Remote sensing vs. field survey data in a post-earthquake context: Potentialities and limits of damaged building assessment datasets

Abstract

Quick building damage assessment following disasters such as large earthquakes serves to establish a preliminary estimation of losses and casualties. These datasets are completed by employing several crowdsourcing initiatives, in which volunteers and collaborators map damaged buildings in a given area at a qualitative damage scale based on a post-earthquake aerial or satellite image. Automating this process is a temptation and a technical issue, but manual interpretation remains essential, with the identification of moderate and lateral damage being the key and limiting factor. Following the Haiti 2010 earthquake, many studies were completed by crossing multilayer data gathered from different sources (satellite, aerial, and field survey). These works created a building damage dataset that enabled the construction of different sets of empirical vulnerability functions. In the present study, we proposed to review the datasets used for the damage assessment again, investigate how they can be managed for understanding urban damage patterns, and quantify the potentialities and limits of the sets.A high-resolution map of damage in Port-au-Prince was used to obtain a deducted map of intensity and was then compared to more detailed post-earthquake investigations such as the microzonation of the city (Belvaux et al., 2018). These detailed post-earthquake investigations, in which array microtremor measurements are performed for characterization of the subsurface soil, contribute to a better understanding of local variations in intensity. Subsequently, a retro damage scenario was run, considering the different sets of vulnerability functions (using the RISK-UE methodology vulnerability indexes) fitted with empirical vulnerability functions. Using the characterization of the exposure on a remote sensing basis, the results fit the heaviest damage well (building collapse), but they overestimated moderate damage states compared to the observations. However, is an aerial image based dataset sufficiently exhaustive for moderate damage, which is mostly visible from a lateral or internal point of view? Finally, we suggested some range of adjustments that can be applied to a vulnerability assessment originating from remote sensing data such that it can be used more accurately in the detection of urban damage, even for moderate damage degrees.