SCALE
Paul Davies
Ove Arup & Partners Ltd, Sheffield, UK.
paul-d.davies@arup.com
Abstract
One of the recommendations in the Pitt Review concluded that surface water management plans (SWMP) should
provide the basis for managing local flood risk. The government‟s response was to recommend that SWMPs
should be produced where surface water risk is high. Six pilot studies were undertaken and several Local
Authorities have since produced SWMPs, but to date all the studies undertaken have adopted a significant
degree of simplification due to either time or technological limitations. A SWMP was commissioned by Bristol
City Council in 2010 with one of its aims being to produce a SWMP to the highest level of accuracy available.
The initial phase of work has now been completed and this paper describes some of the issues that are present
when creating a city sized model. The model created was an integrated urban drainage (IUD) model. The
logistics of running a model of this size are quite impressive with millions of calculations taking place to achieve
a single simulation. The initial results have indicated that „static‟ flood maps can fail to identify the movement of
hazards through the catchment during a flood event. This work has significant implications for all authorities
responsible for contingency planning in major conurbations.
Keywords
Innovation, technology, integrated urban drainage, flood modelling, 2D overland flow
1. INTRODUCTION
Bristol city has a population of over 433,000 incorporating an urbanised area of over 150 square kilometres. It is
currently the sixth largest city in England. The continuing urbanisation within the area has had some negative
results, one of which is more frequent flooding of the city. The city‟s domestic drainage requirements are served
by approximately 50,000 sewers. The system includes the Northern Storm Water Interceptor, a 4-5m tunnel,
which was commissioned in 1962 for flood alleviation purposes.
Following the floods of 2007 and the resulting Pitt review, the Government accepted Pitt‟s recommendation that
Surface Water Management Plans should be undertaken by local authorities. The Pitt Review has also resulted in
the new Flood and Water Management Act 2010, which significantly revises the powers and duties of local
authorities and the Environment Agency. As a consequence of these changes, Bristol City Council in fulfilling
its new role, commissioned a SWMP of the city in 2010. The scope of works for the study specified a very high
level of detail for the overland flow and flood analysis.
Bristol also suffered severe flooding in November 1703 when a storm surge pushed the tide up the Avon,
allegedly covering half of Bristol, in October 1882 when three inches of rainfall fell within 48 hours and in
March 1889 when a rapid thaw of thick snow was followed by continuous rainfall for 36 hours.
Starting from the 1483 flood till the 1999 event the chronology bar in Figure 2 shows the increasing frequency of
flooding in Bristol. The EA estimate that currently 2200 properties are at risk of flooding and that without
intervention this figure will rise to 14000 based on future predictions. Based on the weighted damage
calculations for a 100 year event this represents an increase in potential damages from £59 million to £376
million.
The Department for the Environment and Rural Affairs (DEFRA) identified Bristol as one of the 10 highest risk
areas susceptible to surface water flooding in the UK. Arup were commissioned by Bristol City Council to
develop a surface water management plan which would identify areas for prioritisation of investigation and to
help deliver a comprehensively informed strategic plan. This paper is based on the results from phase 1 of this
work that included building an IUD model for the whole of Bristol and identifying high priority risk areas.
The IUD pilot summary report dated June 2008 stated “An assumption of nearly all the modelling approaches
was that surface water flooding comes from surcharged sewers, overloaded beyond their conveyance capacity.
Direct surface runoff resulting in pluvial flooding as a result of inadequate drainage systems was not
represented because the modelling software used to represent the sewer systems does not readily represent flow
generation and routing above ground.” Most of the pilot studies included some form of modelling and a couple
of the studies had some very comprehensive models. However none of the studies actually utilised a model that
integrated rivers, sewers and overland flow in a fully interactive manner. Some had all the elements but the
interaction was through simplified mechanisms and in some instances only one way flow was allowed. These
limitations mean that only known mechanisms could be represented and even then only in a restricted manner.
Real catchments have numerous interactions that are rarely recorded anywhere, so only being able to represent
some of the interactions must have an impact on the mass balance of the flow across the catchment.
The IUD pilot summary report dated June 2008 also stated “An argument can be made for ignoring the
operation of the underground drainage system altogether by assuming that an extreme event has completely
filled it at an early stage and most runoff remains on the surface. The approach can be modelled by applying
rainfall inputs directly to 2D models of the urban area which formula to generate runoff across the surface.”
This statement assumes that real catchments act in a simplified manner. It infers that systems all fill up during
extreme events. It infers that they all fill up simultaneously and, it infers that when they are full they effectively
stop working. None of these inferences are valid. Real systems can cover significant geographic areas and all
rainfall will have special and temporal variation throughout any event. Some part of the system can be totally
inundated whilst other parts may still be operating below capacity. The timing of flow through a system is one of
the critical factors in determining the capacity of any part of the system. It is feasible that an identical event to
one that causes flooding, which travels across the catchment in a different direction will not cause flooding. Also
just because a system is full, does not mean that it stops working. Although the storage capacity of the system is
used up, the conveyance capacity of the system is virtually unaffected; in fact the conveyance capacity may
actually be increased due to presence of higher hydrostatic heads driving more flow through the system.
I believe that both statements above are the result of trying to produce a report that needed to deal with some
very complicated mechanisms and significant gaps in our knowledge in a very limited timescale. Some of the
conclusions given in the summary reports are based on grossly simplified approaches and have not aided in
advancing our understanding of flooding. Although I still believe the work done was exemplary, we missed
some real opportunities to better understand how urban catchments behave during flooding.
To date the vast majority of the SWMPs undertaken have opted to significantly simplify the catchments under
consideration. The main starting point for most of these studies is that the underground assets do not contribute
any significant storage during extreme events; therefore they are excluded from the study. On the face of it, this
would seem to be a reasonable assumption as it followed the advice given in the output from the IUD pilot
studies. It is also an assumption that allows the level of modelling to be significantly simplified. One such
simplification that has been used in several areas is to simply deduct a percentage of the rainfall as an allowance
for the water draining into the sewers and assume everything else flows across the surface.
The aim was to create an integrated urban drainage (IUD) model incorporating a digital ground model of the
whole catchment to a four metre resolution. Use this model to identify all the surface and subsurface flow routes
within the catchment and simulate a set of extreme events to be determined by the MET office. The results
would give a flood hazard map of the whole of Bristol including depth and velocity that could be used down to
almost single property level. Developing this model was phase 1 of the SWMP and phase 2 would be using this
model to examine potential mitigation measures. The work was completed within the timescale allocated and we
believe that Bristol is pleased with the model produced. A project of this nature will inevitably highlight issues
and some of the more relevant ones are given below.
4.1 Data
Gathering the data for a project such as this can be a daunting prospect and for this study over 150 separate
datasets were collected from numerous bodies including the Environment Agency, Wessex Water and Bristol
City Council themselves. The High Definition Digital Terrain Model (DTM) was generated by manipulating a
LiDAR data set of over 680 million data points together with extruded polygons of all the buildings from OS
MasterMap to be able to achieve the four metre resolution required. The 4m calculation grid created 15 million
surface calculation points that would be analysed simultaneously together with the underground sewer system.
Wessex Water supplied the 33,000 pipe Drainage Area Study (DAS) model that was converted and imported into
the IUD model. As the model utilised a variable timestep down to 1 second to ensure the maximum stability, this
resulted in the possibility of up to almost 1.3 trillion calculations for the simulation of a 24 hr event. The 3
extreme storm scenarios supplied by the MET office were all simulated as 24 hr events.
Managing this amount of data and simply handling the amount of result information required very tightly
controlled data handling procedures and much of the processing had to be done „off the grid‟ to remove some of
the delays caused by normal network traffic.
4.3 Hydrology
The MET office was asked to supply a selection of storm criteria that would coincide with their early warning
procedure for Bristol. These storms were a 100mm 3 hour, a 100mm 6 hour, 200mm 3 and 6 hour. A 1 in 100
year return period was also specified.
As the total catchment being considered covered over 150 square kilometres we needed to resolve the issue of
which duration would be critical. The problem with critical durations is that they change based on what part of
the catchment is under consideration. Different catchments within Bristol would clearly have different critical
storm durations as catchment gradients and urbanisation vary. Rather than running a large set of different storm
durations on an extremely complex model, where one single run would take over a week to complete. We
needed to represent the worst case scenario across the whole catchment simultaneously. We chose to develop
one profile encompassing all the critical durations for the catchment to act as a Superstorm for the catchment.
The original concept of a Superstorm was derived by Microdrainage over 15 years ago for use in analysing the
spill frequencies of Combined Sewer Overflows (CSOs) in sewer networks. At the time this type of analysis
utilised thousands of rainfall events to work out how many times a year a CSO would spill. The Superstorm was
devised to identify the highest rainfall at any single timestep within a series of rain events and then combine
them all into one design storm, hence the name Superstorm. We combined synthetic design storms from a 15
minute to 1440 minute durations for the required return periods, preserving the peak Intensity as well as the total
volume of all the storms to match maximum critical depths. In fact the Superstorm is designed to balance
volume not just across the storm as a whole but also at every sub-duration or time step within the storm as well.
This creates a storm that produces a runoff that is „critical‟ throughout the whole network simultaneously,
reducing the number of simulation runs required by a factor. Superstorms were derived for the following return
periods: 1-in-30 year, 1-in-100 year, and 1-in-200 year.
5. LESSONS LEARNED
5.1 Resolution of model
The client was aware that the original requirement of a 4m grid across the model was a very challenging target.
We would suggest that it is unlikely that a fixed grid of lower resolution than 4m can currently be used to cover
city sized catchments that need to incorporate IUD methods. However at the speed the software and hardware is
improving we do not expect this limitation will be around long. Although we managed to meet this challenge, we
believe that a more selective use of variable grid sizes will produce even higher accuracy results than a fixed
grid. We would recommend reducing grid sizing in areas of interest and increasing them in areas that are not at
risk.
However the Figure 4 below shows that when the underground system is included the flow is transferred down
the catchment, using the underground system as a flow conduit between the two areas.
Although most practitioners have always been aware of this possibility, some have argued that as the
underground system would fill up before major flooding occurred there would be little capacity left to reduce
further flooding, therefore it could be ignored as storage. Whilst this basic statement may be valid, it ignores the
fact that even when the underground system is full, is still has the ability to act as a conduit between adjacent
areas and even when it is full, it will still move flow around.
The result of this transference system was seen in part of the Bristol catchment, where we compared the results
both with and without sewers. The flooded area at the downstream end of the catchment increase and the flooded
area at the upstream end reduce, when the sewers are present within the model. These differences were not
insignificant and represented an order of magnitude at both the upstream and downstream areas.
Whilst the over prediction at the upstream end may have blighted some potential development land, it was
deemed as critical especially as it was in line with the precautionary principle. However the under prediction at
the downstream end of the catchment is serious and could in some instances even be life threatening.
The vast majority of flood mapping undertaken in the UK at this time takes no account of this phenomenon and
the results of this study show that the guidance about ignoring underground system given following the IUD
pilot studies to be misleading and potentially hazardous.
Acknowledgements
The author wishes to acknowledge the Bristol City Council (especially Steven Sodek) for commissioning Arup
to undertake this work, the Environment Agency and Wessex Water for their support in providing data and to
Microdrainage for providing advanced releases of their software and their quick response when changes to the
software were required.
References
[1] Bryant, EA and Haslet, SK, “Was the AD 1607 coastal flooding event in the Severn Estuary and Bristol
Channel (UK) due to a tsunami?”, Archaeology in the Severn Estuary,13, 2002,163 – 167.
[2] Image from Wikimedia Commons and is in the public domain because its copyright has expired