<<

. 5
( 13)



>>

The accuracy and the coverage of data available for these past pe-
riods are limited. However, the model simulations for 9000 years ago,
described above, and those for other periods in the past have demon-
strated the value of such studies in the validation of climate models.
A third way in which models can be validated is to use them to predict
the effect of large perturbations on the climate. Good progress is being
achieved with the prediction of El Ni˜ o events and the associated climate
n
anomalies up to a year ahead (see earlier in the chapter). Other short-term
perturbations are due to volcanic eruptions, the effects of which were
mentioned in Chapter 1. Several climate models have been run in which
the amount of incoming solar radiation has been modi¬ed to allow for the
effect of the volcanic dust from Mount Pinatubo, which erupted in 1991
(Figure 5.21). Successful simulation of some of the regional anomalies
of climate which followed that eruption, for instance the unusually cold
winters in the Middle East and the mild winters in western Europe, has
also been achieved by the models.15
102 Modelling the climate




Modelling of tracers in the ocean

A test that assists in validating the ocean compo- of tritium (in tritium units) in a section of the west-
nent of the model is to compare the distribution of a ern north Atlantic Ocean about a decade after the
chemical tracer as observed and as simulated by the major bomb tests and the distribution as simulated
model. In the 1950s radioactive tritium (an isotope by a twelve-level ocean model. Similar comparisons
of hydrogen) released in the major atomic bomb have been made more recently of the measured up-
tests entered the oceans and was distributed by the take of one of the freons CFC-11, whose emissions
ocean circulation and by mixing. Figure 5.20 shows into the atmosphere have increased rapidly since the
good agreement between the observed distribution 1950s, compared with the modelled uptake.




Figure 5.20 The tritium distribution in a section of the western north Atlantic Ocean approximately one
decade after the major atomic bomb tests, as observed in the GEOSECS programme (a) and as
modelled (b).



In these three ways, which cover a range of timescales, con¬dence
has been built in the ability of models to predict climate change due to
human activities.


Comparison with observations
More than ¬fteen centres in the world located in ten countries are cur-
rently running climate models of the kind we have described in which the
circulations of the atmosphere and the ocean are fully coupled together.
Some of these models have been employed to simulate the climate of the
last 150 years allowing for variations in aspects of natural forcing (e.g.
solar variations and volcanoes) and the increases in the concentrations
of greenhouse gases and aerosols.
Comparison with observations 103



Figure 5.21 The
predicted and observed
changes in global land and
ocean surface air
temperature after the
eruption of Mount
Pinatubo, in terms of
three-month running
averages from April to June
1991 to March to May
1995.




An example of such simulations is shown in Figure 5.22, where the
observed record of global average surface air temperature is compared
with model simulations taking into account in turn natural forcings, an-
thropogenic forcings (i.e. the increase in greenhouse gases and aerosols)
and the combination of natural and anthropogenic forcings. Although
the simulations in Figure 5.22 are based on one model, similar results
have been obtained from many models.
Three interesting features of Figure 5.22 can be noted. Firstly, that the
inclusion of anthropogenic forcings provides a plausible explanation for
a substantial part of the observed temperature changes over the last cen-
tury (especially for the latter part of the century), but that the best match
with observations occurs when both natural and anthropogenic factors
are included. In particular it is likely that changes in solar output and
the comparative absence of volcanic activity were important variations
in natural forcing factors during the ¬rst part of the twentieth century.
Secondly, the model simulations show variability up to a tenth of a degree
Celsius or more over periods of a few years up to decades. This variabil-
ity is due to internal exchanges in the model between different parts of
the climate system, and is not dissimilar to that which appears in the ob-
served record. Thirdly, due to the slowing effect of the oceans on climate
change, the warming observed or modelled so far is less than would be
expected if the climate system were in equilibrium under the amount
of radiative forcing due to the current increase in greenhouse gases and
aerosols.
There remains, however, a large amount of natural variability in both
the observations and the simulations and much debate has taken place
over the last decade or more about the strength of the evidence that
global warming due to the increase in greenhouse gases has actually
104 Modelling the climate



Simulated annual global mean surface temperatures
(a) Natural (b) Anthropogenic
1.0 1.0




Temperature anomalies ( C)
model model
Temperature anomalies ( C)




observations observations
0.5 0.5


0.0 0.0


’0.5 ’0.5


’1.0 ’1.0
1850 1900 1950 2000 1850 1900 1950 2000
Year Year


Figure 5.22 Annual global (c) All forcings
mean surface temperatures
1.0
Temperature anomalies ( C)




simulated by a climate model model
observations
compared with observations
0.5
for the period 1860“2000.
The simulations in (a) were
done with only natural
0.0
forcings “ solar variation and
volcanic activity; in (b) with
’0.5
only anthropogenic forcings “
greenhouse gases and
sulphate aerosols; and in (c)
’1.0
with both natural and 1850 1900 1950 2000
anthropogenic forcings Year
combined. The simulations are
shown in a band that covers
the results from four runs with
the same model and therefore
been observed in the climate record. In other words has the ˜signal™ that
illustrates the range of natural
variability within the model. can be attributed to global warming risen suf¬ciently above the ˜noise™
of natural variability? The Intergovernmental Panel on Climate Change
(IPCC) has been much involved in this debate.
The IPCC™s ¬rst Report published in 199016 made a carefully worded
statement to the effect that, although the size of the observed warming
is broadly consistent with the predictions of climate models, it is also of
similar magnitude to natural climate variability. An unequivocal state-
ment that anthropogenic climate change had been detected could not
therefore be made at that time. By 1995 more evidence was available
and the IPCC 1995 Report17 therefore reached the cautious conclusion
as follows.
Comparison with observations 105



Our ability to quantify the human in¬‚uence on global climate is currently
limited because the expected signal is still emerging from the noise of
natural climate variability, and because there are uncertainties in key
factors. These include the magnitude and patterns of long term natural
variability and the time-evolving pattern of forcing by, and response to,
changes in the concentrations of greenhouse gases and aerosols, and land
surface changes. Nevertheless, the balance of evidence suggests a
discernible in¬‚uence on global climate.

Since 1995 a large number of studies have addressed the problems of
detection and attribution18 of climate change. Better estimates of natural
variability have been made, especially using models, and the conclusion
reached that the warming over the last one hundred years is very un-
likely to be due to natural variability alone.19 In addition to studies using
globally averaged parameters, there have been detailed statistical studies
using pattern correlations based on optimum detection techniques ap-
plied to both model results and observations. For example, Figure 5.23
shows a comparison between simulated and observed estimates of zonal
mean temperature change as a function of altitude. Taking these studies




(b)
(a) Observations
GSO
-1.3 -1.2
-0.9
100
100 -0.2
-2.2 0.0 0.0
200 -1.0 0.2
200 0.0
-0.4
0.4 -0.2
-0.2 0.4
300
Pressure (hPa)




Pressure (hPa)




300
0.0 0.6
0.2
0.0 -0.2 0.4
0.2
0.4
500
0.0
0.2 0.4
500 0.0
0.2
0.4
0.2
0.4 0.2
0.2
0.4
800
0.2
0.4 800
90 o N 60 o N 30 o N 30 o S 60 o S 90 o S
90 o N 60 o N 30 o N 30 o S 60 o S 90 o S 0
0
Latitude Latitude

-1.2 -0.8 -0.4 0 0.4
-1.2 -0.8 -0.4 0 0.4

Figure 5.23 Simulated and observed zonal mean temperature change as a
function of latitude and height. The changes plotted are the differences between
the 1986“1995 decadal average and the twenty-year average from 1961 to
1980. The contour interval is 0.1 —¦ C. (a) simulated changes taking into account
increases in carbon dioxide, sulphate aerosols and the effect of observed
changes in stratospheric ozone; (b) observed changes. A common pattern of
stratospheric cooling and tropospheric warming is evident in the observations
and both model experiments; the stratospheric cooling is partially due to the
increase of carbon dioxide (see also Figure 4.2(b)) and the reduction in ozone
(see Chapter 3).
106 Modelling the climate



into account, the conclusion reached in the IPCC 2001 Report20 is as
follows.

In the light of new evidence and taking into account the remaining
uncertainties, most of the observed warming over the last 50 years is likely
to have been due to the increase in greenhouse gas concentrations.

Con¬dence having been established in climate models in the ways we
have outlined in the last two sections, these models can now be used
to generate projections of the likely climate change in the future due
to human activities. Details of such projections will be presented in the
next chapter.
Before leaving comparison with observations, I should mention
some recent work relating to the warming of the ocean that adds fur-
ther con¬rmation to the picture that has been presented. In Chapter 2,
the effect of an increase of greenhouse gases was expressed in terms
of radiative forcing or, in other words, a net input of heat energy into
the earth-atmosphere system. Most of this extra energy is stored in the
ocean. The amount of this extra energy has been estimated from measure-
ments of the temperature increase in the ocean at different locations and
depths down to 3 km. From 1957 to 1994 the estimate is (19 ± 9) — 1022
joules, the uncertainty relating mostly to inadequate sampling of some
large ocean areas.21 This amount is a little over 0.5% of the solar energy
received by the Earth over this period. Within the limits of uncertainty, it
agrees well with the estimates of radiative forcing presented earlier and
also with model estimates of ocean heat uptake.22



Is the climate chaotic?
Throughout this chapter the implicit assumption has been made that
climate change is predictable and that models can be used to pro-
vide predictions of climate change due to human activities. Before
leaving this chapter I want to consider whether this assumption is
justi¬ed.
The capability of the models themselves has been demonstrated so
far as weather forecasting is concerned. They also possess some skill in
seasonal forecasting. They can provide a good description of the current
climate and its seasonal variations. Further, they provide predictions
which on the whole are reproducible and which are reasonably consistent
between different models. But, it might be argued, this consistency could
be a property of the models rather than of the climate. Is there any other
evidence to support the view that the climate is predictable?
Regional climate modelling 107



A good place to look for further evidence is in the record of cli-
mates of the past, presented in Chapter 4. Correlation between the
Milankovitch cycles in the Earth™s orbital parameters and the cycles
of climate change over the past half million years (see Figures 4.4 and
5.19) provides strong evidence to substantiate the Earth™s orbital varia-
tions as the main factor responsible for the triggering of climate change.
The nature of the feedbacks which control the very different amplitudes
of response to the three orbital variations still need to be understood.
Some 60 ± 10% of the variance in the record of global average tem-
perature from paleontological sources over the past million years occurs
close to frequencies identi¬ed in the Milankovitch theory. The exis-
tence of this surprising amount of regularity suggests that the climate
system is not strongly chaotic so far as these large changes are con-
cerned, but responds in a largely predictable way to the Milankovitch
forcing.
This Milankovitch forcing arises from changes in the distribution
of solar radiation over the Earth because of variations in the Earth™s
orbit. Changes in climate as a result of the increase of greenhouse gases
are also driven by changes in the radiative regime at the top of the
atmosphere. These changes are not dissimilar in kind (although different
in distribution) from the changes that provide the Milankovitch forcing.
It can be argued therefore that the increases in greenhouse gases will
also result in a largely predictable response.


Regional climate modelling
The simulations we have so far described in this chapter are with global
circulation models (GCM) that typically possess a horizontal resolu-
tion (grid size) of around 300 km “ the size being limited primarily by
the availability of computer power. Weather and climate on scales large
compared with the grid size are described reasonably well. However, at
scales comparable with the grid size, described as the regional scale,23
the results from global models possess serious limitations. The effects
of forcings and circulations that exist on the regional scale need to be
properly represented. For instance, patterns of precipitation depend crit-
ically on the major variations in orography and surface characteristics
that occur on this scale (see Figure 5.24). Patterns generated by a global
model therefore will be a poor representation of what actually occurs on
the regional scale.
To overcome these limitations regional modelling techniques have
been developed.24 That most readily applicable to climate simulation
and prediction is the Regional Climate Model (RCM). A model covering
an appropriate region at a horizontal resolution of say 25 or 50 km can
108 Modelling the climate




Figure 5.24 (a) Representation of the Philippines in RCMs with resolutions of
25 km and 50 km and in a GCM with 400 km resolution; (b) patterns of
present-day winter precipitation over Great Britain, (i) as simulated with a
300 km resolution global model, (ii) with 50 km resolution regional model,
(iii) as observed with 10 km resolution.



be ˜nested™ in a global model. The global model provides information
about the response of the global circulation to large-scale forcings and
the evolution of boundary information for the RCM. Within the region,
physical information, for instance concerning forcings, is entered on the
scale of the regional grid and the evolution of the detailed circulation
is developed within the RCM. The RCM is able to account for forcings
on smaller scales than are included in the GCM (e.g. due to topography
or land cover inhomogeneity, see Figure 5.24) and can also simulate
atmospheric circulations and climate variables on these smaller scales.
The future of climate modelling 109



A limitation of the regional modelling technique we have described
is that, although the global model provides the boundary inputs for the
RCM, the RCM provides no interaction back on to the global model. As
larger computers become available it will be possible to run global mod-
els at substantially increased resolution so that this limitation becomes
less serious; at the same time RCMs will acquire an ability to deal with
detail on even smaller scales. Some examples of regional model simu-
lations are given in Chapter 6 (Figure 6.10).
Another technique is that of Statistical Downscaling that has been
widely employed in weather forecasting. This uses statistical methods to
relate large scale climate variables (or ˜predictors™) to regional or local
variables. The predictors from a global circulation climate model can
be fed into the statistical model to estimate the corresponding regional
climate characteristics. The advantage of this technique is that it can
easily be applied. Its disadvantage from the point of view of simulating
climate change is that it is not possible to be sure how far the statistical
relations apply to a climate-changed situation.


The future of climate modelling
Very little has been said in this chapter about the biosphere. Chapter 3 re-
ferred to comparatively simple models of the carbon cycle which include
chemical and biological processes and simple non-interactive descrip-
tions of atmospheric processes and ocean transport. The large three-
dimensional global circulation climate models described in this chapter
contain a lot of dynamics and physics but no interactive chemistry or
biology. As the power of computers increases, global dynamical and
physical circulation models that couple in the biological and chemical
processes that make up the carbon cycle and the chemistry of other gases
are now being developed. Before very long we can expect that models
will be available that are fully interactive and comprehensive in their
inclusion of dynamical, physical, chemical and biological processes in
the atmosphere, the ocean and on the land.
Climate modelling continues to be a rapidly growing science. Al-
though useful attempts at simple climate models were made with early
computers it is only during the last ten years or so that computers have
been powerful enough for coupled atmosphere“ocean models to be em-
ployed for climate prediction and that their results have been suf¬ciently
comprehensive and credible for them to be taken seriously by policy
makers. The climate models which have been developed are probably
the most elaborate and sophisticated of computer models developed in
any area of natural science. Further, climate models that describe the
natural science of climate are now being coupled with socio-economic
110 Modelling the climate



information in integrated assessment models (see box in Chapter 9,
page 237).
As the power of computers increases it becomes more possi-
ble to investigate the sensitivity of models by running a variety of
ensembles that include different initial conditions, model parameteri-
sations and formulations. A particularly interesting project25 involves
thousands of computer users around the world in running state-of-the-
art climate prediction models on their home, school or work computers.
By collating data from thousands of models it will generate the world™s
largest climate modelling prediction experiment.
A great deal remains to be done to narrow the uncertainty of model
predictions. The ¬rst priorities must be to improve the modelling of
clouds and the description in the models of the ocean“atmosphere inter-
action. Larger and faster computers are required to tackle this problem,
especially to enable the resolution of the model grid to be increased, as
well as more sophisticated model physics and dynamics. Much more
thorough observations of all components of the climate system are also
necessary, so that more accurate validation of the model formulations can
be achieved. Further, regional climate modelling techniques will develop
rapidly as they are applied to a wide variety of situations. Very substan-
tial national and international programmes are underway to address all
these issues.


Questions

1 Make an estimate of the speed in operations per second of Richardson™s
˜people™ computer. Do you agree with the estimate in Figure 5.1?
2 If the spacing between the grid points in a model is 100 km and there are
twenty levels in the vertical, what is the total number of grid points in a
global model? If the distance between grid points in the horizontal is halved,
how much longer will a given forecast take to run on the computer?
3 Take your local weather forecasts over a week and describe their accuracy
for twelve, twenty-four and forty-eight hours ahead.
4 Estimate the average energy received from the Sun over a square region
of the ocean surface, one side of the square being a line between northern
Europe and Iceland. Compare with the average transport of energy into the
region by the North Atlantic Ocean (Figure 5.16).
5 Take a hypothetical situation in which a completely absorbing planetary
surface at a temperature of 280 K is covered by a non-absorbing and
non-emitting atmosphere. If a cloud which is non-absorbing in the vis-
ible part of the spectrum but completely absorbing in the thermal infrared
is present above the surface, show that its equilibrium temperature will be
235 K (=280/20.25 K).26 Show also that if the cloud re¬‚ects ¬fty per cent of
Notes 111



solar radiation, the rest being transmitted, the planet™s surface will receive
the same amount of energy as when the cloud is absent. Can you substantiate
the statement that the presence of low clouds tends to cool the Earth while
high clouds tend towards warming of it?
6 Associated with the melting of sea-ice which results in increased evapo-
ration from the water surface, additional low cloud can appear. How does
this affect the ice-albedo feedback? Does it tend to make it more or less
positive?
7 Work out the total energy received by the Earth from the Sun over the thirty-
seven-year period from 1957 to 1994; allow for that lost by re¬‚ection and
scattering to space. What precise proportion of this is: (1) the total radiative
forcing over this period due to increased greenhouse gases (see for instance
Figure 3.8) and (2) the energy absorbed by the ocean as derived by Levitus
et al. (page 106)? Comment on your results.
8 It is sometimes argued that weather and climate models are the most sophis-
ticated and soundly based models in natural science. Compare them (e.g. in
their assumptions, their scienti¬c basis, their potential accuracy, etc.) with
other computer models with which you are familiar both in natural science
and social science (e.g. models of the economy).



Notes for Chapter 5
1 Further information regarding the subject of this chapter can be found in the
following texts:
Houghton, J. T. 1991. The Bakerian Lecture, 1991: The predictability of
weather and climate. Philosophical Transactions of the Royal Society, Lon-
don, A, 337, pp. 521“71.
Houghton, J. T. 2002. The Physics of Atmospheres, third edition. Cambridge:
Cambridge University Press.
Houghton, J. T., Jenkins, G. J., Ephraums, J. J. (eds.) 1990. Climate Change:
the IPCC Scienti¬c Assessments. Cambridge: Cambridge University
Press.
Houghton, J. T., Callander, B. A., Varney, S. K. (eds.) 1992. Climate Change
1992: the Supplementary Report to the IPCC Scienti¬c Assessments. Cam-
bridge: Cambridge University Press.
Houghton, J. T., Meira Filho, L. G., Callander, B. A., Harris, N., Kattenberg,
A., Maskell, K. (eds.) 1996. Climate Change 1995: the Science of Climate
Change. Cambridge: Cambridge University Press.
Houghton, J. T., Ding, Y., Griggs, D. J., Noguer, M., van der Linden,P. J.,
Dai, X., Maskell, K., Johnson, C. A. (eds.) 2001. Climate Change 2001: the
Scienti¬c Basis. Cambridge: Cambridge University Press.
McGuf¬e, K., Henderson-Sellers, A. 1997. A Climate Modelling Primer,
second edition. New York, Wiley.
Trenberth, K. E. (ed.) 1992. Climate System Modelling. Cambridge:
Cambridge University Press.
112 Modelling the climate



2 Richardson, L. F. 1922. Weather Prediction by Numerical Processes.
Cambridge: Cambridge University Press. Reprinted by Dover, 1965.
3 For more details see, for instance, Houghton, The Physics of Atmo-
spheres.
4 For more detail see:
Chapter 13 in Houghton, The Physics of Atmospheres.
Palmer, T. N. 1993. A nonlinear perspective on climate change. Weather,
48, pp. 314“26.
Palmer, T. N. 1999. Journal of Climatology, 12, pp. 575“91.
5 An equation such as y = ax + b is linear; a plot of y against x is a straight
line. Examples of non-linear equations are y = ax2 + b or y + xy = ax + b;
plots of y against x for these equations would not be straight lines. In
the case of the pendulum, the equations describing the motion are only
approximately linear for very small angles from the vertical where the
sine of the angle is approximately equal to the angle; at larger angles
this approximation becomes much less accurate and the equations are
non-linear.
6 More detail in McAvaney, B. J. et al. 2001. In Houghton, J. T., Ding, Y.,
Griggs, D. J., Noguer, M., van der Linden, P. J., Dai, X., Maskell, K., Johnson,
C. A. (eds.) Climate Change 2001: The Scienti¬c Basis. Contribution of
Working Group I to the Third Assessment Report of the Intergovernmen-
tal Panel on Climate Change. Cambridge: Cambridge University Press,
Chapter 8.
7 See, for instance:
Cane, M. A. 1992. In Trenberth, K. E. (ed.) Climate System Modelling.
Cambridge: Cambridge University Press, pp. 583“616.
McAvaney et al., Chapter 8, in Houghton, Climate Change 2001.
Federov, A. V et al. 2003. How predictable is El Ni˜ o? Bulletin of the
. n
American Meteorological Society, 84, pp. 911“919.
8 Folland, C. K., Owen, J., Ward, M. N., Colman, A. 1991. Prediction of
seasonal rainfall in the Sahel region using empirical and dynamical methods.
Journal of Forecasting, 10, pp. 21“56.
9 Information from Folland, C., Hadley Centre, UK.
10 Xue, Y. 1997. Biospheric feedback on regional climate in tropical north
Africa. Quarterly Journal of the Royal Meteorological Society, 123, pp.
1483“1515.
11 Associated with water vapour feedback is also lapse rate feedback which
occurs because, associated with changes of temperature and water vapour
content in the troposphere, are changes in the average lapse rate (the rate of
fall of temperature with height). Such changes lead to this further feedback,
which is generally much smaller in magnitude than water vapour feedback
but of the opposite sign, i.e. negative instead of positive. Frequently, when
overall values for water vapour feedback are quoted the lapse rate feedback
has been included. For more details see Houghton, The Physics of Atmo-
spheres.
Notes 113



12 Stocker, T. F. et al. 2001. Physical climate processes and feedbacks. In
Houghton, J. T., Ding, Y., Griggs, D. J., Noguer, M., van der Linden, P. J., Dai,
X., Maskell, K., Johnson, C. A. (eds.) Climate Change 2001: The Scienti¬c
Basis. Contribution of Working Group I to the Third Assessment Report of
the Intergovernmental Panel on Climate Change. Cambridge: Cambridge
University Press, Section 7.2.1.1.
13 Lindzen, R. S. 1990. Some coolness concerning global warming. Bulletin of
the American Meteorological Society, 71, pp. 288“99. In this paper, Lindzen
queries the magnitude and sign of the feedback due to water vapour, es-
pecially in the upper troposphere, and suggests that it could be much less
positive than predicted by models and could even be slightly negative. Much
has been done through observational and modelling studies to investigate
the likely magnitude of water vapour feedback. More detail can be found
in Stocker, T. F. et al. 2001. Physical climate processes and feedbacks. In
Houghton, Climate Change 2001, Chapter 7. The conclusion of that chapter,
whose authors include Linzden, is that ˜the balance of evidence favours a
positive clear-sky water vapour feedback of a magnitude comparable to that
found in simulations.
14 For more details see McAvaney, B. J. et al. 2001. Chapter 8 in Houghton,
Climate Change 2001.
15 Graf, H.-E. et al. 1993. Pinatubo eruption winter climate effects: model
versus observations. Climate Dynamics, 9, pp. 61“73.
16 See Policymakers summary. In Houghton, J. T., Jenkins, G. J., Ephraums, J.
J. (eds.) 1990. Climate Change: the IPCC Scienti¬c Assessment. Cambridge:
Cambridge University Press.
17 See Summary for policymakers. In Houghton, J. T., Meira Filho, L. G.,
Callander, B. A., Harris, N., Kattenberg, A., Maskell, K. (eds.) 1996. Cli-
mate Change 1995: the Science of Climate Change. Cambridge: Cambridge
University Press.
18 Detection is the process demonstrating that an observed change is signif-
icantly different (in a statistical sense) than can be explained by natural
variability. Attribution is the process of establishing cause and effect with
some de¬ned level of con¬dence, including the assessment of competing
hypotheses.
19 For this and other information about detection and attribution studies see
Mitchell, J. F. B., Karoly, D. J. 2001. Detection of climate change and attri-
bution of causes. In Houghton, Climate Change 2001, Chapter 12.
20 Summary for policymakers. In Houghton, Climate Change 2001.
21 See Levitus, S. et al. 2000. Science, 287, pp. 2225“9; and Levitus, S. et al.
2001. Science, 292, pp. 267“70.
22 Gregory, J. et al. 2002. Journal of Climate, 15, 3117“21.
23 The regional scale is de¬ned as describing the range of 104 to 107 km2 .
The upper end of the range (107 km2 ) is often described as a typical
sub-continental scale. Circulations at larger than the sub-continental scale
are on the planetary scale.
114 Modelling the climate



24 For more information see Giorgi, F., Hewitson, B. 2001. Regional climate
information “ evaluation and projections. In Houghton, Climate Change
2001, Chapter 10.
25 See www.climateprediction.net
26 Hint: recall Stefan™s blackbody radiation law that the energy emitted is pro-
portional to the fourth power of the temperature.
Chapter 6
Climate change in the twenty-¬rst
century and beyond




The last chapter showed that the most effective tool we possess for the
prediction of future climate change due to human activities is the climate
model. This chapter will describe the predictions of models for likely
climate change during the twenty-¬rst century. It will also consider other
factors that might lead to climate change and assess their importance
relative to the effect of greenhouse gases.



Emission scenarios
A principal reason for the development of climate models is to learn
about the detail of the likely climate change this century and beyond.
Because model simulations into the future depend on assumptions re-
garding future anthropogenic emissions of greenhouse gases, which in
turn depend on assumptions about many factors involving human be-
haviour, it has been thought inappropriate and possibly misleading to
call the simulations of future climate so far ahead ˜predictions™. They
are therefore generally called ˜projections™ to emphasise that what is be-
ing done is to explore likely future climates which arise from a range of
assumptions regarding human activities.
A starting point for any projections of likely climate change into the
future is a set of descriptions of likely future global emissions of green-
house gases. These will depend on a variety of assumptions regarding
human behaviour and activities, including population, economic growth,
energy use and the sources of energy generation. As was mentioned in
Chapter 3, such descriptions of future emissions are called scenarios.


115
116 Climate change in the twenty-¬rst century and beyond



26
Scenarios
Scenarios
A1B
A1B
25 A1T
24
A1T
A1FI
A1FI




N2O emissions (Tg N)
CO2 emissions (Gt C)




A2
A2
B1
B1 22
20 B2
B2
IS92a
IS92a
20
15


18
10

16
5
2000 2020 2040 2060 2080 2100
2000 2020 2040 2060 2080 2100

Scenarios
150
Scenarios A1B
A1B A1T
A1FI
A1T
1000 A2
CH4 emissions (Tg CH4)




A1FI
SO2 emissions (Tg S)



B1
A2
B2
B1 IS92a
B2 100
IS92a
800



50
600



2000 2020 2040 2060 2080 2100
2000 2020 2040 2060 2080 2100
Year Year




A wide range of scenarios was developed by the IPCC in a Special Report
Figure 6.1 Anthropogenic
on Emission Scenarios (SRES)1 in preparation for its 2001 Report (see
emissions of carbon dioxide,
methane, nitrous oxide and
box below). It is these scenarios that have been used in developing the
sulphur dioxide for the six
projections of future climate presented in this chapter. In addition, be-
illustrative SRES scenarios,
cause it has been widely used in modelling studies, results are also pre-
A1B, A2, B1 and B2, A1F1 and
sented using a scenario (IS 92a) taken from a set developed by the IPCC
A1T. For comparison the IS
in 1992 and widely described as representative of ˜business-as-usual™.2
92a scenario is also shown.
Details of these scenarios are presented in Figure 6.1.
The SRES scenarios include estimates of greenhouse gas emissions
resulting from all sources including land-use change. Estimates in the
different scenarios begin from the current values for land-use change in-
cluding deforestation (see Table 3.1). Assumptions in different scenarios
vary, from continued deforestation, although reducing as less forest re-
mains available for clearance, to substantial afforestation leading to an
increased carbon sink.
Emission scenarios 117




The emission scenarios of the Special Report on Emission Scenarios (SRES)

B1 Storyline
The SRES scenarios are based on a set of four dif-
ferent story lines within each of which a family of
The B1 storyline and scenario family describes a
scenarios has been developed “ leading to a total of
convergent world, with the same global population
thirty-¬ve scenarios.3
that peaks in mid century and declines thereafter as
in the A1 storyline, but with rapid change in eco-
A1 Storyline nomic structures towards a service and informa-
tion economy, with reductions in material intensity
The A1 storyline and scenario family describes
and the introduction of clean and resource-ef¬cient
a future world of very rapid economic growth, a
technologies. The emphasis is on global solutions to
global population that peaks in mid-century and de-
economic, social and environmental sustainability,
clines thereafter, and the rapid introduction of new
including improved equity, but without additional
and more ef¬cient technologies. Major underlying
climate-related initiatives.
themes are convergence among regions, capacity
building and increased cultural and social interac- B2 Storyline
tions, with a substantial reduction in regional differ-
The B2 storyline and scenario family describes a
ences in per capita income. The A1 scenario family
world in which the emphasis is on local solutions to
develops into three groups that describe alterna-
economic, social and environmental sustainability.
tive directions of technological change in the en-
It is a world with a continuously increasing global
ergy system. The three groups are distinguished by
population, at a rate lower than in A2, intermediate
their technological emphasis: fossil fuel intensive
levels of economic development and less rapid and
(A1FI), non-fossil fuel energy sources (A1T), or a
more diverse technological change than in the B1
balance across all sources (A1B) “ where balance
and A1 storylines. While the storyline is also ori-
is de¬ned as not relying too heavily on one partic-
ented towards environmental protection and social
ular energy source, on the assumption that similar
equity, it focuses on local and regional levels.
improvement rates apply to all energy-supply and
end-use technologies.
From the total set of thirty-¬ve scenarios, an
illustrative scenario was chosen for each of the six
A2 Storyline
scenario groups A1B, A1FI, A1T, A2, B1 and B2.
The A2 storyline and scenario family describes a All should be considered equally sound. It is mostly
very heterogeneous world. The underlying theme for this set of six illustrative scenarios that data are
is self-reliance and preservation of local identi- presented in this chapter.
ties. Fertility patterns across regions converge very The SRES scenarios do not include additional
slowly, which results in a continuously increasing climate initiatives, which means that no scenarios
population. Economic development is primarily re- are included that explicitly assume implementation
gionally oriented and per capita economic growth of the United Nations Framework Convention on
and technological change more fragmented and Climate Change or the emissions targets of the
slower than other story lines. Kyoto Protocol.
118 Climate change in the twenty-¬rst century and beyond



The next stage in the development of projections of climate change is
to turn the emission pro¬les of greenhouse gases into greenhouse gas
concentrations (Figure 6.2) and then into radiative forcing (Table 6.1 and
Figure 6.4(a)). The methods by which these are done are described in
Chapter 3, where the main sources of uncertainty are also mentioned.
For the carbon dioxide concentration scenarios these uncertainties, espe-
cially those concerning the magnitude of the climate feedback from the
terrestrial biosphere (see box on page 40), amount to a range of about
’10% to +30% in 2100 for each pro¬le.4
For most scenarios, emissions and concentrations of the main green-
house gases increase during the twenty-¬rst century. However, despite the
increases projected in fossil fuel burning “ very large increases in some
cases “ emissions of sulphur dioxide (Figure 6.1) and hence the concen-
trations of sulphate particles are expected to fall substantially because
of the spread of policies to abate the damaging consequences of air pol-
lution and ˜acid-rain™ deposition to both humans and ecosystems.5 The
in¬‚uence of sulphate particles in tending to reduce the warming due to in-
creased greenhouse gases is therefore now projected to be much less than
for projections made in the mid 1990s (see the IS 92a scenario for sul-
phur dioxide in Figure 6.1). The other anthropogenic sources of particles
in the atmosphere listed in Figure 3.8 will also contribute small amounts
of positive or negative radiative forcing during the twenty-¬rst century.6


Model projections
Results which come from the most sophisticated coupled atmosphere“
ocean models of the kind described in the last chapter provide fundamen-
tal information on which to base climate projections. However, because
they are so demanding on computer time only a limited number of results
from such models are available. Many studies have also therefore been
carried out with simpler models. Some of these, while possessing a full
description of atmospheric processes, only include a simpli¬ed descrip-
tion of the ocean; these can be useful in exploring regional change.
Others, sometimes called energy balance models (see box on page 121),
drastically simplify the dynamics and physics of both atmosphere and
ocean and are useful in exploring changes in the global average response
with widely different emission scenarios. Results from simpli¬ed mod-
els need to be carefully compared with those from the best coupled
atmosphere“ocean models and the simpli¬ed models ˜tuned™ so that, for
the particular parameters for which they are being employed, agreement
with the more complete models is as close as possible. The projections
presented in the next sections depend on results from all these kinds of
models.
Model projections 119



Figure 6.2 Atmospheric
1300
concentrations of carbon
Scenarios
1200
dioxide, methane and
A1B
CO2 concentration (ppm)




1100 A1T nitrous oxide resulting
A1FI
1000 from the six illustrative
A2
SRES scenarios and from
900 B1
the IS 92a scenario.
B2
800
IS92a Uncertainties for each
700
pro¬le, especially those
600 due to possible carbon
feedbacks, have been
500
estimated as from about
400
’10% to +30% in 2100
300
1980 2000 2020 2040 2060 2080 2100

4000
Scenarios
A1B
3500 A1T
CH4 concentration (ppb)




A1FI
A2
3000 B1
B2
IS92a
2500


2000


1500
1980 2000 2020 2040 2060 2080 2100

500
Scenarios
A1B
A1T
N2O concentration (ppb)




450
A1FI
A2
B1
B2
400
IS92a


350



300
1980 2000 2020 2040 2060 2080 2100



In order to assist comparison between models, experiments with
many models have been run with the atmospheric concentration of carbon
dioxide doubled from its pre-industrial level of 280 ppm. The global aver-
age temperature rise under steady conditions of doubled carbon dioxide
120 Climate change in the twenty-¬rst century and beyond



Table 6.1 Radiative forcing (W m’2 ), globally averaged, for
greenhouse gases from the year 1750 to 2000 and from SRES scenarios
to 2050 and 2100

Greenhouse gas Year A1B A1T A1FI A2 B1 B2 IS 92a

CO2 2000 1.46
2050 3.36 3.08 3.70 3.36 2.92 2.83 3.12
2100 4.94 3.85 6.61 5.88 3.52 4.19 4.94
CH4 2000 0.48
2050 0.70 0.73 0.78 0.75 0.52 0.68 0.73
2100 0.56 0.62 0.99 1.07 0.41 0.87 0.91
N2 O 2000 0.15
2050 0.25 0.23 0.33 0.32 0.27 0.23 0.29
2100 0.31 0.26 0.55 0.51 0.32 0.29 0.40
O3 (trop) 2000 0.35
2050 0.59 0.72 1.01 0.78 0.39 0.63 0.67
2100 0.50 0.46 1.24 1.22 0.19 0.78 0.90
Halocarbons 2000 0.34
2050 0.49
2100 0.57

Data from Ramaswamy, V. et al. 2001. Radiative forcing of climate change. In
Houghton, Climate Change 2001, Chapter 6. Data selected from Tables 6.1 and
6.4. For 2050 and 2100 for the halocarbons, all scenarios make the same as-
sumptions.


concentration has become known as the climate sensitivity.7 The Inter-
governmental Panel on Climate Change (IPCC) in its 1990 Report gave
a ˜best estimate™ of 2.5 —¦ C for the climate sensitivity; it also considered
that it was unlikely to lie outside the range of 1.5 —¦ C to 4.5 —¦ C, a range
that encompasses the results of the best coupled atmosphere“ocean gen-
eral circulation models (AOGCMs). The IPCC 1995 and 2001 Reports
have con¬rmed these values. Reasons why there remains this range of
uncertainty in the estimates from climate models were explained in Chap-
ter 5. The projections presented in this chapter follow the IPCC 2001
Assessment.8


Projections of global average temperature
When information of the kind illustrated in Figures 6.1, 6.2 and 6.4(a) is
incorporated into simple or more complex models, projections of climate
change can be made. As we have seen in earlier chapters, a useful proxy
Simple climate models

In Chapter 5 a detailed description was given of The most radical simpli¬cation in the simpler mod-
general circulation models (GCMs) of the atmo- els is to remove one or more of the dimensions so
sphere and the ocean and of the way in which they that the quantities of interest are averaged over lat-
are coupled together (in AOGCMs) to provide sim- itude circles (in two-dimensional models) or over
ulations of the current climate and of climate per- the whole globe (in one-dimensional models). Such
turbed by anthropogenic emissions of greenhouse models can, of course, only simulate latitudinal or
gases. These models provide the basis of our pro- global averages “ they can provide no regional in-
jections of the detail of future climate. However, formation.
because they are so elaborate, they take a great deal Figure 6.3 illustrates the components of such a
of computer time so that only a few simulations can model in which the atmosphere is contained within
be run with these large coupled models. a ˜box™ with appropriate radiative inputs and out-
To carry out more simulations under differ- puts. Exchange of heat occurs at the land surface
ent future emission pro¬les of greenhouse gases (another ˜box™) and the ocean surface. Within the
or of aerosols or to explore the sensitivity of fu- ocean allowance is made for vertical diffusion and
ture change to different parameters (for instance, vertical circulation. Such a model is appropriate
parameters describing the feedbacks in the atmo- for simulating changes in global average surface
sphere which largely de¬ne the climate sensitiv- temperature with increasing greenhouse gases or
ity), extensive use has been made of simple climate aerosols. When exchanges of carbon dioxide across
models.9 These simpler models are ˜tuned™ so as to the interfaces between the atmosphere, the land and
the ocean are also included, the model can be em-
agree closely with the results of the more complex
ployed to simulate the carbon cycle.
AOGCMs in cases where they can be compared.




Figure 6.3 The components of a simple ˜upwelling“diffusion™ climate model.
122 Climate change in the twenty-¬rst century and beyond



10
(a)
A1FI
9 A1B
A1T
8 A2
B1
7 B2
Forcing (W m’2)



IS92a
Model ensemble
6
all SRES
envelope
5

4

3

2

1

0
1800 1900 2000 2100
Year


(b) 7
A1FI
Several models
A1B
6 all SRES
A1T envelope
Temperature Change ( C)




A2
B1
5 Model ensemble
B2
all SRES
IS92a envelope
4


3


2

Bars show the
1 range in 2100
produced by
several models
0
1800 1900 2000 2100
Year

Figure 6.4 Simple model results. (a) Anthropogenic globally averaged radiative
forcing, based on historical information about greenhouse gas and aerosol
concentrations to the year 2000 (see Figure 3.8) and the SRES scenarios to the
year 2100. The shading shows the envelope of forcing that encompasses the full
range of thirty-¬ve SRES scenarios. (b) Historic anthropogenic global mean
temperature change and future changes for the SRES scenarios and the IS 92a
scenario calculated using a simple climate model tuned to seven AOGCMs (with
climate sensitivity in the range 1.7 to 4.2 —¦ C). The darker shading represents the
envelope of the full set of thirty-¬ve SRES scenarios using the average of the
model results (mean climate sensitivity is 2.8 —¦ C). The lighter shading is the
envelope including all seven model projections (the range of model results for
each scenario is also shown by the bars on the right hand side).
Projections of global average temperature 123



for climate change that has been widely used is the change in global
average temperature.
The projected rise in global average temperature due to the increase
in greenhouse gases and aerosols from pre-industrial times is illustrated
in Figure 6.4(b). It shows an increase of about 0.6 —¦ C up to the year 2000
and an increase ranging from about 2 —¦ C to about 6 —¦ C by 2100 “ that
wide range resulting from the very large uncertainty regarding future
emissions and also from the uncertainty that remains regarding the feed-
backs associated with the climate response to the changing atmospheric
composition (as described in Chapter 5).10
Compared with the temperature changes normally experienced from
day to day and throughout the year, changes of between 2 —¦ C and 6 —¦ C
may not seem very large. But, as was pointed out in Chapter 1, it is in
fact a large amount when considering globally averaged temperature.
Compare it with the 5 —¦ C or 6 —¦ C change in global average temperature
that occurs between the middle of an ice age and the warm period in
between ice ages. The changes projected for the twenty-¬rst century
are from one-third to a whole ice age in terms of the degree of climate
change!
The rate of change of global average temperature projected for the
twenty-¬rst century is in the range of 0.15 —¦ C to 0.6 —¦ C per decade.
These might seem small rates of change; most people would ¬nd it
hard to detect a change in temperature of a fraction of a degree. But
remembering again that these are global averages, such rates of change
become very large. Indeed, they are much larger than any rates of change
the climate has experienced for at least the past 10 000 years as inferred
from paleoclimate data. As we shall see in the next chapter, the ability
of both humans and ecosystems to adapt to climate change depends
critically on the rate of change.
The changes in global average temperature shown in Figure 6.4(b)
from the IPCC 2001 Report are substantially greater than those shown
in the IPCC 1995 Report. The main reason for the difference is the much
smaller aerosol emissions in the SRES scenarios compared with the
IS 92 scenarios. This is illustrated by the temperature in 2100 shown
in Figure 6.4(b) for the IS 92a scenario which is similar to that for the
SRES B2 scenario even though the carbon dioxide emissions at that date
for IS 92a are ¬fty per cent greater than those for B2.
In many of the modelling studies of climate change, the situation
of doubled pre-industrial atmospheric carbon dioxide has often been in-
troduced as a benchmark especially to assist in comparisons between
different model projections and their possible impacts. Since the pre-
industrial concentration was about 280 ppm, doubled carbon dioxide is
about 560 ppm. From the curves in Figure 6.2 this is likely to occur
124 Climate change in the twenty-¬rst century and beyond



sometime in the second half of the twenty-¬rst century, depending on
the scenario. But, other greenhouse gases are also increasing and con-
tributing to the radiative forcing. So as to achieve an overall picture more
easily, it is often convenient to convert other greenhouse gases to equiv-
alent amounts of carbon dioxide, in other words to amounts of carbon
dioxide that would give the same radiative forcing.11 The information
in Table 6.1 enables the conversion to be carried out. For instance, the
increases in the greenhouse gases other than carbon dioxide (including
ozone) to date produce about eighty per cent of the radiative forcing due
to the increase in carbon dioxide to date (see Figure 3.8). This propor-
tion will drop substantially during the next few decades as the growth in
carbon dioxide becomes more dominant in nearly all scenarios. Refer-
ring to Figure 6.4(a), and noting that doubled carbon dioxide produces a
radiative forcing of about 3.7 W m’2 , it can be seen that doubling of the
equivalent carbon dioxide amount from pre-industrial times will occur
between 2040 and 2070 depending on the scenario.
Now, in Figure 6.4(a), select one of the scenarios, say A1B, and
note that the radiative forcing for this scenario reaches that equivalent
to doubled carbon dioxide (i.e. 3.7 W m’2 ) in about 2040. Then move to
Figure 6.4(b) and note that in 2040 the temperature rise is about 1.7 —¦ C.
This is only just over half the 2.8 —¦ C (the value for climate sensitivity
used for the results presented in Figure 6.4(b) see ¬gure caption) that
would be expected for doubled carbon dioxide under steady conditions.
As was shown in Chapter 5, this difference occurs because of the slowing
effect of the oceans on the temperature rise. But this means that, as the
carbon dioxide concentration continues to increase, at any given time
there exists a commitment to further signi¬cant temperature rise which
has not been realised at that time.


Regional patterns of climate change
So far we have been presenting global climate change in terms of likely
increases in global average surface temperature that provide a useful
overall indicator of the magnitude of climate change. In terms of regional
implications, however, a global average conveys rather little information.
What is required is spatial detail. It is in the regional or local changes
that the effects and impacts of global climate change will be felt.
With respect to regional change, it is important to realise that, be-
cause of the way the atmospheric circulation operates and the interactions
that govern the behaviour of the whole climate system, climate change
over the globe will not be at all uniform. We can, for instance, expect
substantial differences between the changes over large land masses and
over the ocean; land possesses a much smaller thermal capacity and
Regional patterns of climate change 125



so can respond more quickly. Listed below are some of the broad fea-
tures on the continental scale that characterise the projected temperature
changes; more detailed patterns are illustrated in Figure 6.5(a). Refer-
ence to Chapter 4 indicates that many of these characteristics are already
being found in the observed record of the last few decades.
r Generally greater surface warming of land areas than of the oceans
typically by about forty per cent compared with the global average,
greater than this in northern high latitudes in winter (associated with
reduced sea ice and snow cover) and less than forty per cent in south
and southeast Asia in summer and in southern South America in winter.
r Minimum warming around Antarctica and in the northern North
Atlantic which is associated with deep oceanic mixing in those areas.
r Little warming over the Arctic in summer.
r Little seasonal variation of the warming in low latitudes or over the
southern circumpolar ocean.
r A reduction in diurnal temperature range over land in most seasons
and most regions; night time lows increase more than daytime highs.
So far we have been presenting results solely for atmospheric tem-
perature change. An even more important indicator of climate change
is precipitation. With warming at the Earth™s surface there is increased
evaporation from the oceans and also from many land areas leading on
average to increased atmospheric water vapour content and therefore also
on average to increased precipitation. Since the water-holding capacity
of the atmosphere increases by about 6.5% per degree celsius,12 the in-
creases in precipitation as surface temperature rises can be expected to be
substantial. In fact, model projections indicate increases in precipitation
broadly related to surface temperature increases of about three per cent
per degree celsius.13 Further, since the largest component of the energy
input to the atmospheric circulation comes from the release of latent
heat as water vapour condenses, the energy available to the atmosphere™s
circulation will increase in proportion to the atmospheric water content.
A characteristic therefore of anthropogenic climate change due to the
increase of greenhouse gases will be a more intense hydrological cycle.
The likely effect of this on precipitation extremes will be discussed in
the next section.
In Figure 6.5(b) are shown the projected changes in the distribution of
precipitation. Although on average precipitation increases there are large
regional variations and large areas where there are likely to be decreases
in average precipitation and also changes in its seasonal distribution. For
instance, at high northern latitudes there are large increases in winter and
over south Asia in summer. Southern Europe, Central America, southern
Africa and Australia are likely to have drier summers.
126 Climate change in the twenty-¬rst century and beyond




Change in temperature for scenario A2
(a)


60N




30N




EQ




30S




60S




180 120W 60W 0 60E 120E 180

Change in temperature relative to model™s global mean Change in global mean temperature (°C)
Much greater than average warming
Greater than average warming 0 1 2 3 4 5 6 8 10
Less than average warming
Inconsistent magnitude of warming
Dec-Jan-Feb
Jun-Jul-Aug


Figure 6.5 Projections for the SRES scenario A2 for the period 2071“2100
relative to 1961“1990 from an ensemble of nine different ocean“atmosphere
general circulation models. (a) The annual mean change of temperature in —¦ C
shown by the shading. The boxes show an analysis of inter-model consistency in
regional relative warming (i.e. warming relative to each model™s global average
warming) for winter and summer seasons. Regions are classi¬ed as showing
either agreement on warming in excess of forty per cent above the global mean
annual average (much greater than average warming), agreement on warming
greater or less than the global mean annual average (greater or less than average
warming) or disagreement amongst models on the magnitude of regional
relative warming (inconsistent magnitude of warming). A consistent result from at
least seven of the nine models is de¬ned as being necessary for agreement. The
global mean average warming of the models used are in the range 1.2 to 4.5 —¦ C
for this scenario. (b) The boxes show an analysis of inter-model consistency in
regional precipitation change for winter and summer seasons. Regions are
classi¬ed as showing agreement on change of greater than +20% (large
increase), between +5% and +20% (small increase) and more than ’20% (large
decrease), or disagreement (inconsistent sign). A consistent result from at least
seven of the nine models is de¬ned as being necessary for agreement.
Regional patterns of climate change 127




Change in precipitation for scenario A2
Change in precipitation for scenario A2
(b)
(b)


60N




30N




EQ




30S




60S




180 120W 60W 0 60E 120E 180


Change in precipitation
Large increase
Small increase
No change
Small decrease
Large decrease Dec-Jan-Feb
Inconsistent sign Jun-Jul-Aug


Figure 6.5 (Cont.).



Much natural climate variability occurs because of changes in, or
oscillations between, persistent climatic patterns or regimes. The Paci¬c,
North Atlantic Anomaly (PNA “ that is dominated by high pressure over
the eastern Paci¬c and western North America and which tends to lead to
very cold winters in the eastern United States), the North Atlantic Oscil-

<<

. 5
( 13)



>>