What governs the location of the Southern Ocean deep winter mixing in CESM

Transcription

1 NSF NCAR WYOMING SUPERCOMPUTER CENTER DOE SCIDAC FUNDED PROJECT What governs the location of the Southern Ocean deep winter mixing in CESM Justin Small Dan Whitt Alice DuVivier Matt Long Acknowledging: Bill Large, Kevin Trenberth, John Fasullo, Peter Sullivan Frank Bryan, Baylor Fox Kemper, Qing Li, Ian Grooms

2 Background/Motivation Deep winter mixing importance to mode water formation zones Subantarctic Mode Water, large regions of low PV (Hanawa and Talley 2001) Important region for heat and anthropogenic carbon exchange with atmosphere. CMIP5 model bias (Sallee et al 2013) Winter mixed layer too shallow Linked to freshwater forcing & surface salinity biases, heat flux bias CESM model bias (Weijer et al. 2012, Alice DuVivier Feb OMWG meeting) Winter mixed layer too shallow salinity stratification important improvements seen with high resolution ocean Possible deficiencies in mixed layer model, improvements in non local mixing term? (Large, Sullivan, Patton)

3 Proposed controlling factors J. Clim Air sea buoyancy flux McCartney et al 1977, 1982, Cerovecki et al 2011, 2013 Enhanced at high resolution Lee et al Ekman buoyancy flux Rintoul and England 2001 look at Temp. only, interannual variability 3. Enhanced mixing through Langmuir McWilliams et al. 1997, Belcher et al. 2012, McWilliams 2014 Li et al. 2016, 2017 most effective in summer 4. Eddy mixing processes at ocean fronts? Sallee et al 2006, 2008 In early season, deepest mixed layers occur close to fronts and in eddies Factors in red are discussed in this presentation

4 Seasonal Cycle AUSTRAL SUMMER AUSTRAL WINTER MIXED LAYER DEPTH MIXED LAYER DEPTH m SURFACE HEAT LOSS, POSITIVE: OCEAN LOSES HEAT m SURFACE HEAT LOSS, POSITIVE: OCEAN LOSES HEAT ZONAL WIND STRESS Wm 2 Wm 2 ZONAL WIND STRESS Note change of color bar for mixed layer depth between seasons Nm 2 Nm 2

5 Resolution dependence of MLD, SHF MLD in forced models high and low res. JAS, density criterion Left: ARGO, m m m Left: Forced low res, GIAF, contours of 200m, 400m MLD from ARGO overlaid Left: Forced high res, GIAF, contours of 200m, 400m MLD from ARGO overlaid

6 Factor 1: Air sea heat flux and buoyancy flux

7 Animation of SST and latent heat flux, highres coupled model Note large latent flux north of ACC during cold air outbreak, both in summer and winter In summer, loss of heat from latent and sensible flux is counteracted by solar radiation In winter, solar radiation much weaker and the heat flux terms combine to form strong heat loss Equatorward of ACC

8 Annual mean Large Yeager 2009 Trenberth Fasullo 2017 Sign convention: positive implies ocean loses heat Forced low resolution ocean (g40.001) Forced high res ocean (M. Long) Wm 2

9 Net surface heat flux: Low res. POP minus Trenberth Fasullo estimate SHF, JAS, LR OBS Wm 2 Net surface heat flux: High res. POP minus Trenberth Fasullo estimate m Sign convention: positive implies ocean loses heat SHF, JAS, HR OBS Wm 2 Low res POP surface heat loss is 50% 100% too small in dark blue regions of top plot

10 Mixed Layer Depth: Low res. POP minus ARGO observations MLD, JAS, LR OBS m Mixed Layer Depth: High res. POP minus ARGO observations MLD, JAS, HR OBS m

11 Downstream effect of air sea heat loss on MLD

12 ARGO MLD, , JAS ARGO: potential density (400m) density (surface): JAS m ARGO: Potential density at 400m minus density at surface, JAS Highly stratified Stratification weakens kgm 3 Water parcel in Agulhas initially highly stratified with warm and salty surface. Under effect of surface heat loss, it cools downstream. At some point, the cold salty water is dense enough to convect. Put another way, ocean heat convergence balances heat loss in the Agulhas, but it is not enough to balance heat loss in south east Indian Ocean.

13 ARGO: Potential temperature at 400m minus temperature at surface, JAS Low res POP: Potential temperature at 400m minus temperature at surface, JAS ARGO and high res POP show rapid reduction of thermal stratification in south west Indian Ocean. Low res POP has very slow reduction of stratification. High res POP: Potential temperature at 400m minus temperature at surface, JAS

14 Sensitivity of high resolution coupled run to smoothing of air sea fluxes See next slide

15 DIFFERENCE: HI RES COUPLED MINUS HI RES COUPLED CESM RUN with SST smoothed for coupling JAS seasonal mean HMXL difference m JAS seasonal mean Total surface heat flux difference (positive cools ocean) Wm 2

16 Factor 2: Ekman buoyancy flux and relationship to Air sea buoyancy flux Ekman buoyancy flux westerly winds drive northward Ekman flow of dense water over light Note that Ekman buoyancy flux has a significant part due to salinity gradients counteracting temperature gradients Note that air sea buoyancy flux is dominated by heat flux (freshwater flux has weak influence) Air sea buoyancy flux (Large and Nurser 2001, Cerovecki et al. 2011) Rintoul and England 2001 Sallee et al 2006, 2008, 2010 They only consider temperature not buoyancy b is buoyancy and B is buoyancy flux

17 High res JAS mean Ekman buoyancy flux High res Air sea buoyancy flux m 2 s 3 Last color level is +/ 5e 8, interval of 0.5e 8 HI RES ECO, WINTER MEAN High res Air sea buoyancy flux + Ekman m 22 ss 3 3 Winter JAS mean fields with contours of JAS mean HMXL overlaid (200m contour) m 2 s 3

18 Model bias in surface heat flux, SST and SSS and how they are related Sea surface heat flux (SHF) biases governed by SST differences Sea surface salinity (SSS) difference structure is similar to SST differences: low res model (LR) is too cold and too fresh in ACC region. Circumglobal effect of missing the Agulhas Retroflection?? Hi res model (HR) has better surface salinity, temperature distribution, especially in Indian Ocean Presumably due to better flow patterns (e.g. Agulhas and Return Current) and air sea heat fluxes (?) I only show surface temp and salinity subsurface properties and their effect on MLD more complicated and discussed by Alice DuVivier (OMWG Feb. 2017)

19 SST, JAS, LowRes OBS SST, JAS, HighRes OBS SST, JAS, HighRes LowRes

20 SHF, JAS, LR OBS Wm 2 Sign convention: positive implies ocean loses heat SHF, JAS, HR OBS Wm 2 SHF, JAS, HR LR Wm 2 Heat flux feedback is about 40W/m2/deg. C in model

21 SSS, JAS, LR OBS PSU SSS, JAS, HR OBS PSU SSS, JAS, HR LR PSU

22 PV distribution and implication for modewaters

23 Potential vorticity section, Indian Ocean, September CESM g case ECO 1.0 deg CESM g case ECO 0.1 deg WOA deg climatology Eg. McCartney s potential vorticity (Q=f/gN^2) in the Indian sector. SAMW too stratified and not sufficiently voluminous in CESM Compared to WOA13 and WOCE 0.1 degree reduces stratification & increases volume Model also has too strong stratification and Q in S E Pacific

24 Summary Deep mixed layers in austral winter form in narrow band north of ACC CESM with high resolution ocean reproduces spatial distribution of deep mixed layers Maximum surface heat loss also occurs in this band Smoothing out the surface fluxes leads to reduction in mixed layer depth Air sea buoyancy flux much larger than Ekman buoyancy flux and co located with deeper mixed layers Low resolution model has incorrect surface T/S characteristics in and downstream of Agulhas Return Current too cold and too fresh SAMW is too stratified and not sufficiently voluminous in CESM; high resolution is slightly closer to observations. Results appear to confirm hypothesis of Lee et al (2011) that continuous and strong cooling of water parcels in eastward flow just north of ACC leads to deep convection Reasonable for Indian Ocean but does the downstream influence extend to S E Pacific?

25 Discussion: Way forward PV budget Related to density stratification budget of Lee et al 2011 Tracking particles and buoyancy loss Investigate new high resolution air sea flux products

26 Extra Slides

27 Proposed controlling factors 1. Atmosphere forcing of ocean surface boundary layer processes Air sea buoyancy flux Enhanced mixing through Langmuir 2. Coupled ocean circulation and boundary layer processes Ekman buoyancy flux Boundary layer processes impact down stream mixed layers (Lee et al.) Feedbacks between ocean circulation and boundary layer mixing (via PV) Mesoscale processes at ocean fronts, mesoscale air sea interaction In early season, deepest mixed layers occur close to fronts and in eddies Factors in red are discussed in this presentation

28 Discussion How to use high res results to improve standard resolution models Air sea fluxes T/S distribution Flow field

29 JAS mean Large Yeager 2009 Trenberth Fasullo 2017 Wm 2 Forced low resolution ocean (g40.001) Wm 2 Forced high res ocean (M. Long) Wm 2 Wm 2

30 MLD differences expressed as ratios e.g. (LR OBS)/OBS Ratio Ratio Ratio

31 Low res mean Ekman buoyancy flux Last contour level is +/ 5e 8, interval of 0.5e 8 Low res Air sea buoyancy flux m 2 s 3 Low RES ECO, winter MEAN Low res Air sea buoyancy flux + Ekman m 2 s 3 Winter mean fields with contours of JAS mean HMXL overlaid (200m contour) Low res Air sea buoyancy flux minus Ekman m 2 s 3 m 2 s 3

32 Annual mean sea surface salinity (top) and bias re Levitus (bottom) HIGH RES COUPLED LOW RES COUPLED

33 Sea surface salinity bias from a recent control forced POP run, 1deg.

34 Downstream effect of air sea heat loss on LMD J. Clim. 2011

35 JAS seasonal mean HMXL May mean PD (color) m (1./1000)*Kgm 3 HI RES COUPLED CESM RUN: JAS seasonal mean mixed layer depth and the May stratification (0) (400m)

36 ARGO MLD, , JAS ARGO: potential density (400m) density (surface): JAS m ARGO: Potential density at 400m minus density at surface, JAS Highly stratified Stratification weakens kgm 3 Water parcel initially highl with warm a surface. Und surface heat downstream i t th

37 Seasonal Cycle Seasonal cycle of: Mld Surface heat flux Wind stress From high res coupled model as an example Note that winter deep mixed layers more aligned with SHF than with wind stress Summer mixed layers aligned with wind stress

38 Relationship of MLD to SHF: model biases and resolution differences Following slides show differences (model minus obs, HR model minus LR model) : SHF (surface heat flux): Obs = Trenberth and Fasullo 2017 MLD Obs = Argo Note approximate spatial correspondence between MLD and SHF differences.

39 Next slide shows fractional change in HMXL, SHF Note that there are regions in the north of the ACC in the low res model where surface heat loss is 50% 100% weaker than observed

40 SHF differences expressed as ratios e.g. (LR OBS)/OBS Ratio Ratio Ratio

41 CESM 0.1 degree forced eco run 5 yr September Climatology CESM 1.0 degree forced eco run 5 yr September Climatology McCartney s PV, Q=f/gN 2 WOA13 September climatology at 90 W WOCE 90 W (Jan & Mar 1993)

42 McCartney s PV, Q=f/gN 2 on an isopycnal 26.5 kg/m 3 Models kg/m 3 CESM 0.1 o eco 5 yr Sep Clim CESM 1.0 o eco 5 yr Sep Clim Light mode water kg/m 3 in Indian WOA13 climatology Dense mode water kg/m 3 in Pacific

43 Estimates of observed air sea heat flux and buoyancy flux Sign convention: positive implies ocean loses heat Trenberth and Fasullo 2017 Large and Yeager 2009 Wm 2 Left: Annual mean net surface heat flux from Trenberth, Fasullo product, merging CERES TOA radiation measurements and ERA I energy transports. Positive denotes ocean heat loss. Years Contour interval is 25W/m2. Right: Annual mean net surface heat flux from CORE (Large and Yeager 2009), averaged years Sum of latent, sensible, shortwave net and longwave net. Positive denotes ocean heat loss. Contour interval is 25W/m2