"Climate Changes Affect Even the Smallest Fish"
Challenges in science and engineering Summer 2000 A publication of the Arctic Region Supercomputing Center vol. 8 no. 1 Climate Changes Affect Even the Smallest Fish Albert Hermann Researchers at the Alaska Fisheries Sciences Cen- Joint Institute for Study of ter and the Pacific Marine Environmental Labora- Atmosphere and Oceans/ tory in Seattle, Washington are modeling ocean Pacific Marine circulation and the biology of pollock and salmon Environmental to learn more about how climate change will affect Laboratory two species of fish that many humans depend on for food. Sarah Hinckley Climate change affects many aspects of life on Alaska Fisheries Earth. Biological processes are tightly linked to Science Center an intricate ecosystem, allowing changes in cli- Dale Haidvogel mate to affect even the smallest fish in the ocean. Rutgers University “Many things affect fish and fish stocks,” says Al Hermann of the Pacific Marine Environmen- Models produced from Hermann and Hinckley’s data show the locations Peter Rand tal Laboratory. “Climate change and fishing are of fish larvae on February 20 (above) and May 10 (below), five days North Carolina and 85 days after release, respectively. The paths of larvae released at 10 just two of these. We’re looking at how the chang- meters are shown in green, while the paths of the larvae released at 40 State University ing climate affects commercially important fish.” meters are shown in purple. Bathymetry is rendered as a grey surface. The Phyllis Stabeno Hermann and his fellow researchers are us- viewpoint looks to the north toward Prince William Sound. Pacific Marine ing several types of models in order to limit the Environmental number of unknowns in their research. These Laboratory models include information on ocean circula- tion, temperature, salinity, predators, prey and the fish themselves. Specifically, the researchers are using individual-based models (IBMs) to look at fish on a more detailed level, rather than as a group. Previous fish models have used an Eulerian approach to modeling, which only gave information about fish in a broad perspective. Such models are sometimes inefficient when studying why fish are reacting to certain events in the environment. “There is a popular saying in the fisheries- oceanography community that ‘the average fish is In This Issue dead,’” says Hermann. “Only the rare, lucky fish survives. With an individual-based approach, you Climate Change ....................................... 1 can say more about what made that particular fish Ocean Circulation ................................... 2 successful. Concentrations of the average individual Viewpoints ............................................... 3 are not as informative.” Hermann’s project is part of the Global Ocean ARSC Bits and Bytes ............................... 3 Ecosystems Dynamics (GLOBEC) program, orga- Global Warming ...................................... 4 nized by oceanographers and fisheries scientists to address the question of how global climate change ARSC Currents ........................................ 6 may affect abundance and production of animals Staff Highlights ........................................ 7 in the sea. Hermann works in conjunction with biologist Who We Are ............................................ 7 Sarah Hinckley of the Alaska Fisheries Sciences Visualization Gallery ................................ 8 see Climate Changes on page 2 Ocean T’ (°C) -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 v ’ = 0.3 m/s Circulation 0 x=120m Dr. Eric Skyllingstad Eric Skyllingstad and Hemantha -10 Oregon State Wijesekera of Oregon State University are -20 University working to improve understanding of z(m) small-scale mixing processes in the coastal -30 Dr. Hemantha ocean environment. They hope their -40 Wijesekera project will increase the accuracy of coastal Oregon State -50 mesoscale prediction models by adding University physically-based approximations to one- -60 0 50 100 dimensional mixing parameterizations. y(m) In any modeling process, the par- ticular elements of the process must be Temperature and velocity structure derrived from the large eddy simulation models run by Skyllingstad and Wijesekera. developed to improve the reliability and success of larger models. These re- searchers are working to do just that—and in the process driven turbulence. Depending on wind direction, currents aid the Department of Defense in creating better oceanic generated by wind forcing can cause deformation at the research models. mixed-layer base and ocean bottom, which in turn causes Accurate measurement of turbulent mixing of increased vertical mixing. The researchers will use large-eddy oceanic waters is essential for understanding the basic simulation models to predict the movement of the ocean cur- dynamics of coastal circulation. Recent oceanic obser- rents in these two boundary regions. The data they produce vations on the Oregon coast, along with modeling will be tested against field data collected during the summer results, suggest a strong link between mesoscale coastal of 1999 along the Oregon shelf. circulation and turbulent mixing rates. Skyllingstad and Because the mixing processes Skyllingstad and Wijesekera Wijesekera hope to determine the structure and study are much smaller than the resolution of typical coastal statistics of turbulence for a domain representing the ocean models, their research will be useful in producing more inner- and mid-shelf region off the coast of Oregon. accurate larger-scale models. For more information on this Cold, up-welled water is warmed by solar energy as it project, contact Eric Skyllingstad at email@example.com. rises, and by heat mixed vertically through wind- and wave- Climate Changes... cont. Center, also in Seattle. Hinckley concentrates on the Spring 2000 Vol. 8 No. 1 biological aspects of the fish and their prey in the models. “The IBM is coupled with physical and nutrient/phy- Challenges is a publication of the Contact us at: toplankton/zooplankton models,” says Hinckley. “This Arctic Region Supercomputing Center voice: (907) 474-6935 allows us to produce prey in the model that are both University of Alaska Fairbanks fax: (907) 474-5494 temporally and spatially distributed—so not everyone’s 910 Yukon Drive, Suite 106 Visit our web site at: getting the same amount of food.” P.O. Box 756020 http://www.arsc.edu Such coupled models with fine spatial resolution Fairbanks, Alaska 99775-6020 allow the researchers to ensure the models are more true Frank Williams .................................................................. Director to life. This calls for a greater number of grid points and smaller time-steps, which requires both the processing Barbara Horner-Miller ..................................... Associate Director/ power of supercomputers and the storage capacity of the User Services Director StorageTek Silo. Jenn Wagaman .............................................. Publications Specialist Hermann and Hinckley’s projects are part of the GLOBEC program, and NOAA’s Fisheries and Ocean- L.J. Evans ............................................... Public Affairs Coordinator ography Coordinated Investigations (FOCI) Programs. Their work will help fisheries managers make better informed decisions about fishing management and pre- TM diction. For more information on this project, go to http:/ /www.pmel.noaa.gov/~hermann/. The University of Alaska is an affirmative action/equal opportunity employer and educational institution. 5/00 2 Viewpoints Welcome to ARSC’s spring edition of Challenges. You will meet user consultants Tom Baring, Derek Bastille will notice a significant change in the appearance of and Shawn Houston, and learn more about how we this issue. This update in format allows us to give ad- work to meet our users’ needs. ditional credit to the scientists and organizations who The universities’ Board of Regents just approved a use our resources as well as present more in-depth ar- new mission statement for the University of Alaska ticles about the research underway at ARSC. These Fairbanks. It reads: The University of Alaska Fairbanks, changes are just part of ARSC’s commitment to help as the nation’s northernmost Land, Sea, and Space Grant our users achieve success with their computational and university and international research center, advances and visualization tasks. disseminates knowledge through creative teaching, More significantly, we provide our users with ex- research, and public service with an emphasis on Alaska, perts who can assist them in moving the North and its diverse peoples. This toward their goals. Just two examples Our unique position at dovetails well with our mission here of this appear in this issue of Chal- UAF allows us to at ARSC: to support high performance lenges—Roger Dargaville’s CO2 mod- computational research in science and eling and the animation created by Bob encourage northern engineering with an emphasis on the Andres displaying global consumption research, while supporting Arctic and the high latitudes. of fossil fuels since 1750. ARSC re- We are proud to be part of the search liaison Guy Robinson worked important Department of University of Alaska as well as the with Dargaville to identify the best Defense projects. Department of Defense High Perfor- platform for his code. Visualization mance Computing Modernization specialist Roger Edberg helped Andres develop Program. Our unique position at UAF allows us to effective graphical representations for a complex data encourage northern research while supporting set. The resulting visualization is displayed in the visu- important Department of Defense projects. alization gallery on page 8—a new feature in Challenges. On page 7 of this issue, you will find Staff Highlights, a new column to introduce our diverse staff. Here, you Frank Williams Director ARSC bits and bytes ❖ ARSC hosted two visualization access lab open houses searchers. Seminars included Managing Tape Based Ar- this spring as part of the UAF Engineering chives, by Gene Harano of the National Center for At- Department’s Engineering Week, and the UAF Col- mospheric Research; Computational Science and Technol- lege of Science, Engineering and Mathematics’ Science ogy Research at the California Institute of Technology by Potpourri. ARSC staff met with over 500 UAF stu- Jim Pool of the California Institute of Technology; The dents, faculty and community families. State of HPF, OpenMP and MPI, by Professor Barbara Chapman of the University of Houston; and Effective ❖ ARSC welcomed four new student lab assistants in Parallel Programming in Advanced ZPL, by Professor Larry March and April. Georgina Blamey—ocean sciences; Snyder of the University of Washington. This panel was Eric Guglielmo—chemistry and pre-med; Joseph chaired by ARSC fellow Bill Buzbee. Sheedy—physics; and Nathan Zierfuss—computer sci- ence. They will be on hand in the visualization access labs to help users with hardware and software issues. ❖ Ben Barton and Steve Munk joined the ARSC staff as systems analysts, to provide support of desktop office systems, and visualization systems, hardware and tools. Barton was the first computer art graduate at UAF, and Munk is currently pursuing a business degree with a minor in computer science. ❖ ARSC convened a technology panel at the center in December to present topics and discuss high- perfor- ARSC Technology Panel members (from left) Larry Snyder, Barbara mance computing with staff and interested UAF re- Chapman, Jim Pool, Barbara Horner-Miller, Bill Buzbee and Gene Harano. Photo by LJ Evans. 3 Tracking The Effects of Global Climate Warming Dr. Roger Dargaville Researchers at the University of Alaska Fairbanks Dr. David McGuire Institute of Arctic Biology are combining forces James Long to study the carbon cycle in the Arctic in search UAF Institute of of a better understanding of the effects of global Arctic Biology warming. Historically, cold climate conditions have led to substantial carbon storage in Arctic soils, leading to the potential for large amounts of carbon to be released into the atmosphere if the climate warms. Temperatures have been warming in the Arctic for the past 30 years. This warming may be accelerated by a kind of vicious circle—warmer climate causing increased release of carbon dioxide (CO 2) and increased CO 2 caus- ing warmer weather. Scientists often look to the Arctic for indica- tions of the effects of global warming. The Arc- tic is a bellwether area for climate change and global warming. Climate models suggest that the impact of global warming will occur first and be most pronounced in the Arctic. “We’re trying to find the balance of the glo- Carbon Exchange (Gigato bal carbon cycle by determining where fossil fuel emissions are going and how the natural fluxes -0.4 from the land and ocean are behaving,” says Roger Dargaville, a postdoctoral researcher, working with scientist David McGuire at the In- stitute of Arctic Biology. Climate change isn’t caused solely by humans. Researchers are using a variety of processes to solve the global climate change puzzle. And and current data to map the ecological processes of the carbon cycle. Dargaville uses his Natural fluctuations in ocean and atmospheric Together, these pieces of information can be tested against each other to bring scientists c dynamics have generated warm and cold climate the model created by Dargaville. trends for millions of years. Although the evi- dence seems clear that increased use of fossil fuels by humans is having unprecedented effects on current global climate, it is difficult for scientists to predict how these changes will The code that Dargaville runs models CO 2 con- affect terrestrial carbon storage in the future. centration in the atmosphere. The researcher inputs Dargaville and McGuire want to understand how data generated by weather forecasting computers and the climate changes resulting from fossil fuel uses information about wind direction and intensity emissions are affecting the terrestrial carbon to drive the CO 2 through the atmosphere. The model exchange with the atmosphere. This understanding is runs for the entire globe at horizontal resolutions of important for predicting future trends in climate. 64 by 64 and 24 vertical levels. By running 16 time Dargaville is using “inversion” modeling with steps in each day from 1978 to 1995, the computer National Oceanic Atmospheric Administration must solve the transport equations 5.5 million times atmospheric CO 2 data and Robert Andres’ fossil per model-run. Dealing with this quantity of data fuel data set (see Challenges back cover) to esti- requires the supercomputers and mass storage systems mate the terrestrial and oceanic carbon exchange at ARSC. There are many variables in Dargaville’s with the atmosphere. model, which means the code is memory-intensive. “It’s similar to trying to locate a pollution In addition, the model produces masses of data which source through modeling,” says Dargaville. “If must be stored in the ARSC StorageTek Silo. you run a range of simulated sources and look at Da r g a v i l l e u s e s t h e A R S C J 9 3 2 t o r u n h i s the effects, you can find which simulation models. He then compares his results with those of matches the effects you’re getting in real life— Dave McGuire, also at the Institute of Arctic Biology. which will show you the source.” McGuire uses a PC to conduct forward-model simula- 4 time step can be made available to other mod- els that could then give us feedback for the cal- culation of the next time step.” McGuire’s model takes into account the be- havior of both the vegetation and soil in the ecosystem under study. Called the Terrestrial Ecosystem Model (TEM), this model simulates carbon uptake by the vegetation and carbon re l e a s e f r o m t h e s o i l b y d e c o m p o s i t i o n , creating a good data set to test against Dargaville’s results. It runs in an historical sense, and can be checked against Dargaville’s inversion to reduce uncer tainties in both scientists’ results. The TEM examines the details of the biogeochemistry in the ecosys- tem, while Dargaville’s model estimates the larger scale CO 2 exchange with the atmosphere. “We mainly compare the estimates of total carbon exchange going on between the terres- trial ecosystems and the atmosphere with onnes Carbon Per Year) Dargaville’s model,” says McGuire. “The units are typically grams of carbon per meter squared 0.0 +0.3 per month for a particular region of interest.” Over large areas the fluxes of CO 2 can add up to millions of tons of this greenhouse gas. McGuire examines the response of a variety of variables in his model, including fire, land dres looks at historical data and its spread across the planet. McGuire uses historical use and CO 2 concentration. There are many storical data to create models to predict activity in the carbon cycle into the future. closer to accurate predictions. Above, a map charts carbon exchange as measured by environmental factors that can influence the c a r b o n c yc l e , f i r e b e i n g a n e s p e c i a l l y important one. If the climate warms, fire events could increase and cause a net release of car- bon from terrestrial ecosystems. On the other hand, if the climate warms in the Arctic, more tions of the carbon exchange between the terrestrial bio- trees may grow, which in turn would tend to s p h e re a n d t h e a t m o s p h e re . Du e t o t h e store more carbon. nature of the biosphere, each model grid can be solved The business of climate prediction is very individually, making the memory requirements much less uncertain. “You need to know what the cur- than those of Dargaville’s transport model, and allowing rent status of the carbon cycle is before you can McGuire to run his model on make predictions,” says a s t a n d a rd P C . Howe v e r, Dargaville. “This is what master’s degree candidate Jim We’re trying to find the balance of m a k e s this kind of Long ported the code to research so difficult.” ARSC’s T3E, allowing more the global carbon cycle by Results from Dargaville’s simulations to be run faster determining where fossil fuel and McGuire’s work will be use- and for longer time periods. emissions are going and how the ful not only in understanding “The original model climate change in the would visit a grid cell and per- natural fluxes from the land and Arctic, but in predicting climate form all time step calculations ocean are behaving. fluctuations over the rest of the before proceeding to the next world and determining policies grid cell to do the same,” says governing human use of fossil Long. “My restructured version visits each grid cell to fuels. For more information on this project, con- calculate one time step so the complete results of a single tact David McGuire at firstname.lastname@example.org. 5 ARSC Currents Communication: The Secret to ice Success polar desert/alpine tundra moist tundra Guy Robinson The key to the success of boreal woodlands ARSC any parallel computa- boreal forest Research Liaison/ tion is distributing both temperate forest MPP Specialist work and data between temperate grassland many processors but still permitting them to co- operate on the task at hand. As in real life, once the work has been di- vided between the avail- A map of Arctic vegetation types produced by Dave McGuire (see Challenges, page 4) adds to the information researchers able workers they must have to solve the puzzle of global warming. cooperate. In parallel computing this is achieved by the exchange of messages that contain either oceans, land, atmosphere and ice cover to create a the necessary information or a pointer to the location of coupled climate model. This kind of coupling of- this information in some global space. The complexity of fers the advantage of bringing together established the division of tasks varies greatly between different prob- proven code to allow tailoring of the different parts lems and determines what communication is necessary. to the specific needs of the research group involved. In embarrassingly- or naturally-parallel cases, there is However, this isn’t as simple as it sounds. Getting relatively little interaction between processors once the models to work together and exchange data in- tasks are distributed. In more advanced cases, where the volves a great deal of human communication re- problem is broken up into many small pieces (domain garding each individual model. decomposition), there may be a frequent need to exchange With all of this communication between com- data. One of the tasks of the programmer is similar to puters and models, it is important not to forget that of the director of a play or the manager of a sports information exchange between the scientists team—making sure the activities of a group work together themselves. Many of today’s models use data de- smoothly to achieve set goals. rived from other models to reduce the computa- The above principles have been known and successfully tional effort. An example of this is featured on page employed for the past two decades, leading to many ad- 2 in this issue of Challenges. In this research, near- vances in scientific and engineering research. By continu- shore regions of the ocean are being studied in ally refining their models, researchers have progressed to detail to obtain a better understanding of the study increasingly complex problems by building on the mixing processes. Parameters gathered from this work developed by others. The cover article of this Chal- research, from both numerical and experimental lenges features an example of how, activity, will be used inother models by combining a discrete event that need to simulate these mixing model of population dynamics As the problems being processes over wider areas and longer and a simulation of ocean currents studied become more time-scales—as might be considered in a single code, accurate predic- tion of responses to outside events complex, communication is in climate codes. prediction simulations and weather becomes possible. occurring at higher levels The above examples only touch on As the problems being stud- with many independent the importance of communication and ied become more complex, com- the differing levels at which it occurs munication is occurring at higher models exchanging data. between research scientists, between levels with many independent computers and between individual pro- models exchanging data. In such cessors on a single computer. In the fu- cases, several distinct models are actually combined to ture it is likely there will be more communication create what might be referred to as a supermodel—a com- and increased sharing of data and ideas as the scope bination of many smaller models encompassing multiple and multidisciplinary nature of research continues variables. An example might combine separate models of to expand. 6 Staff Highlights In the field of supercomputing, perhaps one of the most valuable people a researcher can meet is the person at the help desk. The user services staff at the Arctic Region Supercomputing Center works with users every day, an- swering questions from “How do I log-on?” to “How do I optimize my code?” Tom Baring, Derek Bastille and Shawn Houston share responsibility to help users out of programming binds and ensuring that all users can log on to the proper ma- chines. In addition, Houston maintains the internal and external web for the center by writing cgi scripts and port- ing documents. The user services staff (from left) Shawn Houston, Tom Baring and Derek Bastille, provide support for both T3E and J932 users. Photo by LJ Evans. “My hobby is computers,” says Houston, “So work is all play and no work to me. The best part about my job is that there are new problems to solve every day.” it would use multiple processors effectively,” says Problem solving is what the user services team does Baring. “We got a big speed-up just in time for her best. Baring, Bastille and Houston are constantly testing important demonstration.” their computing knowledge to help users make the most What is the strangest request that’s ever passed through of ARSC’s resources. user services? “It’s really rewarding when you discover the solution “I once got an urgent call: ‘We have a film crew to an especially complex problem,” says Bastille. In addi- stranded on Wrangell Island, Alaska. The weather’s get- tion to consulting, Bastille is in charge of maintaining ting worse, they’re almost out of food, we’re calling from the user database and creating new accounts. New Zealand, we found you on the web and we need In addition to staffing the help desk, Baring help organizing a helicopter rescue,’” says Baring. writes and edits the T3E Newsletter with colleague Although user services staff are not experts in heli- Guy Robinson. copter rescues, they’re always happy to help. They can be “I once tweaked a user’s tsunami code on the J90 so reached at (907) 474-5102 or email@example.com. Who we are and what we do... The Arctic Region Supercomputing Center supports ARSC staff research in science and engineering with an emphasis Specialists at ARSC provide expertise in visual- on high latitudes and the Arctic. ARSC is a part of the ization, massively parallel supercomputing, storage, vector parallel supercomputing, code optimization DoD High Performance Computing Modernization and networking. Program and the University of Alaska Fairbanks. The close relationship of ARSC with the University Hardware of Alaska extends the center’s expertise to include ARSC operates a 272 processor, 450 MHz CRAY T3E specialty areas of the university’s research institutions. System named Yukon with 68 gigabytes of distributed These include ice, ocean, and atmospheric coupled mod- memory and 522 gigabytes of disk storage. The T3E eling; regional climate modeling; global climate change; System has a peak potential parallel performance of over permafrost, hydrology and arctic engineering; magneto- 230 gigaflops. spheric, ionospheric and upper atmospheric physics; ARSC also operates a 12-processor CRAY J932se vec- petroleum and mineral engineering; and arctic biology. tor parallel supercomputer with eight gigabytes (one gigaword) of shared memory and 482 gigabytes of disk Communication storage. The J932, named Chilkoot, provides peak poten- Connectivity to ARSC is provided by an OC12 tial parallel performance of 2.4 gigaflops. (622Mbps) extension to the Seattle Pacific/Northwest Visualization hardware includes numerous Silicon GigaPoP where direct peering provides access to the De- Graphics workstations, a Pyramid Systems ImmersaDesk fense Research and Engineering Network, Internet2’s and a professionally-equipped video editing studio. Abilene network, the vBNS network, and the commod- ARSC data storage resources include a ity Internet. StorageTek robotic tape silo with a capacity in ex- cess of 300 terabytes. 7 Visualization Gallery 1896 1995 Dr. Robert Andres Institute of Northern Engineering Dr. Roger Edberg Arctic Region Supercomputing Center 0 5 10+ 0 5 10+ Mass CO2 Mass CO2 (106 tonnes per 1-degree cell) (106 tonnes per 1-degree cell) Annual CO2 production from fossil fuels is mapped in an animation showing the spread of CO2 across the planet through time. Fossil fuel emissions emerge as a small patch over Europe in 1896 (left) and spread across the planet (right) by 1995. Such data is useful to both researchers and world leaders when making decisions about regulations surrounding global warming issues. Historical Use of Fossil Fuels Many researchers believe humans are the largest contributors to global warming. Through the burning of fossil fuels to heat homes, run automobiles or produce cement, humans add to the amount of carbon dioxide released into the atmosphere. In order to create fair laws and regulations about emissions, governments around the world need reliable data showing how much carbon dioxide is being released and from which areas of the world. The visualizations above were produced by Bob Andres of the UAF Institute of Northern Engineering, with the help of ARSC visualization specialist Roger Edberg, to produce data for the Oak Ridge National Laboratory database. The labora- tory provides unbiased data on global warming for researchers and government officials around the world. The visualization is a 1-degree latitude by 1-degree longitude grid showing carbon dioxide mass released around the globe based on fossil fuel use. The information from each year from 1750 through 1997 is strung together into an animation, showing the growth of the world’s fossil fuel use through the Industrial Revolution and into today. Arctic Region Supercomputing Center Non-Profit Organization University of Alaska Fairbanks U.S. Postage 910 Yukon Drive, Suite 106 P A I D P.O. Box 756020 Fairbanks, A l a s k a Fairbanks, Alaska 99775-6020 Permit No. 2 Address Service Requested