OSFAC:: LANDSAT MOSAIC

Image mosaic information^*

Band combination: All images use Landsat bands 547 in RGB.
All tiffs are 24-bit.

The full-resolution mosaic was split into 4 tiles (NW, NE, SE, SW), each less than 500 MB in size, in order to aid usability and portability. Each tile contains embedded georeferencing information and an associated tfw world file, and so can be easily pieced together in GIS or image processing packages. The included JPG format file indicates the tile borders with cyan lines (it also shows the cutlines between different scenes, and the CBFP landscape borders for reference).

The 57 m and 100 m reduced resolution mosaics were created by averaging all pixel values at the 28.5 m resolution that fall into the extent of the output pixel; they are NOT simply nearest-neighbor resampled versions of the 28.5 m mosaic. ENVI's "pixel aggregation" resizing function was used to do this.

Georeferencing information

UTM Zone nnS projection, WGS-84 ellipsoid.
All tiff files are geotiffs, with embedded georeferencing.
tfw world files are also included.

Mosaicking methodology

Orthorectified images were acquired from the University of Maryland's Global Landcover Facility (glcf.umiacs.umd.edu). The orthorectification process is described in "NASA's Global Orthorectified Landsat Data Set", CJ Tucker, DM Grant, JD Dykstra, Photogrammetric Engineering & Remote Sensing 70(3):313-322 (March 2004).

PCI Geomatica (v 9.1) Orthoengine was used to mosaic the images. There are two manual steps in this process:

defining cutlines, and
matching color.

Cutlines were drawn to select the best non-cloudy scene portion for the overlap region between adjacent scenes. Also, if more than one scene was available for given path/row that had significant cloud cover, cutlines were used to replace the cloudy portions on one date with clear imagery from another date.

After cutlines were selected, polygons were drawn on overlapping scene areas to compute radiometric matching. For these polygons, both scenes should be cloud-free and not include significant change. Using these radiometric tie-points, Orthoengine automatically calculates a linear stretch to match one scene to the other. Note that finding good radiometric tie points can be difficult or impossible if either scene contain clouds or haze for most of the overlapping area. Also, the different sun-angles on different dates leads to further radiometric differences than can be difficult to completely correct for.

After Orthoengine creates the mosaic, two additional steps were taken:

cross-track radiometric balancing, and,
final image stretching. When mosaics were more than 2 scenes across (east-west), there was often a noticeable darkening from west to east. When noticeable, this was corrected using the Cross Track Illumination Correction function in the ENVI image processing software package (Research Systems Inc), with a 1-degree (eg linear) stretch across the mosaic. Lastly, a final image stretch was applied to optimize contrast in all areas of the mosaic. Note that more informative stretches are possible for any local region of the mosaic, but these do not produce a useful stretch for the entire mosaic.

GIS Layers

Two sets of GIS layers are displayed in the posters:

CBFP Conservation Landscapes: the displayed boundaries are the more narrowly defined set of boundaries; a more loosely defined set is also available, and printed on some older posters in this series (but not for Sangha-Tri National Landscape).
IUCN's World Database of Protected Areas (WDPA), 2003 version. More information available at: http://sea.unep-wcmc.org/wdbpa/download/wdpa2004/WDPA_Info/WDPA%20Consortium.html

Project summary

These Landsat mosaics and mosaic posters were produced by Dan Slayback (SSAI/Biospheric Sciences Branch, NASA Goddard Space Flight Center, Greenbelt MD) in collaboration with the Geography Department at the University of Maryland, College Park, in the framework of the Central African Program for the Environment(CARPE)/Congo Basin Forest Partnership(CBFP). Please visit carpe.umd.edu and www.cbfp.org for more information.

Contact info for the CARPE-NASA/UMD project

Mosaic production & reproduction:

Dan Slayback, dan.slayback@gsfc.nasa.gov

CARPE/UMD project coordinator, in the US:

Alice Altstatt, aaltstat@umiacs.umd.edu

CARPE/UMD project, in-country (DRC) contact:

Didier Devers, ddevers@glue.umd.edu

CARPE information: carpe.umd.edu

CBFP information: www.cbfp.org

^* Toutes les informations sur les mosaïques landsat sont collectées sur le CD de distribution.