Evaluation of ocean circulation models for the Bering Sea and Aleutian Islands Region

1. Evaluation of ocean circulation models for the Bering Sea and Aleutian Islands Region Albert J. Hermann1 and David L. Musgrave2 National Oceanic & Atmospheric Administration, Pacific Marine Environmental Laboratory, 7600 Sand Point Way NE, Seattle, WA 98115 USA, (206) 526-6495, Albert.J.Hermann@noaa.gov. 2 School of Fisheries and Ocean Sciences, Insitute of Marine Science P.O. Box 757220, Fairbanks AK 99775-7220 USA, (907) 474-7837, musgrave@ims.uaf.edu2

2. This workshop explored the present and future state of ocean circulation modeling and biological modeling of the Bering Sea and Aleutian Island (BSAI) and the North Pacific Major topics: • 1) the present state of knowledge concerning the BSAI • 2) the various types of circulation models which could be applied to the BSAI, with some assessment of their strengths and weaknesses • 3) existing physical and biological models • 4) adequacy of present forcing and bathymetry datasets • 5) current status and future prospects for data assimilation • 6) modeling needs of managers for this region • 7) a timetable over which we might expect the development of improved models

3. 1. Present state of knowledge of the BSAI region • Highly productive • A big, broad shelf • Passes: • wide, narrow, shallow, deep • Canyons • Cross-shelf flux • Powerful tides • A deep basin • Ice

4. A few highlights…. • Flows through passes – spatially variable, strongly mixed, very important to biology • Ice-edge blooms with possible “Oscillating Control” • Distinct shelf regimes via tidal mixing • Getting warmer, less ice! • PDO was significant, but now other modes more important

5. 2. Classes of Ocean Circulation Models • Pure tidal models • Quasi-gestrophic models (simpler physics) • Primitive equation models (hydrostatic but otherwise include all physics) • Terrain-following coordinates • Z-coordinate • Layered coordinate • Unstructured grid

6. 3. Existing physical and biological models

7. 3.1 Atmospheric models • National Center for Environmental Prediction (NCEP) hindcasts • Assimilated atmospheric data • Easy to get! • Biases include shortwave radiation (not enough stratus clouds) • ECMWF hindcasts • Better for Europe, not necessarily better for Bering • Commercial product • Community Climate Systems Program • Offers global hindcasts and climate forecasts • Better shortwave radiation (b/c assimilates cloud data/climatology) • Regional models • ETA – downscales NCEP nowcasts • NARR – downscales NCEP hindcasts • MM5 – general tool for downscaling global winds

8. 3.2. Ice models • Hibler model and its decendents: • Thermodynamics: • Melt/freeze, snow layers • Dynamics • Viscous-plastic solid • Vary in number of ice/snow layers

9. 3.3 Circulation models • Maslowski group • N Hemisphere model based on MOM/POP • Chao group • N Pacific model based on ROMS • Wang group • Bering Sea model based on POM • Curchitser/Hermann group • Northeast Pacific model based on ROMS

10. 3.4. NPZ Models • 3.4.1. 1D models • NEMURO – NPZD and fish • Saury and herring • Merico (2001) – NPZD • phytoplankton succession • 3.4.2. 3D models • Run both “online” and “offline” • Examples: • Wang/Diehl NPZ model of Bering • Hinckley et al NPZ models of CGOA • Powell/Hermann NPZ for Northeast Pacific

11. 3.5 Individual-based Models • Float-tracking plus behavior • Very useful for individual fish species • Can run online or offline • Examples (from CGOA) • Hinckley et al pollock model • Rand salmon model

12. 3.6 Aggregated models • Consider entire food web • ECOPATH – look at steady state • ECOSIM – add time • ECOSPACE – add space and time • Can benefit from coupling with NPZD

13. 3.7 Fisheries models • MSVPA • Other stock assesment models

14. 4. Adequacy of models, forcing and bathymetry • IDEALLY we would like • Uniformly fine scale resolution or adaptive space/time resolution • Numerically accurate/convergent • Handle tides, subtidal flows, mixing all together • Perfectly accurate bathymetry • Perfectly accurate forcing; hincasts and forecasts

15. 4.1 Numerics and resolution • Terrain-following coordinates • Tend to overemphasize bathymetry • Need smoothed bathymetry • Great for surface and bottom boundary layers • Z-coordinate • Less accurate/convergent numerics • Consistent vertical spacing near the surface • Distort bottom topography into stair steps • Layered coordinate • Good for the deep ocean • bad for the shallow ocean (can’t do tidal mixing) • Unstructured grid • Potentially powerful • Hard to implement • Relatively untested in the Bering Sea • Danger of predetermining answer with choice of grid

16. Crucial elements to get right… • Inflow/outflow through the Aleutian passes • sets conditions in the Southeast BS • Outflow through the Bering Strait • Tidal mixing on the shelf • Ice!

17. 4.2 Ice • Hibler-based models are probably sufficient for the Bering Sea (no multiyear ice) • Major uncertainties arise from shortwave radiation forcing; need to improve

18. 4.3 Atmospheric forcing • NCEP probably OK for winds (except for Aleutians) • NCEP shortwave is badly biased • CCSM is promising • Need better bulk flux algorithms (e.g. to relate wind speed to wind stress) • Extended range mesoscale forecasts are impossible • Long range forecasts/scenarios are useful • Downscaling is needed!

19. 4.4 Freshwater discharge • Important in a few areas • Data is essentially nonexistent!

20. 4.5 Tides • Existing models can handle tidal and subtidal dynamics simultaneously • This is crucial for the Bering Sea, as the two interact • Tidal phasing may be biologically important, so want to get it right.

21. 4.6 Bathymetry • OK on the shelf • Need more data in the canyons • Need much more data in the passes • USGS eventually digitizing the Bering Sea charts

22. 4.7 NPZ models • Need to get: • Pelagic/benthic gradients, north-south and cross-shelf • Green Belt at the shelf break • Important prey species for fish • Jellyfish? • IRON and other nutrients

23. 4.8 IBM models • Need better data on fish movement and behaviour • Need better data on space/time distribution • Groundfish surveys have been very useful for modelers.

24. 4.9 Aggregated models • Could benefit from NPZ results • Aggregate NPZ by space, water mass, or biological regime?

25. 4.10 Fisheries models • As with aggregated models, could make more spatially explicit

26. 4.11 Model coupling • IDEAL integrate model might include: • IBMs of Multiple species and life stages • NPZ with multiple size classes • Feedback between IMB and NPZ! • Long time scale simulations • Web-accessible output and graphics

27. 5. Status, Needs and Prospects for Data Assimilation • T,S, nutrients in passes would be powerful constraint • Skill assesment is difficult to do well • Existing physical assimilation capabilities • 3D variational assimilation (ROMS, Chao et al.) • “Weak constraint” blend of data and model • Useful for nowcasts, not as good for dynamical analysis • Inexpensive to run • 4D variational assimilation (ROMS, Moore et al.) • “Strong constraint” adjustment of IC and BC for hindcasts • Can be used for sensitivity analysis, indices! • Can be expensive to run • Possibilities for biological model optimization • 4D variational assimilation (in ROMS) • Genetic algorithms

28. Data sources • SMMR sattelite for ice cover • TOPEX/POSEIDON/AVISO altimetry • No information on the shelf • Long term moorings • Nice long time series should be continued • Long spatial correlation scales make these representative of broad areas on the shelf • XBT data very sparse prior to the 70s • Hydro/mooring data sparse for the Western Bering Sea • BASIS program in the Eastern Bering • Global circulation models for ICs and BCs

29. 6. Needs of managers • Mandates from • National Environment Protection Act • Marine Mammal Protection Act • Endangered Species Act • Stellar sea lion • Sea otters • Fur seals • Right whales • Fin whales • Predictions of 5-10 years are of special interest • Issues include • Bycatch • Indirect effects of fishing • Phys-bio interactions • Hindcasts of circulation and biology can help establish likely response to future change • Need better indices!

30. 7. Estimated timetable of new model products and projects • See the report!

31. Summary I • The ideal circulation model would adequately and simultaneously resolve all the relevant scales of motion and phenomena in the BSAI, e.g. • flows through the Aleutian Passes • seasonal ice • tidal mixing on the shelves. • None of the present modeling approaches can rapidly and simultaneously capture all of these features for extended time periods on today’s computers • continuing advances in computer technology are expected to expand the limits of feasible simulations, at least doubling the possible spatial resolution for such runs before 2010. • Both nested approaches with structured grids, and variable resolution approaches with unstructured grids, appear promising ways forward.

32. Summary II • Present ice model algorithms appear adequate for the Bering Sea. • The accuracy of circulation hindcasts for the BSAI are limited by the paucity of data, especially as regards the passes. • Long-term moorings and systematic hydrographic surveys, in conjunction with altimeter data, will help rectify this deficiency • Effective mathematical approaches are now available in community model codes for effective assimilation of such data into hindcasts and nowcasts. Computer resources are still a limiting factor in the application of some of these codes. • The atmospheric forcing datasets also have outstanding issues (e.g. biased shortwave radiation estimates), which limit the hindcast skill of BSAI simulations, and of ice in particular.

33. Summary III • The ideal scientific/management biological model might include • multiple species and multiple life stage components • specific species treated using spatially explicit IBMs • coupled to multi-compartment NPZ and circulation models • Proper feedback among different components especially challenging • Intermediate step: focus on coupling spatially explicit NPZ with spatially aggregated food web models. • For all models, longer time scales are needed to aid in ecosystem-based management. • Data gaps are even larger for the biology than for the physics of the BSAI • sustained surveys (e.g. the NMFS groundfish surveys) have yielded much useful data for the quantification of food webs.

34. Summary IV • More collaborative development of both physical and biological models is recommended, as they will require substantial human resources. • Human time to examine and interpret the output can be just as limiting as computer hardware • One way to ease the development and interpretation of such multi-investigator models is to provide easy access to model output through web-based software.

35. FIN! • http://halibut.ims.uaf.edu/SALMON/BSIAModelWorkshop

36. Features of the Bering Sea • A big, broad shelf • Passes: • wide, narrow, shallow, deep • Canyons • Cross-shelf flux • Powerful tides • A deep basin • Ice • High production • Lots of fish! • Climate change

37. Foci of the workshop • review existing modeling efforts in the BSAI • assess strengths and weakness of the various types of ocean circulation models in accurately representing circulation, mixing and exchange due to • forcing mechanisms (winds, tides, ice formation, river runoff) • topographic features (coastline, shelf break, Aleutian Island passes) • evaluate various monitoring and process studies that would improve the accuracy of the models • describe pathway for using these models to develop products that would be useful for resource managers and users.