10 th International Conference on Hydroinformatics HIC 2012, Hamburg, GERMANY PLUVIAL FLOOD MODELLING AND HAZARD ASSESSMENT FOR LARGE SCALE URBAN AREAS ALBERT S. CHEN, SLOBODAN DJORDJEVIĆ Centre for Water Systems, College of Engineering Mathematics & Physical Sciences, University of Exeter, Harrison Building, North Park Road, Exeter, EX4 4QF, United Kingdom The pluvial flooding of five boroughs in south east London, as the consequence of climate change, was analysed in the study using the rainfall from a stochastic model in 2D hydraulic modelling. Computational efficiency of modelling was improved by omitting minor rainfall events from the simulated rainfall data according to a filtering scheme based on the assumed piped drainage capacity in the study area. The infiltration rate was used to mimic the function of sewer network in such a large urban area. The modelling results were aggregated to the postcode area unit to help local stakeholders understand flood risk in the context of combined hazards from different extreme weather events. INTRODUCTION The IPCC's Fourth Assessment Report (AR4) [1] has concluded that global warming is evident from observations of various indices. Extreme weather events (EWEs) such as floods, heat waves and droughts are projected to be more frequent and intense over the 21st century. The Stern Report on the Economics of Climate Change [2] identifies that even if we could stop all greenhouse gas emissions tomorrow our climate would continue to change due to global warming driven by over a century of manmade emissions. As a consequence, the risk of EWEs is likely to continue increasing over the next half century and today's weather extremes are probable to become tomorrow's norms. Bates et al. [3] report a significant change in precipitation attributes since the 1970s and states that heavy rainfall events are very likely to increase over many areas. Following the extreme flood events in 2007, the Pitt Review [4] urged the UK government to support and encourage community activities likely to improve flood resilience as a means of reducing flood risk. As the consequence, the Environment Agency (EA) of England and Wales developed the Surface Water Flooding Maps (SWFM) to indicate the areas prone to pluvial flooding. The SWFM only considered single scenario of pluvial flooding, which was 200 year return rainfall of 6.5h duration, and the function of sewer systems in urban area was not taken into account either. The assumption might be adequate to examine the worst scenario, however, it tended to overestimate the flood depths and extents for other 'more frequent' events, due to ignoring the function of drainage networks. In addition, spatial and temporal variations of rainfall need to be taken into account, such that flooding due to the runoff generated by the torrent rainfall occurring in nearby upstream catchment. On the other hand, the exceed runoff within a catchment could be adequately modelled.
The study presented in this paper is a part of the multi-disciplinary project Community Resilience to Extreme Weather' (CREW), which focuses on assessing the combined probability and impacts of extreme weather events (flooding, heat waves, strong wind, drought, subsidence and lightning) in the future. The study area in the CREW project includes five boroughs in the South East London Resilience Zone (SELRZ). For the purpose of runoff modelling, this area was extended to the catchment boundary. To assess the pluvial flooding impact over such a large study area due to the climate change, the synthetic precipitation data generated by a stochastic rainfall model was used as input to the hydraulic. The results were aggregated within a postcode area, via the originally developed index termed Hazard Number (HN), in order to enable assessment of combined hazards from other extreme weather events defined on the same spatial scale. METHODOLOGY The Urban Inundation Model (UIM) [5], a 2D non-inertial overland flow model that can be driven by a spatially and temporally varying rainfall input was used in this study, to simulate the pluvial flooding. The computational expense of the required hydraulic simulations is however prohibitively high and so a methodology is devised to identify the events that provide significant extreme events. A filtering scheme was therefore developed in order to identify events that are likely to lead to flooding, thus enabling focusing on hydraulic simulation of only those major events. Rainfall input The pluvial flooding uses the rainfall data generated by a spatial-temporal stochastic rainfall model, which was fitted to observed or future rainfall statistics for the SELRZ, as the inputs. The statistical model produced a long time series (~100 years) of hourly rainfall data with 5 km grid resolution for reflecting the weather condition of a specific future scenario. A nested stochastic rainfall disaggregator, conditioned by the synthetic hourly- 5km rainfall data, was applied to capture the rapid change of rainfall in urban areas. This models rainfall at the finer spatial/temporal resolutions (2km/min). Further details about the rainfall generator model can be found in [6]. Event filter In the study, 114 grid cells were used in the model such that each cell has a different long rainfall series. Most of time the rainfall series is dry weather without rainfall. The majority of the wet records had rainfall intensities smaller than 10 mm/hr, for which the soil infiltration and the urban drainage system should be capable to convey the runoff. To focus on the heavy rainfall events that may incur flooding, an event filter was developed for selecting the major events from the rainfall data. The filter first selected the events based on the rainfall intensity in the time series of each rainfall grid. The rainfall series was independent for each grid such that the filter further integrated the selected events for individual grids into catchment events for modelling.
Legend Thames Borough Boundary Study Area Boundary gauges2009.txt Events 15-min grid centre #* hourly grid centre Figure 1 The grids of the statistical rainfall model for the SELRZ Drainage and infiltration The study area is highly urbanised and the existing sewer systems play a role in conveying surface runoff. To model the sewer network for such a large area, a huge effort to prepare the data and run the simulation would be required. Due to limited resource for collecting the information, an alternative approach was adopted by assuming that the function of drainage systems can be mimicked by setting the infiltration rate without storage capacity limit in the UIM. The rate was determined based on the BS EN 752 [7], which defines the standard sewers should be able to cope with 1 in 20 year event without flooding occurring. Rainfall intensity of 1 in 20 year return period 30min duration storm was set as the infiltration rate in urban areas. For rural areas, the Hydrology of soil type (HOST) [8] data was used to determine the soil infiltration rate. RESULTS AND DISCUSSION The rainfall events selected by the event filter were used for hydraulic simulations. Like most other hydraulic models, the UIM can provide flood depths map as shown in Figure 2 for various scenarios of the study area. The flood depth is then overlapped with the buildings and postcode areas from the Ordnance Survey Mastermap, shown in Figure 3, to identify the potential hotspots that can be hit by pluvial flooding. Although the buildings in the region in the middle of Figure 3, which has flood depths less than 0.1m and surrounded by flooding, are not directly inundated in this particular scenario, the residents might still be affected by flooding due to the interruption of traffic, the cut of energy and water supplies,
etc. Hence, a further analysis is required to inform stakeholders with an overview of flood impact at the local community level. Figure 2 Maximum simulated flood depths in SELRZ for various scenarios Figure 3 The map of flood depths, buildings and postcode areas at community level The index Hazard Number (HN) was then proposed to reflect the flood condition over a postcode area. This spatial scale was chosen in order to enable consideration of combined
hazards from different types of extreme weather (all to be defined on the same spatial scale). The HN is determined by the converting the following equation HN=2 (a+b+c) (1 c) (1) where, a is the maximum flood depth parameter, determined by the maximum flood depth D max within a postcode area a= 0 for D max <= 0.1m a= 1 for D max >= 0.6m Linear interpolation for 0.1m < D max < 0.6m b is the average flood depth parameter, determined by the average flood depth D avg within a postcode area b= 0 for D avg <= 0.1m b= 1 for D avg >= 0.6m Linear interpolation for 0.1m < D avg < 0.6m c is the flooded area ratio parameter, determined by the area ratio F area_ratio of flooded extent within a postcode area c= 0 for F area_ratio <= 0.2 c= 1 for F area_ratio >= 0.5 Linear interpolation for 0.2 < F area_ratio < 0.5 The lower bounds of D max and D avg for setting parameters a and b were set as 0.1m, which is lower than curb height, by assuming such flood depth has little impact to traffic and buildings. The upper bounds of D max and D avg for setting parameters a and b were set as 0.6m by considering the flood depth would cause substantial damage and it would be unsafe for pedestrian or vehicles to move. The upper and lower bounds of F area_ratio were adopted based on sensitivity analysis of modelling results. Figure 4 shows an example of testing. Figure 4a is the flood depth map, and Figure 4b and 4c are the HN maps calculated using 0.1 and 0.2 as the lower bound of F area_ratio for determining parameter c, respectively. The flood extent of the flood depth map and the HN maps in the two circled regions are compared. The HN map in Figure 4b, which used 0.1 as the lower bound of F area_ratio, obviously over-describes the flood extents, comparing to Figure 4a. By changing the lower bound to 0.2, the HN map in Figure 4c shows better agreement of flood extent to Figure 4a. According to Eq. (1), a high HN indicates a postcode area that is likely to be affected by high flood depths and a wider flood spread. Floods with HN lower than one have negligible consequences. The HN maps were then produced to illustrate the postcode areas that could be affected by pluvial flooding at community level. Figure 5 shows an example of the HN map at community level. The postcode areas in the aforementioned region, which was surrounded by flooding in Figure 3, were assigned a HN between 2 and 6. The indices show that buildings in these areas could also suffer impact of flooding although
they are not directly inundated. The map shows that the HN describes the flooding conditions in the region better than only using the flood depths next to a building. (a) flood depth map (b) HN map with 0.1 as the lower bound of F area_ratio (c) HN map with 0.2 as the lower bound of F area_ratio Figure 4 The comparison of flood extent between flood depth map and HN maps Within the CREW project, the web-based tool What-If Scenario Portal (WISP) was developed to display the key outputs of the project as they relate to selected scenarios for three stakeholder groups, namely Householders, Small and Medium-Sized Enterprises (SMEs) and Decision makers. The flood depth and the HN data were integrated with other research outputs from other CREW partners into the WISP. By simply inputting the postcode, stakeholders can find out the hazard information from the user-friendly graphical interface of the WISP for the areas they are interested in. The WISP also provides a function that allows stakeholders to estimate the vulnerability to the combination of multiple hazard types of the study area based on the user-defined weighting factors. Further information about the WISP can be found via http://www.extreme-weather-impacts.net/flexviewer/.
The HN was also applied to a larger spatial aggregation unit Lower Layer Super Output Area (LSOA) with further analysed information for investigating the relationship between flood risk, the real estate prices and the local employment [9]. Figure 5 The HN map for postcode areas at community level CONCLUSIONS The study demonstrates the analysis of pluvial flooding using spatio-temporal varied rainfall data from a stochastic model to account for the influence of climate change. The long time series of rainfall data was screened by an event filter to select the major events for modelling. An alternative approach was applied to reflect the function of drainage system and soil infiltration in modelling for the large scale study area. The modelling outputs were further integrated as the HN for hazard assessment to describe the flood situations in postcode area unit. The application enables the stakeholders to query the hazard information easily by inputting the postcode in the WISP web based tool and to have better understanding of the flood impact in the future. ACKNOWLEDGMENTS The research presented in this paper is supported by the UK EPSRC funded SWERVE - Severe Weather Events Risk and Vulnerability Estimator programme package (Grant EP/F037422/1) under the CREW - Community Resilience to Extreme Weather project (Grant EP/F036795/1). Thanks are due to the Ordnance Survey and the National Soil Resource Institute for the provision of digital map data and soil type data respectively. The authors also appreciate the high performance clusters the ASTRAL at Cranfield University HPC and the ZEN University of Exeter for provision of computational resources. The authors are also grateful to Alex Nixon from the Greater London Authority and Roger Street from UKCIP for their support and guidance throughout the CREW project; Aidan Burton, Stephen Blenkinsop and Hayley Fowler from Newcastle University for provision of the stochastic rainfall data; Stephen Hallett from Cranfield University for his
leadership of the the CREW project; and Gwilym Pryce and Yu Chen from Glasgow University for developing the application of the Hazard Number concept in investigating the influence of flood risk on house prices and employment. REFERENCES [1] Bernstein L., et al., "Climate Change 2007 - Synthesis Report", in, Intergovernmental Panel on Climate Change, Geneva, (2008). [2] Stern N., "Stern Review on the Economics of Climate Change", in, Cabinet Office - HM Treasury, Cambridge, (2007). [3] Bates B., Kundzewicz Z.W., Wu S., Palutikof J., "Climate Change and Water", in, Intergovernmental Panel on Climate Change, Geneva, (2008), pp. 210. [4] Pitt M., "The Pitt Review: Lessons learned from the 2007 floods", in, Cabinet Office, London, (2008), pp. 505. [5] Chen A.S., Hsu M.H., Chen T.S., Chang T.J., "An integrated inundation model for highly developed urban areas", Water Sci Technol, 51 (2005) 221-229. [6] Burton A., Glenis V., Bovolo C.I., Blenkinsop S., Fowler H.J., Chen A.S., Djordjevic S., Kilsby C.G., "Stochastic rainfall modelling for the assessment of urban flood hazard in a changing climate", in: BHS Third International Symposium, Managing Consequences of a Changing Global Environment, British Hydrological Society, Newcastle, UK, (2010). [7] British Standards Institution, "Drain and sewer systems outside buildings - Part 4: Hydraulic design and environmental considerations", in, British Standards Institution, (1998). [8] Boorman D.B., Hollis J.M., Lilly A., "Hydrology of soil types: a hydrologically based classification of the soils of the United Kingdom", in, Institute of Hydrology, Wallingford, (1995), pp. 137. [9] Chen Y., Fingleton B., Pryce G., Chen A., Djordjević S., "Implications of Rising Flood Risk for Residential Real Estate Prices and the Location of Employment", Journal of Property Research, in press (2012).