1. Department of Geological and Atmospheric Sciences
2nd RASM Workshop
2012 Spring
Brandon J. Fisel
Iowa State University
Department of Geological and Atmospheric Sciences
Monterey, CA May

2. overview Sea ice – atmosphere interactions
recent, rapid sea-ice decline
Model circulation regimes exist
Goal
sea-ice area changes
affects persistent circulation regimes
implications for predictability of future
atmospheric circulation
seasonal T extremes
Interactions between sea ice and arctic atmospheric circulation is of great interest to the scientific community because of sea-ice decline. Model circulation regimes exist. As sea-ice area changes, this affects the persistence of model dynamic circulation regimes which have implications for the predictability of future atmospheric circulation.
2nd RASM May 2012

3. methodology WRF-ARW 3.1 CORDEX domain 50 km resolution Analysis region
Physical parameterizations appropriate for the Arctic provided by Cassano et al., 2011 and M. Seefeldt (post-doc at Colorado), collaborator, who extensively tested model skill parameterizations and groups of parameterizations in the Arctic. Polar modifications include the changes made and accepted by Hines et al., which includes fractional sea-ice concentrations. Coordinated Regional Downscaling Experiment (CORDEX). CORDEX domain is large enough so that the polar vortex is not constrained by the lateral boundary conditions. The red triangle represents the area where sea-ice concentrations were fractional in the fractional ensemble and binary in the binary ensemble. Area outside of the red triangle had binary sea ice in all simulations. During the simulation period, the largest area of fractional sea-ice concentrations occurred in the area of the red triangle allowing the largest differences between binary/fractional simulations.
Analysis region
2nd RASM May 2012

4. methodology (cont…) ECMWF (ERA-Interim); 6-hr 1 Ensemble; 16 members
SST and sea ice; prescribed; 12-hr
1 Ensemble; 16 members
June – December, 2007
1 or 2 regimes
Model-based clustering
model selection; BIC
model fit; penalty for model complexity
Time windows
> synoptic time frame; 7 & 11 days
Atmospheric forcing data made available from ECMWF every 6 hours at T255 resolution. Sea-ice concentrations of 10% or less set to 0% to account for retrieval errors and potential melt ponding (melt ponds, especially in summer have an open-ocean signature and it is difficult to distinguish between ice-covered and open-ocean). Ensemble members were staggered 12 hours apart and were spun-up for 2 weeks so they would be independent of their initial conditions and reach distinctly different atmospheric states by June 15, when we began our analysis. We looked for when 1 or 2 regimes occurred in our target region using model-based clustering, a common technique for indentifying similar components of multivariate data. Model selection provided by the BIC parameter, which provides a summary of model fit and includes a penalty for the number of parameters in the model. The BIC determines when 1 or more regimes is favored for time windows, 7 and 11 days. Circulations regimes that persist longer than synoptic time frame so that a new synoptic state can enter into the region. Any one day could have two regimes and is expected to persist as long as any 500 hPa state.
2nd RASM May 2012

5. methodology (cont…) Seasonal T extremes
JJA and OND; 1st and 99th percentile
regions; ANC, AS1 and SC
land points only
ANC SC
AS1
2nd RASM May 2012

6. regimes 1 regime 2 regimes; significantly separated
Regimes were common and behavior made up of 1/2 separated/2 trends. When some members trend towards higher pressure and others trend towards lower pressure.
2 regimes; differing trends
2nd RASM May 2012

7. regimes (cont…) significantly separated high MSLP low MSLP differing
trends
c) high MSLP
d) low MSLP
Regimes were common and behavior made up of 1/2 separated/2 trends. When some members trend towards higher pressure and others trend towards lower pressure.
2nd RASM May 2012

8. regimes (cont…) Model variability Month 75th – 25th [hPa] June(15-30)
6.1
July
4.9
August
8.1
September
5.7
October
10.1
November
9.0
December
9.8
< 8 hPa
I first wanted to understand when model variability was small in our simulations. Inter-quartile differences (75th - 25th) of daily MSLP. Wider range of variability for OND than JJA and September. Supports the concept that there is a difference in summer and winter circulations and that the summer circulation evolves more slowly than the winter circulation (could be due to weaker temperature differences, which would result in a weaker jet stream). Assuming the distribution is normally distributed, the IQ difference amounts to 50% of the data and isn’t affected by outliers.
> 8 hPa
2nd RASM May 2012

9. regimes (cont…) BIC2 > BIC1 Month 7-Day 11-Day June(15-30) 93% 100%
July
71%
87%
August
84%
97%
September
33%
30%
October
74%
94%
November
63%
83%
December
96%
We wanted to know how common multiple regimes were in our simulations. Percentage of days when BIC prefers 2 regimes to 1 regimes. Point out slight tendency of multiple regimes preferred in JJA than OND, but JJA and OND are similar in that there are multiple regimes common for both periods. Point out that the largest difference occurred in September during the sea-ice minimum.
2nd RASM May 2012

10. regimes (cont…) Regime streaks
JJA: tendency to remain in one mode of regime behavior
Slower evolving atmospheric flow
Season
Mean length [days]
JJA (7-Day)
7.9
JJA (11-Day)
24.0
OND (7-Day)
6.0
OND (11-Day)
15.0
Knowing when model variability is small and that multiple regimes were common in our simulations, we wanted to understand the persistence of multiple regimes in our simulations, so we computed regime streaks (consecutive days) when multiple regimes are favored to 1 regime. The persistence of multiple regimes to be preferred to 1 regime during JJA vs. OND suggests that the model has a tendency to remain in one mode of regime behavior during JJA. This also suggests that the atmospheric flow during summer evolves more slowly.
2nd RASM May 2012

11. regimes (cont….) 2-Regime JJA 2-Regime OND
How persistent do days fall into a 1 or 2-regime window?
2-Regime OND
More persistent days a calendar date falls into a 2-regime window.
The number of days a calendar date falls into a 2-regime (7-day) window binned by season (JJA & OND). A calendar date may fall into a 2-regime window as it moves in time from 0x to 7x, with 7x describing very long periods of persistent 2-regime behavior.
2nd RASM May 2012

12. extremes 2-Regime JJA [1st P] 2-Regime OND [99th P]
Cold extremes preferred in 2-regime behavior.
2-Regime OND [99th P]
Warm extremes preferred in 1-regime behavior.
2nd RASM May 2012

13. summary Multiple dynamic regimes Implications
persistent, multiple regimes (JJA)
less persistent, multiple regimes (OND)
persistent, 1-regime (September)
Implications
changing sea-ice area affects regime persistence
future predictability of atmosphere
future decline in summer sea ice
more persistent regime behavior?
More persistent multiple regimes in JJA than in OND, which saw less persistent multiple regimes. The persistence in JJA may be due to specified SST’s, preventing ocean-atmosphere interactions and constraining the variability of the model’s atmospheric circulation. OND, which has a larger interactive ice surface, may allow for more variability in the model’s atmospheric circulation. September witnessed the largest change in regime behavior, whereby there was more persistent 1-regime behavior, which may be due to a large open-ocean area that may greatly constrain the variability of the model’s atmospheric circulation, allowing for more persistent regime behavior. This work suggests that changing sea-ice area affects the predictability of atmospheric circulation, which has implications on the future predictability of atmospheric circulation. Many studies have suggested that future reductions in summer sea ice may continue, which may affect the persistence of multiple regimes, whereby the atmosphere is more predictable. Uncertainties due to prescribed SST’s that constrain model variability.
2nd RASM May 2012

14. summary (cont…) Temperature extremes Future work
tendency for more cold (warm) extremes with 2(1)-regime behavior
Future work
regimes: coupled model; modeled vs. prescribed SST’s; similar behavior?
T extremes: 20-year simulation (uncoupled WRF, RASM)
change in energy/moisture budgets allow for a more extreme, future arctic?
2nd RASM May 2012

15. questions ?
2nd RASM May 2012

16. regimes Ensemble MSLP Model Estimated Mean 95% CI
Shading is where about 95% of realizations from each regime should fall into, based on model fit.

17. regimes

18. regimes

19. regimes

20. regimes

21. regimes

22. regimes

23. regimes

24. regimes

25. regimes

26. regimes

27. regimes

28. regimes

29. regimes 2-Regime September
The number of days a calendar date falls into a 2-regime (7-day) window binned by season (JJA & OND). A calendar date may fall into a 2-regime window as it moves in time from 0x to 7x, with 7x describing very long periods of persistent 2-regime behavior.
More persistent 1-regime behavior vs. JJA and OND.
2nd RASM May 2012

30. extremes 2-Regime Sept. [1st P] 2-Regime Sept. [99th P]
Cold extremes preferred in 2-regime behavior.
2-Regime Sept. [99th P]
Warm extremes preferred in 1-regime behavior.
The number of days a calendar date falls into a 2-regime (7-day) window binned by season (JJA & OND). A calendar date may fall into a 2-regime window as it moves in time from 0x to 7x, with 7x describing very long periods of persistent 2-regime behavior.
2nd RASM May 2012