Best Practice Urban Pluvial Flood Modelling
Research based on the 2007 Hull floods, a UK major flood event
Following a request from The Association of British Insurers to the UK Environment Agency to increase the robustness and extent of pluvial flood modelling Ambiental undertook a robust pluvial modelling test to set a best practice standard on modelling inputs.
The test set out to determine which set of inputs gave the highest level of predictive accuracy for urban pluvial flood modelling. Inputs analysed included a variety of topographic surface representations, various topographical resolutions, and different approaches to employing building information within models.
The Study Approach
A variety of Digital Terrain Models (DTM) and Digital Surface Models (DSM) were created, these included:
- DSMs – buildings, trees and other features included
- ‘Bald earth’ DTMs with buildings removed
- DTMs with additional building data (including heights and roof features) but excluding vegetation
Each of these were then resampled to different horizontal grid resolutions (2, 3, 4 & 5m for LiDAR; 5m for IfSAR) to understand resolution effects. Hydrogical data was based on gauge data from the University of Hull which showed over 110mm of sustained rainfall with rates of over 6mm/hour – an event estimated to be in excess of a 150 year return period based on the CEH flood estimation handbook.
To enable binary classification of a buildings’ flood status building-level information was extracted from modelled flood depth grids using GIS analysis and comparison made with observed flood data to determine:
- Correct wet predictions: buildings correctly predicted to flood
- Correct dry predictions: buildings correctly predicted not to flood
- False wet predictions: buildings predicted to flood which didn’t flood
- False dry predictions: buildings predicted not to flood which did flood
Predictive ability of each simulation was assessed using a measure of fit which penalises over- and under-prediction of flooded buildings. To identify those model simulations which exhibit a significant degree of ‘skill’ in predicting flooded and non-flooded buildings (as opposed to those correct due to chance) the Kappa statistic was used.
Primarily due to the overprediction of flooding caused by artefacts in DSM’s these performed less well than either DTM’s. Of the DTM approaches, simulations based on DTM + buildings slightly underperformed bald earth DTM due to the effect adding detailed building information has upon floodplain storage. LiDAR was proven significantly better than IfSAR for building-level pluvial modelling due to poorer representation of flow paths and accumulation areas with IfSAR data.
To ensure optimum predictions of the likelihood of individual buildings flooding in urban pluvial situations the following inputs should be used:
- A LiDAR DTM using a ‘bald earth’ approach
- When using DTM rather than DSM a grid resolution of 5m was sufficient to produce high quality predictive results
- When using DSM data a grid resolution o 3m should be employed
- Wherever possible validation as well as pluvial flood modelling should be carried out at the building scale
“We found that including building data actually reduces model effectiveness.”
David Martin, Technical Director