You are on page 1of 40

Modelling Historic Flooding

of the Avon in Bath


By: Thomas Hewlett
Supervised By: Thomas Kjeldsen

My thanks go to Thomas Kjeldsen of the University of Bath, for his help and patience on this project,
and to Daniel Brown of Bath in Time for his help with historical photography.

Abstract
In order to make predictions for flooding analyses, knowledge of historical flooding is required. With this
in mind this paper looks into the creation of a model generate such data for the River Avon in Bath.
Through the model, analysis of the river was also performed.

The River Avon was drastically changed in between 1966 1974 as part of the Greenhalgh flood protection scheme. Prior to this only minor changes had been made, this allows the utilisation of the model developed in this paper for other historical floods. Given the time period of the changes, the 1960 flood has
been chosen for this analysis to represent the historical floods.

Two weighted Mannings coefficient models for both the year 1960 and 2014, two sub-sectioned models,
for the same two years, and a gradually varying flow model for 1960 have been created. They provide a
comparison across the implementation of the flood defence scheme. These models have been used to
analyse the passage of water in the Avon through the centre of Bath, and how this has changed over time.
Each model comprises of at least nine cross sections, which have been drawn from surveys and historical
photography. These sections run from Churchill Bridge to Bath Library. The paper also includes individual
analysis of each cross section.

The normal flow models found Old Bridge to prevent water leaving the city. A maximum normal flow under the bridge was calculated to be 228m3/s, historical flood discharge records commonly exceeded this.
The bridge was also liable to blockage, which was found liable to reduce flow capacity by 25%.
The gradually varying flow analysis establishes the high bed at HaPenny Bridge in 1960 at the Primary reason behind the widespread flooding upstream.
The Greenhalgh scheme has been found to be very effective in nearly all aspects. The replacement of Old
Bridge with Churchill Bridge and the dredging the river around HaPenny Bridge being of particular merit.

2 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Contents
Introduction

Literature Review

Importance of Historic Data in Probability analysis

Flow Equations

Previous Case Studies

The Avon River in Bath

History of the Development of Bath

History of Flooding

11

Proposed Flood Protection schemes

11

Actioned Flood Protection scheme The Greenhalgh Scheme

12
14

Normal Flow Solver


Calculations

14

Limitations

15

Sub-section Model vs Weighted Mannings Model

16
17

Creation of Cross Sections


Selection of Mannings Coefficients

17

Method of Creation

18

Gradient

18

Creation of Cross Sections and Normal Discharge Profiles

20

Comparison of Normal Flow Results

28

Critical Flow at Old Bridge

30

Gradually Varying Flow (GVF) Curve

31

Assumptions

31

Calculations

31

Results

33

Effect of GVF Flood Depths

34

Comparison with points of known depth

34

Corrections to the model

35

Additional Models and Conditions


Old Bridge Blocked by timber

36
36

Conclusions

37

References

39

T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 3

Introduction
The city of Bath sits on the River Avon, in the south west of
England. It is a UNESCO World Heritage site and contains
many buildings of architectural and historical significance.
On the December 5th 1960 Bath experienced the highest
flood in living memory. Southgate, Dolemeads and Widcombe were submerged as seen in Figure 1. In response to
this the Greenhalgh flood scheme was commissioned in
1966 and finished in March 1974 (Greenhalgh,1979).
Floods have been recorded In Bath since its founding however the associated damages and costs of have increased
as the city has expanded.

Better prediction through better modelling


In order to find trends in river flow, historical flow data is
necessary in a consistent form. Unfortunately historic flow
data only dates back to 1939, when St Jamess measuring
station was built. Flood marks show the height of historic
floods prior to that period, but no flow data is available.

Intended scope of work


This paper intends to model the Avon and its floodplain in
central Bath. Sections will be taken between Churchill
Bridge and Bath Central Library marked on Figure 2. These
will then be examined based on normal flow and in the
1960 case expanded upon into a gradually varying flow
(GVF) curve for the river.
I will also look at unique cases, notably the 1882 flood
when an arch of Old Bridge was blocked by timber.

Aims and objectives

Figure 1 Swans on Southgate Street [Bath in


Time, 1960]
Bath Library

Pulteney
The Rec

Bath

Dolemeads

Southgate
Churchill Bridge
Widcombe

Figure 2 Map of Bath (Google,2014)

This paper is setting out with the aim of providing a comprehensive comparison between the river in 1960
and 2014. Further to this, It will look into the effect of these changes on the Avons flow. The model created will provide incite into how the river reacts under high discharges in the event of a flooding.
Through generation of a 1960 model, more flooding simulations can be calculated with minimal work, allowing improved estimates for historical flood data.
I also hope to provide a comparison of the positive an negative factors of the different modelling techniques.

4 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Literature Review
Risk of Flooding
The ultimate aim of modelling the river is to provide data in order to better predict future floods through
trends. Historical data is needed to avoid miscalculation of probability, which may occur with a limited
data set, as the period of data available may be flood rich or flood free (Reed, 2001).
At Present there are 1,100 properties at risk of a 1 in 100 year flood in Bath , 1,800 if climate change is
taken into account. (Environment Agency, 2012). This 1 in 100 year flood has been calculated to have a
discharge of 294 m3/s (Jerry, 2014).

Modelling Equations
Mannings Equation
Mannings equation (Equation 1) is commonly used
to calculate average flow velocity in a channel.

In Prague( see Case Studies below) Chows modified


Manning Equation (Equation 2) is used.

Equation 1; Mannings Equation

Equation 2; Chows Modified Mannings Equation

Where

Where

These methods have two main issues when applied to natural channels such as the Avon. The first issue is
Mannings coefficient; which is not constant across the section. The second is that the non-rectangular
nature of the channel requires multiple iterations to be calculated in order for discharge and flow velocity
values to be plotted. The complication of using Mannings equation leads to more simplified sections being assumed.
In order to overcome the Mannings coefficient issue an average Mannings roughness can be used
(Washington State Department of Transport, 2007). Other issues of Mannings equation are due to the
assumptions made, namely that the flow is steady and uniform. Whilst this is a reasonable assumption at
most points, in reality flow varies in time and space. In this regard CH2M Hills ISIS software is much more
proficient.
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 5

Sub-section analysis, Figure 3, is a more accurate form of calculation. This involves;


1) The subdivision of the channel into varying slices dependant on the Mannings number and shape of the bed .

2) Calculating the area and wetted perimeter for each strip.

3) Calculating the velocity and discharge Q,


for each individual section is worked out .

4) Summing the sub-section discharges to


produce a whole river discharge.

Figure 3; Sub-section Analysis


This form of analysis allows for variable flow velocities within the river, as commonly seen in flooding cases. However it works on the assumption that there is no shear stress across fluid boundaries and is dependant on where these boundaries are placed within the cross section. In reality the flow velocity will
not jump but form a smooth D shaped curve across the river width.

The General Approach to Manning Coefficients


A additive Mannings coefficient can be created for a flood plain (Jerret, 1985 & Herget n.d.) using the
General Approach where by a typical Mannings coefficient is taken for the material in the channel (n0).

Equation 3General Approach


Manning's Coefficient

Small additions are then made for; the effect of irregular cross
sections (n1) , variations in the channel along its length (n2),
effects of obstructions (n3) and the vegetative state of the channel (n4). This is then multiplied by a factor related to the
meandering of the river within the valley (m). Giving Equation 3.

Utilising this Mannings coefficient, velocity and discharge can be calculated.

ISIS Modelling
ISIS 1D Flood modeller is software produced by CH2M Hill. It is based on the Mannings equation and the
Saint-Venant equations, amongst other analyses. The Saint-Venant equations allow for the calculation of
storm surges and other non-steady flows.
6 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Previous Case Studies


Prague 2013
In 2013 a study into historical flooding in Prague was conducted by Elleder, Herget, Roggenkamp and Neissen in which they reconstructed peak discharges during flood events in Prague on the river Vltava.
(Elleder, 2013)
The study in Prague had similar issues to those faced in Bath. Prague is a city which had expanded onto
The descriptions of floods have traditionally been described
in relation to Brad (The Bearded Man), shown in figure 4,
close to Karlv most (Charles Bridge). Flood depths were
described in reference to the highest point on his face that
the flood reached, or cubits (590mm) above the top of the
head. This is a good comparison to the Avon where historical
depths are known (both at Pulteney and HaPenny Bridge).
Prague and Bath have a similar development pattern. They
have urbanised the flood plain from a point slightly away
from the river and therefore constrained the river to its banks over time. The buildings providing an area
of minimal flow due to the near infinite roughness, Elleder et al. have excluded sections of the city where
the roads run perpendicular to the river for this reason. A similar consideration must be taken in Bath, in
particular the areas of Widcombe and Southgate.

Figure 4; Bradc (L.Elleder, 2013)

There are several weirs and bridges in the Vltava, both upstream and downstream of the data point.
These are important as they act as bottlenecks and form backwater curves behind them. Critical flow is
known to be reached over the top of these weirs allowing gradually varying flow curves to be calculated.
Unfortunately in Bath the only points at which the river was constricted in this manner are Old Bridge and
Pulteney Weir.
The Prague reconstruction process was done from historical sources including maps, etchings and drawings (L. Elleder, 2013) and as such act as proof that this can be repeated in Bath. A general approach
Mannings coefficient was used in Prague to calculate the flow and velocity.
Some winter floods have been excluded in the study due to blockages from ice flows, this will not be taken into account in Bath due to the temperate climate, however the 1882 flood should be discounted as a
similar effect occurred due to escaped timber, from a yard which no longer exists.

The River Ahr, 2014


In 2014 a case study was published on the River Ahr in West Germany by J.Herget.
Whilst similar to Prague there are a few differences with the River Ahr case study (Herget, 2014).The primary difference is the use of multiple measurement points along the river which allows the flow curve to
be examined along the length of the river and validate the results. (Herget, 2014). This is appropriate for
Bath as two flood depths are known along the Avon. This allows for results to be checked in Bath and
gives the possibility to look at flood surges and variable discharge over time.
In the River Ahr, Herget and Roggenkamp took both an approximated mean roughness and then a upper
and lower bound for each of the following four areas; non built up, northern river bank, river channel and
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 7

southern river bank. They also had approximated depths of two floods from photographs of a flooded
road tunnel. From this they calculated the discharge. Removing areas of the urbanised floodplain from the
model has been done in line with Prague.
The model was then confirmed by the readings of peak discharge. Although a 5 -7 % underestimate was
found, this was considered to be acceptable.

The Avon River in Bath


The Avon catchment is mainly rural above HaPenny Bridge, but includes the city of Bath and the towns of
Chippenham and Malmsbury. There are also several large villages and towns along the Avon, including the
frequently flooded Bradford-upon-Avon.

Figure 5 Avon Catchment for Bath ultrasonic (National Environmental Research Council, 2012)
The catchment above the Bath ultrasonic measuring station, 2012, is 3.2 % urban 11.2% Woodland 33.5%
Arable/Horticultural and 46.1% grassland. (National Environmental Research Council, 2012). The rural nature of the area should attenuate and delay the peak discharge following rain and avoid flash flooding.
This allows the utilisation of photography in this project as discharges will not have been rapidly fluctuating.
As Shown in Figure 5, there is an area of high rainfall in the south of the catchment as Atlantic air passes
over the Mendips , giving SAAR (Seasonally Adjusted Annual Rate) values up to 1200mm per annum although the majority of the catchment receives between 650mm and 850mm annually.

History of Development in Bath


Roman
Bath was founded, as Aquae Sulis, in AD60 when the romans built baths and a temple in the Avon valley.
This led to the first bridge over the Avon in this area, thought to be near the site of Cleveland Bridge, to
the North East of Bath (Buchanan, 1989). A bridge may have been built near the site of Churchill Bridge,
as seen from the street layout, however no roman masonry has been found (Buchanan, 1989).
8 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Medieval
As Bath grew it began to establish itself onto the east bank of the river and mills also began to appear on
the south bank in modern day Widcombe. The first Old Bridge, at the end of Southgate, was built in
1340. Prior to this it is a believed that a wooden beam and post structure may have existed. A weir at
Pulteney is also shown to have existed (Guilmore, 1717).

Georgian Period (1714-1830)


The Georgian period saw huge growth in Bath. The population boomed from 2,000 to over 30,000 people
in the city. The spa town grew into a popular resort for the rich and wealthy. This period also saw a lot of
bridge building, including New Bridge (1754), Pulteney Bridge (1774) and Cleveland Bridge (1827). Old
Bridge was also replaced in 1754 (Buchanan,1989). The city expanded onto the eastern edge of the river.
It should be noted however that the Avon was still mostly unconfined on its south and west banks, with
the exception of Widcombe and Pulteney (Masters,1808).

Modern Period (1830-2015)


Despite Baths famous Georgian
period coming to an end, continued growth has caused the river
to become more enclosed by the
city developments have occurred
along the length of the river, particularly along Upper and Lower
Bristol Road. A lot of this development started out as industrial,
including Bath Printworks, the
Bath 1694
Gas Works and the Weston Industrial Estate .In the last ten years
Bath 2014
there has been a trend of conFigure 6 Size of Bath in1964 and 2014 based on Gilmores Map of Bath
verting this land into housing
(Guilmore,1717)
and blocks of flats, including the
new Riverside development;
this increases the impact and costs of flooding in western Bath. Figure 6 shows Baths enormous growth
over the Georgian and modern periods (1694, black, to present, green).
Bath and North East Somersets (BaNES) population has continued to grow at a steady rate, Figure 7. The
population of City of Bath was 88,859, as of the census in 2011. (GB Historical GIS, 2014) This has led to
an increase in urbanisation and gentrification of the centre of Bath.

T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 9

Population (thousands)

200
180
160
140
120
100
80
60
40
20
0
1800

1850

1900
Year

1950

2000

Figure 7 Population of BaNES 1800-2015 (Bath and North East Somerset Council, 2015)
As bridges are built they tend to constrict the river through their abutments and piers, this was particularly apparent at Old Bridge (1340-1965), which had five piers. Bath now has 16 bridges over the Avon between the Twerton Sluice Gates and the upstream limit of the 1960s scheme, below Cleveland Bridge
(Buchannan, 1989), with those in central Bath shown in Figure 8 below.

(1)

(2)
(12)

(3)
(4)

(11)

(5)

(6a

(6b)

(10)
(8)

(9)

(7)
1)

Winsor Bridge

5)

Midford Road Bridge

9)

HaPenny Bridge

2)

Twerton Bridge

6)

Churchill Bridges

10)

St Jamess Bridge

3)

Victoria Bridge

7)

Old Bridge

11)

North Parade Bridge

4)

Green Park Bridge

8)

Skew Bridge

12)

Pulteney Bridge

Figure 8 Map of the bridges of Bath (Buchannan,1989)

10 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

History of Flooding
Historical flood marks are recorded on the abutment of HaPenny Bridge, a footbridge between
Widcombe and Bath Spa railway station. These flood marks date from March 1867 through to
December 1960, with serious floods in 1882, 1894 (two events), 1947 and 1960, these have been
summarised below in Table 1.
Table 1 Bath Flood Records (Greenhalgh 1979)

Year

Month

1809

Depth above crest of


Pulteney Weir.
imperial
m
126

3.810

1823

November

133

4.038

1866

January

83

2.515

1867

March

80

2.438

1873

March

36

1.067

1875

July

46

1.372

1875

November

79

2.362

1877

November

66

1.981

1882

October

126

3.810

1888

November

73

2.210

1889

March

96

2.896

66

1.981

1891
1894
1932

May

1933
1947

March

910

2.997

Discharge
cusecs

m3/s

12,950

366.7

12,063

341.6

8,300

235.0

7,270

205.9

10,017

283.6

Proposed Flood Protection Schemes


Increased level of development has meant removing the natural flood plain and confining the
river to its channel. In turn this has led to higher water levels, faster reaction times, and greater
costs when the river floods. In reaction to this several flood prevention schemes were proposed
historically, including schemes in 1824, 1882, 1896 and 1936 as detailed in Table 2.
Bath and North East Somerset Council have recently proposed an additional scheme to help protect the city down stream of Churchill Bridge. Proposals include a conveyance strip parallel to
Green Park Road, the lowering of Green Park in order to create a floodplain and general improvements to the riverbank as far as the Bath Riverside Development (Black and Veatch, 2013).

T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 11

Table 2 Summary of Flood Protection schemes (Greenhalgh 1979)


Lead Engineer

Year

Estimated Cost
(adjusted)

Key Aspects

T. Telford

1824

50,000 (3.8m)

Single span replacement of Old Bridge


Cuts and channel improvements

A. Mitchell

1877

13,200 (1.0m)

Replacement of weirs at Pulteney, Twerton Kelston,


Saltford, Swineford and Keynsham with batteries
of Iron Hatches
These act like weirs normally but could be raised during flooding
Each at a cost of approximately 2,200

Code, Son &


Matthews

1882

106,545 (8.9m)

Deepening the river between Bathhampton lock and


Kelston Lock
Lowering and installing sluices at Pultney Weir, Upper Twerton Lock and Kelston Lock.
Sheet piling, widening and deepening Weston Cut.
Construction of another bridge over the Weston cut

G. Remington

1896

69,300 (6.4m)

Diversion of Floodwater via a five mile long tunnel


from Limply Stoke(seven miles up stream) to
Twerton
Construction cost offset by sale of Bath stone

H. Mercer

1936 1953

760,000
(14.8m)

Predecessor of the 1960 Greenhalgh Scheme


Similar to Code son and Matthews scheme
Dredging of channel

Actioned Flood Defence Proposal The Greenhalgh Scheme


Following the 1960 flood new protection measures were built under the supervision of Frank Greenhalgh,
the chief engineer, and author of a paper on the flood protection scheme. The scheme radically reduced
the risk of flooding in the city to a 1 in 450 year flood (Jerry, 2014).
The scheme included; sheet piling along the sides of the channel in the centre of Bath, the redesign of
Pulteney Weir, new sluice gates at Twerton and the removal of Old Bridge (Greenhalgh, 1979). The
Greenhalgh scheme was designed to a physical model at a 1:240 plan scale and 1:60 vertical scale. This
allowed pieces to be replaced to predict the effects of future works. A physical model approach was used
as this study predated computer modelling.

Pulteney Weir
A weir at Pulteney has been recorded since the 17th century (Davis, 2006) when it ran diagonally across
the river. Under the Greenhalgh scheme , Pulteney Weir was reconstructed as per Figure 9. It is now a
horse shoe in plan with a sluice gate on the east side. The sluice gate is designed to open when the river
level exceeds 1.5m above the top of the weir (Greenhalgh, 1979).
The weir was chosen over a quarter circle by an arts commission, and has been awarded the Civic Trust
Award.
12 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Figure 9 Greenhalgh's Pulteney weir scheme (Greenhalgh, 1979)

Sheet Pile River Wall


Below Pulteney Weir the river bank has been subject to new concrete walls on the east bank and a concrete embankment on the west side. Below St Jamess Bridge there has been large scale use of sheet piling; this retains the bank and allows for a greater depth of flow to occur(Greenhalgh,1979).

Replacement of Old Bridge


The replacement of Old Bridge was a critical part of the scheme. Old Bridge had previously been an obstruction in the event of flooding, especially when it became clogged with debris. The most prominent example of this occurred when timber escaped from a yard upstream and blocked the arches in the flood of
1882 (Bath in Time, 1882).
In order to solve this issue Old Bridge was replaced by the Churchill Bridges, one road bridge to the west
of Old Bridge and a foot and services bridge to the east. These bridges are single span and whilst the channel was slightly narrowed for their construction, there is no risk of blockages and the narrowing has less
effect on the flow velocities and discharges of the Avon (Greenhalgh, 1979).

Twerton Works
At Twerton two new radial sluice gates were put in to replace the old system of weirs. This allows for the
flow rate to be altered and moves flooding out of the city centre and onto the industrial and greenfield
land to the west. The river was also drastically straightened to increase the flow rate through this area, as
can be seen in Greenhalgh's plan, Figure 10 below.

Figure 10 Greenhalgh's Twerton Scheme (Greenhalgh, 1979)


T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 13

Normal Flow Spreadsheet


For the basic analysis a spreadsheet was created in Microsoft Excel. This sheet took the cross sections and
calculated area, wetted perimeter, normal flow velocity, normal discharge and critical flow for a range of
depths. From this it is possible to follow the depths of flows upstream presuming hydraulic continuity
holds.

Calculations
Area and Wetted Perimeter Each sub-section creates a quadrilateral defined by four points; Bottom Left
(BL), Bottom Right (BR), Top Left (TL) and Top Right (TR). A Water Surface Level (L) is also defined for each
case.
There are then seven cases possible for each sub-section as defined in Table 3. Through use of a nested
IF functions the sheet decides on the case and defines an area and wetted perimeter for each subsection. The sum of these equals the value for the entire section. The wetted perimeter values have also
been calculated locally for the bed and the bridge for use in the calculation of velocity due to their differing roughness's.

Table 3 Calculations of Area and Wetted Perimeter for any subsection


Case

Water Level

Area

Wetted Perimeter

Xtl ,Ytl
Xtr ,Ytr

Xbr ,Ybr
Xbl ,Ybl

14 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Case

Water Level

Area

Wetted Perimeter

Hydraulic Radius has been calculated locally to each subsection, using Equation 4.
Where:

A = Area

RH = Hydraulic radius

P = Wetted perimeter

Equation 4 Hydraulic Radius Calculation


Velocity has been calculated locally to each subsection and as a, more general, weighted Mannings coefficient case, using Equation 5. This equation weights the Manning's coefficient based on that of both the
bridge and the bed for each subsection.
Where:
P = Wetted perimeter

RH = Hydraulic radius

n = Mannings coefficient

S = Slope

Equation 5 Flow Velocity Calculation


Discharge: has been calculated through Equation 6 for both cases.
Where:

V = Velocity

Q= Discharge

A = Area

Equation 6 Discharge Calculation

Limitations
1)

The spreadsheet can only solve the area beneath a series of straight lines. This lead to approximations being made about the shape of the cross section.

2)

The spreadsheet requires identical X coordinates for both the bridge and bed in order to calculate
area, these can be generated through interpolation.

3)

Perfectly vertical and horizontal lines cannot be used, as they cause division by zero errors within
the spreadsheet. This was solved by the use minute (<1x10-9m) alterations to coordinates.

4)

The connections between Bridge Piers were automatically added to the wetted perimeter. This was
solved by manually changing the wetted perimeter to zero on a case by case basis.
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 15

Comparison of Sub-section & Weighted Mannings Models


Two methods of calculating the discharge have been used: Sub Section and weighted Mannings.
The benefit of sub-section analysis, as discussed in the literature review, are easily seen especially at
Pulteney Weir and Parade Gardens. In these sections the mean flow velocity decreases as the depth increases beyond a flood threshold height as shown in Figure 11. This locally caused deceleration then decreases the discharge of the entire section of the weighted Mannings profile, Figure 12.

Figure 11 (left) Average velocity across CS3 (2014) for varying depths and Figure 12 (right) the effect of the
velocity decrease on the discharge of the weighted Manning model (dashed) as opposed to the sub section

This is clearly a incorrect, the normal discharge will only increase as depth increases, with the exclusion of bridges and overhangs. The use of sub-section analysis helps minimise this error.
The actual case is of course more complicated than this, with bridge sections being a prime example. Here the flow does not vary across the width of the channel so much as it contours away
from the friction surfaces, as shown in Figure 13. For this contour model laminar flow is presumed, Mannings equation requires the presumption of turbulent flow.

Figure 13: Illustration of flow velocities across a channel in a weighted Manning model (left) ,a sub-section model
(centre) and real life (right)

16 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Cross-Section (CS) Generation


The river has been split into nine cross sections between Bath Library and Churchill Bridge, each one of
these cross sections has been chosen for its position across an extreme point, be it narrow and constricted, or on a wide flood plane. The nine cross sections chosen were : 1) Bath Library 2) Pulteney Bridge 3)
Pulteney Weir 4) Parade Gardens and The Rec 5) North Parade Bridge 6) St Jamess Bridge 8) HaPenny
Bridge 9) Skew Bridge 10) Old/ Churchill Bridge. These are shown below in Figures 14 and 15.

CS1

CS1
CS3
CS2

CS3

CS2

CS4

CS4
CS5

CS5

CS6

CS9
CS8

CS7

Figure 14 Map of Cross Sections 1960


(Ordinance Survey, 1960)

CS9

CS6
CS8

CS7

Figure 15 : Map of Cross Sections 2014


(Google, 2015)

The cross section profiles have been created through the analysis of historical photos, maps and surveys.
The BaNES ISIS model (BaNES,2014) and the 1954 Cross Sections (Bristol Avon River Board, 1954) were
the most commonly used as they offer standardised information on the river section. Any discrepancies
were amended In line with historical photography from Bath in Time and investigation of the river itself.

Selection of Manning's Coefficients


Manning's coefficients have been selected according to normal values as stipulated in Ven Te Chows
Open Channel Hydraulics (Chow, 1959). These values have been chosen based on known factors of the
channel and their banks, and are explained individually for each of the cross sections. Typical values used
in the model are given in Table 4, overleaf.
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 17

Table 4 Commonly Used Mannings Coefficients


(Chow 1959)
Surface

Mannings
Coefficient

Clean Channel Bed


Dredged Channel Bed
(after weathering)

0.030
0.022

Sheet piling

0.022

Pointed Masonry

0.015

Rough Masonry

0.025

Trowel Finished Concrete

0.013

Short Grass

0.030

Dense Brush (Summer)

0.100

Dense Brush (Winter)

0.070

Trees (below foliage)

0.100

In the ISIS model used by BaNES, a Mannings


coefficient of 0.040 has been used uniformly for
the bed and a value of 0.070 has been used for
vegetated flood plains. These values, whilst
generally correct, are broad approximations. As
such the model created for this paper has
aimed to improve on these values where supplementary information exists. It should be noted than sub-sections have been defined by a
change in Mannings coefficient as well as a
change in shape for a channel.
For foliage and vegetation the winter value has
been used in the 1960 model, as the flood was
in December. For the 2014 model the maximum
of the two values has been taken, typically this
is the summer when the foliage is thicker. This
has been deemed acceptable as floods have
been known as late as May (1932) and we are
looking for a worst case scenario.

Method for Creation of Cross Sections


The limitations of the basic analysis solver led the design of the Cross Sections. The sections were created
in AutoCAD, to start with all known values were imported into the drawing. Any gaps in information usually utilised from 3rd party cross sections was then put in from the Bath In Time photography. This information is less accurate but vital for the creation of the model. Any floodplains that fell were not visible in
the historic photography were modelled using information drawn from maps.
When the sections were completed in AutoCAD, the X coordinates of significant points were taken from
each cross section and the coordinates from both sections were taken and entered into a spread sheet to
be analysed.

Calculation of Gradient
The solver only calculates Steady Uniform Flow, this is an issue for this section of the Avon, as at points,
the gradient is adverse. In order to approximate the flow a least square regression trend line (LSR) can be
drawn through the known depths, as shown in Figure 16 and 17, and a slope derived from this. In order to
best approximate the actual gradient of the river. Four methods were compared:
1) LSR of lowest point of the Bed [L1]. This has issues with local deep spots, notably at Skew Bridge. While
the river may typically be shallow, this is an extreme case and does not represent the overall gradient
properly.
18 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

2) LSR of the average bed height [av.]. Here the opposite Issue was found , shallow sections that were
modelled with their flood plains were taken as much higher than they realistically were under normal
flow conditions.

Figure 16 2014 Gradient Calculations

Figure 17 1960 Gradient Calculations


3) LSR of the average depth of the bottom 10 values of the bed height [L10]. This was to take a more realistic typical bed height, excluding the flood plains on either side and keeping the gradient to the channel.
4) The Gradient taken by BaNES, 1:1800. This gradient models a longer stretch of the river as opposed to
the steeper section seen in central Bath.

Table 5 Gradient Results


2014
BaNES ISIS 0.00056
L.1 LSR
0.00211
L.10 LSR
0.00193
Av. LSR
0.00248
Value Taken 0.00217

1960
N/A
0.00271
0.00241
0.00201
0.00237

In Both 1960 and 2014 the gradients of all three calculated values
were relatively similar, as seen in Table 5. An average of these three
values was used in the model, as it was the best compromise. The
BaNES gradient is a quarter of the slope of the other gradients and as
such it has been discounted.

T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 19

CS 1 : Bath Library & Waitrose


1960: Figure 18 shows the west bank to be lightly vegetated and at approximately

25 degrees. There is

a stone wall which has a Manning's coefficient of 0.017 . The cross section has been ended at buildings at
the top of the bank.
The east bank has a short wall made from sheet piling followed by an approximate 2m flat bank and then
a vertical masonry face. The Manning's coefficient of the bed has been taken as natural . There is no
floodplain so the section has been ended at the height of the opposing bank.

Figures 18 Photography of the North Side of Pulteney Bridge Prior to Greenhalgh scheme (Bath in Time,1960
and Bath in Time, 1972)

2015: The main difference between 1960 and 2015 is the foliage and confinement of the west bank. This
now has buildings closer to the river, Excluding this the river has remained unchanged.
The discharge results, Figure 19, for this section show little differentiation between 1960 and the modern
era. The main changes in discharge here are caused by the seasonal variation of foliage roughness especially on the west bank.

Figure 19 Discharge Vs Depth For CS1 superimposed over the section (not to scale horizontally)
20 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

CS2: Pulteney Bridge


1960: The channel has traditionally been shallower than
it is today. In the 1960s it was undredged and as such
had a higher Mannings coefficient. The river bed has
been created by scaling Figure 20. The Bridge itself is
made from Bath stone, a local oolitic limestone and
pointed with lime mortar. The original piers were also
made from the same material.

2015: As part of the Greenhalgh scheme , the river has


been deepened by dredging . In order to facilitate this
the western pier has been underpinned in concrete
(Greenhalgh, 1979). The bridge arches have remained
the same.

Figure 20 Cross section of Pulteney Bridge


(Adams, 1768 )

The shape of the discharge is as expected, given similarities to open flow in a pipe. The optimal normal discharge is found at 23.44 metres over Datum (mOD) for
the 2014 sub-section analysis. This, along with the shape of the curve, is consistent to that of a critical discharge of a pipe, considering the increased roughness of the base of this section.
Figure 21 shows considerable differences in discharge occurring here due to the channel dredging, showing the effectiveness of this strategy. At peak discharge the difference is around 350 m3/s, similar to an
additional 1888 flood.
Pulteney Bridge can be seen to have minimal problems with flood discharges, typical flood cases of
around 350 m3/s do not even come to the arch springs.

Figure 21 Discharge Vs Depth for CS2 superimposed over the Section (not to scale horizontally)
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 21

CS 3 : Pulteney Weir
1960: In 1960 the weir ran diagonally across the river from the corner
of parade gardens as seen in Figure 23. The weir was 17.02 metres
above ordinance datum (mOD) and flanked by a masonry wall.
On the west bank a masonry wall goes to 19.8 mOD before becoming
the columned Parade as shown on the left of Figure 22. This has been
modelled through an increased roughness at this point, (N=0.1) based
on the roughness of a stand of trees.

2015 The weir has changed dramatically since 1960. It has been rede-

Figure 22 The Parade (Wilkinson,


2011)

signed as a horseshoe and lowered. On the east side of the river there
is a sluice gate separated from the main channel by a wooded Island. In
order to keep the model consistent the section steps back upstream to
the same point at which the 1960 section stops. A radial sluice gate,
was added as part of the Greenhalgh scheme. The sluice gate is designed to be raised when the river overtops the weir by more than
1.5m. The model is being used to calculate worst case flooding situations and as such the spread sheetcalculates flow for the sluice gate has
Figure 23 Pulteney Weir in 1960
been fully open.
(Bath in Time, 1960),
The effect of the sluice gate is a apparent in Figure 24, drastically increasing the discharge of the section. Known historical flood discharges (Table 1, Literature Review) would
not require the sluice gate to be fully opened (if the river underwent normal flow at this point). The
weighted Mannings model is clearly affected by the introduction of the main weir and the central wooded
island. Compared to Pulteney Bridge the discharges of this section, compared to depth, are extremely
large especially in the modern section.

Figure 24 Discharge Vs Depth Super for CS3 imposed over the Section (not to scale horizontally)
22 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

CS 4 : The Rec, and Parade Gardens


1960: In 1960 The Rec was largely undeveloped, Figure 25, this allowed
for water to move across it. On the east bank there was a short masonry wall topped with a grassy slope and a line of trees, as seen in Figure
26.
On the west side, Parade Gardens were protected by the height of the
bank, which at this time was heavily vegetated.
Figure 25 Arial Veiw of
CS4 ,1971 (Bath in Time,1971)

There is no data for the river bed in 1960 at this location and as such
the 2014 section was taken as a best approximation.

2015: A secondary flood wall has been built to protect The Recs rugby pitches and the leisure centre, this has confined the river to its
channel and onto the promenade. The river wall in itself has been
changed and the slope is now steeper and constructed from rough masonry. The flood defence wall has been modelled as a rise in floodplain.

On the west side there is a grassy bank topped with vegetation. Above
this Parade gardens act as a flood plain. The Parade Gardens are mostFigure 26 View North from North ly grass but with some dispersed trees (n= 0.030). There is a vertical
masonry wall at the far side. Further downstream there is a flood wall
Parade Bridge in 1964
on west bank however this is not present at this cross section.
(Ollis,1964)
The main difference in the sections lies with the flood protection of The Rec. In 1960 the utilisation of this
area allowed huge discharges, as shown in Figure 27, to move slowly across this plain . The Rec also acted
to store excess water and attenuate peak flows. The changes to Manning's coefficients have had little
effect to the flow in comparison to the loss of the floodplain.

Figure 27 Discharge Vs Depth Super for CS4 imposed over the Section (not to scale horizontally)
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 23

CS 5 : North Parade Bridge


1960: Whilst the bridge was originally a cast iron structure but in 1937 it was clad in masonry . The underside of the Bridge is very smooth masonry whilst the edges have a higher roughness. The bed had not
been dredged as of 1960, and is considered clean and straight. (Bristol Avon River Board, 1954).

2015: There have been very minimal changes to the cross section of the river here since the implementation of the Greenhalgh scheme. However the river bed has been dredged and smoothed, changing the
shape and roughness (BaNES, 2014).
The discharges are comparably much lower at this point compared to upstream. This is due to the confined nature of the bridge. The ramps up onto this bridge prevent water flow around the ends, and as
such the bridge acts as a regulator to the flow on the floodplain, enhancing the attenuation properties.
The adjustments to the bed have negligible effect on the flow through the bridge as can be seen in Figure
28.

Figure 28 Discharge Vs Depth Super for CS5 imposed over the Section (not to scale horizontally)

CS 6: St Jamess Bridge
1960: In 1960 the main water way under St Jamess Railway Bridge was beneath the main arch . The side
arches were used primarily for footpaths on the banks, but also as flood relief. The arch is rough masonry.
As of 1960 the bed was natural.

2014: As part of the 1960s flood protection scheme vertical sheet piling was put in underneath the piers.
The channel was dredged and widened, with sheet pile walls supporting the banks. These sheet piles have
a Mannings coefficient of 0.022. The sunken boat at this location has not been modelled.
Figure 29 shows little variation in the discharge at this location between 1960 and 2014. The opening up
of the side arches whilst expensive has added little capacity compared to the central section. In reality this
is more for the benefit of the channel widening immediately upstream of the section.
24 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Figure 29 Discharge Vs Depth Super for CS6 imposed over the Section (not to scale horizontally)

CS 7: Hapenny Bridge
HaPenny Bridge is a vital section for this project as this is where the historic water marks are engraved.

1960: The cross section of the bridge has remained unchanged. However there used to be a tow path at
approximately 17.5m, this confined the river horizontally. The river also used to be very shallow and high
(Bristol Avon River Authority, 1959). The river walls and bridge piers were made of smooth masonry. Historically a large amount of residential property backed on to the river in this area, with a floor level of
around 17-18m.

2015: As part of the Greenhalgh scheme extensive sheet piling was put in on the southern bank between
St Jamess Bridge and Skew Bridge. There has also been sheet piling of the northern bank, but to a lesser
depth. The river has been dredged, deepening it and reducing the Mannings coefficient. The residential
area on the south bank has been demolished and a second river wall constructed as part of the A36 realignment.
The effect of deepening the river on discharge can clearly be seen in Figure 30, increasing the discharge
considerably at this section. The nearly identical shapes of the curves shows other changes are insignificant to the section.

Figure 30 Discharge Vs Depth Super for CS7 imposed over the Section (not to scale horizontally)
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 25

CS 8: Skew Bridge
1960: The river was confined by a pointed masonry
wall on both sides (Figure 31) The bed is a natural
channel.
The Bristol Avon River authority drawing, Figure 32,
show a localised area of deep water to the north of the
central pier, its proximity to the pier believed to be an
effect of scour.
Whilst this section has been used for the bed, photography clearly shows the banks much more confined
than in this section. It is believed that the section runs
perpendicular to the river and as such is not constrained by the continued railway viaduct. This has
been amended in the model cross sections.

Figure 31 looking East at Skew Bridge (Bath in


Time,1960)

2014: The southern banks masonry wall has been replaced by a deeper sheet pile wall. Dredging has
evened out the channel bed protecting the central pier
and changing bed roughness (Greenhalgh,1979 and
BaNES,2014).

Figure 32 Bathymetric section (Bristol Avon River


Board 1954)

Minimal changes to normal discharge have been observed in Figure 33. The difference between the
weighted Manning's and sub-section models seen in Figure 33 are mainly due to friction on the vertical
faces, in this case sub-section analysis will clearly be an over estimate due to the large vertical faces are
having little to no effect on the overall section velocity. This is due to the sub-section adjacent to the face
being of minimal area. This is a disadvantage of the sub-section model type.

Figure 33 Discharge Vs Depth Super for CS8 imposed over the Section (not to scale horizontally)
26 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

CS 9: Old Bridge/ Churchill Bridge


1960: Old Bridge was a five arch span, each arch had a
span of 5.5 to 6 metres with an approximate 2 metre wide
pier. The bridge was made from pointed Bathstone and
the bed was as of its natural roughness. In order to model
this bridge two models were produced. Originally the
model up to the level of the highest, central soffit was
made, the 1960 flood verged on over topping the bridge .
As such a second model was built including the flood plane
to either side and the top of the deck, as seen in Figure
Figure 34 Old Bridge 1964 (Bath in Time, 1964)
34, a Mannings roughness for asphalt (n= 0.017) was
used across this surface. This allowed the modelling of the full depth of flow.

2014: Damage occurred to the arches of Old bridge in the 1960 flood (Greenhalgh, 1979). As extensive
repair was already necessary, for hydraulic purposes, Old Bridge was removed and Churchill Bridge was
built 49m downstream. This new bridge was built as a single span to avoid the problems previously faced
with Old Bridge and the bed was cleaned and dredged as part of the works. In order to facilitate an economical bridge the river was narrowed at this point through large masonry river walls. Given the relative
discharges of the bridges and the known height of the 1960 flood the top of Churchill Bridge was not modelled.
Figure 35 clearly shows the difference between the two bridges. It is important to note the larger discharge at lower depths, as this is where flooding is prevented. Of the known flood discharges (Table 1)
three of five are greater than the maximum, normal, under-arch discharge (239m3/s) for Old Bridge. The
narrowing of the river at Churchill Bridge means this area can still experience constriction but the effect is
minor compared to that of Old Bridge.

Figure 35 Discharge Vs Depth Super for CS9 imposed over the Section (not to scale horizontally)
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 27

Normal Flow Results


Figures 36 and 37 have been produced to show how discharge curves change along the length of the river.
These give clear indications of the constrictions on the flow. The effect of flood planes and confined channels can also be seen here through the distribution of lines for each section.

Old Bridge: Figure 36 clearly shows the effect of Old Bridge. A 1 in 100 year flood has a discharge of
294m3/s (Jerry,2014) , Old Bridge has a maximum normal flow discharge of 238 m 3/s through its arches,
beyond this the water level is forced to rise rapidly onto the surrounding area.

HaPenny Bridge: Figure 36 shows a rise over HaPenny Bridge in particular with flows below 238 m3/s. In
1960 this rise in bed height was obviously creating a backlog in the flow. A comparison between Figures
36 and 37 shows the dredging and sheet piling have had a significant effect on the flow of the river, typically lowering the surface by approximately a meter for equivalent discharges.

Skew & North Parade Bridges: Issues can be seen at these locations during periods of extremely high
flow rate, in Figure 37. However given the upstream channels of both these points minimal, damage will
occur even under these conditions.

Pulteney Weir Sluice Gate: Figure 38 illustrates the effectiveness of the Pulteney Weir Sluice Gate, It
means the weir is no longer critical to flow height as it can be seen to be in 1960 (Figure 36). A better
model of actual flow depths at this location, with a variable sluice gate, has been provided in Figure 38.
This model shows how the weir has little effect on the flow level upstream when the sluice is in operation.

Figure 36 Long section of normal flow water surface height for varying discharges in 1960
28 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Figure 37 Long section of normal flow water surface heights for varying discharges in 2014

Figure 38 The effect of modelling a variable sluice gate (Solid) vs modelling a fully open sluice gate
(Dashed) on the water height at Pulteney weir.
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 29

1960 Flood Conditions


Old Bridge at Critical Flow
Given its position as the choke point in the flow of
the Avon it would be expected that critical flow
would be found under Old Bridge during a flood. It
has been presumed that it acts in a similar manner
to a Venturi flume.
From photography (Figure 39) the level is of the
flow known under Old Bridge, on the day of the
floods, to be 16.5moD (Figure 40). Under the assumption the flow is critical, it is possible to calculate discharge using Equation 7.

Figure 39 Old Bridge 5/12/1960 (Whiteway, 1960)

Figure 40 Flow Height of 16.5m imposed over CS9 1960 cross section
Equation 7 Calculation of Critical Discharge
Where
Q: Critical Discharge

g = 9.81m/s2

A: = area of Flow

B = free water surface

The model generated for normal flow was used to find values for area of flow (85.02m2) and free water
surface (10.13m), this gives a critical flow of 772 m3/s, far more than that of the normal discharge at this
depth (192m3/sweighted Manning's) or (237m3/ssub section).
The depth on the upstream face of the bridge is unknown, however the depth is known further upstream
at HaPenny Bridge (Figure 37). Between the known depth and known flow a gradually varying flow model
can be produced. It should be noted no flow across the flood plane has been accounted for given the water level seen in Figure 39.
30 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Gradually Varying Flow Profiles


Assumptions
In order to create a gradually varying flow model there are several assumptions that need to be made:
1) A uniform channel roughness; this has been completed by creating a length weighted average of the
Mannings coefficient for the entire the cross section.
2) Typical section; previous cross sections have been designed to test constrictions and extreme cases.
This is inappropriate for the calculation of gradually varying flow as it models a very constricted channel,
with the exception of CS4. In order to approximate the channel, secondary sections have been created for
Old and Pulteney Bridges, without piers in the bed. All other cross sections at bridges have had the bridge
arch removed.
3) Prismatic channel; the channel has been modelled as a constant cross section for the length of the section (the distance upstream to the next cross section).
4) Given the lack of an upstream section, CS1 has utilised the average lowest point LSR gradient, as calculated for the normal flow model (S0 = 0.00201).
5) The gradient from Parade Gardens to Pulteney Weir has been taken as to be the same as from North
Parade Bridge to Parade Gardens. The profile of this section is that of a mild river bed with a sudden drop
at the weir.

Calculations
Table 6 Free Surface Calculations
Case

Free Surface

2
3

The free surface can be found for each sub-section


using the same seven cases as in the normal flow
calculations, see Table 6, the sum of these then
give the free surface for the entire river. The critical flow can then be calculated using Equation 7
on the previous page.

Slope Calculations

In order to calculate the GVF profile, the direction


from which the flow depth is being driven needs
to be known. This can be derived from the depth
of the normal and critical flow and the known
height of the flow at any point.

5
6
7

Free Surface Calculations

A summary of these calculations has been produced in Table 7, with a green line representing
the location of the known depth at HaPenny
Bridge.
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 31

Table 7 Summary of inputs into the gradually varying flow model


X1 (m)
CS9a to CS9b
40
CS9b to CS8
52
CS8 to CS7
147
CS7 to CS6
311
CS6 to CS5
510
CS5 to CS4
804
CS4 to CS3
910
CS3 to CS2
1064
CS2 to CS2b 1112
CS2b to CS1 1130

X2 (m)
52
147
311
510
804
910
1064
1112
1130
1232

Z1 (m)
12.04
12.01
11.45
13.63
11.26
12.31
12.53
17.6
13.52
13.72

Z2 (m)
12.01
11.45
13.63
11.27
12.31
12.53
17.6
13.52
13.73
14.86

S0
-0.00293
-0.00587
0.013281
-0.01189
0.003564
0.002086
0.032922
-0.085
0.011183
0.011183

Y0 (m)
N/A
N/A
6.05
N/A
6.91
7.46
6.37
N/A
4.84
3.87

Yc (m)
4.46
4.74
7.46
5.57
6.01
4.90
4.72
2.71
5.99
5.32

Slope
Adverse
Adverse
Steep
Adverse
Mild
Mild
Mild
Adverse
Steep
Steep

Driven From
Downstream
Downstream
Upstream
Downstream
Downstream
Downstream
Downstream
Downstream
Upstream
Upstream

Profile Modelling Calculations


The energy profile over a short section can be modelled as in Figure 41, from this model Equation 8 can be
derived. Equation 8 allows us to model the variation of specific energy in the flow over distance travelled .
Specific energy is defined in Equation 9, and has units of metres.

Figure 41 Energy over a short flow


Where

g = 9.81m/s2

Y = Depth

E = Specific Energy

v = flow velocity

S0 & sf as above

Equation 8 Change in specific energy over distance & Equation 9 Definition of specific energy
Over short distances Mannings equation is taken to be true. This allows the friction slope, Sf ,to be calculated using Equation 10. Equation 8 can then be put into the finite form and rearranged to give you the
change in X for a given change in Y (Equation 11). An average friction slope is taken between the two
points.

Where

Rh = hydraulic radius

n = Mannings No:

x = horizontal dist.

Equation 10 Friction Slope as defined by Mannings Equation


& Equation 11 Change in X as defined by calculated values
32 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Creation of the Profile


The flow depth on December 5th 1960 is known accurately at HaPenny Bridge (CS7) as shown in Figure
42. From here it is possible to extrapolate flow depths up to CS2 and down to CS8. The flow depth has also
been taken from Old Bridge upwards to CS2.

Y = 7.18m

Figure 42 Cross section, HaPenny Bridge (CS7) 5/12/1960

Results
The results are summarised in Table 8 and the profiles in Figure 43. The results follow typical GVF curves
as expected for the given sections.

CS1

CS3
CS2

CS4

CS5

CS6

Derived from CS7

CS7

CS8

CS9

Derived
from CS9

Figure 43 Gradually Varying Flow Profile


Table 8 Gradually Varying Flow Profile
CrossSection
9
X (m)
40
Depth, Y (m)
4.56
Surface Height (mOD) 16.5

9b 8 (left) 8 (right) 7
52
147
147
311
5.48
6.87
6.50
7.18
17.48 18.32 17.95 20.80

6
510
10.39
21.65

5
804
9.33
21.64

4
910
9.33
21.86

3
2
1064 1112
4.27 8.85
21.88 22.37

The water level in Figure 43 shows the huge change in surface height around HaPenny Bridge, caused by
the raise in bed level. A large proportion of the gain in surface height and depth occurs between CS8 and
CS7, through an S2 curve. As this is driven by the up stream water level this implies that HaPenny Bridge
is the critical section in avoiding up stream flooding.
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 33

This notion is reinforced by the minimal gain in surface height beyond this section and that neither normal
nor critical flows of any upstream sections even approach the surface height at CS7.
Whilst Old Bridge does not seem to be having a large effect on water level, once the area around HaPenny Bridge is dredged, it would rapidly become a problem. If no section at CS7 was included, the river
would be either mild or adverse from Old Bridge to Pulteney Bridge, the depth would therefore be driven
from the downstream end for nearly the entire length of this model.

Effect of GVF Flood Depths


In the city of Bath in 2015, the area marked in green on Figure
44 would have been flooded under the heights found for the
1960 In the GVF Curve model.
This map has been calculated based on topographic data from
Google Earth. The water surface heights of the GVF curve have
then been expanded perpendicular to the river, and where the
land is lower than the surface height it has been marked as
flooded.
Whilst a large proportion of the area covered is playing fields
and will require little repair, Dolemeads and Widcombe Infants
School also fall under the flooded area, and would be far more
costly,
Historically there would also have been flooding damage to
buildings backing onto the river which were demolished for the
A36. There is also photographic evidence that the floods extended into the lower end of Southgate and the river side at
Dolemeads. These locations have been marked in red in Figure
44.

Figure 44 Flood map under GVF Flow


(Google, 2015)

Comparison of the model with points of known depth.


In order to check the model, flow depths are derived at Pulteney weir from photographic evidence
(Figures 45 and 46) using the following procedure:
The distance between the two store room windows is known to be 12.75m, from this the distance from
the soffit of the arch to the water level can be found by scaling off the photograph. Whilst there is a slight
error due to photographic skew this is minimal and has been considered insignificant. The soffit height is
known to be 23.52mOD, and the distance from the soffit to the water surface has been found to be 2.83m.
This gives a surface height of 20.69m and a depth, Y of 5.83m.
This depth is considerably different to that derived from that in the model, this could be due to a number
of factors;

The slope assumed between Pulteney Weir and Pulteney Bridge has been taken as an average slope
as opposed to a sudden rise at the weir. This adverse slope has lead to a sudden increase in depth
and surface height upstream.

34 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

20.69mOD
Y = 5.83m
12.75m
2.83m

Figure 45 Photograph (Bath in Time), & Figure 46 Cross section, Pulteney Bridge (CS2) 5/12/1960

The confined nature of the sections has led to a faster rise in water surface level along the profile
than had the river been modelled using a larger number and more typical sections.

The presumption of there not being energy lost to eddy currents is incorrect and significant.

The non prismatic nature of the channel is significant.

The photograph, whilst taken on the same day as the flooding, may not have been taken at the time
of peak discharge.

Differences in GVF depths at Skew Bridge (CS8)


At Skew Bridge there is a convergence of two separate GVF curves, which have a discrepancy of 370mm.
The curve derived from the depth under Old Bridge is higher than that derived from Ha'Penny, which indicates that any timing issues with the Old Bridge photograph are insignificant.

Corrections to Discharge.
In this model critical flow has been assumed at Old Bridge, this assumption may have been incorrect. The
flow has previously been calculated at 354 m3/s which seems more inline with historical flooding data
(Jerry,2014).
If T. Jerrys discharge is taken, both the normal flow height (16.73 mOD) and critical flow (17.57mOD) at
HaPenny Bridge drop beneath the level of the historical flood mark (20.81mOD). This would cause an S1
curve to develop between HaPenny and Skew Bridge and a more flat profile to be adopted. More over it
now means that the flow curve is being dictated from downstream, at Old Bridge. The effects on the upstream water level will be minimal as the depths found here were already considerably higher than the
normal or critical depths in these locations. changes in surface height are minimal in the critical flow model, these small changes will decrease slightly under this decreased flow.

Corrections to the model


The model has some discrepancies, in order to attempt to rectify these several methods could be employed :

The creation of a larger number of more typical sections. This would be preferable however overly
time consuming as it would require the building of an entirely new model.

Applying a scaling factor either to the total surface height or the change in surface height. A scaling
factor to the total height should not be used as it may cause the model to break hydraulic principals.
Scaling the change in depth through an empirical factor would be preferable, as it would provide a
more realistic curve, consistent with typical GVF profiles.
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 35

Creation of a 2014 model


Due to a lack of a known depth, or a local constriction whereby critical depth could be taken, it is impossible to calculate a gradually varying flow curve. All the sections are either mild or adverse, meaning that the
flow depths for these sections would be driven from downstream. The next known constriction below
Churchill Bridge is Twerton sluice gate, approximately three kilometres away.

Additional Models and Conditions


Old Bridge Blocked by Timber,1882
In 1882 timber escaped from an upstream yard and became
stuck under Old Bridge, as seen in Figure 47, exacerbating an
already bad flood.
In order to simulate this type of blockage the middle arch of
Old Bridge was removed, this has then been compared to the
unblocked bridge for both critical and normal discharges for
given heights in Figure 48.
Normal flow has dropped considerably with the blockage, this is
especially relevant when considered against historic flood data,
where all previous known discharges would now cause flooding.

Figure 47 Old Bridge 1882 (Bath in


Time, 1882)

Figure 48 Discharge vs depth Comparison (blocked vs unblocked)

Critical discharge has suffered minimal effects, this is due to the effect of the decreased area being mitagated by the decreased free water surface. As discussed in GVF critical flow here is unlikely.
It should be noted that this flood should be excluded from any analyses of flood probability or trend due
to its unusual and unrepeatable nature.
36 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

Conclusions
The Greenhalgh Scheme
The Greenhalgh scheme has been shown to be necessary in order to avoid widespread flooding in Widcombe, Southgate and Dolemeads. This has been proved by the levels found from the 1960 gradually varying flow curve. The scheme can be broken down into its most important components:

HaPenny Bridge Channel Depth


The bed height at HaPenny Bridge can clearly be seen as the main reason behind the rise in water level in
the Avon through Bath. This is best shown through the critical flow GVF curve, however the normal flow
profiles also confirm this hypothesis. The extensive dredging here under Greenhalgh has removed this
blockage and associated rise in water level.

Old Bridge
The removal of Old Bridge is vital, especially if the lower discharge is used in the GVF model. The bridge
prevents water leaving the city. The maximum normal flow under the arches of Old bridge was 228m 3/s
this is smaller than the discharge for the 1888, 1894 and 1947 floods. Old Bridge traditionally caused
Southgate to flood.
The secondary issue with Old Bridge was its liability to blockage, The blockage of the central arch only was
liable to reduce the maximum normal discharge to 171m3/s , 75 % of the unblocked bridge.

Pulteney Weir
The effects of Pulteney Weir are only clearly seen when the river is low however as the water rises the
effect minimises. The modern sluice gate is very effective at preventing the weir from creating a backlog.

Normal Flow Modelling Techniques


Averaged Slope Weighted Mannings Equation model
An averaged Mannings coefficient provides a quick easy way to estimate flow. However it is subject to
difficulties when dealing with floodplains and localised areas of high roughness.

Averaged Slope Sub-section Mannings Equation Model


Division of the channel into subsections provides better estimates for wide flat varying sections, as usually
seen on floodplains and natural channels . However this model is more time consuming to create and had
difficulties with the effects of vertical elements being underrepresented in the flow.

Recommendations
In future modelling it is recommended that the channel be divided up into large sub sections based on
the nature of the channel shape and roughness at that location. Each of these sub sections should then
use an independent weighted Manning's analysis to calculated flow. This method would avoid both issues
found with flood plains in the weighted Mannings analysis and vertical faces in the subsection analysis.

T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 37

Limitations
These methods, whilst a quick and easy way to gauge where a flow is constrained, do not predict the actual water level and as such cannot be used to calculate the effect of flooding. They are also been based of
an averaged gradient and so ignore the effects of local bed slope on depth.

Gradually Varying Flow Curve


This method is highly dependant on accurate information specific to the flood being modelled; discharge
and surface height. Difficulties in this model arose from an error in calculating the latter.
Other difficulties found whilst creating this model were a lack of points in the Avon where critical flow
must occur in 2014. Locations of critical flow are vital in order for depth to be calculated in the case of predicting the effects of the flow. This makes it very difficult to use the model to predict flow depths in the
future.
Sections generated for the normal flow analyses can be used in the generation of the gradually varying
flow curve, however they should be supplemented with additional more typical sections.

Future Models
I believe that three more models should be created and or analysed:
A gradually varying flow model of the Avon in 2014. This model has already been created in ISIS by BaNES.
In order to gain a comparison between 1960 and 2014 it should be run under the same discharge as the
1960 flood. This would give a better comparison of the effects of the river on the flooding of Bath.
A re-run of gradually varying flow calculations should be done based on a more accurate discharge for the
river on December 5th 1960. This is vital in order to accurately model the flood.
The model should be run for other flood marks and points of known discharge and height. This could be
done with little alteration to the model, as few changes occurred to the river prior to the 1960 flood
scheme, and would produce information of use for statistical modelling.

38 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

References
Greenhalgh, F., 1979. Bath Flood Protection Scheme, Bath: s.n.
Bath in Time. (n.d.). Swans on Southgate Street, Bath during the floods of december 1960. Retrieved 02
12, 2015, from Bath in Time: http://www.bathintime.co.uk/image/1087626/swans-on-southgatestreet-bath-during-the-floods-of-december-1960
Google. (2015). Google Maps - bath. Retrieved 04 03, 2015, from Google Maps: https://
www.google.co.uk/maps/@51.3753717,-2.3797541,14z
Reed, B. a., 2001. The Use of Historical data in Flood Frequency estimation. Wallingford: Centre for Ecolo
gy and Hydrology.
Enviroment Agency. (2012). Bristol Avon Catchment Management Plan. Exeter: Enviroment Agency.
Jerry, T., 2014. Flood Frequency Analysis for the River Avon at Bath, Bath: University of Bath.
Washington State Department of Transport, 2007. Washington State Department of Transportation Hy
draulics Manual. [Online]
Available at: http://www.wsdot.wa.gov/publications/manuals/fulltext/M23-03/Chapter4.pdf
[Accessed 06 12 2014].
Jerret, R. (1985). Determination of Roughness Co-efficients For Streams in Colorado. Lakewood: U.S. Geological Survey.
J. Herget, L. E. H. M. A. N. &. T. R., n.d. Reconstructing Peak Discharges of historic flood levels in Urban Ar
eas. [Online]
Available at: http://tolu.giub.uni-bonn.de/herget/pages/poster/reconstructing-peak-discharges.pdf
[Accessed 10 12 2014].
L. Elleder, J. H. T. R. a. A. N., 2013. Historic floods in the city of Prague - a reconstruction of peak discharge
es for 1481-1825 based on documentary sources. Hydrology Research, Issue 44.2, pp. 202-213.
Herget, T. R. a. J., 2014. Reconstructing Peak River Discharges from Historic Floods of the River Ahr, Ger
many. Erdkunde, 68(1), pp. 49-59.
National Enviromental Research Council, 2012. Centre for Ecology & Hydrology - 53022 Avon at Bath Ultra
sonic. [Online]
Available at: http://www.ceh.ac.uk/data/nrfa/data/spatialdata.html?53022
[Accessed 4 12 2014].
Buchanan, R. A., 1989. The Bridges Of Bath. 1st ed. Bath: s.n.
Guilmore, J. (1717). Joseph Gilmore's Map of The City of Bath 1694-1717 - detail. Retrieved 03 22, 2015,
from Bath in Time: http://www.bathintime.co.uk/image/208730/joseph-gilmores-map-of-the-cityof-bath-1694-1717-detail
Masters, C. H. (1808). The City of Bath by Charles Harcourt Masters. Retrieved 03 11, 2014, from Bath in
Time: http://www.bathintime.co.uk/image/264598/the-city-of-bath-by-charles-harcourt-masters1808
Bath and North East Somerset Council. (2015). Bath and North East Somerset Council - Population. Re
trieved February 15, 2015, from http://www.bathnes.gov.uk/services/your-council-and-democracy/
local-research-and-statistics/wiki/population
T. Hewlett; Modelling Historic Flooding of the Avon in Bath | 39

GB Historical GIS / University of Portsmouth. (2014). Bath and North East Somerset UA through time |
Population Statistics | Total Population, A Vision of Britain through Time. Retrieved Febuary 15th,
2015, from http://www.visionofbritain.org.uk/unit/10168442/cube/TOT_POP
Black & Veatch. (2013). B&NES - Bath Flood Risk Management Project. Bath: Bath and North East Somer
set Council.
Davis, G. a. B. P. 2., 2006. A History of Bath: image and reality.. Lancaster: Carnegie.
Bath in Time. (1882). Old Bridge and Full Moon Hotel 1882. Retrieved 04 06, 2015, from Bath in Time:
http://www.bathintime.co.uk/image/150047/old-bridge-and-full-moon-hotel-1882
Ordinance Survey. (2014). Geo Refrenced Maps - 1960 Ordinance Survey . Retrieved 04 13, 2015, from National Library of Scotland: http://maps.nls.uk/geo/explore/print.cfm#zoom=15&lat=51.3816&lon=
-2.35682&layers=10
B&NES. (2014). ISIS Model of the Avon in Bath. Bath.

Bristol Avon River Board. (1954, May 13). Sections of the River Avon Victoria Bridge to Pulteney Weir.
Bath.
Chow, V. T. (1959). Open Channel Hydraulics. Illinois: The Blackburn Press.
Bath in Time. (1960). The rear of Pultney Bridge, Bath. Retrieved 03 16, 2015, from Bath in Time: http://
www.bathintime.co.uk/image/1099590/the-rear-of-pulteney-bridge-bath-c-1960s
Bath in Time. (1972, 10 31). View from the Beaufort (now Hilton) hotel site looking towards the rear of
Pulteney Bridge, 31 October 1972. Retrieved 03 15, 2015, from Bath in Time: http://
www.bathintime.co.uk/image/1119484/view-from-the-beaufort-now-hilton-hotel-site-lookingtowards-the-rear-of-pulteney-bridge-31-october-1972
Adams, R. (1768). Pulteney Bridge, Bath c. 1768. Retrieved 02 05, 2015, from Bath in Time: http://
www.bathintime.co.uk/image/1116023/pulteney-bridge-bath-c-1768
Wilkinson, M. (2011). Trip to bath. Retrieved 04 06, 2015, from this other world: http://www.thisotherworld.co.uk/
UK5.html

Bath in Time. (1960). Fishing on the Pulteney Weir, c.1960s. Retrieved 04 05, 2015, from Bath in Time:
http://www.bathintime.co.uk/image/1099732/fishing-on-the-pulteney-weir-c-1960s
Bath in Time. (1971). 1971 Aerial veiw of Pultney weir. Retrieved 03 12, 2015, from Bath in Time: http://
www.bathintime.co.uk/image/202418/1971-aerial-view-of-pulteney-weir-april
Ollis, I. (1964). Pulteney Bridge and the riverside, 1964. Retrieved 02 30, 2015, from Bath in Time: 5)
http://www.bathintime.co.uk/image/953393/pulteney-bridge-and-the-riverside-1964
Whiteway, J. (1960, 12 5). Floods at Old Bridge, Bath,1960. Retrieved 03 21, 2015, from Bath in Time:
http://www.bathintime.co.uk/image/1099007/floods-at-the-old-bridge-bath-c-1960

40 | T. Hewlett; Modelling Historic Flooding of the Avon in Bath

You might also like