LAGRANGIAN DATA ASSIMILATION LADA OF SURFACE DRIFTERS TO
- Slides: 5
LAGRANGIAN DATA ASSIMILATION (LADA) OF SURFACE DRIFTERS TO SUPPORT OCEAN AND COUPLED MODEL INITIALIZATION Presenter: Luyu Sun (University of Maryland, College Park, US) Collaborators: Steve Penny, Matt Harrison, Laura Slivinski, Elaine Spiller and Amit Apte 1
What is a Surface Velocity Program (SVP)/ Global Drifter Program (GDP) surface drifter? • Surface float: • • • centered at 15 m, measure mixed layer currents in the upper ocean. OBS. : GPS location and T/S etc. Motivation of assimilating drifter locations • High-frequency. (~every hour) • Low cost. • Under-exploited (Assimilating TS measurement). • Trajectories are reflections of the underlying flow dynamics. • Augmented method (x=(x. F, x. D)) to assimilate drifter locations directly (Ide et al. 2012). 2 (Figures from the website of Global Drifter Program)
Application to double-gyre model (MOM 4 p 1): Free Run/TS obs only/La. DA with ALL (OSSEs) • • • 3 Model configuration: ¼ degree, 50 z*coordinate, constant wind stress. Analysis states: T/S/U/V. LETKF: 40 ens. Oneday DA window. • • The estimation of the state is improved not only at the surface, but throughout the upper 1000 m. Better than assimilating only TS measurement. (Sun and Penny 2019)
Application to the Gulf of Mexico Regional Model (MOM 6) Free run/Control/LETKF-La. DA (OSSEs) • Model configuration: 1/24 degree, 75 z*coordinate, time-variate atmos forcings: T 2 m, U 2 m, V 2 m, Precipitation rate etc. (20 CRv 3) • Analysis states: T/S/U/V. • LETKF: 26 ens. One-day DA window Forecast Error: average from h=1 m to h=1200 m • • 4 Impact of TS obs. (Free Run vs. Control): Positive effect in TS; Marginal effect in UV TS obs. + drifter locations (Control vs. La. DA): Positive effect in T and UV. Marginal effect in S.
Application to the Coupled Model System (MAOOAM) • QG-model: two layers of atmos. ; one layer of ocn. (De Cruz et al. 2016) • Analysis states: atmos/ocn streamfunctions, atmos/ocn T • Compare to Eulerian DA, i. e. assimilating “proxy” u, v: La. DA works better in estimating both atmos and ocn states. • Comparing Free Run/La. DA with only x, y/La. DA with x, y, T/DA with all ocean modes: 1) La. DA works better in estimating psi_o; 2) Assimilating x, y, T outperforms in
