Inside this Issue

• Expanding Tecnology Aids Volcano Early Warning and Detection Activities
• Digital Dog Mushing
• Bering Sea Community Modeling
• Reclaiming History with Technology
• A Summer of Invention, Innovation and Immersion in HPC
• Forward Momentum in Computational Science

Dr. Kate Hedstrom, ARSC
Dr. Georgina Gibson, School of Fisheries and Ocean Sciences, UAF
Dr. Al Hermann, Pacific Marine Environmental Laboratory, National Oceanographic and Atmospheric Administration
Dr. Enrique Curchitser, Lamont-Doherty Earth Observatory, Columbia University

Story by Lorien Nettleton

The Bering Sea is one of the most significant fisheries in the United States. Multitudinous pollock populations make it the largest single-species fishery in the world. Accounting for 30 percent of national sea-catch income, the US fleet alone hauls in 2.5 billion pounds of pollock annually, and there is high confidence in the sustainability of this resource.

However, little is currently understood about how variables such as nutrients, temperature and ocean circulation cumulatively impact the biology of this system, and how that may affect the annual fish harvest. Pollock populations may be strongly affected by zooplankton fluctuations, which are, in turn, affected by environmental conditions and timing, climatic variability or human activity. Changes to any of these factors could cause significant economic consequences to the fishing industry. Knowledge of the inter-relationships of such variables is crucial to understanding when and where adverse situations may occur.

To generate a more intimate understanding of the physical and biological processes that govern the maintenance or decline of fish populations, the Alaska Ocean Observing System (AOOS) has drawn on the cumulative expertise of collaborators at the University of Alaska Fairbanks (UAF) and abroad to undertake a community modeling effort, distributing the investigation of specific mechanisms thought to link ecosystem responses with environmental variability.

Building the Model

ARSC oceanographic specialist Kate Hedstrom is deeply involved in the continued individual and coupled modeling efforts. The Regional Ocean Modeling System (ROMS), a free-surface, hydrostatic, primitive equation ocean model, is the foundation of the AOOS modeling project. Hedstrom has contributed to the development of ROMS for much of her career as the model has evolved into a multi-field ocean simulation tool. ROMS contains a range of features necessary for modern ocean modeling including high-order advection schemes; accurate pressure gradient algorithms; several sub grid-scale parameterizations; atmospheric, oceanic, and benthic boundary layers; biological modules; coupled sea-ice and data assimilation.

ROMS was initially optimized for vector computers, and then ported to OpenMP. Hedstrom helped debug two different MPI implementations of parallel ROMS so that it can be run on any of ARSC’s supercomputers, including Iceberg, ARSC’s IBM p655+, p690+ system.

To verify the accuracy of ROMS in each system, the physical components—circulation, heat flux, currents, atmospheric features and water budgets—were modeled independently and validated before being used as forcing files for the increasingly complex coupled simulations that followed.

In December 2006, Hedstrom submitted a coupled ice-ocean model of the northeast Pacific to Iceberg. The simulation ran within a spatially-nested grid, starting with a global ocean circulation model with a resolution of ten kilometers per grid point, and covering a 47-year time period, from 1958 to 2004. Prior to the late 1970s, observable data for the sea ice was limited to atmospheric data from weather stations, making validation more difficult. But once the simulation reached into the late 70s, satellite imagery became available showing ice coverage to which the final 30 years of the run could be compared. In this manner, researchers were able to assess the earlier decades for accuracy based on comparisons with the last 30 years of satellite data.

The large scale of the simulation provides a holistic view of the circulation and thermal processes of the ocean. The results of the global ocean circulation model are used as boundary conditions for a grid centered on the northeast Pacific at l0 km resolution, providing greater detail of currents and water budget movement. The 10 km grid becomes a boundary condition for a still more focused Bering Sea simulation at 4 km resolution, which researchers hope will do a better job of representing the bathymetry to achieve better results through the passages as well as with circulation through the Aleutian passages. The nested domain will also include tides, which are extremely important in the Bering Strait.

After gaining confidence in the accuracy of the ocean models and in the circulation, bathymetry and heat-flux aspects, the simulation was ready for the inclusion of ice data as a forcing file. Enrique Curchitser, from Lamont-Doherty Earth Observatory in New York, provided the external files required by the model for forcing and boundary conditions. The coupled sea ice model was coded and tested by Paul Budgell, a physical oceanographer at the Institute of Marine Research in Norway. The thermodynamic ice data came from a Kantha/Mellor model that simulates and describes the heat flux and thermal properties of developing ice, and the resulting interaction with the hydrodynamics of the sea. The dynamic model, adapted from an elastic-viscous-plastic (EVP) model developed by Los Alamos National Laboratory in New Mexico, shows how ice physically reacts to the wind by drifting, shifting and pushing into itself.

Modeling Fish Populations

Al Hermann, of the Pacific Marine Environmental Laboratory (PMEL), part of the National Oceanic and Atmospheric Administration (NOAA), is looking at variations in observed fish populations. By using models of circulation and plankton and fish dynamics in the Pacific Ocean, from Southern California’s coast up to the Gulf of Alaska and into the Bering Sea, Hermann seeks to uncover physical reasons for the apparent relationship between shifting rates in phytoplankton and zooplankton populations and the fish stocks in various regions of the Pacific.

Hermann’s method involves running Hedstrom’s numerical model output with included physical processes, and exploring adjustments that might increase the accuracy of the model. The groups then collaborate to form suggestions for adjustments that can be made to improve the output. PMEL also runs its own version of the models over the same domains with differing boundary conditions, the results of which are made available to researchers who seek to use larger simulations for forcing fields in high-resolution, small-scale models.

Biological Models

ARSC Postdoctoral Fellow Georgina Gibson (left) and ARSC Oceanographic Specialist Kate Hedstrm discuss their work. The researchers are both contributing model data to the Alaska Ocean Observing System (AOOS). Photo by Leone Thierman Ice area (above) and volume (below) for the NCEP-forced run (red) and CCSM-forced run (blue). Note: the ice area and volume doesn't go to zero in the summer due to the background minimum levels required by the model.

UAF School of Fisheries and Ocean Sciences (SFOS) research staff and ARSC postdoctoral fellow, Georgina Gibson, is developing and enhancing a biological component of ROMS to improve understanding of ecosystem and fisheries dynamics in the AOOS regions. A change has been observed in the recruitment of juveniles to the pollock fishery in the southeast Bering Sea, depending on whether it is a cold year or a warm year. Gibson’s research aims to explore a number of alternate hypotheses proposed to explain this inter-annual variability through modeling of the biological elements of the Bering Sea. Her work is closely affiliated with Hedstrom’s research as well as the work of Hermann.

Through a series of one-dimensional models, Gibson seeks to describe the biological cycles of each trophic level in the Bering Sea. Nutrients, phytoplankton, zooplankton and pollock, as well as the detritus and benthic settlements, will all be independently modeled. Once the models satisfactorily simulate the biology, they will be coupled to simulate the cycling of nutrients as producers are eaten by primary and secondary consumers and nutrients are cycled back through the environment through decomposition and settling of organic matter to the benthic zone. Here, Gibson will seek to correct the automatic re-distribution of nutrients found in models that assume a certain return of nutrients to the system, by adding another biological layer at the benthic zone. It is known, for instance, that certain benthic consumers, such as crabs, act as storage for the nitrates found in the settling detritus before cycling the nutrients back into the ecosystem. Yet most models do not account for the biology responsible for this nitrogen cycling; they simply factor in the results afterward.

The individual biology models are coupled to a one-dimensional version of ROMS. Once Gibson is satisfied in the simulated cycling of nutrients through the food chain, the individual models will be rendered in 3D and coupled to a 3D implementation of ROMS, which will run in parallel on ARSC machines.

ROMS developments
Because of its reputation as a well-defined coastal short-term simulation as well as the broad community of researchers supporting it, was recently extended for use in conjunction with NOAA’s Community Climate System Model (CCSM) atmospheric application. The addition of ROMS to NOAA’s suite of simulations will further diversify the operational global forecasts for tide, ocean circulation and wave behavior simulations.