Deciphering the Geology of Northwest Yemen by Remote Sensing
Michael Lewis – Dec 2009
ES 775, Advanced Image Processing, Emporia State University

Table of Contents
Abstract and goals
Introduction to the Area of Interest including a brief overview of the local geology
Overview of publicly available resources
Methods of enhancement
Products
Conclusion

NCOL.BMP
An ancient volcanic field in Northwest Yemen.

Abstract and Goals

Complex geology and conflict coexist in many places around the world. Simple analysis using remote sensing tools and techniques can reveal a great deal on information with minimal local knowledge. In concert with even limited knowledge of the area the products acquired from image processing remote data can reveal complex and intricate detail.

Arab tribal regions such as Yemen, Somalia, and Iraq have been a hotbed of both conflict and resource exploration in recent years. Rapidly changing political landscapes require the ability to understand the physical setting of regions that have traditionally had only limited scope exploration. Using publicly available images from the Landsat program, I will attempt to identify and classify the area using automated processes.

Introduction to the Area of Interest (AOI)

Northwest Yemen is rugged country. The Arabian shield is fractured in multiple faults while Quaternary and Tertiary volcanism and intrusions have pushed through sedimentary rocks of the Jurrasic and Paleozoic creating a jigsaw pattern of varied landscapes. This combined with the rapid changes in elevation, and with it rainfall, creates a landscape of remarkable variety. The area of interest includes the capital of Sana’a, a metropolitan of 1.7 million, ancient volcanoes, and desert ergs.

I began my inquiry into this area when I saw what appeared to be an almost perfect complex impact crater near the village of Shaharah yet I could find no mention of it in any literature. As I began researching the area further and analyzing imagery, more of these structure appeared and I realized I was looking at an ancient volcanic field. This is an example of the importance of Image processing and analysis.


Yemen Geology around Sana.jpg

Figure 1 Yemen Geological Survey and Mineral Resources Board map  of basic rock and geologic landform types with the footprint of the Landsat data used in analysis overlaid in red.

 

Overview of Publicly Available Sources

To analyze this area I used publicly available Landsat 5 and Landsat 2 data from the United States Geological Survey website. The Landsat 2 Multi-spectral Scanner data was originated on October 6th, 1975. It is 79m resolution consisting of 4 bands designated

Landsat 1-3 MSS Spectral Bands

Band

Wavelength (micrometers)

Resolution

Band 4

0.5-0.6

79

Band 5

0.6-0.7

79

Band 6

0.7-0.8

79

Band 7

0.8-1.1

79

 

Landsat 5 Thematic Mapper (TM) data was captured October 23rd, 1998.  From comparing the wavelengths and resolutions of the Multi-Spectral Scanner (MSS) instrumentation with the Thematic Mapper data there is a much deeper penetration into the Infrared by TM as well as a narrower focus on spectral signatures in each band. I will be using the MSS data to create a ratio analysis layer that will accentuate the types and surface expression of the AOI while I will use the TM data to attempt an isocluster analysis that will attempt to accomplish the same goal.

Landsat 4-5 TM Spectral Bands

Band

Wavelength (micrometers)

Resolution

Band 1

0.45-0.52

30

Band 2

0.52-0.60

30

Band 3

0.63-0.69

30

Band 4

0.76-0.90

30

Band 5

1.55-1.75

30

Band 6

10.40-12.50

120 (resampled at 60)

Band 7

2.08-2.35

30

 

Methods of Enhancement

The first method of enhancement will be using the MSS data to create a ratio composition. To do this haze must be corrected for, this can be accomplished by comparing the minimum value of bands 1 and 2 (4 & 5) with band 4 (band 7). Since energy traveling at the infrared wavelength will tend to be unaffected by haze that should be used as the baseline. None of the images I used required correction for haze.

I began by stretching each image with 2% saturation at the ends and performed a scalar +1 calculation to prepare for ratio dividing procedures. At this point I use the suggested ratio composition for geology using MSS data in a ½, ¼, ¾, composition. To do this I run an overlay dividing the stretched images to achieve the desired component ratio. I then run a histogram on each divided overlay and continue to narrow in on a stretch that will provide only the significant spread of data using .01 cohorts. Stretches were generally performed in the 0.01 to 1.4 range.

MSS12.jpg

This ratio of Bands 1 and 2 appear nearly black with very little information until a histogram is run. The data in a ratio is as rich as an image but is compressed in a small range and must be stretched before its true value can be realized.

The same data after stretching becomes

And the image reflects the exaggerated values.

12.jpg

Applying this to all three ratios and creating a composite in MSS gives us the ratio composite image.

Products

 

In this we can plainly see the blue of vegetative color, bright yellow of recent (Quaternary) volcanic flows, and the pale yellow of Paleozoic sediment rock. The MSS instrumentation does not have high resolution however so next I attempt the same ratio using Landsat 5 Thematic Mapper data.

FALSE COLOR RATIO.BMP

While this ratio is better at differentiation features in moist environments it has lost some of the geological differentiation visible in the MSS ratio composite. As noted in the graph band wavelength table, there is a difference in the MSS and TM band sensitivity. The higher resolution does provide some noticeable differences such as the defined agriculture on the western edge in a flood plain.

 

To better realize an image that would differentiate geology I attempted a ratio suggested in the online remote sensing tutorial by Nicholas M. Short Sr., a 1/7, 4/2, 3/1 composite.

 L5 17 42 31 COMP.BMP

Here the Volcanic flow becomes a salmon pink in the lower center of the image, the City of Sana’a (lower center right) becomes a magenta, and the widespread ancient basalt becomes a marine green, interestingly some rocky portions of the north central and northeast take on a purple hue signifying a variation from the surrounding rock not visible in other composites. Additional inspection with a catalogue of signatures or ground study could reveal what that difference is.

Ratios are not the only the only products available for identifying structures. A false color composite properly adjusted to maximize contrast can be very useful in understanding the composition and land usage of a region.

SANA'A FC.BMP 

While the other composites may prove a useful tool for rock and structure differentiation, the similarity of the false color to natural color combined with the boldness of red to represent vegetation presents a great deal of information to the analyst quickly. Vegetation is present in large swaths only in the mountains in the southwest and south of Sana’a, sand becomes dominate in the northeast and fills many of the central wadi valleys in the central portion. Some farming exists in the large central drainage and south and west appear to have experienced recent volcanism.

  From the above images We have been able to glean a reasonable amount of information to be able to understand the basic layout of the region, we have identified many features and differentiated some without yet being able to identify them. Running an isocluster analysis will allow me to narrow down the wide variety of the area to the significant land-cover types. 10 major classes appear in the isocluster analysis and by cross referencing features that are known to their class, the following image can be made.

 

Conclusion

Using only a basic map of the nation’s geology and remotely sensed data we were able to create a fairly detailed and descriptive image of the region. Each of the images has advantages in furthering the identification of features in an area while the collection adds to the full understanding of the region. The images can be used in conjunction to reveal even more data. Using the isocluster, made from evaluating the density of clusters in comparison to the other images and basic geological map, the area of rock differentiated in the Landsat 5 ratio composite in the northeast can now be identified as containing more or being more similar to Precambrian rock than the surrounding giving it the distinctive signature.

 

 

References

Short Sr., Nicholas M, 2005. Remote Sensing Tutorial. http://rst.gsfc.nasa.gov/start.html. Accessed 11/01/09.

Yemen Geological Survey & Mineral Resources Board. http://www.ygsmrb.org.ye/geo_of_yemen.htm. Accessed 11/01/09.

Earth Resources Observation and Sciences Center. United States Geological Survey. http://edcwww.cr.usgs.gov/#/. Accessed 11/01/2009.

 

 

.