Cartography in Early Warning and Crisis Management

Visualization for enhanced early warning

3504.2 - Modeling and representation of spatial uncertainty in risk maps.

Monday, July 3
1:50 PM - 2:10 PM
Location: Virginia B

Cartography is an integral part of natural disaster management. Whether during the phases of prevention, crisis management, or damage assessment, spatial representations are mobilized at all stages of the risk management cycle. More specifically during the prevention phase, regulatory maps are developed to constrain urban planning according to the delineation of danger areas associated with one or more natural hazards. In these maps, danger areas are represented with crisp boundaries whereas the risk is inherently uncertain. Indeed, spatial uncertainty in the delineation of danger areas can be involved by several parameters: the intensity of the hazard (e.g. the height of a marine submersion), the methods used to model its impact (e.g. with or without attenuation according to land penetration), or the quality of the data used to delineate danger areas (e.g. the resolution of a DTM). In this context, this paper seeks to understand how uncertainty is considered in the realization of risk maps and how it is spatially represented. Based on an analysis of existing risk maps, this paper also tends to propose a model in order to parameterize uncertainty representations in the delineation of danger areas associated to marine submersion hazard.
To address these issues, a review in the domain of spatial uncertainty representation is preliminarily conducted. It shows that several contributions have proposed methods to represent spatial uncertainty, such as Fisher (1991) or MacEachren et al. (2005). Some of these methods have been experimented on risk maps (Arnaud and Davoine, 2009) and this paper discusses how these methods can be adapted to represent uncertainty in risk maps. To assess how spatial uncertainty is considered in the elaboration of risk maps, an analysis of a sample of 46 maps in France is conducted. Three types of natural hazards are concerned: flooding (15 maps), marine submersion (17 maps) and avalanches (14 maps). Results of the analysis show that the uncertainty in the delineation of danger areas is mainly related to the intensity of the hazard. Uncertainty is generally associated to a hazard level (“low”, “average”, “high”, “very high”) and no major stylistic trends can be highlighted to represent uncertainty. Indeed, the analysis of these maps show that the color is mainly used to differentiate the hazard level, combined with the value and the texture. Only avalanches maps propose a standardized symbolization, associated to a representation of uncertainty using textural elements. Based on this analysis, a library of the main styles used to represent uncertainty is extracted. Finally, a framework is proposed to model spatial uncertainty representation in the delineation of danger areas, and an experiment is proposed to illustrate it. The case study focuses on the marine submersion hazard involved by a potential tsunami in the island of Martinique (French lesser Antilles). To model danger areas, the intensity of the hazard is defined by a 10 meters high marine flood, using two different methods to simulate marine submersion (horizontal projection, and with attenuation according to penetration in land), and four altimetric datasets (LITTO3D, BDALTI, SRTM30 and SRTM90). The danger areas delineated are combined, and uncertainty representations (proposed in the literature or extracted from the analyzed risk maps) are applied, in order to propose alternative representations of marine submersion hazard according to the final user needs.
Perspectives of this work deals with the implementation of further methods to model marine submersion (using local conditions of roughness) and other representations of uncertainty in risk maps. The final goal is to develop a mapping environment allowing continuous transitions between crisp representations of regulatory maps, and uncertain representations of hazardous areas.

Jean-François Girres

Université Paul-Valéry Montpellier 3 - UMR GRED

He is an Assistant Professor in Geomatics at the University Paul-Valéry Montpellier, and member to the GRED laboratory. His research topics deal with spatial data quality, multi-scale representations and geovizualisation.


Send Email for Jean-François Girres

Send Email for Milan Konecný


3504.2 - Modeling and representation of spatial uncertainty in risk maps.

Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Modeling and representation of spatial uncertainty in risk maps.