The researchers used imagery from a high resolution commercial satellite, QuickBird, to assess the damage to buildings that resulted from an earthquake of magnitude 6.8 that struck the coast of Algeria on May 21, 2003. The study focused on the two cities of Boumerdes and Zemmouri. Using photographs from before and after the event, researchers coded buildings as either a 1-5 based on level of destruction. The higher the number, the greater the damage. They also noted the location of tents constructed to house refugees. This study proves the efficacy of remote damage assessment in the field of disaster management.
- Satellite photos cover a large area at one time
- More manageable than on-the-ground classification efforts
- Hard to compare images taken from different angles and at different resolutions
- Shadows can affect interpretation
The study revealed that destruction classifications based on 2 images (from before and after the event) were more stable and consistent across five raters than those based on just the photo from after the earthquake. The precision level was 80% meaning that the raters were in agreement about the level of damage to a structure for 80% of the buildings. Similarly, the number of damaged buildings was greater using both sets of photos than just the ones depicting the earthquake's aftermath. The use of both sets of images was more crucial for buildings with low-grade damage.
The authors also compared their satellite-aided results to those from a team of Algerian engineers conducting classification of building damage on the ground in Boumerdes. The comparison indicated that overall, people using satellite images to assess damage tended to under-estimate the degree of structural damage. The authors did not quantify this effect, but noted that it warrants future study.