The PANCROMA Landsat Point Spectrum Generator (PSG) is a Landsat spectral analysis utility that allows you to interrogate a Landsat band data set and to generate a spectrum that corresponds to points that you select from the set. This can be a powerful tool for target discrimination, mineral exploration or land use analysis when combined with the Landsat Spectral Analyzer described in a previous White Paper. The PSG can input both Landsat 8-bit Digital Number (DN) multispectral band files or 16-bit Landsat Reflectance band files. This article will describe the DN version, although the use of the Reflectance version is similar. See the PANCROMA Instruction Manual for further information on the Reflectance version.
The use of the PSG is best illustrated by an example. Consider the Landsat data set represented by the band1 image below.
Let's say that we are interested in measuring the snow pack coverage for the subsetted region in the image. Snow pack measurements are often needed in order to estimate the amount of runoff in the spring and to determine the water supply potential for down gradient areas. We can estimate the coverage using the PSG, followed by using the Landsat Spectral Analyzer.
We first start by opening the six Landsat visible and infrared band files, that is bands 1,2,3,4,5 and 7 by selecting 'File' | 'Open' as usual. Next select 'Spectral Analysis' | 'Landsat Point Spectrum Generator' | 'Six DN Bands'. When you do so the band1 image will be generated, along with the Point Spectrum Data Entry Form, as shown below.
Since we want to measure the snow pack, we will start by generating its spectrum. We can do this by successively clicking the cursor on the snowy areas in the band file. Some of the locations that I selected are shown as red dots in the first image above. Notice that there are areas of highly reflective snow and areas in shadow, as well as areas of thinner coverage. Click on all types that you are interested in measuring. You can sample as many as twenty points. Note that the coordinates of your selection appear in the text boxes and a running count is kept of your selections.
When you have collected your sample, click 'OK'. The TOA reflectance form will appear. This is the same form that is described in the previous Section. As before, you must enter the Acquisition Date and Solar Elevation Angle from the Landsat metadata so that TOA reflectances can be computed. When you click 'OK' on this form, the spectrum will be generated.
The plot shows all of the computed TOA reflectances versus band number. Now you must analyze the plot. Notice that the snow pack reflectances are uniformly low for bands 5 and 7 and higher for the other bands. Note also that the other bands do not show such uniformity as in bands 5 and 7. Our objective is to determine the appropriate settings for the Spectral Analyzer by inspecting this plot. I decided that I would try specifying that bands 5 and 7 not exceed 0.10, while requiring that band1 be at least 0.20 and the rest of the bands at least 0.10.
So the next step is to select 'Close Graphics Windows and Reset'. Select 'File' | 'Open' again and then re-open the six band files as before, i.e. bands 22.214.171.124.5 and 7. . Next select 'Spectral Analysis' | 'Landsat Spectral Analyzer' | 'Six DN Bands'. When the Spectral Criteria Form appears, enter the criteria from your analysis of the Point Spectrum. In this example the settings are as shown below.
(Note the radio button selections on the six enabled bands.) Now select 'OK'. The TOA Reflectance Data Form will again appear. Re-enter the Solar Elevation Angle and Acquisition Date and click 'OK. After a bit of computation a grayscale image will appear with the areas that match your specified criteria highlighted in red. It may take a bit of trial-and-error in order to dial in the settings that you feel best outline your areas of interest. The results for this example are shown in the image below.
PANCROMA computed that the coverage was 11.355%. Comparison of the highlighted image with the band1 image indicates that the coverage is about right.
Our second example looks at an urban scene on the west side of Chicago, USA. The RGB Landsat image for the area of interest is shown below.
Lets say that we are conducting a study of open space growth/degradation over time and we want to measure the percent open space coverage in this image. It is hard to guess how much that might be just in looking at the image. We will use the Landsat PSG and LSA to do this.
As before, we open the six band files, bands 1,2,3,4,5 and 7. The grayscale band 1 image is displayed as shown below. As before, we will click on the obvious open space areas of interest. Some of these are indicated on the image
I selected twenty such points in total. Now select 'Spectral Analysis' | 'Landsat Point Spectrum Generator' | 'Six DN Bands' as before. The spectrum that resulted is shown in the image below. Note that this spectrum is much more precisely defined than the one for the snow reflectances. Snow is somewhat problematic because it is so highly reflective and subject to slope angle and shadow in mountainous areas. The point scatter for each band is relatively low for this spectrum.
I analyzed this spectrum and decided on my settings for the LSA. So the next step is to reset by selecting 'Close Graphics Window and Reset'. Select 'File' | 'Open' again and then re-open the six band files as before, i.e. bands 1,2,3,4,5 and 7. Next select 'Spectral Analysis' | 'Landsat Spectral Analyzer' | 'Six DN Bands'. When the Spectral Criteria Form appears, enter the criteria from your analysis of the Point Spectrum. Mine looked like those shown in the data entry form below.
I clicked "OK', prompting another form requesting the Solar Elevation Angle and Acquisition Date as before. After entering this data, the LSA computed the open space coverage as in the image below.
The LSA computed that the open area coverage was 35.336%. Most of the coverage is in several large areas that appear to be parks. However a significant amount of the open space consists of smaller plots, areas alongside highways, etc.
For comparison, I generated a BGR=234 false color composite image, i.e. using band2, band3 and band4 in the blue, green and red channels to make an RGB false color composite image. I then used the PANCROMA K-Cluster algorithm to classify the colors. This method also highlighted the target areas, as shown in the image below:
This image showed a similar result to the PSG image. Its coverage was 27.29%, slightly less than the PSG image predicted but reasonably close. Of course my selected target was fairly easy as these areas are highly reflective in the near infrared (NIR) band, as can be seen from the PSG spectrum above. The PSG method has the advantage of using all of the spectral bands and as a result should have greater discriminating power.
Another useful tool is the Spectral Analyzer™ using the Euclidean Distance plot utility. This can be used for vegetation species identification. The following image shows a visible spectrum image of cultivated fields. Let's say that we are interested in identifying all of the fields that are growing the type of crop that is in the irrigation circle identified in the image. It is not very easy to do so by visual inspection of the image.
The first step in the identification procedure is to generate the spectrum of the target using the Point Spectrum Generator™ as described above. This is done by using the utility and then clicking on several points within the target irrigation circle. When I did so I generated the spectrum shown below.
The final step in the procedure is to use this spectrum and the Spectral Analyzer™ to identify all vegetation with similar spectra in the image. PANCROMA™ will automatically load the spectrum into the Spectral Analyzer form. Select the 'Euclidean Distance' option from the Spectral Analysis menu selection. The form with my spectrum loaded in looked as follows.
The Euclidean Distance plot is shown in the image below. The vegetation with the closest spectral match are shown in red.
There are many other uses for the PSG and the Spectral Analyzer. These include crop coverage analyses and mineral exploration. The PANCROMA Landsat Point Spectrum Generator provides a more discriminating method for performing such measurements than band combinations or even band ratios and provides a convenient method for quantifying the coverage.
PANCROMA also offers Point Spectrum Generators for SPOT, ASTER, EO-1 ALI and Hyperion hyperspectral data. A spectrum generated from the nine bands of an EO-1 ALI image is shown below. These tools offer a wide variety of options for earth study and analysis.