In this study we will use a Tahoe land cover map as a guidance. We will also two different maps one from 1982 and one from 1992. This is so we can access and see the land changes over the ten years. There will be a series of different methods to properly perform this. These will include a series of different map classification and charts and some scatter plots.

Spectral Signatures

We can see from the graph that deep water has a very low spectral profile and low reflectance. The depth of the water greatly affects the amount of radiation that is being reflected through the atmosphere. There would be very low transmitted energy because of the depth of water. This is what gives it its unique signature. However, the reflectance is at its maximum at the blue end of the spectrum and decreases as wavelength increases. As for the Shallow water it does have an overall reflectance. This because of the low depth of the water so more light is reflected. He water could also be turbid which as some sediment suspension which increases the reflectance in the red end of the spectrum. Clear water reflects very little reflectance, but turbid water is reflects more because of the chlorophyll content of a water, for example plankton and algae. This gives it a slight green signature. If we can compare the Tahoe area from 1982 to 1992, it is harder to make out the shallow area from the bare land around it. This might mean that the water has dried up because of global warming. Its spectral signal in 1992 is similar to that of the bare rock. And we studied the same points on both maps.

Because agriculture can display different unique signatures, we are able to tell if the land is using a different type of crop and land use. Forestry also shares many characteristics with agriculture. They are both low in the red and blue region due to the absorption of chlorophyll. But we can see they both reach their peak on the green region of the spectrum. This is because of the cellular structure of the leaves. Because of these characteristics we are able to easily make out agriculture and forestry. It has a high NIR but low visible reflectance. The bare rock signature profile rates highly in all regions.

Comparing the two charts we see that there is an increase in some values and decrease in others. But overall the pixel values are the same over the ten years.

Feature plots for Tahoe 1982

Viewer 1 Viewer 2

Viewer 3, 4 and 5 Viewer 6

When viewer 3, 4 and 5 were calculated and appeared on the screen. There scatterplot were identical.

Feature plots for Tahoe 1992
Viewer 1 Viewer 2

Viewer 3 Viewer 4

Viewer 5 Viewer 6

Signature Mean Plot
We have just gone through the spectral values of this multispectral scene. Now with this information we are able to combine our data about the spectral profiles of single pixels and pairs of bands by selecting different land cover uses.

Signature mean plot for Tahoe in 1982

Signature Editor 1992

Signature Mean Plot for Tahoe in 1992

Signature Editor 1992

We can see that a few spectral areas have different wavelengths and they’re having been changes from the last ten years. In the case of bare rock we see in the red band the value has gone down slightly. Whereas in the green and blue region increased by a lot. The green went from 0.165 to 0.653 over the course of ten years. And the blue region has experienced and even higher increase. It went from 0.165 to 0.852.
We now study the change in value for shallow water. The green and blue bands are 1.00 and 1.00 in 1982. And in 1992 they are 1.00 in green and .0995. So there is not much of a change over the ten years. But we do see that in the red value that in 1982 is 0.00 and then increase to 1.00, which is a massive increase.
Forestry increases over the ten year in both the red and blue region and decreases in the green region. In 1982 green was 1.00 and decreased to 0.531 in 1992. Red went from having no value to going up to 0.590. We then the blue value going from 0.00 to 3.55
Not much of a change in the Deep Water class. Red and Green values are still 0.00. But the blue value dropped from 1.00 to 0.151 over the course of ten years.
We will now look at the last one which is bare rock. In 1982 the values were Red 0.00, Blue 0.00 and the green value was at 1.00. Then in 1992 the red value was 0.

Raster attribute table and ISODATA map for 1982

I personally believe this is not the most sufficient why to classify information. It is too difficult to distinguish from the different land uses. As we can see from the map above. Deep water can be easily made out from the other classes, the same goes for forestry. But when it comes to identifying the bare rock and agriculture it is very difficult compared to the original land cover map for Tahoe. As their clusters are spread out unevenly beside each other. You can’t find a definite shape of agriculture, which can be easily seen in the original map. Another problem that arises from this classification process is that you can’t tell the difference between shallow and deep water. Because shallow water does not have a unique cluster at all. The shallow water area of the map is now the same colour as deep water.

ISODATA map and attribute table for 1992

This map does appear to be a clearer version from the Tahoe 1982 map. We can now safely assume that the shallow lake that was there in 92 has being almost dried up. And that there is now less agriculture in the area and more bare rock. The forestry of the area has now decreased also. But there are still similar problems from the last set of results.

Signature Merge Plot Editor
This process will allow us to compare Tahoe 1982 and Tahoe 1992, by signatures generation. We do this by crating polygon shapes over a certain land cover. When we get the sginature profile of all of them and merge them together and use that one. The reults from the process gave use quite similar results to the first signature method we used, spectral signature. The table below and graph points show us this.

Merge Plot and Graph for 1992

Parallelepiped Classifer

Parallelepiped Classifer Map for 1982

This clissification process is a lot more better than the previous excersise. But some problems still arise. Fortunatly we are now able to clearly distinguish between shallow and deep water. We can see the light blue along the edges of the dark blue. Using the original land cover map for guidance we can make out areas of agriculture on the right side of the map. But I am unsure as to why agriculture appears in other places where is shouldn’t be. This process seems to show that there is more bare rock than there’s meant to be and less foestry. It seems this method of clasification is good for distinguishing shallow and deep water.

Parallelepiped Classifer Map for 1992

This map is similar to the previous but because of the colors it is hard to distinguish between shallow and deep water. We can make out though that the shallow lake at the top right apears to be almost dried up. There is less forestry in the area as well agriculture. And there is now a greater area of bare rocks.

Maximum Likelihood Classifier

Maximum Likelihood Classifier Map for 1982

So far I think this is the better classification process compared to the last two. We can easily make out the shallow and deep water classes. And the bare rock class is now seems to be more defined and correct. The only major problem I see from it is that the agriculture class is spead unevenly arcoss the forest area which doe not make sense. But you can clearly make out the agricultureally in the bottom left, that was previously seen in the origianl land cover map. The results are quite similar in the 1982 map and we can make out the land changes over the ten years.
Maximum Likelihood Classifier Map for 1982