Generic MW Scientific Testbed Results

1
Global ocean state estimation
Ross Hoffman (for Rui M Ponte) Atmospheric and
Environmental Research, Inc.
group at MIT-AER Carl Wunsch (PI),
Constantinos Evangelinos, Gael Forget,
Patrick Heimbach, Charmaine King, Matthew
Mazloff, Diane Spiegel Sergey Vinogradov, Nadya
Vinogradova IOOS MAST Workshop, Arlington,
Virginia, July 2008
2
Outline

• Background and goals of the ECCO-GODAE project
  (MIT-AER component)
• Brief description of present global state
  estimates
• Observations
• Ocean model
• Optimization/optimal control problem
• Some science and research applications
• What lies ahead

3
An early view
Taken from C. Wunsch, in "A Celebration in
Geophysics and Oceanography 1982. In Honor of
Walter Munk on his 65th birthday." C. Garrett
and C. Wunsch, Eds., SIO Reference Series 84-5,
March 1984
4
25 years later
WOCE
Argo
T/P, Jason
GRACE
Feasible and efficient synthesis of all available
information in data and models
5
Ocean state estimation in a climate context

• Global estimates that best represent our
  knowledge of ocean physics and dynamically
  interpolate all the information available in the
  disparate data sets
• Consistent and stable descriptions of the
  three-dimensional, large-scale ocean circulation
  at seasonal and longer time scales
• Preserving climate parameters of relevance such
  as heat and freshwater for balanced diagnostics
  of property fluxes, exchanges with the
  atmosphere, and climate dynamics

6
ECCO-GODAE project
Estimating the Circulation and Climate of the
Ocean - Global Ocean Data Assimilation Experiment

• Approaching 10 years of efforts in global ocean
  state estimation
• Initially a consortium of MIT, Scripps
  Institution of Oceanography and the Jet
  Propulsion Laboratory
• Continued as ECCO-GODAE since May 2004 with new
  partners and various objectives
• MIT-AER component dedicated to combining most
  available observations and a state-of-the-art
  ocean model using advanced least-squares
  optimization methods to provide a kinematically
  and dynamically consistent estimate of the state
  of the ocean for the modern instrumental period

7
Observations
sum 100(obs) 800(forcing) million
individual elements
8
Southern Elephant Seals as Oceanographic Samples
(SEaOS)

• CTD-type observations from seals in SO

• Sea Mammal Research Unit,
• University St. Andrews, UK,
• British Antarctic Survey
• Courtesy M. Meredith

9
Gravity Recovery And Climate Experiment(GRACE)

Estimates of global mean bottom pressure values
can serve as constraints on mean freshwater
flux into the oceans (seasonal cycle, other
timescales?)
Ponte, Quinn, Wunsch Heimbach (2007, GRL)
10
Time-dependent GRACE
GRACE (JPL) GRACE (GFZ) ECCO-GODAE
Bottom pressure seasonal cycle (annualsemiannual)
average over 50S-60S
Ponte, Quinn, Wunsch Heimbach (2007, GRL)
11
The MIT general circulation model (MITgcm)
Parallel implementation of a general-purpose
grid-point algorithm for a Boussinesq or
non-Boussinesq fluid, hydrostatic or
non-hydrostatic, in curvilinear coordinates

• z-level or pressure vertical coordinates (ocean -
  atmosphere isomorphism)
• nonlinear free surface
• finite-volume formulation with partial cells
• various subgrid scale physical parameterization
  schemes
• thermodynamic/dynamic sea-ice model
• ocean biogeochemical model
• global grid topology

12
ECCO-GODAE setup

• (Version 2)
• 1 degree horizontal resolution
• covering 80N to 80S
• 23 vertical levels
• subgrid scale parameterizations
• covers 1992 to 2006 (soon through 2007)

• forcing by 6-hourly atmospheric analysis
  (air-sea fluxes of heat,
• freshwater and momentum)
• Version 3 forcing with atmospheric state,
  includes sea ice model

13
Least-squares optimization
14
Objective (cost) function
Initial conditions Model-data misfits Surface
forcing control variables

• Large efforts needed to best define all the
  weights and
• avoid over-fitting or under-fitting the
  observations
• Currently using diagonal matrices

15
Optimization procedure

• Run model forward in time for the full period
• Evaluate the cost function
• Use adjoint model to determine sensitivity of the
  cost function to all the control parameters
• Adjust the control parameters in the minimization
  direction
• Iterate procedure until a satisfactory solution
  is achieved
• Simple in principle, not so easy in practice.

Dont try this at home!
16
Control variables

• 3-dimensional initial conditions
• temperature, salinity
• 2-dimensional 2-day averaged surface forcings
• Version 2
• heat flux, freshwater flux
• zonal/meridional windstress
• Version 3
• surface air temperature, specific humidity,
  precipitation
• downwelling shortwave radiation
• zonal/meridional wind speed
• 3-dimensional internal model parameters
  (experimental)
• mixing coefficients (Stammer, 2005, JPO)
• eddy stress parameterization (Ferreira, Marshall
  Heimbach, 2005, JPO)
• bottom topography (Losch Heimbach, 2007, JPO)

17
A large-scale optimal control problem
18
After all the hard work

• a solution for the period of reference
  (1992-present) that is close in a least-squares
  sense to all available in situ and satellite
  observations and fully consistent with the model
  dynamics and the atmospheric forcing fields,
  without any spurious sinks or sources of
  momentum, heat or other properties

19
Altimeter ECCO-GODAE
20
Comparing to data (cost analysis)
Argo floats (temperature)
Altimeter (sea level)
21
Decadal sea level trends (1993-2004)
ECCO-GODAE v2.216
Altimeter
weights

• Altimeter data

Ponte, Wunsch Stammer (2007, JAOT)
Wunsch, Ponte Heimbach (J. Climate, 2007)
22
Effects of optimization
after optimization
before optimization
23
Decadal steric patterns

• Vertical partition (steric trends)
• Temperature effects

??T
Wunsch, Ponte Heimbach (J. Climate, 2007)
24
Trend in global mean sea level
Wunsch, Ponte Heimbach (J. Climate, 2007)
25
Sea surface temperature tendency

• ECCO-GODAE v2.216 Data

26
What controls sea surface temperature?
Annual cycle
27
Ocean dynamics and SST

• Ratio of advection to
• SST tendency

• Importance of advection in tropical areas and
  western boundary currents
• Advection becomes increasingly important at
  longer time scales

28
Atlantic Meridional Overturning
29
Decadal variations in Atlantic poleward heat and
mass transports
3-month averaged volume transport (106 m3/s1 SV)
Wunsch Heimbach (2006, J. Phys. Oceanogr.)

• complex structure of variability in space and
  time
• serious sampling/aliasing issues expected

30
Science and research applications

• Decadal sea level variability in relation to
  steric and mass contributions, forcing, etc.
  (Wunsch, Ponte Heimbach, J. Climate, 2007)
• Decadal variations in Atlantic poleward heat and
  mass transports and changes in meridional
  overturning (Wunsch Heimbach, 2006, J. Phys.
  Oceanogr.)
• Climatological seasonal cycle (Vinogradov, Ponte,
  Heimbach Wunsch, J. Geophys. Res., 2008)
• Mean climatology for the modern instrumental
  period (In preparation)

31
Other users and applications

• Long-term biogeochemical tracer calculations (S.
  Khatiwala, LDEO)
• MIT Darwin Project (Follows et al.)
• marine ecosystem modeling ocean biogeochemical
  cycle
• CLIMODE CLIvar MOe Water Dynamic Experiment
  (Marshall et al.)
• Investigation of sub-tropical 18o mode water
  formation dynamics
• CODAE Central California ODAE (C. Edwards, UCSC)
• Study of the Monterey Bay Central California
  coast circulation
• Mean Dynamic Topography uncertainties (F.
  Fossepoel, IMAU, Utrecht)
• Uncertainty estimate in preparation of the
  Gravity Field and Steady-State Ocean Circulation
  Explorer (GOCE) satellite mission
• Global gravitational model development (N.
  Pavlis, SGT)
• Sponsored in part by GRACE and National
  Geospatial Intelligence Agency (NGA)
• Ocean mass changes and Earth rotation (J.
  Nastula, Polish Academy of Sciences)
• Ocean circulation magnetic field signals (Manoj
  et al., GFZ, Potsdam)
• Liming the ocean increases CO2 absorption (D.
  Harvey, U. Toronto)
• Uses upwelling velocities to investigate pCO2
  increases in response to CaCO3
• Thermodynamic analysis of ocean circulation (N.
  Wienders, FSU J. Nycander, MISU, Sweden)
• Streamfunction-based calculations of MOC
  energetics
• Global instability calculations (Shafer Smith,
  NYU)

32
Current work on version 3

• Addition of new data types
• Tide gauges
• Coastal altimetry
• Bottom pressure (in situ and GRACE)
• Surface drifters
• SST (daily/weekly currently monthly)
• Forcing
• Inclusion of atmospheric pressure
• Time-varying runoff
• Cost function
• New implementation of constraints on global mean
  sea level, net surface freshwater flux
• Adjustment of weights

33
Towards ECCO-GODAE version 4

• Various forward/adjoint model developments
• global grid and topography
• nonlinear free-surface
• real surface freshwater flux
• non-Boussinesq formulation
• adjoint of global grid topologies
• Data
• Arctic datasets
• port cost function on new grid topology
• Cost function
• new control parameters (eddy coefficients)
• improved weights, nondiagonal elements

34
Summary

• Framework for global ocean state estimation using
  most available data, MITgcm and advanced
  optimization procedures and focusing on
  large-scale, climate variability
• Dynamically-consistent solutions of MITgcm that
  are a best fit to altimetry, hydrography,
  scatterometry and other data
• Estimates publicly available at
  www.ecco-group.org for science and research on
  the nature of the large-scale ocean circulation
  and its role in climate (decadal variability, sea
  level rise, meridional overturning circulation,)
• Evolving solutions as new datasets become
  available, model formulations are improved,
  estimates of weights are adjusted, and
  optimization continues

35
ECCO-GODAE products

• http//www.ecco-group.org/