Modelling, mapping and visualisation of flood inundation uncertainties

Sammanfattning: Flood maps showing extents of predicted flooding for a given extreme event have wide usage in all types of spatial planning tasks, as well as serving as information material for the public. However, the production processes that these maps undergo (including the different data, methods, models and decisions from the persons generating them), which include both Geographic Information Systems (GIS) and hydraulic modelling, affect the map’s content, and will be reflected in the final map. A crisp flood boundary, which is a common way of representing the boundary in flood maps, may therefore not be the best representation to be used. They provide a false implication that these maps are correct and that the flood extents are absolute, despite the effects of the entire modelling in the prediction output. Hence, this research attempts to determine how flood prediction outputs can be affected by uncertainties in the modelling process. In addition, it tries to evaluate how users understand, utilise and perceive flood uncertainty information. Three main methods were employed in the entire research: uncertainty modelling and analyses; map and geovisualisation development; and user assessment. The studies in this work showed that flood extents produced were influenced by the Digital Elevation Model (DEM) resolution and the Manning’s  used. This effect was further increased by the topographic characteristic of the floodplain. However, the performance measure used, which quantify how well a model produces result in relation to a reference floor boundary, had also biases in quantifying outputs. Determining the optimal model output, therefore, depended on outcomes of the goodness-of-fit measures used. In this research, several ways were suggested on how uncertainties can be visualised based on the data derived from the uncertainty assessment and by characterising the uncertainty information. These can be through: dual-ended maps; flood probability maps; sequential maps either highlighting the degrees of certainty (certainty map) or degrees of uncertainty (uncertainty map) in the data; binary maps; overlain flood boundaries from different calibration results; and performance bars. Different mapping techniques and visual variables were used for their representation. These mapping techniques employed, as well as the design of graphical representation, helped facilitate understanding the information by the users, especially when tested during the evaluations. Note though that there were visualisations, which the user found easier to comprehend depending on the task given. Each of these visualisations had also its advantages and disadvantages in communicating flood uncertainty information, as shown in the assessments conducted. Another important aspect that came out in the study was how the users’ background influence decision-making when using these maps. Users’ willingness to take risks depended not only on the map, but their perceptions on the risk itself. However, overall, users found the uncertainty maps to be useful to be incorporated in planning tasks.