Evaluating the effect of data and data uncertainty on predictions of flood inundation
In the light of uncertain climate change, there is a need to assess flood risk outside the realm of current experience. Flood inundation models allow river discharge upstream to be related directly to flood extent downstream. They are, therefore, potentially very useful tools that can be used in a variety of real and “what-if” scenarios. However, flood inundation models are limited by the availability and use of appropriate data.
In this thesis, the effect of data and data uncertainty on the prediction of flood inundation was assessed, using the two dimensional model, LISFLOOD-FP. Simulations were conducted on two sites in the United Kingdom. Specifically, the objective of the research was to use LISFLOOD-FP to (i) assess the suitability of elevation data available nationally in the UK (and via equivalents globally) for use in the prediction of flood inundation and (ii) assess the role of land cove in predicting inundation extent. Research assessed (i) several sources of elevation data and the effect that they had on the prediction of inundation extent, (ii) the effects of prediction uncertainty in interpolated elevation data and (iii) the use of remotely sensed land cover to predict spatially distributed friction coefficients on the floodplain and model sensitivity to them.
At the scale of modelling undertaken, the most accurate predictions of inundation extent were obtained using photogrammetric elevation. The effects of prediction uncertainty in interpolated elevation data were found to increase both with distance downstream and throughout the simulation. Remotely sensed estimates of land cover were found to be useful for generating spatially distributed friction coefficients. However, predicted flood inundation using LISFLOOD-FP was insensitive to floodplain friction.