5.3 Idealized OSSE Evaluation of Ocean Profiler Assimilation for Improving Mesoscale Ocean Analysis and Prediction

Tuesday, 9 January 2018: 11:00 AM
Room 14 (ACC) (Austin, Texas)
George R. Halliwell Jr., NOAA/AOML, Miami, FL; and V. H. Kourafalou, R. Atlas, and M. Le Henaff

The ability of ocean observing systems to improve the analysis and prediction of mesoscale variability is rigorously assessed in the western North Atlantic Ocean using the OSSE (Observing System Simulation Experiment) framework. Taking advantage of the high resolution, three-dimensional representation of truth provided by the Nature Run, impacts are quantitatively assessed as a function of wavenumber using spectrum analysis. Satellite altimetry is the only existing ocean observing system designed to substantially correct the ocean mesoscale. However, OSSE analysis results demonstrate that altimetry correction is only significant over wavelengths > 200 km, even with four available altimeters. Idealized OSSEs are then conducted to quantitatively assess how densely ocean Conductivity-Temperature-Depth (CTD) profilers should be deployed to provide significant additional correction to mesoscale structure, particularly over smaller mesoscale wavelengths not corrected by altimetry. Significant additional analysis error reduction resulting from daily profiler assimilation over a four-month time interval is confined to wavelengths that exceed the Nyquist value, defined as twice the nominal horizontal separation distance of the deployed profilers. Consequently, significant error reduction over wavelengths between 100 and 200 km requires that profilers be separated by ~0.5 degrees. Adding profiler assimilation to other ocean observing system components further increases mesoscale predictability time scales by roughly one-third, which amounts to an additional 7-10 days for ocean dynamic height and 2-3 weeks for upper-ocean thermal fields. The dependence of mesoscale error reduction on key design features of the ocean data assimilation system, particularly the estimated background error covariance matrix and the observation localization procedure, is demonstrated.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner