1. Eddies and Ocean Biogeochemistry Andreas Oschlies IFM-GEOMAR

2. The Biological Pump: Traditional 1D View Sea surface z(mix)~z(euph) CO 2, O 2 organic matterinorganic nutrients nutrients,  CO 2 z ``relatively constant´´ C:N:P:-O 2

3. Biological Pump in 3D Z(euphot. zone) Z(winter mixed layer) CO 2, O 2 (1) (2) particulate and dissolved organic matter low lats high lats

4. Z(euphot. zone) Z(winter mixed layer) CO 2, O 2 newly-remineralised dissolved inorganic matter (3a) (1) (2) particulate and dissolved organic matter low lats high lats Biological Pump in 3D

5. Z(euphot. zone) Z(winter mixed layer) CO 2, O 2 newly-remineralised dissolved inorganic matter (3a) (1) (2) particulate and dissolved organic matter low lats high lats (3b) newly-generated inorganic matter deficit (Oschlies & Kähler, 2004) Biological Pump in 3D

6. Potential of the biological pump Present-day sea-surface nitrate concentrations mmol/m 3 Mean profile (Conkright et al., 1994) Controls are not fully understood

7. Potential of the biological pump Present-day sea-surface nitrate concentrations mmol/m 3 Mean profile (Conkright et al., 1994) Controls are not fully understood Subtropical deserts

8. Surface Chlorophyll Biogeographical Provinces

9. Nitrate distribution Subtropical nitrate “bowl” Observed biological production requires net supply across NO 3 =const surfaces. Gauss’s theorem: mean advection cannot contribute to transport across mean iso-surface.  diapycnal mixing  Ekman transport anomalies  eddy stirring (isopycnal & dia-nutrial)

10. Where have eddies come into play? Apparent observational discrepancy in oligotrophic subtropical gyres: Large-scale biogeochemical estimates of export production >> local direct measurements Still not fully resolved despite several decades of research

11. Subtropical desert conundrums Simulated NO 3 supply to euphotic zone Eastern basin: ~factor 12 high rates of O 2 consumption, (Jenkins, 1982: ~0.6 mol N m -2 yr -1 )

12. Subtropical desert conundrums Simulated NO 3 supply to euphotic zone Eastern basin: ~factor 12 high rates of O 2 consumption, (Jenkins, 1982: ~0.6 mol N m -2 yr -1 ) low 15 NO 3 uptake, diff. NO 3 supply (Lewis et al., 1986: ~0.05 mol N m -2 yr -1 )

13. Subtropical desert conundrums Simulated NO 3 supply to euphotic zone Eastern basin: ~factor 12 high rates of O 2 consumption, (Jenkins, 1982: ~0.6 mol N m -2 yr -1 ) low 15 NO 3 uptake, diff. NO 3 supply (Lewis et al., 1986: ~0.05 mol N m -2 yr -1 ) low simulated NO 3 supply (Oschlies, 2002: ~0.02 mol N m -2 yr -1 )

14. Subtropical desert conundrums Simulated NO 3 supply to euphotic zone Eastern basin: ~factor 12 high rates of O 2 consumption, (Jenkins, 1982: ~0.6 mol N m -2 yr -1 ) low 15 NO 3 uptake, diff. NO 3 supply (Lewis et al., 1986: ~0.05 mol N m -2 yr -1 ) low simulated NO 3 supply (Oschlies, 2002: ~0.02 mol N m -2 yr -1 ) Bermuda: ~ factor 4 high rates of 3 He supply (Jenkins, 1988 : ~0.6 mol N m -2 yr -1 ) lower sedimentation rates (Michaels et al.,1994:~0.15 mol N m -2 yr -1 )

15. Subtropical desert conundrums Simulated NO 3 supply to euphotic zone Eastern basin: ~factor 12 high rates of O 2 consumption, (Jenkins, 1982: ~0.6 mol N m -2 yr -1 ) low 15 NO 3 uptake, diff. NO 3 supply (Lewis et al., 1986: ~0.05 mol N m -2 yr -1 ) low simulated NO 3 supply (Oschlies, 2002: ~0.02 mol N m -2 yr -1 ) Bermuda: ~ factor 4 high rates of 3 He supply (Jenkins, 1988 : ~0.6 mol N m -2 yr -1 ) lower sedimentation rates (Michaels et al.,1994:~0.15 mol N m -2 yr -1 ) Substantial interannual variability (Lipschultz, 2001; Oschlies, 2001)

16. Subtropical desert conundrums Simulated NO 3 supply to euphotic zone Eastern basin: ~factor 12 high rates of O 2 consumption, (Jenkins, 1982: ~0.6 mol N m -2 yr -1 ) low 15 NO 3 uptake, diff. NO 3 supply (Lewis et al., 1986: ~0.05 mol N m -2 yr -1 ) low simulated NO 3 supply (Oschlies, 2002: ~0.02 mol N m -2 yr -1 ) Bermuda: ~ factor 4 high rates of 3 He supply (Jenkins, 1988 : ~0.6 mol N m -2 yr -1 ) lower sedimentation rates (Michaels et al.,1994:~0.15 mol N m -2 yr -1 ) Substantial interannual variability (Lipschultz, 2001; Oschlies, 2001) Discrepancy based on different tracers! Role of conversion factors?

17. Where have eddies come into play? Large-scale biogeochemical estimates of export production >> local direct measurements –Trace metal contamination, sediment trap problems,… => underestimated local production rates?

18. Where have eddies come into play? Large-scale biogeochemical estimates of export production >> local direct measurements –Trace metal contamination, sediment trap problems,… => underestimated local production rates? –Unintended tradition of undersampling!  under-representation of episodic eddy events?

19. Evidence for episodic nutrient supply Section Azores – Cape Farewell (Strass, 1992)

20. Evidence for episodic nutrient supply by eddies (McNeil et al., 1999) Bermuda Testbed Mooring Time series 4 months, eddy time scale 15 days.

21. Eddy pumping concept (vertical one) (McGillicuddy et al., 1998)

22. On the relevance of eddy pumping “vertical flux of nutrients induced by the dynamics of mesoscale eddies is sufficient to balance the nutrient budget” “Eddy pumping and wintertime convection are the two dominant mechanisms transporting new nutrients into the euphotic zone” Nutrient flux by eddy pumping “is more than an order of magnitude higher than the diapycnal diffusive flux as well as … vertical transport due to isopycnal mixing”.

23. First counterargument: Statistics Should we have missed the important events? Undersampling of episodic events  under-representation of episodic events? chance of under-representation = change of over- representation

24. Undersampling = under-representation? Hawaii Ocean Timeseries Site (Karl et al., 2003) T, z = 0m T, z = 200m

25. (Siegel et al., 1999) SLA >3 years BATS vs Topex-Poseidon Undersampling = under-representation?

26. Counter argument 2: high estimates based on models Falkowski et al. (1991), eddy off Hawaii: infer ~< 20% enhancement of large-scale primary production by eddies. –based on direct fluorescence measurements of primary production

27. Counter argument 2: high estimates based on models Falkowski et al. (1991), eddy off Hawaii: infer ~< 20% enhancement of large-scale primary production by eddies. –based on direct fluorescence measurements of primary production McGillicuddy et al. (1998), eddy off Bermuda: infer ~100% enhancement of large-scale nutrient supply. –based on nitrate-density relationship and inconsistent model assumptions

28. Altimetry-based eddy-pumping estimate (Siegel et al., 1999) SLA z euph Estimated nitrate supply

29. Altimetry-based eddy-pumping estimate (Martin & Pondaven, 2003) Based on mutually exclusive assumptions: All eddy events contribute to local nitrate flux (wave-like eddies) 100% of upwelled nutrient is taken up locally (slow growth at base of euphotic zone  water must be trapped in moving eddy) Plausible efficiency more likely 20-25%

30. Model-based assessment Spring bloom in eddy-resolving model (1/9x2/15 degrees) ecosystem model, (Oschlies & Garcon, 1999) (Oschlies, 2002)

31. Model statistics at BATS mean NO 3 rms vert. displ. of iso-NO 3 surfaces Corr 2 (SSH,Z(NO 3 )) Corr 2 ( ,NO 3 )) BATS (1/9) o model ~ correct amplitude for lifting of iso-NO 3 surfaces

32. Model-based assessment at BATS mg Chl/m 3 mmol NO 3 /m 3 Chlorophyll Nitrate cumulative NO 3 supply SSH 0.05mol/m 2 after spring (Lipschultz (2001): 0.07 in 1992, 0.04 in 1993) Siegel et al. (1999) method would predict 6 times too much NO 3 supply associated with eddy event (1). (Oschlies, 2002)

33. Counter argument 3: Recharging issues Eddy-pumping process. recharging time

34. Eddy-pumping process. recharging time Sinking is diapycnal transport Counter argument 3: Recharging issues

35. Eddy-pumping process. recharging time Sinking is diapycnal transport Recharging of nutrients on shallow isopycnals matters. Counter argument 3: Recharging issues

36. Eddy-pumping process. recharging time Sinking is diapycnal transport Recharging of nutrients on shallow isopycnals matters. Recharging requires diapycnal nutrient transport (local or remote). Counter argument 3: Recharging issues

37. Eddy-pumping process. recharging time Sinking is diapycnal transport Recharging of nutrients on shallow isopycnals matters. Recharging requires diapycnal nutrient transport (local or remote). Bottleneck is diapycnal transport rather than isopycnal uplift! (Oschlies, 2002) Counter argument 3: Recharging issues

38. What about eddy-resolving models? Idealised models (frontal dynamics)

39. Idealised models (Levy et al., 2001) Large local impacts > 100% increase in regional production. Often run in spin-up or spin-down mode. Representative of steady-state large-scale mean?

40. “We can never do merely one thing” (Hardin, 1985) SST New Production(Mahadevan & Archer, 2000) New production increases with finer and finer resolution. No convergence seen, yet.

41. (Mahadevan & Archer, 2000) 0.36 o C cooling over 120/2=60 days BATS:Hawaii for MLD = 50m: 14 W/m 2 18 W/m 2 for MLD = 100m: 28 W/m 2 36 W/m 2 120 day mean higher NO 3 supply lower SST Heat transport constraint on nutrient transport? “We can never do merely one thing” (Hardin, 1985)

42. Basin-scale models (i) Spring bloom in eddy-resolving model (1/9x2/15 degrees) ecosystem model, (Oschlies & Garcon, 1999) (Oschlies, 2002) permitting

43. Surface heat “flux correction” 1/9 o model, forced by ECMWF 1989-93 ERA 1/3 o model, forced by ECMWF 1989-93 ERA About 25W/m 2 additional heating required to reproduce observed SSTs, (little less at higher resolution: ML restratification by eddies) (Oschlies, 2002)

44. Eddy-induced stratification Heat flux required to balance eddy-induced stratification of ML. Eddies stratify and heat the surface ML in most areas. (Oschlies, 2002) Baroclinic instabilities in ML generate stratification light warm dense cold (Nurser & Zhang, 2000) y z

45. Sub-mesoscale heatflux Inferred from altimetry: Positive everywhere! (Fox-Kemper & Ferrari, 2008) NH winter SH winter

46. Simulated North Atlantic spring bloom 1/3 x 2/5 degrees 1/9 x 2/15 degrees Eddy resolving looks “better”. Is there a significant net impact? Surface chlorophyll (mg/m 3 ) (Oschlies, 2002)

47. Small difference in oligotrophic subtropical gyre. Some difference at gyre’s margins. Simulated annual NO 3 supply into upper 126m (mol m -2 yr -1 ) eddy resolving eddy permitting viscous (Oschlies, 2002)

48. Eddy impacts on mean nutrient supply total supply by eddies Eddy supply by vertical excursions (includes “eddy pumping”) Eddy supply by lateral stirring (exceeds vertical eddy contribution over large parts of subtropical gyre!) (Oschlies, 2002)

49. Role of lateral stirring Supply by lateral stirring might reach larger distances for organic nutrients with longer lifetimes/slower utilisation rates (Lee & Williams, 2000)  =3 months  =1 year

50. Basin-scale models (ii) McGillicuddy et al. (2003): Nutrient transport model –z = 0-104m: N uptake rate =  min(Q N,L) –z > 104m: Remineralisation: 1/  [ NO 3 obs (x,y,  0 ) – NO 3 ] –Sensitivity experiments:  = 10, 30, 60 days (results only shown for  = 10 days, though) NO 3 00 (McGillicuddy et al., 2003)

51. Basin-scale models (ii): NP pathways 1/10 o degree model (McGillicuddy et al., 2003) For  = 10 days, vertical advection by eddies dominates nutrient supply!

52. Overestimated eddy pumping by rapid restoring. Recharging issue: 10 days realistic? How different are results for  = 30 days, 60 days? time

53. Conclusions (i) Eddy pumping occurs, but mean impact has often been grossly overestimated because of inconsistent assumptions about time scales.

54. Conclusions (i) Eddy pumping occurs, but mean impact has often been grossly overestimated because of inconsistent assumptions about time scales. Vertical eddy pumping cannot resolve observational discrepancy.

55. Conclusions (i) Eddy pumping occurs, but mean impact has often been grossly overestimated because of inconsistent assumptions about time scales. Vertical eddy pumping cannot resolve observational discrepancy. Lateral stirring by eddies at least as important.

56. Conclusions (i) Eddy pumping occurs, but mean impact has often been grossly overestimated because of inconsistent assumptions about time scales. Vertical eddy pumping cannot resolve observational discrepancy. Lateral stirring by eddies at least as important. Other processes overlooked previously? –Eddy/wind interactions –Submesoscale

57. Eddy/wind interactions Theory: Induces vertical circulation at eddy’s margin (Martin & Richards, 2001; Mahadevan et al., 2008, comment on McGillicuddy et al., 2007)

58. Simulated impact of eddy-wind interaction (Eden & Dietze, 2009) Simulated new production (mmol C m -2 d -1 ) without wind-current interaction with wind-current interaction Difference ~ 5% reduction Also: reduced EKE (10-50%), reduced energy input by wind

59. Submesoscale upwelling? Ubiquitous, large associated vertical velocities (Martin & Richards, 2001)(Levy et al., 2001) PRIME eddy, c.i. 5m/day Heat flux constraint on nutrient fluxes?

60. First results from steady-state basin-scale simulation 1°1/3° 1/9° 1/27°1/54° SSS, yr 100 1/27 o – 1/9 o : Less phytoplankton, Less new production! (courtesy Marina Levy)  PHY  NP

61. Conclusions (ii) Eddy/wind interactions have small (negative) impact on nutrient supply. Submesoscale variability may reduce nutrient supply (ML restratification). Can we pump nutrients without pumping heat? –Useful constraints from surface heat fluxes? –Depends on correlations of T and NO 3 in thermo- /nutricline.

62. Why should we still want to resolve eddies in biogeochemical models? Isopycnal stirring important to get oxygen minimum zones right (“shaddow zones”)

63. Ventilation of OMZs K iso zonal K iso meridional (Eden & Greatbatch, 2009)

64. Ventilation of OMZs K iso zonal K iso meridional (Eden & Greatbatch, 2009) Dissolved O 2 along 23 o W, 4/3 o model K iso =0 m 2 /s K iso =2000 m 2 /s  mol/l <5 >50

65. Impacts on species composition? Lima et al. (2002): larger phytoplankton favoured at higher eddy activity Hansen & Samuelsen (2009): more diatoms, less flagellates at finer resolution off Norway (run for 1.5 years) (run for 200 days)

66. Why should we still want to resolve eddies in biogeochemical models? Isopycnal stirring important to get oxygen minimum zones right (“shaddow zones”) Sensitivity of Southern Ocean CO 2 uptake to past & future climate change. Eddies as means to structure marine ecosystems? Impact on species composition?

67. Thank you!

68. Why should be want to resolve eddies? Plots look so much nicer… It’s very expensive! –Computational efforts: 1 x (1/10) o ~ 1000 x 1 o –OK for process studies –Impact on mean properties?

69. Idealised models (ii) PHY ZOO PP (Spall & Richards, 2000) Large local impacts < 10% increase in regional production. Equilibrium reached?

70. Simulated impact of eddy-wind interaction (Eden & Dietze, 2009) Simulated EKE (cm 2 s -2 ) without wind-current interaction with wind-current interaction Difference ~ 10-50% reduction

71. Analogy with heat budget (Greatbatch et al., 2007) Gauss’s theorem: mean advection cannot contribute to net transport across closed mean-isosurface

72. Analogy with heat budget (Greatbatch et al., 2007) Export of organic matter Gauss’s theorem: mean advection cannot contribute to net transport across closed mean isosurface

73. Sensitivity to isopycnal mixing Typical range in coarse-resolution models (has little effect on density and velocity fields) Which diffusivity is “correct”, if any?