Drivhuseffekt og global opvarmning - circle-2

Transcription

Drivhuseffekt og global opvarmning - circle-2
Uncertainties in different elements of
the cause-impacts-adaptation chain
(From the perspective of RCMs and data)
Ole B Christensen
Danish Climate Centre
Danish Meteorological Institute
Annual maximum daily
precipitation
Overview
• Model uncertainty, and how to make
it smaller
• The more detailed questions, the
more uncertainty!
• We have uncertainty, but we still
have a quantitative message
• Lesson learned from PRUDENCE and
ENSEMBLES
– The simple-minded approach towards
model weighting
Climate of the future
AR4, relative to 1980-1999
Categories of uncertainties
• Scenario What will happen in the future
• Structural Models disagree
– Sensitivity Warming per CO2 increment
– Model Different models have different regional results
• Statistical
Even the same model does not agree…
• Critical thresholds
If the Copenhagen public
transportation system breaks down, and Arland has control
tower problems, and my plane has a technical problem,
then the meeting agenda breaks down
• The ”unknown unknowns”
Categories of uncertainties:
What to do
• Scenario
– We cannot do predictions
– Pattern scaling, though not for
everything (sea level!)
• Structural
– Sensitivity
• Better models; super-ensembles
– Model
• Better models; higher resolution; superensembles-based probability distributions
• Statistical
– Ensembles of model runs
Example: Extreme precipitation
Average;
99%,
99.5%,
Average;
90%,
95%,
99%,
99.5%,
99.9%
Average;
90%,
95%,
99.9%
Average; 90%,
90%, 95%,
95%, 99%,
99%, 99.5%,
99.5%, 99.9%
99.9%
Uncertainty on extremes
Present
Future
Uncertainty on extremes
Relative change exp1 (%)
Relative change exp2 (%)
Number of points in Europe with a positive signal
Black line: Land points
Orange line: All points
PRUDENCE: It all depends on what
we look at
PRUDENCE domains
Temperature change – sources
of spread
100
90
DJF
80
% variance
70
60
RCM
Scenario
Forcing
Member
50
40
30
Depends on
driving model
20
10
0
1
2
3
4
5
6
7
8
subdomain
100
90
JJA
80
% variance
70
60
RCM
Scenario
Forcing
Member
50
40
Also on RCM
and scenario
30
20
10
0
1
2
3
4
5
subdomain
6
7
8
Déqué et al. 2007
Precipitation change – sources
of spread
100
90
DJF
80
% variance
70
60
RCM
Scenario
Forcing
Member
50
40
30
Driving GCM
and RCM
20
10
0
1
2
3
4
5
6
7
8
subdomain
100
90
JJA
80
% variance
70
60
RCM
Scenario
Forcing
Member
50
40
30
RCM quite
important
20
10
0
1
2
3
4
5
subdomain
6
7
8
Déqué et al. 2007
ENSEMBLES: More data, more
systematic approach
-and some model weighting
Public data access:
http://ensemblesrt3.dmi.dk/
ENSEMBLES GCM-RCM Matrix
Global model
METO-HC METO-HC METO-HC
Regional
Standard Low sens. Hi sens.
inst.
METO-HC
2100
2100*
2100*
MPIMET
MPIMET
Standard
MPIMET MPIMET
Ens.m. 1 Ens.m. 2
IPSL
2100*
1
2100
3
2100*
1
2100
2100*
2100
ICTP
2100
SMHI
2100*
2100*
1+4
2100*
1
2100*
2100*
2100*
3+1
2100
1
2050
C4I
2100*
2
2050 (A2)*
GKSS
1
2050*
2050*
1
2050*
CHMI
1
2050*
OURANOS**
2050*
VMGO**
2050*
Total (19512050)
5
Total
number
2
2050*
2100
KNMI
METNO
MIROC CGCM3
4
2100
DMI
UCLM
NERSC
2100
(late 2010)
CNRM
ETH
CNRM
1
1
2
2
6+2
1+1
0+1
2
3
3
0+1
1
25+5
Red: Online now; *: non-contractual runs; **:affiliated partners without obligations; underscore: 50km resolution;
(in parantheses): Expected date. For partner acronym explanations, see the participant list. NOTE that all partners
also did an ERA-40 driven analysis 1951(1961)-2000
Summer temperature
11 RCM simulations covering
1961-2099 with OK quality
Summer precipitation
Summer precipitation
extremes
1 yrv
5 yrv
30 yrv
RCM weighting scheme
f1: large scale circulation based on a weather regime classification
f2: meso-scale signal based on seasonal temperature and
precipitation analysis
f3: probability density distribution match of daily and monthly
temperature and precipitation analysis
f4: extremes in terms of re-occurrence periods for temperature and
precipitation
f5: trend analysis for temperature
f6: representation of the annual cycle in temperature and
precipitation
Total weight from multiplication of the individual weights:
-This could roll down the model avalanche!
Resulting weights
2y return value of storm surge
Largest change around 30cm
Relevant projects wrt. integrated
modelling
• PESETA Bottom-up analyses of
impacts. Contribution to EU Green
Paper on adaptation
• ClimateCost Bottom-up including
mitigation. Adaptation benefits, cost
of inaction…
Summary
• Many uncertainty sources. We need to:
– Formally quantify what we know and don’t
know.
• E.g.: Both positive and negative summer
precipitation changes, but most likely negative
– Communicate it
• Model data are abundant. Methods for
interpretation of ensemble data are
required
• Weights: Might work…
• Integrated model work is necessary