Updated Flood Map for Surface

Water
National Scale Surface Water
Flood Mapping Methodology

Final Report version 1.0 (May 2013)

Title here in 8pt Arial (change text colour to black) i
We are the Environment Agency. We protect and improve the
environment and make it a better place for people and wildlife.
We operate at the place where environmental change has its
greatest impact on people’s lives. We reduce the risks to people
and properties from flooding; make sure there is enough water
for people and wildlife; protect and improve air, land and water
quality and apply the environmental standards within which
industry can operate.
Acting to reduce climate change and helping people and wildlife
adapt to its consequences are at the heart of all that we do.
We cannot do this alone. We work closely with a wide range of
partners including government, business, local authorities, other
agencies, civil society groups and the communities we serve.

Published by:

Environment Agency
Horizon House, Deanery Road
Bristol BS1 5AH
Email: enquiries@environment-
agency.gov.uk
www.environment-agency.gov.uk

© Environment Agency 2011 Further copies of this report are available

from our publications catalogue:
All rights reserved. This document may be http://publications.environment-
reproduced with prior permission of agency.gov.uk or our National Customer
the Environment Agency. Contact Centre: T: 03708 506506
E: [email protected].

ii Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
Acknowledgements
This report has been produced with input from the following individuals:
 Neil Hunter, Rob Berry, Amanda Crossley, Duncan Faulkner, Andrew
Gubbin, Rosalind Rogers and Simon Waller, JBA Consulting

 Ali Cotton, Elliot Gill, Dan Stansfield and Yong Wang, Halcrow

 Matt Horritt, Horritt Consulting

 Alastair Duncan, Jo Diamond, Shirley Greenwood, Karl Jeans, Kate Marks

and Nikki Richardson, Environment Agency

It has also benefited considerably from technical reviews and discussions with the
following individuals:
 Chris Digman, MWH
 Jamie Margetts, Clear Environmental Consultants
 Michael Adams, Adam Baylis, Mike Steel, Mark Whitling and Paul Wyse,
Environment Agency
 David Graham and Sue Humm, Gloucestershire County Council
 Andrew Stone, Rhondda Cynon Taf County Borough Council
 Francis Comyn, Rochdale Metropolitan Borough Council

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology iii
Contents
1 Introduction 1

2 Data preparation 3
2.1 DP1 - LIDAR/IfSAR composite digital terrain model 3
2.2 DP2 - Ordnance Survey MasterMap and other land cover datasets 4
2.3 DP3 - Soil maps 4
2.4 DP4 - FEH depth-duration-frequency model parameters 4
2.5 DP5 - Local drainage information from LLFAs and WaSCs 5

3 Flood modelling 6
3.1 MOD1 - Improve DTM representation of key topographic controls 8
3.2 MOD2 - Develop initial rainfall hydrology 9
3.3 MOD3 - Run JFlow+ to check model set-up and edit DTM 20
3.4 MOD4 - Finalise DTM and spatially-varying model inputs 22
3.5 MOD5 - Finalise rainfall hydrology 23
3.6 MOD6 - Run final JFlow+ models 23

4 Flood mapping 24
4.1 MAP1 - Post-process depth, velocity and hazard rating model outputs 24
4.2 MAP2 - Undertake flood map validation and confidence rating 37
4.3 MAP4 - Deliver national scale surface water flood map 48

5 Comparison of Environment Agency surface water flood mapping

methodologies 49

References 52

List of abbreviations 54

Table 1 Manning’s n roughness values by Ordnance Survey MasterMap Feature Code (reproduced from Capita
Symonds/Scott Wilson (2010, p23)) 22
Table 2 Quadrant analysis formula 27
Table 3 Hazard to People classific ation using FD2320/FD2321 hazard rating formula (Depth x (Velocity + 0.5) + Debris
Factor, where debris factor = 0.5 if depth <= 0.25m and debris factor = 1 if depth > 0.25m) 30
Table 4 Combinat ion of model routing and effectiv e rainfall uncertainties into final model output uncertainty - example
values 38
Table 5 Validation evidence for the pioneer locations 41
Table 6 Results of property validation for flooded properties recorded in 2007 summer floods in suburb of Cheltenham
covered by InfoWorks ICM model 46
Table 7 Results of property validation for internal flood incidents recorded in RCT, 2002-2011 47
Table 8 Comparison of Environment Agency surface water flood mapping methodologies 49

Figure 1 Stages in producing the Updated Flood Map for Surface Water 2
Figure 2 Total rainfall depths (in mm) evaluated using FEH DDF model parameters at the centroids of 5km x 5km
modelling grids (shown in red, superimposed on the greyscale DTM) 5
Figure 3 Monitor point w ater level and velocity time series for Test 8A 7
Figure 4 Map of critical storm durations in Birmingham (reproduced from Faulkner, 2010)) 10
Figure 5 Urban-rural land cover mask used to determine net rainfall-runoff 13
Figure 6 Urban runoff processing to generate net rain (1 in 30 probability, 1 hour duration rainfall event) 14

iv Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
Figure 7 Sketch of surface runoff and throughflow processes 17
Figure 8 Example application of different net rainfall hyetographs according to urban-rural classific ation for two adjacent
modelling tiles 19
Figure 9 Example differences betw een total and net rainfall 20
Figure 10 A DTM before and after manual editing 20
Figure 11 Areas (shown in red) which could need a DTM edit and require further investigation 21
Figure 12 Flood mapping data outputs 24
Figure 13 Overview of flood mapping process 25
Figure 14 Overview of Step 1 - Blending grids together 26
Figure 15 Quadrants for flow direction analysis 27
Figure 16 Areas filled in (blue) 31
Figure 17 Areas removed (red) 31
Figure 18 Areas on diagonal with an area > 96m2 31
Figure 19 Overview of Step 2 - Creation of cleaned outline 32
Figure 20 Flow direction at 1:500 on a 2m grid 34
Figure 21 Flow direction at 1:1,000 on a 5m grid 34
Figure 22 Flow direction at 1:3,000 on a 10m grid 34
Figure 23 Overview of Step 3 - Clipping of grids to the outline 35
Figure 24 Overview of Step 4 - Creation of outputs 37
Figure 25 Data used to derive the confidence star rating and final map output 40
Figure 26 Comparison between known flooding locations (surface water in purple polygons, unknown sources in green)
and national scale map outputs (1 in 30 probability) for two locations in Greater Manchester 42
Figure 27 Comparison between AStSWF (top, 1 in 200 probability), FMfSW (middle, 1 in 30 probability) and latest
national scale map (bottom, 1 in 30 probability) for Manchester city centre 43
Figure 28 Comparison between latest national scale map (top, 1 in 100 probability) and InfoWorks ICM (bottom, 1 in
100 probability). Properties flooded in 2007 are shown in red. 44
Figure 29 Velocities predicted by the InfoWorks ICM (top) and national scale mapping models (bottom) in Cheltenham,
displayed in two classes: blue <1 m/s and red >1 m/s 45
Figure 30 National scale map for 1 in 30 probability, w ith reported flood incidents as red points, for two locations in RCT 47

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology v
1 Introduction
In August 2012, the Environment Agency appointed a contractor team of JBA
Consulting, Halcrow and Horritt Consulting to deliver new national scale surface water
flood mapping for all England and Wales by February 2013. This mapping is intended
to form the basis of a new product, the Updated Flood Map for Surface Water, that will
show the worst case flood extents, depths, velocities (both magnitude and direction of
flow) and hazard rating associated with the 1 in 30, 1 in 100 and 1 in 1,000
probabilities, in order to:
 help Lead Local Flood Authorities (LLFAs) whose areas cover part or all of a
Flood Risk Area to meet their obligations for Flood Hazard and Risk Mapping
under the Flood Risk Regulations 2009;
 provide all LLFAs with better quality datasets to support local flood risk
management and spatial and emergency planning activities;
 provide a consistent view of surface water flood risk across England and
Wales to support the Environment Agency's Strategic Overview/Oversight
role, for example, by facilitating better decision making and more effective
targeting of the available resources through the Flood Defence Grant-in-Aid
(FDGiA) and Long Term Investment Strategy (LTIS) processes;
 produce mapping that is useful, locally-credible and accessible to the public
to help them understand the hazards of flooding from surface water;
 provide models and parameterisation datasets that the Environment Agency
and LLFAs will want to re-use and develop further for local studies.

This report sets out the final methodology used to produce the new national scale
surface water flood mapping (the "how") and should be read in conjunction with the
Guidance on surface water flood mapping for Lead Local Flood Authorities
(Environment Agency, 2012a), which sets out the political context and drivers for the
project (the "why"). The methodology described herein has been developed and
refined in consultation with Environment Agency technical specialists, external peer
reviewers (Dr Chris Digman (MWH) and Jamie Margetts (Clear Environmental
Consultants)), the Project Board, three "pioneer" LLFAs (Greater Manchester,
Gloucestershire and Rhondda Cynon Taf) and delegates who attended the LLFA
Capacity Building Workshops hosted by Defra in November 2012.
This report is structured according to the three main stages of producing the new
national scale surface water flood map shown in Figure 1:
1. Data preparation (Section 2)

2. Flood modelling (Section 3)

3. Flood mapping (Section 4)

The "LLFA review" component of Stage 4 is described in Reviewing the national scale
surface water flood maps (Environment Agency, 2013a) while the approach to "map
updating" is currently a work in progress. Finally, Section 5 provides a summary of the
methodological differences between the new flood maps and the previous Areas
Susceptible to Surface Water Flooding (AStSWF) mapping and the Flood Map for
Surface Water (FMfSW).

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 1
 Figure 1 Stages in producing the Updated Flood Map for Surface Water

2 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
2 Data preparation
The following datasets are of key importance for producing high quality surface water
mapping products. These data were acquired, stored and processed by the contractor
project team in accordance with the agreed data management plan.

2.1 DP1 - LIDAR/IfSAR composite digital terrain model

The modelling and mapping was undertaken on an updated version of the Environment
Agency’s LIDAR/NEXTMap composite digital terrain model (DTM). This DTM provides
a continuous description of "bare earth" topography across England and Wales at a
horizontal grid resolution of 2m. It includes all the LIDAR data from the recent “urban
areas greater than 3km²" campaign but does not include any LIDAR data held by
LLFAs or Water & Sewerage Companies (WaSCs). Vertical accuracies for the LIDAR
and NEXTMap DTMs are quoted as ±0.15m and ±1m respectively, although the
accuracy of Environment Agency LIDAR data has been closer to ±5cm since 2005 (A
Duncan, pers. comms.). However, it should be remembered that these data and
accuracy statistics will only reflect a "snapshot" of the landscape at the time of data
capture and where the landscape has subsequently changed (e.g. due to urban
development, natural change, landfill, coastal realignment), then the potential for
anomalies is much greater. A GIS file is provided with the final mapping data that
shows which areas are covered by LIDAR data.
The first stage in producing the composite DTM was to resample the underlying terrain
data – LIDAR data of 2m, 1m, 0.5m or 0.25m resolutions and NEXTMap data of 5m
resolution – to a common 2m resolution. This resampling was done using the bilinear
method within ArcGIS Spatial Analyst. The resampled data was then mosaiced
together into a single DTM, with the LIDAR data taking precedence in areas of
common coverage. In order to remove any potential steps at the boundaries of the
LIDAR and NEXTMap data, the mosaicing process incorporated a "blending" step
which utilised distance-weighted averaging to provide a smooth transition at the DTM
edges. Blending is applied across an "overlap distance" of 100m.
The same technique was used to smooth the edges of overlapping LIDAR surveys,
albeit with a much shorter overlap distance: a 30m overlap was used for 2m data, a
25m overlap for 1m data, a 20m overlap for 0.5m data and a 15m overlap for 0.25m
data. These values were selected after testing various overlap sizes over many
different types of terrain and have been found to provide the best fit overall. They have
been used for all LIDAR composite products since 2007 without any significant
problems (A Duncan, pers. comms.).
It should be noted, however, that blending does not resolve discrepancies in the
underlying data – for example, large woodland tracts which have not been removed
from the NEXTMap DTM. In these situations, a 100 metre wide “ramp” was produced
from the more realistic elevations of the LIDAR DTM up to the false heights (top of
woodland canopy) recorded in the NEXTMap DTM. These errors will not be overcome
without re-processing all the NEXTMap data with new surface feature removal
algorithms or flying the whole of England and Wales with LIDAR.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 3
2.2 DP2 - Ordnance Survey MasterMap and other land
cover datasets
Ordnance Survey (OS) MasterMap Topography data (11 June 2011 release) were
used to produce maps of building footprints, road layouts and manmade areas using
the following attribute queries:
 Building footprints:

("FEATCODE" = 10021) AND ("DESTERM" <> 'Archway')

 Road layouts:

("FEATCODE" = 10172) AND ( ("MAKE" = 'Manmade') OR ("MAKE" =

'Multiple') OR (("MAKE" = 'Natural') AND ("DESTERM" = 'Track')) )

 Manmade areas:

"MAKE" <> 'Natural'

These maps were used as the basis for positively reinforcing important topographic
controls on flow within the DTM (see Section 3.1) and defining the spatial variation in
runoff and infiltration rates and hydraulic roughness (see Sections 3.2 and 3.4
respectively).

2.3 DP3 - Soil maps

The Environment Agency provided the HOST (Hydrology Of Soil Types) mapping
(Boorman et al., 1995), which includes the BFIHOST (Baseflow Index) and SPRHOST
(Standard Percentage Runoff) parameters on a 1km grid, under licence from the
National Soil Resources Institute (see http://www.landis.org.uk/data/sshydrology.cfm
for further details).
Allowing for discretisation errors, these data are as per the Flood Estimation Handbook
(FEH) CD-ROM and were used in the calculation of runoff in rural areas (see Section
3.2.3).

2.4 DP4 - FEH depth-duration-frequency model parameters

England and Wales was subdivided into approximately 7,100 5km x 5km "tiles" that
provided the basis for rainfall estimation and subsequent hydraulic modelling. For each
tile, a model of rainfall depth-duration-frequency (DDF) was constructed using
parameters available from the FEH CD-ROM at the tile centroid and the techniques
outlined in Volume 2 of the FEH (Faulkner, 1999). Each DDF curve was then used to
calculate a tile-specific total rainfall depth for a storm of given duration and probability
(see Figure 2).
The DDF parameters (C(1km), D1(1km), D2(1km), D3(1km), E(1km), F(1km)) were
extracted automatically from a specially-licensed copy of the FEH CD-ROM version 3.
A further FEH catchment descriptor, PROPWET, was also extracted for use in the rural
runoff calculations.

4 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
 Figure 2 Total rainfall depths (in mm) evaluated using FEH DDF model
parameters at the centroids of 5km x 5km modelling grids (shown in red,
superimposed on the greyscale DTM)

It is acknowledged that this approach ignores the spatial variation in rainfall that is
known to occur across areas smaller than a 5km x 5km tile. As such, it may over
predict the amount of rainfall across large areas, and therefore give a very much worst
case estimate of runoff, which for some large sub-catchments may over play the flood
risk being predicted. The choice of a 5km x 5km tile therefore represents a
compromise between the computational limits of the hydraulic modelling software and
the need to describe broadscale variations in rainfall patterns without oversampling the
underlying DDF methods and data.

2.5 DP5 - Local drainage information from LLFAs and

WaSCs
Approximately 10% of LLFAs provided information on drainage rates, critical storm
durations and/or infiltration/runoff rates in response to a data request issued by the
Environment Agency in Summer 2012 (Environment Agency, 2012b). Following basic
sense checks, these data were processed into suitable formats for incorporating within
the rainfall hydrology calculations and hydraulic modelling. Where data could not be
used "as supplied", inconsistencies were resolved via one-to-one discussions.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 5
3 Flood modelling
The flood modelling component of the project was undertaken on a 2m resolution grid
for all England and Wales. Although almost certainly an overspecification in many
areas (e.g. unpopulated rural locations), a "2m everywhere" approach was felt by the
Project Board to be the most efficient and expedient way to offset technical and political
concerns over which areas should be modelled at higher resolution. It is, however,
acknowledged that the use of sub-2m resolution grids/meshes is now relatively
common in 2D urban drainage modelling (J Margetts, pers. comms.). Here finer
resolution modelling was not considered compatible with either the available DTM data
(which would not support further disaggregation) or the project programme/resource
constraints.
Simulations were performed using the JFlow+ 2D hydraulic model (Lamb et al., 2009;
Crossley et al., 2010a, 2010b). JFlow+ solves the Shallow Water Equations using a
finite volume formulation that combines the Riemann based solver of Roe with an
upwind treatment of the source terms. The model is both conservative and shock
capturing, and maintains water at rest over irregular topography. JFlow+ is
implemented on a regular grid using the supplied DTM and does not require any
secondary grid generation process. This simplifies the model set up and allows for
direct interpretation of the model results relative to the DTM.
JFlow+ has been designed with the emphasis on easy set up and model specification.
Models are configured using databases, and this provides a highly ordered means to
store significant quantities of data. The modelling engine is controlled through a web
interface, ensuring that projects and models can be set up, run and monitored from any
location in which an internet connection is available.
JFlow+ has been benchmarked using the test cases proposed by the Environment
Agency in the Science Report SC080035/SR2, Benchmarking of 2D Hydraulic
Modelling Packages, and the results have been submitted to the Environment Agency.
Results for Test 8A which considers rainfall and point source surface flow in an urban
area are shown in Figure 3 (note that the axes used are consistent with those in the
SC080035/SR2 report). These plots demonstrate the ability of JFlow+ to deliver robust
velocity data for direct rainfall applications.
JFlow+ is commercially available now for use by the Environment Agency or any LLFA,
or by any other party on behalf of the Environment Agency or any LLFA. The latest
version of the software includes all the modifications required to undertake the
modelling work described here.

6 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
Water level (depth) time series Velocity time series
Test 8A, point 7 Test 8A, point 7

Test 8A, point 9 Test 8A, point 9

Figure 3 Monitor point water level and velocity time series for Test 8A

The flood modelling approach was based on the "direct rainfall" concept where net or
"effective" runoff volumes applied to each grid cell in the hydraulic model are routed
across the DTM surface, identifying flooding pathways and areas where ponding will
occur. The approach has been successfully implemented by JBA to produce both the
previous Environment Agency surface water flood maps and is widely accepted as an
appropriate method for analysing higher magnitude, lower probability storms where
subsurface drainage systems are likely to be overwhelmed and/or inlet capacities
exceeded (Defra, 2010).
As for the Areas Susceptible to Surface Water Flooding (AStSWF) mapping and the
Flood Map for Surface Water (FMfSW), England and Wales was subdivided into
approximately 7,100 5km x 5km tiles (or model domains). These modelling tiles
included a 500m buffer with adjacent tiles (so are actually 6km x 6km) to ensure that
modelled areas overlap sufficiently and "edge effects" are not visible in the final maps.
For expediency, tile boundaries were aligned to the British National Grid rather than
surface water catchments, and this may be significant for underestimating runoff
response from missing (i.e. "off-tile") upland areas or inputs to the natural and
manmade drainage network that prevent their discharge further downstream,
particularly during longer storm events. However, with no national map of surface
water catchments available (or reliable means of deriving one automatically) and the
use of relatively short storm durations (see Section 3.2.1), there was little option but to
acknowledge and accept this compromise. As noted in Section 2.4 above, the choice
of a 5km x 5km tile also represented a trade-off between the computational limits of the
hydraulic modelling software and the need to describe broadscale variations in rainfall
patterns without oversampling the underlying DDF methods and data.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 7
3.1 MOD1 - Improve DTM representation of key
topographic controls

3.1.1 Buildings
As directed by the Project Statement of Requirements (SoR), the OS MasterMap data
processed during step DP2 was used to explicitly raise the ground level within building
footprints (according to the bare earth DTM) by approximately 0.3m. An upstand
height of 0.3m was selected because flooding at this depth will certainly exceed the
level of any damp-proof course and result in property flooding in many cases.
This requirement was based on the conclusions of the FMfSW Improvements Pilot
Studies (Halcrow/JBA Consulting, 2012), which demonstrated the importance of
modelling the "deflection effect" of buildings on surface water flows. However, contrary
to the recommendations of Allitt (2009, pg12), buildings are also represented in such a
way that there can be flow through them once the depth exceeds the height of the
upstand. This is consistent with evidence from steeper areas where water is observed
to pass through, rather than around, buildings during flood events (N Rookes, A Stone,
pers. comms.). It was therefore felt that the upstand approach provided the best
compromise currently available for building representation in national scale surface
water modelling.
To ensure that buildings deflect flow but do not cut into the DTM, even on the steepest
slopes, the following logic was applied:

—
upstand building + 2ff + 0.3), ax(Z bu[kfi?lff ) + 0.3

Where:

Accordingly, each building footprint was assigned a horizontal "floor level" that will
ensure more consistent results within individual properties and also minimise the
occurrence of partially flooded buildings in the final mapping that can complicate
property counting approaches.

3.1.2 Roads
Better definition of kerb features is standard practice in high resolution urban flood
modelling (Hankin et al., 2008; Allitt, 2009; WaPUG, 2009; Defra, 2010; Halcrow,
2011). The representation of the road network, which is known to preferentially collect
and route storm water when it rains, was therefore improved within the DTM. Road
surfaces, selected from OS MasterMap data using the criteria shown in Section 2.2,
were lowered by 0.125m (the height of a British Standard kerb) to better delineate
these important pathways in the hydraulic modelling and mapping. However, this
approach may overestimate the routing effect of roads in rural areas where there are
fewer kerb stones or where the kerb height is substantially less because the road has
been resurfaced.

8 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
Both the topographic modifications described above were quick and easy to apply and
ensured that the principal flood pathways between buildings and along roads were
better represented in the regular 2m model grids. However, detailed urban drainage
modelling often shows that it is subtle changes in local topography that can significantly
affect the ultimate direction and extent of the flooding, particularly during higher
probability events where depths may be low. As such, the inability to represent other
important urban features explicitly within the DTM, such as walls, fences, drop kerbs
and speed bumps, should be recognised as a limitation (albeit understandable for a
national scale approach).

3.2 MOD2 - Develop initial rainfall hydrology

As part of this step, two pieces of work relating to the rainfall hydrology were
undertaken:
1. Conservative rainfall estimates were produced to allow DTM editing (see
Section 3.3) to be started as soon as possible. Total rainfall depths were
calculated for a 3 hour, 1 in 1,000 probability event on a tile-by-tile basis and
then scaled across a 50% summer profile assuming 100% runoff to produce
a hyetograph that will be applied uniformly across each hydraulic model
domain.

2. New methods were investigated for better representing the spatial variation
in rainfall runoff rates resulting from different urban densities, as well as the
urban/rural split. While a new approach was developed for use in rural areas
(described in Section 3.2.3), it was not possible to improve on the urban loss
model used in the previous FMfSW given the time and data available. This
method is therefore described again for completeness in Section 3.2.2.

3.2.1 Specification of design events

Rainfall probabilities
As directed by the Project SoR in order to meet the requirements of the Flood Risk
Regulations 2009, the modelling and mapping work was undertaken for the 1 in 30, 1 in
100 and 1 in 1,000 rainfall probabilities. Unlike previous Environment Agency surface
water flood maps, mapping was not produced for the 1 in 200 probability.

Rainfall storm durations

Originally the Project SoR specified that a single rainfall storm duration of 3 hours
should be applied everywhere, except where LLFAs requested that alternative values
should be used. However, choosing a single, representative critical storm duration is
difficult, because any modelled area will include a number of sub-catchments of
different size, steepness and shape. Our physical understanding indicates that a short
duration storm (around 1 hour) better represents the type of event that leads to surface
water flooding, but there is also evidence that longer storms may be critical in flatter
areas (e.g. Halcrow, 2007; Faulkner, 2010).

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 9
Figure 4 shows how the distribution of critical durations is strongly linked to the
topography. In low-lying areas, near to rivers, the critical duration is long because
surface runoff drains into these areas from larger catchments which have a longer time
of concentration. There are also isolated areas elsewhere with a long critical duration,
which tend to be topographic depressions where water will pond. On hill slopes the
critical duration is generally short because the greatest flood depth arises from high
intensity rainfall.
It is also recognised that the critical duration for flood extent and depth may not be the
same as for velocities (which will increase for shorter, more intense storms).

Figure 4 Map of critical storm durations in Birmingham (reproduced from

Faulkner, 2010))

10 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
Therefore, in response to the technical concerns of the peer reviewers as well as
numerous requests for additional storm durations from LLFAs, the Project Board took
the decision to model storm durations of 1, 3 and 6 hours for each rainfall probability
(i.e. 9 rainfall scenarios in total). Use of multiple storm durations should also mean that
the final mapping is more representative of a 1 in x chance flood, rather than only
representing a 1 in x chance rainfall event.
An alternative to the conventional design storm is the "superstorm" concept (Osborne,
2012) which is a long-duration storm that includes within it the critical conditions for
shorter duration events. It is intended to avoid the need to run large numbers of
different storm durations. It has been applied on some Surface Water Management
Plans (SWMPs) and Integrated Urban Drainage (IUD) studies but the long duration of
the superstorm proved prohibitive in terms of model run times here.

Storm profile
Total rainfall depths for each rainfall probability/storm duration were then scaled across
a standardised storm profile to produce design hyetographs. Two standard profiles are
typically recommended: a 75% winter profile, for rural catchments, and a 50% summer
profile, for urban catchments. The summer profile is more peaked than the winter
profile, because of the prevalence of intense convective storms during the summer, so
the intensity is greater in the middle of the storm. Intuitively, therefore, the summer
event is more likely to be critical for surface water flooding and it has been selected for
previous surface water mapping studies in England and Wales, Scotland and Ireland.

Areal reduction factor

An areal reduction factor (ARF) was not applied to the rainfall because, unlike for fluvial
studies, there is no defined catchment into which the rainfall is falling. The direct
rainfall approach to surface water mapping can be viewed as merging together many
separate point rainfall events during a single model run. Admittedly, in reality the depth
of surface water flooding at any location will depend on rainfall over an area which is
draining towards that location and local downstream conditions. The size of this
contributing area will vary from point to point, but in general it is likely to be small,
because when there is a large area draining into a point, the point generally coincides
with a river, in which case the flooding is fluvial rather than pluvial. For a rainfall
duration of 3 hours, the ARF is above 0.97 for an area of 2km 2 and so it would seem
reasonable to ignore the areal reduction effect for surface water mapping. The surface
water mapping methodology divides an area into 5km x 5km tiles for the purpose of
running JFlow+. This division is not relevant to the setting of ARF values because it is
not likely that any points will receive runoff from the entire 25km 2 square.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 11
3.2.2 Urban runoff calculation

Introduction: how does surface water flooding occur in urban areas?

In urban areas, serviced by underground sewerage and highways drainage
infrastructure, surface water flooding occurs when either:
 Runoff cannot reach inlets (e.g. road gullies and downspouts) to the drainage
system; or

 Drainage systems or their inlets are at capacity and no further flow can enter;
or

 Drainage systems are at capacity and flood through gullies, manholes and
sometimes within properties through domestic plumbing connections.

In addition, urban areas can be crossed by watercourses, often in culvert, which also
flood when capacity is exceeded. These watercourses receive flows from outside of
the urban area but also from within it via surface water drainage outfalls or combined
sewer overflows or from pluvial flows entering directly along the bank.
The combined effect is a highly complex flooding environment where even close
observation of flooding events is uncertain to identify the full mechanism. Complex
integrated urban drainage models are used to represent the interaction of these
different processes and represent the state of the art in predicting urban surface water
flooding. Unfortunately it is not feasible to model flooding at a nationwide scale using
these methods because simulations take a long time and models are demanding of
detailed asset data. Therefore, the Updated Flood Map for Surface Water uses a
simplified approach which can approximate these processes over large geographical
areas.

Description and justification of the chosen method

The chosen method generates a net rainfall for urban areas which allows for an urban
runoff coefficient (70%) and a default drainage capacity (12 mm/hr). Net rainfall is then
assumed to flow and pond freely across the urban area to predict flooding according to
topographic controls.
Urban areas were defined using an analysis which considered the land use (based on
OS MasterMap information) for grid squares 250m by 250m. Each square was
screened to determine the percentage coverage of "manmade" landuse; this included
all buildings, roads, paths and other hard-standing. Where the manmade landuse was
greater than 50% of the square, the whole square was determined as "urban" and
urban runoff rules were applied. All other squares were defined as "rural" and the
methods defined in Section 3.2.3 were applied. This method proved successful in
delineating the edge of the built-up area whilst excluding large areas of green space
(e.g. parks) within cities. Figure 5 is an extract of the results of this analysis for part of
the town of Skipton in North Yorkshire. Urban 250m grid squares are coloured red.
Rural runoff calculation methods were applied in green squares.

12 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
 Figure 5 Urban-rural land cover mask used to determine net rainfall-runoff

A runoff coefficient of 70% was chosen for urban areas. This is a good average runoff
coefficient for built-up areas including gardens and green verges and a mix of city
centre and more suburban land uses. The FEH catchment descriptor assumes a 70%
coefficient for urban areas. It is the value for urban runoff used in previous national
surface water flood maps in England and Scotland. Viessman and Lewis (2003), a
standard hydrology text, quote city centre runoff coefficients in the range 70-95% and
suburban runoff coefficients in the range 50-70%. This is also in line with average
runoff coefficients in calibrated and verified sewerage hydraulic models.
Making an allowance for drainage systems is more challenging because urban
drainage systems vary so much in nature and their effectiveness in different event
durations is linked to very local characteristics such as the arrangement and capacity of
road gullies and whether drainage is via combined or separate sewerage. This
information is not available in a consistent manner over large areas. Therefore, an
approach used in the earlier Flood Map for Surface Water was re-used. The method
combined estimates of service level (or standard of protection from flooding) for
drainage systems (between 1 in 5 and 1 in 30 years, centred around 1 in 10 years) with
estimates of critical storm duration (0.5 to 2 hours) with estimates of percentage
impermeable area (30% to 80%) with estimates of DDF rainfall parameters to
determine typical drainage system capacity expressed in mm/hr using a Monte Carlo
method. This is the rate of rainfall in urban areas which is typically conveyed in
drainage systems. The range was between 5mm/hr and 54mm/hr with a mode of
12mm/hr. The mode value was adopted as a typical drainage removal rate. In areas of
known low or high drainage capacity, alternative values of 6mm/hr or 20mm/hr were
substituted.

Assumptions and limitations

Whilst very robust and appropriate for the wide-area assessment of surface water flood
risk, the selected method contains a number of limitations which should be noted when
interpreting flood maps.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 13
The selected method assumes that runoff generation is even across urban areas and
even though large green spaces are accounted for, the method does not differentiate
between the higher runoff rates likely in city centres (e.g. >80%) and the lower rates
found in more suburban areas (e.g. <60%). Therefore, the approach risks
underestimating city centre runoff and over estimating suburban runoff.
It also assumes that drainage is principally provided by underground sewers and that
their capacity controls the onset of surface water flooding. This is recognised as a
simplification because the role of inlet control structures is likely to be important for
short duration, high intensity rain events. It is probable that surface water flooding
sometimes occurs before underground drainage systems reach their capacity.
Therefore, the chosen method probably underestimates net rainfall (causing flooding)
for at least the 60 minute (short duration) events.
The prediction of net rainfall (and hence flooding) for the 1 in 30 rainfall probability is
especially sensitive to the assumption for drainage capacity. This is because the
drainage system controls most of the flow volume for this size of event and small
changes in actual capacity would result in significantly greater or less flooding.
Prediction of flooding for the 1 in 100 and 1 in 1,000 probability events are less
sensitive to assumptions about drainage capacity because the drainage system plays a
less significant role in conveying storm water. However, and especially for the 1 in 100
year event, the drainage system can still play an important role in determining the exact
location and timing of flooding. In cities where large underground tunnels convey storm
water in directions counter to the natural topography the simplified approach described
here is unlikely to accurately predict the location of flooding.

Implementation
Figure 6 illustrates the changes made to rainfall data to generate a net rainfall which is
applied in urban areas to predict surface water flooding. First a factor of 0.7 is applied
to account for losses to infiltration (effective rain). Second, 12 mm/hr is removed to
account for losses to the drainage system (net rain). In the illustration below, the
scenario is the 1 in 30 probability, 1 hour duration rainfall event which has a total gross
rainfall depth of 31.79 mm, a total effective rainfall depth of 22.25 mm and a total net
rainfall depth of 11.81 mm.

Duration (minutes)
0 10 20 30 40 50 60 70
0

20

40
rain rate (mm/hr)

60 Gross rain (mm/hr)

80
Effective rain (mm/hr)

Net rain (mm/hr)

100

120

140

Figure 6 Urban runoff processing to generate net rain (1 in 30 probability, 1 hour

duration rainfall event)

14 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
3.2.3 Rural runoff calculation

Introduction: how does surface water flooding occur in rural areas?

In marked contrast to fluvial flood estimation, there has been little research into
methods for estimating design runoff rates for surface water flood modelling. There
have been numerous investigations of runoff processes during flood conditions, but
many have been focused on the eventual role of runoff in producing stream flow, rather
than the way in which surface water flooding is generated. Even the meaning of runoff
can be hard to pin down: it is often used to refer to movement of water through the soil
and subsoil (throughflow) in addition to overland flow.
A simple conceptual model of surface water flooding would envisage it as resulting
from overland flow, due either to infiltration excess (rainfall at a rate higher than can
infiltrate) or saturation excess (rainfall on saturated ground). However, flooding that
occurs where throughflow emerges at the ground surface (e.g. due to topographic
factors or soil heterogeneity) could also be classed as surface water flooding.
Approaches to the estimation of runoff vary from physically-based models (often
calibrated from measurements of soil properties at individual points or in laboratory
studies) to conceptual or statistical methods in which runoff is taken as one of the
components of the process of converting rainfall to river flow. There are pros and cons
to all types of approaches.
We have reviewed recent research literature and guidance documents relevant to
surface water flood modelling (e.g. WaPUG, 2009; Burton et al., 2010; Defra, 2010;
Hurford et al., 2012; Osborne, 2012) and project reports including the runoff calculation
methods used for pluvial flood maps in Scotland (2011) 1 and Ireland (2010) 2. Methods
used for calculating rural runoff include:
 a blanket percentage runoff irrespective of soil type;
 the FSR/FEH rainfall-runoff method, applied in a variety of ways;
 Horton's equation for infiltration;
 a combination of initial loss (depression storage) and infiltration.

We believe that the approach outlined below is a significant step forward from methods
that have previously been applied for national-scale surface water mapping. However,
it does have some limitations and make some assumptions, as discussed in the
following sections.

Description and justification of the chosen method

Our approach for calculating losses, and hence runoff, in rural areas used the losses
model from the Revitalised Flood Hydrograph (ReFH) rainfall-runoff method (Kjeldsen
et al., 2005). An identical approach was applied for calculation of runoff within the
green portions of urban areas.

1 HALCROW, 2011. Deriv ation of a National Pluv ial Flood Hazard Database. Report to SEPA.
2 HR WALLINGFORD, 2010. National Pluv ial Screening Project f or Ireland. Report EX633 5 f or Office of Public Works.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 15
This is an improvement on the percentage runoff model used in the FSR/FEH rainfall-
runoff method. Losses are controlled by a single parameter, the maximum soil
moisture storage capacity (Cmax), which can be estimated from two FEH catchment
descriptors, BFIHOST and PROPWET. It is also necessary to specify an initial
condition, the starting value of soil moisture, Cini, which can also be estimated (when
simulating a design flood event) from catchment descriptors.
The ReFH losses model calculates the volume of runoff at each time step as a function
of the current soil moisture content, so that the percentage runoff increases as the
rainfall continues. This increase is fairly minor for short duration storm events.
Although ReFH was designed for fluvial rather than surface water studies, the same is
true of other methods that have been previously applied for national-scale surface
water mapping. For example, the Environment Agency's current flood map for surface
water uses a runoff percentage for rural areas (39%) derived from consideration of
standard percentage runoff values used in the FSR/FEH rainfall-runoff method for
estimating fluvial design flows (JBA Consulting, 2009).
Methods such as ReFH (or FSR/FEH rainfall-runoff) have an advantage over more
physically-based techniques of calculating runoff (or infiltration) in that they are
designed and calibrated from well-established long datasets with national coverage
(rainfall, river flow and potential evaporation). No equivalent datasets are available for
parameters such as infiltration or surface runoff which are generally measured over
small scales and short durations at experimental sites or in the laboratory.
ReFH was developed not only to model runoff from rainfall but specifically to do so in
the context of estimating design floods. When used in conjunction with a specified
rainfall depth, duration and profile and initial soil moisture, the ReFH model has been
calibrated to produce a design river flow with a specified return period.
An alternative to ReFH, as envisaged from the outputs of the FMfSW Improvements
Pilot Studies (Halcrow/JBA Consulting, 2012), would be to calculate the volume of
runoff from the standard percentage runoff estimated from HOST data, SPRHOST.
We suggest that the approach outlined above is preferable, for the following reasons:
 Calculation of runoff from SPRHOST comes from the FSR/FEH rainfall-runoff
method which has now been superseded by the ReFH method.
 SPR is considerably more difficult to calculate than BFI, and thus the
relationship between SPR and HOST data is less certain than that between
BFI and HOST. For this reason, recent research by CEH has actively sought
to avoid using SPRHOST as an explanatory variable in regression models
(Kjeldsen et al., 2008).
 ReFH allows for the proportion of runoff to increase through the storm, as the
soil wets up.

Assumptions and limitations

The application of FEH/ReFH methods for modelling surface runoff is consistent with a
recent recommendation from Defra/Environment Agency R&D project SC090031 to
use such methods for estimation of greenfield runoff for development control (Faulkner
et al., 2012). Essentially this extension of such methods relies on the assumption that
the relationship between rainfall and river flow continues to follow a similar form as the
catchment becomes smaller and smaller until it no longer contains a watercourse, at
which point the flow is occurring over or close to the surface of the ground. In the
present case, only one part of the ReFH method is being used: the losses model,
which determines the volume of flow. At first sight it may seem a reasonable

16 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
assumption that the relationship between the volume of rainfall and the volume of
surface runoff is similar to the relationship between the volume of rainfall and the
volume of river flow, given that nearly all pluvial flood water can be expected to
continue down the catchment and enters a river to become fluvial flood water.
However, the assumption does have some potential weaknesses:
 There may be additional losses occurring during or after surface water
flooding and before the runoff reaches the river (see Figure 7);

 Conversely, there may be additional sources of quick flow for fluvial floods,
such as interflow, which are not associated with runoff (see Figure 7);

 The definition of runoff used in the ReFH model (and other conceptual
models) depends on the way the fluvial flood hydrograph is separated into
baseflow and quick flow, and does not have any direct physical
interpretation.
The assumption is difficult to test due to the lack of extensive long-term records of
surface runoff.

Rain

Runoff
SW
Continued flooding
runoff Losses

River
More
Interflow losses?

Baseflow

Figure 7 Sketch of surface runoff and throughflow processes

The way in which we implemented the ReFH losses model was to pre-calculate net
rural rainfall before applying it within the hydraulic model. A drawback of this approach
was that there was no allowance for ongoing infiltration in places where water ponds
for long periods, for example, in topographic depressions. Currently JFlow+ does not
represent mechanisms such as infiltration or evaporation and thus there is the potential
to over-estimate the duration and depth of flooding in some locations, particularly
during prolonged events.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 17
The ReFH losses model has a known limitation on highly permeable soils
(BFIHOST>0.65)3. The regression model tends to underestimate the maximum soil
moisture storage capacity, Cmax, in such areas4, which (in isolation) would result in
overestimation of percentage runoff. The value of the initial soil moisture, Cini, may
also be unrealistic 5. Difficulties representing the complex flood-generation processes
on highly permeable catchments are to be expected for design event methods of flood
estimation, and the validity of the FSR/FEH rainfall-runoff method is also limited in such
cases6. In areas where BFIHOST exceeds 0.65 (principally chalk and limestone
geology), it is hoped that LLFAs will review the results of the method and suggest
refinements where necessary, for example setting a minimum percentage runoff.

Implementation
To apply the ReFH losses model for simulation of design floods, it is necessary to
specify values for the rainfall depth (see Section 2.4), duration and temporal profile and
the initial soil moisture, Cini.
The ReFH model can be run with either a summer or winter design event. The
summer event has a greater depth of rainfall but a lower value of Cini. The rainfall
profile is also more sharp-peaked, so the intensity is greater in the middle of the storm.
Intuitively, the summer event is more likely to be critical for surface water flooding but
the summer design event in ReFH (in particular, for the present application, the
regression equation relating Cini to catchment descriptors) was developed using only
seven catchments and so its validity is uncertain. Further testing was not possible
within the project timescales and so the recommendation in the ReFH report to use the
summer event had to be followed.
The rainfall depth for input to ReFH is calculated by applying a seasonal correction
factor to the depth derived from the FEH rainfall frequency model. In the case of the
summer event, the seasonal correction factor is quite close to 1, but we appl ied it
anyway.
For design runs of the ReFH model, the initial soil moisture Cini is scaled down by an
adjustment factor, , which was introduced to ensure that fluvial flood growth curves
produced by the ReFH model matched those estimated directly from flood peak data.
For our purposes we were interested only in the losses model, not the other aspects of
ReFH. It was not appropriate to apply the  factor in this case, which is only relevant
for modelling fluvial flood peaks.
The parameter Cmax and initial soil moisture Cini are estimated from two FEH
catchment descriptors: BFIHOST and PROPWET. Both of these descriptors are
simply spatial averages of gridded data and thus their definition is not restricted to
areas that comprise river catchments: they can be evaluated for any geographic area.

3 FAULKNER, D.S. AND BARBER, S., 2009. Performance of the Rev italised Flood Hy drograph Method. Journal of Flood Risk Management 2, 254-
261.
4 KJELDSEN, T.R., STEWART, E.J., PACKMAN, J.C., FOLWELL, S. AND BAY LISS, A.C., 2005. Revitalisation of the FSR/FEH Rainfall-Runoff
Method. Defra R&D Technical Report FD1913/TR.
5 FAULKNER, D.S, ROBB, K. AND HAYSOM, A., 2008. Return Period Assessment of the Summer 2007 Floods in Central England. BHS 10th
National Hy drology Symposium, Exeter, September 2008, 227-232.
6 WEBSTER, P., 1999. Factors Affecting the Relationship Between the Frequency of a Flood and its Causative Rainf all. In: L. Gottschalk, J.-C.
Oliv ry, D. Reed & D. Rosbjerg, eds. Hydrological Extremes: Understanding, Predicting, Mitigating (Proceedings of the IUGG99 Symposium HS1,
Birmingham, July 1999). IAHS Publication. No. 255. Wallingf ord: IAHS Press, 1999, 251-257.

18 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
BFIHOST data was extracted from the HOST mapping at the native 1km resolution
(see Section 2.3), where as PROPWET values were obtained from FEH catchment
descriptors previously extracted during step DP4 (see Section 2.4). PROPWET is
derived from MORECS data evaluated on a 40km x 40km grid and so it will take
identical values for sets of nearby 5km tiles.
Net rainfall hyetographs were therefore calculated on a 1km grid and then applied
within JFlow+ according to the urban-rural land cover mask shown in Figure 5 to
produce runoff for all rural areas or green spaces within urban areas (see Figure 8).

Figure 8 Example application of different net rainfall hyetographs according to

urban-rural classification for two adjacent modelling tiles

Figure 9 shows example results from the ReFH losses model: net rainfall hyetographs
for a 1.1-hour storm in rural areas with different soil types.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 19
 18
16
14
Rain (mm) 12
10
8
6
4
2
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Gross rainfall Time (hours)
Net rainfall: impermeable, wet soils
Net: Moderate permeability and wetness
Net: permeable, dry soils

Figure 9 Example differences between total and net rainfall

3.3 MOD3 - Run JFlow+ to check model set-up and edit

DTM
The composite DTM developed in steps DP1 and MOD1 will need further processing to
provide a suitable topographic basis for direct rainfall modelling. Manual editing is
required to provide flow paths through "flyover" features that present unrealistic barriers
to known flow routes. These features include road and railway embankments, bridges,
subways, and tunnels, and, unless edited, can cause runoff to back up and flood a
larger area "upstream" of the obstruction. Similarly, areas "downstream" of the
obstruction may be unrealistically shown as being free from flood risk (see Figure 10).

Figure 10 A DTM before and after manual editing

20 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
The approach adopted requires that a model run is completed and the results from the
model run used to identify where water is retained. Edits are then made to the local
DTM values and a further model run is undertaken. Further edits may be necessary
and the process is repeated until a satisfactory output is achieved. Achieving
satisfactory results in rural and urban areas can typically require between 2-3 and 5-10
iterations respectively. All edits are then checked and signed-off by experienced staff.
A range of datasets were used to support this work, including OS MasterMap and
1:10,000 Scale Raster mapping, vertical aerial photography (Google Earth, Bing Maps)
and 360-degree, street-level imagery (Google Street View), but as can be seen from
Figure 11, DTM editing is a very time-consuming and highly subjective task that
requires experienced judgement to determine which blockages should be removed and
how.
In total, approximately 91,000 edits were made to the DTM as part of this project.
Information useful for this task also exists within asset registers held by various
organisations (e.g. Highways Agency, Network Rail and local authority highway and
drainage departments) but it was not possible to collect and collate these data within
constraints of the project programme. It is hoped, however, that this information will be
used to check the mapping at key asset locations during the LLFA review process.

Figure 11 Areas (shown in red) which could need a DTM edit and require further
investigation

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 21
3.4 MOD4 - Finalise DTM and spatially-varying model
inputs
Following step MOD3, the DTM was now ready to undertake the final model runs for
each of the three flood probabilities required.
As well as topography, maps that describe the spatial variability of hydraulic roughness
were also required. Here these maps were based on the land cover information
provided by the OS MasterMap data which was then converted to Manning’s n values
using look-up tables developed during the recent Drain London project 7 (see Table 1
below).

Table 1 Manning’s n roughness values by Ordnance Survey MasterMap Feature

Code (reproduced from Capita Symonds/Scott Wilson (2010, p23))
Feature Descriptive Group Comment Manning's n value
Code
9999 0.035
10021 Building 0.015 (Depth <=
30mm)
0.500 (Depth > 30mm)
10053 General Surface Residential yards 0.040
10054 General Surface Step 0.025
10056 General Surface Grass, parkland 0.030
10062 Building Glasshouse 0.015 (Depth <=
30mm)
0.500 (Depth > 30mm)
10076 Land; Heritage And 0.500
Antiquities
10089 Water Inland 0.035
10093 0.040
10096 0.040
10099 0.040
10111 Natural Environment Heavy woodland and 0.100
(Coniferous / forest
NonConiferous Trees)
10119 Roads Tracks And Manmade 0.020
Paths
10123 Roads Tracks And Tarmac or dirt tracks 0.025
Paths
10167 Rail 0.050
10172 Roads Tracks And Tarmac 0.020
Paths
10183 Roads Tracks And Pavement 0.020
Paths (roadside)
10185 Structures Roadside structure 0.030
10187 Structures Generally on top of 0.015 (Depth <=
buildings 30mm)
0.500 (Depth > 30mm)

7 CAPITA SY MONDS AND SCOTT WILSON, 2010. Drain London - Data and Modelling Framework. Report version 1.0 (10 December 2010).

22 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
Feature Descriptive Group Comment Manning's n value
Code
10193 0.040
10203 Water Foreshore 0.040
10210 Water Tidal water 0.035
10217 Land (unclassified) Industrial yards, car 0.035
parks

As can be seen from Table 1, the Manning's n value associated with certain
MasterMap Feature Codes varies with depth. The conceptual argument for this
approach is presented in Capita Symonds/Scott Wilson (2010) and is implemented
here at the request of the Environment Agency.

3.5 MOD5 - Finalise rainfall hydrology

The sensed-checked local information on drainage rates and infiltration/runoff
coefficients provided by LLFAs was used to update the parameters of the rainfall-runoff
methods described in Sections 3.2.2 and 3.2.3.

3.6 MOD6 - Run final JFlow+ models

Using the inputs derived and finalised in the previous steps, the final model runs were
undertaken for each of the nine rainfall scenarios. As requested in the Project SoR, the
following outputs were produced for each hydraulic model run:
 Maximum flood depth, velocity and hazard rating
 Direction of flow at maximum hazard rating and at maximum velocity.

To ensure that the peaks in flood depth, velocity, and hazard rating are captured during
the simulation, a model simulation period of the storm duration plus 3 hours was used
throughout (i.e. if the storm duration was 6 hours then the hydraulic model was run for
9 hours).
As part of the quality assurance (QA) process, model diagnostic outputs such as
numerical convergence and mass balance were checked. Where QA checks did not
satisfy predetermined criteria based on the contractor project team's extensive
experience of similar modelling studies, the model run was flagged for user intervention
and re-run when resolved.
On project completion, a full set of electronic model input and output files in widely
supported formats (e.g. .xls, .csv or .txt) will be provided for re-use in JFlow+ or
alternative 2D hydraulic models.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 23
4 Flood mapping
GIS post-processing is required to improve the usability of the model output data for
identifying areas and receptors at risk of surface water flooding. Here the specification
for the flood mapping (i.e. post-processing thresholds, size of wet and dry ‘islands’ and
topology rules) was developed in close consultation with future users of the data. A
number of initial options were developed during face-to-face sessions with Environment
Agency technical specialists and the three pioneer LLFAs, which were then discussed
and voted on by attendees at the nine Capacity Building Workshops hosted by Defra in
November 2012. As well as reaching a strong consensus on what the final product
should look like, the process also highlighted the appetite of risk management
authorities to re-use and develop further the models and parameterisation datasets
within local studies.
The Project SoR also required pragmatic, easily understood methods for validating and
rating the confidence in the final flood map products to be developed. It is intended
that the default confidence scores produced via the automated methods described in
Section 4.2 will be confirmed or updated by LLFAs using locally-held flood risk
knowledge and data during the subsequent review phase.

4.1 MAP1 - Post-process depth, velocity and hazard rating

model outputs

4.1.1 Process overview

The flood map production process takes the JFlow+ modelling outputs and processes
them into usable maps to aid decision making processes.
The JFlow+ output data can be summarised by output types, rainfall probabilities and
storm durations (see Figure 12). In total, when combined together, there are 45
different outputs that require post processing.

Figure 12 Flood mapping data outputs

24 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
The flood mapping process has been split into four stages to facilitate the efficient post
process of data. These stages are shown in Figure 13 and are discussed in more
detail in subsequent sections:
1. Blend (Section 4.1.2)

2. Clean (Section 4.1.3)

3. Clip (Section 4.1.4)

4. Output (Section 4.1.5)

Figure 13 Overview of flood mapping process

In addition to splitting the post-processing into stages, the maps are produced on a
50km x 50km tile basis (as per the OS 50k reference frame, i.e. SK1, SK2, SK3, SK4,
etc) to facilitate the transfer of the maps and to keep the maps to a practical size for
end users. There are 93 50km x 50km tiles which cover England and Wales.
Map processing has been undertaken using the tools and functionality built into ESRI
ArcGIS. The process was automated using the ArcPy scripting language, which was
parallelised to increase processing performance. The entire process was further
automated through a job scheduler which controlled the processing of maps and the
allocation of jobs to processing machines.

4.1.2 Step 1 - Blending grids together

Combination (blend) of JFlow model tiles

Each JFlow+ model is run on a 6km x 6km tile which includes a 0.5km overlap of
results with neighbouring tiles. This generates an overlap with adjoining tiles so that
the results from each model can be combined.
Step 1 of the mapping process is to combine (blend) the JFlow+ model outputs for
each OS 50km tile. The process is described in Figure 14. For the blend process the
model output tiles are selected for the OS tile to be processed. In addition a set of
surrounding edge model tiles from adjacent OS tiles are also selected to ensure that a
smooth blend between OS tiles can be achieved.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 25
 Figure 14 Overview of Step 1 - Blending grids together

The approach to blending grids together varies depending on the output type:
 Depth, velocity and hazard output types:

To create a complete depth, velocity and hazard flood maps, we use a

weighted blend on the overlapping areas to join results together. The output
cell value of the overlapping areas is a blend of values that overlap; this
blend value is determined from an algorithm that is weight based and
dependent on the distance from the pixel to the edge within the overlapping
area.

 Flow direction output types:

For the creation of flow direction grids, a blend methodology is not

appropriate. A flow direction of 359o and 1o would result in an average of
179o in the central portion of the overlap, or some weighted variation in other
parts. A quadrant based averaging approach is instead used, which uses
different formulas based on the quadrant where the flow directions in each
overlapping tile lie. Figure 15 shows the quadrants and Table 2 shows the
formulae used to calculate the average flow direction in each case.

26 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
 Figure 15 Quadrants for flow direction analysis

Table 2 Quadrant analysis formula

Scenario Layer 1 (L1) Layer 2 (L2) Direction calculation
quadrant quadrant
1 1 1 (L1 + L2) / 2
2 1 2 (L1 + L2) / 2
3 1 3 If L2 - L1 < 180:
(L1 + L2) / 2
Else:
L2 + ((360 – L2 + L1) / 2)
4 1 4 If ((L1 + (360 – L2)) / 2) + L2 >= 360:
((L1 + (360 – L2)) / 2) + L2 – 360
Else:
(L1 + (360 – L2)) / 2 + L2
5 2 1 (L1 + L2) / 2
6 2 2 (L1 + L2) / 2
7 2 3 (L1 + L2) / 2
8 2 4 If L2 - L1 < 180:
(L1 + L2) / 2
Else:
L2 + ((360 – L2 + L1) / 2) - 360
9 3 1 If L1 - L2 < 180:
(L1 + L2) / 2
Else:
L1 + ((360 – L1 + L2) / 2)
10 3 2 (L1 + L2) / 2
11 3 3 (L1 + L2) / 2
12 3 4 (L1 + L2) / 2
13 4 1 If ((L2 + (360 – L1)) / 2) + L1 >= 360:
((L2 + (360 – L1)) / 2) + L1 – 360
Else:
(L2 + (360 – L1)) / 2 + L1
14 4 2 If L1 - L2 < 180:
(L1 + L2) / 2
Else:
L2 + ((360 – L2 + L1) / 2) - 360
15 4 3 (L1 + L2) / 2
16 4 4 (L1 + L2) / 2
Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 27
Each of the outputs of the blending process is stored in a master spatial database. In
total the process creates 45 blended output type grids. The file naming convention for
the outputs is:
XXX_YYY_HHH_T_MOS

Where:
XXX = OS 50k tile reference, e.g. TR3

YYY = P1000, P100, P30

HHH = 1HR, 3HR, 6HR

T = D (depth), V (velocity), H (hazard), VDMH (flow direction at maximum

hazard), VDMV (flow direction at maximum velocity)

Example:
SD1_P1000_1HR_D_MOS
TR3_P100_3HR_V_MOS

Maximum outputs
For each of the blended output types and rainfall probability grids, a grid of maximums
across the different storm durations is calculated to identify the worst case result in
every grid cell. These maps are then clipped in Step 3 of the process to give the final
outputs which appear on the LLFA review website (www.ufmfsw.com).
For the depth, velocity and hazard grids, a simple cell by cell maximum is calculated.
For the flow direction grids, the maximum value for each grid cell is selected from the
same storm duration that produced the worst case magnitude result.
In total the process creates 15 maximum output type grids. The file naming convention
is:
XXX_YYY_MAX_T_MOS

Where:
XXX = OS 50k tile reference, e.g. TR3

YYY = P1000, P100, P30

T = D (depth), V (velocity), H (hazard), VDMH (flow direction at maximum

hazard), VDMV (flow direction at maximum velocity)

Example:
SD1_P1000_MAX_D_MOS
TR3_P100_MAX_V_MOS

28 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
Critical storm duration
As part of the flood mapping process, critical storm duration outputs are generated.
These are useful in providing a visual check of the where the values from the maximum
outputs are obtained from and they are also used in the creation of the maximum flow
direction grids.
The critical storm duration process is undertaken on the depth, velocity and hazard
output types and in total it generates 9 maximum output type grids. The file naming
convention for the outputs is:
XXX_YYY_CSD_T_MOS

Where:
XXX = OS 50k tile reference, e.g. TR3
YYY = P1000, P100, P30
T = D (depth), V (velocity), H (hazard)
Example:
SD1_P1000_CSD_D_MOS
TR3_P100_CSD_V_MOS

4.1.3 Step 2 - Creation of cleaned outline

Threshold grids
The outputs which are generated from Step 1 contain complete coverage of flooding
and do not possess a conventional wet/dry flood outline. To create an outline, a
"cookie cutter" approach is used where a thresholded depth, velocity or hazard rating
grid is used as a mask to extract results from the other two datasets. A number of
possible options were considered:
 where surface water flooding is greater than 0.1m deep, extract the
corresponding velocity and hazard rating results

 where surface water flooding is quicker than 0.25m/s, extract the

corresponding depth and hazard rating results

 where surface water hazard rating is greater than 0.555 (see Table 3 below),
extract the corresponding depth and velocity results

 where surface water hazard rating is greater than 0.575 (see Table 3 below),
extract the corresponding depth and velocity results

 where surface water hazard rating is greater than 0.75 (i.e. in the "danger for
some" category, see Table 3 below), extract the corresponding depth and
velocity results

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 29
 Table 3 Hazard to People classification using FD2320/FD2321 hazard rating
formula (Depth x (Velocity + 0.5) + Debris Factor, where debris factor = 0.5 if
depth <= 0.25m and debris factor = 1 if depth > 0.25m)
Depth (m)
0.05 0.10 0.20 0.25 0.30 0.50 0.75 1.00 2.00 3.00 4.00 5.00
0.05 0.528 0.555 0.610 0.638 1.165 1.275 1.413 1.550 2.100 2.650 3.200 3.750
0.10 0.530 0.560 0.620 0.650 1.180 1.300 1.450 1.600 2.200 2.800 3.400 4.000
0.20 0.535 0.570 0.640 0.675 1.210 1.350 1.525 1.700 2.400 3.100 3.800 4.500
0.25 0.538 0.575 0.650 0.688 1.225 1.375 1.563 1.750 2.500 3.250 4.000 4.750
0.30 0.540 0.580 0.660 0.700 1.240 1.400 1.600 1.800 2.600 3.400 4.200 5.000
0.40 0.545 0.590 0.680 0.725 1.270 1.450 1.675 1.900 2.800 3.700 4.600 5.500
Velocity (m/s) 0.50 0.550 0.600 0.700 0.750 1.300 1.500 1.750 2.000 3.000 4.000 5.000 6.000
0.75 0.563 0.625 0.750 0.813 1.375 1.625 1.938 2.250 3.500 4.750 6.000 7.250
1.00 0.575 0.650 0.800 0.875 1.450 1.750 2.125 2.500 4.000 5.500 7.000 8.500
2.00 0.625 0.750 1.000 1.125 1.750 2.250 2.875 3.500 6.000 8.500 11.000 13.500
3.00 0.675 0.850 1.200 1.375 2.050 2.750 3.625 4.500 8.000 11.500 15.000 18.500
4.00 0.725 0.950 1.400 1.625 2.350 3.250 4.375 5.500 10.000 14.500 19.000 23.500
5.00 0.775 1.050 1.600 1.875 2.650 3.750 5.125 6.500 12.000 17.500 23.000 28.500

Hazard Rating Key Hazard to People Classification

Less than 0.75 Very low hazard – caution
0.75 to 1.25 Danger for some – includes children, the elderly and the infirm
1.25 to 2.0 Danger for most – includes the general public
More than 2.0 Danger for all – includes the emergency services

In consultation with stakeholders, it was decided that the maps should be thresholded
based on a hazard rating of 0.575. This covers shallow flows ≥ 1m/s and all depths ≥
0.15m. This value was used to create a polygon from the hazard output type rasters.
To aid the processing of this data individual isolated pixels ≤ 4m 2 (1 cell) were removed
from the analysis. This process creates 3 feature classes per OS 50k tile and they are
called:
XXX_YYY_MAX_H_MOS_MASK

Where
XXX = OS 50k tile reference, e.g. TR3
YYY = P1000, P100, P30
Example:
SD1_P1000_MAX_H_MOS_MASK
TR3_P100_MAX_H_MOS_MASK

Clean up of outline
Several algorithms have been used to clean the flood polygons generated from the
hazard threshold:
 Fill of areas smaller than 48m 2 (Figure 16)

 Removal of polygons less than 96m 2 unless they touch a larger than 96m 2
area polygon (Figure 17)

 Polygons kept where they touch other polygon(s) where their combined
areas are greater than 96m 2 (Figure 18)

30 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
 Figure 16 Areas filled in (blue)

Figure 17 Areas removed (red)

Figure 18 Areas on diagonal with an area > 96m2

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 31
This process creates 3 feature classes per OS 50k tile and they are called:
XXX_YYY_MAX_H_MOS_OUTLINE

Where:
XXX = OS 50k tile reference, e.g. TR3
YYY = P1000, P100, P30
Example:
SD1_P1000_MAX_H_MOS_OUTLINE
TR3_P100_MAX_H_MOS_OUTLINE

Figure 19 Overview of Step 2 - Creation of cleaned outline

32 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
4.1.4 Step 3 - Clipping of grids to the outline

Clipping of rasters
A constant raster clipping mask is generated from the flood outline and this is clipped to
the OS 50k tile boundary and the coastline. The mask is then used to clip the blended
grids for maximum depth, velocity, hazard and flow direction x 2, and critical storm
duration for depth, velocity and hazard.
This process is shown in Figure 20. It creates 24 grids for each tile and the clipped
grids are called:
XXX_YYY_ZZZ_T_MOS_CLIP

Where:
XXX = OS 50k tile reference, e.g. TR3

YYY = P1000, P100, P30

ZZZ = MAX or CSD

T = D (depth), V (velocity), H (hazard), VDMH (flow direction at maximum

hazard), VDMV (flow direction at maximum velocity)

Example:
SD1_P1000_MAX_V_MOS_CLIP
TR3_P100_MAX_VDMH_MOS_CLIP

Resampling of flow direction

Some thinning of the 2m data is required to produce useful maps of flow direction at
different zoom scales. Figure 20, Figure 21 and Figure 22 show that at:
 1:500 scale a 2m grid is appropriate

 1:1,000 scale a 5m grid is appropriate

 1:3,000 scale a 10m grid is appropriate

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 33
 Figure 20 Flow direction at 1:500 on a 2m grid

Figure 21 Flow direction at 1:1,000 on a 5m grid

Figure 22 Flow direction at 1:3,000 on a 10m grid

34 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
On completion of clipping, the flow direction grids are re-sampled from 2m to 5m, 10m
and 20m grids for their anticipated future use. This creates a further 18 grids for each
tile and they are called:
XXX_YYY_MAX_T_MOS_CLIP_SSS

Where:
XXX = OS 50k tile reference, e.g. TR3

YYY = P1000, P100, P30

T = D (depth), V (velocity), H (hazard), VDMH (flow direction at maximum

hazard), VDMV (flow direction at maximum velocity)

SSS = resampled grid resolution, either 5M, 10M, 20M

Example:
SD1_P1000_MAX_ VDMV_MOS_CLIP_5M
TR3_P100_MAX_VDMH_MOS_CLIP_20M

Figure 23 Overview of Step 3 - Clipping of grids to the outline

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 35
4.1.5 Step 4 - Creation of outputs
Flood maps are output from the master spatial database into the following formats:
 ESRI File GeoDatabase (version 9.3.2)

 GeoTIFF images of clipped grids (compatible in either ESRI or MapInfo

software)

 ASCII files of clipped grids (compatible in either ESRI or MapInfo software)

Figure 24 shows an overview of the datasets that are outputted to each format. These
117 files are produced on an OS 50k tile basis (i.e. SK1, SK2, SK3, SK4, etc). In
summary there are:
 27 x blended raster grids of depth, velocity, hazard for the 1 in 30, 1 in 100
and 1 in 1,000 probabilities and 1hr, 3hr and 6hr storm durations

 9 x maximum grids of depth, velocity, hazard for the 1 in 30, 1 in 100 and 1 in
1,000 probabilities

 9 x critical storm duration grids of depth, velocity, hazard for the 1 in 30, 1 in
100 and 1 in 1,000 probabilities

 18 x blended raster grids of direction of flow at maximum velocity and

direction of flow at maximum hazard for the 1 in 30, 1 in 100 and 1 in 1,000
probabilities and 1hr, 3hr and 6hr storm durations

 6 x maximum grids of direction of flow at maximum velocity and direction of

flow at maximum hazard for the 1 in 30, 1 in 100 and 1 in 1,000 probabilities

 3 x hazard polygon masks for the 1 in 30, 1 in 100 and 1 in 1,000

probabilities

 3 x cleaned hazard polygon masks for the 1 in 30, 1 in 100 and 1 in 1,000
probabilities

 24 x clipped 2m grids of depth, velocity, hazard, direction of flow at maximum

velocity and direction of flow at maximum hazard for the 1 in 30, 1 in 100 and
1 in 1,000 probabilities and 1hr, 3hr and 6hr storm durations

 18 x resampled clipped grids of direction of flow at maximum velocity and

direction of flow at maximum hazard for the 1 in 30, 1 in 100 and 1 in 1,000
probabilities at 2m, 5m and 20m

Due to the number of grids and the size of files, each dataset occupies a sizeable
amount of storage space. The average file sizes for an OS tile with complete coverage
are:
 ESRI File GeoDatabase - 137Gb

 GeoTIFF images of clipped grids - 4Gb

 ASCII files of clipped grids - 83Gb

Approximately 15Tb of storage space is required to hold the complete set of maps.

36 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
 Figure 24 Overview of Step 4 - Creation of outputs

4.2 MAP2 - Undertake flood map validation and confidence

rating
The aim of the validation and confidence mapping work was to provide users with
guidance and information which allows them to view the map with an appropriate level
of confidence. This was a technically challenging aspect of the project. Previous
projects have not provided a clear lead in how surface water model predictions can be
validated to assess confidence in their outputs, and as noted in the project start up
meeting "effectively communicating the level of confidence/certainty in the new
mapping will be difficult".
The system used for classifying confidence is based on a star rating from 1 to 5 stars,
with 1 star representing lowest confidence, and 5 stars representing the highest. For
the national scale flood mapping, the star rating is expected to be in the range 1-3, with
a 4 star rating possible if detailed local modelling is included in the Updated Flood Map
for Surface Water. This means future improvements in national scale mapping and
local modelling can be included in the same system as confidence moves towards a 5
star rating.
The star rating system is also being used as part of a project to map confidence in
maps of flooding from rivers and the sea, as described in Environment Agency (2013b).
Using a single system for flooding from rivers, the sea and surface water will make
understanding confidence easier for users and allow them to understand the
differences in confidence in maps of different sources of flooding.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 37
A star rating has been derived for the national scale mapping. This varies between
locations, and is produced as a grid of 50m x 50m cells. The confidence star rating is
based on the quality of input data, topography (steep/flat) and the land cover type
(urban/rural). Validation has been carried out for the 3 pioneer locations, to confirm
that the national scale mapping is capable of adequately reproducing observed
flooding. More details of how the star rating has been derived and validation results
can be found in Environment Agency (2013c).
Users will have the opportunity to review and change the star rating, and provide an
appropriate star rating for local modelling uploaded as part of the Updated Flood Map
for Surface Water. Detailed guidance for users on how to do this, and appropriate uses
of the surface water flood maps based on the confidence star rating can be found in
Environment Agency (2013a).
An overview of the confidence rating system, and the validation process used to inform
it, are given in the following sections.

4.2.1 Confidence rating

Based on the findings of the FMfSW Improvements Pilot Studies (Halcrow/JBA
Consulting, 2012), the method recognises that the sources of uncertainty in model
outputs are in two broad classes: related to the effective rainfall input into the model,
and related to how water is routed by the model. These uncertainties are used to
derive the confidence star rating; where uncertainty is higher, the confidence star rating
will be lower.
Effective rainfall uncertainty arises from uncertainty in the total rainfall (derived from
FEH depth-duration-frequency relationships) and the infiltration/drainage model.
Uncertainty will be greater in urban areas where no local drainage information has
been supplied, where the default drainage rate of 12 mm/hr is used. For areas where
local drainage capacity information has been provided, or for rural areas which are
unaffected by drainage systems, uncertainty is lower.
Uncertainty in how water is routed by the model is based on the key influences of DTM
source (LIDAR or NEXTMap), topographic steepness, and influence of buildings.
Uncertainty will be higher in areas with NEXTMap-derived DTM, because the
NEXTMap data has less vertical accuracy than LIDAR. Uncertainty will be higher in
flatter topography because errors in the DTM will be greater in relative terms than the
underlying topography which influences patterns of flooding; for steeper areas DTM
errors matter less because the topographic variations are larger. Uncertainty will be
higher in rural areas where the chief control on flow paths is the DTM, and lower in
urban areas where building and street patterns have a significant influence on flow
paths.

Table 4 Combination of model routing and effective rainfall uncertainties into

final model output uncertainty - example values
Urban LLFA Urban Default Rural
DTM NEXTMap Flat
Steep
LIDAR Flat
Steep

38 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
In practice, these effects have been combined in a scoring system, based on sensitivity
analysis and the outputs of previous surface water mapping projects (see Environment
Agency 2013c) for more details). The inputs to the scoring system are:
 Polygons of LIDAR and NEXTMap coverage

 Polygons of urban areas derived from the Office for National Statistics
England and Wales Urban Area Boundaries (2001) dataset

 A raster map of topographic steepness, derived by calculating the

topographic slope, in degrees, from a 25m resolution DTM, then taking the
median value of this slope over 5km x 5km tiles. Steep areas are defined as
those giving a median value greater than 2°. This gives a map which is
representative of steeper and flatter areas across England and Wales.

 Polygons of areas where LLFAs have provided local drainage information

These data sources, and the results of applying the scoring system, are shown in
Figure 25. The map reflects what would be expected from our understanding of the
factors affecting confidence in national scale map outputs. There are large areas of 3
star rated outputs in Wales, the West Country and the North West, where topography is
steeper and there is good LIDAR coverage. There are large areas of 1 star output in
the South East, where NEXTMap coverage coincides with flatter topography.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 39
 Figure 25 Data used to derive the confidence star rating and final map output

40 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
4.2.2 Validation
The purpose of the validation component of this project was to assess how well the
new national scale map reproduces true flooding and how the well the discrepancies
between the national scale map and observed flooding are reflected by the confidence
rating. This process has been limited by difficulty of accessing good quality validation
data sets; finding validation data for the 10 locations required in the Project SoR was
not possible.
Effort has instead been focussed on producing a robust but easily followed local review
process, to allow LLFAs and other partners to provide feedback on the quality of
national scale mapping outputs based on their understanding of local flooding issues.
The review process is described in more detail in Environment Agency (2013a).
Limited validation has been undertaken for the 3 pioneer locations of Greater
Manchester, Gloucestershire and Rhondda Cynon Taff (RCT), using both qualitative
methods and quantitative analysis of flooded properties identified by the national scale
map. The results are described fully in the separate validation report (Environment
Agency, 2013c); an overview and summary of the conclusions is given below.
Evidence for improvements in the national scale map outputs over previous versions
(AStSWF and FMfSW) is also included.
Table 5 summarises the validation data available for the three pioneer locations.
Gloucestershire and RCT have some good quality evidence; data for Greater
Manchester is less useful, with only data from Bolton being useable.

Table 5 Validation evidence for the pioneer locations

Location Data Source Number and Date
Greater Manchester Polygons of flooding 62 records from 1946-2009
recorded in Bolton, with
source noted as surface
water
Polygons of wet spots 29 records, no dates
recorded in Bolton
Gloucestershire Points of recorded flood 3424 possible surface
incidents water incidents, summer
2007
Output grid from InfoWorks Approximately equivalent
ICM model for 1 in 100 to summer 2007 event
probability event
Rhondda Cynon Taff Point records of flooded 946 points from 2002-2011
buildings
Points records of external 1874 points from 2002-
flooding 2011

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 41
Greater Manchester
The lack of good quality evidence for Greater Manchester makes it difficult to draw
conclusions about model performance. The national scale map correctly identifies
some flooding locations, but not all and there are areas of significant flooding in the
map not associated with a known flooding location. An example is shown in Figure 26;
the results are similar for the rest of Bolton.

Figure 26 Comparison between known flooding locations (surface water in

purple polygons, unknown sources in green) and national scale map outputs (1
in 30 probability) for two locations in Greater Manchester

There is little evidence for an improvement in results compared to previous

generations. The results for FMfSW and the latest national scale map are very similar
(see Figure 27), predicting isolated areas of flooding which roughly coincide in the two
maps. This is to be expected, given the similarity of the modelling approaches used in
the two generations. AStSWF predicts much more flooding, as is expected since it
makes no allowance for drainage, and uses a more extreme design storm.
Most of Bolton and Manchester are assigned a confidence rating of 3*, with some
areas of 2*, by the confidence rating method described in the previous section. The
validation data is of insufficient quality to assess whether this is appropriate.

42 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
Figure 27 Comparison between AStSWF (top, 1 in 200 probability), FMfSW
(middle, 1 in 30 probability) and latest national scale map (bottom, 1 in 30
probability) for Manchester city centre
Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 43
Gloucestershire
National scale map outputs have been validated against a high quality InfoWorks ICM
model of Cheltenham. This model represents sub-surface drainage through a linked
sewer model, with hydraulics based on physical properties (pipe network location and
dimensions) rather than an assumed level of performance. It should therefore provide
a good test of national scale map quality as independent validation evidence.
Results are shown in Figure 28. The correspondence between the two models is
close, and both pick out groups of buildings which were reported as flooding in 2007
(which had a probability of approximately 1 in 100). The flood extent for the ICM
outputs appears larger, but this is mainly due to large areas of very shallow water
which are not regarded as significant.
A direct comparison of depths predicted by the two models has also been made. The
national scale mapping predicts significantly deeper water than InfoWorks, and the two
sets of point depth predictions match approximately within a factor of 2. The two
models also predict approximately the same areas of lower and higher velocity (Figure
29).

Figure 28 Comparison between latest national scale map (top, 1 in 100

probability) and InfoWorks ICM (bottom, 1 in 100 probability). Properties flooded
in 2007 are shown in red.

44 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
 Figure 29 Velocities predicted by the InfoWorks ICM (top) and national scale
mapping models (bottom) in Cheltenham, displayed in two classes: blue <1 m/s
and red >1 m/s

The dataset of properties flooded in summer 2007 allows a quantitative comparison of

properties predicted as flooding to be made, with the results summarised in Table 6 for
floods recorded in Cheltenham. Properties were identified as flooded by buffering the
building footprint by 2m; taking the maximum depth around the buffered footprint; and
thresholding this at 0.3m. Property counts are sensitive to buffering and thresholds
applied, and further investigation of property counting methods is being undertaken at
the time of writing.
The national scale map correctly identifies 76 out of 100 properties that were observed
to flood, but there is also significant overprediction with a further 813 properties
predicted as flooding which were not reported as doing so. This result is consistent
with previous studies, which also found significant overestimation of property counts by
factors of 10 or more. The national scale map and InfoWorks ICM perform similarly;
InfoWorks ICM identifies fewer of the flooded properties and halves the number of
properties mapped as flooded that have no record of flooding.
The previous versions of the Environment Agency's surface water flood maps show
similar levels of performance, both identifying >80% of flooded properties with
significant overprediction in the number of flooded properties. In the case of FMfSW
this will be partly due to it modelling a 1 in 200 probability event rather than the 1 in 100

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 45
event observed. This indicates that identifying flooded properties is relatively easy for
this site.
The area covered by the InfoWorks ICM model has been assigned a confidence rating
of 3 stars for the majority of its area. This agrees well with the outputs of the validation
study, which indicates that the national mapping should be associated with a high
confidence. For the wider Cheltenham area where property count validation has been
carried out, the confidence rating is approximately half 2* and half 3*. These
confidence ratings are broadly appropriate given the number of properties correctly
identified.
In summary, the national scale map does well in reproducing observed patterns of
flooding (including at individual properties), with performance equalling that of a more
detailed local model. The map appears to overestimate numbers of flooded properties
by a factor of 10.

Table 6 Results of property validation for flooded properties recorded in 2007

summer floods in suburb of Cheltenham covered by InfoWorks ICM model
Hits Misses Mapped Total Total
but no predicted observed
record of
flooding
National scale map 76 24 813 889 100
1 in 100
InfoWorks ICM 56 44 413 469 100
1 in 100
AStSWF 84 16 995 1079 100
Less category
FMfSW 94 4 1680 1774 100
1 in 200

Rhondda Cynon Taff

Surface water flood incidents have been recorded over a period of approximately 10
years in RCT, and these should correspond most closely with the 1 in 30 probability
event. However, Figure 30 shows that there is little correlation between the national
flood map and the locations of flood incidents, which are scattered throughout the
urban areas with no obvious clustering. The rest of RCT shows similar patterns.
Table 7 shows the results for property count validation for RCT. All 3 generations of
modelling show similar performance, identifying ~1/3 of the 1,000 flooded buildings
correctly. The models identify 20-30 thousand properties in the 1 in 30 probability flood
outline, meaning that on average ~10 thousand properties might be expected to flood
in a 10 year period. The models therefore appear to be overpredicting the number of
flooded properties by a factor of ~30 (although the exact figure may change with any
updated property counting method as discussed in the previous section).
The majority of RCT is assigned a confidence rating of 3 stars, reflecting the steep
topography and LIDAR DTM. The validation outputs indicates that this is not
appropriate, given the discrepancies between observed and predicted patterns of
flooding.
In summary, the performance of the national scale map in RCT is not as good as for
Cheltenham, either in terms of predicting patterns of flooding, or property counts.

46 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
 Table 7 Results of property validation for internal flood incidents recorded in
RCT, 2002-2011

Hits Misses FA Total Total

predicted observed
National scale map 367 572 26326 26693 939
1 in 30
AStSWF 319 620 24586 24905 939
Less category
FMfSW 294 645 21359 21653 939
1 in 30

Figure 30 National scale map for 1 in 30 probability, with reported flood

incidents as red points, for two locations in RCT
Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 47
4.3 MAP4 - Deliver national scale surface water flood map
At the end of the project, the contractor team will deliver all national scale surface water
flood mapping output data and default confidence ratings for the coverages specified in
the Project SoR to the Environment Agency. All data will be supplied with appropriate
EU INSPIRE-compliant metadata.
The contractor team will also provide:
 A full set of electronic model input and output files in JFlow+ format
 A full set of electronic model input files in widely supported formats (e.g. xls,
.csv or .txt) for re-use in alternative 2D hydraulic models by/for you, LLFAs or
their contractors. All model grids (e.g. DTM, rainfall-runoff, Manning's n) will
be supplied in ESRI File GeoDatabase, GeoTIFF and ASCII raster formats.

48 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
5 Comparison of Environment
Agency surface water flood
mapping methodologies
A summary of the methodological differences between the AStSWF, FMfSW and the
new national scale surface water flood mapping is provided in Table 8 below.

Table 8 Comparison of Environment Agency surface water flood mapping

methodologies

Item AStSWF FMfSW uFMfSW

Modelling approach
Hydraulic 2D overland flow 2D overland flow 2D overland flow
modelling modelling modelling modelling
Hydrological Direct Rainfall Direct Rainfall Direct Rainfall
modelling approach with no approach with approach with
allowances made for allowances for the allowances for the
the sewer network and sewer network and sewer network and
infiltration: infiltration: infiltration:
- England and Wales
classified as either
rural or urban based
on manmade area
coverage as defined in
OS MasterMap
Topography Layer on
250m x 250m grid.
- In rural areas, ReFH
losses model is
applied and is
parameterised using
NSRIs ‘SERIES
Hydrology’ data.
- In urban areas, a
default loss rate of
12mm/hr and 70%
runoff coefficient is
applied.
- Runoff parameters
adjusted by local
drainage information
where available
Software JFlow-DW (diffusion JFlow-DW (diffusion JFlow+ (Shallow
wave-based) wave-based) Water Equation-
based)

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 49
Item AStSWF FMfSW uFMfSW
Hydraulic model parameters
Digital Infoterra bare earth Bare earth Bare earth
elevation LIDAR and LIDAR/NEXTMap
LIDAR/NEXTMap
model GeoPerspectives DTM composite DTM at 5m
composite DTM at 2m
provided in 2007. horizontal resolution
horizontal resolution
provided by provided by
Geomatics in 2010.
Geomatics in 2012.
Grid size 5m regular grid 5m regular grid
2m regular grid
Representation Not represented Represented explicitly
- Use of an
of buildings as unfloodable objects
approximately 0.3m
“up-stand” and depth-
in the DTM. Building
footprints, as defined
varying roughness
in OS MasterMap
coefficients within the
data, raised by 5m.
OS MasterMap
building footprint (as
per Drain London
specification).
- Complex rules were
developed to ensure
that building footprints
did not “cut” into the
DTM but did always
protrude at the
upstream face of the
building.
Representation DTM was manually DTM was manually DTM was manually
of structures edited in over 5,000 edited in over 40,000 edited in over 91,000
locations to improve locations to improve locations to improve
flow through ‘flyover’ flow through ‘flyover’ flow through ‘flyover’
features, such as features, such as features, such as
rail/road embankment rail/road embankment rail/road embankment
culverts, bridges etc. culverts, bridges etc. culverts, bridges etc.
Representation N/A N/A Areas of the DTM
of other covered by the road
features network were lowered
by 0.125m.
Manning’s n 0.1 0.1 rural, 0.03 urban Varied by OS
values MasterMap
Topography Layer
Feature Code as per
Drain London
specification
Mass balance Not recorded ±1% 0% (JFlow+ is mass
conservative by the
nature of their
numerical formulation)
End of Dynamic stopping Dynamic stopping Rainfall event duration
simulation condition. Models will condition. Models will + 3hrs
criteria stop running if the stop running if the
number of wet cells is number of wet cells is
unchanged over a 1 unchanged over a 1
hour period. hour period.

50 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
Item AStSWF FMfSW uFMfSW
Rainfall and hydrology
Design rainfall FEH depth-duration- FEH depth-duration- FEH depth-duration-
frequency parameters frequency parameters frequency parameters
defined on a regular defined on a regular defined on a regular
5km grid: 5km grid: 5km grid:
- 1 in 200 rainfall - 1 in 30 and 1 in 200 - 1 in 30, 1 in 100, 1 in
probability. rainfall probabilities. 1,000 rainfall
- A storm duration of - A storm duration of probabilities.
6.25hrs was used for 1.1hrs was used for all - Storm durations of 1,
all scenarios. scenarios. 3 and 6hrs were used
- 50% summer storm - 50% summer storm for all scenarios.
profile used profile used - 50% summer storm
- No aerial reduction - No aerial reduction profile used
factor applied factor applied - No aerial reduction
factor applied
Inflows from None None None
outside of
study area
Downstream Free overflow Free overflow Free overflow
boundary
conditions
Validation of results and sensitivity testing
Validation High-level evaluation Undertaken for 11 Undertaken for three
of the potential uses areas using historical pilot areas using
and quality of the observations and local historical observations
AStSWF. Some modelling data (see and local modelling
qualitative comparison Halcrow, 2010). data (see Environment
against historical Agency, 2013c).
observations and local
modelling data (see
Halcrow, 2008).
Sensitivity None None None
testing

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 51
References
ALLITT, R. 2009. Coupled 1D - 2D Modelling in Urban Areas. WaPUG User Note No
40. Report version 1.0 (November 2009).
BOORMAN, D.B., HOLLIST, J.M. AND LILLY, A., 1995. Hydrology Of Soil Types: A
Hydrologically-Based Classification of the Soils of the United Kingdom. Institute of
Hydrology Report No. 126.
BURTON, A., GLENIS, V., BOVOLO, C.I., BLENKINSOP, S., FOWLER, H.J., CHEN,
A.S., DJORDJEVIC, S. AND KILSBY, C.G., 2010. Stochastic Rainfall Modelling for the
Assessment of Urban Flood Hazard in a Changing Climate. BHS Third International
Symposium, Newcastle.
CAPITA SYMONDS AND SCOTT WILSON, 2010. Drain London - Data and Modelling
Framework. Report version 1.0 (10 December 2010).
CROSSLEY, A., LAMB, R., WALLER, S. AND DUNNING P., 2010a. Fast 2D Flood
Modelling using GPU Technology. European Geosciences Union, General Assembly,
Vienna, Austria, May 2010.
CROSSLEY, A., LAMB, R., WALLER, S. 2010b. Fast solution of the shallow water
equations using GPU technology, BHS International Symposium, Newcastle, UK, July
2010.
DEFRA, 2010. Surface Water Management Plan Technical Guidance.
ENVIRONMENT AGENCY, 2010. Flood Map for Surface Water Property Count
Method. Report version 1.0 (10 November 2010).
ENVIRONMENT AGENCY, 2012a. Guidance on Surface Water Flood Mapping for
Lead Local Flood Authorities. Report version 1.0 (26 June 2012).
ENVIRONMENT AGENCY, 2012b. Making Use of Local Information in National Scale
Surface Water Flood Mapping. Report version 1.0 (11 June 2012).
ENVIRONMENT AGENCY, 2013a. Reviewing the national scale surface water flood
maps. Report version 1.0 (28 February 2013).
ENVIRONMENT AGENCY, 2013b. Measuring Confidence in NaFRA Outputs - Stage 1
Method Development. Report version 1.0 (22 March 2013).
ENVIRONMENT AGENCY, 2013c. Updated Flood Map for Surface Water - National
Scale Surface Water Flood Mapping Confidence and Validation. Report version 1.0 (12
March 2013).
FAULKNER, D.S., 1999. Flood Estimation Handbook. Volume 2: Rainfall Frequency
Estimation. Institute of Hydrology, 110 pages.
FAULKNER, D.S., 2010. Hydrological inputs to a broad-scale model of pluvial flooding.
BHS International Symposium, Newcastle, UK, July 2010.
FAULKNER, D.S. AND BARBER, S., 2009. Performance of the Revitalised Flood
Hydrograph Method. Journal of Flood Risk Management 2(4), 254-261.
FAULKNER, D.S., KJELDSEN, T.R., PACKMAN, J.C. AND STEWART, E., 2012.
Estimating Flood Peaks and Hydrographs for Small Catchments: Phase 1.
Environment Agency Science Report SC090031/R1.

52 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology
FAULKNER, D.S, ROBB, K. AND HAYSOM, A., 2008. Return Period Assessment of
the Summer 2007 Floods in Central England. BHS 10th National Hydrology
Symposium, Exeter, UK, September 2008, 227-232.
HALCROW, 2007. Hull Level 2 Strategic Flood Risk Assessment (2007). Report to Hull
City Council.
HALCROW, 2008. Areas Susceptible to Surface Water Flooding Assessment. Final
Report to the Environment Agency of England and Wales.
HALCROW, 2010. Flood Map for Surface Water Final Evaluation Report. Final Report
to the Environment Agency of England and Wales.
HALCROW, 2011. Derivation of a National Pluvial Flood Hazard Dataset. Report to
SEPA.
HALCROW AND JBA CONSULTING, 2012. Flood Map for Surface Water
Improvements – Pilot Studies Final Evaluation Report. Report version 4.0 (9 May
2012).
HANKIN, B., WALLER, S., ASTLE, G. AND KELLAGHER, R., 2008. Mapping space for
water: screening for urban flash flooding. Journal of Flood Risk Management 1(1), 13-
22.
HURFORD, A.P., PARKER, D.J., PRIEST, S.J. AND LUMBROSO, D.M., 2012.
Validating the Return Period of Rainfall Thresholds Used for Extreme Rainfall Alerts by
Linking Rainfall Intensities with Observed Surface Water Flood Events. Journal of
Flood Risk Management 5(2), 134-142.
HR WALLINGFORD, 2010. National Pluvial Screening Project for Ireland. Report
EX6335 for Office of Public Works.
JBA CONSULTING, 2009. SC080029 - Refining the Data Quality and the Methodology
for Mapping Surface Water Flood Risk. Report to Environment Agency.
KJELDSEN, T.R., JONES, D. A. AND BAYLISS, A.C., 2008. Improving the FEH
Statistical Procedures for Flood Frequency Estimation. Environment Agency Science
Report SC050050.
KJELDSEN, T.R., STEWART, E.J., PACKMAN, J.C., FOLWELL, S. AND BAYLISS,
A.C., 2005. Revitalisation of the FSR/FEH Rainfall-Runoff Method. Defra R&D
Technical Report FD1913/TR.
LAMB, R., CROSSLEY, A. AND WALLER, S., 2009. A Fast Two-Dimensional
Floodplain Inundation Model. Proceedings of the ICE - Water Management 162(6),
363-370.
OSBORNE, M., 2012. Design Storms - Have We Been Getting It Wrong All This Time?
WaPUG Spring Conference, Birmingham, UK, May 2012.
VIESSMAN, W. AND LEWIS, G.L., 2003. Introduction to Hydrology (Fifth Edition).
Prentice Hall, 612 pages.
WAPUG, 2009. Integrated Urban Drainage Modelling Guide.
WEBSTER, P., 1999. Factors Affecting the Relationship Between the Frequency of a
Flood and its Causative Rainfall. In: L. Gottschalk, J.-C. Olivry, D. Reed & D. Rosbjerg,
eds. Hydrological Extremes: Understanding, Predicting, Mitigating (Proceedings of the
IUGG99 Symposium HS1, Birmingham, July 1999). IAHS Publication. No. 255.
Wallingford: IAHS Press, 1999, 251-257.

Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology 53
List of abbreviations
ARF Areal Reduction Factor
AStSWF Areas Susceptible to Surface Water Flooding
BFIHOST Baseflow Index (estimated from soil type)
DDF Depth-Duration-Frequency
DTM Digital Terrain Model
FDGiA Flood Defence Grant-in-Aid
FEH Flood Estimation Handbook
FMfSW Flood Map for Surface Water
FSR Flood Studies Report
HOST Hydrology Of Soil Types
IfSAR Interferometric Synthetic Aperture Radar
IUD Integrated Urban Drainage
LIDAR Light Detection And Ranging
LLFAs Lead Local Flood Authorities
LTIS Long Term Investment Strategy
NRD National Receptors Dataset
OS Ordnance Survey
QA Quality Assurance
ReFH Revitalised Flood Hydrograph
SoR Statement of Requirements
SPRHOST Standard Percentage Runoff (estimated from soil type)
SWMP Surface Water Management Plan
WaSCs Water & Sewerage Companies

54 Updated Flood Map for Surface Water - National Scale Surface Water Flood Mapping Methodology