Transcript Klimatologie & Hydrologie II

Eidgenössische Technische Hochschule Zürich
Swiss Federal Institute of Technology
European Heat Waves
in a Changing Climate
Christoph Schär and Erich Fischer
Atmospheric and Climate Science, ETH Zürich
http://www.iac.ethz.ch/people/schaer
Thanks to: Peter Brockhaus, Christoph Buser, Cathy Hohenegger,
Sven Kotlarski, Hans-Ruedi Künsch, Dani Lüthi, Elias Zubler
1
WCRP / UNESCO Workshop on Extreme Climate Events, Paris, September 27-29, 2010
Schär, ETH Zürich
European heatwave 2003
July 20-27 2010 temperature anomaly
(NASA)
(Reto Stöckli, ETH/NASA)
August 2003 temperature anomaly
Russian heatwave 2010
10 y
100 y
1000 y
2
10 y
mean
Schär, ETH Zürich
Return
period
100 y
1000 y
(Schär et al. 2004, Nature)
“There was nothing similar to this on the
territory of Russia during the last one
thousand years."
Alexander Frolov
Head of the Russian Meteorological Center
Contents
Prologue
Basic considerations
Heatwave impacts
Scenarios and observations
Key uncertainties
Epilogue
3
Schär, ETH Zürich
Prologue
Climate change
is much more certain than
some people believe!
The impacts of climate change
are much more uncertain than
most people believe!
4
Schär, ETH Zürich
Predicting the water cycle: a key challenge
Physics is complicated and
rather poorly understood:
Depends upon multi-scale and multiprocess interactions:
• From micrometer to gigameter.
• Involves dynamics, aerosols,
clouds, radiation, convection,
land-surface processes, etc.
5
Schär, ETH Zürich
Contents
Prologue
Basic considerations
Heatwave impacts
Scenarios and observations
Key uncertainties
Epilogue
7
Schär, ETH Zürich
Extreme events defined
Mean warming
Probability
CTRL
SCEN
Temperature
cold extremes
Extreme events are defined in
statistical terms, as events that
deviate strikingly from the
statistical mean.
8
Schär, ETH Zürich
warm extremes
Climate change implies changes in
frequency of extremes.
Generally needs to account for changes
of the whole statistical distribution
(i.e. mean, variability, skewness, etc)
Frequency of daily summer temperatures
CTL: 1961-1990
SCN: 2071-2100
9
Schär, ETH Zürich
France
Iberian Peninsula
Increase in variance
Increase in skewness
PRUDENCE, CHRM (ETH) model
Fischer and Schär 2009; Clim. Dyn.
Validation of daily summer temperatures
OBS: 1961-1990
CTL: 1961-1990
France
Iberian Peninsula
Regional variations captured!
10
Schär, ETH Zürich
PRUDENCE, CHRM (ETH) model
Fischer and Schär 2009; Clim. Dynam, Observations from Haylock et al. 2008
Contents
Prologue
Basic considerations
Heatwave impacts
Scenarios and observations
Key uncertainties
Epilogue
11
Schär, ETH Zürich
Impacts of the summer 2003 in Europe
Agricultural losses:
12.3 Billion US$
(SwissRe estimate)
Shortage of electricity,
peak prices on spot market
(EEX, Leipzig)
Serious problems with
- freshwater resources (Italy)
- forest fires (Portugal)
- freshwater fish (Switzerland)
Heat excess mortality:
estimates 35’000 to 70’000
August 2003 temperatures relative to 2000-2002, 2004
(Reto Stöckli, ETH/NASA, MODIS)
12 Schär, ETH Zürich
Excess mortality in France
Normalized mortality = mortality 2003 / longterm mean
+200%
+150%
+100%
+50%
0%
-50%
Note: no harvesting effect
Date: August 1 - November 30, 2003
13
Schär, ETH Zürich
(INSERM 2004)
Key health impact factors
Not only peak temperatures, but:
(1) Length of heat wave (accumulation effect)
=> study heat wave duration
(2) Daytime AND nighttime temperatures (sleep deprivation)
=> study number of hot days and nights
(3) Relative humidity (heat stress)
=> study “apparent temperature” or “heat index”
14
Schär, ETH Zürich
Contents
Prologue
Basic considerations
Heatwave impacts
Scenarios and observations
Key uncertainties
Epilogue
15
Schär, ETH Zürich
Regional climate scenarios
ENSEMBLES Project:
Transient experiments: 1950-2100
Greenhouse gas scenario: A1B
Available models (period 1950-2100):
6 GCM / RCM pairs:
GHG scenario
GCM
Coupled AOGCM
(~120 km)
RCM / Group
CLM / ETH
HadCM3Q0
RCM
(25 km)
HadRM / HC
HadCM3Q16
RCA3 / C4I
RACMO / KNMI
ECHAM5
REMO / MPI
RCA / SMHI
16
Schär, ETH Zürich
(EU-Project ENSEMBLES, Coordinated by John Mitchel, Hadley Centre
Changes in mean and variance
2071-2100 versus 1961-1990
JJA T2m change
99th percentile [K]
JJA mean warming [K]
Change in JJA daily
variability [K]
Strongest 99th percentile increase to the north of strongest mean warming
Good agreement with PRUDENCE results (Kjellström et al. 2007, Fischer and Schär 2009)
Differences between mean and 99th percentile is due to variability changes
17
Schär, ETH Zürich
ENSEMBLES, mean of 6 models
Fischer and Schär 2010, Nature Geoscience
Changes in variability
2070-2099 versus 1961-1990 (A2 scenario)
Interannual var
[%]
Intraseasonal var
[%]
Seasonal var
Diurnal temp range
[%]
[K]
=> Pronounced variability increases on all time-scales <=
18
Schär, ETH Zürich
Schär et al. 2004; Vidale et al. 2007; Fischer and Schär 2009 (mean of 8 PRUDENCE RCMs)
Observed increases in temperature variability?
Analysis of 54 high-quality
homogenized temperature
records from 1880-2005.
Finds a statistically significant
increase in peak temperatures.
Finds a statistically significant
variability signal.
Geographical pattern of trends is
consistent with climate change
scenarios:
• variability increases has
maximum in Central Europe
• warming has maximum in
Iberian Peninsula
19
Schär, ETH Zürich
(Della Marta et al., 2007, JGR
Observed relationship between mean and variance?
Correlation between mean and variance
over Europe, summer
(ECA&D data set)
variance
mean
Temporal evolution of mean and variance,
Summer daily maximum temperature
(La Rochelle, France)
Evolution of mean and variance
correlate!
Warm periods have
higher variability
20
Schär, ETH Zürich
(Parey et al., 2010)
(Yiou et al., 2009)
Do peak temperature explain health impacts?
JJA T2m change
99th percentile [K]
Not really!
Use concept of “apparent temperature”
or “heat index” instead (Steadman 1979):
21
Schär, ETH Zürich

Accounts for effects of
temperature AND humidity.

Measure of the effort needed to
maintain human body temperature under
shaded and ventilated conditions.

Based on a biophysical model that accounts
for the effects of radiation, clothing,
heat-transfer, sweating, etc.
Changes in apparent temperature
1961-1990
2021-2050
2071-2100
Number of days with apparent temperature ≥ 40.6°C
(large heat stroke risk with extended exposure)
Dramatic increases in low-altitude Mediterranean (river basins and coasts)
22
Schär, ETH Zürich
ENSEMBLES, mean of 6 models, scenario A1B
Fischer and Schär 2010, Nature Geoscience
Affected regions
2071-2100
Affected river basins
Tejo
Ebro
Rhone
Po
Tiber
Danube
Geographical pattern
is consistent across
all model chains and
health indices
considered, but
amplitude strongly
depends upon model
Affected towns
Lisbon
Seville
Cordoba
Marseille
Milan
Roma
Napels
Budapest
Belgrade
Bucharest
Thessalonica
Athens
Number of days with apparent temperature ≥ 40.6°C
23
Schär, ETH Zürich
(Fischer & Schär, 2010, Nature Geoscience)
Contents
Prologue
Basic considerations
Heatwave impacts
Scenarios and observations
Key uncertainties
Epilogue
24
Schär, ETH Zürich
Role of model physics
Perturbed physics ensemble
Testing the key physics parameters at 2xCO2
(53 simulations, 2.5x3.75 deg).
Increase in July temperature extremes [factor]
(99th percentile of Tmax)
25
Schär, ETH Zürich
Ensemble range
(80% range)
(Barnett et al. 2006)
Role of land-surface processes
Summer 2003
Observations
Control simulation
Simulation with prescribed
mean soil moisture evolution
Number of hot days (Tmax > 90th percentile)
26
Schär, ETH Zürich
(Fischer et al. 2007,
see also Zaitchik et al. 2005, Seneviratne et al. 2006, Vidale et al. 2007, Fischer and Schär 2009)
Role of convection …
27
Schär, ETH Zürich
… and resolution
1 km
100 km
(Figure: Elias Zubler)
25 km
CR RCM
RCM
GCM
[mm/h]
Explicit
Parameterized
convection
convection
WET soil
CTL
DRY
-28
Schär, ETH Zürich
Soil-moisture precipitation
feedback crucially depends
upon representation of
moist convection!
[UTC]
+
[UTC]
+
[UTC]
o
--
IFS ECMWF
+
[UTC]et al, 2009; Brockhaus
[UTC]
[UTC]
(Hohenegger
et al. 2010)
Role of model biases
Special role of biases for analysis of extremes:
•
biases in statistical distribution beyond biases in mean (i.e. in variability)!
•
impacts often depend on absolute thresholds => biases particularly crucial!
Model biases are state dependent!
Assumptions about bias changes matter!
Analysis of ERA-40 driven RCM
simulations (ENSEMBLES)
Bayesian methodology using ENSEMBLES
results (2021-2050) with TWO bias assumptions:
(Christensen et al. 2008)
29
Schär, ETH Zürich
PDF
Observed monthly temperature [deg C]
Alps
PDF
Temperature bias [K]
constant bias | constant relation | joint estimate
Scandinavia
T [K]
T [K]
(Buser et al. 2009; Buser et al. in press)
Contents
Prologue
Basic considerations
Heatwave impacts
Scenarios and observations
Key uncertainties
Epilogue
30
Schär, ETH Zürich
Attribution of extreme events to climate change
Probability of 2003-like heat waves
in natural versus anthropogenic climates
Associated
liability issues
natural
anthropogenic
Attributable to human influence
(in a probabilistic sense).
Might ultimately lead to liability claims.
Easier with heat waves than other extremes.
31
Schär, ETH Zürich
(Report by Munich Re, 2010)
(Stott et al. 2004; see also: Schär and Jendritzky 2004; Allen and Lord 2004
Summary
Observations and models show relevance
of warming AND changes in variance.
Peak temperatures:
- changes not co-located with changes in mean,
- but affected by variance changes.
Health impacts:
- changes not co-located with changes in peak temperature,
- most significant impact in low-altitude river basins and along coasts,
- Pattern robust but amplitude uncertain.
Major uncertainties related to
- soil-moisture precipitation feedback (land-surfaces and convection),
- bias assumptions.
32
Schär, ETH Zürich