Skip to Main content Skip to Navigation
Theses

Prise en compte de l'incertitude dans l'expertise des risques naturels en montagne par analyse multicritères et fusion d'information

Abstract : Rapid mass movements hazards such as mountain rivers (torrents) and snow avalanches put humans and material assets at risk with dramatic consequences. Risk reduction is reached through structural and non-structural measures such as zoning control, preventive information. In a context of lacking knowledge about the natural phenomena, expertise is required to provide analyses for decision and risk management purposes using multi-disciplinary quantitative or qualitative approaches. Expert judgements depend on the availability, quality and uncertainty of the available information that may result from measures, historical analysis, testimonies but also subjective, possibly conflicting, assessments done by the experts themselves. At the end, phenomenon scenarios and decisions may very well rely on very uncertain and conflicting information without being able to really know what was completely true, imprecise, conflicting or simply unknown in the hypotheses leading to the result. At the same moment, traceability and shared decision-aid tools are expected from the stakeholders for better understanding and use of expertise results in an integrated risk management framework able to consider both technical, environmental and social aspects of decision. In that context, this work analyses relations between information and decision including the integration of results in information systems. On one hand, it contributes to improve traceability and quality description of the expertise process and, on the other hand, it provides and analyzes multi-criteria aid-decision tools able to consider information imperfection (uncertainty, imprecision) resulting from conflicting sources. Firstly, we analyze the ability of multi-criteria decision analysis to help risk assessment, including a diagnosis of existing empirical approaches developed to combine the risk components evaluations. A methodology is proposed to transform those existing models mainly related to lists of criteria (one-level hierarchy, meaning no hierarchy) in real hierarchical decision models. This allows to elicit in a clearer way the relative preferences chosen by the analysts between individual or grouped criteria. On this basis, any hierarchical model can be transformed in an another one, keeping the same decision results and highlighting some new preferences views on the problem such as role and importance of scenarios based on aggravated criteria in the decision. Secondly, a framework mixing evidential reasoning and multi-criteria decision analysis ER-MCDA) combines qualitative and quantitative criteria on which more or less reliable and conflicting sources provide uncertain and imprecise evaluations. The Analytic Hierarchy Process (AHP) is used both to model the problem in a conceptual way and to elicit preferences between key criteria. Fuzzy Sets and Possibilities theories are used to transform quantitative and qualitative criteria into a common frame for Dempster-Shafer Theory (DST) and Dezert-Smarandache Theory (DSmT). In that context, DSmT proposes more valuable modeling principles for vague, imprecise and uncertain information and conflict management. As an example, information fusion allows merging different uncertain and imperfect pieces of information to identify the sensitivity of an avalanche prone area. A simplified decision sorting problem based on a snow-avalanche risk management problem shows how the use of multi-criteria decision analysis principles and information fusion can be used to characterize and take information quality or imperfection into account for decision purposes. This methodology can be extended to spatial applications such as hazard and risk zoning maps. Mixing information fusion and multi-criteria decision points out some important issues such as basic belief assignments elicitation, conflict identification and management, fusion rule choices, validation of results but also specific needs to make a difference between importance and uncertainty in the fusion process. At the end, multi-criteria decision-aid methods can be viewed as a valuable and alternative support for risk components evaluation. Traceability through easily understandable methodologies (hierarchical approaches, fuzzy numbers) and information imperfection assessments can be viewed not only as ways to improve the global quality of the expertise and decision processes but also as creating a new space for interaction between expertise and society. In addition to their technical interest, they may become unexpected but valuable support to an integrated risk management approach.
Document type :
Theses
Complete list of metadata

Cited literature [570 references]  Display  Hide  Download

https://hal.inrae.fr/tel-02592979
Contributor : Migration Irstea Publications <>
Submitted on : Friday, May 15, 2020 - 4:48:54 PM
Last modification on : Thursday, April 15, 2021 - 4:58:05 PM

File

pub00028329.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-02592979, version 1
  • IRSTEA : PUB00028329

Citation

J.M. Tacnet. Prise en compte de l'incertitude dans l'expertise des risques naturels en montagne par analyse multicritères et fusion d'information. Sciences de l'environnement. Doctorat Sciences et Genie de l'Environnement Ecole des Mines de Saint-Etienne, 2009. Français. ⟨tel-02592979⟩

Share

Metrics

Record views

175

Files downloads

319