Earlier this summer a new global digital elevation model data set derived from the ASTER satellite system was jointly released by Japan's Ministry of Economy, Trade and industry (METI) and NASA The GDEM dataset was created from the 1.3 million scene ASTER VNIR archive which completely covers the Earth's land surface between 83N and 83S latitudes. The GDEM dataset is distributed in 30 meter per pixel resolution (see below), and is presented in 1 x 1 degree tiles. File format is GeoTIFF only and HDF versions are not available. 3DEM, Global Mapper, and GEOTIFF4 can read these files directly but not BLACKART which only has an HDF reader. (I may fix this if I get around to it.)

A Quality Assessment file accompanies each GDEM file. This file is a weird kind of thing that encodes pixel-by-pixel data in Geotiff format. The files can be read directly by 3DEM, MICRODEM and GEOTIFF4. (BLACKART can only read ASTER .HDF format but I may fix that soon.) Positive pixel values indicate the number of ASTER scenes from which data was extracted to create the GDEM pixel. Negative values indicate the alternative source dataset used to fill in voids or areas of corrupt data.

GDEM (apparent) resolution is 1 arc-second or approximately 30m at the equator. File size is 3601 by 3601 pixels. Coverage is from 83 degrees N latitude to 83 degrees S latitude, including over 22,000 tiles in total. As a result, each GDEM tile covers a considerably larger area than ASTER DEM tiles, which were about 60km square. The coordinate system has reverted to latitude and longitude from the UTM system used for ASTER data previously. Metadata is now included in an accompanying XML file rather than the former text file which makes it easier for a machine to read it but harder for a human. An example GDEM is shown to the right in the first figure.

The procedure for downloading the data is very similar to that for ASTER data and is described at the NASA EOS Website so I will not describe it here. In order to downloaded the data navigate to the NASA WIST website NASA WIST Website. It is still necessary to create an account, order the datasets through the user interface, and wait for the download link to be sent to you by email. It is a complete mystery why the NASA administrators make obtaining the data so cumbersome. If any commercial venture tried to conduct business using the WIST site they would be out of business in a week! You always get the impression that US government agencies have never been and never will be completely comfortable with allowing public access to their data. I am not complaining however as the USA is one of the very few companies that shares its data at all.

Although the release of global ASTER DEM data represents a step forward, it is not for the reasons that you might think. The most important benefit is that the area of coverage of global medium resolution DEM data has been significantly expanded to include the bands between 60N and 80N and 56S and 80S. In addition, GDEM offers an alternative data source for high relief areas where SRTM has large amounts of missing data. This is possible because the ASTER instrument senses the IR spectrum while SRTM is a radar instrument. (ASTER can also have missing data, but for different reasons. As a result, they may not occur in the same place.)

However, the most surprising thing about the data is that apparently does not represent a resolution improvement over SRTM, despite the vastly increased sampling rate (30m for GDEM versus 90m for SRTM). GDEM README file states: "Also, while the elevation postings in the ASTER GDEM are at 1 arc second, or approximately 30 m, the detail of topographic expression resolvable in the ASTER GDEM appears to be between 100 m and 120 m." (!) This confusing sentence does not make much sense, despite explanations in two GDEM references. Recall that the source of ASTER DEM data is the band 3N (nadir) and 3B (backward) bands of the L1A product. These images have a stated resolution of 15m. Conceivably there could be some loss of fidelity in the computation of the DEM image from the two stereo pairs. A degradation from 15m to 120m seems like a lot of loss from such a conversion. The other alternative is that the data have been purposely degraded. This is not as unlikely as it sounds. NASA could have released 30m SRTM data for its entire area of coverage, not just the USA. This was a political decision the stated purpose of which was to avoid offending US allies and other countries that objected to publication of terrain data of unacceptably high (to them) resolution.

In addition to the resolution issues, the ASTER folks have expended a lot of effort to determine the vertical accuracy of the dataset. The stated accuracy is 7-14m stated in terms of standard deviation. This should mean that 99% of GDEM data will fall within +/- 42m maximum of the actual elevation. Vertical accuracy was determined by subtracting GDEM elevation values from those of other datasets, including the NED and SRTM plus comparison with control points. A lot of attention has been focused on anomalies as well. These can result from sensing and processing errors and may take the form of pits, spikes and "mole runs". Further errors can occur around bodies of water and data may be completely missing around areas of persistent cloud cover. Further information regarding vertical and horizontal accuracy is available in the ASTER Global DEM Validation Summary Report

Although any addition to the archive of free GIS data is welcome, GDEM is a disappointment as a result of its restricted resolution. Unfortunately, the GIS user community must continue to wait for global DEM data that allows the creation of reasonably detailed terrain models. Perhaps this will be improved in future release versions of GDEM.

[Postscript: I accidentally learned that additional if not all products listed on the WIST website can also be downloaded for no charge. For example, I was able to search and download L1A images as well as the GDEMs. Unfortunately, L1A coverage is sparse and is not available for most of the earth's surface. However, this is a welcome return of this data to the global supply of free high resolution imagery.]

[Follow up: I encountered two problems when attempting to order ASTER L1A data from the WIST site. Firstly, the coverage was extremely sparse. I selected a large region of the African continent and only got two granules returned. Secondly, when my order notification arrived by email, it stated that it was cancelled because the images were no longer available! I am now working on another source of free ASTER L1A imagery. If successful it will be the subject of an article in next month's publication.]