 and 3.0 percent for Landsat-5 images. The observed differences between Landsat and MODIS reflectance values were within the uncertainty levels allowed by the instrument specifications of the two systems for 96% of the Landsat 7 images and 93% of the Landsat 5 images. Most of the RMSD values larger than the uncertainty levels allowed by the instrument specifications were attributed to 1) clouds that moved between Landsat and MODIS acquisitions, 2) saturation of the Landsat detectors over snow-covered areas, 3) observations that saturated the Landsat system but not the MODIS system, and 4) problematic MODIS data. These surface reflectance data sets were the first global Landsat surface reflectance products that have been validated comprehensively using MODIS data. They are available for free download at the Global Land Cover Facility (www.landcover.org).
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2011
- Bibcode:
- 2011AGUFMIN21A1411F
- Keywords:
-
- 0430 BIOGEOSCIENCES / Computational methods and data processing;
- 0480 BIOGEOSCIENCES / Remote sensing;
- 0520 COMPUTATIONAL GEOPHYSICS / Data analysis: algorithms and implementation;
- 1632 GLOBAL CHANGE / Land cover change