Download `Future` climate and impacts - Climatic Research Unit

INFORMATION SHEET ON FUTURE CLIMATE AND IMPACTS IN THE URBAN CASE STUDIES: ATHENS,
GREECE
Summary
►
►
►
►
►
►
►
Daily minimum and maximum air temperatures in Athens increase during the period 1950-2050.
The number of hot days and nights also show a sharp increase during the 21st century.
Precipitation is increasing slightly in the present climate, while it is decreasing significantly in the
‘mid-century’ climate. Precipitation extremes do not show a clear trend.
For Attica and the surrounding forest areas, a significant increase in peri-urban forest fires risk is expected.
Mortality associated with high temperatures could increase significantly, especially for the period 2071-2100.
A gain in energy demand is expected for the colder half of the year (November to April), especially during the latter
part of the 21st century. For the warmer half of the year (May to October), an increasing trend is evident which is
double in the hot summer months of July and August when the demand for air conditioning peaks.
An increase in ozone episodes is evident, while at the end of the century ozone exceedance days could increase by
almost a month.
1. Introduction
This information sheet is the third in a series on
the present and future impacts of climate change
for the Athens urban case study. The first
information sheet focused on observed climate
(Giannakopoulos et al., 2009). It showed that
mean air temperature exhibits a tendency
towards warmer years, with significantly warmer
summer maximum temperatures of about 1.8 °C
since the late 19th century. The number of hot
days/nights shows a ‘virtually certain’ increasing
trend, especially in the last decade. A slight
increasing trend in total rainfall has been
observed.
The
second
information
sheet
(Giannakopoulos et al., 2010) reported that
higher forest fire risks are direct consequences of
increases in maximum temperature and
decreases in rainfall and relative humidity during
the summer. Heat waves have readily discernible
health impacts as substantial heat-related deaths,
occurred at very high temperatures for Athens.
Energy mainly peaks in winter; however, a
second significant peak is also apparent in
summer. In the city of Athens, ozone enhances
with temperature although there are still high
ozone concentrations at lower temperatures.
This information sheet focuses on climate
scenarios for the present and future periods in
Athens (1950-2050), based on a set of the most
recent model simulations including for the first
time a coupling of the Mediterranean Sea with
the atmosphere at a high spatial resolution. The
set of atmospheric variables simulated by the
models are presented and compared to observed
ones, if available. Biogeophysical impacts are
investigated using future projections of the
observed relationship between climate and
biogeophysical variables.
For the Athens urban case study,
biogeophysical and vulnerability indicators
based on system thresholds were calculated
using regional climate model output both from
the CIRCE project and from the earlier EU
ENSEMBLES project (van der Linden and
Mitchell, 2009). The response of biogeophysical
and social systems to past extremes in climate
can be used as an analogue for the potential
consequences of climate change in a region.
Future climate indicators are presented for:
►
►
►
►
►
►
►
►
Maximum temperature
Number of hot days and nights
Annual rainfall
Rainfall extremes
Peri-urban forest fires
All-cause daily mortality
Energy consumption
Ozone exceedance days
1
Climate models
The CIRCE project climate model runs (Information sheet on future climate: climate projections for
the CIRCE case studies, http://www.cru.uea.ac.uk/projects/circe/projections_models.html) are used to
investigate present and future climate changes in the Greater Athens Area for the period 1950-2050.
After the year 2000, the A1B scenario for greenhouse gas emissions is used (Nakićenović and Swart,
2000). The CIRCE climate models are of differing spatial resolution and their gridpoints around
Athens are shown in Figure 1.
INGV (Global model) ~80 km
ENEA (Regional model), 30 km
IPSL (Global and regional model) 30 km.
The IPSL model used is IPSL2
MPI (Regional model), 25 km
CNRM (Stretched grid global model), 50 km
Figure 1: The Athens area covered by the atmospheric components of the five CIRCE climate models and their
horizontal resolution. Red star indicates NOA meteorological station
2
Mean annual maximum and minimum temperature
What is it?
Mean annual maximum and minimum temperature (Tx and Tn) were used as core climate
indicators for the case studies. Observed and projected mean annual Tx and Tn temperature anomalies
(Figure 2) for Athens are calculated using output from the ENEA, MPI and IPSL models, for the
period 1950-2050. Furthermore, the projected changes (for 2021-2050 with respect to 1961-1990) in
these indicators are presented in Section 5: Uncertainties.
Athens_Mean Annual Tmax anomalies
Athens_Mean Annual Tmin anomalies
4.0
obs
ipsl
model mean
3.0
2.0
4.0
enea
mpi
obs
ipsl
model mean
3.0
2.0
1.0
enea
mpi
1.0
0.0
0.0
1
2050
2030
2010
1990
1970
1950
1890
2050
2030
2010
1990
1970
1950
1930
-4.0
1910
-3.0
-4.0
1890
-2.0
-3.0
1930
-1.0
-2.0
1910
-1.0
Figure 2: Observed and projected mean annual maximum (left) and minimum (right) temperature anomalies
for Athens calculated using output from the ENEA, MPI and IPSL models, 1950-2050.
What does this show?
For temperature, the observed trends and climate
change signal are stronger than for precipitation.
Figure 2 shows observed and simulated (ENEA,
MPI and IPSL models) mean annual maximum
and minimum temperature for Athens, as
anomalies from the 1961-1990 baseline. In
general, the projected changes appear as an
extension of the observed positive trends, while
the long-term 1950-2050 trend for both Tx and
Tn is 0.26°C/decade. For Athens maximum
temperature, the projected changes do not extend
much beyond the upper range of the
observations. The influence of the hot summer of
2007 is evident in the 2007 observed Tx value
for Athens which appears as an outlier with
respect to the observed series but falls centrally
within the model range by 2050.
Why is it relevant?
This indicator is of prime importance, especially
for a Mediterranean city such as Athens, since it
determines the thermal comfort and cooling
demands. There is a strong relationship between
the stress experienced by organisms exposed to
high temperatures and daily temperature
extremes. Moreover, the rate of energy demand
depends on the ambient temperature and hence,
summer maximum temperature is a valuable
estimator of changes in energy consumption for
air conditioning of buildings. Changes in
temperature may have critical implications in an
urban area for surface water resources, periurban forestry, infrastructure, industry, and most
notably population heat stress and health
(Giannakopoulos et al., 2009).
3
Hot days and nights
What is it?
Hot nights act synergistically with hot days to contribute to human discomfort during a heat wave.
Here, the threshold of the 95th percentile of Tx and Tn for the 1961-1990 period for each of the
National Observatory of Athens (NOA) observational records and model (ENEA, MPI. IPSL and
model mean) projections were used to define a very hot day and a very hot night, respectively.
Athens_Number of Days with Tx>95 thp
100
obs
ipsl
model mean
90
80
70
Athens_Number of Days with Tn>95 thp
100
obs
ipsl
model mean
90
enea
mpi
80
70
60
50
60
40
40
30
30
20
10
20
0
0
enea
mpi
50
2050
2040
2030
2020
2010
2000
1990
1980
1970
1960
1950
1940
1930
1920
1910
1900
1890
2050
2040
2030
2020
2010
2000
1990
1980
1970
1960
1950
1940
1930
1920
1910
1900
1890
10
Figure 3: Observed and projected annual number of very hot summer days (Tx95) - left; and very hot summer
nights (Tn95) – right, for Athens, calculated using output from the ENEA, MPI and IPSL models, 1950-2050.
2050
2040
2030
2020
2010
2000
1890
2050
2040
2030
2020
2010
2000
1990
1980
1970
1960
1950
1940
1930
1920
-20
1910
-20
1900
-10
1890
0
-10
1990
10
0
1980
20
10
1970
30
20
1960
40
30
1950
50
40
enea
mpi
1940
60
1930
50
obs
ipsl
model mean
70
enea
mpi
1920
obs
ipsl
model mean
60
1910
70
Athens_Number of Days with Tn>95 thp anomalies
80
1900
Athens_Number of Days with Tx>95 thp anomalies
80
Figure 4: Observed and projected anomalies of left: annual number of very hot summer days (Tx95), and
right: very hot summer nights (Tn95) for Athens, using output from ENEA, MPI and IPSL models, 1950-2050.
What does this show?
Similar features to mean temperatures, are seen
for temperature extremes – very hot days and
very hot nights (Figures 3 and 4), although the
inter-annual variability is greater and the upper
range of the observations is not exceeded very
frequently in the future except for the IPSL
model.
Why is it relevant?
An increase in the frequency of heat waves has
been observed, particularly in southern Europe
(IPCC, 2007). These events intensify in urban
areas such as Athens, and are a familiar feature
of Greek summers. Heat-wave days have
negative effects on human comfort and
contribute significantly to heat stress especially
if associated with high levels of humidity. In
addition, stagnant air masses encourage the
build-up of air pollutants, which act
synergistically with higher air temperature to
increase mortality. Warm nights following a hot
day and accompanied by high humidity can be
particularly uncomfortable to urban residents.
4
Annual total rainfall
What is it?
Annual total precipitation was also selected as a climate indicator (Figure 5). Observed and projected
annual total rainfall anomalies for Athens are calculated using output from the ENEA, MPI and IPSL
models, for the period 1950-2050.
Athens_Total Annual Rainfall
1200
obs
ipsl
model mean
1000
Athens_Total Annual Rainfall anomalies
700
enea
mpi
500
300
800
100
600
-100
400
-300
obs
ipsl
model mean
-500
enea
mpi
2050
2030
2010
1990
1970
1950
1930
-700
1890
2050
2040
2030
2020
2010
2000
1990
1980
1970
1960
1950
1940
1930
1920
1910
1900
1890
0
1910
200
Figure 5: Observed and projected total annual rainfall (left) and total annual rainfall anomalies (right) for
the Athens case study calculated using output from the ENEA, MPI and IPSL models, 1950-2050.
What does this show?
The CIRCE ensemble mean changes for
rainfall for the period 1950-2050 in Athens are
negative, about 77 mm in absolute terms, while
the change in percentage terms is -16 %. The
inter-model spread is quite large. The MPI
model indicates a particularly large decrease in
annual precipitation of -43% for Athens.
However, the magnitude of change does not
vary systematically from model to model as is
the case of temperature. The large inter-annual
variability of precipitation makes it difficult to
identify statistically significant trends in
observed or simulated precipitation-based
climate indicators.
Why is it relevant?
Rainfall is a limiting factor in biological
systems, so the examination of total rainfall on
an annual and seasonal basis is important,
particularly for Mediterranean regions.
Changes in annual total rainfall may have
important implications for the availability of
water resources, which in turn can affect
agriculture, forestry and water supply.
5
Precipitation extremes
What is it? Figure 6 shows projected changes for three indices in Athens calculated using daily output
from the ENEA, MPI and IPSL models for the period 1950-2050: the maximum dry spell length
(consecutive dry days: cdd), heavy precipitation (the 90th percentile of daily precipitation: pq90) and
maximum three-day precipitation (px3d), where a threshold of > 0.5 mm defined a rain day.
Consecutive Dry Days anomalies
200.0
90th percentile of daily precipitation anomalies
30.0
150.0
20.0
100.0
10.0
50.0
0.0
0.0
-50.0
-10.0
-100.0
2050
2040
2030
2020
2010
1990
2050
2000
1980
ipsl
mean model
1970
1960
1950
2050
2040
2030
2020
2010
-30.0
2000
1990
1980
enea
mpi
2040
-20.0
ipsl
mean model
1970
1950
-200.0
1960
enea
mpi
-150.0
maximum 3-days precipitation anomalies
200.0
150.0
100.0
50.0
0.0
-50.0
-100.0
2030
2020
2010
2000
1990
1980
ipsl
mean model
1970
1950
-200.0
1960
enea
mpi
-150.0
Figure 6: Projected precipitation extremes indices for Athens calculated using daily output from the ENEA,
MPI and IPSL models: the maximum dry spell length (consecutive dry days: cdd), heavy precipitation (the
90th percentile of daily precipitation: pq90) and maximum three-day precipitation (px3d).
What does this show?
It is difficult to draw clear messages even
concerning the direction of change for
precipitation extremes, particularly given the
strong inter-annual variability of precipitation.
ENEA and IPSL models suggest a slight
increase in the maximum three-day precipitation,
while MPI exhibits a decrease. For consecutive
dry days, all models indicate an increase. In
general, the intermodel spread is quite large and
the signal of change cannot be considered robust
in contrast to the fairly robust projections for
longer dry periods across the Mediterranean,
particularly in summer, emerging from other
studies (Beniston et al. 2007, Giorgi and
Lionello 2008, May 2008, Giannakopoulos et al.
2009). These studies tend to focus on the end of
the century and it may be that any signal of
change has not yet emerged from the noise due
to natural variability by 2021-2050. See Section
5 for further discussion of uncertainties.
Why is it relevant?
Changes in precipitation extremes are
particularly relevant for urban areas. An increase
in the incidence and intensity of rainfall events
may provoke natural disasters such as localised
flash flooding in urban areas. Intense rainfall
events lead to greater erosion rates and a higher
risk of flash flooding in urban and peri-urban
areas with occasional serious loss of life and
property.
6
Peri-urban forest fires
What is it?
Future fire risk for the Greater Athens Area and
the surrounding forest area in the Attica
peninsula was assessed using the Canadian Fire
Weather Index (FWI, Van Wanger 1987). This a
numerical rating of fire intensity calculated using
daily maximum temperature, relative humidity,
wind and precipitation. Although the FWI was
originally developed for Canadian forests,
several studies have shown its suitability for the
Mediterranean basin (e.g., Viegas et al. 1999).
The FWI was estimated for future periods using
daily output from six ENSEMBLES RCMs
(KNMI-RAMCO2, CNRM, ETHZ, MPI, METO
and METNO) simulations. Ensemble-mean
changes (with respect to the 1961-1990 baseline)
for two 30-year future periods (2021-2050 and
2071-2100) were calculated. Model output
covers Europe at a resolution of about 25 km, for
the A1B emissions scenario. Unlike the CIRCE
model runs these are not coupled atmosphereocean regional models. Future FWI projections
have also been investigated with two CIRCE
models simulations (ENEA and MPI) for the
mid-century future period 2021-2050.
Figure 7: Projected changes (2021-2050 minus 1961-1990) in the number of days with fire risk FWI > 15
(left) and extreme fire risk FWI > 30 (right), based on the mean of six ENSEMBLES RCM (KNMI-RAMCO2,
CNRM, ETHZ, MPI, METO and METNO) simulations.
7
Figure 8: Projected changes (2021-2050 minus 1961-1990) in the number of days with fire risk FWI > 15
(left) and extreme fire risk FWI > 30 (right), based on the average of two CIRCE models (ENEA and MPI)).
What does this show?
In order to identify a relationship between FWI
values and fire occurrence for the Attica region,
meteorological data for the period 1983-1990
were obtained from the Greek National
Meteorological Service and daily fire frequency
and area burnt in the regions of the
meteorological stations were obtained from the
Forest Research Institute of Athens. Fire
occurrence was found to be low for FWI less
than about 15 and tends to increase with
increasing FWI. A threshold of FWI = 30 was
used to define extreme fire risk (Giannakopoulos
et al. 2010). For Attica and the surrounding
forest areas, fire risk is projected to increase in
both future periods. For the near-future period
2021-2050, the number of days with fire risk
(FWI>15) increases by up to 17-20 days or more
in the north western parts of the study area,
while extreme fire risk (FWI>30) is more
spatially variable, with an increase of about 10
days in the vicinity of Athens (Figure 7). By the
end of the century (2071-2100), the increase is
50 days (for FWI>15) and 40 days for extreme
fire risk (not shown). The same analysis based
on the two CIRCE simulations shows a different
spatial distribution for the FWI. In peri-urban
regions, the future fire risk for FWI>15 is larger
in the Great Athens Area than for the
ENSEMBLES simulations. In contrast, extreme
fire risk shows smaller increases for the CIRCE
simulations compared to the ENSEMBLES
simulations (Figure 8).
Why is this important?
Peri-urban forest fires are highly sensitive to
climate change since fuel moisture is affected by
precipitation, relative humidity, air temperature
and wind speed. The contribution of
meteorological factors to fire risk is simulated by
various non-dimensional indices of fire risk. The
destruction of forests is of great concern, since
this has many side effects, e.g. floods, soil
erosion and consequent loss of fertility.
Furthermore, peri-urban forests fires play a
fundamental role in regulating the air
temperature and wind circulation in the
surrounding city. Thus peri-urban forest fires
aside from the general side effects may also
contribute to an increase of temperature in the
city during the summer months and an
intensification of the urban heat island.
8
All-cause daily mortality
What is it?
A linear model was used to project heat-related mortality to the future climate (2021-2050 and 20712100) using temperature output from the RACMO2 RCM ENSEMBLES simulation for the A1B
emissions scenario. An ‘adaptation’ or acclimatisation factor of 1°C per 30 years (Dessai, 2002) was
included to allow for physiological and behavioural adjustment to higher temperatures. The same
model was also applied for the CIRCE simulations (ENEA, MPI and IPSL) for the 2021-2050 period.
Figure 9: Excess deaths (right axis; red and dark blue bars for present and future climate with adaptation,
respectively) and daily temperature frequencies (left axis; light blue bars) in Athens. Top: for the future
periods 2021-2050 (left) and 2071-2100 (right), using the RACMO2 RCM ENSEMBLES simulation; bottom:
for the future period 2021-2050 derived using the ENEA, MPI and IPSL CIRCE models.
What does this show?
Even after adjustment for adaptation, there are
significant increases in excess heat-related
mortality for the periods 2021-2050 and
especially for the period 2071-2100 (Figure 9).
This increase is shown by the shift to the right
and the longer tails of the simulated dark-blue
bars compared to the red observed bars. There
are significant differences in projected mortality
between the three CIRCE models. The IPSL
model fails to capture the high summer
temperatures recorded in Athens, and thus shows
much less excess mortality in the future. ENEA
and MPI have a much better fit to observations.
Why is it relevant?
Heat waves have readily discernible health
impacts because they result in a large number of
deaths and affect relatively large, heterogeneous
areas simultaneously. Not all heat waves have a
similar impact on mortality; the intensity,
duration and timing of the event are particularly
important. Illnesses recognisable as the direct
results of exposure to prolonged periods of high
temperature are heatstroke, heat exhaustion, and
heat cramps. Heat-related mortality in large
urban areas disproportionately affects the
elderly, young people with pre-existing illness,
and low-income groups.
9
Energy consumption
What is it?
Future energy demand in Athens was estimated using temperature output from the RACMO2 RCM
ENSEMBLES simulation for the A1B emissions scenario and a simple extrapolation of the non-linear
trend between temperature and energy load from the present day (Giannakopoulos et al., 2010). It is
assumed that technology use (UNDP, 2002) is the same in the control (1961-1990) and future periods
(2021-2050 and 2071-2100). The same model was also applied for two CIRCE simulations (ENEA,
MPI and IPSL).
Figure 10: Athens energy demand (bars, right axis) and daily maximum temperature (triangles, left axis) for
the colder (blue bars) and warmer (red bars) halves of the year for 1961-1990, 2021-2050 periods for, top:
the RACMO2 RCM ENSEMBLES simulation; bottom: CIRCE simulations (ENEA, MPI, IPSL).
What does this show?
Figure 10 (top) shows that for the colder half of
the year (November to April), energy demand
decreases, especially during the latter part of the
21st century. The saving in energy demand is
about 2% for 2021-2050 and about 5% for 20712100. For the warmer half of the year (May to
October), an increasing trend is evident: a 5%
increase in demand for 2021-2050 and 15% for
2071-2100. This increases doubles in the hot
summer months of July and August when the
demand for air conditioning peaks.
The energy demand projections based on CIRCE
simulations for 2021-2050 (Figure 10, bottom)
10
also indicate a decrease in energy demand during
the cold period and an increase during the hot
period. In contrast to the ENSEMBLES
simulations, the projected change in energy
demand is lower during the warm period (not
exceeding +2%) compared to the cold (about 2% as in the ENSEMBLES simulations).
Why is it relevant?
Consumption of electricity is particularly
sensitive to weather (especially temperature),
since large amounts of electricity cannot be
stored and thus the electricity that is generated
must be instantly consumed. Average daily
electricity demand in most European countries
shows a single peak during winter months
(Hekkenberg et al. 2009). In Europe, only Spain,
Portugal, Italy and Greece currently show an
additional peak during summer months
(Giannakopoulos
and
Psiloglou
2006,
Hekkenberg et al. 2009).
Knowledge of the energy characteristics for an
urban area such as Athens has important policy
implications for the power industry. Power
planning and management needs to take key
climatic, demographic and economic factors into
consideration to address the pressing issue of
energy saving. This will help each local city
government to establish policies to encourage
the replacement of old inefficient electrical
equipment with energy saving equipment. If
feasible, introducing renewable energy systems
should also be encouraged in the light of
increased levels of energy demand due to
climate change. Climate change may, in turn,
result in summer electricity demand peaks, seen
already in Athens, affecting in this way,
generation capacity, maintenance scheduling and
pricing.
The peak in cooling energy demand falls in the
dry season. Low water supply can reduce energy
production from hydroelectric plants, as well as
from conventional power plants, which require
water for cooling and for driving the turbines. As
a result, production may not match energy
demand in the warm period of the year.
Additional capacity may need to be installed
unless adaptation / mitigation strategies or
stricter building regulations to improve
insulation are established. Adaptation measures
such as building codes could help minimize the
increase in energy demand during heat-waves.
11
Ozone exceedance days
What is it?
The relationship between daily eight-hour maximum temperature and ozone concentration for Athens
is shown in Giannakopoulos et al. 2010. The probability of ozone exceedance days (defined as days
with maximum 8-hour average ≥ 60 ppb, i.e., according to EU Directive 2008/50/EC) increases over
the temperature range 17 to 38°C, mainly due to high activity of photo-chemical processes. Figure 11
(top) depicts the ozone probability distribution for temperatures above 17°C using output from one of
the ENSEMBLES RCMs (the KNMI model) in order to estimate the potential impact of increasing
future temperatures on ozone exceedance days in the Greater Athens Area. The same approach was
also applied for the CIRCE simulations (ENEA, MPI and IPSL) (Figure 11, bottom).
Figure 11: Ozone probability distributions for temperature above 17 °C, for top: the observed (1990-1999)
(black), the 2021-2050 (green) and the 2071-2100 (red) periods for KNMI model. Bottom: the observed
period (1990-1999) (green), and three projections for period 2021-2050 (grey – ENEA; red – IPSL; blackMPI). Values are plotted at the mid-point concentration of each 5ppb ozone concentration bin.
12
What does this show?
Figure 11 (top) indicates that the number of
ozone exceedance days is projected to increase.
For the future period 2021-2050, the percentage
increase is about 8%, amounting to about seven
extra ozone exceedance days per year. For the
end of the 21st century (2071-2100), the
percentage increase is about 30%, which is
almost an additional month (about 27 days) of
ozone exceedance days per year. The same
analysis for the CIRCE models simulations
indicates that for the period 2021-2050 ozone
exceedance days are expected to increase by 4
days for MPI, 8 days for ENEA and 34 days for
IPSL (Figure 11, bottom).
Why is it relevant?
It is already known that day-to-day changes in
weather affect both the severity and duration of
pollution episodes. In a future climate-change
world, faster chemical reactions, increased
4. Socio-economic trends
Since climate is not the only driver of change,
future changes in non-climate drivers and
vulnerabilities can accentuate or diminish future
climate change impacts. For urban case studies,
population
growth,
GDP
and
energy
consumption are key issues. In Greece, the urban
population is generally projected to increase by
30% by the mid century (Figure 12a). In
contrast, a low annual growth rate for Athens
indicates that the population will not increase
significantly in the future (Figure 12b).
Total energy consumption is projected to
increase 2.5% by 2030s. However, the domestic
sector is estimated to increase 40% (Table 1).
Given the increase in urban population, rising
domestic energy consumption mainly concerns
biogenic emissions, and stagnation may
contribute to an increase in the occurrence of
ozone pollution episodes. Ozone episodes are
closely linked to adverse effects on human
health, vegetation and ecosystems, and there are
documented relationships between ozone
exceedances and hospitalization. It has
previously been shown that daily maximum
temperature variation accounts for much of the
influence of meteorological variables on ozone
(Lin et al. 2001, Bloomer et al. 2009). This
strong correlation of ozone episodes with
temperature is primarily associated with surface
air ventilation, since high temperatures are
typically a result of stagnant air. The ozonetemperature association also depends on local
ozone production chemistry and on temperaturedependent biogenic emissions (Jacob et al.
1993).
cities. The only sector indicating future
decreases in future energy consumption is the
rural sector, explained by an increase in
urbanization and decrease in rural population.
The increase in energy consumption appears to
be in line with the increase in GDP (Figure 12c).
IMF projections for the Greek GDP take into
account the current economic recession since
GDP is estimated to decline slightly in 2010 and
2011. IMF projections show an increase
thereafter although the current economic crisis
could be deeper and therefore last longer than
forecast. Given the rise in GDP, life expectancy
is also on the rise, reaching or exceeding 80
years of age by 2050 (Figure 12d).
Table 1. Projected energy consumption per activity sector. Source: Hellenic Ministry of Environment, Energy
and Climate Change ‘Analysis of energy scenarios for the attainment of 20/20/20 goals’.
Total Energy Consumption (ktoe)
per sector
Rural
2010
2015
2020
2025
2030
1065
1045
1033
1044
1051
Industry
4300
4192
4486
4936
4729
Transportation
8355
8757
9368
10018
10521
Domestic
5753
6009
6865
7544
8089
Tertiary
2059
2193
2436
2680
2884
21532
22195
24187
26222
27274
Total
13
3
2.5
2
Urban annual growth rate (%)
9000
Urba n popul a tion
Annual rate of change of
percentage urban (%)
8000
Athens popul a tion
7000
Rura l popul a tion
6000
1.5
5000
1
4000
0.5
3000
2040-2045
2030-2035
2020-2025
2010-2015
2000-2005
1990-1995
1980-1985
1970-1975
1960-1965
1950-1955
0
Figure 12a: Annual urban population growth rate,
Greece. Source: World Population Prospects: The
2006 Revision and World Urbanization Prospects:
The 2007 Revision, http://esa.un.org/unup
23500
23000
22500
Gross domestic
product per capita,
current prices
Gross domestic
product, current prices
(in billions)
0
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050
Figure 12b: Urbanisation (urban and rural
population) Source: World Population Prospects: The
2006 Revision and World Urbanization Prospects: The
2007 Revision, http://esa.un.org/unup
100
260
90
255
80
250
70
240
21500
1000
265
245
22000
2000
235
Life expectancy at birth for both sexes (years)
60
50
World
40
Greece
5. Uncertainties
Uncertainties are present in all stages of the
CIRCE
case-study
assessment
from
observations, to emissions, models, projections,
impacts and adaptive response. Bias is evident in
comparisons of observed and simulated (ENEA,
MPI and IPSL models) maximum and minimum
annual temperature (Figure 13) and annual
precipitation (Figure 14) for the Athens case
study. The cold bias is most clearly evident for
minimum temperature in all model output.
Maximum temperatures are better simulated
especially for the ENEA model. Precipitation
appears generally well simulated for Athens. The
models, however, are not able to capture extreme
years: There are no simulated values comparable
2085-2090
2070-2075
2055-2060
2040-2045
2025-2030
2010-2015
1995-2000
Figure 12c: Gross domestic product in billions (red
line, right axis) and Gross domestic product per
capita (green line, left axis) for Greece. Source: IMF
World Economic Outlook Database, April 2010
1980-1985
2008 2009 2010 2011 2012 2013 2014 2015
1965-1970
230
1950-1955
30
21000
Figure 12d: Life expectancy at birth in Greece and the
world. Source: UN World Population Prospects: The
2008 Revision, http://esa.un.org/unpp
to the exceptionally high total annual
precipitation observed in Athens in 2003. This is
expected as CIRCE model runs are not
reanalysis-forced runs so one would not expect
year-to-year consistency, although some extreme
years are evident in the simulations. To minimise
the effects of bias, CIRCE projections are
generally presented in the case-study work as the
difference between the future (2021-2050) and
control (1961-1990) periods. In the case of
extreme events (Figures 15, 16 and 17), biases in
the absolute magnitude of events (but not in the
shape of the distribution) are minimised by using
percentile thresholds calculated from the model
control period rather than from observations.
14
Athens_Annual Maximum Temperature
30
25
20
15
10
obs
ipsl
model mean
5
enea
mpi
2050
2040
2030
2020
2010
2000
1990
1980
1970
1960
1950
1940
1930
1920
1910
1900
1890
0
Athens_Annual Minimum Temperature
30
obs
ipsl
model mean
25
enea
mpi
20
15
10
5
2050
2040
2030
2020
2010
2000
1990
1980
1970
1960
1950
1940
1930
1920
1910
1900
1890
0
Figure 13: Observed and projected mean annual
maximum (top) and minimum (bottom)
temperatures for Athens calculated using output
from the ENEA, MPI and IPSL models, 1950-2050
Athens_Total Annual Rainfall
1200
obs
ipsl
model mean
1000
enea
mpi
800
600
400
200
2050
2040
2030
2020
2010
2000
1990
1980
1970
1960
1950
1940
1930
1920
1910
1900
1890
0
Figure 14: Observed and projected total annual
rainfall for the urban case studies using output from
the ENEA, MPI and IPSL models, 1950-2050.
It is important to consider the ensemble range as
well as the ensemble mean. The precipitationbased indicators for the urban case studies
(Figure 17) are based on only three of the
CIRCE models: ENEA, MPI and IPSL – so
caution is needed. Caution is also needed in
using the biogeophysical and vulnerability
indicators for Athens. While the fire-risk
analysis is based on six RCM runs from the
ENSEMBLES project, the air pollution, human
health and energy demand analyses are all based
on a single ENSEMBLES RCM. However, these
use projected changes in temperature which are
generally more robust than precipitation. All the
CIRCE models (and ENSEMBLES RCMs)
agree in the direction of temperature change: an
increase in both mean and high temperature
extremes, although there is some uncertainty in
terms of the magnitude of change.
While the CIRCE models generally agree in
projecting a decrease in mean annual
precipitation (Figure 5), the uncertainties are
greater at the seasonal scale and particularly with
respect to extremes (see Figure 17). This is in
part due to the higher inter-annual variability of
precipitation, but also to the higher spatial
variability of precipitation and the finer spatial
scale of the underlying physical processes –
particularly with respect to convective events.
The large inter-annual variability of precipitation
makes it difficult to identify statistically
significant trends in observed (Giannakopoulos
et al. 2010) or simulated (Figure 5) precipitationbased climate indicators.
For temperature, the observed trends and
climate change signal are stronger. Figure 2
shows observed and simulated (ENEA, MPI and
IPSL models) mean annual Tx and Tn for the
urban case studies as anomalies from the 19611990 baseline to correct for the model biases
shown in Figure 13. In general, the projected
changes appear as an extension of the observed
positive trends. Most notably for Athens Tx, the
projected changes do not extend much beyond
the upper range of the observations. The
influence of the hot summer of 2007 is evident
in the 2007 observed Tx anomaly for Athens
which appears as an outlier with respect to the
observed series but falls centrally within the
model range by 2050. Similar features are seen
in the case of temperature extremes – very hot
days and very hot nights (Figure 3), although the
inter-annual variability is greater and the upper
range of the observations is not exceeded that
frequently in the future.
15
seasonal tn95n
60
ensemble
enea
ingv
ipsl
mpi
cnrm
50
40
30
20
10
0
DJF
MAM
JJA
SON
Figure 15: Projected changes (2021-2050 minus
1961-1990) in the number of very hot nights
(Tn95n) based on CIRCE model data for Athens.
30.00
summer tx95n
25.00
20.00
15.00
10.00
5.00
0.00
enea
ingv
ipsl
mpi
cnrm
ensemble
Figure 16: Projected seasonal changes (2021-2050
minus 1961-1990) in the number of very hot
summer days (Tx95n) based on CIRCE model data
for Athens. The ensemble-mean change is shown,
together with changes for five individual models.
65.0
55.0
cdd (days)
45.0
pq90 (mm/day)
35.0
px3d (mm)
25.0
15.0
5.0
-5.0
-15.0
-25.0
-35.0
enea
ipsl
mpi
Figure 17: Projected changes in three precipitation
extremes indices for Athens calculated using daily
output from the ENEA, IPSL and MPI models: the
maximum dry spell length (consecutive dry days:
cdd), heavy precipitation (the 90th percentile of
daily precipitation: pq90) and maximum three-day
precipitation (px3d). A threshold of > 0.5 mm was
used to define a rain day
The quantitative impacts assessments undertaken
for the Athens case study are based on statistical
rather than process-based modelling. Most of the
statistical models are linear. It may be reasonable
to assume linearity and stationarity given certain
underlying assumptions in the model – which
should be made clear to users. In the case of the
energy demand model for Athens, for example,
it is explicitly assumed that technology use is the
same in the control and future periods. Thus no
allowance is made for a potential drive towards a
more energy efficient economy or for other
socioeconomically-driven trends. A somewhat
different approach is taken in the Athens health
risk work, where
an
adaptation or
acclimatisation factor of 1°C per decade is used
in an attempt to account for behavioural and
physiological changes.
It should also be noted that, when
constructing the impacts models, absolute
thresholds of temperature were used. This choice
gives a clear picture of the underlying impact to
relevant stakeholders. However, the projected
impact of future climate is very much dependent
on the model’s bias in comparison to
observations. As an example, the IPSL model
has the largest cold bias so it projects virtually
no increase in mortality when health impacts are
considered. Another option would be to work
with percentile rather than absolute thresholds.
In this case, IPSL would show the highest
increase in mortality since it is the model that
presents the greatest differences between future
and baseline climate. Obviously, each option
presents its advantages and disadvantages but in
the Athens case study we have chosen to work
with absolute thresholds.
Ideally, for the Athens and other urban case
studies, impacts and vulnerability indicators
would also be considered at high spatial
resolution. In terms of the health implications of
high ozone days, for example, the distribution of
vulnerable groups (the young and elderly, those
with pre-existing conditions, outdoor workers)
may vary spatially across the city and particular
‘hot spots’ of pollution may occur, for example
downwind of industrial areas.
Thus there are uncertainties inherent to all
stages of the CIRCE case-studies integrated
assessments. The uncertainties in the climate
16
projections are perhaps the easiest to quantify
and demonstrate to stakeholders. Nonetheless, it
is important to consider all aspects of uncertainty
in the context of adaptation decision making.
6. Integrated assessment
The Greater Athens area, i.e., the city with its
suburbs and the Attica peninsula in general,
comprises almost half of the Greek population
(circa 5 million inhabitants) and consequently
represents a major part of the economic, social,
cultural and administrative activities of the
country. The occurrence of a severe heat wave in
Athens can therefore paralyse all socioeconomic
sectors and culminate in a substantial increase in
energy demand (for air conditioning needs) and
a deleterious effect on human health (due to the
high population density). In Athens, urbanization
constitutes an additional aggravating factor since
the urban fabric retains night temperatures at
high levels and hinders biological cooling.
Additionally, poor air quality contributes to the
increased health risks in Athens. Persistent high
summer temperatures also increase significantly
peri-urban forest fire risk (Giannakopoulos et al.,
2009).
Table 2 provides a summary of climate
indicators values based on the CIRCE model
climate simulations for the ‘present’ 1961-1990
and ‘mid-century’ 2021-2050 periods, as well as
the changes between the two periods. Table 3
provides a summary of climate trends based on
the CIRCE model simulations for these same
periods. The negative, though not statistically
significant, temperature trends for the present
climate are in contrast to the statistically
significant increasing trends simulated for the
mid-century climate. Table 4 summarises the
biogeophysical
and
social
vulnerability
indicators for the ‘mid-century’ 2021-2050, and
the ‘end of century’ 2071-2100 periods based on
numerical simulations derived from the
ENSEMBLES and CIRCE models.
Table 2: Mean values and trends of some key variables simulated by three coupled models for the Athens area
(Figure 1) based on ensemble averages, for the present, 1961-1990 and ‘mid-century’, 2021-2050 periods for
a set (ENEA, MPI, IPSL) of numerical simulations. Mean ‘long-term’ changes between 1961-1990 and 20212050 are also shown.
Climate indicator
present climate
(1961-1990)
mid-century
(2021-2050)
Long-term changes from 19611990 to 2021-2050
T max (°C)
20.2±0.3
21.8±0.2
+1.3±0.3
T min (°C)
11.4±0.1
12.9±0.1
+1.6±0.2
Hot days (days)
18±1
39±2
+21±1
Hot nights (days)
18±1
49±2
+30±2
Precipitation (mm)
365±11
288±12
-77±16
17
Table 3: Summary of trends for the present 1961-1990, ‘mid-century’ 2021-2050 and ‘long-term’ 2061-2050
periods, based on a set (ENEA, MPI, and IPSL) of numerical simulations.
Climate indicator
present climate trend
(1961-1990)
mid-century trend
(2021-2050)
long-term trend
(1961-2050)
T max (°C/decade)
-0.07±0.07
+0.62±0.081
+0.27±0.021
T min (°C/decade)
-0.03 ±0.06
+0.59±0.071
+0.26±0.021
Hot days (days/decade)
+1.3 ±1.0
+8.2±1.51
+3.5±0.31
Hot nights (days/decade)
+1.4±1.2
+10.6±1.61
+5.1±0.31
Precipitation (mm/decade)
+8.7 ±1.3
-23±101
-12.3±2.11
1
denotes that the trend is significant at the 95% level. The statistical significance of the trends was examined
using the Student’s t-test.
Table 4: Summary of differences for the biogeophysical and social vulnerability indicators between the ‘midcentury’ 2021-2050 or the ‘end of century’ 2071-2100 and the present climate period based on numerical
simulations derived from CIRCE and ENSEMBLES models.
Biogeophysical /
Social Vulnerability Indicator
Fire risk (days/year)
Extreme fire risk (days/year)
Excess deaths
Energy demand – cold season
Energy demand – warm season
Ozone exceedance days
CIRCE
mid-century
(2021-2050)
trends
ENSEMBLES
mid-century
(2021-2050)
trends
ENSEMBLES
end of century
(2071-2100)
trends
(↑) 20
(↑) 14
(↑) 20
(↑) 10
(↑) 50
(↑) 40
(↑) 63%1
(↑) 70%
(↑) 60%
(↓) 2%
(↑) 2%
(↓) 2%
(↑) 5%
(↓) 5%
(↑) 15%
(↑) +6days2
(↑) 8% (+7days)
(↑) 30% (+27
days)
Impact
significant
increase in periurban forest fires
increase in heatrelated mortality
increase in
energy
consumption
increase in
ozone episodes
1
: Except for IPSL that fails to capture high summer temperatures. Future excess deaths are calculated with 1°C
adaptation for the mid-future and 2°C for the end of the century.
2
: Average from ENEA and MPI. The projected IPSL increase is +34 days.
Acknowledgements
CIRCE (Climate Change and Impact Research: the Mediterranean Environment) is funded by the Commission of
the European Union (Contract No 036961 GOCE) http://www.circeproject.eu/. The CIRCE model date were
provided by INGV-CMCC (Gualdi Silvio, Enrico Scoccimarro); MF (Florence Sevault, Clotilde Dubois); IPSLCNRS (Laurnet Li); ENEA (Alessandro Dell'Aquila, Adriana Carillo); MPIM-HH (Alberto Elizalde Arellano).
18
References
► Bloomer, B. J., Stehr, J. W., Piety, C. A., Salawitch, R. J., and Dickerson, R. 2009. Observed relationships of
ozone air pollution with temperature and emissions, Geophysical Research Letters, 36.
► Giannakopoulos C. and B.E. Psiloglou, 2006. Trends in energy load demand for Athens, Greece: weather and
non-weather related factors, Climate Research, 31, 97-108.
► Giannakopoulos C., LeSager, P. Bindi, M. Moriondo, M. Kostopoulou, E. & Goodess, C. 2009. Climatic
►
►
►
►
►
►
►
►
►
►
►
►
►
►
changes and associated impacts in the Mediterranean resulting from a 2ºC Global Warming, Global and
Planetary Change, 68( 3), 209-224.
Giannakopoulos C., M. Hatzaki, and E. Kostopoulou, 2009. Information sheet on observed climate indicators
for the urban case studies: Athens, M. Agnew and and C. Goodess editors, Climatic Research Unit, School of
Environmental Sciences, University of East Anglia, Norwich, UK.
Giannakopoulos C., M. Hatzaki, K. Varotsos, and E. Kostopoulou, 2010. Biogeophysical and social
indicators: Urban case studies information sheet: Athens M. Agnew and and C. Goodess editors, Climatic
Research Unit, School of Environmental Sciences, University of East Anglia, Norwich, UK
Good P., M. Moriondo, C. Giannakopoulos and M. Bindi, 2008. The meteorological conditions associated
with extreme fire risk in Italy and Greece: relevance to climate model studies, Int. J. Wildland Fire, 17, 1-11.
Hekkenberg M, Benders RMJ, Moll HC, Schoot Uiterkamp AJM. 2009. Indications for a changing electricity
demand pattern: the temperature dependence of electricity demand in the Netherlands. Energy Policy,
37:1542–51
IPCC, 2007. Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II
to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, M.L. Parry, O.F.
Canziani, J.P. Palutikof, P.J. van der Linden and C.E. Hanson, Eds., Cambridge University Press, Cambridge,
UK, 976 pp
Jacob, D. J., Logan, J. A., Gardner, G. M., Yevich, R. M., Spivakovsky, C. M., Wofsy, S. C., Sillman, S., and
Prather, M. J., 1993: Factors Regulating Ozone Over the United States and Its Export to the Global
Atmosphere, Journal of Geophysical Research, 98, 14817-14826
Lin, C., Jacob, D. J., and Fiore, A. M. 2001. Trends in exceedances of the ozone air quality standard in the
continental United States, 1980-1998, Atmospheric Environment, 35, 3217-3228
Moriondo M., P. Good, R. Durao, M. Bindi, C. Giannakopoulos, J. Corte-Real, 2006. Potential impact of
climate change on fire risk in the Mediterranean area, Climate Research, 31, 85-95.
Nakićenović, N., and R. Swart (eds.) 2000. Special Report on Emissions Scenarios. A Special Report of
Working Group III of the Intergovernmental Panel on Climate Change. Cambridge University Press,
Cambridge, United Kingdom and New York, NY, USA, 599 pp.
Psiloglou, B.E., C. Giannakopoulos, S. Majithia, M. Petrakis, 2009, Factors affecting electricity demand in
Athens, Greece and London, UK: A comparative assessment, Energy, 34, 1855–1863
UNDP (United Nations Development Programme), 2002, Energy for sustainable development: a policy
agenda. UN, New York.
van der Linden P. & Mitchell, J.F.B. (eds.) (2009). ENSEMBLES: Climate Change and its Im-pacts:
Summary of research and results from the ENSEMBLES project. Exeter EX1 3PB, UK: Met Office Hadley
Centre, FitzRoy Road, 160pp. http://ensembles-eu.metoffice.com/docs/Ensembles_final_report_Nov09.pdf
Van Wagner CE, 1987. Development and Structure of the Canadian Forest Fire Weather Index System.
Canadian Forestry Service, Forestry Technical Report 35.
Viegas DX, Bovio G, Ferreira A, Nosenzo A, Sol B, 1999. Comparative study of various methods of fire
danger evaluation in southern Europe. International Journal of Wildland Fire 9, 235–246.
Authors: Christos Giannakopoulos, Maria Hatzaki, Kostas V. Varotsos, Basil Psiloglou, Anna Karali
Contact: Christos Giannakopoulos: [email protected].
Editors: Maureen Agnew ([email protected]) and Clare Goodess ([email protected]), Climatic Research
Unit, School of Environmental Sciences, University of East Anglia, Norwich, UK
Date: August 2011
19