Automated glider tracking of a California undercurrent eddy using the extended Kalman filter

Abstract

Automated feature tracking and vehicle navigation have the potential to facilitate autonomous surveys of ocean eddies by increasing sampling quality and/or decreasing operator workload. During an observational campaign in late 2013 and early 2014, methods for automated tracking were used to direct multiple ocean gliders during persistent surveys of a California Undercurrent eddy in Washington and British Columbia, Canada, coastal waters over a 3-month period. Glider observations of depth-averaged currents in the ocean's upper kilometer and vertical separation of selected isopycnals were assimilated into a simple model describing eddy position, size, strength, and background flows using an extended Kalman filter. Though differing in detail from observations, results show the assumed eddy structure was sufficient to describe its essential characteristics and stably estimate eddy position through time. Forecast eddy positions and currents were used to select targets automatically to guide multiple gliders along transects through the eddy center as it translated. Transects performed under automated navigation had comparable or better straightness and distance from the eddy center when compared to navigation based on manual interpretation of the eddy scale and position. The tracking results were relatively insensitive to model choices at times when the eddy was well sampled, but they were more sensitive during sampling gaps and redundancies or rapid eddy translation. Overall, the results provide evidence that automated tracking and navigation are feasible with potential for widespread application in autonomous eddy surveys.