Validating Assessments of Seismic Damage Made from Remote Sensing

Abstract

Assessments of damage following the 2010 Haitian earthquake were validated by comparing three datasets. The first, for 107,000 buildings, used vertical aerial images with a 15–25 cm spatial resolution. The second, for 1,241 buildings, used Pictometry images (oblique angle shots with a resolution of about 10 cm taken in four directions by aircraft). The third dataset, for 142 buildings, used ground observations. The ground observations confirmed the tendency of remote sensing to underestimate the proportion of heavily damaged and collapsed buildings, and the difficulty of making remote assessments of moderate or low damage. Bayesian statistics and sample surveys made from Pictometry images and ground observations were used to improve remote damage assessments from vertical images. The possibility of developing standard factors to correct remote assessments is discussed. The field exercise pointed to the need to produce an internationally agreed-upon set of damage definitions, suitable for postdisaster needs assessments as well as for other uses.