A float at the surface is pushed by the wind on its exposed part and by waves, as well as by the current. The difference between its motion and that of the water is called slip. Open-ocean practice controls slip with a large drogue: the Surface Velocity Program drifter hangs a holey-sock drogue centred at fifteen metres beneath a small surface float, and its slip was calibrated against current meters by Niiler et al. (1995) as a function of wind speed and of the ratio of drogue drag area to that of the other parts. The instrument and its data are described by Lumpkin and Pazos (2007). For coastal work the reference design is the drifter developed for the Coastal Ocean Dynamics Experiment, in which four fabric vanes span the upper metre of the water column and only a small antenna housing breaks the surface (Davis, 1985). Trials with drogued and undrogued units in the Mediterranean confirmed that wind-induced slip is small for the former and large for the latter (Poulain et al., 2009).

Johnson et al. (2003) described a compact drifter for the nearshore zone, lakes and estuaries built around a logging receiver in a sealed tube with a parachute-style drogue. Schmidt et al. (2003) built a surf zone drifter whose submerged body and damping plate kept it from surfing shoreward on breaking waves. MacMahan et al. (2009) examined whether inexpensive handheld receivers were accurate enough to measure surf zone currents and concluded that, with post-processing, they were. Mobile telephone networks have been used to return positions from coastal drifters in real time (Ohlmann et al., 2005), and a biodegradable drifter was later designed so that hundreds could be released without recovery (Novelli et al., 2017). Student-built drifters have also contributed to regional studies of coastal currents (Manning et al., 2009).

Citizen science has a substantial literature of its own. Bonney et al. (2009) described a model in which scientists frame the question and the protocol and volunteers collect the data, and that is the model adopted here. Kosmala et al. (2016) reviewed the evidence on the quality of volunteer data and found that it can match professional data where the protocol is simple, training is given, the instruments are standardised and the records are checked. Each of those conditions is built into the procedure below.

No single geometry suits every site, and the first decision is to match the drifter to the depth and wave conditions. Table 1 compares the options. For open bays and inner-shelf water deeper than about two metres, a vaned drifter after Davis (1985) is proposed: a central tube carrying four arms from which fabric panels hang, with small floats at the arm ends. For surf zones and very shallow estuaries a short ballasted tube with a damping plate is more suitable. Both can be made from polymer pipe, sailcloth and fishing floats. The tracking unit should log position at intervals no longer than one minute, hold charge for a full day, and sit in a waterproof housing at the top of the tube. Every unit carries a number, a contact telephone and a line of text asking finders to report it.

Before any release, all units are switched on and left together at a fixed point with a clear sky for at least an hour. The scatter of the logged positions about their mean gives the noise of each receiver, and a unit that wanders much more than the others is withdrawn. The same test fixes the shortest interval over which a velocity can sensibly be computed: the displacement over that interval must be several times the position noise. This simple stationary test follows the approach used to assess handheld receivers for surf zone work (MacMahan et al., 2009).

Campaigns are planned around the tide. A complete picture needs releases on the flood and on the ebb, at springs and at neaps, and in at least two contrasting wind conditions. Within a campaign, drifters go into the water in clusters of three or more, a few tens of metres apart, at release points agreed on a chart beforehand. Clusters serve two purposes. Agreement among neighbours guards against a single faulty track, and the rate at which a cluster spreads is itself a measurement of dispersion. The release points should include the features of local concern, such as a river mouth or an outfall, and should be repeated from campaign to campaign so that results can be compared.

Volunteers are organised in boat teams of at least two, each with a team leader who has attended a half-day briefing. The briefing covers the purpose of the work, how to switch on and seal a unit, how to lower it without tangling the vanes, and how to complete the log sheet, which records unit number, time, position of release and recovery, wind estimated on the Beaufort scale, and anything unusual. Safety rules are those of the host organisation and take precedence over the science: lifejackets are worn, no boat goes out alone, and a campaign is cancelled if the forecast exceeds agreed limits. A shore coordinator holds the list of units in the water and confirms that each has been recovered.

Tracks are trimmed to the interval between release and recovery using the log sheets. Position fixes that imply impossible speeds are removed and short gaps are interpolated, in the manner long used for the global drifter data (Hansen and Poulain, 1996). Velocities are computed by centred differences over the interval established in the stationary test. Where wind was measured, a slip correction proportional to wind speed may be applied, with a coefficient taken from the literature for the geometry in use and stated in the record. Tracks are then plotted by tidal phase and the velocities averaged in spatial bins to give maps of mean flow and its variability. For clusters, the growth of mean separation between pairs with time gives a relative dispersion estimate; the statistics and their pitfalls are set out by LaCasce (2008), and the technique was applied to large releases in the Gulf of Mexico by Poje et al. (2014).

Each campaign ends with a return of results to those who did the work: track maps displayed at the club or school, with a plain account of what they show. Data are archived with their metadata, including drifter type, logging interval and corrections applied, in an open format so that other users can judge their fitness for a given purpose.