Transcript Lysbilde 1 - Universitetet i Bergen

Arctic climate change –
structure and mechanisms
Nils Gunnar Kvamstø,
Input from: Øyvind Byrkjedal, Igor Ezau, Asgeir
Sorteberg, Ivar Seierstad and David Stephenson
Arctic zonal temperature anomalies
(within 60º-90ºN latitudinal zone)
• Winter, summer, and annual anomalies, 1881-2003 period
• All linear trends significant at the 0.01 level
• (available from CDIAC, Lugina et al. 2003, updated).
2
Courtesy P.Groisman
Northern Hemisphere temperature
anomalies
• Winter, summer, and annual anomalies, 1881-2003 period
• All linear trends significant at the 0.01 level
• (available from CDIAC, Lugina et al. 2003, updated).
3
Courtesy P.Groisman
Arctic vs. Global Change
DJF Zonal mean Ts anomalies
Johannessen et al. 2003
4
∆Ts
DJF
MAM
JJA
SON
5
Vertical structure
Hartman (1994)
Seasonal cycle of Arctic
temperature profiles
Inversion
Hartman (1994)
Vertical structure of recent Arctic warming
DJF
JJA
Graversen et al 2008, Nature
MAM
SON
8
Cross-section cold air outbreak, arctic front,
Shapiro & Fedor 1989
Isentropes
height
Sea
Ice
9
Vertical structure
Hartmann and Wendler J. Clim (2003)
10
SAT is heavily sensitive to the relative strengths
of surface inversions
Change in mean winter temperature from 1957-58 to 2003-04 for decoupled (left)
and coupled (right) PBL cases. After Hartmann and Wendler (2003).
11
POLAR AMPLIFICATION
• GHG forcing considered to be quite
uniform, why polar amplification?
• Ice-albedo feedback
• Cloud feedback
• ”Dynamic feedback”
12
Fixed albedo experiment –> Albedo feedback
Hall (2004)
13
Fixed cloud experiment -> Cloud feedback
Vavrus (2004)
14
Ghost forcing -> Dynamical feedback
Alexeev, Langen, Bates (2005)
15
Ghost forcing -> Dynamical feedback
Alexeev, Langen, Bates (2005)
16
___ ENSEMBLE MEAN
SRES A1B (CO2 ENDS AT 700 ppm)
8
10
2 Projected changes
6
4
2
0
ºC
4-10ºC
1920
1940
1960
1980
2000
2020
2040
2060
2080
CHANGES IN ARCTIC TEMPERATURES
FROM 15 CLIMATE MODELS
Sorteberg and Kvamstø (2006)
17
Why is the spread so large?
• Insufficient formulation of processes in
GCMs?
• Internal atmospheric variability?
• Differences in external forcing (GHG,
aerosols)?
18
LARGE DIFFERENCES IN PROJECTED CLIMATE
CHANGE EVEN WHEN SAME FORCING IS USED:
19 CMIP2 MODELS : ZONAL TRENDS IN T2m
YEAR 31-60 (ºC/DECADE)
Is this spread entirely due to different models?
Sorteberg and Kvamstø (2006)
19
BCM SPREAD vs MULTIMODEL SPREAD
ANNUAL 5 MEMBER ENSEMBLE MEAN T2m CHANGE
YEAR 1-30 (C)
20
Sorteberg and Kvamstø (2006)
BCM ENSEMBLE SPREAD IN ANNUAL T2m ZONAL
MEAN TEMPERATURE CHANGE RELATIVE TO
MULTIMODEL SPREAD (%)
YEAR 1-30
60%
40%
20%
21
Sorteberg and Kvamstø (2006)
Role of internal variability w.r.t. multi model spread
Temperature
Precipitation
22
Sorteberg and Kvamstø (2006)
Ensemble mean change
Year 61-80
<∆T>
<∆P>
23
Sorteberg and Kvamstø (2006)
Ensemble spread
Year 61-80
σ∆T
σ∆P
24
Sorteberg and Kvamstø (2006)
Signal to noise ratio
Year 61-80
S/N; T
S/N; P
25
Sorteberg and Kvamstø (2006)
Spreads dependence on ensemble size
95% confidence in annual means:
<ΔT>±0.2K
<ΔP>±0.1mm/day
What contributes to the large Arctic T variability?
26
Sorteberg and Kvamstø (2006)
CHANGE IN ICELANDIC LOW AT 2CO2
DJF
DJF:
ARCTIC TEMP CHANGE
27
THE ICELANDIC LOW: A MAJOR PLAYER ATMOSPHERIC
HEAT TRANSPORT INTO THE ARCTIC
28
Surface air temperature change
(AR4)
DJF (1954 – 2003)
A2
B2
29
Kattsov, Walsh
Can we trust projected changes?
(even with large ensemble sizes)
•
•
•
•
Generally too cold troposphere
Too warm SAT
Underestimation of precipitation
Systematic biases in surface pressure
distribution (Beaufort high)
• Model problems connected to poles
(Randall et al. BAMS, 1999)
30
T2m is a heavily used climate parameter.
How is the ABL represented in GCMs?
HIRLAM and ARPEGE comparison with Sodankylä Data http://netfam.fmi.fi/
• Models are missing cold events – model SAT is too warm
• Climate variability, diurnal cycle and blocking events are underpredicted
31
Mixing profiles in NERSC LES (dashed) and ARPEGE –
large discrepancy in shallow Arctic PBLs
32
A model resolution problem
An analysis of observations and LES data shows
that the standard closure type in todays GCMs e.g.
 
 u w
  u 
 k  
z
z  z 
are not applicable on vertical resolutions > 10-50m
H: If implemented correctly it should work well
33
Test of H
hPa
90L:
• 90 vertical layers
• 70 layers increased resolution
from 600hPa and below
•10m resolution in the lowest 60 m
31L:
• 31 vercikal layers (standard)
• lowest layer at ca 70m
Far too costly –
Alt: use analytical functions
34
Simulated vertical temperature profile vs observed data (SHEBA)
90L
31L
Obs
35
Response in Surface Temperature by season (90L-31L)
djf
mam
Moderate improvement.
Local processes important, but
large-scale dynamics is playing
a significant role as well!
jja
son
36
Data analysis
Daily SLP anomalies
in Bergen
Highpass filtered SLP
variance (2-10d)

     const
 
2
Monthly storminess Y:
GLM:
Y  ~ Gamma(, )
where
5
p
i 1
j 1
ln( )     0   i xi    j y j  t
Predictors:
Seierstad et al (2007)
Seasonality
Local SLP +
9 leading PCs
37
Can teleconnection patterns provide additional
explanation for variations in storminess?
ΔY (%) due
to 1σ change
in predictors
Yes! But, restricted to
local, mostly high
latitude areas.
Seierstad et al (2007)
38
Given limited resources, modellers
have to make priorities
39
Response Surface flux for DJF (90L-31L)
sensible
latent
40
Complexity of the Arctic Climate System
•Winds (days – weeks)
•Ocean Currents (years to
decades)
•Rivers (years to
decades)
•Terrestrial cryosphere
(centuries and longer)
This is a highly nonlinear coupled system
Macdonald et al., 2003
Thank you for your attention!
42