Principal Component Analysis

...

[pic 6]

Figure 6

When comparing Figure 4 and Figure 3. The above feature in Figure 6 represents bare rock in Figure 3. Using the colour slider, the turquoise colour above is a RGB combination of 99,203,205 which is represented as 39% red, 80% green and 100% blue. From this we can see that the blue factor and green factors are predominant. This means that this feature is represented by both component two, which is represented by green, and component three, which is represented by blue.

This is reiterated by the fact that the main factors in component three are band four and band six. Band four represents vegetation slopes and band six represents moisture content of soil and vegetation. This coincides with the feature as it is representative of a slope.

[pic 7]

Figure 7

When comparing Figure 4 and Figure 3. The above feature represents landmass in Figure 3. Using the colour slider, the green colour above is a RGB combination of 34, 241, 73 which is represented as 13% red, 95% green and 29% blue. From this we can see that the green factor is predominant. This means that this feature is represented by component two (since component two is represented by green).

This is reiterated by the fact that the main factor of component two is band five. Band five shows information that is used to emphasize biomass content. This coincides with the fact that the above Figure 7 is a land portion.

[pic 8]

Figure 8

When comparing Figure 4 and Figure 3. The above feature in Figure 8 represents vegetation in Figure 3. Using the colour slider, the RGB combination for the colour of the above feature is 179,50,156 which is represented by 70% red, 20% green and 61% blue. From this we can see that the red and blue factors are predominant, however the red factor is more predominant than the blue. This means that this feature is represented mostly by component one and secondly by component three.

This is reiterated by the fact that the main factors for component one is band five and band six. Band five and band six represents biomass and vegetation which is present in this feature.

[pic 9]

Figure 9

When comparing Figure 4 and Figure 3. The above feature in Figure 9 represents very dense vegetation in Figure 3. Using the colour slider, the RGB combination for the yellowish colour in Figure 9 is 245, 229,76 which is represented by 96% red, 90% green and 30% blue. From this we can see that the red and green factors are predominant. This shows that the feature in Figure 9 is represented by component one and component three.

This is reiterated by the fact that the main factors in component one is band five and band six and the main factors in component three are band four and band six. Band four represents vegetation slopes and band six discriminates moisture content of soil and vegetation. Band five and band six represents biomass and vegetation which is present in this feature. This coincides with the above feature because these bands show the density of the vegetation shown in the above feature

[pic 10]

Figure 10

When comparing Figure 4 and Figure 3. The above feature in Figure 9 represents a water body in Figure 3. Using the colour slider, the RGB combination for the above feature is 255,255,255 which is represented by 100% red, 100% green and 100% blue. This is because the feature is represented by white. It comprises of all three components.

In Figure 3, it can be seen that this water body is very dark which means that it absorbs more than it reflects. It requires a combination of all three components to shown in the false colour composite.

Conclusion

Principal component analysis is a very useful technique to determine the amount of variation spread across the different bands used. The results of the analysis were the components and the amount of variation represented by each component. In gaining this, we can see which components don’t have much impact on the variation of the data. The conclusion drawn from the principal component analysis was that the first three components held 99,03% of the variance in the data cumulatively therefore this shows that the remaining components aren’t really necessary.

The principal component analysis also was used to calculate the Eigen vectors which were used to show which bands were significant factors of each component.

In creating a colour composite using the first three components, we were able to visualise the component data. From the images, it can be seen that component one represents dense vegetation or vegetation that shows as dark green on the natural colour composite, component two represents areas of landmass that aren’t typically covered with green vegetation and component three represents bare surfaces such as the bare rock face feature in the image.

In using principal component analysis in conjunction with the Landsat bands; it helps provide a better understanding of what the different bands represent as well as which bands have the most variation within the data.