Constraining the mesoscale field
Gregg Jacobs(1) , Jay Shriver(1) , Kirk Whitmer(1) , Ole Martin Smedstad(2) , and Harley Hurlburt(1)
Naval Research Laboratory,
NRL Code 7320,
Stennis Space Center, MS 39529,
(2) Planning Systems Inc, PSI, Stennis Space Center, MS 39529, United States
The history of altimeter sensors provides a valuable opportunity to examine the challenges in observing the ocean mesoscale features and using those observations to constrain a numerical forecast model. A wide range of twin model experiments have been conducted using 1 or more altimeters to gauge accuracy in reconstructing ocean sea surface height. Many of these involve optimal interpolation systems that are inherently linear. However, the ocean dynamics are nonlinear, and small errors grow in time. Additionally, numerical models often have biases in circulation or energy of the mesoscale field. These pitfalls are avoided in this study by using direct satellite observations. To evaluate the accuracy with which multiple altimeters can constrain the nonlinear growth of the mesoscale field, we first use a 3 year assimilative 1/32 degree resolution global ocean model assimilating Jason-1, ENVISAT and GFO observations from 2001-2003. The model is the NRL Layered Ocean Model (NLOM) forced by NOGAPS surface stress and heat fluxes. Additional assimilation experiments are performed using only GFO observations and then only GFO and ENVISAT observations. Finally, a third experiment is performed assimilating all three altimeter data sets. The key difference between the two experiments assimilating 3 altimeter data sets is that they start from slightly different initial conditions. Due to the nonlinear behavior of the mesoscale flow, small perturbations grow in time. One of the three altimeter assimilation experiments is declared "truth". The difference between the two experiments using 3 altimeters provides a measure of the ability of these systems to constrain the mesoscale.
Individual snapshots or spatial plots of the RMS error indicate the largest errors occur in the western boundary currents and the associated eastward extensions. The error variance to signal variance is examined spatially for the 1 through 3 sensor assimilation cases. With only one altimeter, the unresolved variability is in the typical range of 75% to 100%. With 3 altimeters, the results are much improved, particularly in mid latitudes. Zonal average error statistics reveal the improvement from of additional sensors as well as the importance of the observations. There are instances in time and areas in the world where a non-assimilative wind forced model has higher skill than the experiment using 1 altimeter. In these areas and times the dominant processes are wind driven deterministic variations and occur primarily in the equatorial regions.