Assessment of the Contribution of Crowd Sourced Data to Post-Earthquake Building Damage Detection

Abstract

This paper compares the results of building damage detection based on Crowd Sourced (CS) data, image processing of remotely sensed (RS) data and predictive modelling with institutional spatial data (Spatial Data Infrastructure - SDI). In particular, it focuses on the contribution of Crowd Sourcing to detecting post-earthquake building damages, while also considering the integration of Crowd Sourced with two other data sources (RS and modelling). To simulate CS data submission following the 2003 earthquake in Bam City (Iran) a survey was administered to the population which experienced the earthquake. The results obtained from this and two other sources are compared with the Actual Earthquake (AE) data by cross-tabulation analysis and McNemar's Chi Square Test. When assessed against AE data, the average accuracy levels of assessments based on the use of RS data and CS data integrated with each RS data and predictive modelling and with both, show a statistically significant increase relative to the predictive modelling. While this research does not provide for a full assessment of the value of CS data alone and in fact finds it slightly inferior to predictive modelling, it suggests that Crowd Sourcing could be a useful source of information, especially if combined with other sources.