Transcript Challenge and directions for improving GCM simulations of the

Challenge and directions for
improving GCM simulations of
the monsoon
Julia Slingo and Andrew Turner
Challenge 1: Multi-scale Processes
Asian and Australian Monsoons are dominated by the effects of
convection organised on a wide range of space and time scales
(diurnal cycle, tropical cyclones, monsoon depressions, MJO,
BSISO, convectively coupled equatorial waves…)
Increasing evidence of multi-scale interactions involving:
 Coupling between dynamics and physics on wide range of
scales within components of the climate system
 Coupling on wide range of scales between components of
the climate system
Increasing evidence that multi-scale interactions affect:
 Mean state of the climate system
 Low frequency variability of the climate system
Scale interactions are fundamental to the
tropical climate system
From THORPEX/WCRP Workshop on Organised Convection and the MJO
Challenge 2: Air-sea interaction and
coupling with the ocean
• Increasing evidence that many aspects of monsoon
variability involve air-sea interaction and coupled processes:
Implies that atmosphere-only models may not be appropriate
for monsoon studies.
• Indian Ocean may play a much more significant role than
previously thought: Implies the need for more detailed
evaluation of Indian Ocean in coupled models.
• Diurnal cycle in the ocean mixed layer may be important for
the mean state and for intraseasonal variability: Implies that
higher vertical resolution in the upper ocean may be needed.
High-frequency, observed SST forcing and
the intraseasonal oscillation
 Objective
To determine the influence of high
frequency SSTs on intraseasonal
monsoon variability.
 SST forcing dataset
Feb. 2005–2006 reanalysis
from the Met Office GHRSST
project.
Assimilates satellites (e.g.,
TRMM) and in situ buoys.
Available as daily analyses
at 1/20° spatial resolution.
Substantial intraseasonal
(30-70 day) variability
during the monsoon.
Standard deviation of 30-70 day
SSTs for June – September.
Line contours give percentage of
variability explained.
High-frequency, observed SST forcing and
the intraseasonal oscillation
HadAM3 ensembles

“Daily” ensemble forced by daily GHRSST
SST product.

“Monthly” ensemble forced by monthly mean
GHRSST (following AMIP II method)

N144 (1.125°x0.875°) and 30 vertical levels,
beginning 1 Feb.

30 ensemble members

Difference between the ensembles shows the
influence of sub-monthly SSTs.
Seasonal-mean rainfall

Sub-monthly SST variability projects onto the
ensemble-mean, seasonal-mean rainfall.

Differences are small but statistically
significant.
Difference in ensemble-mean,
JJAS-mean rainfall, taken as
the Daily ensemble mean minus
the Monthly ensemble mean.
High-frequency, observed SST forcing and
the intraseasonal oscillation
Intraseasonal variability

Significant increase in 30-70
day variability in Bay of
Bengal and Arabian Sea.

Spatial pattern of increases is
broadly consistent with
regions of high 30-70 day
variability in GHRSST SSTs.

No coherent northwardpropagating signal from the
equatorial Ocean to the
Indian peninsula – lack of
coupling?
Difference in ensemble-mean
standard deviation in 30-70 day
filtered JJAS rainfall.
High-frequency, observed SST forcing and
the intraseasonal oscillation
Intraseasonal variability


Daily ensemble contains
much stronger power at
intraseasonal (30-50
day) periods.
Sub-monthly SST
variability can increase
the variability of rainfall
at much longer
timescales.
Ensemble-mean 1D wavelet transforms
of Bay of Bengal rainfall
Daily
Ensemble
Monthly
Ensemble
Spatial variability of intraseasonal modes
30-60day 10-20day
Percentage variance explained by each band of total intraseasonal
variance of U850 wind anomalies:
ERA-40
HadCM3
HadCM3FA
The spatial pattern of explained variance is better simulated in
HadCMFA, especially in the 30-60 day band.
Temporal variability of intraseasonal modes
Lag-regressions of U850 against reference timeseries (85-90E, 5-10N),
showing westward (10-20) and northward (30-60) propagation
HadCM3
HadCM3FA
30-60day
10-20day
ERA-40
Northward propagating modes on 30-60 day timescales show no
improvement in HadCM3FA.
Mixed layer depth anomaly active and
break composites
HadCM3FA
Break
Active
HadCM3
Mixed layer model studies of the diurnal
cycle: Sensitivity to vertical resolution




Bernie et al. 2005
1m resolution (CTR)
gives good simulation of
diurnal and
intraseasonal variability
10m resolution of most
ocean models will not
resolve diurnal
variability of SST
Intraseasonal variability
is ~0.4°C less than CTR
Implies 40%
underestimate of the
strength of air-sea
coupling
Diurnal Coupling with the Ocean:
Impact on the annual mean climate
HadAM3 coupled to
OPA with high vertical
ocean resolution – 1
meter in near surface
layer:
HDC: Hourly coupling
HDM: Daily coupling
Dan Bernie, Eric Guilyardi, Gurvan Madec, Steve Woolnough & Julia Slingo
Amplitude of SST diurnal cycle in HadOPA (L300)
DJF
JJA
MAM
SON
Dan Bernie, Eric Guilyardi, Gurvan Madec, Steve Woolnough & Julia Slingo
Note large
seasonality in
the amplitude of
the diurnal cycle
for the northern
Indian Ocean.
Is this a crucial
component of
the premonsoon high
SSTs?
Challenge 3: Influence of basic state
errors on monsoon variability
A very interesting talk
The effect of heat flux adjustments
The effect of heat flux adjustments
More in session 4….
Challenge 4: Sensitivity to resolution
Recent developments in UK Climate Models
Atmosphere
Ocean
Flux
HadCM2
HadCM3
HadGEM1
HiGEM
NUGAM
1994
1998
2004
2005
2006
~300km
~300km
~150km
~90km
~60km
19 levels
19 levels
38 levels
38 levels
38 levels
2.50 x 3.750
1.250 x 1.250
10 x 10 (1/30)
1/30 x 1/30
(1/30 x 1/30)
20 levels
20 levels
40 levels
40 levels
(40 levels)
Yes
No
No
No
1
4
40
400
(No)
Adjustment?
Computing
Earth
Simulator
JJA precipitation minus CMAP
HiGEM
HiGAM
HadGEM
HadGAM
Tropical Precipitation
Errors
JJA 2004
© Crown copyright 2006
Dry
Wet
Page 20
Probability density function of central
relative vorticity for tropical cyclones
Distribution shifted to
higher intensities as
resolution is increased
135 km model
90 km model
60 km model
Observed hurricanes/typhoons
seen to have vorticities (spin)
between 10-70 x10-5 s-1
x10-5
Atmosphere Only
• Atmosphere-only model fails to simulate MJO
• HiGEM is a significant improvement on HadCM3 (and HadGEM1)