Docstoc

TOMS

Document Sample
TOMS Powered By Docstoc
					                  A Community
                Terrain-following
                 Ocean
                   Modeling
                    System

Hernan G. Arango, Rutgers University
     (arango@imcs.rutgers.edu)

   Tal Ezer, Pricenton University
    (ezer@splash.princeton.edu)

                    FTP File: TOMS.tar
            COLLABORATORS
 Bennett et al. (FNMOC; OSU)
 Chassignet / Iskandarani et al. (RSMAS)
 Cornuelle / Miller (SIO)
 Geyer (WHOI)
 Hetland (TAMU)
 Lermusiaux (Harvard)
 Mellor (Pricenton)
 Moore (U. Colorado)
 Shchepetkin (UCLA)
 Signell (SACLANT; USGS)
     OTHER COLLABORATORS

 Chao / Song (JPL)

 Preller / Martin (NRL)



 Naval Operational Community

 POM Ocean Modeling Community

 ROMS / SCRUM Ocean Modeling Community
                OBJECTIVES
 To design, develop and test an expert ocean
  modeling system for scientific and operational
  applications
 To support advanced data assimilation strategies
 To provide a platform for coupling with operational
  atmospheric models (like COAMPS)
 To support massive parallel computations
 To provide a common set of options for all coastal
  developers with a goal of defining an optimum
  coastal/relocatable model for the navy
                 APPROACH
 Use state-of-the-art advances in numerical
  techniques, subgrid-scale parameterizations, data
  assimilation, nesting, computational performance
  and parallelization
 Modular design with ROMS as a prototype
 Test and evaluate the computational kernel and
  various algorithms and parameterizations
 Build a suite of test cases and application
  databases
 Provide a web-based support to the user
  community and a linkage to primary developers
           CHALLENGE


“The complexity of physics, numerics,
  data assimilation, and hardware
  technology should be transparent
     to the expert and non-expert
               USER”
     TOMS KERNEL ATTRIBUTES
 Free-surface, hydrostatic, primitive equation
  model
 Generalized, terrain-following vertical coordinates
 Boundary-fitted, orthogonal curvilinear, horizontal
  coordinates on an Arakawa C-grid
 Non-homogeneous time-stepping algorithm
 Accurate discretization of the baroclinic pressure
  gradient term
 High-order advection schemes
 Continuous, monotonic reconstruction of vertical
  gradients to maintain high-order accuracy
            Dispersive Properties of Advection

              5/2                             Parabolic
                                               Splines
               2
K(k) • x



                                      10
                                  6               Vs
              3/2                         8
                                      4         Finite
               1                  2            Centered
                                              Differences
              1/2


                      /4   /2       3/4
                            kx
          TOMS SUBGRID-SCALE
           PARAMETERIZATION

 Horizontal mixing of tracers along level, geopotential,
  isopycnic surfaces

 Transverse, isotropic stress tensor for momentum

 Local, Mellor-Yamada, level 2.5, closure scheme

 Non-local, K-profile, surface and bottom closure
  scheme
        TOMS BOUNDARY LAYERS
 Air-Sea interaction boundary layer from COARE
  (Fairall et al., 1996)
 Oceanic surface boundary layer (KPP; Large et al.,
  1994)
 Oceanic bottom boundary layer (inverted KPP;
  Durski et al., 2001)
Boundary Layer Schematic

                                           1.   ABL
     L
     o                                     2.   SBL
     n
     g
     w                         O
                                           3.   BBL
                                   H E
     a
     v
                           H
                               O
                                       v
                                       a
                                           4.   WCBL
     e                                 p
                           H       H
        TOMS BOUNDARY LAYERS
 Air-Sea interaction boundary layer from COARE
  (Fairall et al., 1996)
 Oceanic surface boundary layer (KPP; Large et al.,
  1994)
 Oceanic bottom boundary layer (inverted KPP;
  Durski et al., 2001)

 Wave / Current / Sediment bed boundary layer
  (Styles and Glenn, 2000)
 Sediment transport
             TOMS MODULES

 Lagrangian Drifters (Klinck, Hadfield)

 Tidal Forcing (Hetland, Signell)
Gulf of Maine M2 Tides




                          Surface
                         Elevation
                           (m)
             TOMS MODULES

 Lagrangian Drifters (Klinck, Hadfield)

 Tidal Forcing (Hetland, Signell)

 River Runoff (Hetland, Signell, Geyer)
                  Hudson River Estuary
                                                30

             -5
                                                25




                                                     Salinity (PSS)
            -10
Depth (m)




                                                20

            -15
                                                15

            -20
                                                10

            -25
                                                5

                  5   10       15     20   25
                           Distance (km)
             TOMS MODULES

 Lagrangian Drifters (Klinck, Hadfield)

 Tidal Forcing (Hetland, Signell)

 River Runoff (Hetland, Signell, Geyer)

 Biology Fasham-type Model (Moisan, Shchepetkin)

 EcoSim Bio-Optical Model (Bissett)
              TOMS TESTING

 Systematic evaluation of numerical algorithms via
  robust test problems

 Data/Model comparisons

 Study optimal combination of algorithmic options
  for various coastal applications

 Documentation of testing procedures
          TOMS CODE DESIGN
 Modular, efficient, and portable Fortran code
  (F77+, F90)
 C-preprocessing managing
 Multiple levels of nesting
 Lateral boundary conditions options for closed,
  periodic, and radiation
 Arbitrary number of tracers (active and passive)
 Input and output NetCDF data structure
 Support for parallel execution on both shared- and
  distributed -memory architectures
     TOMS PARALLEL DESIGN
 Coarse-grained parallelization
      PARALLEL TILE PARTITIONS




8x8




                    Ny
                         } Nx
    TOMS PARALLEL DESIGN
 Coarse-grained parallelization
 Shared-memory, compiler depend directives
  MAIN (OpenMP standard)
 Distributed-memory (MPI; SMS)
 Optimized for cache-bound computers
 ZIG-ZAG cycling sequence of tile partitions
 Few synchronization points (around 6)
 Serial and Parallel I/O (via NetCDF)
 Efficiency 4-64 threads
    TOMS DATA ASSIMILATION
 Nudging
 Optimal Interpolation (OI)
 Tangent linear and Adjoint algorithms
 4D VARiational data assimilation (4DVAR) and
  Physical Statistical Analysis System (PSAS)
  algorithms
 Inverse Ocean Modeling System (IOMS)
 Ensemble prediction platform based on singular
  value decomposition
 Error Subspace Statistical Estimation (ESSE)
ESSE Flow                                                                          ^
                                                                                  Y0/N                   ESSE Smoothing via
                                                                                                                                                +

 Diagram
                                                                                                        Statistical Approximation
                                                                                  E0/N                                                         +
                                                                                  0/N
         Field         +          Central                                                                           -         -                  Performance/
     Initialization   Y0                         ^
                                                 Ycf(-)
                                  Forecast                                                                                                         Analysis
                                                                       Most
                                                                                                                                                   Modules
                                                                     Probable
                                                                     Forecast
                                                                       ^
                                                                      Y mp(-)               Shooting
                                                                                                             Synoptic       + Measurement
                                                                                                               Obs                                  A Posteriori
 Historical,                                                                                                                       Model
                                                        Sample                                                          +                            Residules
  Synoptic,                                            Probability                                                                         -           dr (+)
 Future in                                              Density                  Select                 -    Measurement
Situ/Remote                                                                      Best                          Model
Field/Error                                                                     Forecast
                                                                                                         Data             Measurement
Observations                                                                                                                                          OA via
                                                          Mean                                         Residuals             Error
    d0R0                                                                                                    ^                                         ESSE
                                                                     Ensemble                           d-CY(-)            Covariance
                                                                       Mean
        Options/                                                        ^
                                                                     eq{Y j(-)}                                 +                   +
                                                                                                                                                      Gridded
       Assumptions                                                                                                      Minimum
                                                                                  ^                                                                   Residules
                                                                                  Y(-)                                   Error
                                                                                                            +           Variance
                                                                                                                                               ^
                                           j=1
                                                 Y0
                                                  1                   ^1
                                                                      Y -
                                                                           -                                                                   Y(+)
                                                                                                                                                            ^
                                                                                                                                                            Y(+)
                              +                        Scalable                        +                            Within Error
                           Perturbations         Y0
                                                  j     Parallel      ^ -          +       SVDp   E(-)               Subspace
 Error Subspace                                                       Yj                          (-) +
  Initialization       +    +                          Ensemble          -
                                                                      Yq -
                             +/-                 Yq    Forecast       ^                                              (Sequential
                      E0                   j=q    0
                                                                                                                    processing of
                      0
                                                                           Normalization                            Observations)
                              uj(o,Ip)                Continuous
                           with physical                                                                                                   Adaptive
                                                      Time Model
     Key                    constraints               Errors Q(t)     Convergence
                                                                                                                                            Error
                                                                                                                                           Subspace            Ea(+)
                                                                        Criterion                           Peripherals           E(+)
    Field                                                             Continue/Stop                                               (+)
                                                                                                                                           Learning            a(+)
                                                                                                             Analysis
    Operation                                                           Iteration                            Modules
    Assumption                                                          Breeding
     PRESSURE GRADIENT FORCE

 Density Jacobian Class (Blumberg and Mellor,
  1987; Song 1998; Song and Wright 1998)
    More Accurate
    Error vanishes with linear density profiles
 Pressure Jacobian Class (Lin 1998; Shchepetkin
  and McWilliams, 2001)
    JEBAR consistent
    Conserve Energy
Seamount Test Case




(64 x 64 x 20) dx = dy = 8 km
    Second Order Advection Scheme
   Models with 2nd order advection scheme
            POM            ROMS


Surface
Elevation
Anomaly




Stream
Function
Anomaly
      Advection Schemes in ROMS
           (Seamount Case)
                                     V




Second Order
  Centered
                Third Order
               Upstream Bias

                               Fourth Order
                                 Centered
           Pressure Gradient Errors

            POM                   POM
U (cm/s)                       (6th order)


              ROMS
V (cm/s)




                     X (km)
Relative CPU per time step


Percentage
           RESULTS (YEAR 1)

 Build TOMS from ROMS prototype
   Mellor-Yamada, level 2.5
   Passive and active open boundary conditions
   Tidal forcing
   River runoff
   Lagrangian drifters
   Data assimilation
 Inter-comparison between POM and ROMS
   Evaluation of time-stepping, advection, and
     pressure gradient algorithms
 Initial development of TOMS web site
Initial web page: www.aos.princeton.edu/WWWPUBLIC/ezer/TOMS
            TRANSITION PATHS

 To Be Determined !!!


 Potential Users:
            NAVO
            FNMOC
            NOAA
            USCG
                PUBLICATIONS

 Chassignet et al., 2000: Damee modeling review
 Ezer, 2000: Mixed-layer evaluation
 Ezer and Mellor, 2000: POM Damee application
 Haidvogel et al., 2000: ROMS Damee application
 Malanotte-Rizzoli et al., 2000: ROMS Damee
 Mellor, 2001: Improved turbulence scheme
 Mellor et al., 2001: Generalized vertical coordinate

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:6
posted:10/18/2011
language:Basque
pages:34