A Digital Elevation Model of McDonald Island derived from GeoEye-1 stereo imagery captured 19 May 2012

Full description

This dataset consists of:
1 GeoEye-1 stereo imagery of an area of approximately 100 square kilometres including McDonald Island, captured 19 May 2012
2 A Digital Elevation Model (DEM) derived from the GeoEye-1 stereo imagery; and
3 Image products derived from the most vertical dataset of the stereo imagery and orthorectified using the DEM.
4 Contours generated from the DEM.

The DEM was produced at a 1 metre pixel size and is available in ESRI grid, ESRI ascii and BIL formats.
The DEM and image products are stored in a Universal Transverse Mercator zone 43 south projection, based on the WGS84 datum.

The image products are geotiffs as follows.

McDonald_Island_BGRN.tif: GeoEye-1 4-band multispectral (vis blue, green, red and Near Infrared), 2 metre resolution.

McDonald_Island_PAN.tif: GeoEye-1 panchromatic, 0.5 metre resolution.

McDonald_Island_PS_BGRN.tif: GeoEye-1 pansharpened, 4-band multispectral (vis blue, green, red and Near Infrared), 0.5 metre resolution.

McDonald_Island_RGB.tif: GeoEye-1 pansharpened, natural colour enhancement, 0.5 metre resolution.

Lineage

Progress Code: completed

Statement: The DEM and image products were created by Geoimage (www.geoimage.com.au) for the Australian Antarctic Division. The stereo imagery swathes were relatively cloud-free and their parameters are given in the Appendix of a report available for download from a Related URL. The stereo pairs were processed in SocetSet using tie points without any height information. The DEM was masked to only include well correlated points. The stereo imagery is poorly correlated in the south-western part of McDonald Island. This is due to the high amount of terrain shadow. GeoEye provided a previously captured stereo pair to assist in the process, however because this dataset also had high levels of shadow, results were only slightly improved. Because of the volcanic nature of the island, other datasets such as the Shuttle Radar Topographical Mission (SRTM) DEM were of no use with remedial processing on the datasets. Finally, it was decided that regions of poorly correlated points would be filled using a minimum curvature algorithm developed by Geoimage. Elevation data within this area will not be particularly accurate. No ground control was available to include in the creation of the DEM. Geoimage estimate the horizontal spatial accuracy of the datasets is within 4 metres of its true position and the relative vertical accuracy of the DEM should be less than a metre. When Geoimage have ground control points to check DEMs against, they typically see vertical shifts of 2 to 3 metres. The final DEM product was used to systematically orthorectify the most vertical dataset of the stereo imagery. The most vertical dataset used is a single image which GeoEye tiled into two images: po_872370_0010000 and po_872370_0010001. The Source ID code for this dataset is 2012051904383581603031601904, product ID 001. The dataset was captured at 04:38 GMT (UTC) on 19 May 2012. The orthorectification was performed within PCI OrthoEngine using cubic convolution methodology. XY control for the orthorectification was derived from satellite ephemeris data used within a satellite modelling module. The z control was derived from the DEM. The GeoEye-1 pansharpened 4-band multispectral image was derived by pansharpening the GeoEye-1 4-band multispectral image using the GeoEye-1 panchromatic image. The GeoEye-1 4-band multispectral image and the GeoEye-1 panchromatic image had been orthorectified separately before being used to produce the pansharpened product. The GeoEye-1 pansharpened, natural colour enhancement image is a colour enhanced 3-band (RGB) version of the GeoEye-1 pansharpened, 4-band multispectral image. The GeoEye-1 pansharpened, natural colour enhancement image is more noisy than the GeoEye-1 4-band multispectral image. This is a result of the pansharpening process, and is caused by the high levels of "glint" in the water. The colour enhancement "stretch" tends to highlight the noise. This noise is also evident in the GeoEye-1 pansharpened, 4-band multispectral dataset.