Abstract's details

Constrained scales in ocean forecasting

Gregg Jacobs (Naval Research Laboratory, United States)


Joseph D'Addezio (Naval Research Laboratory, USA); Brent Bartels (Perspecta, USA); Pete Spence (Perspecta, USA); Max Yaremchuk (Naval Research Laboratory, USA); Bob Helber (Naval Research Laboratory, USA); Scott Smith (Naval Research Laboratory, USA); Clark Rowley (Naval Research Laboratory, USA)

Event: 2019 Ocean Surface Topography Science Team Meeting

Session: Science III: Mesoscale and sub-mesoscale oceanography

Presentation type: Type Oral

Contribution: PDF file


As we look to understand the submesoscale circulation through SWOT in combination with numerical models, we must first understand the scale limitations imposed by existing altimeter space-time resolution on present model systems. The approach is through evaluation relative to dense real world data provided by 1000 drifters deployed in January 2016 as part of the LAgrangian Submesoscale ExpeRiment (LASER) campaign in the Gulf of Mexico that persisted for 3 months and sampled eddy features over a wide range of mesoscale and submesoscale features. All available satellite altimeter observations are assimilated into a 1 km resolution ocean model. In this situation, the ocean model generates features that are smaller than the resolution of existing altimeter data. We define constrained scales as those in which the forecast system has skill, and we determine these by successively filtering small-scale variability from model experiments to reach a minimum error relative to the ground truth data that were not assimilated. We also vary the decorrelation scales of the assimilation system (a second order auto-regressive correlation function) to determine the decorrelation scale that produces the smallest forecast trajectory errors. We find the constrained scales are larger than those defined by a Gaussian filter with e-folding scale of 58 km or ¼ power point of 220 km. The decorrelation scale of 36 km used in the assimilation provides lowest trajectory errors. Filtering unconstrained variability from the model solutions reduces 24-hour trajectory errors by 20%. Multi-scale analyses have been proposed to best assimilate future SWOT observations. These systems update the background successively by using existing nadir altimeter observations with SWOT to first correct the mesoscale and then correct the smaller submesoscale field. The results here provide a breakpoint in assimilation length scale separating the first and second steps of the multi-scale analysis procedure.

Oral presentation show times:

Room Start Date End Date
The Forum Wed, Oct 23 2019,16:30 Wed, Oct 23 2019,16:45
Gregg Jacobs
Naval Research Laboratory
United States