Climate modelling uncertainties

1
(No Transcript)
2
VARIABILITY AND TRENDS IN THE ARCTIC
CLIMATE SIMULATED WITH THE BERGEN CLIMATE MODEL
A. Sorteberg, H. Drange, N.G.Kvamstø, T. Furevik

3
MOTIVATION

• LARGE DIFFERENCES IN PROJECTED CLIMATE CHANGE IN
  THE ARCTIC
• WHY ?
• INTERMODEL DIFFERENCES
• MODEL FORMULATIONS
• COMPLEXITY
• RESOLUTION
• CLIMATE SYSTEM UNPREDICTABILITY
• UNCERTAINTIES RELATED TO INTERNAL CLIMATE
• VARIABILITY

4
MOTIVATION
CMIP2 MODELS ZONAL TRENDS IN 2m TEMPERATURE
YEAR 31-60 (C/DECADE)
HOW MUCH OF THE SPREAD IS DUE TO
INTERMODEL DIFFERENCES AND HOW MUCH IS DUE TO
CLIMATE SYSTEM UNPREDICTABILITY ?
5
EXPERIMENTAL SETUP

• Both the ocean and atmosphere model are
  initialized in different phases of the AMOC
  from a 300 year control integration
• The coupled system are integrated for 80 years
  with the same flux adjustment terms as
  in the control integration.
• CO2 is increased with 1 per year starting with
  the same CO2 concentration as in the control
  integration (353 ppm)

6
Bergen Climate Model (Furevik et al. Clim Dyn
2003)
Atmospheric model (ARPEGE/IFS)
Coupler (OASIS)
Ocean model (MICOM)
7
MODEL CONFIGURATION

• The atmospheric grid (red dots) has a resolution
  of T63
• (2.8 by 2.8) and 31 levels from the surface
  to 0.1 hPa
• The ocean grid (blue dots) has a resolution of
  0.8 by 2.4 at the Equator, gradually
  transforming to approximate square grid cells
  towards the poles (Mercator projection) and 26
  isopycnic vertical layers

8
NAO SIGNAL IN BCM CNTRL
NAO VARIABILITY
MSLP(Gibraltar)-MSLP(Iceland)
9
CHANGES IN NAO IN CMIP2 SIMULATIONS
CHANGES IN NAO
MSLP(Gibraltar)-MSLP(Iceland)
10
AMOC IN BCM CNTRL
AMOC VARIABILITY
11
AMOC IN CMIP2 SIMULATIONS
AMOC Change Relative to CTRL Mean -2.2 Sv
12
SEAICE CNTRL AND CMIP2
Purple Control integration White At
doubling of atm CO2
SEA ICE EXTENT
September
March
13
BCM SPREAD vs MULTIMODEL SPREAD
ENSEMBLE MEAN ZONAL TRENDS IN 2m TEMPERATURE
YEAR 31-60 (C/DECADE)
14
CHANGES IN ENSEMBLE SPREAD DUE TO AVERAGING
PERIOD
HOW SENSITIVE IS THE BCM ENSEMBLE SPREAD TO
AVERAGING PERIOD ? TRENDS BASED ON LAST 20
YEARS vs TRENDS BASED ON ALL 80 YEARS
15
CHANGES IN ENSEMBLE SPREAD DUE TO AVERAGING
PERIOD
HOW SENSITIVE IS THE BCM ENSEMBLE SPREAD TO
AVERAGING PERIOD ?
?t20 yrs
0 20 40
60 80 YEARS
16
CHANGES IN ENSEMBLE SPREAD DUE TO AVERAGING
PERIOD
LONGER SAMPLING OF INTERNAL VARIABILITY
SHOULD INCREASE THE SIGNAL TO NOISE RATIO AND
REDUCE THE SPREAD NON-LINEARITIES IN THE
CLIMATE PROJECTIONS MAY INCREASE THE SPREAD
17
CHANGES IN ENSEMBLE SPREAD DUE TO AVERAGING
PERIOD
THE IMPORTANCE OF ACCOUNTING FOR INTERNAL
VARIABILITY
CHANGE IN BCM ENSEMBLE SPREAD OF ZONAL MEAN
2m TEMPERAURE TREND RELATIVE TO THE SPREAD BASED
ON 20 YEARS () 65-90N
80-90N
MEAN RELATIVE CHANGE IN SPREAD OF ZONAL
MEAN TRENDS
18
CHANGES IN ENSEMBLE SPREAD DUE TO AVERAGING
PERIOD
THE IMPORTANCE OF ACCOUNTING FOR INTERNAL
VARIABILITY
CHANGE IN BCM ENSEMBLE SPREAD OF ZONAL MEAN
PRECIPITATION TREND RELATIVE TO THE SPREAD BASED
ON 20 YEARS () 65-90N
80-90N
MEAN RELATIVE CHANGE IN SPREAD OF ZONAL
MEAN TRENDS
19
SUMMARY

• CLIMATE CHANGE SIMULATIONS STARTED IN DIFFERENT
  PHASES OF THE AMOC (?AMOC 2-3 Sv) EXHIBIT A
  RELATIVELY LARGE SPREAD IN THE ARCTIC RESPONSE
• If the BCM simulations are realistic
• INTERPRETATION OF THE DIVEREGENCE OF MULTIMODEL
  ENSEMBLES FROM A SINGLE SOLUTION SHOULD BE SEEN
  AS A MANIFESTATION OF BOTH
• REAL INTERMODEL DIFFERENCES
• THE FACT THAT THE MODEL SPREAD MAY PARTLY
  REPRESENT THE FREQUENCY DISTRIBUTION OF THE
  CHAOTIC BEHAVIOUR OF THE CLIMATE SYSTEM

20
www.bjerknes.uib.no/conference2004
21
(No Transcript)