Earthquake in El Salvador
I am a project leader for the El Salvador group of the Tufts' Chapter of Engineers Without Borders. I have traveled to El Salvador twice to work on the water filtration system that the group has implemented in a small community south of the capital. During my first visit, I learned some shocking details about the county's history from one of our contacts in the country named Mike Jenkins. On a routine car ride into the capital for supplies, Mike told me about a large earthquake that had devastated the country many years ago. Fortunately Mike was outside the country when the quake hit, but the devastation he described when he was able to get back into the country was unimaginable. It was this experience that prompted me to research the January 2001 earthquake and write this report. The earthquake was centralized 60 miles southwest of San Miguel,El Salvador and killed 944 people. About 585 of the deaths were caused by large landslides in Nueva San Salvador and Comasagua. The quake also injured 5,565 people, destroyed 8,226 houses, and damaged more than 150,000 buildings. Utilities and roads were damaged or destroyed by more than 16,000 landslides in the hours and days after the event. There were 445 landslides that were blamed for the destruction of 688 houses. The quake, which registered a 7.7 on the Richter scale, was felt from Mexico City to Colombia. Its location relative to the coast of Central Americamay be seen below.
One method for identifying areas of destruction and assessing the need for assistance is remote sensing photography. Pictures are taken of an area before and after a catastrophic event. Areas of interest are then identified and damages are assessed. In the case of the El Salvadorearthquake "before" images of the country were taken on November, 15 2000 and "after" images were taken on February, 19 2001. The images, which span an area of 183 km by 170 km, differed from regular color satellite images and were analyzed using techniques described below.
The satellite that took the pictures was a Landsat 7. This satellite was launched in April 1999 and has an additional Enhanced Thematic Mapper Plus sensor. Compared to older Landsat satellites which can only take color pictures in 30 m resolution, this model has the ability to take panchromatic images with a 15 m resolution and thermal band images with a 60 m resolution. The new sensor allows images to be crisper and cleaner than images taken with just a color camera. To get this new and better image 4 steps must be taken. The first is to convert the original image, based in the RGB (red, green, blue) scale of color, into an image based on HSI. This process is called "encoding." HSI is a representation of color via hue (dominant wavelength of color), saturation (purity of color relative to grey), and intensity (total brightness of the color). The next step is to change the spatial resolution of the converted image to match its higher quality resolution. This step is known as "manipulation." The next step is to convert the modified HIS image back into an RGB image. This step is known as "decoding." The final step is to cross lace the modified image with the panchromatic image to yield a higher resolution picture.
Once the images are converted they are ready to be analyzed. The images taken after the event are compared to those taken before. Areas of change are found by manually scanning the images. New software is available to automate this task but is not yet precise enough to be used on a large scale. Once these areas are determined the images are loaded into a GIS system where information about housing communities and population distributions are available. By overlaying the visual information with the GIS database information it is possible to estimate the extent of the destruction that is found (deaths, destroyed houses, economic effects, etc). A picture of an identified landslide (left) and a picture of that same map imported into GIS (right) may be seen below. A full analysis of this process in relation to the 2001 quake may be found in a paper entitled "Detection of Damage Due to the 2001 El Salvador Earthquake Using Landsat Images" by Miguel Estrada, Masayuki Kohiyama. This study was done at the InstituteofIndustrial Science at The University of Tokyo and may be found at: http://www.crisp.nus.edu.sg/~acrs2001/pdf/182estra.pdf
An alternate method of assessing the damage from an earthquake would be to manually travel to all the affected areas and asses the damage on site. This type of analysis is also done but on a much smaller scale and at a much slower speed. Processing satellite images is a quick and easy way to estimate the proper level of on ground assistance needed. The biggest question I have about this process is the about the speed a photograph may be captured of an effected area. In the study from The University of Tokyo the comparative image was taken a month after the initial quake. This is a function of the trajectory of the satellite and its job queue, yet I wonder if a system could be developed to capture images of an effected area weeks or even days after a catastrophic event. This would greatly add to the effectiveness of the disaster response and possibly save countless lives. I would also question the effectiveness of solely relying on satellite imagery analysis, especially in a third world country. Many of the homes in the community that I stayed in were made from a combination of adobe brick and sheet metal. It would seem to me that assessing the damage of a house whose materials so closely resemble that of the ground and scrap metal would be difficult, though for large scale events (like landslides) this system does seem like a good fit.