« Previous | Contents | Next »
Listen
2 STUDY METHODOLOGY
2.1 STUDY STAGES
Three stages of work were undertaken to support the
preparation of this report. They were:
- Analysis of Weather Event Implications
- Identification of Potential Trends in Climate
Change in Weather Event Categories
- Evaluation of Potential Trends in Terms of Impact
on the Road Network
2.2 ANALYSIS OF WEATHER EVENT
IMPLICATIONS
In order to identify the weather events of particular
importance to this study, an initial analysis was
undertaken to consider how weather events influence aspects
of design and operation of the road network. The results of
this assessment are presented in section 3 of this report.
This analysis was reviewed at a workshop involving key
stakeholders, to confirm that the areas identified and the
further assessments proposed were appropriate.
2.3 IDENTIFICATION OF PREDICTED TRENDS IN
CLIMATE CHANGE IN WEATHER EVENT CATEGORIES
2.3.1 Introduction
Extensive information is available in relation to
predicted trends in climate change. However, this is
recognised as an area of research where the understanding
and modelling of predicted behaviour is continuously being
updated and improved. To confirm that the information used
for this study was appropriate, evaluation of existing data
sources was undertaken to determine any additional
requirements. The methodology underpinning climate change
prediction is described more fully below.
2.3.2 Background
In 1997 the Department of Environment, Transport and the
Regions (
DETR) established the
UK Climate Impacts Programme (
UKCIP). This provides an umbrella
organisation to oversee and facilitate integration of the
numerous recent and on-going sectorial and regional studies
of climate change impacts in the
UK. Underpinning the
UKCIP is the development of climate
change scenarios for the
UK by the Climate Research Unit at the
University of East Anglia and the Hadley Centre of the Met
Office. The Hadley Centre develops and runs the models used
to generate predicted future trends in weather taking
account of climate change. Their work provides an in-depth
analysis of possible future changes to
UK climate over the next 100 years and
is generally accepted as the most authoritative guidance
available within the
UK.
In 2002 the
UK Climate Impacts Programme published
the latest set of Climate Change Scenarios for the United
Kingdom (
UKCIP02). The climate model results
contained within this report were from the Met Office's
Hadley Centre, based upon findings from the latest regional
climate model
HadRM3. This model adopts the
Intergovernmental Panel on Climate Change (
IPCC) scenarios of change, based upon a
range of possible emission outputs. All of these scenarios
can be considered equally likely to occur, and are termed
'Low', 'Medium-Low', 'Medium-High' and 'High'.
The
UKCIP02 report details the latest
scientific understanding and modelling results for climate
change predictions, which are presented for three
'time-slices' or thirty year periods centred about the
2020's, 2050's and 2080's. However, there are a number of
sources of uncertainty, due to emissions scenarios,
scientific uncertainties and natural variability, which
should be borne in mind when interpreting the predicted
future
UK climate. These are discussed in
Chapter 7 of
UKCIP02, and are summarised briefly
below. Some of these uncertainties account for the
differences between the 1998 and 2002
UKCIP scenarios and the Regional Climate
Change Scenarios for Scotland report (Hulme et al, 2001),
as also discussed below.
Global climate models (
GCMs) are currently the only scientific
tool available for predicting changes in future climate
and, in particular, the large scale patterns of change.
These models require a large computing resource and hence
are run at relatively coarse resolution. For example, the
latest version of the Hadley Centre global model,
HadCM3, represents the British Isles
with a horizontal resolution of approximately 300km, which
means that only two grid squares cover Scotland. For
greater detail, and for impact studies, higher resolution
regional models can be 'embedded'. This means that a
regional climate model (
RCM), which represents a limited area of
the globe at higher resolution, takes all its information
for the surrounding area from a global climate model. The
version of the Hadley Centre regional model used to produce
the
UKCIP02 scenarios,
HadRM3, has a horizontal resolution of
50km. The regional model prediction shows much more detail
than the global model and is able to better represent
extremes.
An example of this is presented in Figure 2.1, which
shows a comparison of observed and modelled summertime
rainfall over the UK, with the rainfall being categorised
by intensity. The modelled figures are produced from both a
global model, such as that used in
UKCIP98, and a regional model of the
resolution used for the
UKCIP02 scenarios. At low intensities of
rainfall, less than 10 mm/day, the
GCM predicts amounts comparable with
observations and the
RCM. However, as the threshold is
increased the ability to predict intense rainfall is
greatly reduced in the
GCM, while the
RCM compares well with observations into
the 30 mm/day range.
Figure 2.1 - Comparison of Observations,
Regional Climate Model (
RCM) predictions and Global Climate
Model (
GCM) predictions for
UK summer rainfall. Source: Met
Office.
Regional models are able to better capture extreme
weather because not only can they represent smaller scale
weather features but they also have a much better
representation of orthography. For example, the mountains
of Scotland cannot be represented in the global model at
all as the entire country is represented by two grid boxes.
In the
UKCIP02 scenarios, the 50km
RCM permits the mountains of Scotland to
be partially resolved, resulting in increased spatial
variation of rainfall. The model used for the British Irish
Council (
BIC) report (Jenkins et al, 2003) was a
single run of a 25 km
RCM, the finer resolution resulting in
an ability to represent the varied topography of Scotland
more accurately. The increased resolution also permitted
some of the Scottish islands to be resolved and hence
included in climate modelling studies for the first time.
This can be seen in Figure 2.2 where average winter
precipitation amounts (mm/day) by the 2080's under the
Medium-High emissions scenario are shown for the 50km and
25 km models.
Figure 2.2 - Comparison between
HadRM3, 50km model, left, and the
BIC 25km model, right, for winter
precipitation amounts (mm/day) by the 2080's with the
Medium-High emission scenario. Source: Met
Office.
It is clear from the above figure that the winter
precipitation increases predicted in the Scottish mountains
and along the west coast are much better resolved in the
higher resolution model, with greater spatial detail.
Patterns and scale of change predicted by this model are
comparable with those in
UKCIP02. This model has not as yet been
used in studies of extremes, except for a limited study of
changes predicted in extreme rainfall and temperature for
some of the Scottish Islands (Shetland, Orkney and the
Western Isles, Jenkins et al 2003). In this study it was
shown that the predicted changes in extremes are comparable
with those derived from the
UKCIP02 scenarios over the mainland.
This study will therefore refer to both of these models
where their outputs are available.
2.3.3 Climate Change Modelling
As discussed previously,
UKCIP02 provides four climate scenarios;
Low, Medium-Low, Medium-High and High. To provide guidance
on predicted climate scenarios, climate change trends have
been examined for the Medium-High scenario where this
information is available. This reflects the greater level
of detail that is available for this scenario from the
models.
Every climate model, global or regional, produces a
wealth of data. To achieve maximum value this data must be
used in the context of the uncertainties, as detailed in
Section 2.3.4 of this report. It should also be noted that
not all climate model data has been validated against
observed climate. In many cases it is not possible to
validate the modelled present day climate because
comparable observed data sets of sufficient length do not
exist. The availability of satellite and radar data will
continue to help resolve this issue as the length of
datasets available from these sources increases. The large
array of modelling studies which have been completed and
published do, however, increase confidence and
understanding of model predictions and the degree of
uncertainty.
As a regional model is forced at its geographical
boundaries by data from the global model, any large scale
patterns of change are established in the regional model
from this data. Hence regional models are run for
time-slices to provide regional details lacking in the
global model studies. In addition, techniques exist to
scale the predicted patterns of change at seasonal and
monthly average periods to other 'time-slices' or periods
of time, based upon changes predicted by the global model.
However, it is not possible to scale extreme events and
their frequency of occurrence in the same way, especially
the very low frequency events with high return periods.
Nevertheless, a useful insight into the evolution of
extreme events, from the present day to the end of the
century, can be gained by combining knowledge of the
extremes predicted by the regional model with knowledge of
longer term variability and changes in large scale flow
from the global model.
All daily data are available from the global model,
HadCM3, from the present day to 2100 and
beyond in some cases, under the four emissions scenarios.
Some sub-daily data is available, including that relating
to precipitation. However, as Scotland is represented by
two grid boxes in the global model this is likely to be of
little practical use for this study. The
RCM, both 50km and 25km, has been run
for two thirty-year periods, a present day time-slice (1961
to 1990) and a future time-slice (2071 to 2100). These
represent the current and projected future climates and
daily model data is available throughout both of these
periods. The use of the 1961 to 1990 period for the present
day climate is consistent with other recent studies on
climate change. Limited validation of
HadRM3 sub-daily precipitation was
completed as part of the recent
UKWIRCL/10 study, 'Climate Change and the
Hydraulic Design of Sewerage Systems'. A validation of
12-hourly precipitation against observed rainfall gauge
data, including sites in Scotland, showed that
HadRM3 successfully captured the
observed regional variation in rainfall and many of the
characteristics of its statistical distribution.
2.3.4 Uncertainties in Climate
Modelling
Three areas of uncertainty are outlined below:
- Emissions Scenarios, which vary depending on how
future emissions of gases are considered likely to
occur
- Scientific Uncertainty, which underlies the ability
of the current modelling capability and knowledge of
climate behaviour to provide a reasonable estimate of
future trends
- Natural Variability, which results in changes over
time
Emissions Scenarios
Future anthropogenic emissions of gases which alter
climate will depend upon the way in which society evolves.
There is no way of knowing how population, technology,
economic growth, etc, will develop and thus 'scenarios' of
future development have to be constructed. The
IPCC Special Report on Emission
Scenarios (
SRES,
IPCC, 2000) considered four plausible
'narrative storylines'. From these it produced details of a
wide range of possible future emissions scenarios, where
each scenario associated with a particular 'storyline' is
considered to be a member of that 'family'. Each 'family'
generated a 'marker scenario', which was considered to be
representative of that 'storyline'. This is illustrated in
Figure 2.3, which shows the
CO2 emissions for each of these 'marker
scenarios'.
Figure 2.3 - Emissions of CO2 (giga-tonnes
carbon per year) in four emissions scenarios. Source:
Hadley Centre Technical Note 44.
This set of marker scenarios were used in the
UKCIP02 report. As a rough guide, the
'Medium-High' emissions scenario can be considered to
approximate to a 'business as usual' evolution. However, it
must be noted that no likelihood of occurrence can be
placed upon any of the storylines of emissions scenarios.
Although the scenarios diverge very quickly, all are
plausible and while they may not be equally probable they
are possible, therefore none can be discounted. This is why
predictions should always be qualified by saying which
scenario was used in producing the relevant result.
As predicted climate changes depend upon the scenario
used it is not surprising that a range of possible futures
results. For example, when considering the global mean
temperature change predicted by a single global model it
becomes obvious that in fact the choice of scenario
actually has very little impact over the next fifty years.
It is only in the second half of this century that
predictions begin to diverge. This is because of the large
thermal inertia of the climate system and the long lifetime
of
CO2. This means that the changes which
the global climate will experience over the next few
decades are already programmed into the climate system as a
result of the emissions of recent decades.
While using a range of scenarios appears to introduce
uncertainties into predictions of future climate, in this
instance it actually adds a level of certainty to predicted
changes by particular models up to the middle of the
century. In the case of the Hadley Centre's global climate
model,
HadCM3, the global mean warming is
predicted to be approximately 1¡C under all scenarios by
2040. After this, predictions diverge and uncertainty due
to evolution of emissions and choice of scenario manifests
itself. As previously stated, probabilities cannot be
ascribed to these scenarios but the suite of scenarios
permits the range of possible futures to be scoped.
Scientific Uncertainties
The ability of models to represent predicted climate
change is one of the largest areas of scientific
uncertainty. For example, the
IPCC Third Assessment Report (
IPCCTAR, 2001) included predictions from
over thirty models of varying complexity. Each model was
run using the same
SRES scenarios but each model predicts a
different future, and some of the differences are large,
much greater than the range introduced by the choice of
emissions scenario.
These differences are due to the way the models each
represent the globe, the processes which are included and
the manner in which they are parameterised. Some models
have a finer resolution than others and include more
complex physical processes. It is not possible at this time
to say how credible each model is because evaluation is not
simple and must always include some level of subjectivity.
It is therefore not presently possible to discount any of
the models and even the more extreme predictions could be
underestimating what will occur. This could be due to the
existence of a vital but as yet undiscovered feedback
process, which is not represented in the current
models.
The range of futures suggested by the different models
is one of the largest uncertainties in prediction of
climate change. However, when model predictions are
consistent, increased confidence can be placed in the
result. For example, all models predict global warming
under all scenarios. Difficulties arise when there is a low
level of consistency between model results. This is
illustrated in Figure 2.4 where changes in winter
precipitation over the
UK, as predicted by nine different
global climate models, are compared. The models presented
here all predict an increase in winter precipitation over
Scotland, but the range of predicted change is quite large,
from just greater than 0% to over 50%. The outputs should
not be confused with the predictions of finer scale
regional models, such as that used to provide the
UKCIP02 predictions. These outputs also
demonstrate the effects of natural variability, as three of
the predictions, references
UKMO A2,
UKMO A2#2 and
UKMO A2#3, are from the Hadley Centre
Model
HadCM3.
Figure 2.4 - Change in winter precipitation (%)
in the
UK, from present day to 2080's,
drawn from nine global climate models, using the
Medium-High emissions scenario. Source: Met
Office.
Given the nature of weather patterns across the
UK it is perhaps unsurprising that it is
a challenging area to model. For example, a slight shift in
the location of a pressure system over the Atlantic can
mean that storms may track to the north or south rather
than travelling across Scotland. Work is continuing to
validate the representation of current climate by the
models and to find plausible reasons for the different
climate prediction outputs by models.
The impact of scientific uncertainties can be
demonstrated by comparing the results of the
UKCIP98 report, the 2001 report by Hulme
et al and the
UKCIP02 report. Each of these
publications was based upon Hadley Centre modelling results
but in each case a different model was used. For example,
comparison of winter precipitation shows a predicted
increase over Scotland, and the increase was seen to
persist throughout the year, with the summer months also
experiencing more rainfall under most of the emissions
scenarios by the 2080's. It should be noted however that
these results came from a global climate model in which
Scotland was represented by only two model grid
squares.
The
UKCIP98 report was followed in 2001 by
'An exploration of regional climate scenarios for Scotland'
(Hulme et al 2001) presenting results from
HadRM2, the Hadley Centre's regional
version of the model used in
UKCIP98. Although this model was
scientifically the same as the model used in
UKCIP98 it had a much greater horizontal
resolution of fifty kilometres, so spatial detail of the
predicted climate changes over Scotland became
possible.
It was found that results were broadly similar to those
found in the
UKCIP98 report, although the spatial
resolution permitted an east-west contrast in precipitation
changes to be identified. In particular, the greatest
increase in precipitation was predicted to occur over the
Western Highlands, which saw increased precipitation
throughout the year. However, the model also predicted
decreased rainfall over eastern Scotland during summer
months, a result not seen in the
UKCIP98 report, as the global model
could not resolve these spatial differences. The study also
found relatively little difference in the predicted change
to precipitation return periods between the regional and
global models, although the intensity of rainfall is
greater in
HadRM2.
The results contained within the
UKCIP02 report came from the Hadley
Centre's latest regional climate model,
HadRM3, based upon the global model
HadCM3, the successor to the model used
in
UKCIP98. Although
HadCM3 is scientifically more complex
than its predecessor, it cannot be argued that it is
scientifically more valid than
HadCM2. The models include different
parameterisations and thus predict slightly different
regional patterns of climate change. This is to be
expected. The reports also use slightly different time
periods (2081 to 2100 in
HadRM2 versus 2071 to 2100 in
HadRM3) and different emissions
scenarios.
All of the differences are discussed in the
UKCIP02 report, but two major
differences are noted here. The first of these is in
summertime rainfall across Scotland, with the latest model
predicting a widespread drying when
HadRM2 indicated generally wetter
summers. The second of these is in autumn rainfall across
Scotland, with the latest model predicting no change from
the present when
HadRM2 indicated greatly wetter
conditions. While some of the differences can be explained
by natural variability, the major contributing factor is
the different large scale circulation patterns established
in each model. Although many general features of the
simulated climate are similar between the models, each
predicted climate is modulated locally and seasonally by
changing patterns of large scale air flow.
The Hadley Centre model,
HadCM3, is globally respected among the
climate modelling community and is an established world
leader. Results from the model generally fall within the
mid-range of predictions rather than the extremes. In this
study the impacts have been assessed in the context of the
results from the Hadley Centre models. However, a full
appreciation of the confidence which can be placed in any
prediction, which is beyond the scope of this report,
should consider the full range of
IPCC models.
Each of the model integrations present one possibility
of what the future may hold, each being completed with one
of a range of scenarios of emissions. Fundamentally,
however, the model remains the same and uncertainty due to
choices made within the scientific parameterisations
remain. One way to combat this uncertainty is to run a
large 'ensemble' of integrations, where each aspect of the
physics can be tested within a realistic range of possible
values, thus each model in the ensemble has a different but
equally plausible representation of the climate system. As
an integration is completed a solution is produced, giving
one point in the range of possible outcomes. Each
integration will generate its own solution, thereby
creating more points within the range of possible
solutions. When enough integrations have been completed a
frequency distribution of solutions begins to emerge. This
is termed a 'physics' ensemble and the output is a
probability distribution of change in the quantities of
interest, such as winter rainfall over Edinburgh.
The technique is described graphically in Figure 2.5.
The left panel represents the outcome of nine
GCM integrations with slightly
different, all equally plausible, physics, resulting in a
range of 35% reduction to a 5% increase in the predicted,
hypothetical, quantity. The right panel is the probability
distribution function derived from a large ensemble of the
GCM, predicting the same range but a
mean predicted value of 10% below present day values.
Figure 2.5 - Graphical representation of a
physics ensemble. The left panel illustrates the
'current' method of climate prediction and the right
panel presents the outcome using a physics ensemble.
Source: Met Office.
This technique is relatively new, as it requires massive
computing resources in order to be successful. An ensemble
of models is currently being run at the Hadley Centre but,
while early indications are encouraging, there are
insufficient members in the ensemble at present to make
reliable probability predictions possible.
Natural Variability
The Earth's climate varies naturally, with the climate
system's internal variability providing year to year and
decade to decade change. It is highly likely that at some
time in the future there will be periods when this natural
variability combines with anthropogenic climate change to
produce a period of extreme warming or summer drying. It is
equally likely that the two factors will combine at another
time to produce a relatively cold or dry winter. For this
reason it is important to consider natural variability when
looking at predicted mean climate change.
The impact of this is demonstrated in Figure 2.6, which
shows the evolution of winter mean precipitation over
Scotland from one integration of
HadCM3, the global model, forced with
the Medium-High emissions scenario. Although there is a
trend towards wetter winters, there are still individual
years or short periods when winter precipitation is below
the model's present day climate, even towards the end of
the period when average changes typically exceed 20%. Note
that this is one evolution of the model and is purely to
demonstrate the impact of natural variability. It is not a
year on year prediction of rainfall and should not be taken
to be a forecast.
Figure 2.6 - Winter mean precipitation change
(%), relative to 1961 to 1990 average, over Scotland,
Medium-High emissions scenario, one
HadCM3 ensemble member. Each bar is
a winter mean, the red curve is a longer term running
mean, and * denotes a model year which is 'record
breaking'. Source: Met Office.
One technique to address the uncertainty due to natural
variability as simulated within climate models is to run an
ensemble of integrations. In much the same way as a physics
ensemble works, the starting conditions of a model can be
'perturbed' slightly, which sets a model off on a slightly
different, but equally plausible, predictive pathway. This
technique effectively introduces new 'weather' to the
starting conditions, by means of a small but feasible
alteration of starting conditions. The difference which
results from predictions by members of such an ensemble, in
which each member has identical physics, is the modelled
representation of natural variability.
The majority of mapped results presented in the
UKCIP02 report are for an ensemble mean.
Three integrations with slightly perturbed starting
conditions were completed for the control period (1961 to
1990) and for the 2080's time-slice with the Medium-High
emissions scenario. This means that three sets of thirty
year integrations are available for each period. In this
way the simulated variability of the modelled present day
climate can be more fully assessed and the impact of
climate change on variability more fully captured than
could be achieved with a single ensemble member.
The single 25km integration used for the
BIC report is not part of an ensemble
and hence a number of uncertainties exist due to the lack
of other ensemble members. However, in some ways it can be
considered a fourth ensemble member to the
UKCIP02 integrations, and results from
this model should be considered within the context of the
UKCIP02 results.
2.4 EVALUATION OF PREDICTED TRENDS IN TERMS OF
IMPACT ON THE ROAD NETWORK
The purpose of the study was to establish what the
likely impacts of climate change would be on the Scottish
road network, to enable identification of appropriate
responses. To achieve this objective, the trends identified
were considered in relation to the current guidance on the
use of weather event data in the design and operation of
the road network. This consideration also included
consultations with parties responsible for managing the
road network. In order to provide a broad range of
geographical experience, together with views on roads
ranging from major urban motorways to rural single
carriageways, this consultation was undertaken with the two
parties responsible for maintaining the North-West,
North-East, South-West and South-East trunk road units on
behalf of the Scottish Executive. In addition, in response
to emerging findings from the study process, consultations
were undertaken with
NADICS.
2.5 CLIMATE CHANGE MODELS USED IN THIS
STUDY
Table 2.1 provides a list of the models used to assess
the predicted trends in climate change presented in this
report.
Table 2.1 - Climate Change Models Used in this
Study
Weather Variable | Climate Change Models Used |
|---|
Temperature | HadRM3: 50km and 25km
Grid |
|---|
Precipitation | HadRM3: 50km and 25km Grid,
HadRM2 |
|---|
Snow | HadRM3: 50km and 25km
Grid |
|---|
Wind | HadRM3: 50km Grid |
|---|
Fog | HadRM3: 50km Grid |
|---|
Coastal Flooding | HadRM3: 50km Grid |
|---|
« Previous | Contents | Next »