Global Ensemble Generation Using Perturbed Observations in the Navy Coupled Ocean Data Assimilation System (NCODA)

Clark David Rowley1, Sergey Frolov2, Michael Stokes3, Patrick J Hogan4, Mozheng Wei4 and Craig H Bishop5, (1)US Naval Research Laboratory, Oceanography Division, Washington, DC, United States, (2)US Naval Research Laboratory, Monterey, CA, United States, (3)Northshore High School, Slidell, LA, United States, (4)Naval Research Laboratory, Oceanography Division, Stennis Space Center, MS, United States, (5)Naval Research Lab, Monterey, United States
Abstract:
A perturbed-observation analysis capability has been developed for the Navy Coupled Ocean Data Assimilation system (NCODA). The resulting analysis is used to represent analysis error in the initial conditions of a global ocean forecast ensemble using the Hybrid Coordinate Ocean Model (HYCOM).

For cycling with HYCOM, the NCODA system performs a 3D variational analysis of temperature, salinity, geopotential, and vector velocity using remotely-sensed SST, SSH, and sea ice concentration, plus in situ observations of temperature, salinity, and currents from ships, buoys, XBTs, CTDs, profiling floats, and autonomous gliders. Sea surface height is assimilated through synthetic temperature and salinity profiles generated using the Modular Ocean Data Assimilation System (MODAS) historical regression database with surface height and surface temperature as inputs.

Perturbations to the surface observations use random samples from a normal distribution scaled by the observation error standard deviation, which combines estimates of instrument and representation error. Perturbations to the synthetic profiles are generated by supplying the perturbed surface inputs to the MODAS system, resulting in correlated profile changes with vertical correlations associated with historical uncertainty about thermocline depth and gradients. For in situ profile observations, representation error is much larger than instrument error, so a technique is implemented to create correlated perturbations associated with large, mesoscale errors.

Initial results from a cycling regional analysis show the resulting analysis perturbations have scales and amplitudes consistent with short term forecast error covariances. Results using the perturbed observation analysis in regional and global cycling forecast systems will be presented.

Back to: Developments and Ocean Applications of Data Assimilation, Uncertainty, and Sensitivity Analyses I