Predictive and investigative modelling of
flood hazard in Welsh river catchments
The River Dee at Bangor-on-Dee
FINAL REPORT
March 2006
River Basin Dynamics and Hydrology Research Group,
University of Wales, Aberystwyth
Project funded by:
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Table of contents for Volume 1
Page
List of Figures
iii
List of Tables
iv
Table of contents for volume 2
v
Executive Summary
1
1.
OPENING STATEMENT
3
2.
REPORT STRUCTURE
3
3.
BACKGROUND TO THE RESEARCH
3
3.1
Flood risk maps
4
3.2
The use of numerical modelling in flood hazard assessment
7
3.3
The use of geomorphological data in flood hazard assessment
3.3.1 Examples of geomorphic change affecting Welsh
floodplains
3.3.2 Geomorphological archives of flooding
8
10
14
4.
PROJECT AIMS AND OBJECTIVES
15
5.
SELECTION OF CATCHMENTS AND REACHES FOR STUDY
15
6.
RESEARCH METHODOLOGY: an integrated geomorphological
modelling approach to flood hazard assessment
17
6.1
6.2
6.3
Data collection and preparation
6.1.1 GIS data sets
6.1.2 Hydrological and climatic data
6.1.3 Assessment of present–day patterns and recent changes
in flood magnitude and frequency in the study catchments
6.1.4 Assessment of present–day patterns and recent changes
in flood hazard in the study reaches
18
18
19
19
19
Field–based geomorphological investigations of study reach
landforms and sediment sequences
6.2.1 Geomorphological mapping
6.2.2 Sediment observations and dating
20
20
22
Modelling methodology
6.3.1 Key features of the CAESAR model
6.3.2 Hydrological calibration
6.3.3 Structure of model set up
6.3.4 HEC GeoRAS modelling
22
22
23
24
25
i
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
6.3.5 Model validation
7
26
SUMMARY RESULTS
27
7.1
Current patterns and recent trends in flooding and flood hazard
27
7.2
Morphological investigations of fluvial landforms and
sediment sequences
7.2.1 Evidence for geomorphic change in the study reaches
7.2.2 Use of LiDAR
7.2.3 Overview of 14C dating results
29
30
35
35
7.3
Summary of CEASAR modelling results
7.3.1 Modelled flood inundation areas
7.3.2 Modelled patterns of erosion and deposition
7.3.3 Modelled impact of climate change
7.3.4 Modelled impact of landuse change
7.3.5 Modelled impact of man-made structures on flood hazard
7.3.6 Pros and cons of the CAESAR dynamic modelling
programme
37
37
40
41
41
42
42
8.
CONCLUSIONS
42
9.
RECOMMENDATIONS
44
10.
ACKNOWLEDGEMENTS
44
REFERENCES (a consolidated list of references for both volumes of the report is
included in Volume 2)
ii
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
List of Figures in Volume 1
Page
Figure 1.
Map of the Afon Teifi floodplain west of Lampeter
5
Figure 2.
Meander migration on the River Severn near Caersws
10
Figure 3.
Changes in channel bed elevation at three cross-sections near
Caersws, River Severn.
11
Figure 4.
Lateral channel migration on the River Tywi, near Llandeilo
12
Figure 5.
Avulsion on the River Tywi near Llanwrda
13
Figure 6.
Cross-sections through river terraces in the upper Dyfi valley
13
Figure 7.
Changes in the area of river gravels exposed in the upper
River Severn, between ~1890 and 2002
14
Figure 8.
Major events and processes to have influenced Welsh rivers
15
Figure 9.
Location of study catchments
16
Figure 10. Outline of the geomorphological modelling approach to flood
hazard assessment
17
Figure 11. High resolution LIDAR DEM of the River Dee at Bangor-on-Dee
18
Figure 12. HEC-RAS modelled present-day flood hazard zones on the River
Dee at Bangor-on-Dee (river flows from west to east)
20
Figure 13. (a) LIDAR derived geomorphological map; and (b) core logs and
14C sample levels and dates. Tregaron study reach of Afon Teifi
21
Figure 14. Conceptual structure of the CAESAR model
23
Figure 15. Modelled discharge for different m and p values
24
Figure 16. Overview of structure of CAESAR
25
Figure 17. Schematic of HEC-GeoRAS
26
Figure 18. Image of HEC-GeoRAS results for the reach around Bangor on
Dee
27
Figure 19. Flood magnitude – frequency curves for the River Severn at
Abermule
28
Figure 20. Recent changes in (A) flood magnitude and (B) flood inundation
Extent
28
iii
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 21. Geomorphological map of the ‘Roman Road’ site, Lampeter
30
Figure 22. Radargram of the Roman Road site, Lampeter
31
Figure 23. Geomorphological map of the Dyfi valley study reach.
31
Figure 24. Patterns of channel change in the Dyfi valley
32
Figure 25. Geomorphological map of the Red House / Ty Mawr study site,
Caersws
33
Figure 26. Cut offs on the River Severn near Caersws
33
Figure 27. DEM of the River Severn study reach at Welshpool
34
Figure 28. LIDAR derived geomorphological map of the Afon Dyfi study reach
35
Figure 29. CPDF plots of radiocarbon dated fluvial units
36
Figure 30. Graph of areas inundated for different magnitude floods for the
Tregaron reach of Afon Teifi
38
Figure 31. Inundation areas for a 220 m3 s-1 flood for the Tregaron reach,
Afon Teifi
38
Figure 32. Inundation areas for a 400 m3 s-1 flood for the Dyfi valley study
Reach
39
Figure 33. Present- day flood hazard zones within the Afon Dyfi study reach
39
Figure 34. Patterns of erosion and deposition within the Caersws reach
40
Figure 35. Cumulative sediment yield plots for the Teifi catchment
41
List of Tables in Volume 1
Page
Table 1. Comparison of ‘fixed boundary’ and ‘dynamic’ modelling methods
9
Table 2. m and p values for catchment simulations
24
iv
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Table of contents for Volume 2
Appendix 1: Study catchments and reaches
Appendix 2: Geomorphological and modelling methods
Appendix 3: Current patterns and recent trends in flooding and flood hazard in the
study catchments and reaches
Appendix 4: Field-based geomorphological investigations of landforms and
sedimentary sequences in the study reaches
Appendix 5: CAESAR modelling results
Appendix 6: Recent changes in flood frequency and magnitude in Welsh river
catchments
Appendix 7: Dating and correlating Late Pleistocene and Holocene alluvial
sequences in Welsh river catchments
Appendix 8: A geomorphic re-appraisal of the River Severn Roundabout GCR site
Appendix 9: Technical Communication: High resolution interpretive geomorphological
mapping of river valley environments using airborne LIDAR data.
Appendix 10: Flooding-related research in Wales: some recent developments
v
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Executive Summary
Background
 Pilot project to investigate the potential of dynamic modelling methodologies to
predict future flood hazard for Welsh river catchments in the context of climate
change
 Response to perceived deficiencies with existing ‘indicative floodplain’ mapping
methods, the length of flood records and the concept of the ‘100 year’ event
 Based on the premise that Welsh rivers migrate, erode and deposit and that
these process strongly influence the location and severity of flooding
 Cellular Automaton Evolutionary Slope and River (CAESAR) modelling software
has been shown to be effective in simulating patterns of river erosion and
deposition for use in predicting flood hazard
 Co-funded by the Welsh Assembly Government, Environment Agency Wales,
Countryside Council for Wales, British Geological Survey and the River Basin
Dynamics and Hydrology Research Group (University of Wales, Aberystwyth)
Methodology
 Seven river reaches were studied in detail located within the catchments of the
rivers Dee, Dyfi, upper Severn and Teifi
 Critical assessment of the flood record and trends for each catchment
 Detailed examination of fluvial landforms for each of the study reaches utilising
LiDAR data, aerial photographs, walkover surveys, ground penetrating radar and
drilling to assess scales, types and rates of change
 Radiocarbon dating of fluvial deposits in each of the study reaches as means of
calibrating erosional and depositional events and processes
 Application of CAESAR modelling method to each of the study catchments and
reaches for three different climate change scenarios (i.e. no change, increased
winter precipitation, and increased precipitation throughout the year)
 Generation of maps for each of the study reaches illustrating the CAESAR
predicted patterns of flooding for each of the three climatic scenarios to allow
comparison with existing EA ‘indicative floodplain’ outlines
Key findings
 With the exception of the Afon Dyfi, analysis of flood records and trends have
revealed marked increases in the magnitude (Dee and Severn) and frequency
(Dee, Severn and Teifi) of flood events since the mid 1980s
 Analysis of current trends suggests that, at specific sites, flood events of a
magnitude previously thought to occur on average once every 100 years are now
likely to occur on average once every 20 years on the Afon Teifi, every 7 years
on the River Dee, and every 3.5 years on the River Severn
 Welsh rivers and valley floors are confirmed to have undergone profound
changes within historical times, but also within the last few decades; the use of
LIDAR is key to the better analysis of floodplain landforms and processes
 In the past, periods of more frequent flooding and valley floor modification
coincided with changes to the Welsh climate, or in landuse, or both
 CAESAR models confirm that deposition and erosion within channels and on
adjacent floodplains, will exert a strong control on the location of future flooding
by lower magnitude events and on the local depth of water, speed of flow and
degree of hazard for all flood events, irrespective of inundation area
1
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1



CAESAR models reveal a complex response to future climate change scenarios for some reaches it predicts increases in areas prone to flooding by moderate
size events of 5 to 15%, but for reaches where it predicts erosion to become the
dominant process, a reduction of as much as 35% may occur
CAESAR models show that where valley sides and high terraces limit the flood
prone area of a reach all ‘realistic’ flood events above a certain magnitude
inundate broadly the same area of the valley floor, only in lowland and coastal
setting will predicted increases in flood magnitude lead to significant increases in
flood inundation areas
The current CAESAR software does not adequately model the lateral migration
of river channels and the impact this may have on future flood hazard
Principal conclusions
 Significant increases in the frequency and magnitude of floods affecting some
Welsh rivers are a response to climate change and set to continue, rendering
concepts such as the ‘100 year’ and ‘1000 year’ flood event meaningless as an
underpin to medium term planning policy and flood defence strategies
 During periods of increased flooding, and of climate and landuse change, Welsh
rivers significantly modify the form of their valley floors
 Hydraulic modelling procedures which view valley floors as fixed and unchanging
cannot provide realistic indications of the patterns, depths, speeds of flow and
hence the hazards posed by future floods
 Dynamic modelling software such as CAESAR predicts and takes account of
changes to the floodplain surface and, hence, provides a more realistic indication
of areas where the medium- and long-term hazards of flooding are likely to be
greatest
 The refinement and application of dynamic modelling software such as CAESAR,
underpinned by the landform and flood analysis procedures developed as part of
the project, is a prerequisite of effective and focused flood defence strategies and
flooding-related planning policy in Wales
Recommendations
 Improve the capability of CAESAR to model lateral migration of river channels
and its impact on future flood hazard
 Adapt and test the combined CAESAR and landform-based flood assessment
methodology to lowland and tidal reaches of Welsh rivers (e.g. River Usk)
 Investigate the use of the CAESAR methodology for prioritising the construction
of flood defences and for gauging the efficacy of different forms of flood defence
structure
 Encourage assessors of flood hazard in Wales to test and adopt a combined
landform-based flood assessment and dynamic modelling methodology in the
production of a ‘future flood hazard map’ for Wales for use in landuse planning
and in any forthcoming revision of TAN 15
 Ensure LIDAR data is available for all Welsh river catchments to underpin future
flood hazard-related research
2
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
PREDICTIVE AND INVESTIGATIVE MODELLING OF FLOOD
HAZARD IN WELSH RIVER CATCHMENTS
1.
OPENING STATEMENT
This report documents the results of a 3-year pilot project to research the potential of
using geomorphological data and fluvial modelling to provide realistic estimates of future
flood hazard in Welsh river valleys. The project is based on the simple and
incontestable premise that rivers migrate, deposit and erode and that over time these
processes can significantly alter the form of a valley floor. In Welsh river valleys, as
elsewhere, these processes, coupled with predicted changes in the climate and possibly
in land use, will influence the location and severity of future floods, possible markedly
so. Most modelling procedures currently used to define flood prone areas fail to take
these factors into account and hence the results they provide are of only limited value
as indicators of future hazard. Only by understanding the nature of these landform
changes in river valleys, and by the use of refined modelling methods which take them
into account, can more realistic and informed estimates of future flood hazard, and the
risks it poses, be made. Such estimates are essential for sustainable and cost-effective
medium- and long-term landuse planning and flood defence strategies in Wales.
The project was co-funded by the Welsh Assembly Government (WAG),
Environment Agency Wales (EAW), Countryside Council for Wales (CCW), British
Geological Survey (BGS) and the River Basin Dynamics and Hydrology Research
Group (RBDHG) (University of Wales, Aberystwyth). The research was
undertaken, and the report written and compiled by the River Basin Dynamics
and Hydrology Research Group assisted by the British Geological Survey. The
findings of the project and the conclusions and recommendations made in this
report are advisory. They are not necessarily endorsed by the Welsh Assembly
Government, Environment Agency Wales, or the Countryside Council for Wales.
2
REPORT STRUCTURE
This report comprises two volumes and a CD. Volume 1 is a stand-alone synopsis of
the project which provides the background and scientific rationale, as well as outlining
key aspects of the methodology and results, followed by the conclusions and
recommendations arising from the research. Volume 2 comprises ten technical
appendices, which provide detailed information on the project methodology and its
interpreted results, together with a consolidated list of references. The CD presents the
results of the river valley landform and hydraulic modelling, which formed the central
part of the project, as a series of maps with explanatory keys.
3
BACKGROUND TO THE RESEARCH
The severe flooding events of recent years have focused both public and political
attention on the shortcomings in planning and building control policies in relation to
development on floodplains in Wales. Climate-based predictions that both magnitude
and frequency of flood events are set to increase in the future provided additional
impetus for a reassessment of central government policy given to Welsh planners and
3
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
developers alike. Consequently, in 2001, the Welsh Assembly Government (WAG)
initiated a revision of their Technical Advice Note (TAN) 15 ‘Development and flood
risk’, which would supplement policy statements in Planning Policy Wales, published in
March 2002. Following a lengthy period of consultation and compilation the new TAN,
which utilises a radical new special approach, was published in 2004. Arising from the
advice it received during its compilation, the WAG has encouraged and funded a
number of important research and data-gathering initiatives designed not only to
underpin the procedures and guidance offered by the new TAN, but also to inform future
debate and to pave the way for further revision.
At a seminar in 2001, BGS and RBDHRG highlighted the deficiencies in existing
floodplain mapping and modelling methods, and with the use and interpretation of flood
records. They proposed a pilot project which would seek to integrate an understanding
of the rapidly changing nature of Welsh river floodplains, with modelling software
designed to predict the likely impact of these changes on future flood hazard. The
project gained approval and funding from its participating organisations, and work
began, in November 2002.
3.1
Flood risk maps
Central to the policy set out in the new TAN 15 was the compilation of a map which
delineates those areas adjacent to Welsh rivers and coastlines which are at real risk of
being flooded – now or at some time in the future. The underlying concept was to
distinguish zones deemed to be at high risk of flooding, from areas of medium to low
risk, and from areas at little or no risk. The TAN gives a clear indication as to the types
of development that should be permitted in, or precluded from, each zone and of the
costly investigative, remedial and precautionary measures required to be undertaken by
any developer seeking to build in high-risk areas. Hence, the accuracy of the zonal
boundaries shown by the TAN 15 map, and the criteria and data sets used to define
them was of fundamental importance to the WAG advisors, and was the main focus for
investigation by its researchers.
The EA are statutory consultees in the planning process and responsible for issuing
flood warnings and for the construction and maintenance of flood defences. Maps
issued by the EA in the 1990s for many catchments in mid Wales delineated areas of
the main trunk valleys deemed to be at highest risk of flooding. When public flood
warnings were issued, in the form or amber and red alerts, they related to the likelihood
of imminent flooding for these particular sectors of the valley floor. The outline of these
‘at risk’ areas was based on a variety of different techniques and data sets. Some
reaches had been surveyed in detail (Section 105 surveys) and hydrological models
used to calculate areas likely to be inundated by annual floods. However, for large
portions of many river catchments throughout Wales, but most notably in rural settings,
the flood prone outline was based on historical records and anecdotal data.
Unsurprisingly, the large flood events of the mid to late 1990s exposed the deficiencies
in this generation of flood inundation maps and saw large areas outside the depicted
high risk limits flooded, in some cases on several occasions.
Waters et al. (1997), as part of a BGS project to investigate earth science factors
relevant to land use planning in the Teifi valley, contrasted the then extant EA ‘at risk’
zone with BGS maps of alluvial landforms. In many of the Teifi reaches surveyed by
BGS, the area mapped as active floodplain (shown as ‘alluvium’) was significantly wider
than the high-risk zone shown on the contemporary EA maps (Figure 1). Moreover, the
4
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
outline of the floodplain mapped by BGS more closely matched areas reported by
landowners to have been submerged during recent winter floods, suggesting that the
landforms surveyed by BGS presented a more accurate and realistic depiction of flood
prone areas. These findings would underpin a BGS assertion that the landforms created
by the rivers themselves, fashioned over hundreds and in some case many thousands
of years, were a more accurate indicator of long-term flood susceptibility.
Figure 1. Map of the Afon Teifi floodplain west of Lampeter showing the extent of BGS
alluvium compared with the area for which the EA issued flood warnings in the mid
1990s (adapted from Waters et al., 1997)
The damaging flood events of the 1990s galvanised on-going efforts to produce more
accurate and more relevant ‘indicative floodplain’ maps. This effort coincided with the
availability of improved digital elevation models (DEMs) for the UK landscape, including
Wales. It fuelled the development of enhanced modelling procedures which integrated
digital datasets of other relevant catchment properties, including rainfall, ground
permeability, landuse and vegetation cover. Underpinning these developments was the
concept of defining the ‘hundred year return level’ – the need to focus on establishing
which areas would be inundated by floods of a magnitude that occur, on average, once
every hundred years. This was viewed as an appropriate time scale in the context of
5
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
sustainable landuse planning and hence as the standard to which flood defences should
be built.
One of the early products of this phase of research was the generation of the digital
‘Flood risk map for England and Wales’ by the Institute of Hydrology (IH) (Morris and
Flavin,1996). This was the first attempt to portray all areas at risk from the 100-year
flood event using a single uniform modelling technique. However, as significant an
advance at it was, the IH map had several acknowledged drawbacks. These included
some of the basic assumptions made within the modelling procedure (discussed in
detail below), but also the relatively coarse grid (50 m 2) used to calculate and portray
the data. Nevertheless, the EA subsequently adopted the IH data as they began to
upgrade their maps in accord with the 100-year event concept. Thus, at the time of their
research undertaken on behalf of the WAG, Thompson et al. (2003) noted that the EA
maps for Wales depicting the 100-year return level were hybrid in character, utilising
Section 105 mapping where available, but elsewhere relying on recorded events or the
IH data. A smooth outline to the flood risk envelope would give way to an obviously
stepped outline wherever the IH data was utilised. Thompson et al. (2003) provide a
detailed and authoritative critique of the contemporary EA maps for Wales and the
methodologies and data they were based on; they recognised some key deficiencies:
1)
2)
3)
The estimates of what constituted the 100-year event were based on gauging
records, which for many Welsh river catchments extended over just the last
few decades. Recent flood events had consistently exceeded these estimates
suggesting that the available records were insufficient to extrapolate a
realistic measure of the 100-year event.
The maps took no account of predicted 21st century changes to the Welsh
climate and therefore their use in the context of medium to long term landuse
planning was limited.
The mixture of datasets used to compile the maps meant there was built-in
inconsistency and inaccuracy. Though the Section 105 mapping provided a
detailed topographical base to estimate the present day flood prone areas,
the historical records relied upon in many areas were of unverified accuracy.
The DTM used by the IH and the resulting jagged outline to the flood risk
envelope was also viewed as too imprecise for landuse planning purposes.
The widespread recognition that the available gauging data was not providing realistic
estimates for the 100-year event prompted a gradual shift by the EA and the flooding
research community, towards the use of the replacement expression ‘1% annual
probability’. Thus the indicative floodplain outline, which remained the same, was now
viewed as defining the limits of flood events of such magnitude that there was only a 1%
chance of them occurring in any one year. It was a cosmetic change that failed to
address the key deficiencies in the methodologies and data. To circumvent these
issues, the EA, first in England as part of PPG 25, and more recently in Wales to
service TAN 15, has produced an ‘extreme flood limit’ map. This seeks to portray areas
liable to flooding once every 1000 years, or with a 0.1% chance of being flooded in any
one year. The hope was that this extended flood limit will enclose all areas likely to
become flood prone as the predicted effects of climate change take effect later in the
century, thus hopefully providing a failsafe outline to inform landuse planning and flood
defence strategies.
6
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
3.2
The use of numerical modelling in flood hazard assessment
Numerical modelling is a well–established tool for assessing the extent and limits of
inundation produced by flooding. For a given flood event within a river of known
geometry and slope, hydraulic simulation models have proved to be successful in
reconstructing the known limits and extents of observed inundation events. Most of
these models are underpinned by simple 1– or 2–dimensional flow equations (e.g.
Chezy or Manning’s equation), which are used to estimate the water surface elevation
produced by the peak flow magnitude along a reach. The inundation limits of the
simulated event are estimated by draping the flood water surface onto a digital elevation
model (DEM) representation of the surrounding floodplain.
The validity of this approach in estimating present–day flood limits has been indicated
by studies comparing simulated and observed flood limit data, the latter being derived
from sources including visual accounts, aerial photography or remotely sensed data.
The recent advent of precise, high–resolution valley–floor topographic data, such as the
UK Environment Agency’s LiDAR data set, has resulted in significant improvements in
the accuracy of flood limit simulations.
In some cases, hydraulic models have also been employed to assess future flood
hazard within a reach. Here, the magnitude of future flood events entering a reach is
normally estimated using a rainfall–runoff simulation model of catchment hydrological
response with respect to anticipated climatic or land use change scenarios. However, a
major shortcoming of this modelling approach is the underlying assumption that channel
morphology and floodplain topography are fixed and not liable to change over time – it
can be described as ‘fixed boundary’ modelling (Brewer et al., 2005).
The view that rivers and floodplains are geomorphologically invariable over time has
been overturned by field evidence from Welsh valley–floors (see below). In the past,
many rivers and floodplains across Wales have undergone profound morphological
changes, modifying the extent and limits of flooding, and therefore flood hazard, across
a wide range of time scales. In view of this, the ‘fixed boundary’ catchment modelling
approach is likely to be severely limited in relation to strategic flood hazard assessment
(Table 1).
In the last decade, advances in the understanding of catchment dynamics and rapid
developments in geo–computing technology have facilitated the development and
implementation of landscape evolution models. These ‘dynamic’ catchment models
recognise that climate and land use, water and sediment transport on slopes and in
channels, and valley–floor topography are inter–connected, non–stationary properties of
catchment systems (Table 1).
Because landscape evolution models can simulate morphological change in rivers and
on floodplains, they are potentially valuable for assessing the extent to which
geomorphological controls affect flood hazard. One dynamic model – CAESAR (Cellular
Automaton Evolutionary Slope and River model) – has proved to be particularly
successful for simulating how the effects of environmental changes over time affect
catchment and valley–floor geomorphology (Coulthard et al., 2002; Coulthard et al.,
2005). The CAESAR model represents the landscape as a mesh of uniform grid cells,
each containing data for land and water elevation, water discharge, vegetation cover
and grain size distribution. Uniquely, CAESAR models processes across the whole
catchment area, simulating not only the transmission of flood waters, but also changes
7
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
in slope, river and floodplain morphology arising from erosion and deposition as water
and sediment are transferred between hill slopes, tributary streams and main river
channels.
A rainfall–runoff model drives simulations of overland and river flow across the
catchment area and this forms the basis for calculations of river erosion, sediment
transport and deposition in each grid cell. Changes in land elevation resulting from
these processes, and from others associated with landsliding and soil creep, are
updated simultaneously at short time intervals. As all CAESAR processes operate
simultaneously within the same regular grid, feedback mechanisms between them are
automatically integrated. This is a key feature of the model, and allows inputs of rainfall,
vegetation cover and topography to drive simulations of landscape development at all
time and space scales.
3.3
The use of geomorphological data in flood hazard assessment
The application of geomorphological data in flood hazard assessment is underpinned by
the basic principle that ‘the past is the key to the future’. Floodplains are constructed of
landforms and sediments which themselves are a product of past changes in flooding
and river behaviour. Geomorphological study of this natural archive can yield unrivalled
information on past changes in flooding and river behaviour, and thus provides a sound
basis for estimating future trends in flood hazard.
Geomorphological insight is particularly important because the form – or morphology –
of rivers and floodplains, exerts a strong influence on flood hazard. In the short–term,
erosion and sedimentation during a single flood event can cause morphological
changes in the river and on the floodplain. This will modify the position and capacity of
the channel and the elevation of the floodplain, locally altering the extent and limits of
inundation produced by a later flood of the same magnitude.
Over long time scales, vertical and lateral river channel changes have left behind a
legacy of characteristic landforms (such as river terraces and abandoned former
channel courses termed ‘palaeochannels’) and deposits which result from, and
determine, long–term patterns of flood hazard. In this respect, Welsh floodplains contain
a potential archive covering the period of time since the end of the last glacial episode –
a time scale of over 11,500 years.
The integration of geomorphology in hazard assessment is a fundamental requirement
in order to provide more accurate forecasts based on the best available data. In this
respect, geomorphological studies can provide unrivalled information on two inter–
related themes:


Characterising channel and floodplain morphological responses to past
climatic/land use change, permitting informed assessment of the likely
morphological impacts of future anticipated environmental changes;
Extending the flood record to periods that lie beyond the reach of instrumental
analysis and historical data, thus improving awareness of variability in long –
term flooding and reducing uncertainty in relation to the magnitude and frequency
characteristics of important extreme events.
8
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Table 1. Comparison of ‘fixed boundary’ and ‘dynamic’ modelling methods (after Brewer
et al., 2005)
‘FIXED BOUNDARY’ MODELS
(e.g. ISIS)
Under-pinned by the premise
that the topography of river
catchments,
and
their
floodplains and channels is
‘fixed’ (ISIS sediment only
alters channel topography)
‘DYNAMIC’ CATCHMENT MODELS
(e.g. CAESAR )
View river catchments and floodplain
systems as evolving and dynamic.
Testify to periods of relative stability
punctuated by intervals of rapid change:
 Evidence that many river terraces
and floodplains are recent in origin,
dating to the last 2000-500 years
 Evidence of major changes in
channel morphology in Wales over
last 50-60 years
Changes are a response to:
 Intra-catchment
effects
e.g.
anthropogenic
 Regional and global effects e.g.
climatic, neotectonic
Require detailed topographic
data (ground surveys, DTM,
LiDAR)
Require catchment
parameters to be defined e.g.
run-off, infiltration,
evapotranspiration,
anthropogenic effects
Require detailed topographic data (ground
surveys, DTM, LiDAR)
Require catchment precipitation and landcover data
Requires geomorphological mapping
Requires intra-catchment and regional
calibration of alluvial surfaces, both
topographically and in terms of age
Requires the interpretation of catchment
changes in terms of past events and
processes
Allows modelling for various Allows modelling for various climatic/rainclimatic/rain-fall scenarios
fall scenarios
Only long term variables are In addition to climatic and anthropogenic
climate and/or land-use
effects, floodplain topography and channel
morphology are treated as important long
term variables
Is underpinned by retrospective catchment
modelling (retro-validation)
Allows more realistic and accurate
predictions of flood risk and channel
change
9
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 2. Meander migration on the River Severn near Caersws, using repeat surveys
The field–based geomorphological approach combines interpretative morphological
mapping and sediment studies, including dating (e.g. using radiocarbon analysis), to
reconstruct the spatial and chronological evolution of a floodplain system. For recent
centuries, survey data may be supplemented by information from available
documentary records of flood events. Maps and aerial photographs provide a further,
valuable source of information on river channel changes over recent decades.
3.3.1 Geomorphic change in Welsh floodplains
Changes in flood hazard may arise from lateral and vertical changes in the position of a
river channel. These channel changes can be progressive or rapid, and represent an
important control on flood hazard patterns across a spectrum of space and time scales.
Repeat field surveys conducted on the upper River Severn suggest that progressive
lateral channel migration at a meander bend site is proceeding at a rate of ~13 m per
year (Figure 2).
A local change in flood hazard is taking place: the river is lengthening its course and
reducing its gradient, slowing the transmission of floods and thus reducing the
magnitude required to overtop the banks. On the same river, flood hazard has also
altered due to vertical shifts in the river bed recorded between 1987 and 2003 (see
Figure 3). Erosion and deposition has significantly changed the elevation of the bed at
three locations, raising it by over 0.5 m at one site. The general trend suggests that the
capacity of this river has reduced and, hence, flood hazard is increasing.
10
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
2003
Cross-section 1
1987
125
Elevation (m)
124.5
124
123.5
123
Lowest level of river bed in 2003
Lowest level of river bed in 1987
122.5
0
20
40
60
80
100
120
Distance (m)
2003
Cross-section 10
1987
124
123.5
Elevation (m)
123
122.5
122
121.5
Lowest level of river bed in 2003
121
Lowest level of river bed in 1987
120.5
0
10
20
30
40
50
60
70
80
Distance (m)
2003
Cross-section 20
1987
123
122.5
Elevation (m)
122
121.5
121
120.5
Level of river bed in 1987
120
Level of river bed in 2003
119.5
0
20
40
60
80
100
120
140
160
Distance (m)
Figure 3. Changes in channel bed elevation at three cross-sections near Caersws, River
Severn. Red areas represent erosion; blue areas represent deposition.
11
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 4. Lateral channel migration on the River Tywi, near Llandeilo, from historical
maps and aerial photographs
Maps and aerial photographs have been used to trace progressive lateral channel
changes over the last 160 years on a reach of the River Tywi (Figure 4). This river
shifted from a straight course in 1841 to a meandering route in 2002. This has clearly
modified the topography of the floodplain, reduced river slope, and is thus likely to have
altered both the extent and the limits of inundation for a flood of a given size.
A complete change of course (termed ‘avulsion’) of the River Tywi at Llanwrda, between
1985 and 1992, resulted in the switching of this channel to a position very close to a
railway line (Figure 5). Both meander ‘cut offs’ and channel ‘avulsions’ tend to occur in
response to large flood events, and demonstrate the capacity for lateral changes to
promote a sudden change in the extent and limits of flood inundation.
Many Welsh valleys contain a ‘staircase’ of river terraces – past floodplains produced by
lateral channel migration and subsequently abandoned during phases of sustained river
down–cutting. In the Dyfi valley, up to eight river terraces are found, the highest up to
~20 m above the modern river, and on most, the evidence for past lateral migration is
preserved in the form of palaeochannels (Johnstone, 2004; Brewer et al., 2005) (Figure
6).
These examples demonstrate that vertical and lateral river system dynamics have led to
significant, localised changes in flood hazard across a number of Welsh floodplains over
recent years and decades. Once formed, river terraces and palaeochannels exert
important controls on subsequent flood hazard, defining the outer limits of, and
preferred route ways for, flood waters. Over centuries and beyond, river channel
changes shape the valley–floor, exerting a strong control on the large–scale pattern of
flood hazard throughout a river valley.
12
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 5. Avulsion on the River Tywi near Llanwrda.
Figure 6. Cross-sections through river terraces in the upper Dyfi valley interpreted in the
context of the events outlined in Figure 8 (after Johnstone, 2004; Brewer et al., 2005)
13
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
3.3.2 Geomorphological archives of flooding
While studies that focus on river floodplain morphological dynamics exploit the landform
record, sedimentary archives are the key to unlocking the history of flood magnitude
and frequency and flood–associated patterns of flood hazard. The character of coarse
channel deposits and fine floodplain deposits laid down during a flood is to a large
extent a product of the power, or magnitude, of the event.
Over time, changes in event magnitude may be preserved in the nature of accumulated
sediments. Thus, establishing a sediment chronology can form a strong basis for
assessing patterns of flood hazard arising from changes in the magnitude and
frequency of flooding.
Historical maps and sequential aerial photographs have been used to assess how the
extent of exposed river gravel – a measure previously related to flood magnitude and
associated bank erosion rates – has varied across Wales over the last ~100 years
(Gittins, 2004). A very detailed record is available for the River Severn, showing a cyclic
pattern with a peak in the 1940s, and a rapid increase since the 1980s (Figure 7). This
pattern most likely reflects climatic variations on winter flooding. The data indicate that if
wetter and warmer winters continue into the 21 st century, flood magnitudes could return
to or surpass those generated by large snow melt events during severe cold winters of
the mid 20th century.
450,000
Active Gravel Area (m 2)
400,000
350,000
300,000
250,000
200,000
150,000
100,000
50,000
0
1890
1903
1948
1975
1981
1984
1992
2000
2002
Figure 7. Changes in the area of river gravels exposed in the upper River Severn,
between ~1890 and 2002
The short-term variability in flooding seen in the historical reconstruction is also
apparent in a new long–term record of flooding activity, based on a synthesis of
radiocarbon dates taken from river sediments in Wales and elsewhere in GB (Macklin et
al., 2005). The results suggest that flooding activity in Wales has been highly
changeable over the last ~11,500 years, typically varying in cycles covering several
centuries. These cycles link to a succession of changes in the Welsh climate extending
back to the last Ice Age and the withdrawal of tundra-like conditions, and record periods
of increased precipitation generally coincident with periods of cooler climate (neoglacial
events) (Figure 8). Other factors, such as crustal movements and changes in sea level
also influenced the development of Welsh rivers during this period and, with sea levels
predicted to rise, may do so in the future.
14
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 8. Major events and processes to have influenced Welsh river catchments since
the late Ice Age (Holocene neoglacial event record is taken from Macklin and Lewin,
2003; LIA – Little Ice Age) (after Brewer et al., 2005)
4
PROJECT AIMS AND OBJECTIVES
The primary aim of this study is to evaluate a pilot geomorphological modelling
methodology for assessing flood hazard in Welsh river catchments. In this, field
geomorphological investigations and CAESAR geomorphological modelling will be
integrated in order to estimate past and future changes in flood hazard. Specific
objectives of the study were:

To identify the range of river channel floodplain types in Wales

To establish whether any changes in flooding have occurred in the recent
past and to investigate the causes for any change

To establish the most robust and cost–effective means of identifying the
floodplain for a given flood magnitude

To forecast the effects of climate and land-use change on future patterns
of flood hazard

To produce maps showing zoned assessment of future flood hazard
5
SELECTION OF CATCHMENTS AND REACHES FOR STUDY
The selection of study catchment areas and river reaches was based on the need to
encapsulate different types of catchment, river channel and floodplain systems, but also
to base modelling on areas for which high quality research data exist. Based on these
criteria, four river catchments were selected (Figure 9) – the Teifi and Dyfi (both
unregulated rivers; although there is some regulation of the Teifi’s source at Tefi Pools),
the upper Severn (a semi-regulated river) and the Dee (a regulated river). Seven
reaches were selected for study: two in each of the Teifi (Tregaron and Lampeter),
upper Severn (Caersws and Welshpool-Roundabout) and Dee (Corwen and Bangor-onDee) catchments, and a single site from the Dyfi (Machynlleth).
15
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 9. Location of study catchments
16
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
6
RESEARCH METHODOLOGY: an integrated geomorphological modelling
approach to flood hazard assessment
An outline of the methodological approach adopted in this study is shown in Figure 10.
The project can be broken down into three main phases of work:
(1)
(2)
(3)
Data collection and preparation: collection and integration into GIS of data sets
required for modelling studies; provision of baseline data on present–day flooding
and flood hazard at study sites.
Field-based geomorphological studies: assessment of past flood hazard in study
reaches based on landform and sediment evidence for river floodplain
morphological dynamics and variations in flood activity.
Model–based geomorphological studies: assessment of future flood hazard
based on modelling geomorphological response of study catchments and
reaches to selected environmental change scenarios; investigation of linkages
between environmental change, geomorphological change, and flood hazard.
Figure 10. Outline of the geomorphological modelling approach to flood hazard
assessment
17
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
6.1
Data collection and preparation
6.1.1 GIS data sets
Catchment and reach Digital Elevation Models (DEMs), used to define the present–day
landscape surface in CAESAR catchment model simulations, were constructed from
Ordnance Survey national elevation data for Wales. Within this data set, Wales is sub–
divided into regularly spaced 50 m grid squares containing mean land surface elevation
values accurate to ± 1 m. The data was stored within ARCGIS in GRID file format. Pre–
processing within ARCHYDRO software was necessary to transform the national
elevation data set into CAESAR–compatible catchment DEMs. Where possible, reach
DEMs were based on high–resolution LiDAR (Laser Induced Direction and Ranging)
elevation data, supplied by the Environment Agency. LiDAR data provides precise
valley–floor elevations (± 0.1 m), resolved on a 2 m grid. It represents the best available
topographic data for Welsh floodplains. LiDAR was available for the Dyfi, Teifi and Dee
reaches (Figure 11), but not those in the Severn catchment.
Filtered LiDAR elevation data
2m spatial resolution
N
Flow
Bangor-on-Dee
2km
Figure 11. High resolution LiDAR DEM of the River Dee at Bangor-on-Dee
Owing to the lack of LiDAR data for the upper Severn, a different method was used to
construct reach DEMs for the Welshpool and Caersws reaches. This involved the use of
spot height data (0.1 m interval) and contour data (0.25 m interval), supplied from aerial
photographic analysis by the Environment Agency. In ARCGIS, spot heights and
contours were digitised, combined, and converted into continuous surface elevation
maps in a triangulated irregular network (TIN) format.
Information on channel bathymetry, not available from the LiDAR or digitised data, was
integrated into reach DEMs by the following method. Channel margins were identified
from digital OS LANDLINE feature maps. Available surveyed channel cross sections
were cut into the DEM. River bed elevations between survey sites were estimated by
18
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
linear interpolation using RASTEREDIT software (M. Van de Wiel, RBDHRG). For unsurveyed reaches, channel depth was estimated according to known width–depth
relationships.
6.1.2 Hydrological and climatic data
Gauged river flow data and rain gauge data were supplied for each study catchment by
the Environment Agency. They included daily mean flow records, varying in length
between 34 and 47 years, and peak daily flow records, mostly covering the period since
1990 or 1991. These data sets were used for two purposes in the study: to establish the
present day and recent flood magnitude and frequency characteristics of each river; to
calibrate CAESAR catchment models. Hourly rain gauge data covering the last 7 years
were available for several locations, located either within or close to each study
catchment. These data were used as the climate input for CAESAR catchment
geomorphic simulations.
6.1.3 Assessment of present–day patterns and recent changes in flood magnitude and
frequency in the study catchments
Annual maximum flood series, based on water years (October to September), were
derived from a single flow gauge on each river. For the River Dee at Manley Hall, this
series was derived from peak daily flow data covering the period 1970–2003. For the
remaining rivers, however, peak daily flow records beginning after 1990 were deemed
to be too short for meaningful analysis. For these rivers, annual maximum series were
derived from mean daily flow records (the Afon Dyfi at Dyfi Bridge, Afon Teifi at Glanteifi
and River Severn at Abermule), also covering the period 1970–2003.
In order to assess recent changes in flood magnitude and frequency, the four flow
records were partitioned in time. Based on inspection of the River Dee data indicating
an increase in flooding after 1987, each record was partitioned into two 17-year periods,
1970–1986 and 1987–2003. For the Dyfi flow record, data for 1971–1974 were
incomplete and were substituted with data from the preceding years 1966–1970.
Gumbel frequency analysis (Gumbel, 1958; Dunne and Leopold, 1978) was performed
on each partitioned annual maximum flood series in order to estimate the return period
of a given event. Flood magnitude–return period curves were plotted on a log-probability
(normal) scale using Weibull plotting positions. A straight line was fitted to the curves
using Origin™, and this was used to estimate the size of 5, 10, 20, 50 and 100 year
floods. It should be noted that this is not the method recommended in the Flood
Estimation Handbook (Robson and Reed, 1999) which is currently used by the
Environment Agency (see Appendix 6 for justification of the method used in this study).
6.1.4 Assessment of present–day patterns and recent changes in flood hazard in the
study reaches
The HEC–GeoRAS software package was used to simulate the extent and limits of
inundation produced by different return period floods according to present–day reach
DEM topography (Figure 12). Present day flood hazard zones were defined as areas
lying within the inundation limits of simulated 100 year (e.g. ‘low’ hazard zone) and 10
year (e.g. ‘high hazard zone) post–1987 flood magnitudes. The HEC–modelled ‘low’
hazard limit was compared to that of the 100 year flood as defined by the Environment
Agency Indicative Floodplain Map (IFM). Recent changes in flood hazard were
assessed by calculating the different extent of flooding for each of the 5, 10, 20, 50 and
100 year return period events between the periods 1970–1986 and 1987–2003.
19
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
A
Figure 12. HEC-RAS modelled present-day flood hazard zones on the River Dee at
Bangor-on-Dee (river flows from west to east)
6.2
Field–based geomorphological investigations of study reach landforms
and sedimentary sequences
6.2.1 Geomorphological mapping
Interpretive geomorphological mapping was carried out in order to identify the
distribution of fluvial features preserved on the floodplain surface within each study
reach. Particular attention was given to mapping the following features: valley–floor
margins, modern channel margins, river terrace surfaces, terrace margins,
palaeochannels and tributary alluvial fans, non–fluvial (e.g. glacial) landforms and man–
made structures (e.g. embankments). Geomorphological mapping of the study reaches
utilised high quality LiDAR data followed by field walk over surveys (Figure 13a).
Available maps and aerial photographs were used as a source of information regarding
the sequence of channel and floodplain changes which have occurred within each study
reach since the mid– to late–19th century. Channel margins, gravel bars and islands
identified on each map, or photograph, were all digitised as feature layers using
ARCGIS software. These digitised layers were transformed and scaled onto a common
OS grid coordinate system facilitating visual comparisons by GIS overlay operations.
This enabled time sequence analysis of lateral channel migration rates and gravel bar
dynamics to be carried out.
20
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Core 4
Core 2
Figure 13. (a) LiDAR derived geomorphological map showing river terraces and
palaeochannels; and (b) core logs and 14C sample levels and dates from the Tregaron
study reach of Afon Teifi
21
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
6.2.2 Sediment observations and dating
Sub–surface floodplain sediments were investigated to provide information about the
chronological sequence of flood–related deposition. Geomorphological mapping and
ground penetrating radar surveys were used to select sites that were most likely to be
underlain by thick vertical sequences of fine–grained (clay to sand sized) material, and
were therefore suitable for retrieval by a percussion ‘vibra’ corer with a maximum
penetration depth of ~7 m. Vertical sediment sequences in cores and river banks were
described in terms of several key characteristics at each depth: grain size,
cohesiveness, water content, sediment bedding, colour and organic matter content
(Figure 13b) . These descriptions formed the basis for interpretations of past floodplain
environments and flooding activity. Material required for radiocarbon dating – wood,
charcoal, leaves and seeds – were sampled. Where possible, radiocarbon samples
were taken from layers where a sedimentary change indicated a change in flooding
activity, such as: (1) an up–sequence change from coarse channel gravels to fine–
grained sediments, relating the abandonment of an active channel; (2) an up–sequence
switch from fine to coarse deposits, relating to an increase in flood activity (Figure 13b).
Radiocarbon samples were analysed and dated at the Waikato Radiocarbon Dating
Laboratory, New Zealand.
6.3
Modelling methodology
A detailed account of the technical aspects of the modelling methodology used in the
project is given in Technical Appendix 2.
6.3.1 Key features of the CAESAR model
The CAESAR landscape evolution model used here (Coulthard et al., 2000; 2002;
2005) is based on the cellular automaton concept, whereby the continued iteration of a
series of local process-‘rules’ governs the behaviour of the entire system (Figure 14).
Although these rules are relatively simple and straightforward representations of fluvial
and hillslope processes, their combined and repeatedly iterated effect is such that
complex non-linear geomorphological response can be simulated within the model. Both
positive and negative feedbacks between form and process can emerge.
CAESAR can be run in two modes: a catchment mode, with no external in-fluxes other
than rainfall; and a reach mode, with one or more points where sediment and water
enter the system. In both modes the model requires the specification of various spatially
distributed landscape parameters (initial conditions): elevation, roughness, grainsizes
and vegetation cover. The temporal input requirements (forcing conditions) vary
according to the mode in which the simulation is run. In catchment mode, the model
requires rainfall data for the duration of the simulation; in reach mode, it requires
discharges and sediment fluxes for all inflow points. These temporal data are usually
specified at hourly intervals.
Landscape simulation in CAESAR follows a simple structure (Figure 14), whereby
topography drives fluvial and hillslope processes that determine the spatial distribution
of erosion and deposition for a given time step. This alters the topography, which
becomes the starting point for the next time step. The model uses variable length time
steps, depending on the rates of erosion and deposition occurring within the system
(see below). Outputs of the model are elevation and sediment distributions through
space and time, and discharges and sediment fluxes at the outlet(s) through time.
Additional fluxes at specified points in the catchment or reach can be easily obtained.
22
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 14. Conceptual structure of the CAESAR model
6.3.2 Hydrological calibration
For all the modelled catchments, the hydrological model needed to be calibrated, so
that simulated runoff for a given rainfall event, matched that measured in the field. This
step is important, as the magnitude of a flood will not only determine the area inundated
by flood waters, but also the volume and rates of sediment generated by the flood.
CAESAR is first calibrated to present–day conditions by assessing the ‘goodness of fit’
between modelled and observed (i.e. gauged) flood peaks over a 1- year calibration
period. In CAESAR, river flow volume at a given time is controlled largely by rainfall
intensity – represented by the ‘p’ value in the hydrological model, but peak flood
magnitude is also controlled by the rate of water table fluctuation – represented by a
factor termed the ‘m’ value. Following the method successfully piloted for Welsh
catchments in a previous study (Coulthard and Jones, 2002), calibration involved
running each model according to different combinations of factored rainfall intensity and
‘m’ value, until simulated flood discharges converged with the gauged flood record.
This process is illustrated with regard to part of the Dee catchment. A real hourly rainfall
record (Bala) was used to run the CAEASAR model with different m and p values. The
output was converted to mean daily flow allowing a direct comparison with the actual
flow as recorded at the Druid gauge for the same period. Figure 15 shows the
comparisons of the two ‘best fit’ m and p values: m=0.015 p=0.7 for all discharge
values, m=0.015 p=0.8 for all discharge values >10 m3s-1.
23
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 15. Modelled discharge for different m and p values compared with the actual
Druid gauge record
On the basis of the above runs the final values chosen were m=0.015 and p=0.7.
Though m=0.015 p=0.8 gave the better fit for flows >10 m3s-1, p=0.7 gave model results
which more closely matched the two largest peaks, and so was considered the overall
‘best fit’ value. Table 2 details the m and p values determined by this method for all the
study catchments.
Table 2. m and p values for catchment simulations
Catchment
Dee
Dyfi
Severn
Teifi
m values
0.015
0.014
0.015
0.02
p values
0.7
1.1
1.0
0.45
The ‘m’ value is additionally important in that it appears to link to surface vegetation
within the catchment, and the way this affects the rate at which rain water soaks into the
ground. This underpins another powerful application of the CAESAR software allowing
the impact of changes in land use, from pasture to forestry for example, to be factored
into catchment modelling and, hence, the impact of such changes on future flood
hazard to be gauged (Coulthard and Jones, 2002).
6.3.3 Structure of model set up
As previously described, CAESAR can run in both catchment and reach modes. Figure
16 shows how each set of simulations was constructed. These range from simple
systems such as the Teifi, where there was only one contributing catchment, to the
Severn, where 13 simulations were required to create the input data for one reach
simulation.
24
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 16. Overview of structure of CAESAR model set up for the Severn at the
Roundabout reach
6.3.4 HEC-GeoRAS modelling
Whilst CAESAR can simulate the areas of a floodplain that are inundated with water
during flood events, the ‘standard’ method used by the Environment Agency and other
consultancies to model flood inundation is a 1 dimensional approach, using models
such as ISIS or HEC-GeoRAS. In order to give the results from this study more
relevance, we use HEC-GeoRAS to model flood inundation areas on the topographies
generated by the CAESAR model.
HEC-RAS is a 1 dimensional hydraulic model working on the step backwater approach
developed by the US Army’s engineering corps. It operates by dividing the channel
network up into a series of linked cross sections (Figure 17). Water depths are then
calculated for given discharges at each cross section based on the slope between the
water surface at a given cross section and the section immediately downstream – hence
the name step back water. HEC-GeoRAS is a version of the model that is integrated
within the GIS package ARCVIEW 3.2. Within the GIS, cross sections are determined,
and the elevations for each point along a cross section are calculated from a DEM of
the modelled surface. These points are then exported into HEC-RAS where the water
surface elevations are modelled. HEC-GeoRAS then exports this data back into the
GIS, where outlines of inundated areas and flow depths are determined by subtracting
the water surface elevation from the original elevation of the DEM. These can then be
displayed on top of the DEM providing extents of flood inundation as shown in Figure
18.
25
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 17. Schematic of HEC-GeoRAS, illustrating (a) the TIN (Triangular Irregular
Network) DEM that cross sections are extracted from (b) the structure of the main
channel network and location of cross sections and (c) detail of one of the cross
sections
HEC-RAS does have some limitations. It is one-dimensional, so only simulates
flow occurring at the cross sections and can provide a slightly inaccurate picture of
overbank inundation extents and depths. However, it is simple to use, relatively well
tested and was specifically utilised in this study to allow direct comparison to methods
used by the Environment Agency.
6.3.5 Model validation
The geomorphic evidence gathered as part of the project provides a basis for evaluating
the extent to which the CAESAR reach models are producing realistic patterns of future
river-floodplain change. While it is clearly not possible to validate predictive results, the
field data provide clear pointers as to whether CAESAR is simulating the type of
changes detected within individual reaches in the past, for example patterns of erosion
or deposition, incision and terrace formation and channel migration.
26
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Flood Hazard
Zone
Figure 18. Image of HEC-GeoRAS results for the reach around Bangor on Dee. The
cross sections used by HEC-RAS are shown running across the floodplain
7
SUMMARY RESULTS
7.1
Current patterns and recent trends in flooding and flood hazard
The results of this aspect of the project are presented in detail in Technical Appendix 6.
Analysis and re-interpretation of the flood gauging records has revealed a step change
in the magnitude and frequency of flood events affecting all the study catchments (e.g.
Figure 19), with the exception of the Dyfi, since the mid 1980s. This re-assessment
supports a partitioning of the flood record data for these catchments into two sections an early period prior to 1987 characterised by generally smaller and less frequent
floods; and a later period from 1987 onwards during which the magnitude and frequency
of flooding has increased, in some cases quite markedly.
This change in flooding regime is thought to record increased precipitation associated
with changes in the Welsh climate, but whether this relates either to the predicted longterm effects attendant to global warming, or to shorter term variations driven by cyclical
changes in North Atlantic atmosphere-oceanic circulation is unclear (Macklin et al.,
2005). It is also important to note that even the largest of the most recent floods, both in
the study catchments and elsewhere, are significantly smaller than the ‘rain on snow’
events which affected many UK rivers during the 1940s and 1960s, though this may
change if the predicted 21st climate trend towards warmer, wetter winters is realised.
The partitioning of the flooding records carries significant implications for calculations of
flood return periods for the study catchments. At Glanteifi, in the Teifi catchment, the
partioned data suggests that floods of a magnitude estimated to have occurred on
average once every hundred years prior to 1987, are now predicted to occur on average
once every 20 years. At Manley Hall on the River Dee, the pre-1987 ‘100 year’ flood
event could now occur, on average, once every 7 years; and on the River Severn at
Abermule, once every 3.5 years. However, the significance of these
27
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
1987-2003
Linear fit 1987-2003
1970-1986
Linear fit 1970-1986
1000
800
3 -1
Annual maximum discharge (m s )
600
400
200
100
80
60
40
20
10
0.999
0.99
0.95
0.8
0.6
0.4
0.2
0.05
0.01
1E-3
Exceedence probability
Figure 19. Flood magnitude – frequency curves for the periods 1970–1986 and 1987–
2003, River Severn at Abermule
Figure 20. Recent changes in (A) flood magnitude and (B) flood inundation extent,
plotted against return period for the Afon Dyfi at Dyfi Bridge (left) compared with the
River Severn at Caersws (right)
28
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
calculations is not in the numbers themselves, but in their clear demonstration that flood
hazard changes over time and the implications this carries for the fixed return period
concept as currently employed in the context of flood defence and landuse planning
strategies.
Of the studied catchments, only the Afon Dyfi fails to show evidence of these recent
changes in flood regime (Figure 20). This may reflect the short and steep nature of the
catchment and the way it transmits flood waters to the gauging point at Dyfi Bridge.
Importantly, the Dyfi data shows the dangers of generalisation – suggesting that in
some cases flood hazard can be so strongly determined by a catchment’s internal
geomorphology and hydrology that moderate climatic changes have little additional
impact.
7.2 Morphological investigations of fluvial landforms and sedimentary sequences
The detailed results of this aspect of the project are presented in Technical Appendices
4, 7 and 8. The examination of landforms and flood-related sediments, including the
results of 14C dating undertaken as part of the project, has confirmed that the valley
floors of the study reaches have undergone periods of profound geomorphological
change during both pre-historical and historical times. The data also indicates that
periods of rapid change to valley floor geomorphology have alternated with periods of
relative stability. This cyclicity appears to relate to past changes in the Welsh climate, or
to changes in landuse, or both. In some reaches major changes are evident over the
last few decades, in others the evidence suggests a period of relative stability lasting
from the late 19th century is giving way to one of channel and floodplain modification.
The types of geomorphic change recorded in the study reaches include:
 lateral migration, leading in some reaches to the abandonment and major
re-alignment of river channels
 river down-cutting (incision) and terrace formation
 a change from low sinuosity to highly meandering channel courses
 building-up (accretion) of the floodplain surface by flood-related deposition
of sediment
 changes in the grade of sediment being transported and deposited by the
river and on its floodplain (e.g. gravel and sand giving way to silt and mud)
These changes to the valley floor are brought about by a complex interplay of erosional
and depositional processes operating both within river channels and on adjacent
floodplains. The varying balance between these processes is reflected in the different
styles of valley floor modification displayed by some of the different study reaches, but
also over time within some of individual reaches. It is also clear that the landforms
created by these contrasting processes strongly influence the distribution of areas
currently prone to flooding by small to moderate size events, but also the depth of water
and speed of flow and, hence, the hazards posed by higher magnitude floods. It follows
that future periods of pronounced floodplain modification will impact on flood hazard, the
more so given their clear link to changes in climate and land-use.
The 14C dating results from floodplain sediments, including abandoned former channel
courses, also testify to periodic changes in the frequency and magnitude of flood events
affecting the different study reaches, and also to related changes in the rates of
sediment accumulation. Again the link to periods of past climate and land-use change is
evident.
29
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
The study has also highlighted the impact of man-made structures such as railway and
road embankments on valley floor processes and landforms. In limiting the lateral
migration of channels and acting as a barrier to flood waters, such structures have
clearly influenced the patterns of change across a valley floor and consequently flood
hazard.
7.2.1 Evidence for geomorphic change in the study reaches
The Lampeter reach of the Teifi catchment provides evidence of several periods of
incision leading to the formation of a series of low-relief river terraces both prior to and
after the Roman period, and most recently during the late 19th century. During the preRoman period, the river occupied a low sinuosity channel, but it subsequently evolved a
highly meandering course with evidence of extensive lateral migration (Figure 21).
Figure 21. Geomorphological map of the ‘Roman Road’ site, Lampeter
GPR images show the subsurface structure of this area (e.g. Figure 22). Perversely,
though the incision displayed by this reach of the river has acted to reduce flood hazard
across some of the higher terraces, the development of a meandering course is
associated with a reduction in channel capacity and an increase in flood activity. The
present channel was established before 1946 and testifies to a period of relative
stability, however, evidence of recent bank erosion and channel migration by as much
as 3 m a year, suggests the onset of a period of renewed valley floor modification linked
to the recent increases in flood frequency and magnitude.
The study reach in the Dyfi catchment, upstream from Machynlleth, displays a staircase
of former floodplain surfaces (river terraces) ranging from 20 to 1.9 m above the present
day river channel and testifies to high rates of river down-cutting over the last 11,500
years (Figure 23). Flood hazard has clearly changed significantly during this period and
the formation of the modern floodplain, sited 1.3 m above the river, around 350 years
ago, saw a further reduction in the area inundated by small to moderate sized events.
The river channel currently follows a low sinuosity course, but abandoned former
30
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 22. Radargram of the Roman Road site, Lampeter along the line shown in Figure
21 intersecting an abandoned former meander channel.
Terrace 1
Twymyn
confluence
Terrace 2
Terrace 3
Terrace 4
Terrace 5
Mathafarn
Floodplain
Post-1884 activity
Afon Dyfi
Stream / drainage ditch
#
Llanwrin
Palaeochannel
#
Radiocarbon sample site
Hendreseifion
Dulas (N)
Aberffrydlan
#
#
#
N
Dulas (S)
Kilometres
0
0.5
1.0
1.5
2.0
2.5
Figure 23. Geomorphological map of the Dyfi valley study reach
(after Johnstone, 2004).
31
3.0
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 24. Patterns of channel change in the Dyfi valley between Llanwrin and
Aberffrydlan between 1884 and 1992
channel courses evidence a previously meandering route and the major changes in the
position of the main channel, shown to have occurred since 1884, have further impacted
on local flood hazard. The local railway embankment is also shown to have been
influential in this process (Figure 24). 14C dates from the floodplain sediments suggest
very high rates of deposition (c. 3.5 mm per year) which, if sustained, could see the
floodplain surface raised by as much as 1.75 m in the next 50 years.
The Caersws reach in the upper Severn catchment is one of the most complex and
active investigated during the project. A series of river terraces and abandoned former
channel courses (palaeochannels) provide evidence of both vertical and lateral
movements in the river channel (Figure 25). Up to nine terrace levels have been
identified though many are of low relief and still prone to flooding. 14C dates from the
second highest terrace suggest a 2,500-year history of subsequent terrace formation.
Historic maps and aerial photographs record major changes to the valley floor during
the last 160 years. Rapid lateral shifts have occurred during the 20 th century, meander
bends having been ‘cut off’ at their neck to form a straighter path (Figure 26).
32
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 25. Geomorphological map of the Red House / Ty Mawr study site, Caersws
Figure26. Cut offs on the River Severn near Caersws
Within its Welshpool reach, the modern floodplain of the River Severn displays two
contrasting geomorphological zones: an inner zone, flanking the modern river course,
and an outer zone sited between the modern flood embankment and a series of low
river terraces (Figure 27). The inner zone is characterised by meandering former
channel courses, frequent flooding and by rates of floodplain deposition of as high as 10
mm per year. The outer zone has a complex multi-threaded pattern of abandoned
channels; its surface is several metres lower in elevation than the inner zone and
characterised by much lower rates of deposition.
33
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 27. Contour/spot height based DEM of the River Severn study reach at
Welshpool
14C
dates suggest that the multi-threaded pattern of river channels preserved in the
outer zone is older and once characterised much of the valley floor. The change to a
meandering single channel as seen in the inner zone may have occurred as late as the
early mediaeval period. In the inner zone, high rates of over-bank deposition and its
elevated nature reflect the growth of a natural levee. However, by impounding
floodwaters and promoting deposition within this zone, the modern flood defences have
significantly contributed to this process. Though the current embankments exercise a
major influence on flood hazard, with such high rates of vertical floodplain accretion
prevailing within the inner zone, their future efficacy will be reduced. Over topping of the
embankments results in large areas of the low standing outer zone being flooded. Over
bank deposition within the inner zone is dominated by cohesive clay and silt grade
sediment and this has contributed to the lateral stability of the river’s present-day
course, which, in contrast to the Caersws reach, has remained largely unchanged over
the last 160 years. However, the increasing height difference between the two zones
increases the future risk of avulsion – were the confining levee and embankments to be
breached, the river might abandon its current course in favour a new one sited within
the low ground of the outer zone. Clearly a change of this type would have a major
influence on flood hazard.
The presence of numerous abandoned meandering former channel courses in both the
study reaches in the Dee catchment, at Corwen and Bangor-on-Dee, provide evidence
of major lateral movements of the river channel prior to the early 19 th century, but of no
marked migration since then. This phase of stability, as on the River Severn, may reflect
the deposition and accumulation of cohesive clay and silt on the adjacent floodplain.
However, there is evidence to suggest the onset of a more pronounced phase of bank
erosion and channel modification. Drill cores demonstrate that, historically, as much as
0.5 m of sediment was deposited during a single flood event in some abandoned
channel courses.
34
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
7.2.2 Use of LiDAR
The availability of LiDAR data for several of the study reaches proved to be particularly
valuable. These data allowed very detailed geomorphological maps of river floodplains
to be compiled prior to field investigations. Techniques have been developed as part of
the project to maximise the potential of the data to view floodplain structures at a variety
of different scales (Figure 28). This successful
application of LiDAR data for mapping Welsh flood
study reaches represents an important new
methodology
for
cost–effective
valley
geomorphological mapping. A full account of the
LiDAR–based methodology developed during the
course of the project is due for publication in Earth
Surface Processes and Landforms later in 2006
(Jones et al., submitted). A copy of the original
manuscript submitted for publication is included as
Technical Appendix 9.
Figure 28. LIDAR derived geomorphological map
showing river terraces and palaeochannels on the
Afon Dyfi (height classes are in 0.5 m intervals
7.2.3 Overview of 14C dating results
Details of the 14C sampling programme and its results are presented in Appendix 8. To
investigate broader patterns of geomorphological activity in Welsh river systems during
the Holocene, the 29 radiocarbon dates obtained during this project have been
calibrated and plotted as a cumulative probability density function (CPDF) in Figure
29A. This has been plotted alongside curves produced for the 55 known radiocarbon
dates that come from Welsh river environments and the 539 known from catchments
throughout the UK (Figures 29B and C, respectively).
The two plots of the Welsh data reveal similarly skewed distributions of radiocarbon
dates, with a trend towards increased rates of geomorphic activity and flooding from the
mid-Holocene (c. 5750 cal. BP) and, more particularly, from c. 2000 cal. BP. The most
prominent peaks in the Welsh CPDF plots can be seen at c. 5500, c.4850, c. 2900,
c.1950, c. 1250, c. 1000, c. 650 and c. 500 cal. BP. These dates may be proposed as
periods when Welsh rivers were experiencing high rates of geomorphic activity and
flooding.
The skewed nature of the distribution of radiocarbon dates is once again evident in the
Great Britain plot of 539 dates. There are also a number of instances when peaks in
Welsh geomorphic activity and flooding appear to coincide with patterns found
elsewhere in Britain. These corresponding phases of flooding are most evident at c. 11
200, c. 5500, c. 4850, c. 2900, c. 1950, c. 1250 and c. 650 cal. BP. As well as these
similarities there are also incidences when the British record of flood events appears to
deviate from the record in Wales. Flood events appear in the British sedimentary record
at c. 9500 and c. 3500 cal. BP, for example, but appear to be absent from the Welsh
flood record.
35
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Relative probability
1.0
(A)
CPDF plot of the 29 14C dates collected from
the Dyfi, Teifi, Severn and Dee
0.8
0.6
0.4
0.2
0.0
14000
10000
12000
8000
6000
4000
2000
0
Age (cal. BP)
Relative probability
1.0
(B)
CPDF plot of the 55 14C dates reported in
Welsh river catchments
0.8
0.6
0.4
0.2
0.0
14000
12000
10000
8000
6000
4000
2000
0
2000
0
Age (cal. BP)
Relative probability
1.0
CPDF plot of the 539 14C dates reported
in British river catchments
(C)
0.8
0.6
0.4
0.2
0.0
14000
12000
10000
8000
6000
4000
Age (cal. BP)
Figure 29. CPDF plots of radiocarbon dated fluvial units from (A) the present
radiocarbon dating project, (B) the whole of Wales,
and (C) GB (after Macklin et al., 2005)
36
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
As well as reflecting increased rates of flood activity, the skewed distributions of
radiocarbon dates are also likely to reflect the preferential preservation of more recently
formed fluvial sedimentary units (Macklin et al., 2005). However, in Wales and
elsewhere, the bias of dated events towards the last 4000 years, and particularly the
last millennium, suggests that human induced landscape modification of catchment
areas (i.e. woodland clearances, farming practices) has dramatically altered flood
hazard on downstream valley–floors.
7.3
Summary of CEASAR modelling results
The results of this aspect of the project are presented in detail in Technical Appendix 5
and on the enclosed CD. The CAESAR modelling runs undertaken for all the various
climate scenarios – no change, 20% increase in winter rainfall, and 20 % increase in
annual rainfall – confirm that erosion and deposition within river channels and on their
adjacent floodplains is likely to significantly influence future flood hazard.
7.3.1 Modelled flood inundation areas
There is a varying response between catchments with the future runs causing increases
in flood inundation area of between 5 and 15%. Notably, there are larger, and significant
decreases in inundation area in the Teifi at Tregaron (Figures 30 and 31) and the Upper
Severn reaches. Some reaches (Upper Severn, Tregaron, Corwen) display significant
increases and decreases in inundation area within the same reach. These changes are
strongly linked to erosion (decrease in area) and deposition (increase in area) within
and around the river channel. Incision increases the depth and thus capacity of a
channel to convey flood waters, making it less likely to flood therefore decreasing the
area inundated. River channel sedimentation has the opposite effect, decreasing the
capacity of the channel increasing the likelihood that the banks will be overtopped and a
flood will occur.
Interestingly, in reaches where there was a decrease in flood hazard, the decrease
occurred across the whole range of floods modelled. This is intuitively correct, as
increasing channel capacity through erosion (lateral and/or vertical) will reduce
inundation area across all flood sizes. Conversely, deposition and corresponding
increases in flood inundation area will not affect the largest floods, as they are still
restricted by the main valley wall – as shown for the Dyfi in Figure 32.
The modelled changes in future flood inundation areas are most notable for moderate
size flood events (e.g. Figure 30) and modelled future inundation areas differ little from
those for the present day topography irrespective of the climate model used. In these
reaches, steep valley sides or the fronts of high river terraces form natural limits to flood
inundation for a range of flood magnitudes including some of the largest predicted
events. In such reaches modelled future inundation areas differ little from those shown
on EA indicative floodplain maps (compare Figures 32 and 33). However, this should
not be seen as evidence that, over time, flood hazard will remain constant for such
floods. The increases in flood frequency and magnitude already identified during this
study, coupled to the predicted effects of future climate change, suggest that high
magnitude events, and therefore complete inundation of the flood-prone valley floor,
37
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 30. Graph of areas inundated for different magnitude floods, for the present day
topography, and the topography after runs Climate 1, 2, and 3 for the Tregaron reach of
Afon Teifi
Figure 31. Inundation areas for a 220 m3s-1 flood for present day topography (blue) and
after model run for Climate 2 (red) for the Tregaron reach of Afon Teifi
38
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
Figure 32. Inundation areas on the Dyfi for a 400 m3s-1 flood for present day topography
(blue) and after the model run for Climate 3 (red) – inundation areas are broadly
coincident
Figure 33. Present- day flood hazard zones within the Afon Dyfi study reach upstream
section including EA ‘100 year’ limit (compare right side of Figure 32)
39
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
is likely to occur with greater frequency in the future. Moreover, the changes to
floodplain morphology predicted by the modelling will impact on the depth of water and
speed of flow and, hence, on the hazards associated which such events. However, the
modelling undertaken for the project also counsels against a simplistic extrapolation of
hazard. For reaches in which erosion is predicted to improve the transmission of flood
waters, both the severity and duration of high magnitude flood events may be reduced.
Conversely, in reaches prone to deposition, the risks associated with such floods may
increase. Moreover, in low relief lowland and coastal reaches where the natural limits to
flooding are less well-defined, future changes in the magnitude of flooding may see
significant increases in flood inundation areas.
7.3.2 Modelled patterns of erosion and deposition
The complexity of river channel and floodplain response to varying flood magnitudes
even within a single reach (Figure 34) shows the danger of generalising river behaviour.
Each section of a river needs to be studied and modelled individually. For example, the
Corwen reach stored sediment during the simulations and the flood hazard would
increase if the volumes of sediment input from the upstream catchments were to rise. A
contrasting example is the Upper Severn at Caersws, where a significant amount of
sediment was eroded and exported from the reach. Upstream catchments input large
volumes of sediment into this reach, suggesting that it might be relatively insensitive to
environmental change as it can move sediment through the reach rapidly. Both these
reaches appear adjusted to upstream sediment supply.
Figure 34 Patterns of erosion and deposition within the Caersws reach, demonstrating
how rapidly patterns of erosion and deposition can change within a reach
40
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
7.3.3 Modelled impact of climate change
A primary aim of this project was to determine whether climatic fluctuations would cause
the morphology of river reaches to change, which in turn could then alter flood hazard.
The results presented here show that there can be significant changes in channel
morphology over 50 simulated years, and that these changes can have a large effect on
flood hazard (increasing it by up to 15% and decreasing by 33%). However, the
changes are not necessarily all due to climate. For example, most simulations show
substantial changes by continuing with present day climate, though several runs do
show increased or enhanced changes with increases in precipitation. These changes
are due to the inherent dynamics of river systems.
7.3.4 Modelled impact of landuse change
Catchment-scale modelling results have also provided insights to the interrelationship between sediment yield, climate change and vegetation cover. Simulating
changes in land cover (by varying the ‘m’ value within the modelling equations), for
example a change from woodland dominant to grassland dominant, appear, on their
own, to have little impact on sediment yield (Figure 35). Modelled increases in
precipitation, however, have a greater impact. However, when both effects are
combined, increases in sediment yield predicted by CAESAR are significant, suggesting
that catchment behaviour is sensitive to combined changes in climate and land cover
(and hence land use). These responses highlight a non-linear relationship between
sediment yield and climate change which has important implications for river channel
and floodplain behaviour, and could be an important factor determining changes in the
capacity, elevation and position of river channels, and hence flood risk, over future
decades.
Low ‘m’
Med ‘m’
High ‘m’
Figure 35. Cumulative sediment yield plots for the Teifi catchment, produced by running
CAESAR according to 9 different combinations of climatic (‘p’ value) and land cover (‘m’
value) conditions over a 50 year period
41
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
7.3.5 Modelled impact of man-made structures on flood hazard
In line with the findings of the geomorphological aspects of the study, the CAESAR
modelling appears to confirm that man-made structures will continue significantly to
influence future flood hazard. This carries important implications for the use of the
modelling in the context of flood defence design, suggesting that CAESAR has the
potential to predict the impact and efficacy of different types and locations of flood
defence prior to their construction.
7.3.6 Pros and cons of the CAESAR dynamic modelling programme
The modelling results from CAESAR do not to replicate the patterns of lateral
movement of river channels shown by geomorphology to have been an important
process in all the study reaches. As currently configured CAESAR focuses on modelling
vertical movements of river channels and on the adjacent floodplains. However, the
lateral migration of Welsh river channels, as in the past, is likely to play a significant in
role in determining future flood hazard and its failure properly to predict such changes is
a serious shortcoming of the current CAESAR programming. In lowland and coastal
reaches, storm and tidal surges can only act to increase the flood hazard. However,
though assessing the likely impact of such processes was not part of the remit of the
project, it also falls outside the capabilities of the current CAESAR software.
Until these deficiencies are addressed, the output of CAESAR can only be viewed as
indicative. Nevertheless, the results obtained by the project make it clear that changes
in morphology caused by climate change, or by the continuous action of rivers can lead
to significant changes in the flood inundation areas (+15% -33%) and this factor is
largely ignored by all conventional modelling studies. If we wish to simulate flood hazard
accurately, then we need to either model changes in morphology, as exemplified by
CAESAR, or use the potential rates of change in inundation determined from studies
like this one to create suitable ranges for error and uncertainty.
8
CONCLUSIONS
8.1
The analysis of flood gauging records for the study catchments has revealed
significant increases in the frequency and magnitude of floods affecting some Welsh
rivers since the mid 1980s. This appears to be a response to recent climate change
leading to increased winter precipitation, though whether this records either the
predicted effects of global warming, or ‘normal’ long-term climatic variation is unclear. In
either case, the discovery that flood frequencies and magnitude can change markedly
over time, highlights the difficulty of trying to predict what constitutes an event of a given
return period (or probability) such as the ‘100 year’ and ‘1000 year’ flood events,
undermining the reliance placed on these concepts as an underpin to medium term
planning policy and flood defence strategies.
8.2
All the river reaches examined as part of the project display evidence of
significant geomorphological change over time. This may take the form of incision and
river terrace formation, or channel migration and abandonment, or deposition both
within river channels and on adjacent floodplains. These changes date from the prehistoric period through to the present day. In several of the studied reaches the last
phase of marked modification pre-dated the early 19th century, since when river
channels have remained relatively stable. However, other reaches, notably at Caersws,
42
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
and elsewhere in Wales, display evidence of marked lateral shifts in channel position
within the last few decades.
14C dating results show that, in the past, periods of pronounced valley floor
8.3
modification coincided with periods of increased flood activity associated in turn with
changes in the Welsh climate, or land-use, or both. This suggests that the recent
increase in frequency and magnitude of flood events noted in many of the study
catchments, coupled with predicted changes to the climate, will also herald a period of
renewed floodplain modification.
8.4
Geomorphological changes to Welsh river floodplains exert a strong influence on
the location and severity of flood hazard and will do so in the future.
8.5
In limiting the lateral migration of channels and acting as barriers to flood waters,
man-made structures such as road, railway and flood embankments clearly influence
the patterns of geomorphological change across a valley floor and consequently future
flood hazard. In promoting high rates of deposition adjacent to river channels, some of
these structures may act significantly to increase the future flood hazard for other parts
of the adjacent floodplain.
8.6
The availability and use of LiDAR data significantly enhances geomorphological
investigation of river floodplains. The techniques developed as part of this study to
maximise the potential of LiDAR data to image floodplain morphology has important
applications in studies of present and future flood hazard.
8.7
Hydraulic modelling procedures which view valley floors as fixed and
unchanging, cannot provide realistic indications of the inundation limits, or the depths
and speeds of flow of future floods and, hence, of the risks they may pose. Only
dynamic modelling software such as CAESAR which predicts and takes account of
changes in river channels and floodplains can provide more realistic indications of areas
where the medium- and long-term risks arising from flooding are likely to be greatest.
8.8
CAESAR modelling undertaken as part of the project predicts that changes in
floodplain geomorphology will strongly influence future flood hazard within Welsh river
catchments under a variety of future climate change scenarios including ‘no change’.
Significantly, it suggests that for some reaches the flood hazard posed by small to
moderate sized events will reduce over time, though for other reaches it will increase.
This suggests that the dynamic modelling methodology may have an important role to
play in prioritising flood defence works, as well as in modelling the impact different
forms and locations of defence may have.
8.9
In the study reaches, CAESAR suggests that floods above a certain magnitude
inundate broadly the same areas of the valley floor. However, the frequency, severity
and hence the risks posed by such floods are likely to increase significantly for some
reaches and under some climate change scenarios. Intuitively, these effects are likely to
impact most notably on lowland and tidal settings, but modelling the hydrology of such
reaches is beyond the capability of the current software.
8.10 The CAESAR modelled patterns of channel change suggests that the current
software does not adequately replicate the patterns of lateral migration shown to have
been a major feature of many of the study reaches in the past.
43
Predictive and investigative modelling of flood hazard in Welsh river catchments – VOLUME 1
8.11 The refinement and application of dynamic modelling software such as CAESAR,
underpinned by the landform and flood analysis procedures developed as part of the
project, is a prerequisite of effective and focused flood defence strategies and floodingrelated planning policy in Wales.
9
RECOMMENDATIONS
9.1
There is an urgent need to improve the capability of CAESAR to model lateral
migration of river channels and its impact on future flood hazard.
9.2
Consideration should be given to funding a follow-up project designed to adapt
and test the integrated CAESAR and geomorphic analysis methodology for application
to lowland and tidal reaches of Welsh rivers (e.g. River Usk).
9.3
The use of the integrated CAESAR and geomorphic analysis methodology for
prioritising the construction of flood defences, and for gauging the efficacy of different
forms of flood defence structure, should be investigated.
9.4
Assessors of flood hazard in Wales should be encouraged to test and adopt a
combined landform-based flood assessment and dynamic modelling methodology in the
production of a ‘future flood hazard map’ for Wales for use in landuse planning and in
any forthcoming revision of TAN 15.
9.5
There is an urgent need to have LiDAR data available for all Welsh river
catchments to underpin future flood hazard-related research.
10
ACKNOWLEDGEMENTS
The research undertaken during the course of the project was overseen by a project
steering group, the principal members of the group (and the bodies they represented)
were: Miss Lucy Berry (EA), Dr Paul Brewer (RBDHRG), Dr Stewart Campbell (CCW),
Professor Tom Coulthard (formerly RBDHRG, now University of Hull), Dr Jeremy
Davies (BGS), Mr Tim England (EA), Dr Peter Jones (WAG), Professor Mark Macklin
(RBDHRG), Mr Chris Morgan (WAG), Mr Chris Utley (CCW).
Significant contributions to the project were made by Dr Gez Foster, Miss Anna Jones,
Dr Eric Johnstone, Mr Matt Rowberry and Miss Catherine Swain. The final report has
benefited from the advice and comments of Dr Dick Waters and Professor Martin
Culshaw, both of BGS. The 14C dating undertaken as part of the project was separately
funded via BGS University Collaboration Contract GA/02E/01. Dr Huw Sheppard
assisted with the early phase of the 14C sampling programme.
REFERENCES
A consolidated list of references for both volumes of the report is included at the end of
Volume 2.
44