Identification of earthquake disaster hot spots with crowd sourced data

Abstract

This paper explores the value of the application of Crowd Sourced (CS) data in identification of areas damaged in the aftermath of an earthquake. A survey was conducted to collect CS data based on two stage cluster sampling method from people who experienced the earthquake in Bam city, Iran in 2003. The CS data submission time was considered for data analysis, including continuous, discrete and complete data submission. The CS data reporting on the level of building destruction, the number of fatalities and the number of injuries was used to identify hot spot areas for dispatching response operation teams. To test the value of CS data in identification of hot spots, the results were compared with the Actual Earthquake (AE) data by using of Fuzzy Kappa index, Fuzzy Inference System, and cross tabulation to calculate similarity and dissimilarity, quality and allocation disagreement between them. The similarity and dissimilarity measures indicate that there is a low to moderate similarity between hot spot maps based on the application of CS and those based on the AE data. They suggest that CS data has a moderate potential role in identifying highly damaged areas (hot spots) and low damaged areas (cold spots). The results of this study show that the CS data is better suited for more general determination of hot and cold spot areas than to provide exact locations where the resources could be dispatched. Consequently, we conclude that CS data is useful for decision making process by disaster managers if combined with the other sources of information to allocate the limited resources in affected areas.