Articles | Volume 16, issue 12
Research article
30 Nov 2016
Research article |  | 30 Nov 2016

Analyzing the sensitivity of a flood risk assessment model towards its input data

Hanne Glas, Greet Deruyter, Philippe De Maeyer, Arpita Mandal, and Sherene James-Williamson

Abstract. The Small Island Developing States are characterized by an unstable economy and low-lying, densely populated cities, resulting in a high vulnerability to natural hazards. Flooding affects more people than any other hazard. To limit the consequences of these hazards, adequate risk assessments are indispensable. Satisfactory input data for these assessments are hard to acquire, especially in developing countries. Therefore, in this study, a methodology was developed and evaluated to test the sensitivity of a flood model towards its input data in order to determine a minimum set of indispensable data. In a first step, a flood damage assessment model was created for the case study of Annotto Bay, Jamaica. This model generates a damage map for the region based on the flood extent map of the 2001 inundations caused by Tropical Storm Michelle. Three damages were taken into account: building, road and crop damage. Twelve scenarios were generated, each with a different combination of input data, testing one of the three damage calculations for its sensitivity. One main conclusion was that population density, in combination with an average number of people per household, is a good parameter in determining the building damage when exact building locations are unknown. Furthermore, the importance of roads for an accurate visual result was demonstrated.

Short summary
Adequate flood damage assessments can help to minimize damage costs in the SIDS. Data availability is, however, a major issue in these areas. In order to determine the minimal data necessary for an adequate result, a sensitivity analysis was performed on the input data. This has shown that population density, in combination with an average number of people per household, is a good parameter to determine building damage. Furthermore, a complete road dataset is visually indispensable.
Final-revised paper