Transcript Tropical Cyclones

Tropical cyclones in global climate
models: the role of resolution
P.L. Vidale*, M. Roberts
K. Hodges, ESSC
A. Clayton, M.-E. Demory, J. Donners
with big thanks to:
S. Emori, A. Hasegawa (NIES)
T. Davies and many others at UKMO
L. Bengtsson (ESSC)
L. Shaffrey, I. Stevens, W. Norton, J. Slingo (UK-HiGEM)
*NCAS-Climate, Walker Institute, University of Reading, UK
Evolution of N. Atlantic hurricane frequency in past 100+ years.
2005 was a true record year: 15 hurricanes (incl. Katrina), 27 named storms … and some of the most intense storms in US history. Katrina
damage = 200 bn U$.
In the same region, in the last 2 weeks, Dean (cat 5), Felix (cat 5)
Yet, most GCMs, especially the coarse resolution ones
used for long (e.g. IPCC) integrations,
cannot represent tropical cyclones properly
K. Trenberth
Tropical Cyclones in
decadal (coarse) climate
simulations
Our 3-D TC tracking algorithm,
using 6-hourly, multi-level data.
1.
Compute 850hPa vorticity
– truncate to T42
TCs are warm-core storms
– identify and track, using (weak) vorticity
threshold
2.
Compute multi-level vorticity at T63
3.
Reference tracks back onto T63 grid, at all
vertical levels
4.
Further identify/filter TCs using:
– intensity at 850hPa @ T63
While warm core storm,
vorticity decreases with height
– lifetime > 2 days
– vertical gradient of vorticity
– TC centre must be present at all levels
5.
Reference full-resolution winds, precip
onto tracks
6.
Build storm composite climatologies
7.
Finally, find Extra-tropical Transition by
core reversal (to cold core)
Are the modelled TCs really warm core storms ?
Is the near-core circulation correct; do we see changes with resolution ?
Composite of 100 most intense TCs in 25 yrs
Had-Hi-NUGAM
135km
resolution
60km
MIROC
T106
resolution
Hurricane Mitch
T213
Once we can trust that we are targeting the right phenomena, we want to ask:
can we realistically compute where they originate, where they reach maximum intensity
and where they eventually die ? Do models agree on location ?
Statistics of Tropical Cyclones in AMIP2 integrations: role of resolution
Once we can trust that we are targeting the right phenomena, we want to ask:
can we realistically compute where they originate, where they reach maximum intensity
and where they eventually die ? Do models agree on location ?
Statistics of Tropical Cyclones in AMIP2 integrations: role of resolution
Once we can trust that we are targeting the right phenomena, we want to ask:
can we realistically compute where they originate, where they reach maximum intensity
and where they eventually die ? Do models agree on location ?
Statistics of Tropical Cyclones in AMIP2 integrations: role of resolution
But model formulation matters !
Statistics of Tropical Cyclones in AMIP2 integrations
Impact of resolution on TC Intensity
MIROC
Had-Hi-NUGAM
135km
T106
I
90km
T213
60km
135km
T106
90km
60km
135km
ECHAM5
90km
60km
P
So, despite inter-model differences,
T213 our
main result is quite robust:
a larger number of the more intense
Tropical Cyclones are simulated as the
model resolution
is increased.
T106
W
T213
L. Bengtsson, K. Hodges, ESSC, Reading
Tropical cyclones intensity in a warmer world:
NUGAM and ECHAM5, 20th century and 21s century
135km
ECHAM_T213
ECHAM_T319
90km
60km 60km+4K
135km
90km
60km 60km+4K
← Changes in precipitation
135km
90km
60km 60km+4K
← Changes in wind speed
Key questions for our climate models:
example from high-impact weather in climate models
• General aim in the community: regionalisation of impacts
and their prediction under climate change
• With a 3-model intercomparison, no general consensus on
Tropical Cyclones, but some issues for studying small
scale processes in climate:
–
–
–
–
do any of our GCMs converge ?
how much resolution is enough for treating each problem ?
is there any verification data, e.g. from satellites ?
can we learn from the process-resolving models ?
• With this knowledge, to which degree can we use partially
resolved phenomena as proxies for what happens in
nature and use top supercomputers to study, e.g. for
tropical cyclones:
– globally, in decadal and centennial simulations
– with large ensemble simulations (significant sample size) →
meaningfully study the local impact of extreme events, e.g.
probability of landfall in Miami or Tokyo ?
– in transient and stabilised climate change simulations ?
Hurricane Felix, cat. 5, 2 Sept 2007
Next: a multi-scale GCM TC intercomparison is needed;
an opportunity to derive proxies for studying TCs in low-resolution GCMs
Model
resolution /
type
1-5 km
Case studies
10-20 km
Climate runs
Case studies
40-60 km
Climate runs
Ensemble runs
100-300 km
IPCC-type
climate runs
Participants
MSSG-A (MSRG)
NICAM
(CCSR/FCRGC)
AFES/CFES (AOSG)
MRI
IFS T799
(ECMWF)
MIROC T213
NUGAM
ECHAM5 T319
IFS T319
(ECMWF)
MIROC T63
HadGAM
ECHAM5 T63
IFS T159
(ECMWF)
UK-CASCADE
Benefit to
other models
Realistic processes:
resolved convection
Resolved flow
Partially resolved
storms
Sample size
Variability
Multiple
scenarios
Observations
TRMM, CALYPSO
Obs. case studies
Best Track
ECMWF analysis
ERA-interim
ERA40/JRA25
IPCC
Impacts
WillisRe
Wind gusts, radius of
max wind, storm
surge, forward TC
speed, max. precip.
Winds, intensity,
Precipitation,
fwd. TC speed, storm landfall stats.
surge
Exchange of 6-hourly: U,V,W,T, at multiple vertical levels; sfc. Prec.; MSLP
To upload data to UJCC server in Yokohama: [email protected]
Where next ?
Analysis of coupled
simulations: are the
storms weaker in the
coupled runs ?
Stratification into
Niño/Niña years
Regional analyses
Ocean-Atmosphere
interactions
Cold wake from Katrina and Rita in Gulf of Mexico
SST in Gulf
NASA
Summary and future work
In 2007
•
•
•
•
•
Coupled ultra-high resolution model, to study
interactions (e.g. tropical cyclones on ocean
mixing, ENSO interactions) and extremes;
Used AMIP2 simulations to perform 3-model
intercomparison of TC characteristics and
their resolution dependence;
Resolution seems to mostly affect storm
intensity;
Model formulation seems highly relevant,
especially for geographical distribution;
There is value in resolution, but … it is still
unclear where the models at various
resolution converge on simulating key
processes: value of model
intercomparison;
In 2008
UJCC-HiGEM:
• Tropical Cyclones intercomparison including
CRMs;
• Explore role of high-resolution SSTs, used to
force NUGAM model;
• Extend coupled climate integrations at ultrahigh resolution;
Collaborative:
• CCSR/FRCGC/NIES/ESC/Univ. of Tokyo:
tropical cyclones
• WillisRe: impacts of tropical cyclones
• Analysis of climate change simulations (with
Reading North-Atlantic group);
• Weather and climate variability, with HiGEM
and ESSC;
• Extremes, with HC, Oxford, Edinburgh