http://www.aoml.noaa.gov/phod/dac/dac_meanvel.php

Satellite-tracked SVP drifting buoys (Sybrandy and Niiler, 1991; Niiler, 2001) provide observations of near-surface circulation at unprecedented resolution. In September 2005, the Global Drifter Array became the first fully realized component of the Global Ocean Observing System when it reached an array size of 1250 drifters. A drifter is composed of a surface float which includes a transmitter to relay data, a thermometer that reads temperature a few centimeters below the air/sea interface, and a submergence sensor used to detect when/if the drogue is lost. The surface float is tethered to a holey sock drogue, centered at 15 m depth. The drifter follows the flow integrated over the drogue depth, although some slip with respect to this motion is associated with direct wind forcing (Niiler and Paduan, 1995). This slip is greatly enhanced in drifters that have lost their drogues (Pazan and Niiler, 2000). Drifter velocities are derived from finite differences of their position fixes. These velocities, and the concurrent SST measurements, are archived at AOML's Drifting Buoy Data Assembly Center where the data are quality controlled and interpolated to 1/4-day intervals (Hansen and Herman, 1989; Hansen and Poulain, 1996).

In this study, 6h winds were interpolated onto the drifter positions and used to estimate and remove the slip (Niiler and Paduan, 1995; Pazan and Niiler, 2000; Laurindo et al., 2017). All velocities and SSTs were lowpassed at 1.5 times the local inertial period, or five days if that is smaller, to remove high frequency variability (diurnal, tidal, inertial). If 1.5 times the local inertial period is shorter than one day, the lowpass is done at one day.

Drifters sample regions of the ocean inhomogeneously, which can cause aliased time-mean values if strong seasonal or interannual variations are neglected. To address the seasonal cycle, Lumpkin (2003) developed a methodology to simultaneously decompose the drifter observations into time-mean, seasonal and eddy components using a Gauss-Markov approach that produces formal error bars on all components. Lumpkin showed that this methodology produces significantly different results than standard bin averaging. This methodology was further developed and evaluated using SST observations and products, and simulated drifters in the MICOM model (Lumpkin and Garraffo, 2005). The method produces significantly improved estimates of the mean currents and SST, and simultaneously provides the annual and semiannual amplitudes and phases at a nominal resolution of one degree squared.

To address inhomogeneous interannual sampling associated with ENSO, Johnson (2001) added a component proportional to a five-month lowpassed Southern Oscillation Index, and estimated components in elliptical bins with axes scaled and oriented using the residual variability (i.e., the eddy fluctuations).

In Lumpkin and Johnson (2013), the methodologies of Lumpkin (2003) and Johnson (2001) are combined. The observations are projected onto a time mean, annual and semiannual, SOI, and spatial gradient components within elliptical bins scaled and oriented using eddy fluctuations, and error bars are estimated for all terms.

In Laurindo et al. (2017), slip correction using a spatially-varying coefficient was introduced in order to exploit velocity information from undrogued drifters. This approximately doubles the number of available observations, which allowed the bin sizes to be considerably reduced (one degree radius circles). In addition, 1D spatial fitting was done in the direction that maximizes the explained variance, allowing for greater resolution of cross-stream velocity gradient. These two changes result in significantly better spatial resolution of ocean current structure. In addition, Laurindo et al. examined how the formal standard errors compare to actual errors when using a "toy" dataset of surface currents derived from AVISO and subsampled at the drifter observation locations and times; they found that formals errors underestimate actual errors by approximately a factor of 2.

When these more advanced methodologies are applied to the modern data set of tropical Atlantic drifter observations, many features of the near-surface circulation become apparent which were not resolved by older ship-drift-based climatologies or by SEQUAL/FOCAL drifter trajectories (Lumpkin and Garzoli, 2005). In the Hawaiian Island region, the thousand-kilometer long island wake is revealed at unprecedented detail (Lumpkin and Flament, 2013) Zonally elongated mid-ocean striations are resolved in the zonal currents of all major ocean basins (Laurindo et al., 2017).