A new global surface drifter dataset at hourly resolution

NOAA’s Global Drifter Program maintains a global array of about 1250 surface drifting buoys. These drifting buoys or drifters, primarily tracked by the Argos satellite system, provide observations of near-surface drift at uneven time intervals. The current GDP product consists of drifter positions and velocities estimated at 6-hour intervals by kriging, an interpolation method that utilizes a historic model of the drifter position variogram. The 6-hourly product has provided decades of valuable information on ocean circulation at subinertial time scales. In 2005, the time interval between consecutive drifter positions decreased to 1-2 hours on average due to enhanced satellite coverage. As a result, the post-2005 GDP record contains high-frequency information about oceanic variability and air-sea processes which are smoothed or entirely removed by 6-hourly kriging.

Our goal is to generate a publicly-available global data set of surface drifter positions and velocities at hourly resolution for the post-2005 data. A study of 82 drifters with both GPS and Argos fixes indicates that the Argos positions have an error probability distribution which is better represented by a t location-scale distribution than by a normal distribution. Thus, standard least squares methods for fitting trajectory models to drifter data, with underlying assumptions of normality, are overly sensitive to outlying data, and return inadequate formal estimation errors. As a remedy, we use a maximum likelihood estimator for fitting trajectory models, and prescribe that the position data follow a t location-scale distribution. We show that this new hourly drifter product can be used to study high-frequency oceanic processes such as tides and inertial oscillations.