Deriving community vulnerability indices by analyzing multi-resolution space-borne data and demographic data for extreme weather events in global cities

Abstract

Climate change in the Anthropocene is cited as one of the main reasons for extreme weather events around the world. There is a need for rapid vulnerability assessment models that may be deployed in near-real time to assist stake holders in driving the required mitigation and adaptation efforts to such events, especially in the Global South. We present an approach to derive Community Vulnerability Index (CVI) to flooding by using an integrated geospatial computational model and illustrate the methodology for six global cities. Time-series of multi-resolution space-borne optical and synthetic aperture radar (SAR) data were analyzed to derive land cover metrics from before and after flood imagery for delineating and mapping the land cover and the spatial extent of flash floods. A texture-based analysis was employed on the SAR data to delineate the spatial extent of flash floods, and this was integrated with the land cover metrics derived from the optical Sentinel-2 (Level 2A) orthorectified surface reflectance imagery. Land cover classification based on supervised image classification yielded a minimum overall accuracy of 71% when only SAR data was available after floods, and increased to 87% with the availability of cloudless Sentinel-2 data. A weighted index method, using weights derived through Factor Analysis, was used to assign appropriate weights to the composite variables derived from satellite datasets and demographic datasets for developing the CVI. The model was developed and tested on global cities with contrasting economies and validated by cross checking results with reliable open source surrogate data. The model provides an effective, adaptable, and cost-effective approach for vulnerability assessment and may provide information for policy makers to design appropriate resilience and adaptation efforts.