The mass field is currently restricted to the TS climatology of the SABLAM region' We are examining

1
Implementation of the South Atlantic Bight
Limited Area Model
Brian Blanton1, Rick Luettich1, Eric Sills2,
Alfredo Aretxabaleta1, Harvey Seim1, Cisco
Werner1, Daniel Lynch3, Keston Smith3, Dennis
McGillicuddy4, Jim Nelson5
(1) University of North Carolina at Chapel Hill,
(2) North Carolina Supercomputer Center, (3)
Thayer School, Dartmouth College (4) Woods Hole
Oceanographic Institution, (5) Skidaway Institute
of Oceanography
Abstract
Far-field Prior
Climatology
Forecast
We are constructing a nowcast/forecast
operational system for the middle South Atlantic
Bight (SAB) of the eastern US coast. A suite of
finite element models is used to compute
solutions in a limited-area context, with tidal
and weather-band open-water boundary conditions
determined by a far-field model. Data available
for assimilation include water level from NOS
tide gauges and 2 operating ADCPs from the South
Atlantic Bight Synoptic Offshore Observation
Network (SABSOON). Data are inverted to deduce
open-boundary pressure corrections according to a
cost function. Mass field initialization is
currently restricted to climatology. The general
methodology has been used at sea and for
hindcasting purposes. The modeling system is
being developed using resources at the North
Carolina Supercomputer Center. This represents
its first application in an operational setting.
We show preliminary work on the implementation of
the SAB Limited-Area Model (SABLAM).
The figures below show a posteriori forecast
error growth of alongshelf ETA winds and
operational forecasted alongshelf depth-averaged
velocity at the R2 SABSOON tower. The presence
of the estuaries impacts the overall forecast
error by reducing tidal band misfit. Subtidal
misfit is related to ETA wind errors and
un-modeled processes (e.g., Gulf Stream). The
forecasts are based entirely on the prior initial
condition, the far-field forecasted boundary
conditions, and the ETA model forecasted wind
fields. Currently, the forecast is not impacted
by the weather band boundary condition
adjustments deduced by the data assimilation
system. One method to project the adjustments
into the forecast period is to require the
boundary adjustments to be correlated to the
prior wind stress.
A digital monthly temperature and salinity (TS)
climatology has been developed by objective
analysis of observations from the NODC database.
TS distributions are narrow offshelf and
unorganized on the shelf. Diagnostic solutions
forced by the monthly TS fields and COADS
monthly mean winds provide realistic mean shelf
break flow (Gulf Stream) and a general poleward
mid-shelf flow, particularly during summer
months when the winds and alongshelf pressure
gradient reinforce each other.
Tidal band
Tidal skill is sensitive to the resolution of
estuary and tidal inlet complex along the coast.
Significant improvement at both coastal and
mid-shelf stations (at semidiurnal frequencies)
is achieved by inclusion of estuaries. The left
figure shows the difference in the prior M2
solution with and without the estuaries. The
table gives the rms amplitude (m) and phase (deg)
error at coastal and shelf stations in the SAB.
Alongshelf ETA 10 meter wind error at
Charleston, SC
Alongshelf computed forecast error at R2 (w/
estuary)
Alongshelf computed forecast error at R2 (w/o
estuary)
Weather Band
Mayport FL, detided and band-pass (3-15 days)
observed and computed water level. Weather band
skill is generally .06 m rms misfit.
Future and Conclusions
Data Assimilation
Water level
Observations (left) of water level and
depth-averaged currents over the window 25-30
April, 2003, are gathered from the real-time
SABSOON and NOS networks and assimilated. The
detided mean over the prior (below, left) shows
weak poleward outer shelf flow. The mean over
the posterior indicates that stronger poleward
flow is needed to reduce the prior misfit. The
equatorward shelf break flow and circulation is
not realistic, and the shoreward flow along the
southern boundary indicates a preference of the
regularization toward smooth boundary conditions
along that boundary.

• The mass field is currently restricted to the TS
  climatology of the SABLAM region. We are
  examining density field estimation for the region
  through ensemble Kalman filter predictions and by
  using HYCOM/GODAE products for initialization.
• Additional work is being done to
• improve the cost function, including
• spatially dependent regularization
• weights and connection of boundary
• adjustments to the forecast period.
• The methodology requires significant
  computational resources. It currently takes
  about 12-24 hours on an IBM p690 to compute a 5
  day hindcast followed by an 84 hour forecast on
  the current domain. The balance between
  operational requirements (real-time versus
  hindcasting)
• poses constraints on the amount of data that can
  be assimilated, domain resolution, and
  regularization parameters.
• Current levels of performance may reflect the
  spatial distribution of the observations.
  Although the data are dense in time, they are
  sparse in space and indicate the need for broad
  data coverage, if possible. In light of sparse
  data coverage, the prior hindcastforecast may be
  preferable.
• Additional instrumentation is being deployed in
  the SABSOON tower array to enhance data coverage.
  This includes CODAR as well as ADCP packages.
  The need to gather more frequent observations of
  temperature and salinity, on the shelf is
  pressing, particularly in the vertical.

ADCP
Low-frequency
Mayport, FL, low-pass (15 days) observed and
computed water level. Reduced low-frequency skill
related to sea level variability induced by Gulf
Stream transport modulation. Skill varies
monthly, from 8-24 cm rms. Blaha (1984), Nobel
and Gelfenbaum (1992)
Operational System
Daily Schedule
Data Assimilation Loop
Prior BCs
Data Assimilation Window
Forecast Period
Hotstart ADCIRC BCs
RAMPUP
S
-NNZ
00Z
-MMZ
Turbulence Stratification
Weather band BC Adjustments
Cost Function Components
The boundary condition adjustments deduced by the
inversion (left) show most of the work being done
on the northern part of the domain. The
components of the cost function (right) need not
decrease monotonically.
Wind band Inverter
ICs
FORWARD
Met Fluxes
Model Connections
Tidal Inverter
Expected Misfit Regularization
The far-field model (ADCIRC-2DDI) is forced by
surface momentum flux from NCEPs ETA-12
(AWIP32) and tidal elevations from FES95 along
60W.
Misfit
DATA
AWIP 32
Cost function minimized in DA Loop
References
Misfit Observation Model
B. Blanton, A. Aretxabaleta, et al., Monthly
climatology of the continental shelf waters of
the South Atlantic Bight, JGR, in press, 2003.
J. Blaha, Fluctuations of monthly sea level as
related to the intensity of the Gulf Stream from
Key West to Norfolk, JGR, 1984. T. Lee, et al.,
Observations of a Gulf Stream frontal eddy on the
Georgia continental shelf, April 1977, DSR, 1981.
T. Lee, L. Atkinson, Low-frequency current and
temperature variability from Gulf Stream frontal
eddies and atmospheric forcing along the
southeast U.S. outer continental shelf, JGR,
1983. R. Luettich et al., ADCIRC An advanced
3-D circulation model for shelves, coasts, and
estuaries. Tech Report DRP-92-6, 1992.
D. Lynch, C. Hannah, Inverse model for
limited-area hindcasts on the continental shelf,
JTech, 1999. D. Lynch, F. Werner, 3-D
hydrodynamics on finite elements Part II
Nonlinear time-stepping model, IJNMF, 1991. M.
Noble, G, Gelfenbaum, Seasonal fluctuations in
sea level on the South Carolina Shelf and their
relationship to the Gulf Stream, JGR, 1992. H.
Seim, Implementation of the South Atlantic Bight
Synoptic Offshore Observational Network,
Oceanography, 2000.
Boundary Regularization Terms
ADCIRC provides prior estimates of tidal and
remotely forced boundary elevations to the
limited-area model system (QUODDY/TRUXTON/CASCO)
Boundary Elevation
Boundary Slope
Boundary Tendency
Velocity
Elevation