NASA ’ s Global Orthorectified Landsat Data Set

NASA has sponsored the creation of an orthorectified and geodetically accurate global land data set of Landsat Multispectral Scanner, Thematic Mapper, and Enhanced Thematic Mapper data, from the 1970s, circa 1990, and circa 2000, respectively, to support a variety of scientific studies and educational purposes. This is the first time a geodetically accurate global compendium of orthorectified multi-epoch digital satellite data at the 30to 80-m spatial scale spanning 30 years has been produced for use by the international scientific and educational communities. We describe data selection, orthorectification, accuracy, access, and other aspects of these data. Background Spatial variability exists for our planet’s land surface at dimensions of meters to tens of meters, due to local terrain variability and associated microclimatic influences upon vegetation types and associations. Accordingly, spatial data at tens of meters are required to accurately map many areas, because of the low spatial autocorrelation of land-surface features (Townshend and Justice, 1988; Townshend and Justice, 1990). In addition, a variety of natural and human land-use changes, such as wild fires, deforestation, wetland conversion, and urbanization, represent alterations of landscapes, which also occur at spatial scales of tens of meters. These are important perturbations of the global environment and require similar spatial scale data for quantification. Currently, we lack information regarding where environmental change is occurring, what the changes are, and what the post-change properties of the altered areas are (Townshend et al., 1991). Understanding environmental or land-cover dynamics represents an important challenge in the study of the global environment, because many land-cover changes take place at fine scales of resolution, requiring Landsat-type imagery for accurate measurement. Uses for such data range from biodiversity and habitat mapping for localized areas, to specifying parameters for large-scale numerical models simulating biogeochemical cycling, hydrological processes, and ecosystem functioning. These needs have been recognized in the International Geosphere Biosphere Programme, the World Climate Research Programme, and the International Satellite Land Surface Climatologic Project, among others (Becker et al., 1988; IGBP, 1990; WMO, 1992). Responding to these needs, the Scientific Data Purchase Program of NASA’s Stennis Space Center (http://www.esa.ssc.nasa.gov/datapurchase) has directed the production of global orthorectified and co-registered Landsat Multispectral Scanner (MS), Thematic Mapper (TM), and Enhanced Thematic Mapper (ETM+) data for three periods: the late 1970s, circa 1990, and circa 2000, respectively. Landsat 80-, 30-, and 15-m satellite data are the only record of global land-surface conditions at a spatial scale of NASA’s Global Orthorectified Landsat Data Set Compton J. Tucker, Denelle M. Grant, and Jon D. Dykstra tens of meters spanning the last 30 years. They constitute an indispensable history of land-surface state. Data at these spatial resolutions can provide a high potential mapping accuracy of natural vegetation and alterations to it, if and only if highly accurate scene-to-scene withinand among-date registration is achieved. Otherwise, misregistration errors between or among dates are confused with land-cover change and resulting interpretations are meaningless (Townshend et al., 1992). The best solution to eliminate or minimize misregistration errors is to precision orthorectify each scene within each of the three epochs. The process of orthorectification removes erroneous image displacements caused by the interaction between terrain relief or local elevation changes and sensor orientation variations. The orthorectification process results in remotely sensed image products that possess both the imagebased information of the original satellite data and the geometric information of a geodetically accurate map. When Landsat data are assembled into a mosaic to cover an area of interest, an underlying assumption is that the data have a consistent geometry throughout the image. This has been demonstrated for the MS, TM, and ETM+ instruments with increasing accuracy, respectively (Desachy et al., 1985; Welch et al., 1985; Malaret et al., 1985; Bryant et al., 1985; Storey and Choate, 2000). If the Earth’s surface were the same elevation over the geographical area of interest and the satellites in question were in identical orbits for the period of interest, a Landsat data mosaic could be “tied” together using a linear or affine mapping, and the resulting mosaic would be an accurate representation of the surface with accurate distances among all surface features. When the topography is irregular, as is normally the case, it is necessary to correct localized horizontal displacements created by perspective view distortions around areas of local relief. This process, called orthorectification, combines knowledge of the elevation of each image point with the precise viewing geometry at that point to calculate a horizontal correction to the satellite data. The result is an image product that appears as if the satellite or the viewer is looking normal to the Earth at every location. In such orthogonal views, the horizontal position of any feature directly beneath the viewer would not be effected by local terrain variations. Correcting satellite imagery for variations in topography and/or satellite viewing perspectives is best accomplished by using ground coordinates, also known as ground control points. This enables digital elevation data to be associated with the respective satellite data by matching coordinates. Accurate association of the digital elevation data with the satellite image data is necessary to compensate for topography and/or viewing perspective variations. When the satellite data have been corrected for terrain and/or satellite viewing perspectives, these data are referred to as having been “orthorectified.” In this paper we use orthorectified to mean the satellite data have been corrected for terrain displacements, corrected for any P H OTO G R A M M E T R I C E N G I N E E R I N G & R E M OT E S E N S I N G March 2004 3 1 3 C.J. Tucker and D.M. Grant are with the Laboratory for Terrestrial Physics, NASA/Goddard Space Flight Center, Greenbelt, MD 20771 ([email protected]). J.D. Dykstra is with Earth Satellite Corporation, 6011 Executive Blvd., Suite 40, Rockville, MD 20852. Photogrammetric Engineering & Remote Sensing Vol. 70, No. 3, March 2004, pp. 313–322. 0099-1112/04/7003–0313/$3.00/0 © 2004 American Society for Photogrammetry and Remote Sensing 02-112.qxd 2/4/04 11:36 AM Page 313