Rapid Sea Ice Loss in Climate Model Simulations of a Changing Arctic

Transcription

1 Rapid Sea Ice Loss in Climate Model Simulations of a Changing Arctic Marika M Holland National Center for Atmospheric Research NCAR is sponsored by the Na3onal Science Founda3on

2 September Ice Extent 1 September Arctic Sea Ice Extent km 4 NSIDC Ice Index (Obs) CCSM3 Results Range across 8 Ensemble Members Year (Holland et al., 6)

3 September Ice Extent 1 September Arctic Sea Ice Extent km 4 NSIDC Ice Index (Obs) CCSM3 Results Range across 8 Ensemble Members Year (Holland et al., 6)

4 Changes associated with Rapid Summer Ice Loss (Holland et al., 6) Ocean Heat Transport to Arctic Cloud Cover (Vavrus et al., 1) 6 Terrestrial warming (Lawrence et al., 8)

5 Ice thickness shows large drop associated w/ event March Ice Thickness This change is similar to earlier reductions in th century that had little ice extent change. Increased efficiency of OW production for a given ice melt rate As ice thins, vertical melting more efficiently produces open water Relationship with ice thickness is non-linear

6 Sea ice loss is modified by climate feedbacks Fundamental sea ice thermodynamics gives rise to a number of important feedbacks Balance of fluxes at surface F sw!f sw F LW F SH F LH (1"#)F SW + F LW "$T 4 + F SH + F LH +k %T dh = "q %z dt h s -k s dt/dz h i T 1 T T 3 T 4 -k dt/dz Vertical heat transfer (conduction, SW absorption) F ocn Surface albedo changes modify SW absorption in ice and ocean heat flux Ice loss lowers albedo positive feedback Balance of fluxes at ice base F ocn " k #T #z = "q dh dt

7 Ice mass budgets affected by climate feedbacks Fundamental sea ice thermodynamics gives rise to a number of important feedbacks Grow/Melt h s h i. 1.5 F sw 1.!F sw F LW T 4 Melt F SH -k s dt/dz -k dt/dz F LH.5 T (Following 1 T. Bitz and Roe, 4) T Annual Thickness Balance of fluxes at surface (1"#)F SW + F LW "$T 4 + F SH + F LH +k %T dh = "q %z dt Vertical heat transfer (conduction, SW absorption) F ocn Heat conduction related to vertical temperature gradient Causes ice growth to vary as 1/h Has a stabilizing effect on ice thickness since thin ice grows more rapidly Balance of fluxes at ice base F ocn " k #T #z = "q dh dt

8 Sea Ice Model Schematic Ice Mass Budget Ice volume change Thermodynamic source Divergence

9 1 Mass Budget Change Control Melt Divergence Sea ice mass budget m 5 Thickness Year 4 Ice volume 3 change 1 Thermodynamic source Divergence grow/melt+div Melt+Divergence Thickness Year From Slab Thermodynamic Considerations Grow/Melt Melt Annual Thickness

10 Does this Rapid Ice Loss Result From a Tipping Point? If we initialize simulations after a possible threshold, will ice retreat to summer ice-free conditions even with no continued forcing changes 7 Experiments CO Concentra3ons CO b3.3b.es1 b3.4b.es1 b3.4b b3.4b.es1bcom3 b3.4b.es1bcom4 b3.4b.es1bcom5 A1B Scenario C Obs Year (Holland et al., in prep)

11 With no continued increases in CO, Arctic does not transition to summer ice-free conditions September Ice Extent CO remains at values CO remains at 3 values CO continues to increase 1 Model results suggest Rapid ice loss events in CCSM3 do not result from threshold behavior Stabilizing feedbacks allow a perennial ice pack to remain

12 m m Year Mass Budget Change Commit -1 Melt - Divergence Year Mass Budget Change Control Melt Divergence Year Sea ice mass budgets grow/melt+div grow/melt+div grow/melt+div Melt+Divergence 5. Stabilizing effects of growth and divergence Thickness Thickness

13 .6.4. Heat Flux Change Annual Flwout_aice Net Flux Net Heat Flux Annual Annual Flat_aice Flwdn Incoming Longwave Annual Annual Fswabs_aice Fhnet_aice Ice-Ocean Heat Flux in 5.5 commitment runs Annual Fswabs_aice Reduced net heat flux to sea ice Results largely from: reduced incoming LW radiation Direct influence of CO Annual Fsens_aice.forcing Possible cloud feedbacks -.5 Reduced ice-ocean heat exchange associated with surface albedo feedback Annual Net Flux Control: CO increases CO remains at values CO remains at 3 values Annual Fsens_aice

14 Stability of Seasonally Ice-Free State Performed highly idealized experiments Runs initialized with 1 CCSM3 conditions (seasonally ice free Arctic state) Reductions in CO concentrations applied 7 Experiments CO Concentrations CO b3.3b.es1 b3.4b.es1 b3.4b b3.4b.es1bcom3 b3.4b.es1bcom4 b3.4b.es1bcom5 C Obs A1B Scenario 19 1 Year

15 September Extent March Thickness iextent (1 6 km ) 6 4 (as func3on of 3me) hi (1 13 m 3 ) 3 1 (as func3on of 3me) Time Time iextent (1 6 km ) 6 4 * (as func3on of CO) hi (1 13 m 3 ) 3 1 * (as func3on of CO) CO CO

16 Ann NH extent Ann Ice Extent th Century 1st Century -.4 million km /K Ann Global TREFHT b.es1bcom4 b3.3b.es1 b3.4b.es1bcom5 b3.4b.es1bcom3 Sept Ice Extent Relationship to global surface air temperature Linear relationship of annual NH ice extent and global SAT Relationship with September extent and ice thickness not linear Relationship nearly identical for the ice loss and ice recovery simulations 4 Annual Ice Thickness Sept NH extent 6 4 Ann Arctic hi Ann Global TREFHT Ann Global TREFHT

17 9 8 7 SAT (8-99 minus ) Arctic amplification Latitude Ann Global TREFHT minus Present Day 3 J F M A M J J A S O N D Ann Polar TREFHT o C Arctic amplification has similar characteristics for the loss and recovery of perennial sea ice In CCSM3, the Arctic surface air temperature change is about 3Xs that of the global change

18 Conclusions Climate simulations exhibit rapid ice loss events (RILEs) in the summer Arctic sea ice with associated changes throughout the Arctic system Simulations suggest that these RILEs Are not associated with threshold behavior But instead result from large natural variability superimposed on considerable GHG forced change Simulations initialized from a summer ice free (1) state with applied reductions in CO Recover an Arctic perennial ice pack with some time lag Relationship of sea ice conditions (state, mass budget) to global air temperature is nearly identical for the loss and recovery of perennial sea ice Suggests that loss of perennial sea ice is reversible NOTE: These are highly idealized runs that would require technology to scrub CO from the atmosphere

19 Questions?

20 Extra Slides

21 Sea Ice Model - Dynamics Sea Ice Model - Dynamics! Force balance between wind stress, water! Ice treated as a continuum with an effective stress,rheology internaldescribing ice stress, and stress large-scale thecoriolis relationship between stress and flow! associated with sea surface slope! Force balance between wind stress, water stress, Ice treated as a continuum with an effective internal ice stress, coriolis and stress associated large-scale rheology describing the with sea surface slope! relationship stress and! Ice freely divergesbetween (no tensile strength)! Ice freely diverges tensile! Ice resists convergence and(no shear! flow strength)! Multiple ice categories advectedand withshear same velocity Ice resists convergence field! "u m = #mfk $ u + % a + % o # mg&h + & ' "t Coriolis Coriolis Air Air stress stress Ocean Ocean stress stress Sea Sea Slope Slope Internal Internal Ice Ice Stress Stress (e.g. Hibler, 1979)

22 Thermodynamics Vertical heat transfer Assume e brine pockets are in thermal equilibrium with ice Heat capacity and conductivity are functions of T/S of ice Assume constant salinity profile Assume non-varying density Assume pockets/channels are brine filled brine filled "c #T #t = # #z k #T #z + Q SW! Q SW = " d dz I SW e"#z where I SW = i (1"#)F SW (from Light, Maykut, Grenfell, 3) (Maykut and Untersteiner, 1971; Bitz and Lipscomb, 1999; others)

23 What allows sea ice recovery? Ice Loss - Run with increasing GHG 1 1 m Melt Divergence Mass Budget Change b3.3-4 Melt Divergence Ice Recovery decreasing GHG 4 - Mass Budget Change b3.4b.es1bcom4 Melt Divergence Increased Enhanced Melt Year Year 1 1 m Mass Budget b3.3-4 Enhanced Melt Yrs: Mlt: -.63 Gro:.97 Yrs: 8-99 Mlt: -.66 Gro:.73 Ice loss Month transition Reduced J F M A M J J A S O N D Mass Budget Change b m Mass Budget b3.4b.es1bcom4 Yrs: 81-1 Mlt: -.66 Gro:.7 Yrs: Mlt: -.89 Gro: 1.3 J F M A M J J A S O N D Month Mass Budget Change b3.4b.es1bcom4 Melt

24 b3.4b.es1bcom4 Ice (1 1 m 3 ) Ice Melt (1 1 m 3 ) Ice Divergence (1 1 m 3 ) Ice Volume (1 1 m 3 ) b3.4b.es1bcom Ice Volume (1 1 m 3 ) Melt Divergence b3.4b.es1bcom Ice Volume (1 1 m 3 ) Ice Volume Ice Loss Ice Recovery Mass budget Characteristics as a function of ice state are similar during the ice loss and ice recovery simulations Hints of a difference with ice divergence response (more obvious in other run), but not clear how robust/significant

25 Ice Thickness Distribution 6 5 ITD for April b3.4b.es1 4 % Thickness

26 8. Using Perennial Start-of-Time Ice Melt+Div(black)/(red) Ann Thickness 1.5 Using Perennial Start-of-Time Ice Melt(black)&Div(blue) Anom Ann Thickness

27 Simple Thermodynamics of a Sea Ice Slab! T S F A F Ctop Atmosphere dh q i B dt! = F C " F B! Basal /Melt h i T B F Cbase F B Sea Ice Ocean! F C = k i (T B " T S ) h i!! Conductive Heat Flux (Semtner -layer model) Thickness Year Grow/Melt Melt Annual Thickness (Following Bitz and Roe, 4)