Transcript Frank Selten

How to link large-scale circulation structures to local extremes ?
Frank Selten and Deb Panja
Royal Netherlands Meteorological Institute
“Extreme Associated Functions: Optimally Linking Local Extremes
to Large-scale Atmospheric Circulation Structures”
In discussion in Atmospheric Chemistry and Physics Discussions
Motivation
• Local weather extremes are usually connected
to typical large-scale circulation anomalies
Examples:
extreme rainfall in the UK
daily mean summer temperatures in the Netherlands
Floods in the UK
Average rainfall May – July 2007
Impressions
British always make the best of it …..
UK July rainfall and Z500 anomalies
Teleconnection Z500
Daily JA temperatures in Holland
Z500 anomaly
Z500 anomaly
Motivation
• Local weather extremes are usually connected
to typical large-scale circulation anomalies
• Probability of occurrence of these structures
impact probability of the local extremes
Motivation
• Local weather extremes are usually connected to
typical large-scale circulation anomalies
• Probability of occurrence of these structures
impact probability of the local extremes
• Future probability of local extremes depends on
the response of circulation to the CO2 forcing
Motivation
• Local weather extremes are usually connected to
typical large-scale circulation anomalies
• Probability of occurrence of these structures
impact probability of the local extremes
• Future probability of local extremes depends on
the response of circulation to the CO2 forcing
• Uncertainties in circulation changes lead to
uncertainties in local weather extremes
Identification of circulation
structures that are optimally linked
to local extremes enables:
• model evaluation against observations
• diagnosis of cause of discrepancies; maybe not
due to circulation but clouds, soil-moisture or
radiation deficiencies
• evaluation and intercomparison of simulated
changes in extremes
• dynamical understanding of circulation changes
which enhances faith in simulated changes
Use information on the extremes
• Example: July and August daily mean
temperatures in De Bilt and 500 hPa
geopotential height fields over the EuroAtlantic region
EOF 1 (12.8 %)
EOF 2 ( 11.6 %)
Daily values 1958-2000
No apparent clusters by simple visual inspection …
Temperature anomaly ~ EOF1
Clear dependence …
Extreme Associated Functions
• Linear combinations of first L EOF amplitudes
that have maximum ‘tilt’ r in scatter plot with
local temperature (or wind, rain, …)
n
<..>p
b
an adjustable power to emphasize the more extreme anomalies
time average over positive anomalies only
amplitude of the new pattern
Interpretation: find the pattern that for a one standard deviation change
gives the largest change in the local temperature
Two possibilities: find c’s that maximize r2 by variational analysis:
=0
Or find the least-squares solution of the multiple linear
regression problem:
The solution is:
Comparison
Linear regression T and Z500
Composite of 5 % hottest days
EAF 1
Temperature ~ EAF 1
Robustness
• Taking only the 30% maximum temperature
anomalies leads to the same EAFs
• Varying the power from 1 to 3 leads to
qualitatively similar EAFs
• Choosing a smaller geographical region
leads to the same EAFs
Other patterns ????
• Test: synthetic temperature timeseries
• T(t) = a1(t) + a2(t)
EAF1 : sum of both patterns
Linear regression pattern
Does not reproduce the original patterns as well
Conclusion
• EAFs are a robust method to link large-scale
circulation structures to local extremes; all
contributing patterns are sumarized into one
• Next application: validate climate simulations
for present day and assess changes in
climate scenario simulation
Application to simulated data
“ESSENCE project: a 17 member ensemble of climate
SRES A1b scenario simulations from perturbed initial
conditions using the ECHAM5-MPI-OM model ”
random perturbations in
atmospheric temperatures
(< 0.1 K )
initial state
from preindustrial
control
integration
1850
1950
17 simulations
2000
2100
historical concentrations of GHG according to
GHGs and sulphate
SRES A1b
aerosols.
Streamfunction at 500 hPa
ESSENCE JA 1958-2000
ERA JA 1958-2000
Mean
Standard
deviation
Mean
Mean
Standard
Standard
deviation
deviation
Streamfunction 500 hPa EOFs
ERA
1
2
ESSENCE
Streamfunction 500 hPa EOFs
ERA
3
4
ESSENCE
Streamfunction EAFs
ERA
1
ESSENCE
T2m versus EAF1 amplitude
ERA
ESSENCE
ESSENCE
Climate change in ESSENCE
• Compare 2071-2100 period with 1958-2000
• Average across all 17 ensemble members
Temperature at 2m
ESSENCE JA climate change
Mean
Standard deviation
Streamfunction at 500 hPa
ESSENCE JA climate change
Mean
Standard deviation
Streamfunction 500 hPa EOFs
ESSENCE 2051-2100
1
2
ESSENCE 1958-2000
Streamfunction 500 hPa EOFs
ESSENCE 2051-2100
3
4
ESSENCE 1958-2000
Streamfunction EAFs
ESSENCE 1958-2000
1
2
ESSENCE 2050-2100
EAF 1 Netherlands
present
PDF of EAF amplitudes
present
future
present EAF projected on future
future wrt future climate
present EAF projected on future
wrt future climate
Pattern not changed; mere shift of PDF
future
Netherlands EAF 1
Mean change included
Mean change subtracted
EAF 1 France
present
PDF of EAF amplitudes
present
future
present EAF projected on future
future wrt future climate
present EAF projected on future
wrt future climate
Pattern not changed; mere shift of PDF
future
France EAF 1
Mean change included
Mean change subtracted
EAF 1 Spain
present
PDF of EAF amplitudes
present
future
present EAF projected on future
future wrt future climate
present EAF projected on future
wrt future climate
Pattern bit changed; PDF changes
future
Spain EAF 1
Mean change included
Mean change subtracted
EAF 1 Greece
present
PDF of EAF amplitudes
present
future
present EAF projected on future
future wrt future climate
present EAF projected on future
wrt future climate
Pattern bit changed; PDF changes
future
Greece EAF 1
Mean change included
Mean change subtracted
EAF 1 Romania
present
PDF of EAF amplitudes
present
future
present EAF projected on future
future wrt future climate
present EAF projected on future
wrt future climate
Pattern not changed; PDF changes
future
Romania EAF 1
Mean change included
Mean change subtracted
EAF 1 Moscow
present
PDF of EAF amplitudes
present
future
present EAF projected on future
future wrt future climate
present EAF projected on future
wrt future climate
Pattern not changed; PDF mere shift
future
Moscow EAF 1
Mean change included
Mean change subtracted
EAF 1 Poland
present
PDF of EAF amplitudes
present
future
present EAF projected on future
future wrt future climate
present EAF projected on future
wrt future climate
Pattern not changed; PDF slight change
future
Moscow EAF 1
Mean change included
Mean change subtracted
EAF 1 Hamar
present
PDF of EAF amplitudes
present
future
present EAF projected on future
future wrt future climate
present EAF projected on future
wrt future climate
Pattern not changed; PDF bit changed
future
Hamar EAF 1
Mean change included
Mean change subtracted
EAF 1 UK
present
PDF of EAF amplitudes
present
future
present EAF projected on future
future wrt future climate
present EAF projected on future
wrt future climate
Pattern not changed; PDF bit changed
future
UK EAF 1
Mean change included
Mean change subtracted
Conclusions
• EAF method is an objective, robust tool to
relate local extremes to large-scale
circulation structures
• Useful tool to evaluate climate simulations