![Page 1: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/1.jpg)
GFDL’s high-‐resolu1on seasonal climate predic1on system: modeling, ini1aliza1on and predictability
Xiaosong Yang NOAA/GFDL, Princeton, NJ
Using JPSS Products for Earth System Data Assimila:on, JPSS-‐CPO
Technical Interchange Mee:ng, Jan 30, 2017 Acknowledgements: Gabriel A. Vecchi, T. Delworth, S. Kapnick,, H. Murakami, L. Jia, R. Gudgel, S.D. Underwood, L. Krishnamurthy, A. Rosa1, S.-‐J. Lin, A.T.
WiRenberg, W. Anderson, V. Balaji
![Page 2: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/2.jpg)
GFDL FLOR/HiFLOR: High-‐resolu1on coupled seasonal predic1on system for regional climate and extremes (Vecchi et
al. 2014, Murakami et al. 2015 )
Delworth et al. (2012), Vecchi et al. (2014)
Medium resolu1on (CM2.1)
High resolu1on (CM2.5-‐FLOR)
Precipita1on in Northeast USA
Modified version of CM2.5 (Delworth et al. 2012): • 50km cubed-‐sphere atmosphere • 1° ocean/sea ice (low res enables predic1on work) Contributed to NMME from March 2014
![Page 3: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/3.jpg)
Computa1on cost
CM2.1/LOAR FLOR HiFLOR Atmospheric Grid Cell Size
200 km (~160 miles)
50 km (~30 miles)
25 km (~16 miles)
Compu:ng Cost 1 20 120
Size of 5 Year of Daily Data File (ex: Total Precipita:on)
91 MB 1.5 GB 7.1 GB
Model eleva1on
![Page 4: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/4.jpg)
Tropical cyclones (TCs) and extratropical storms (ETSs) cause weather/climate extremes for natural disasters (winds, flooding, blizzards)
• Hypothesis: Enhanced resolu1on improves simula1on (fidelity)
and predic1on of regional climate & extremes (TCs and ETSs). • Tropical cyclones (TCs)
• GFDL-‐CM2.1 (~200x250km atmosphere/land, 1° ocean/sea ice): No-‐TC • GFDL-‐FLOR (50km atmosphere/land): TC-‐permifng • GFDL-‐HiFLOR (25km atmosphere/land): TC-‐category-‐resolving
• Extratropical storms (ETSs) • All the three models resolve ETSs • Can we s1ll improve ETSs with high-‐res models?
![Page 5: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/5.jpg)
HiFLOR: doubling atmospheric resolu1on of FLOR (cost 6x) allows model to simulate Cat. 4-‐5 TCs (most destruc1ve storms)
Murakami et al. (2015)
Fidelity of TCs
![Page 6: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/6.jpg)
Model fidelity of ETS: Can we do it beYer using high-‐res?
FLOR/HiFLOR improves climatological ETS over North America, North Atlan1c, Eurasia and Southern Ocean.
Contour: ETS climatology; Shaded: Model bias
DJF ETS: Var(filtered 6-‐hourly SLP)
Fidelity of ETS
![Page 7: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/7.jpg)
GFDL-‐FLOR ini1aliza1on: Phase 1, NMME
• Ocean & sea ice ini1alized from CM2.1 V3.1 EnKF Assimila1on • Atmosphere and land ini1alized from ensembles of AGCM driven by
observed SST (i.e., only informa1on contained in SST and radia1ve forcing in atmos/land ICs)
• Uncoupled ini1aliza1on: high-‐res ensemble coupled data assimila1on computa1onally expensive.
• References: Jia et al. (2014, J. Clim.), Msadek et al. (2014, J. Clim.), Vecchi et al. (2014, J. Clim.), Yang et al. (2015, J. Clim.), Murakami et al. (2015, J. Clim.)
![Page 8: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/8.jpg)
Regional ENSO rainfall paRern improved in FLOR (skill up too)
obs
a
CM2.5_FLOR
c
CM2.1
e
obs
b
CM2.5_FLOR
d
CM2.1
f
−0.9 −0.7 −0.5 −0.3 −0.1 0.1 0.3 0.5 0.7 0.9
Figure 2: Observed pattern of precipitation response (in mmday−1 per unit variate) to ENSO
in low latitudes Americas and tropical Asia(a, b); Spatial structure of the most predictable
component of land precipitation (in mmday−1 per unit variate) from CM2.5 FLOR(c, d) and
CM2.1(e, f).
11
(Jia et al. 2015, J. Clim.) Most predictable precip paRern (mm/day)
![Page 9: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/9.jpg)
ETS predictability: Leading Predictable paRern of storm tracks is ENSO-‐related, and is predictable up to 9 month lead 1me.
333
333
6666666
6 6 6 6
6
6
99
9 9999
999 999
9 12
60oE 120oE 180oW 120oW 60oW 0o 80oS
40oS
0o
40oN
80oN a)
0.70.50.30.1
0.10.30.50.7
1990 2000 2010
2
0
2
Year
Tim
e Se
ries
b)
Nino3.4 Obs
Lead 1 5 Lead 6 10
Nov Oct Jul May Mar0
0.2
0.4
0.6
0.8
1
Initial month
c)
Leading Predictable Component (DJF)
FLOR
• Storm reduc1on over
most North America
• Time series highly correlated with NINO3.4
• Skill is comparable with predic1ng ENSO
Yang et al. (2015, J. Clim.)
ACC
Storm track (DJF) : std(24-‐hour difference filtered SLP)
Prediction of ETS
![Page 10: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/10.jpg)
1-‐July Ini1alized Retrospec1ve NA TC forecasts with HiFLOR (25km FV3 atmosphere coupled to 1° MOM5)
! !!"#$%&%'"()*+),")+%-.%'/%.,)01)2345%61,,"372)+5%728%/9:%61,,"372)+%"2%;6)%<-,;6%=;>72;"3$
Prediction of TCs Murakami et al. (2016)
![Page 11: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/11.jpg)
GFDL-‐FLOR ini1aliza1on: Phase 2, research mode
• Ocean & sea ice ini1alized from CM2.1 V3.1 EnKF Assimila1on • Atmosphere and land ini1alized from a coupled FLOR simula1ons
with atmosphere state relaxed to MERRA (CFSR) reanalysis , SST restored towards observed SST
• Predictability source arising from observed ocean/sea ice and atmosphere/land in P2
• References: Jia et al. (2016, J. Clim.), Jia et al. (2017, J. Clim.), Yang et al. (2017, J. Clim.)
![Page 12: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/12.jpg)
Seasonal (MAM) Temperature Predic1on The Role of the Stratosphere
SST+Strat
d
SSTonly
e
180oW 120oW 60oW 0o 60oE 120oE 180oW diff
f
15oN 30oN 45oN 60oN 75oN 90oN
ObsAIC
a
15oN 30oN 45oN 60oN 75oN 90oN
SST−forcedAIC
b
180oW 120oW 60oW 0o 60oE 120oE 180oW 15oN 30oN 45oN 60oN 75oN 90oN
diff
c
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Source: Jia et al. 2017, SubmiGed
P2
P1
Prescribed stratosphere + SST
SSTonly
![Page 13: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/13.jpg)
Flipped precipita1on paRern over the Western US (WUS) during 2015/16 winter VS the canonical El Niño paRern
2015/16 1997/98
Canonical
• The canonical El Niño paRern: wet (dry) over the southern (northern) half over the WUS.
• 2015/16: almost opposite to canonical paRern
• 1997/98: similar to the canonical paRern. mm/day
![Page 14: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/14.jpg)
NMME models (including GFDL-‐FLOR) predicted a canonical El Niño precipita1on paRern over the WUS.
I.C.: 1Nov2015 hRp://www.cpc.ncep.noaa.gov/products/NMME/
![Page 15: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/15.jpg)
FLOR P1 and P2 hindcasts: 2015/16
and 1997/98 2015/16: Flipped paRern from P1 to P2 towards the observa1on 1997/98: P1 and P2 have similar skill.
PaRern Correla1on Coef.
2015/16 1997/98
Obs.
P1
P2
Obs.
P1
P2 Yang et al. 2017, J. Climate, submiRed
![Page 16: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/16.jpg)
Impact of land and atmospheric states on winter precipita1on skill during 1982-‐2016.
The ra:o of the grid points with posi:ve significant correla:on to the total grid points: P1: 17% P2: 55%
P1
P2
![Page 17: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/17.jpg)
Conclusions and discussions
• High-‐res model substan1ally improves the regional climate predic1on skill.
• Skillful seasonal predic1on of extratropical storms, tropical cyclones as well as Category45 hurricanes
• Stratosphere plays important roles in the extratropical surface seasonal climate predic1on, so accurately modeling and observing stratosphere could improve seasonal climate predic1on.
• The appropriate ini1aliza1on of atmosphere/land in addi1on to the ocean could substan1ally improve the seasonal precipita1on predic1on.
• Seasonal predic1on for regional climate is a very challenge problem (modeling and ini1aliza1on).
• BeRer skill is expected if the predic1on model is ini1alized using coupled DA.
![Page 18: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/18.jpg)
Model fidelity of ETS: mean and varia:ons of eddy kine:c energy (EKE) simulated by high-‐res coupled models
FLOR (1990-‐control) improves climatological SH EKE as well as spa1al-‐temporal varia1ons.
m2 /s2
SH baroclinic annular mode (BAM) (Thompson and Barnes, 2014): • Leading EOF mode of daily EKE • Periodicity of 20-‐30 days
Leading EOF daily EKE
BAM spectrum
250-‐mb
Mean EKE at 250 mb
500-‐mb
850-‐mb Subseasonal signal
Fidelity of ETS
![Page 19: GFDL’s’high+resolu1on’seasonal’climate ... - STAR€¦ · ENSO+related,’and’is’predictable’up’to’9’month’lead’1me. 3 3 3 3 6 6 6 6 6 6 6 6 6 6 6 6 9 9 9](https://reader036.vdocuments.net/reader036/viewer/2022071017/5fd0fede446b7f2cf64773e9/html5/thumbnails/19.jpg)
Predicted circula1on anomalies for 2015/16 and 1997/98 from P1 and P2
P2 predicted the observed local blocking high for 2015/16 winter
Obs
P1
P2
2015/16 1997/98 700-‐hPa Height: contour Winds: vector Moisture: shading