7/28/2019 2169-3358-2001-2-865 gnome

1/8

GENER AL NOAA OIL MODELING ENVIRONM ENT (GNO ME) :A NEW SPILL TRAJECTOR Y MODEL

C. J. Beegle-KrauseHazardous M aterials Assessment Division, Office of Response and RestorationNational Ocean Service, National Oceanic and Atmospheric Administration7600 Sand Point Way, NE.Seattle W ashington 98115-6349

ABSTRACT: The General NOAA Oil Modeling Environment(GNOM E) is a standard Eulerian/Lagrangian spill-trajectorymodel designed to meet the needs of planners and expertresponders through three different user modes: Standard, GISOutput, and Diagnostic. Spills are modeled by LagrangianElements (LEs or splots) within continuous flow fields. GNOMEsupports the National Oceanic and Atmospheric Administration(NOAA)/Hazardous Materials Response Division (HAZMAT)standard for Best Guess and Minimum Regret trajectories byproviding information about where the spill is most likely to go(Best Guess solution) and the uncertainty bound (MinimumRegret solution).The public, including spill responders, industry , and students,can use GNOME in Standard or GIS Output mode to preparespill scenario-related products and for intuition building. TheseGNOME modes require a Location File that contains a regionaltrajectory model with a Mini-Expert System to aid in setting upthe model. The Mini-Expert System sets up the trajectory modelbased on user input via dialog boxes. Information sources alsoare provided to help users answer the dialog questions.Responders can use GNOME 's Diagnostic mode to quickly setup custom trajectory models for any area, as HAZMAT doesduring spill response . GNOME's Diagnostic model can acceptcirculation patterns from any hydrodynamic model (from two-dimensional steady-state to three-dimensional time-dependentmodels) with proper formatting.GNOME allows all users to save their work in files and createQuickTime movies. In GIS Output and Diagnostic modes, userscan export the model results to GNOME Analyst to convert thedata from LEs o r splots to oil-concentration contours. Both thesplots and contours can be exported to a GIS system (HAZMATprovides an ArcView extension). HAZMAT presently is creatingLocation Files for U.S. Coast Guard and NOAA priority locationswith a design philosophy to allow users significant control overthe model setup without requiring extensive spill modelingexperience. GNOME, all Location Files, and documentation areavailable for download from NOAA 's Office of Response andRestoration W eb site1 under Aids for Oil Spill Responders.

I ntroduc t ionThe objective of this paper is to acquaint the reader with the

General NOAA Oil Modeling Environment (GNOME)2 and goals

for future GNOME development and Location Files. GNOME isthe latest spill-trajectory model developed by the HazardousMaterials Response Division (HAZMAT), Office of Responseand Restoration1, National Oceanic and Atmospheric A dministration (NOAA). GNOME is a multipurpose trajectory model foruse by both experts and the public via different user modes.GNOME's Diagnostic mode is a spill response trajectorymodel. HAZMAT is aware that many people who are nottrajectory analysts are interested in running spill trajectories forintuition building, drill scenarios, and answering "what ifquestions. For these users, HAZMAT developed Location Files,which are resource files for use with GNOME in Standard or GISOutput modes. For oil spill-related planning, HAZMAT hasdeveloped a separate applicationthe Trajectory AnalysisPlanner (TAP)to summarize thousands of individualtrajectories run in GNOME.Until the basic GNOME algorithms had been tested fully,GNOME was run jointly with HAZMAT's On-Scene Spill Model(OSSM) during all HAZMAT spill responses. The mostnoticeable difference between GNOME and OSSM is thatGNOME has a graphic user interface (GUI, point-and-click),whereas OSSM has a command line interface. Although GN OMEis now used for trajectory support, GNOME's development teamhas prioritized the addition of capabilities not available in OSSM,such as three-dimensional (3-D) trajectories and currents. OSSMis still used for responses that require specialized tools not yetfound in GNOME, such as statistical trajectories.Location Files contain all the necessary files an expert woulduse to set up a trajectory model for a particular region, with theaddition of a Mini-Expert System. The Mini-Expert Systemconsists of a short series of dialogs with questions for the user toanswer and information sources that the user can find on theInternet. Location Files are not meant to be used for spillresponse, since they represent climatological conditions that maysignificantly differ from actual conditions on any particular day.Although a trajectory analysis expert may use some of the files oralgorithms from a Location File during a response, the expertwould adjust the GNOME model to represent conditions duringthe spill using GNOME's Diagnostic mode tools. A summary ofthe differences between the Diagnostic and Standard/GIS Outputmodes is given in Table 1. HAZMAT is available 24 hours a dayfor spill response, including providing trajectory analysisexpertise; Location Files contain information on how to reportspills and contact HAZMAT for spill response.

86 5

7/28/2019 2169-3358-2001-2-865 gnome

2/8

8 66 2 00 1 I N T E R N A T I O N A L O I L S P IL L C O N F E R E N C E

Table 1. Information flow from the user through GNOMEand the available output options.File InputStandard/GIS DiagnosticLocation FileResponses to LocationFile dialogsSpill information

Length and time scale of problemKnowledge of regional physics anduncertaintyMapCurrents (hydrodynamicmodels)Wind forecast(Optional) location fileSpill information

Location save file (usew/ Location File)Quick Time movieGIS output (MOSSformat; GIS outputonly)GNOME analyst files(oil concentrationcontours; GIS outputonly)

Diagnostic save fileQuick Time movieGIS output file (MOSS format)

GNOME analyst files (oilconcentration contours)Color/black-and-white pictureprintoutNote: In Standard and GIS-Output modes, a Location File assiststhe user in setting up the model via a Mini-Expert System(Wizard) that converts the input into model parameters and setup.In Diagnostic Mode, the user is responsible for all model fieldsand parameters; thus, the user is expected to have sufficientocanographie and spill-modeling knowledge to create and usethe model appropriately.GNOME supports the NOAA standard for trajectory output byproviding both the Best Guess and Minimum Regret solutions(Gait, 1998). The Best Guess solution is the trajectory created byassuming that all model inputs are correct. The Minimum Regretsolution is a statistical compilation of trajectories that samplespossible forecast errors in all of the model inputs (wind, currents,horizontal mixing, etc.). This Minimum Regret spill distributionis used to create an uncertainty bound on trajectory forecasts.Because quantitative confidence limits often are unavailable formodel input fields, the GNOME uncertainty bound is generallyconsidered a 90% confidence limit based on experience.Responders can use this information to examine potential spilltrajectories and evaluate the protection of highly valuable orvulnerable resources that merit response even when theprobability of oil contact is low.

G NO ME's traj ec tory mode lGNOME uses the standard Eulerian/Lagrangian approach tospill modeling with the regional physics simulated as Eulerian(continuous) fields within which the oil spill's LagrangianElements (LEs or spill dots called splots) move. The user watchesthe spill evolve over time in the model as the splots move withthe currents and winds. As with all trajectory models, the usermust have an idea of the temporal and length scales involved toappropriately set up the model. The spilled product, amountspilled, and local dynamics will influence the persistence of theproduct and how far it will travel. For example, during spillresponse, the NOAA/HAZMAT trajectory modeling team firstdetermines their view of the temporal and spatial scales of theresponse and sets up the trajectory model accordingly. GNOME

is calibrated to overflight observations twice a day, and forecastsare made for up to 3 days (the temporal boundary betweenweather forecasts and climatology). When creating Location Filesfor GNOME, the development team considers the types andamounts of products that could potentially spill in the area and theregional physics that must be simulated.GNOME is written using the latest object-oriented programming methodologies in the C++ programming language. In thisway, different components of oil spill simulation and physics areall self-contained objects within the application, so they can beadded, modified, or deleted as self-contained units. The physicsin the trajectory simulation is broken apart into "mover" objectsbased on the assumption of linear superposition of mechanics thatmove LEs. This means that during each time step of a trajectorysimulation, each LE has a known starting position and querieseach active mover to find out how far and in what direction theLE would move in that time step. Then these steps are added in avector sum to give the resulting direction and distance, and theLE is moved to the new location.The basic movers within GNOM Ecurrents (with and withouttides), horizontal mixing, and windare explained below in moredetail.Currents. GNOME accepts current data output from hydrodynamic models in a number of different forms. For current patterns(2-D steady state), GNOME has separate formats for finite-element modelssuch as HAZMAT's CATS (Current Analysisfor Trajectories) modeland for a regular rectangular-grid,finite-difference model. For 2-D and 3-D time-dependentcurrents, GNOME has only a custom data format, although morecommon public-domain formats are being investigated, such asnetCDF (network Common Data Format3).Current patterns can be made time dependent in several wayswithin GNOME. The current pattern is multiplied by a time seriesrepresenting the time-dependent physics of interest, such as tidesor changes in river flow rates. A tidal current time series from thenearest tide station can simulate tidal currents. The time seriescan either be entered as a time series of currents or the tidal

harmonic coefficients. Wind-driven currents can be simulatedwith current patterns through GNOME's Wind Componentmover. A single current pattern can be tied to a particularcomponent of the wind time series vectors, or two patterns(generated by orthogonal wind stress directions) can be tied totwo orthogonal components of the wind time series. HAZMATuses linear current patterns when using the Wind Componentmover with two patterns in Location Files.Horizontal mixing. The ocanographie processes that spreadspills horizontally are simulated in GNOME by a random walkafter Csanady (1 973). The user inputs a diffusion coefficient,which is used to calculate random step lengths in the x and ydirections from a uniform distribution. A uniform distributionwas chosen over a standard normal distribution (Apstol, 1969)because it is more conservative for estimating the extent an oilspill will spread.Wind. Currently, wind is simulated only as a spatially constantfield with optional temporal variability. Spatially variable windsare planned but have not been implemented at this time. The usermay input wind data by typing, pointing-and-clicking on a windtarget, or reading in a file. Wind data can be saved as a file to beused with other HAZMAT applications, such as ADIOS(Automated Data Inquiry for Oil Spills) oil weathering model.

Unce rtainty in mode l/data param etersTo model the Minimum Regret solution, GNOME requiresinformation on the amount of uncertainty associated with each

7/28/2019 2169-3358-2001-2-865 gnome

3/8

M O D E L I N G 8 6 7

mover. For example, the horizontal mixing estimate may be offby a factor of 2, or the currents may err by 30% in magnitude.GNOME uses these data to simultaneously run 1,000 separate,single LE spills to sample the uncertainty. Each of theseUncertainty LEs samples a different portion of the uncertaintyspace. One may have winds slightly faster and more to the right, asmaller diffusion, and slightly slower currents. Another mighthave slower winds, a much larger diffusion, and faster currentsslightly to the left. The trajectories of these Uncertainty LEs mapthe domain for the Minimum Regret solution and are used in theGNOME Analyst application to calculate the uncertaintyboundary for the trajectory.

Simulating dif ferent spi l ls and spi l l productsSpill trajectory models usually combine surface physics effectscaused by windssuch as wave stress, wave compression, Stokesdrift (Lighthill, 1978), dispersion, over-washing, surface drift,and Langmuir circulation (Farmer and Li, 1994)into a singlemodel parameter called the Wind Drift Factor (Gait, 1985).

GNOME's default values for the Wind Drift Factor range from1-4%, although any values from 0 -10 0% may be used. Thisallows the user to simulate subsurface tarballs or other pollutantsthat would be unaffected by the wind, or other objects that havegreater wind effects than oil, such as drifting ships, large floatingdebris, or jetsam.GNOME can simulate different spill distributions (point, line,2-D, instantaneous, or leaking over time) through the GUI. Spilldistributions can vary in space (point, line, or sprayeddistributions) and in time (point and line sources simulateinstantaneous spills or spills over time). GNOME uses sixproduc t types and on e non weather ing type for spill simulation s:gasoline, kerosene/jet fuels, diesel, fuel oil No. 4, medium crude,and fuel oil No. 6. The user interface includes spill information,simple weathering and shoreline-contact mass balance, and totalmass balance when more than one spill is included. GNOMEtracks the mass balance of the floating, beached, and evaporatedoil, as well as any oil that has left the model domain. Theevaporation algorithm in GNOME is very simple; useHAZMAT's ADIOS 2 model for a better estimate of oilweathering.

T raj ec tory p roduc tsOnce you have run a spill scenario in GNOME, you can createthe following trajectory pro ducts (see Figure 1 for example),some of which have the option of a single frame or hourly output: Printed Picture contains spill information from the modelrun. Save File allows users to save their work where they left

off: (.lfs) in Standard/GIS Output modes or (.sav) inDiagnostic Mode. QuickTime animation saves hourly pictures from themodel run and puts them into an animation file. NOAA Splot Files display standard trajectory output foruse in GIS systems. GNOME Splot Files display standard trajectory output foruse in GNOME Analyst. GNOME Analyst is theHAZMAT tool for converting splot distributions intocontours of relative oil concentration and an uncertaintybound. Other Support Files exported for other HAZMATsoftware applications include Shoreline Map for use in

GNOME Analyst and W ind Time Series for use in the oil-weathering model (ADIOS).

Location Fi lesLocation Files are resource files for the GNOME applicationthat contain all information pieces for the trajectory model(currents, tides, parameters) and a Mini-Expert System (theWizard) that acts as a facilitator in helping the user customize themodel setup for a specific spill/region. In this section, the authordiscusses the philosophy behind the Location File design,describes the user interface with Wizard and with the types ofsimulated physics, and gives examples of how to use LocationFiles. Currently, NOAA/HAZMAT is creating Location Filesbased on a U.S. Coast Guard priority list of 32 Marine SafetyOffices (MSOs), as needed for testing GNOME development andin collaboration with other federal branches such as the Navy andState Department. Location Files that scheduled for completionby IOSC 2001 are listed below:GNOME Location Files completed or in final stages (as of

10/15/2000):Apra Harbor, GuamCentral Long Island Sound, New York, and ConnecticutColumbia River Estuary, Oregon and WashingtonDelaware Bay, Delaware and New JerseyGal veston Bay, TexasKanoehe Bay, HawaiiMobile Bay, AlabamaPrince William Sound, AlaskaROPM E Sea Area (Persian Gulf)San Diego Bay, CaliforniaSan Juan, Puerto RicoSanta Barbara Channel, CaliforniaSoutheastern MediterraneanTampa Bay, FloridaGNOME Location Files expected by March 2001:Boston, MassachusettsJuneau and Glacier Bay, AlaskaLos Angeles/Long Beach, CaliforniaPortland, MaineSanta Barbara Channel and Santa Maria Basin, CaliforniaStrait of Juan de Fuca, WashingtonAs for any trajectory model, the first considerations in creatinga Location File for a particular area are the time and length scalesof the physics involved. Domains must represent the regionalphysics on a scale large enough to be useful and, at the sametime, balance the model's limitations and the local variability. Forinstance, a single Location File created for the entire Californiacoast would not be feasible because the number of questions forthe user to answer would be prohibitive, and the length scale forspills over a few days could not cover such a long distance. Nouser would want to research and answer questions irrelevant totheir area of interest (e.g., setting the Santa Barbara Channelcirculation pattern makes no sense for a 24-hour spill scenario offHumboldt Bay). Spill scenarios in Standard and GIS Outputmodes are limited to 5 days since the Location Files capture onlyciimatological conditions of a changing ocean.To understand the philosophy underlying the creation ofLocation Files, consider two sets of information (see Figure 2):1. The first set contains all of the regional physics that needto be simulated to create a useful Location File.2. The other set contains available data and informationabout the Location File region from sources outside themodel.

7/28/2019 2169-3358-2001-2-865 gnome

4/8

868 2001 INTERNATIONAL OIL SPILL CONFERENCE

*Model Mode: StandardEstimate for: 1315 12/23/98Prepared : 1621 12/07/98

Scenario Name:Prepared by:Contact Phone:

Drill 6CJ Krause

206 -526-6961This trajectory was created using climatological currents from a G N O M E Location File, and is unlikelyto represent conditions existing at any particular time at the depicted location. Use Location Files onlyto create spill scenarios for training or educational purposes, not for actual spill response.

Location File: Columbia River EstuaryUser SelectionsColumbia River Transport: : HighWind : Variable, 5 kt NNE at 1300 12/23/98Number of spills: 1

Black splots: Best Guess Red splots: UncertaintyTotal Mass Balance for Spills:xx % releasedxx % evaporatedxx% off mapxx% beached

Figure 1. Exam ple of G N O M E trajectory product.

7/28/2019 2169-3358-2001-2-865 gnome

5/8

MODELING 869

RegionalPhysics

Observational dataand information available

to the user

Regional physics are simulated throughparameterizations and modeled fields in theWizard when no user data is available tocustomize the trajectory model setup.

Where these two sets intersect, the Wizard isdesigned to setup the model based onuser input. Dialogs with Help topics are addedto the Location File so the user can passinformation to the Wizard. The Wizard thenconverts the user input values into modelsetup fields.

Location Files contain background informationand references so users can learn moreabout a specific area.

mxuCD

> >3

7/28/2019 2169-3358-2001-2-865 gnome

6/8

8 7 0 20 01 I N T E R N A T I O N A L O I L S P IL L C O N F E R E N C E

! Setting R iver R ow \You can choose an estimate of the total river flow or input flow values for theBonneville Darn and the W illam ette R iver at PortlandThe to tal flow of the Columbia River is primarily controlled by the amount ofw a t e r fowing over the Bonn evil le Dam, and down the W illamette liver Thetransport effects generally take 5-6 hrs to m ake the transit from P ortland toAstoria.The Columbia v er low s: I'll specify Bonn evil le D am & W illame tt e a }

Flow ov erBonneville Dam W illamette Riverat P ortlandkefs 3 kefs 3Jay (1984) considered 200 kefs for Bonneville D am and 90 kefs for the W illametteRiver at Portland the boundaries between low and high flow conditions.

Finding Row Data c t B a c k 3 1 Next Figure 3 . G N O M E dialog box illustrating user selections for setting the river flow rate. The user can select statistical high,medium, or low flow ra tes, or they can find real-time flow values for the main tributaries and the W izard will calculate thetransport rate.

Users also can set up drills or tabletop exercises in GNOME byrunning the already-planned scenario and then making small,iterative adjustments to the winds, timing of the spill, or otherlocal physics. This exercise of figuring out how to adjust themodel to achieve the desired results builds the user's intuitiveknowledge about how different forces affect trajectories and howsmall changes (e.g., shifts in wind speed or direction) cansometimes dramatically influence a trajectory. Trajectorymodelers use this type of physical intuition when adjusting themodel and when making spill hindcasts to match overflightobservations.As mentioned earlier. Location Files are not appropriate for useduring spill response. Location Files deal with climatologicalconditions, while the real ocean on a given day can be highlyvariable. An expert should do trajectory forecasting, andHAZMAT is available 24 hours a day to provide expertise duringspill response throughout the United States and its territories.

Conc lus ionsGNOME is a trajectory model for use by people with a widerange of spill experience, from the interested public to

experienced trajectory analysts. This user diversity is accomplished through three different user modes within GNOME, andancillary Location Files aid in setting up a trajectory model forthe user's particular region. Location Files are region-specifictrajectory models containing all the necessary hydrodynamic dataand a Mini-Expert System (Wizard) to help users set up GNOMEfor their situation. The Wizard queries the user about localconditionswith suggestions on where to find the answers andhow to interpret themand then sets up the model accordingly.Experienced trajectory modelers can use GNOME with their ownhydrodynamic models to create custom trajectory models. Development is moving ahead toward extending GNOME's capabilitiesfrom 2-D time-dependent currents to 3-D time-dependentcurrents. GNOME supports the NOAA standard for trajectory

output of both the Best Guess and Minimum Regret solutions forthe trajectory forecast.GNOME, its User's Manual for Standard and GIS OutputModes, the Location Files, User's Guides, and Example Problemsare all available from the Office of Response and RestorationWeb site4.

A c k n o w l e d g e m e n t sThe author would like to thank the International Oil SpillConference reviewers and Nancy Peacock (HAZMAT technicaleditor) for very thorough comments during manuscript revisionand review.

BiographyC. J. Beegle-Krause has been a physical oceanographer withNOAA/HAZMAT for 3 years, previously was employed inoceanography education (primary through undergraduate level)and oceanography consulting. She earned her B.S. at theCalifornia Institute of Technology, her M.S. in Physical

Oceanography from the University of Alaska-Fairbanks, and herPh.D. in Physical Oceanography from the University ofWashington. Dr. Beegle-Krause's research focus is 3-D modelingof chemical transport in fluids and oil-spill trajectory modeling.She currently serves on the HAZMAT response team (both on-and off-site), is the GNOME project coordinator and primarydesigner and developer of GNOME Location Files, and conductstraining in trajectory modeling.

References1. Apstol, T.M. 1969. Calculus: Volume II. 2nd ed. JohnWiley & Sons, New York, New York.

7/28/2019 2169-3358-2001-2-865 gnome

7/8

M O D E L I N G 8 7 1

2. Csanady, G.T. 1973. Turbulent Diffusion in the Environment. D. Reidel Publishing Company, Boston, MA.3. Farmer, D., and M. Li. 1994. Oil Dispersion by Turbulence and Coherent Circulations. Ocean Engineering.21(6):575-586.4. Galt, J.A. 1985. Ocanographie Factors Affecting the Predictability of Drifting Objects at Sea. Proceedings, Workshop on the Fate and Impact of Marine Debris, Honolulu,Hawaii (November 1984), R.S. Shomura and H.O.Yoshida, eds. Technical Memo NOAA-TM-NMFS-

1 Available on-line at http://response.restoration.noaa.gov .2 Available on-line athttp://response.restoration .noaa.gov/sofiware/gnome/gnome. html.3 A good discussion is available on-line athttp://www . unidata. ucar. edu/cgi-in/mfs/01/packages/netcdf/index. html.4 Available on-line athttp://response.restoration.noaa.gov/softw are/gnome/gnome, html.

SWFC-54. NMFS Southwest Fisheries Science Center, LaJolla, CA.5. Galt, J.A. 1998. Uncertainty Analysis Related to Oil SpillModeling. Spill Sei. Technol. 4(4):231-238.6. Jay, D. 1984. Final Report on the Circulation Work Unitof the Columbia River Estuary Data Development Program: Circulatory Processes in the Columbia River Estuary. Geophysics Program, University of Washington, Seattle, WA.7. Lighthill, J. 1978. W aves in Fluids. Press Syndicate, University of Cambridge, New Y ork, New York.

http://response.restoration.noaa.gov/http://response.restoration/http://response.restoration/http://noaa.gov/sofiware/gnome/gnomehttp://noaa.gov/sofiware/gnome/gnomehttp://www/http://www/http://response.restoration.noaa.gov/software/http://response.restoration.noaa.gov/software/http://www/http://noaa.gov/sofiware/gnome/gnomehttp://response.restoration/http://response.restoration.noaa.gov/

7/28/2019 2169-3358-2001-2-865 gnome

8/8