Facilitating the use of high-resolution EO data to support the development of mapping products for natural disasters

Abstract

Earth observation capabilities are used to respond to major disasters around the world, for humanitarian aid and security. Satellite derived information needs to be used in combination with additional data to be presented in a proper geospatial context for the work of civil protection agencies and relief organizations. This paper aims to reveal a methodology developed to quantitatively evaluate the impact of a natural disaster over a region. The proposed approach was initiated in the frame of GEODIM Project (http://geodim.meteoromania.ro) whose goal is to develop a Romanian downstream emergency response service in order to contribute to current disaster and risk management approach based on Earth observation data. The project is focused on developing experimental processing algorithms and mapping products for natural disasters (floods, earthquakes, landslides) damage assessment in urban areas based on very high resolution optical and SAR satellite imagery acquired worldwide. The usefulness of remote sensing data for natural disasters damage assessment clearly rely on the number of available images, their type and quality and last, but not least, the timeliness of the data sets, or how delayed are the available post disaster images relative to the damaging event. Previous work demonstrated the use of a semi-automated data processing method in order to reveal and determine the area affected by the disaster, considering both qualitative and quantitative approaches. The proposed scenarios consider knowledge discovery from pre and post event EO images by mapping the extracted data features into semantic classes and symbolic representations like ”buildings”, ”vegetation”, ”streets”, ”bare land”, and ”damaged buildings”, etc. In order to fully exploit the high-resolution EO data a method based on patches is proposed to extract relevant contextual information to be further used to build the situation maps.