The Development and Uses of Crowdsourced Building Damage Information based on Remote-Sensing

Sabine Loos, Karen Barns, Gitanjali Bhatacharjee, Robert Soden, Benjamin Herfort, Melanie Eckle, Cristiano Giovando, Blake Girardot, Gregory Deierlein, Anne Kiremidjian, Jack Baker, & David Lallemant

Crowdsourced analysis of satellite and aerial imagery has emerged as a new mechanism to assess post-disaster impact in the past decade. Compared to standard ground-based damage assessments, crowdsourcing initiatives rapidly process extensive data over a large spatial extent and can inform many important emergency response and recovery decisions. We test three approaches to crowdsourcing post-earthquake building damage using 50cm resolution satellite imagery from the 2010 Haiti earthquake. Approach 1 further develops the predominant building-level map-based assessment method that has been implemented in earlier crowdsourcing initiatives. Two novel area-based assessment approaches were also developed, where users rate the level of building damage in an image (Approach 2) and compare building damage between two images (Approach 3). The results of the two area-based approaches show a trend between crowdsourced and “true” field damage severity, which can be improved by weighting high-performing volunteers. Alternative methods, including Bayesian updating and network analysis, are proposed to analyze the paired comparison data from Approach 3. However, Approach 1 did not reach completion, because of the time intensive nature of building-level assessments.

Parallel to the crowdsourcing tests, an extensive ‘demand survey’ of interviews with post-disaster practitioners was conducted to develop a timeline of six key decisions that are dependent on post-earthquake building damage data. The resulting framework can guide future research concerning rapid damage estimates to address decision-makers’ specific, and cross-cutting needs. Considering the results of the crowdsourcing tests and the demand survey, area-based approaches are promising methods to crowdsource building damage, because of its user-simplicity and ability to address specific post-disaster decisions.

Loos, S and Barns, K and Bhattacharjee, G and Soden, R and Herfort, B and Eckle, M and Giovando, C and Girardot, B and Deierlein, G and Kiremidjian, A and Baker, J and Lallemant, D. (2018). The Development and Uses of Crowdsourced Building Damage Information based on Remote-Sensing. Blume Earthquake Engineering Center Technical Report 197. Stanford Digital Repository. Available at: https://purl.stanford.edu/bj915mt6570