Darryl Williams

Influence of the Wind on the Motion of Surface Drifters: Application of a Data Assimilative Model to the Outer Scotian Shelf

Thesis Approved 1996

The dynamics of the upper meter or so of the ocean are complicated by the interaction of current shear, wind shear and the action of waves. An effective model for the motion of surface drifters should take into account: (i) Stokes drift; (ii) the leeway effect, due to wind acting on exposed areas of the drifter; (iii) motion in the wind drift layer, which consists of a 5cm thick surface sublayer above a logarithmic sublayer of about one meter thickness. Drifter trajectories were observed on the outer Scotian Shelf in May 1993. Three types of drifter were deployed: one was drogued at 20m, the other two were surface drifters with different drafts and mast areas. To quantify the various components of the drifter motion, the water column was partitioned into separate layers leading to the following expression for drifter velocity:

Udrifter = Udeep + UEkman + UStokes + Ulee + Ulog + Uerror

Udeep is the background flow and UEkman is the velocity near the top of the Ekman layer. The remaining suB.Sc.ripts refer to the surface corrections for Stokes drift, the leeway effect, and the logarithmic sublayer (the effect of the surface sublayer was assumed negligible). The calculation of UStokes, Ulee and Ulog is straightforward and based on physical principles. The background flow (below 20m) was modelled using the adjoint method to assimilate current data from moored instruments and a shipborne acoustic doppler current profiler. UEkman was modelled statistically, using complex regression. The background flow correction gave the greatest reduction in the variance between model and observation of the drifter velocities. The variance of the error was 53% that of the observed (i.e. uncorrected) velocity variance. Subtracting the surface correction factors further reduced the residual variance to 45%. Complex regression of the corrected drifter velocity on the wind stress reduced the residual variance to 37%.