By John Childs email@example.com
technical editor Paul Burkhardt
Printer Friendly Version
I have discussed the distressing lack of international DEM data extensively in articles and on my website. Until just recently, I knew of only two sources of free international DEM data. The first is the 1 km DTED0 data available from NIMA (and the equivalent GTOPO30 data set available from USGS/EOS). The second was the reverse engineering approach of extracting DEMs from topo maps as described in my earlier article. Of course IKONOS and QuickBird commercial satellite imagery are (or will soon be, in the case of QuickBird) available, but although of excellent quality, this data is quite expensive. (The Digital Globe website quotes prices of $25/square kilometer. This equates to about $3,500 for a 1:24000 USGS quad.)
However, there is another source of international DEM data that has been known to the scientific community and few others until recently: ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer). The quality of this data is excellent, equivalent in resolution to USGS 30m SDTS DEM data.
ASTER is an imaging instrument that is flying on the TERRA satellite launched in December 1999 as part of NASA's Earth Observing System (EOS). ASTER represents a revolution in the remote sensing community because of the availability of its imagery and its superior resolution. ASTER resolution ranges from 15m to 90m, depending on the wavelength. The instrument records in three bands: the Visible and Near Infrared (VNIR ), the Shortwave Infrared (SWIR ), and the Thermal Infrared (TIR ), oriented on the nadir and backward looking. There are 14 spectral bands all together spanning the visible and infrared spectra, so the sensor is susceptible to cloud cover and cannot record images at night.
Because of its off-nadir sensor pointing capability, ASTER can collect the stereo pairs necessary to generate high resolution DEMS (using bands 3N and 3B). Instead of the woeful 1km resolution DTED0 data, ASTER DEMs offer a very respectable 30m resolution. The EOS ground stations archive imagery corresponding to the spectral bands of the ASTER sensors, including the Level 1A data sets. The DEMs are assembled from this data. Once produced, EOS archives the DEMs in its database so that they will be available essentially immediately upon subsequent request. If the DEM you are interested in is not in the archive, all is not lost. You can enter your request for DEM production into a queue. Your DEM will be constructed by EOS from the L1A data. This can take anywhere from two to ten weeks. Another option is to construct the DEMs yourself using the appropriate software. (It is also reportedly possible to order a fly-by acquisition, but this method is not publicized by NASA and is probably by special arrangement for government and academic researchers.)
Why does ASTER represent a revolution in the world of terrain modeling? Because its 30m resolution is the same as the classified NIMA DTED1 data set. NIMA has never released its DTED1 series of DEM data and in fact is doing its best to block NASA from releasing 30m SRTM data from the Space Shuttle. (Quote from the NASA SRTM website:"Following the events of Sept. 11, 2001 our project partners in the National Imagery and Mapping Agency have requested that NASA and JPL not distribute any SRTM data to the scientific community or the public, including shaded relief maps or other visualizations. We look forward to this restriction being lifted soon so data may be made publicly available." NIMA does not feel obliged to explain the connection between the events of September 11 and DEMs of someplace like Mt. Aconcagua in Argentina, for example, or exactly when "soon" might be.)
However, NASA somehow overcame whatever reservations NIMA might have had concerning ASTER. This is (pleasantly) surprising considering NIMA's position on SRTM. It is hard to believe that NIMA does not know about the EOS Data Gateway, despite its obscurity. On the other hand, how can NIMA rationalize its position on the SRTM data in light of the unrestricted availability of ASTER data of similar quality? It is perhaps time for the ASTER client base to expand from its current concentration in the academic and governmental community to the broader public GIS sector. In this way the continued availability of this valuable data source might be assured by the strength of larger numbers.
What does ASTER resolution mean to the terrain modeler? Consider the first image below. It is an ASTER DEM of the Jordan river valley in the occupied West Bank of Israel. The image is approximately 2500 by 2500 postings large, covering an area of about 75 square kilometers. The image to the right is a section of NIMA DTED0 quad N32E035, covering approximately the same area. Because of the inferior resolution of the DTED0 image, it is hard to believe that the coverages coincide very closely. Before ASTER, this was effectively the highest resolution DEM available to the civilian public. This DEM contains 232 by 246 individual elevation postings. The point should be obvious. The ASTER image contains more than 100 times as much data as the DTED0 image.
Despite only being in orbit since February, 2000, ASTER has already accumulated a relatively large archive of imagery. This data is offered for essentially free download from the EOS Data Gateway. However, the procedure for downloading the data is not particularly easy. You have to navigate a somewhat awkward and slow user interface. After placing your order using a four-step process, you must wait anywhere from one hour to two days for your image to become available for archived images, and as long as ten weeks if you request production of a DEM not in the archive. You download the data from a designated anonymous ftp site using an ftp utility like WS_FTP or the DOS command line.
Another obstacle is the file format. ASTER imagery is offered in something called HDF-EOS. HDF is yet another cumbersome, complicated self-describing scientific file format that few applications recognize. Those that do are all generalized HDF readers that are necessarily complex because they must accommodate all types of scientific data sets that may be packaged in HDF format. (In general, HDF data may not even be image data.) The DEM files are also offered in a 16-bit GeoTiff format. This does not represent much of an advantage, because most graphics applications only read the more common 24-bit tiff format.
Even if they could read it, very few applications know what to do with a GeoTiff DEM once they get it, since the format was really designed to accommodate 2D spatial data. (One that does due to recent upgrades is GlobalMapper). ASTER DEMs are also quite large. Each file is about 12 MB in either of the two formats. This makes converting and transporting the files from one application to another a challenge.
Of course I wrote yet another converter, this one called GEOTIFF to help with this problem. This program takes ASTER Geotiff DEMs and converts them to USGS native DEM format. (SDTS would have been a better choice for output because of its binary format, but the SDTS format is very complicated to write, and I already had a USGS DEM writer. So this was the expedient choice.)
Because USGS native DEM format is ASCII, the file size will increase by a factor of three upon conversion. The program must make an O(N^2)) file write operation, which means it will take some time to execute. This makes subsetting of the ASTER data into blocks of something like 500 by 500 elevation values very advisable. GEOTIFF will ask you if you want to subset the data. You would be wise to accept its offer.
Although the metadata is contained within the file, GEOTIFF requires manual georeferencing. I actually deciphered the GeoTiff system of proprietary tags and GeoKeys so parsing the metadata is not a problem, theoretically. There are numerous technical issues, however. USGS ASCII DEM format requires a coordinate transformation from the latitude and longitude contained in the ASTER file to UTM meters in WGS 84 projection. The UTM zone needs to be obtained from a lookup table. The coordinates are in 64-bit IEEE 754 floating point format, which means more programming to decode. So I took the easy way out and require the user to read the metadata from the .hdf.met file and input the appropriate information when prompted.
The program will prompt you for the grid spacing, a corner tie point, and the UTM zone. (Since ASTER images all seem to be oriented north and south, GEOTIFF only requires one tie point.) You will need to convert latitude and longitude projected to the WGS 84 ellipsoid in UTM prior to running GEOTIFF. Don't forget to record the UTM zone because you will need that too.
After you have done the conversion, you can import the ASCII DEM file into virtually any GIS application. I used 3DEM to produce the images below. The first image is a subset of the Jordan river valley DEM. The subset area is marked with a white rectangle on the DEM shown above. I extracted this DEM using the subsetting capabilities of GEOTIFF. The corresponding subset DEM and DTM are shown in the two images below. Note the extremely flat terrain of the Jordan river drainage. It looks like the valley lies on an ancient lake bed (like the Salt Lake valley in the United States) and that at one time the water level was much higher in this region.
The two images below show a view of a mountainous region in the vicinity of the Khyber Pass on the Pakistan/Afghanistan border, extracted from the third DEM above. The view covers an area of approximately 75 square kilometers. (Note the moth-eaten appearance of this particular DEM. The holes represent areas of missing data, often as a result of cloud cover. ASTER imagery, unlike the carefully processed USGS data is very much raw and unprocessed. This characteristic adds to the appeal of ASTER, as you know you are exploring the frontier of this exciting technology.)
This DEM is subsetted from the area corresponding to the white rectangle on the image labeled 'ASTER_DEM20020104110754.TIF' above. The resulting subset DEM and terrain model are shown in the final two images. The impressive resolution of the ASTER DEMs should be apparent from the high quality of the DTM extracted from this small region of the overall DEM. Compare the terrain map of Afghanistan shown in the section on my website entitled ' Afghanistan Maps' (prepared from DTED0 data and covering thousands of square kilometers) to the ASTER subset data which covers a mere 15 by 15 kilometer square.
This article has given just a brief introduction to the exciting world of ASTER imagery. I have left out many details of how to create a DTM from ASTER DEM data. ASTER data sets represent a rich but decidedly complex source of DEM and overlay imagery for the digital cartographer. I will explore the technical challenges of using this complex data set more fully further in subsequent articles. I am not sure how NASA will respond to the increasing interest in this product. Hopefully, its response will be to recognize the considerable demand and accelerate the introduction of more such data to the mapping community.
You can download GEOTIFF and obtain additional information related to digital terrain models from my website at www.terrainmap.com.
About The Author
John Childs is an amateur programmer and cartographer. He maintains a not-for-profit website at www.terrainmap.com where he offers a series of DEM
file translation utilities and tutorials available for free download. Paul Burkhardt is a spatial analysis expert with hundreds of hours of ASTER experience.
More Articles by John:
Entire Article Copyright 2002 John Childs, all rights reserved. Reproduction or redistribution in whole or in part without contacting the author and The Geocommunity is strictly prohibited.