See note at base of this article regarding FREE ASTER L1A/L1B images available in .tif format at the the GLCF ASTER L1A L1B website.

In previous articles we outlined procedures and techniques for acquiring and using ASTER DEM data from the EOS Data Gateway. While the availability of free International DEM data is truly exciting to the terrain modeler, what we need to complement the DEMs and to make useful digital terrain models (DTMs) is equally high quality overlay imagery. Fortunately, such data is available from the same source. This article will look at a few possibilities using the ASTER L1A and L1B data sets.

ASTER L1A/L1B data was not designed to produce overlay material for DTMs at all. The sensors that acquire this type of data are primarily designed to collect reflected passive radiation outside the visible spectrum to assist scientists in studies of deforestation, crop analysis, land use analysis and global climate change. Only some of the ASTER sensors record in the visible spectrum, and the selection of wavelengths has nothing in particular to do with the production of natural appearing images. With a little manipulation this data can be immensely useful to the terrain modeler nevertheless, because it is possible to process the data to achieve something like natural terrain coloration.

ASTER L1A/L1B data, like its ASTER DEM counterpart is free. It's resolution actually exceeds that of the ASTER DEM product. (The resolution of the L1A/L1B VNIR bands are 15m, double that of the DEM product at 30m.) This is the ideal situation for the production of DEM overlays. Data is recorded in 15 spectral bands, providing a wealth of data for many purposes, including image production.

Recall that the ASTER DEMs are actually produced from the L1A product (bands 3N and 3B). As a result, L1A data can match the DEM data exactly, making registration of the DEM to the overlay easier. Also, if you already have the DEM product, you know that the L1A/L1B product exists (although the converse is not true). Although there are similarities between the L1A and L1B data sets, I will use the L1B data for this exercise mainly because the L1B image I have on hand is of better quality for our purpose than the my L1A image. However, the procedure is generally applicable. Let’s walk through the procedure for downloading the L1B data.

Log on to the NASA WIST exactly as if you were searching for ASTER DEMs. Select ‘EOS-EDC‘ for the data center and choose ‘ASTER Level-1B Data Set - Registered Radiance at Sensor‘ for the data set. Set your search parameters by latitude and longitude as explained in the previous section and press ‘Search’.

If data is available for your area of interest, you will see one or several data granules returned in anything from a few minutes to in extremely busy times under an hour. Examine the returned parameters including the latitude and longitude to see if what was returned will be useful to you. If so, order the data as you did for the DEMs. (L1A/L1B data is free just like the DEM data.) (If you thought the ASTER DEM data sets were large at 12Mbytes, brace yourself for a shock: the L1B data sets are approximately 120 megabytes in size! This makes downloading as well as the subsequent image processing steps a challenge. I recently purchased a brand-new eMachines 2.0GHz T4200 processor with a fast video card just to handle ASTER data as the processing burden proved too much for my relatively new HP6730 at 600 MHz.)

Once you have downloaded your data you are ready to produce your images. Besides the souped-up hardware requirements, you will need two essential software tools. The first is Multispec, a satellite data processing application available for free download from Purdue University. The second is Paintshop Pro, or another similar general-purpose image processing program.

First, rename your ASTER L1B file to something shorter than the name that EOS-EDC gives it. (Don’t forget to append .hdf. to the file name that you choose.) It is necessary to shorten the name because Multispec will not recognize the file by its given name because it is too long. For example,my test image of an area of northern Algeria was entitled 'pg-PR1B0000-2001050902_001_001.hdf' I renamed it to 'ASTERL1B_1.hdf. Now, open Multispec. Select ‘file’ ‘open image’ and select the file. Open it. Under ‘HDF Data Set’ choose ‘*ImageData1’. [Editor's note: this step is no longer necessary as the latest revison of Multispec v3.1 can handle long file names]. Leave everything else at the default settings and select ‘OK’. Set the Display Specifications for 1-Channel Color, 8 bits of color, and 256 levels per channel. Select ‘OK’ again. When you do this, a grayscale image of VNIR band 1 will be produced. Save this image as a .tif file by selecting ‘Save Image to Tif As...’.

Now repeat this procedure for ImageData2 and ImageData3N, each time saving the grayscale image as a .tif file. Why, you are asking, are we saving three grayscale images of the same thing? If you have not seen this trick before, prepare to be impressed. We are going to produce an RGB color composite image from the three grayscales. This is possible because even though the images look similar, they are not exactly alike. Each one has recorded reflected radiance in a different wavelength as a pixel value that we just happen to have displayed as a grayscale image. We can subsequently assign the pixel value of the ‘red’ grayscale to the red channel in a 24-bit color model, assign the ‘green’ grayscale pixel value to the green channel, and assign the ‘blue’ grayscale pixel value to the blue channel. This will produce a 24-bit (16 million color) image from the three grayscales. The three grayscales that I produced for my test image are shown at the upper right.

The technique is as follows. Open Paintshop Pro and open the three grayscales. Actually, they are not really grayscales at this point even though they are gray, so convert them to grayscale by selecting ‘Color’, ‘grayscale’ for each image. Now choose ‘Colors’, ‘combine channel’, ‘combine from rgb’. Put one of each of the images into each of the three RGB channels and hit ‘OK’. Your color composite image will be generated automatically. If you have decent data in your VNIR bands, you will see a high-quality false-color image, meaning that the colors that you see will not correspond to reality. (These false-color images are essential to image analysis because they can be adjusted to highlight certain features that scientists are looking for, like projected alfalfa harvest and the like.)

We, however are not interested in this, but would instead like to produce an image that is as close to real as possible. This is done by experimenting with the RGB channel assignment and then color balancing and correcting the result. A good place to start is to assign the shortest wavelength channel ASTER VNIR band to the blue channel and the longest to the red channel. Sometimes you will like the results better by not following this rule. The images presented on this page have the green and blue channels switched from this scheme.

The results of my effort is shown to the right. Beware! This large (450KB) composite image will take a while to view if you do not have a high speed data connection. The image is actually thinned down considerably from the 5MB original, so the excellent quality of these ASTER images can only be imagined until you try this yourself. I have provided a detail in the next image to the right to give an impression of the resolution provided by this data. Although coarse by modern standards (commercial IKONOS images are about 1m resolution) ASTER VNIR imagery at 15m is quite good and in fact would have been considered cutting-edge (and very expensive to buy) just a few years ago.)

This image is just chock-full of interesting features and would take another entire article to analyze fully. The impressive detail should be apparent to even a cursory inspection. It could be improved with redeye correction using BANDPASS but I kind of liked it as is, so I decided against further processing for now.

A second close-up of the water body and inlet river are shown in the image to the lower right. I believe this is the Chatt Melghir reservior south of Beskra. If you do not have a broadband connection, I recommend looking at this instead of the large one. The colors, although close to natural, are still false. This allows the sediment dispersion into the reservoir to be seen clearly. Note also the abundance of man-made features such as field boundaries for example.

This article has provided a brief introduction to the impressive world of ASTER L1A/L1B imagery. Although the images are close to natural, they are decidedly false-color. What is needed to produce accurate color images is sensor bands in the visible spectrum with channels in the red, green and blue wavelengths. Such data is available in the MISR product. We have also not solved the problem we set out to investigate, namely the creation of digital terrain models using L1A overlays and ASTER DEM data. These will be subjects for upcoming articles.

Sadly, effective August 12, 2002 LP-DAAC has started charging USD $60 for formerly free ASTER L1A/L1B images. United States citizens who feel as I do that after paying with our tax dollars for the satellite, launching the satellite, maintaining the groundstation, processing the data and administering the database where the data is archived that perhaps we already own the data may wish to contact LP-DAAC at edc@eos.nasa.gov to express this opinion.

Update, March 3, 2006 FREE ASTER L1A/L1B images are available in .tif format at the the GLCF ASTER L1A L1B website. Since the data is archived in .tif format, it is not necessary to convert from .hdf using Multispec as described in this article. Just pick up the technique after the conversion step to use these files as described in the article.