In a previous article I outlined techniques for producing false color composites from the excellent 15 meter resolution ASTER L1A/L1B satellite images available from LP-DAAC. Unfortunately, these images, while not as expensive as commercial imagery, are no longer free. Amateur terrain modelers looking for no-cost overlay material must look elsewhere. Of course with sufficient funds one can purchase excellent commercial satellite data. However, it is more challenging to find data of adequate quality for free, which is what this website is dedicated to, of course.

What the world needs is something exactly like what ASTER almost was: free, high quality, multi spectral, overlapping band satellite imagery covering the entire surface of the earth. First: the bad news: as far as I can tell, this does not exist (anymore). But the picture is not entirely bleak. There is a lot of interesting data out there. Unfortunately, it is not all available in one place, the quality is not consistent, it is often not multi spectral and total earth coverage is woefully lacking. In subsequent articles, I will review each of these alternative sources individually. The first, best , and probably the most exciting is the free Landsat imagery available from GeoGratis. In addition to introducing GeoGratis (what a great name) I will describe the Landsat 7 file format and discuss techniques for processing the data.

GeoGratis is a Canadian website that offers Landsat coverage for all of Canada and for part of the northern United States. (Why US citizens have to go to a Canadian website for free data that our own government charges us up to $600 per image for is probably a good subject for another article.)

GeoGratis is apparently an initiative of the Canada Centre for Remote Sensing, which in turn is apparently an agency of Natural Resources Canada. Whatever it is, I think every country should have one. It offers a wealth of GIS data including RADARSAT, AVHRR, Aerial Photo Images, Census Data, Watershed Maps and many other data sets besides Landsat. I have not fully explored all the offerings, but the Landsat imagery is the most exciting that I have found thus far. Landsat’s resolution is similar to that of ASTER. Bands 1-5 and band 7 are 30m pixel size, Band 6 is 60m and band 8 (the panchromatic band) is 15m resolution, the same as the best ASTER bands. (However, I find the 15m ASTER bands superior for a few reasons.) Downloading the data is easy. Just navigate the GeoGratis site using your browser (you can use FTP if you like using ‘ftp-geogratis.ccrs.nrcan.gc.ca’ with user name ‘anonymous’ and your email address as password.) GeoGratis supports command-line ftp but does not support passive mode. This is the opposite of the EOS-EDC ftp protocol. There is actually some merit in this mode of access due to the Landsat file structure as will be seen momentarily.

I have not found a catalogue or index of all the images (and there are a lot of them) so it may be necessary to use the EOS-EDC search utility to locate your target image and then navigate to it using the file name.

For my example, I chose file L71003026_02620000827 as my target, which covers an area: UL_CORNER_LAT = 49.8456726; UL_CORNER_LON = -55.9499168 UR_CORNER_LAT = 49.4634933 UR_CORNER_LON = -53.3299141 which is an area of Newfoundland. Important note: GeoGratis apparently offers the Landsat 7 imagery in more than one format. I will describe the procedure for processing the HDF format files, which is the native and more compressed format. The imagery is apparently also available in geotiff format. The geotiff versions are in the directory called 'ortho/' so they may be orthorectified. I am in the process of checking this out. For the time being I will stick with the procedure for HDF format.

Upon entering the 'hdf/' directory and opening the folder for our target file we are presented with eighteen individual data and metadata files. The first files that you will probably want to look at are the browseimage.gif, browseimage.tif, and/or thumbnail.gif files. These are self-explanatory: they give you a quick look at the panchromatic band image.

Once you have that out of the way, you need to find the metadata file. In this case the file is called L71003026_02620000827_MTL.L1G. The file names are all annoyingly truncated but it will be the one that is 64K in size. This file contains important information necessary for subsequent file processing.

In order to process Landsat 7 HDF images, you need to understand the file structure. While technically conforming to the HDF standard, the developers cleverly un-encapsulated the bands. This makes machine-processing the files arguably a little more difficult, but offers several key advantages. First, it is not necessary to download the entire file. You only have to download the bands you are interested in. Those of us using dialup connections will appreciate this feature. The second advantage is even more interesting. Since the bands are written as a sequential series of eight-bit bytes, each file is essentially a flat binary file that is extremely easy to process “manually”. All that is required is the row count and column count and the band files can be read by any variety of applications, including MultiSpec or even my favorite GIS application: Paintshop Pro. The procedures for these two applications are almost identical, but the Paintshop procedure is painfully simple so I will demonstrate using that application.

If you inspect the metadata file somewhere you will see the following lines:

            BAND1_FILE_NAME = "L71003026_02620000827_B10.L1G"
            BAND2_FILE_NAME = "L71003026_02620000827_B20.L1G"
            BAND3_FILE_NAME = "L71003026_02620000827_B30.L1G"
            BAND4_FILE_NAME = "L71003026_02620000827_B40.L1G"
            BAND5_FILE_NAME = "L71003026_02620000827_B50.L1G"
            BAND61_FILE_NAME = "L71003026_02620000827_B61.L1G"
            BAND62_FILE_NAME = "L72003026_02620000827_B62.L1G"
            BAND7_FILE_NAME = "L72003026_02620000827_B70.L1G"
            BAND8_FILE_NAME = "L72003026_02620000827_B80.L1G"
            METADATA_L1_FILE_NAME = "L71003026_02620000827_MTL.L1G"
            CPF_FILE_NAME = "L7CPF20000719_20000930.09"
            HDF_DIR_FILE_NAME = "L71003026_02620000827_HDF.L1G"

This is the file directory relating the file names to the bands. Information relating the bands to the sensor characteristics can be found at NASA’s Landsat Facts page. Immediately above these lines, the following appear:

            PRODUCT_SAMPLES_PAN = 12928
            PRODUCT_LINES_PAN = 12000
            PRODUCT_SAMPLES_REF = 6464
            PRODUCT_LINES_REF = 6000
            PRODUCT_SAMPLES_THM = 3232
            PRODUCT_LINES_THM = 3000

This important information is the row and column count for the various bands. The first set is for band 8, the second set is for bands 1-5 and 7 and the last set is for band 6. For this example, I downloaded band 5, L71003026_02620000827_B50.L1G. (If you plan to download multiple bands using a dialup connection, ftp may be attractive because you can select several files and then let them download overnight.)

After a successful download and unzipping the file I merely opened Paintshop Pro and selected Files of type ‘Raw’. After selecting ‘Open’ I was presented with a screen prompting me for the image width and height. Looking at the information listed above, I found that I needed to enter 6464 in the width edit box and 6000 in the height edit box. I clicked on ‘OK’ and my grayscale band 1 image was generated within seconds. A thinned version of this image can be seen in the first picture shown to the right. Beneath this image are Band 2 and Band 5 images. The spectral range of band 5 at 1.55-1.75 micro meters is the longest of the five 30m bands, so I chose this for the red channel. (I discovered after I created the composite that the visible RGB bands are 4, 3, and 2, respectively. However, the bands I chose seemed to work fine.) The fourth image is a color composite created from the three images. The technique used is exactly as described in the previous ASTER L1A/L1B article. The true-color composite image is shown as the fourth image. The final image shows a detail corresponding to the inset on image four. The impressive detail of Landsat 30m resolution images should be apparent. The band specifications for Landsat are shown below:

              Band 1      0.45-0.52      Visible Blue
              Band 2      0.53-0.61      Visible Green
              Band 3      0.63-0.69      Visible Red
              Band 4      0.75-0.90      Near Infrared
              Band 5      1.55-1.75      Mid Infrared
              Band 6      10.4-12.5      Thermal
              Band 7      2.09-2.35      Short Wave Infrared
              Band 8      0.52-0.90      Visible Green to Near Infrared

An interesting sidebar to the images shown: Bands 1 and 2 appear a little "muddier' than Band 5. This is in large part due to the greater absorption of atmospheric aerosols at the corresponding wavelengths. A clearer false color composite may have resulted from using the visible bands 2, 3 and 4 or even false color bands 3, 4 and 5. I did not experiment with this but I liked the composite image "texture" resulting from the atmospheric effects. The importance of the additional information in multi-spectral imagery cannot be over estimated. The disadvantages of "flat" imagery as compared to the wealth of information in multi spectral imagery will be apparent in subsequent articles that examine other sources of remotely sensed data.

Landsat also presents a high resolution image in band 8. This appears to be a 15m gray scale image. At least that is what I got when I processed it. This can be used stand-alone or to pan sharpen the multispectral band images. See information at www.PANCROMA.com on how to accomplish that.

The detail available is similar to that in the ASTER L1B images highlighted in the previous article, with the qualifications noted above.

I hope that this brief introduction to Landsat 7 processing has shown that it is not necessary to have an expensive GIS application to manipulate this interesting dataset. As usual, the editorial staff at terrainmap (me) has shown how this impressive data set can be accessed with some humble tools and a little know-how.