Workshop on Data Assimilation in Support of Coastal Ocean Observing Systems

PROGRAM

Cooperative Institute for Oceanographic Satellite Studies (CIOSS) College of Oceanic and Atmospheric Sciences

Oregon State University, Corvallis, OR April 3-5, 2007

1

2

Emerging coastal ocean observing systems (COOS) will provide oceanographers with comprehensive long-term time-series observations, facilitating new scientific studies of the ocean physics, biology, chemistry, etc. on the regional (meso-) and coastal (submeso-) scales. Even with the new observing systems, ocean measurements will remain sparse in space and time. In particular, satellite-derived oceanic information (e.g., SSH and SST) has been limited in shelf areas because of insufficient spatial and temporal resolution, clouds, and problems deciphering the electromagnetic signal in the 25-50 km zone near coast. Near real-time coastal ocean models should help improve the utility of observations, providing time- and space-continuous information on the origin and evolution of dynamical structures apparent in the observations. Data assimilation (DA) in the circulation models will provide new tools for data mapping and synthesis, oceanographic analysis, and for designing optimal observational arrays. DA will be used to identify dominant error sources in the models and verify statistical hypotheses about the ocean. Under the sponsorship of the NOAA Cooperative Institute for Ocean Satellite Studies (CIOSS), this workshop is organized to bring together experts in coastal ocean modeling, data assimilation, and satellite data analysis, with the charge to assess the present status of modeling and DA in the coastal ocean and to identify directions for future research. In particular, we will review the theory and implementation of advanced DA methods, exchange information on DA activities in recently funded NOPP-CODAE projects, and discuss opportunities for assimilation of satellite (versus other available) observations in coastal ocean models. The program includes invited talks, contributed posters, and discussions. Organizing committee: Alexander Kurapov: kurapov@coas.oregonstate.edu, 541-737-2865 John S. Allen: jallen@coas.oregonstate.edu P. Ted Strub: tstrub@coas.oregonstate.edu

3

Table of Contents: Participant List 4

Agenda with Oral Presentation Titles 6

List of Poster Titles 9

Oral Presentation Abstracts (in the order given by the agenda) 10

Poster Abstracts (alphabetical order) 22

4

Participant List Name Affiliation Allen, John Oregon State University Arango, Hernan Rutgers U. Barth, Alexander U. South Florida Barth, Jack Oregon State University Bayler, Eric NESDIS-NOAA Beardsley, Robert WHOI Bennett, Andrew Oregon State University Breitlow, Karen Oregon State University Broquet, Gregoire UC Santa Cruz Chassignet, Eric FSU Chua, Boon Oregon State University Cornuelle, Bruce UCSD-SIO Davis, Curt Oregon State University Di Lorenzo, Emanuele Georgia IT Edwards, Chris UC Santa Cruz Egbert, Gary Oregon State University Fiadeiro, Manuel ONR Fiechter, Jerome UC Santa Cruz Foreman, Mike IOS - Canada Frolov, Sergey OHSU Halliwell, George U. Miami Harper, Scott ONR Harrison, Cheryl UC Santa Cruz Hermann, Albert PMEL Herring, H. James Dynalysis of Princeton Hoteit, Ibrahim UCSD-SIO Kim, Sangil Oregon State University Kindle, John NRL Kurapov, Alexander Oregon State University Lermusiaux, Pierre MIT Li, Zhijin (Gene) JPL Lozano, Carlos NOAA/NCEP McWilliams, James UCLA Miller, Robert Oregon State University Mooers, Chris U. Miami Mosca, Cesare Georgia IT Nechaev, Dmitri U. Southern Mississippi Nerger, Lars NASA Ngodock, Hans U. Southern Mississippi

5

O’Brien, James Florida State U. Özkan-Haller, Tuba Oregon State University Panteleev, Gleb U. Alaska, Fairbanks Plant, Nathaniel ONR Powell, Brian UC Santa Cruz Samelson, Roger Oregon State University Schofield, Oscar Rutgers U. Shulman, Igor NRL Smedstad, Ole Martin Planning Systems, Inc. Smith, Scott NRL Snyder, Chris NCAR Spitz, Yvette Oregon State University Springer, Scott Oregon State University Vandehey, Amy Oregon State University Veneziani, Milena UC Santa Cruz Vernier, Guillaume U. Northern Carolina Wei, Eugene NOAA-NOS Wilkin, John Rutgers U. Wilson, Greg Oregon State University Zaron, Ed Portland State U. Zavala-Garay, Javier Rutgers U. Zhang, Zepu U. Chicago

Agenda - Data Assimilation in Support of Coastal Ocean Observing SystemsApril 3 - 5, 2007

LaSells Stewart Center, Corvallis, OR

4/03/2007 Tuesday: Agricultural Science Room

8:00 Continental Breakfast

8:25 Welcome, logistics

Special Morning Session: Overview and DA activities in recently funded NOPP-CODAE projects

8:30 C. Edwards, UC Santa Cruz: The Central California Coast CODAE Project9:00 Discussion

9:10 G. Halliwell, U. of Miami: Impact of Initial and Boundary Conditions Provided by Data-Assimilative Ocean Hindcasts on Nested Simulations of the Florida Coastal Ocean

9:40 Discussion

9:50 R. Samelson, OSU: Boundary Conditions, Data Assimilation, and Predictability in Coastal Ocean Flows10:20 Discussion

10:30 Coffee

Session: DA Theory and Applications I

10:50 J. Kindle, NRL: Global real-time modeling using NCOM and HYCOM in support of GODAE and Regional/Coastal modeling efforts

11:20 Discussion

11:30 P. Lermusiaux, MIT: Adaptive Data Assimilation and Multi-Model Fusion12:00 Discussion

12:10 Lunch

13:00 C. Snyder, NCAR: Ensemble Filtering for the Atmospheric Mesoscale13:30 Discussion

13:40 B. Cornuelle, Scripps, UCSD: Strong Adjoint Sensitivities in Tropical Eddy-Permitting Variational Data Assimilation

14:10 Discussion

14:20 H. Ngodock, NRL: The representer method with nonlinear models: to cycle or not to cycle?14:50 Discussion

15:00 Break (coffee, refreshments, cookies)

Poster Session: First Interstate Bank RoomEmergent break-out meetings: Weyerhauser Room available

17:00 Reception - hosted by CIOSS 6

4/04/2007 Wednesday: Agricultural Science Room

8:00 Continental Breakfast

Special Morning Session: New topics in coastal ocean modeling

8:30 R. Beardsley, WHOI: An Unstructured Grid, Finite-Volume Coastal Ocean Model (FVCOM) System: Validations and Applications

9:00 Discussion

9:10 J. McWilliams, UCLA: The Submesoscale Dynamical Transition in the Ocean9:40 Discussion

9:50 Y. Spitz, OSU: Overview of Data Assimilation in Ecosystem Modeling10:20 Discussion

10:30 Coffee

Session: DA Theory and Applications II

10:50 E. Bayler, JCSDA: Satellite Ocean Data Assimilation at the Joint Center for Satellite Data Assimilation and NOAA

11:20 Discussion

11:30 E. Wei, J. Herring, NOS: Operational Estuarine and Coastal Forecast Hydrodynamic Modeling and Development in the National Ocean Service

12:00 Discussion

12:10 Lunch

13:00 J. Wilkin, Rutgers U.: Predictability of Mesoscale Variability in the East Australia Current System given Strong Constraint Data Assimilation

13:30 Discussion

13:40 E. Di Lorenzo, Georgia IT: Weak constraint 4D variational data assimilation in the inverse Regional Ocean Modeling System (ROMS)

14:10 Discussion

14:20 G. Panteleev, U. of Alaska: Hindcast and Reanalysis of the Circulation in the Bering and Chukchi Seas14:50 Discussion

15:00 Break (coffee, refreshments, cookies)

Poster Session: First Interstate Bank RoomEmergent break-out meetings: Weyerhauser Room available

17:00 Adjourn

7

4/05/2007 Thursday: Agricultural Science Room

8:00 Continental Breakfast

8:30 O. Schofield, J. Wilkin, Rutgers U.: Evolution of the Mid-Atlantic Coupled Observation and Modeling System

9:00 Discussion

9:10 B. Powell, UC Santa Cruz: Real-Time Data Assimilation and Ensemble Prediction System in the Intra-Americas Sea

9:40 Discussion

9:50 J. Barth, OSU: Linking Observations and Modeling in Coastal Ocean Observing Systems10:20 Discussion

10:30 Coffee

10:50 Plenary session discussion (workshop report)

12:00 Adjourn

8

List of Poster Titles: Barth, A.: Ensemble-based assimilation of HF-Radar currents in a West Florida Shelf ROMS model nested in HYCOM Foreman, M.G.G., P.F. Cummins, and J.Y. Cherniawsky: An Assimilating Tidal Model for the Bering Sea Frolov, S., A. Baptista, T. Leen, Z. Lu, R. van der Merwe: Data assimilation in the Columbia River estuary and plume: results of hindcast, forecast, and observation-optimization experiments Hoteit, I., B. Cornuelle, A. Koehl, P. Heimbach: S4DVAR assimilation of HF radar radial surface currents Kim, S., R. Samelson and C. Snyder: Predictability of the Oregon coast by the relative entropy Kurapov, A. L., G. D. Egbert, J. S. Allen, R. N. Miller and S. Erofeeva: Representer-based variational data assimilation in nonlinear coastal ocean models Li, Z., Y. Chao, J. Farrara, X. Wang, J. C. McWilliams, and K. Ide: A three-dimensional variational data assimilation system and observing system experiments for coastal oceans Nerger, L. and W. W. Gregg: State estimation in an ocean-biogeochemical model by assimilation of satellite ocean chlorophyll data Panteleev, G., D. Nechaev, V. Luchin, R. Woodgate, P. Stabeno: Hindcast and Reanalysis of the Circulation in the Bering and Chukchi Seas Shulman, I., C. Rowley, S. Anderson, J. Kindle, S. DeRada, J. Cummings and J. Doyle: Results from glider data assimilation experiments Smedstad, O. M., J. Cummings, A. J. Wallcraft, E. J. Metzger, H. E. Hurlburt, P. J. Hogan, E. P. Chassignet, G. R. Halliwell: Boundary conditions for coastal models from data assimilative HYbrid Coordinate Ocean Model (HYCOM) nowcast/forecast systems Smith, S., H. E. Ngodock and G. Jacobs: The Cycling Representer Method applied to the Navy Coastal Ocean Model Springer, S. R., J. S. Allen, G. D. Egbert, A. L. Kurapov, R. N. Miller, R. M. Samelson: The impact of boundary data on solutions of a nested grid model of the Oregon Coastal Transition Zone Vernieres, G., K. Ide and C. T. Jones: Lagrangian Data Assimilation in the Gulf of Mexico Zaron, E. D.: Baroclinic Tidal Generation in the Kauai Channel Inferred from HF-Radar Zavala-Garay, J., J. Wilkin and H. G. Arango: Predictability of Mesoscale Variability in the East Australia Current System given Strong Constraint Data Assimilation Zhang, Z., D. Beletsky, D. J. Schwab, M. L. Stein: Assimilation of Current Measurements Into a Circulation Model of Lake Michigan 9

Oral Presentation Abstracts: The Central California Coast CODAE Project C. Edwards(1), M. Veneziani(2), B. Powell(2), A. M. Moore(1), J. D. Doyle(3), D. Foley(4), P. Heimbach(5) and C. Wunsch(5) (1)Ocean Sciences Department, University of California, Santa Cruz, CA (2)Institute for Marine Sciences, University of California, Santa Cruz, CA (3)Naval Research Laboratory, Marine Meteorology Division, Monterey, CA (4)NOAA/NMFS, Pacific Fisheries Environmental Laboratory, Pacific Grove, CA (5)Department of Earth Atmospheric and Planetary Sciences, Massachusetts Institute of

Technology, Cambridge, MA cedwards@ucsc.edu The NOPP funded Coastal Ocean Data Assimilation Experiment was intended to understand the sensitivity of the coastal ocean circulation to large-scale fields provided by the Global Ocean Data Assimilation Experiment model output. The CODAE project along the central California coast uses the Regional Ocean Modeling System, driven by atmospheric forcing provided by the Navy's Coupled Ocean Atmosphere Mesoscale Prediction System. At the outer boundaries, which extend from Southern California to Washington, the model is forced by output from the ECCO-GODAE effort (Estimating the Climate and Circulation of the Oceans). We present results from this central California project, specifically how forward sensitivity experiments, adjoint sensitivity experiments, and 4-Dimensional Variational Data Assimilation experiments all provide information on the impact that the basin-scale information has on regional circulation models. Impact of Initial and Boundary Conditions Provided by Data-Assimilative Ocean Hindcasts on Nested Simulations of the Florida Coastal Ocean

G. R. Halliwell(1), V. Kourafalou(1), E. P. Chassignet(2), A. Barth(3), P. J. Hogan(4), O. M. Smedstad(5), J. A. Cummings(6), H. E. Hurlburt(4), R. H. Weisberg(3), L. K. Shay(1), and Ge Peng(1)

(1)MPO/RSMAS, University of Miami, Miami, FL (2)COAPS, Florida State University, Tallahassee, FL (3)College of Marine Science, University of South Florida, St. Petersburg, FL (4)Naval Research Laboratory, Stennis Space Center, MS (5)Planning Systems, Incorporated, Stennis Space Center, MS (6)Naval Research Laboratory, Monterey, CA ghalliwell@rsmas.miami.edu Sensitivity of nested coastal simulations along the Florida coast to initial and boundary conditions is documented as part of a NOPP-CODAE project. Two data assimilative nowcasts are evaluated in comparison to a free-running ocean model for use as the outer model within

10

which nested simulations are performed. The two nowcast products used are obtained from two generations of the Navy ocean nowcast-forecast system being developed at NRL: the original Atlantic optimum interpolation nowcast that assimilated SSH with Cooper-Haines vertical projection, and the new NCODA (Naval Coupled Ocean Data Assimilation) system. The NCODA system was run within a Gulf of Mexico domain as was the free-running simulation that serves as the baseline against which the assimilative products are compared. Free-running simulations driven by COAMPS atmospheric forcing were performed for 2004 and 2005 within a West Florida Shelf (WFS) and a South Florida coastal domain and nested within these three outer models. Multiple observational programs are providing the high-quality observations required for the evaluation effort. All nowcast products and nested simulations use the HYbrid Coordinate Ocean Model (HYCOM). The WFS simulations are run on the same curvilinear coordinate system used by modelers at the University of South Florida to facilitate future comparison between HYCOM and other models. Both the OI and NCODA fields are highly correlated over most of the WFS domain, specifically over most of the continental shelf and in the region directly influenced by the Loop Current, but are poorly correlated over the continental slope region. The nesting boundary in the curvilinear coordinate domain is located too far offshore for the boundary conditions to positively affect the simulated fields over the continental slope. Large sensitivity to initial and boundary conditions exists, but improvement in the simulations resulting from nesting within assimilative hindcasts is difficult to prove. A clear benefit is demonstrated in the South Florida domain where the Florida Current path and Tortugas gyre are correctly reproduced when nested within the NCODA hindcast but not when nested within the free-running model. This is critical if ocean models are to be used to study larval recruitment in the Florida Keys region. NRL nested the NCODA nowcast within a model-generated Atlantic Ocean climatology, which negatively impacted the accuracy with which Florida Current transport is simulated. Boundary Conditions, Data Assimilation, and Predictability in Coastal Ocean Flows R. M. Samelson(1), J. S. Allen(1), G. D. Egbert(1), A. Kurapov(1), R. Miller(1), J. Kindle(2), C. Snyder(3), S. Kim(1), S. Springer(1) and B. J. Choi(1) (1)College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR (2)Naval Research Laboratory, Stennis Space Center, MS (3)National Center for Atmospheric Research, Boulder, CO rsamelson@coas.oregonstate.edu Recent progress under a NOPP-CODAE project on boundary conditions, data assimilation, and predictability of coastal ocean flows is reviewed. This research focuses on numerical model simulations of ocean circulation in the Oregon coastal transition zone (CTZ), the region extending several hundred km offshore from the Oregon coast into the northern portion of the California Current System and the eastern interior North Pacific. The CTZ model is nested in the NRL NCOM-CCS model, and impact of these open boundary conditions on the CTZ model is examined. Atmospheric forcing is obtained from the NRL COAMPS product. Validation of the simulated coastal ocean circulation is provided by comparison with in-situ data sets from regional observational programs during 2000-2003 and with satellite and land-based remote-sensing observations. These comparisons have confirmed that the nested CTZ model yields

11

reasonable physical fields and provides a useful testbed for data assimilation studies. This work has benefited from coordination and collaboration with closely related CTZ modeling supported by NSF-GLOBEC and NOAA-CIOSS. Current work is directed toward determining the impact on the coastal ocean simulations of assimilating observations from satellite remote sensing, including sea-surface heights and temperatures, and from coastal HF radars. Predictability studies have focused on using the relative entropy to quantify predictive information content in ensembles of CTZ simulations, in a similar domain with simplified boundary and initial conditions and idealized forcing. Global real-time modeling using NCOM and HYCOM in support of GODAE and Regional/Coastal modeling efforts J. C. Kindle, S. deRada, I. Shulman and S. Cayulas Naval Research Laboratory, Stennis Space Center, MS kindle@nrlssc.navy.mil Two data-assimilative, real-time global ocean models are currently being run on the High Performance Computing (HPC) systems at the Naval Oceanographic Office. These are the global version of the Navy Coastal Ocean Model (NCOM) and the global Hybrid Coordinate Ocean Model (HYCOM). The NCOM model at 1/8th degree resolution is a U.S. Navy operational model, and the HYCOM at 1/12th degree resolution is currently being run in real-time by the Naval Research Laboratory in preparation for transition to the Naval Oceanographic Office for operational testing at the end of FY07. The NCOM model employs 40 levels: 19 sigma coordinate levels in the upper 137 m and 21 z-levels from 137m to 5500m. The real time system uses Navy Operational Global Atmospheric Prediction System (NOGAPS) 3-hourly wind stresses and heat fluxes. Operationally available sea-surface temperature (SST) and altimetry (SSH) data are incorporated into the Naval Oceanographic Office Modular Ocean Data Assimilation System (MODAS) and Navy Layered Ocean Model (NLOM) analyses with forecasts of SSH and SST. These surface fields are combined with the MODAS synthetic database to yield three-dimensional fields of temperature and salinity for assimilation into global NCOM. The global HYCOM model has 32 hybrid layers in the vertical with z-levels (pressure coordinates) near the surface to resolve the mixed layer, isopycnal layers in the stratified interior and terrain-following (σ) coordinates in shallow water. Three-hourly wind and thermal forcing is provided by the FNMOC NOGAPS fields. Ocean data are assimilated using the Coupled Ocean Data Assimilation (NCODA) system which QC’s the data and performs multi-variate optimal interpolation (MVOI) analyses using a model forecast as the first guess. The global system is run daily in real time with assimilation of satellite altimeter data (ENVISAT, GFO and JASON-1 provided via the Altimeter Data Fusion Center (ADFC) at NAVOCEANO), in situ and Multi-Channel Sea Surface Temperature (MCSST) data, and profiles of temperature and salinity. The current status of these models, details of their assimilation methodologies, and the availability of their output for use by regional and coastal models are discussed. The relative merit of using these global GODAE products as boundary and initial conditions for coastal models versus the use of a regional model forced by a higher resolution atmospheric model is a research issue being addressed by the NOPP project , “Boundary Conditions, Data Assimilation, and Predictability in Coastal Ocean Waters” and by collaborative NRL core projects. This

12

presentation briefly reports on efforts at NRL using our suite of coupled bio-physical models off the U.S West coast. The system presently includes a regional model (Navy Coastal Ocean Model—NCOM) at 9km resolution that receives boundary information from the operational global NCOM and provides boundary values to high resolution NCOM sub-nests for the Monterey Bay. The regional model and Monterey Bay nests include an embedded ecosystem model (COSINE) based on the formulation of Chai et al. (2002); these models are forced by high resolution surface fluxes from a regional COAMPS atmospheric model with resolution as fine as 3km. Adaptive Data Assimilation and Multi-Model Fusion P. F. J. Lermusiaux, O. G. Logutov and P. J. Haley Jr. Massachusetts Institute of Technology, Cambridge, MA pierrel@mit.edu Most data assimilation schemes used for realistic studies approximate generic principles for combining various sources of information. In part because of these approximations, the schemes involve parameters, options and heuristic algorithms whose specifics impact the assimilation results. In the Error Subspace Statistical Estimation system, such specifics can vary with each application and with users inputs. They are also modified with time, as a function of the regional dynamics, available data or other considerations. We review and illustrate several of these adaptable components, within the context of comprehensive real-time ocean observing and prediction systems. Even though much more research is needed, results indicate that error estimates, ensemble sizes, error subspace ranks, covariance tapering parameters and stochastic error models can be calibrated by quantitative adaptation to observational data. Adaptive estimation of model errors and correction of model biases is essential for data assimilation with single models but also for the quantitative comparison of competing models and combination of multiple models. We present a new Bayesian based fusion of multiple model estimates that involves the estimation of uncertainties in dynamical ocean models based on the comparison of past model estimates to measurements, and the subsequent fusion of multiple models in accord with their estimated uncertainties. We apply these results in the context of multi-model real-time ocean forecasting experiments (AOSN-II and MB06) in the Monterey Bay/California Current system. Ensemble filtering for the atmospheric mesoscale C. Snyder National Center for Atmospheric Research, Boulder, CO chriss@ucar.edu

13

Ensemble filtering for atmospheric flows is enjoying rapid progress. For the atmospheric mesoscale, some of the phenomena of most pressing interest are severe convective storms, hurricanes and cyclonic winter storms with their embedded fronts and associated precipitation. Because of the lack of spatial resolution in conventional observing systems, remotely sensed observations are important, especially those from Doppler radars and from satellites. I will provide an overview of key issues, both partially resolved and largely unresolved, in the use of ensemble filters for atmospheric-mesoscale flows. These include (i) the need to restrict the influence of observations to spatially local state variables in order to diminish effects of sampling error, (ii) techniques for representing the uncertainty in lateral boundary conditions and in lower boundary forcing (from land or sea surface), (iii) techniques to account for forecast-model error, which can be particularly large for mesoscale flows and (iv) the obstacles presented by flows with multiple spatial scales. Strong Adjoint Sensitivities in Tropical Eddy-Permitting Variational Data Assimilation B. Cornuelle(1), I. Hoteit(1), A. Koehl(2) and D. Stammer(2) (1)Scripps Institute of Oceanography, UCSD, San Diego, CA (2)Institut für Meereskunde der Universität Hamburg, Hamburg, Germany bdc@ucsd.edu A variational data assimilation system has been implemented for the tropical Pacific Ocean for an eddy-permitting regional implementation of the MIT general circulation model (MITgcm). The model uses realistic topography with parameterizations for the surface boundary layer (KPP) and open boundaries at the south and north, as well as in the Indonesian throughflow. The strong constraint 4DVAR method is used to adjust the model to observations in the tropical Pacific region using control parameters which include initial temperature and salinity, temperature, salinity and horizontal velocities at the open boundaries, and twice-daily surface fluxes of momentum, heat and freshwater. The model is constrained with most of the available datasets in the tropical Pacific, including climatologies, TAO, ARGO, XBT, and satellite SST and SSH data. The forward model runs exhibit strongly growing flow instabilities in the regions of high kinetic energy and low planetary potential vorticity gradient, as expected. The growth of these perturbations is limited by nonlinearities once they reach finite size, meaning that the high linear growth rates do not apply for long time periods. This poses a technical problem for adjoint-based assimilation, which depends on the linearized sensitivities to adjust the controls. Relative to the forward model runs, increased viscosity and diffusivity terms are used in the adjoint model runs to avoid large sensitivities related to the flow instabilities present in the high-resolution model. This talk will discuss some of the technical aspects and show results for 1 year assimilation periods.

14

The representer method with nonlinear models: to cycle or not to cycle? H. E. Ngodock, S. R. Smith and G. A. Jacobs

The Naval Research Laboratory, Stennis Space Center, MS ngodock@nrlssc.navy.mil Variational data assimilation with nonlinear models requires tangent linearization, which may be sufficiently accurate only for relatively short time scales. However, for time intervals beyond the scales of nonlinear event development, the tangent linearization cannot be expected to be sufficiently accurate. The representer method would, therefore, not be able to yield a reliable and accurate assimilation solution. However, the method can be implemented for successive cycles in order to solve the entire nonlinear problem. By cycling the representer method, it is possible to reduce the assimilation problem into intervals in which the linear theory is able to perform accurately. For each cycle, the background needed for the tangent linearization is computed by propagating the nonlinear dynamics using the final solution to the linearized assimilation problem from the previous cycle as the initial conditions. This study demonstrates that by cycling the representer method, the tangent linearization is sufficiently accurate once adequate assimilation accuracy is achieved in the early cycles. The outer loops that are usually required to contend with the linear assimilation of a nonlinear problem are not required beyond the early cycles, because the tangent linear model is sufficiently accurate at this point. The combination of cycling the representer method and limiting the outer loops to one significantly lower the cost of the overall assimilation problem. In addition, this study shows that weak constraint assimilation is capable of extending the assimilation period beyond the time range of the accuracy of the tangent linear model. That is, the weak constraint assimilation can correct the inaccuracies of the tangent linear model and clearly outperform the strong constraint method. Preliminary examples using the Lorenz attractor and a reduced gravity ocean model will be presented. An Unstructured Grid, Finite-Volume Coastal Ocean Model (FVCOM) System: Validations and Applications R. C. Beardsley(1), C. Chen(2) and G. Cowles(2)

1Department of Physical Oceanographic Institution, Woods Hole Oceanographic Institution, Woods Hole, MA 2Department of Fisheries Oceanographic Institution, School for Marine Science and Technology, University of Massachusetts-Dartmouth, New Bedford, MA rbeardsley@whoi.edu A team of University of Massachusetts-Dartmouth -Woods Hole Oceanographic Institution researchers has developed a new prognostic, unstructured-grid, Finite-Volume, free-surface, 3-D primitive equation Coastal Ocean circulation Model (called FVCOM) for physical and coupled physical/biological studies in coastal regions characterized by complex coastlines and bathymetry and diverse forcing. FVCOM is solved numerically by flux calculation using the

15

integral form of the governing equations over an unstructured triangular grid. This approach combines excellent grid flexibility with superior numerical efficiency and code simplicity and provides local conservation of mass, momentum, salt, heat, and tracer. The present version of FVCOM includes the following options and components: (1) choice of Cartesian or spherical coordinate system; (2) generalized terrain-following vertical coordinate; (3) mass-conservative wet/dry point treatment for flooding/drying simulation; (4) General Ocean Turbulent Model (GOTM) modules for optional vertical turbulent mixing schemes; (5) 3-D sediment transport module (based on the U.S.G.S. national sediment transport model) for estuarine and near-shore applications; (6) Generalized (7-group) Biological Module (GBM) to facilitate lower-food web ecosystem applications; (7) water quality model to simulate dissolved oxygen and other environmental indicators; and (8) nudging and Kalman Filter methods for data assimilation. New modules in development include: (1) non-hydrostatic dynamics; (2) nonlinear ice model; and (3) unstructured-grid version of SWAN. FVCOM is written in Fortran 90 with MPI parallelization, and runs efficiently on single and multi-processor machines. FVCOM has been successfully applied to a number of community test problems and estuarine, continental shelf and regional/open ocean studies involving realistic model domains (see http://fvcom.smast.umassd.edu). For hindcast and forecast applications, an integrated coastal ocean model system has been built by coupling FVCOM with the fifth-generation NCAR/Penn State mesoscale meteorological model (MM5) for realistic surface forcing and the addition of advanced data assimilation. Some example applications and model results will be presented. The Submesoscale Dynamical Transition in the Ocean J. C. McWilliams Department of Atmospheric and Oceanic Sciences & Institute of Geophysics and Planetary Physics, UCLA, Los Angeles, CA jmc@atmos.ucla.edu Analogous to the fact that weather phenomena are essential to climate dynamics, oceanic mesoscale eddies with a horizontal scale of about 50 km are essential to the general circulation in the ocean. The prevailing view is that eddies approximately satisfy a diagnostic force balance – hydrostatic in the vertical and geostrophic or gradient-wind in the horizontal – and are both weakly dissipative and nearly adiabatic in the sense of retaining material parcels on isentropic surfaces in stably stratified, interior regions. It is becoming increasingly clear that this view is falsely simplistic; i.e., there is an intrinsic submeoscale current regime, spawned by mesoscale eddies, on a scale of a few km. It is not highly force-balanced, and it affects a significant forward kinetic energy cascade toward microscale dissipation and transports material across isentropic surfaces. Its flow patterns are comprised of lateral density fronts and vortices arising from frontal instabilities. Particularly in association with its frontogenesis, submesoscale flows act to restratify the fluid near vertical boundaries, where boundary-layer turbulence tends to mix and destratify. Illustrations will be shown from upper-ocean measurements and from computational simulations for an idealized boundary current and for Eady's uniform vertical shear flow in turbulent equilibrium balance.

16

Overview of Data Assimilation in Ecosystem Modeling Y. H. Spitz College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR yvette@coas.oregonstate.edu Over the last two decades, data assimilation has been traditionally used in atmospheric and oceanic circulation modeling to estimate initial and boundary conditions. In ecosystem modeling, the largest uncertainties rely in the model parameter and pathway definition. During the last decade, the availability of long term time series observations, such as from the Bermuda Atlantic Time Series (BATS) and the Hawaii Ocean Time series (HOT), North Sea Belgian coastal station 330, and from process oriented studies (e.g. the North Atlantic Bloom Experiment (NABE), Equatorial Pacific experiment (EqPac), has made the application of data assimilation techniques (e.g. variational adjoint method) feasible to determine unknown ecosystem model parameters and their relative importance in controlling ecosystem dynamics. Examples of zero-dimensional (time) and one-dimensional (time and vertical) application of data assimilation including examples of model validation from the U.S. JGOFS Regional Ecosystem Modeling 'Testbed' Project (http://www.ccpo.odu.edu/RTBproject/index.html) will be given. For coastal and basin scale problems (three-dimensional), the computational cost (cpu and memory) of the variational adjoint method becomes quickly prohibitive and uncertainties include ocean circulation, atmospheric forcing and ecosystem model parameters. Kalman filter type of data assimilation has been applied for state variable estimations and can be applied to determine errors of representation. We show an example of the application of the Reduced State Space Kalman filter to determine the error of representation of the modeled circulation of the North Pacific basin that can be applied to the estimate of the error of representation of the coupled circulation/ecosystem model. Satellite Ocean Data Assimilation at the Joint Center for Satellite Data Assimilation and NOAA E. Bayler Joint Center for Satellite Data Assimilation, Camp Springs, MD eric.bayler@noaa.gov The Joint Center for Satellite Data Assimilation is a partnership between the National Oceanic and Atmospheric Administration (NOAA), the National Aeronautics and Space Agency, and the Department of Defense (in particular the Navy and the Air Force). The Joint Center’s goal is to accelerate and improve the quantitative use of research and operational satellite data in weather, climate, and ocean prediction models and environmental analysis. Collaborative and contributive efforts are provided by the partners. Consistent with the increasing availability of satellite ocean observations in the operational environment, the Joint Center’s ocean component is growing. Objectives, plans, and efforts will be discussed. NOAA’s efforts within the

17

partnership will be highlighted. In particular, plans and contributions for enhancing the operational models within the National Weather Service’s National Centers for Environmental Prediction (NCEP) through the use of satellite ocean observations will be reviewed. Links between NCEP’s basin scale ocean models and the National Ocean Service’s coastal, regional, and estuarine scale models will be discussed. Operational Estuarine and Coastal Forecast Hydrodynamic Modeling and Development in the National Ocean Service

F. Aikman III(1), E. Wei(1), R. Patchen(1) and H. J. Herring(2)

(1)Marine Modeling & Analysis Programs, Coast Survey Development Laboratory, National Ocean Service, NOAA, Silver Spring, MD (2)Dynalysis of Princeton, Princeton, NJ eugene.wei@noaa.gov NOAA’s National Ocean Service (NOS) is involved in the application of hydrodynamic models and the development of operational forecast systems in estuaries and the coastal ocean. These models and systems have applications in the support of safe and efficient navigation and emergency response, as well as marine geospatial and ecosystem applications. Nowcast/forecast systems developed in-house for the Chesapeake Bay, the Port of New York and New Jersey, and Galveston Bay are now operational. Models developed elsewhere have recently been transitioned to operational status for the five Great Lakes and for the St Johns River, FL. Forecast systems will be operationally implemented in the future for the Columbia River, Tampa Bay, Delaware Bay, as well as Cook Inlet, AK and elsewhere. Once tested, fully evaluated and deemed accurate by NOS standards, the nowcast/forecast systems are transitioned into the operational environment. The components of a real-time estuarine modeling system are discussed in terms of a “standard” Coastal Ocean Modeling Framework (COMF) which will increase the efficiency of research, development, transition and operations. The COMF includes the essential operational management of observations and forecasts of atmospheric, coastal and riverine inputs, as well as operational quality control and the dissemination of results. The COMF is intended to stimulate a community approach to coastal modeling, providing tools and protocols, and to abide by IOOS and ESMF standards. Some examples of ecological applications of hydrodynamic models will be discussed, as well as the challenge of data assimilation in the near coastal and estuarine environment. In support of our coastal forecast systems, NOS for the last two-years has been running routinely both short-term and long range forecasts for the Gulf of Mexico (http://chartmaker.ncd.noaa.gov/csdl/op/dgom.m.html). The comparison between numerical model forecasts of Gulf of Mexico circulation and observations from drifters and satellite imagery indicates that a large part of the forecast error is due to the representation of satellite information that is assimilated during the model nowcast. NOS and Dynalysis of Princeton are collaborating on a project to correlate the surface height anomaly with the integral density anomaly of each of the water masses in the Gulf, using contemporaneous altimeter and hydrographic cast measurements, and then develop an algorithm for assimilating the temperature and salinity fields in a nowcast/forecast model.

18

Predictability of Mesoscale Variability in the East Australia Current System given Strong Constraint Data Assimilation J. Wilkin, J. Zavala-Garay and H. G. Arango Institute of Marine and Coastal Sciences, Rutgers University, New Brunswick, NJ wilkin@marine.rutgers.edu The objectives of data assimilation in oceanography are typically the reanalysis of a suite of observations for the purposes of hindcast state estimation, and determining initial conditions for model forecasting. In this project we focus on the latter objective and evaluate the Incremental, Strong constraint, 4-Dimensional Variational (IS4DVAR) data assimilation method implemented in the Regional Ocean Modeling System (ROMS) for predictions of mesoscale variability in the East Australia Current (EAC) System. The observations assimilated are daily composites of AVHRR SST, 7-day reanalysis of AVISO SSH anomalies, and high resolution expendable bathythermograph (XBT) temperature profiles from Volunteer Observing Ship (VOS) transects of the Tasman Sea. Considering a 2-year data set for 2001 and 2002, ROMS forecast initial conditions are generated every week by assimilating observations from the 7 days preceding the forecast initial time. Forecast ensembles are produced by adding to the forecast initial conditions so-called optimal perturbations computed from singular vectors of the ROMS Tangent Linear model. Assimilation of satellite data only (SST and SSH) results in relatively poor estimates of the true subsurface ocean state observed by XBTs, and consequently a poor subsurface forecast skill. Furthermore, the modeled circulation shows significant sensitivity to errors in the initial conditions and therefore the uncertainty, or spread, in the forecast ensemble is high. Including the XBTs in the assimilation experiments improves the ocean state estimation in the vicinity of the XBT transects and reduces the sensitivity to errors in the initial conditions resulting in a more skillful ensemble forecast. Motivated by this finding we explore the utility of including subsurface pseudo-observations based on an empirical relationship between subsurface information and satellite observed surface conditions (CSIRO’s “Synthetic XBT” analysis). The preliminary results show that better ocean state estimates and more skillful forecasts are obtained in the entire domain considered. Weak constraint 4D variational data assimilation in the inverse Regional Ocean Modeling System (ROMS) E. Di Lorenzo Georgia Institute of Technology, Atlanta, GA edl@gatech.eduTBA

19

Hindcast and Reanalysis of the Circulation in the Bering and Chukchi Seas G. Panteleev(1), D. Nechaev(2), V. Luchin(3), R. Woodgate(4) and P. Stabeno(5) (1)International Arctic Research Center, University of Alaska, Fairbanks, AK (2)University of Southern Mississippi, Hattiesburg, MI (3)FEBRAS, Vladivostok, Russia (4)University of Washington, Applied Physics Laboratory, Seattle, WA (5)Pacific Marine Environmental Laboratory, Seattle, WA gleb@iarc.uaf.edu We present results of a set of variational data assimilation experiments aimed at the reconstruction of quasi-stationary and time-varying circulations in the Chukchi and Bering Seas. Depending on the assimilated data, these experiments are focused on different regions and time periods. The model used for the reconstruction is designed specifically for efficient variational assimilation of long-term observations in ocean regions governed by flow through open boundaries and by atmospheric fluxes. The circulation in the Chukchi Sea is derived from various sources of observations including two months of velocity, temperature and salinity records from moorings and CTD observations in autumn 1990 (www.frontier.iarc.uaf.edu/~gleb). Assimilation of mooring velocities allows us to quantify volume, heat and salt transports in the Chukchi Sea. The reconstructed circulation pattern reveals periodical reverse of the East Siberian Current and flow through the Bering Strait, which are the important features of the Chukchi Sea circulation. The quasi-stationary circulation in the Bering Sea is recovered from drifter and mooring observations, climatological temperature and salinity data, and climatological surface fluxes of momentum, heat and fresh water. The estimates of volume transports through the Aleutian straits are presented. Several numerical examples show that the reconstructed climatological sea surface height distribution can be effectively used for operational hind-cast of the circulation in the Bering Sea. Experiments intended for evaluation of the Amukta Pass transports in January 2002 are conducted with assimilation of AVISO satellite altimetry anomalies and optimized climatological SSH. The obtained volume transports are compared with low-pass filtered transports through the Amukta Pass calculated from four moorings (Stabeno et al., 2005). Evolution of the Mid-Atlantic Coupled Observation and Modeling System O. Schofield and J. Wilkin Rutgers University, New Brunswick, NJ oscar@marine.rutgers.eduTBA

20

Real-time Data Assimilation and Ensemble Prediction System in the Intra-Americas Sea B. Powell(1), A. Moore(1), H. Arango(2), R. Milliff(3) and E. Di Lorenzo(4) (1)University of California, Santa Cruz, CA (2)Rutgers University, New Brunswick, NJ (3)Colorado Research Associates, Boulder, CO (4)Georgia Institute of Technology, Atlanta, GA powellb@ucsc.edu The Intra-Americas Sea is an interesting oceanic region comprised of deep basins and complex bathymetry and geometry. It is a well-constrained region of the North Atlantic with land mass boundaries along the western and northern extents. The Caribbean region is highly dynamic, but is well sampled by a long, overlapping time series of both satellite and in situ physical oceanographic measurements. Using the Regional Ocean Modeling System (ROMS), we have developed a real-time data assimilation system utilizing both satellite surface observations and in situ ship measurements to generate the best model state for the current day. Utilizing the numerous tangent-linear solutions from the data assimilation system, we generate a set of orthonormal perturbations to apply to the generated initial conditions to generate a forward, two- week predictive ensemble. This assimilative/prediction system is now running automatically in an experimental operational capability aboard a ship making regular trips across the region. We will be discussing the algorithmic setup and preliminary results from this experiment. Linking Observations and Modeling in Coastal Ocean Observing Systems J. Barth College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR barth@coas.oregonstate.edu Making progress on dynamical understanding of coastal ocean processes and building predictive capability requires both ocean observations and models. Observing and modeling, and observers and modelers, have not always been in lock step. Data assimilation is, of course, one obvious way to move forward. After a brief review of some past, present and future coastal ocean processes programs, I'd like to discuss future needs for coastal ocean observing and modeling, and, most importantly, how our community can work together.

21

Poster Abstracts: Ensemble-based assimilation of HF-Radar currents in a West Florida Shelf ROMS model nested in HYCOM A. Barth, A. Alvera-Azcarate and R. H. Weisberg Ocean Circulation Group, University of South Florida, FL abarth@marine.usf.edu A West Florida Shelf (WFS) model is constructed by nesting the Regional Ocean Model System (ROMS) in the Atlantic Hybrid Coordinate Ocean Model (HYCOM) to include both local and deep-ocean forcing, particularly the Gulf of Mexico Loop Current (LC). An ensemble simulation of WFS ROMS model was carried out under different wind forcings in order to estimate the error covariance of the model state vector and the covariance between ocean currents and wind. Radial currents measured by the CODAR antennas near St. Petersburg and Venice, FL are assimilated. The wind stress is included in the state vector and the assimilation of surface currents also produces a correction of the wind stress. First results of WFS ROMS model assimilating surface currents using this ensemble show an improvement of the model currents not only at the surface but also at depth. An Assimilating Tidal Model for the Bering Sea M. G. G. Foreman, P. F. Cummins and J. Y. Cherniawsky Institute of Ocean Sciences, Department of Fisheries and Oceans, Sidney, BC, Canada ForemanM@pac.dfo-mpo.gc.ca Representers are used to assimilate tidal harmonics computed from TOPEX/POSEIDON altimetry into a barotropic, finite element model of the Bering Sea. Accuracy is evaluated though comparisons with independent bottom pressure gauges. The model is used to estimate energy fluxes through each of the Aleutian Passes and Bering Strait and to construct an energy budget for the major tidal constituents. The finite element model does not conserve mass locally and this is shown to give rise to an additional term in the energy budget whose contribution is significant for the prior model, but which can be reduced with the assimilation. Diurnal tidal dissipation is shown to be larger than semi-diurnal and the effects of the 18.6 year nodal modulation on tidal mixing are estimated.

22

Data assimilation in the Columbia River estuary and plume: results of hindcast, forecast, and observation-optimization experiments S. Frolov(1), A. Baptista(1), T. Leen(2), Z. Lu(2) and R. van der Merwe(2)

(1)Department of Environmental and Biomolecular Systems, Oregon Health and Science University, Portland, OR (2)Department of Computer Science and Electrical Engineering, Oregon Health and Science University, Portland, OR frolovsa@gmail.com Data assimilation (DA) plays a central role in emerging costal observatories. However, a wide application of many existing DA methods is hampered, among other things, by the limited computational resources available to coastal observatories. To address the need for fast, model-independent, non-linear data assimilation methods, we recently proposed a non-linear extension to the reduced-space Kalman filter (Lu, et al. in prep., Frolov, et al. submitted). The computational efficiency of the new method comes, in part, from the use of neural network model surrogates (van der Merwe, et al. in press) that execute forward simulations three orders of magnitude faster than the traditional numerical circulation codes. In this poster, we report on progress applying our DA methods to assimilate in-situ measurements of the Columbia River estuary and plume. The assimilation adds considerable skill in representing several processes of interest, including: non-linear tides, salinity intrusion, heat exchange with the atmosphere, size and direction of the freshwater plume, and stratification of the coastal ocean. We also demonstrate the aptitude of the new method for “real-time” forecasts. Finally, we use the error statistics predicted by the Kalman filter to guide the design of the observation system in the estuary. The data assimilation method is very fast: ~300 time faster than “real-time” and ~50 faster than the forward model. The results of the data assimilation experiments suggest that mathematically-principled, fast, and model-independent data assimilation, in a complex coastal environment like the Columbia River estuary and plume, is possible. 1. Lu, Z., T.K. Leen, S. Frolov, R. van der Merwe, and A.M. Baptista (in prep.) Sequential data

assimilation with a Sigma-point Kalman filter on low-dimensional manifold. 2. Frolov, S., Z. Lu, R. van der Merwe, T.K. Leen, and A.M. Baptista, (submitted). Fast data

assimilation with a non-linear Kalman filter and model surrogates: application to an estuarine circulation, Ocean Modeling.

3. van der Merwe, R., T.K. Leen, Z. Lu, S. Frolov, and A.M. Baptista (in press). Fast Neural Network Surrogates for Very High Dimensional Physics-based Models in Computational Oceanography, Neural Computation.

23

S4DVAR assimilation of HF radar radial surface currents I. Hoteit(1), B. Cornuelle(1), A. Koehl(2) and P. Heimbach(3) (1)Scripps Institute of Oceanography, UCSD, San Diego, CA (2)Institut für Meereskunde der Universität Hamburg, Hamburg, Germany (3)Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA ihoteit@ucsd.edu We consider the problem of combining High Frequency (HF) radar data with a coastal hydrodynamic model of the San Diego region using an adjoint-based strong-constraint (S4DVAR) assimilation method. The HF radar provides near-real-time time radial velocities from 3 sites covering the region offshore of San Diego Bay. The hydrodynamic model is the MITGCM with 1km horizontal resolution and 40 vertical layers. The domain is centered on Point Loma, extending 117 km offshore and 120 km alongshore. The reference run (before assimilation) is initialized from a single profile of T and S and forced with wind data from a single shore station and with zero heat and fresh water fluxes. The S4DVAR method is used to adjust the model to hourly HF radar radial velocity observations by adjusting initial temperature, salinity, and velocity, twice-hourly temperature, salinity and horizontal velocities at the open boundaries, and twice-hourly surface fluxes of momentum, heat and freshwater. The assimilation system is described and assimilation results are presented and discussed. Predictability of the Oregon coast by the relative entropy S. Kim(1), R. Samelson(1) and C. Snyder(2) (1)College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR (2)National Center for Atmospheric Research, Boulder, CO skim@coas.oregonstate.edu The uncertainty and predictability of dynamical systems can be quantified by the relative entropy, which gives a measure of predictive information content based on ensemble distributions. The concept of the relative entropy is explored in primitive-equation simulations of ocean circulation along the Oregon coast. The ensemble members are obtained by perturbing the initial temperature field with either white noise or specified Fourier components, and then advancing the simulations for 60 days with either constant southward wind forcing, or with periodic wind forcing that has a 5- day period and a southward mean. The relative entropy consists of two components, the "signal" and "dispersion" terms, and is computed here under the assumption that temperatures are mutually independent and have Gaussian distributions. The signal term is found to dominate, giving a different result than would be obtained using potential predictability, which focuses on the covariance of the ensemble distribution. When the simulation length is increased, the ensemble spread varies in the opposite manner to the

24

dispersion in the relative entropy. This contrasts with results from simple models, for which the dispersion often varies with the ensemble spread. Some of these results depend also on the particular choices made for the definition of the climatological distributions. Representer-based variational data assimilation in nonlinear coastal ocean models A. L. Kurapov, G. D. Egbert, J. S. Allen, R. N. Miller and S. Erofeeva College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR kurapov@coas.oregonstate.edu Our research has been focused on advanced data assimilation in coastal ocean circulation models. To study aspects of data assimilation for flows in which variability is dominated by instabilities and eddy interactions we have implemented the variational representer-based method (Bennett, 2002) with a shallow-water model of circulation in the nearshore surf zone and with synthetic data. The response of the model to the steady forcing, representing the effect of breaking waves, depends on the magnitude of the friction coefficient. If the friction coefficient is low, the flow is prone to shear instabilities that evolve with time in an equilibrated waves regime, or a more strongly nonlinear irregular flow regime. We have run data assimilation experiments in both nonlinear regimes. As a result of data assimilation, both the initial conditions and forcing are corrected. In the strongly nonlinear case, the representer method can converge to an accurate solution over a time interval that is substantially longer than the validity limit of the TL model or the eddy time scale. To achieve that, time variability in the forcing (on eddy time scale) is allowed by a choice of the forcing error covariance, to represent scales of observed flow variability in the forcing correction. To obtain accurate estimates of both the state and the forcing, we implement a forcing error covariance with a temporal correlation separated into an O(1) steady and small amplitude time-variable parts. By that means, only small temporal variations near the time-mean inverse forcing are allowed. Our experience with the shallow water model is being utilized as we approach variational data assimilation in a model of stratified shelf circulation. This work in progress involves testing of the Regional Ocean Modeling System (ROMS) tangent linear and adjoint codes (Moore et al., Oc. Mod. 2004, Di Lorenzo et al., Oc. Mod., 2007) and the Inverse Ocean Modeling System (IOM) (Chua and Bennett, Oc. Mod. 2001). A three-dimensional variational data assimilation system and observing system experiments for coastal oceans Z. Li(1), Y. Chao(1), J. Farrara(1), X. Wang(1), J. C. McWilliams(2) and K. Ide(2)

(1)Jet Propulsion Laboratory, Pasadena, CA (2)University of California, Los Angeles, CA zhijin@pacific.jpl.nasa.gov

25

A three-dimensional variational data assimilation (3DVAR) system has been developed for the Regional Ocean Modeling System (ROMS), and it is called ROMS-DAS. This system provides a capability of predicting meso- to small-scale variations with temporal scales from hours to days in the coastal oceans. To cope with the particular difficulties that result from complex coastlines and bottom topography, unbalanced flows and sparse observations, ROMS-DAS utilizes several novel strategies. These strategies include the implementation of three-dimensional anisotropic and inhomogeneous error correlations, application of particular weak dynamic constraints, and implementation of efficient and reliable algorithms for minimizing the cost function. The ROMS-DAS system was applied in field experiments for Monterey Bay during both 2003 (Autonomous Ocean Sampling Network - AOSN) and 2006 (MB06). These two experiments involved intensive data collection from a variety of observational platforms, including satellites, airplanes, High Frequency radar, Acoustic Doppler Current Profilers, ships, drifters, buoys, autonomous underwater vehicles (AUV), and particularly a fleet of undersea gliders. By using these observed data sets, various data assimilation experiments and observing system experiments (OSE) were performed. The major results will be presented. State estimation in an ocean-biogeochemical model by assimilation of satellite ocean chlorophyll data L. Nerger and W. W. Gregg

Goddard Earth Sciences and Technology Center NASA Goddard Space Flight Center, Greenbelt, MD

lnerger@gmao.gsfc.nasa.gov

Chlorophyll data from the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) is assimilated into the three-dimensional global NASA Ocean Biogeochemical Model (NOBM) for the period 1998-2004. The assimilation is performed by a multivariate configuration of the SEIK filter, which is an ensemble-based Kalman filter scheme. The filter is simplified by the use of a static error covariance matrix. It operates with a localized analysis and is amended by an online bias correction scheme. The multivariate assimilation is applied to update the four phytoplankton groups of the model as well as simulated nutrient fields. The chlorophyll estimates of the model can be improved by the assimilation such that they outperform the assimilated SeaWiFS data. However, the results are less clear for the nutrients where the bias estimation is required for stability but reduces the assimilation improvements. Hindcast and Reanalysis of the Circulation in the Bering and Chukchi Seas G. Panteleev(1), D. Nechaev(2), V. Luchin(3), R. Woodgate(4) and P. Stabeno(5) (1)International Arctic Research Center, University of Alaska, Fairbanks, AK (2)University of Southern Mississippi, Hattiesburg, MI (3)FEBRAS, Vladivostok, Russia (4)University of Washington, Applied Physics Laboratory, Seattle, WA (5)Pacific Marine Environmental Laboratory, Seattle, WA

26

gleb@iarc.uaf.edu We present results of a set of variational data assimilation experiments aimed at the reconstruction of quasi-stationary and time-varying circulations in the Chukchi and Bering Seas. Depending on the assimilated data, these experiments are focused on different regions and time periods. The model used for the reconstruction is designed specifically for efficient variational assimilation of long-term observations in ocean regions governed by flow through open boundaries and by atmospheric fluxes. The circulation in the Chukchi Sea is derived from various sources of observations including two months of velocity, temperature and salinity records from moorings and CTD observations in autumn 1990 (www.frontier.iarc.uaf.edu/~gleb). Assimilation of mooring velocities allows us to quantify volume, heat and salt transports in the Chukchi Sea. The reconstructed circulation pattern reveals periodical reverse of the East Siberian Current and flow through the Bering Strait, which are the important features of the Chukchi Sea circulation. The quasi-stationary circulation in the Bering Sea is recovered from drifter and mooring observations, climatological temperature and salinity data, and climatological surface fluxes of momentum, heat and fresh water. The estimates of volume transports through the Aleutian straits are presented. Several numerical examples show that the reconstructed climatological sea surface height distribution can be effectively used for operational hind-cast of the circulation in the Bering Sea. Experiments intended for evaluation of the Amukta Pass transports in January 2002 are conducted with assimilation of AVISO satellite altimetry anomalies and optimized climatological SSH. The obtained volume transports are compared with low-pass filtered transports through the Amukta Pass calculated from four moorings (Stabeno et al., 2005). Results from glider data assimilation experiments I. Shulman(1), C. Rowley(1), S. Anderson(1), J. Kindle(1), S. DeRada(2), J. Cummings(1) and J. Doyle(3) (1)Oceanography Division, Naval Research Laboratory, Stennis Space Center, MS (2)Jacobs Sverdrup, Stennis Space Center, MS (3)Marine Meteorology Division, Naval Research Laboratory, Monterey, CA igor.shulman@nrlssc.navy.mil Results of glider data assimilation during the Autonomous Ocean Sampling Network (AOSN) experiments are presented. The modeling system consists of an implementation of the Navy Coastal Ocean Model (NCOM) in the Monterey Bay area. The model receives open boundary conditions from a basin-scale NCOM-based the California Current System, and surface fluxes from the Coupled Ocean-Atmosphere Mesoscale Prediction System (COAMPS) atmospheric model at 3 km resolution. The data assimilation component of the system is a version of The Navy Coupled Ocean Data Assimilation (NCODA) system which is used for assimilation of the glider data into the NCOM model of the Monterey Bay area. The NCODA is a fully 3D multivariate optimum interpolation system that produces simultaneous analyses of temperature, salinity, geopotential, and vector velocity. Impact of glider data assimilation on the model predictions of surface and subsurface properties are discussed. Assimilation of the glider data

27

significantly reduced biases and RMS errors in predictions of surface temperature and salinity at mooring locations. Also, assimilation of the glider data provided a better agreement with observations (for example, with aircraft measured SSTs) of the model predicted and observed spatial distributions of surface properties. Moorings observations of subsurface temperature and salinity show sharp changes in the thermocline and halocline depths during transitions from upwelling to relaxation and vice versa. The non-assimilative run shows these transitions in subsurface temperature; however, changes are not as defined as in observations. For salinity, the non-assimilative run significantly differs from the observations. However, the glider data assimilating run is able to show comparable results with observations in deepening (shallowing) of thermocline as well as halocline depths during upwelling (relaxation) events. Boundary conditions for coastal models from data assimilative HYbrid Coordinate Ocean Model (HYCOM) nowcast/forecast systems

O. M. Smedstad(1), J. Cummings(2), A. J. Wallcraft(2), E. J. Metzger(2), H. E. Hurlburt(2), P. J. Hogan(2), E. P. Chassignet(3) and G. R. Halliwell(4) (1)Planning Systems Inc., Stennis Space Center, MS (2)Naval Research Laboratory, Stennis Space Center, MS (3)Florida State University, Tallahassee, FL (4)University of Miami, Miami, FL smedstad@nrlssc.navy.mil The Navy Coupled Ocean Data Assimilation (NCODA) system has been implemented in HYCOM. NCODA is a multivariate optimal interpolation scheme (MVOI) that assimilates satellite track data, available MCSST and in situ observations, including profile data from BT's and ARGO floats. NCODA is also able to assimilate other types of data, e.g. profiles from gliders or velocity observations. The 24 hour forecast from HYCOM is used as a first guess for the NCODA analysis. This analysis is used to incrementally update the HYCOM forecast variables. A NCODA analysis is performed every 24 hours. Currently two systems based on HYCOM/NCODA are running in real time, a 1/12 degree global model and a 1/25 degree Gulf of Mexico model. Both models are forced by the 0.5 degree Navy Operational Global Atmospheric Prediction System (NOGAPS) atmospheric fields. The systems run a 4 day hindcast to pick up late arriving observations. A daily 7 day forecast is performed with the Gulf of Mexico system, while a 5 day daily forecast and a weekly 30 day forecast is planned for the global model. The prediction systems provide boundary conditions for higher resolution coastal models. An accurate representation of the oceanographic fields at the open boundaries of a coastal model is important for a successful coastal ocean prediction system. Results from the HYCOM systems will be discussed as well as examples of using HYCOM as boundary conditions for coastal models.

28

The Cycling Representer Method applied to the Navy Coastal Ocean Model S. R. Smith, H. E. Ngodock and G. A. Jacobs The Naval Research Laboratory, Stennis Space Center, MS scott.smith@nrlssc.navy.milSee presentation abstract by H. E. Ngodock. The impact of boundary data on solutions of a nested grid model of the Oregon Coastal Transition Zone S. R. Springer, J. S. Allen, G. D. Egbert, A. L. Kurapov, R. N. Miller, R. M. Samelson College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR springer@coas.oregonstate.edu We have developed a nested model of the Oregon Coastal Transition Zone for studying interactions between the shelf and the open ocean in the northern reaches of the California Current system. To explore improvements that arise from nesting in GODAE products, we have assessed the impact of different sources of boundary data (Levitus climatology, data-assimilating NCOM-CCS and non-assimilating NCOM-CCS) on the nested model forward solutions for the summer upwelling season of 2001. Evaluations are made by comparison with the extensive data collected in the vicinity of Heceta Bank during the COAST experiment, with coastal tide gauges, and with TOPEX sea surface height observations. Differences in coastal sea level can be traced to differences in the southern boundary data applied to the model, and the impact of boundary data on modeled sea surface height increases with increasing depth on the shelf. Similarly, the source of boundary data is increasingly important to modeling alongshore barotropic currents as one moves from the inner (<50 m depth) to outer (> 150 m depth) shelf regions, although all solutions reproduce variability of the currents reasonably well. Off the shelf, circulation is dominated by the surface-intensified coastal upwelling jet. Instability of this jet, combined with episodic relaxation of upwelling favorable winds, induces the formation of eddies with a scale of tens of km. The number and timing of these eddies is sensitive to both the source of the boundary data and the implementation of the nested boundary condition. In the subjectively chosen best model run, comparison of the largest eddy with TOPEX sea surface height observations suggests that the model represents well the size of the Cape Blanco eddy and the timing of its creation. These results suggest that the impact of boundary conditions in the Oregon Coastal Transition Zone is best assessed offshore of the coastal jet, but that the predictability of eddies must be considered in comparing model results with data. Lagrangian Data Assimilation in the Gulf of Mexico G. Vernieres(1), K. Ide(2) and C. T. Jones(1) (1)Department of Mathematics, University of North Carolina at Chapel Hill, NC (2)University of California, Los Angeles, CA

29

gvernier@email.unc.edu We present our ongoing research on a Lagrangian data assimilation method applied to the Gulf of Mexico. In particular, we are interested in the efficiency of the method to estimate the shedding of eddies from the loop current. The model used is a simple multi-layer reduced gravity model of the region of interest with an imposed inflow/outflow condition. We will present some preliminary results based on the assimilation of the position of synthetic drifters, in which the control run represents the shedding of an eddy. Baroclinic Tidal Generation in the Kauai Channel Inferred from HF-Radar E. D. Zaron Department of Civil and Environmental Engineering, Portland State University, Portland, OR ezaron@coas.oregonstate.edu A data-assimilating 3-dimensional primitive equations model is used to study baroclinic tidal generation in the Kauai Channel on the Hawaiian Ridge. The data assimilation system is implemented using the IOM -- the Inverse Ocean Model -- a software system that implements variational data assimilation systems for arbitrary numerical models and observing systems. By combining the dynamical constraints provided from the model with the surface currents from the HF-radar array, we quantify the impact of mesoscale currents on tidal processes in the channel. Non-tidal, mesoscale currents modify the flow fields and reduce the baroclinic tidal generation compared to what would occur without the non- tidal currents. The surface expression of the internal tide can be detected by satellite altimetry, and these data are used to validate the inferences from the HF-radar data. Research is ongoing to investigate temporal variability in the tidal--mesoscale interactions. Predictability of Mesoscale Variability in the East Australia Current System given Strong Constraint Data Assimilation J. Zavala-Garay, J. Wilkin and H. G. Arango Institute of Marine and Coastal Sciences, Rutgers University, New Brunswick, NJ jzavala@marine.rutgers.edu One of the many applications of data assimilation is the estimation of adequate initial conditions for model forecasts. In this work we evaluate to what extent the Incremental, Strong constraint, 4-dimensional Variational (IS4DVAR) data assimilation can improve prediction of mesoscale variability in the East Australia Current (EAC) System. The observations considered in the assimilation experiments are daily composites of AVHRR SST, 7-day reanalysis of AVISO SSH anomalies, and temperature sections from high resolution expendable Bathythermograph (XBT). Considering all available observations for years 2001 and 2002, ROMS forecast initial

30

conditions are generated every week by assimilating the available observations from the 7 days prior to the forecast initial time. It is shown that assimilation of just surface information (SST and SSH) results in poor estimate of the true ocean state (as depicted by the XBTs) and therefore poor forecast skill of subsurface information. In addition, the resulting circulation is highly sensitive to errors in the initial conditions and therefore the uncertainty (spread) in an ensemble forecast is high. Including the XBTs in the assimilation experiments improves the ocean state estimation in the vicinity of the XBT transects and also reduces the sensitivity to errors in the initial conditions resulting in a more skillful ensemble forecast. Motivated by this finding we explore the utility of including some pseudo-observations based on an empirical relationship between surface and subsurface information known as synthetic XBT (provided by CSIRO). The preliminary results show that better ocean state estimates and more skillful forecasts are obtained in the entire domain considered. Assimilation of Current Measurements Into a Circulation Model of Lake Michigan Z. Zhang(1), D. Beletsky(2), D. J. Schwab(3) and M. L. Stein(4) (1)Center for Integrating Statistical and Environmental Science, University of Chicago, Chicago, IL (2)CILER, SNRE, University of Michigan, Ann Arbor, MI (3)Great Lakes Environmental Research Laboratory, NOAA, Ann Arbor, MI (4)Department of Statistics, University of Chicago, Chicago, IL zpzhang@uchicago.edu We present a method for assimilating current measurements into a two-dimensional circulation model of Lake Michigan, which is based on the Princeton Ocean Model and is driven by observed winds. Measurements are assimilated by updating the stream function of the velocity field via kriging interpolation; therefore the physical constraint of a non-divergent flow is satisfied in the updated field. Moreover, the use of stream function avoids a component-wise treatment to the velocity vector. The coastal constraint is represented by the stream function being constant along the coastline, and is implemented by incorporating pseudo coastal data into the interpolation. This eliminates the need to construct complex spatial covariance models. The method also accommodates measurement uncertainties. Results show that the assimilations successfully meld measurements into simulations, with the effect propagating in space and time. In addition, the critical role of the covariance model in the method is demonstrated.

31