Anthropogenic litter in the aquatic environment has been reported in many studies around the globe as a major threat to natural ecosystems (Barnes et al., 2009; Wilcox et al., 2015). Plastics are the most common form of this anthropogenic litter and they tend to accumulate at the sea surface, shorelines and the seafloor (Ioakeimidis et al., 2017). Over 250,000 tons of marine plastic litter is floating at the ocean surface (Eriksen et al., 2014). It is distributed by complex surface currents and winds that result in complex pathways to the accumulation zones in the open sea and coastal areas. Scientific evidence-based studies on the distribution of floating marine litter (FML) as well as understanding the processes influencing these pathways are vital steps toward pinpointing potential accumulation zones. International, regional and local environmental agencies have been echoing the need for sustainable, robust and affordable marine technology to improve benchmark datasets related to FML (GESAMP, 2015; G20, 2017; UN-SDG14, 2017). Already, optical sensors capable of collecting crucial benchmark information have shown promising results in identifying and monitoring FML in the Great Pacific Garbage Patch (Garaba et al., 2018).

In the last four decades, the rising occurrence of marine litter has been documented in the North Sea (Dixon and Dixon, 1983; Vauk and Schrey, 1987; Galgani et al., 2000; Claessens et al., 2011; Van Cauwenberghe et al., 2013; Schulz et al., 2015a,b). Thiel et al. (2011) estimated about 32.4 litter items per km², with a significantly higher density in coastal regions. More recently, Kammann et al. (2018) found at least 16.8 litter items per km² on the seafloor of the North Sea. In both studies the dominant type of litter was from plastics. Prevailing westerly winds and the cyclonic residual circulation in the North Sea transports FML from the English Channel along the coast of the southern North Sea into the German Bight (Vauk and Schrey, 1987). As a result, marine litter is washed ashore along the German Bight (Schulz et al., 2015a,b). Neumann et al. (2014) showed from numerical simulations a seasonal trend in the distribution of FML toward the coast which was strongly driven by wind patterns. The seasonal trend for the southern North Sea is consistent with recent numerical simulations that also revealed the west coast of Denmark and the Skagerrak as major sink regions for FML (Gutow et al., 2018). Unfortunately, a comprehensive performance test of these models is still missing due to a lack of sufficient Lagrangian data in the North Sea region. This lack of datasets resulted in van Sebille et al. (2012) and Maximenko et al. (2012) excluding the North Sea from their global marine litter transport models. Clearly, there is an urgent need for high quality surface drifter data to validate particle tracking and hydrodynamic models for the North Sea and the German Bight.

The implementation of a Lagrangian particle tracking model in shallow waters, such as the German Bight, that are strongly affected by diurnal and semi-diurnal tides is challenging and needs high-resolution hydrodynamical models in coastal areas. Such a numerical model requires in-situ Lagrangian observations for calibration, validation and tuning purposes especially for complex sub-mesoscale processes that have a huge impact on the dispersion of particles at the ocean's surface. Generic particle tracking models tend to neglect two important mechanisms along the coast, consisting of (a) washing ashore or beaching of litter and (b) re-floating of stranded litter back into the ocean. Beaching and refloating are very complex physical processes, let alone their numerical representation. These processes depend on the bathymetry of the coastline, wind conditions, waves, tides, litter type, size, and buoyancy (Yoon et al., 2010). The investigation of the interaction between complex coastal processes and the transport of FML is therefore a crucial step toward reliable marine litter transport models (Critchell et al., 2015).

In tracking the pathways of FML, ongoing research can be complemented by high quality in situ Lagrangian measurements using advanced satellite-tracked surface drifters (Maximenko et al., 2012; van Sebille et al., 2012). This is especially true of the very important questions, such as where marine litter comes from and ends up, must be answered with the aid of numerical models, which must in turn be validated with Lagrangian observations.

Recent experiments with surface drifters have fundamentally enhanced our understanding of complex surface current systems, including the prediction of pathways and accumulation areas of FML (Maximenko et al., 2012; van Sebille et al., 2012; Poje et al., 2014; Miyao and Isobe, 2016; D'Asaro et al., 2018), improved search and rescue models (Breivik et al., 2013) as well as the modeling and forecast of oil spill dispersion (Reed et al., 1994; Liu et al., 2011; Poje et al., 2014). A very popular design evolved in the context of the Surface Velocity Program (SVP) in the early 1980s. It consists of a surface buoy containing a telemetry device and a subsurface drogue. The drogue is connected with a tether to the surface buoy and is centered 15 m beneath the water line (Sybrandy et al., 1991). Drogues are used in various drifter designs to reduce the wind slippage and the Stokes drift due to wave motion (Niiler et al., 1995). Another popular drifter design was developed within the Coastal Ocean Dynamics Experiment (CODE). The drifter consists of a vertical tube, which contains the electronics, and is surrounded by four drag producing vanes. The drifter was designed to follow the current in the top meter of water from the surface (Davis et al., 1982; Davis, 1985). Recently, over 1,000 biodegradable surface drifters were deployed in the Gulf of Mexico for the investigation of oil spill dispersal (Novelli et al., 2017; Laxague et al., 2018). For a detailed review of the history and the state of the art of drifter developments, see Lumpkin et al. (2017) and the references therein.

Maximenko et al. (2012) used a transition matrix, derived from a global dataset of drifter trajectories, to study the distribution of marine litter and identified five major oceanic accumulation zones which were confirmed by in situ observations (Ryan, 2014; Lebreton et al., 2018). van Sebille et al. (2012) implemented a more representative source function of litter and found another garbage patch in the Barents Sea which was also confirmed by recent in situ observations (Bergmann et al., 2016). Other authors have combined trajectories of surface drifters and particle tracking methods to investigate the distribution of FML in the Mediterranean Sea (Carlson et al., 2017b; Politikos et al., 2017; Zambianchi et al., 2017).

Within the framework of the project “Macroplastics Pollution in the Southern North Sea—Sources, Pathways and Abatement Strategies” funded by the German Federal State of Lower Saxony, we developed a robust, low-cost and state-of-the-art drifter. The objectives of our drifter were firstly to adapt and customize a CODE drifter design for use in shallow water where litter can be beached and resuspended. The original CODE drifters have tethered drogues that can get stuck in shallow water and, once beached, are not easily re-floated. The second objective is to contribute toward high quality datasets useful for investigations of sub-mesoscale dynamics of surface velocity fields in strong tidally influenced shallow water areas. To this end, we tested and used seven of our prototypes to expand the much needed in situ knowledge of Lagrangian observations in the German Bight. We also discuss the surface velocity field derived from our drifter measurements regarding the tidal and the wind-induced surface velocity components and provide insights on coastal processes influencing the retention times and the beaching behavior of the drifters.

A total of four drifters were deployed in March 2017 and 3 drifters were deployed in October 2017 from RV Heincke and RV Senckenberg. These deployments were done in 3 pairs, at the same time, with an initial separation of <1 m except for drifter 7. Deployment locations were chosen to be in the coastal area near the major shipping lane as a possible source for FML in the German Bight. Metadata related to these drifter deployments, beaching and drifter displacement are summarized in Table 1. The lifetimes of the drifters deployed in March 2017 ranged from 22 to 39 days.

All drifters launched in October 2017 had lifetimes of 15 days. The average meridional speeds of the drifters deployed in March 2017 ranged from 0.019 to 0.051 ms⁻¹ and the average zonal speeds ranged from 0.013 to 0.052 ms⁻¹. In October 2017, average meridional speeds were in the range of 0.026 to 0.033 ms⁻¹ whilst average zonal drifter speeds were from 0.051 to 0.089 ms⁻¹. Wind direction distributions for the March (Figure 5A) and October (Figure 5B) deployments are presented as a wind rose histogram in direction of the blowing wind. The general wind direction for the March deployment was mostly southeastwards, with wind speeds ranging from 5 ms⁻¹ to 12 ms⁻¹ and maximum speeds reaching 16.8 ms⁻¹. Wind speeds in October were higher, ranging from 8 ms⁻¹ to 14 ms⁻¹ and peak speeds of up to 24.3 ms⁻¹. The predominant wind direction in October was toward the northeast.

The individual trajectories of the drifters deployed in March are shown in Figure 6 with a different color for consecutive five-day intervals. The corresponding colored arrows illustrate the average wind conditions for these time intervals to give an overview of the wind forcing in these time periods. Drifter 1 (D1) and Drifter 2 (D2) were deployed on the 13th of March 2017 at the border of the German Exclusive Economic Zone to the northwest of Borkum (Figures 6a,b). In the beginning, the drifter pair moved eastward due to stable westerly winds. The further course of the trajectories is mainly influenced by semi-diurnal tidal oscillations and wind forcing. Due to principal westerly winds during the deployment interval, D1 beached at the North Frisian coastline 33 days after the release with a displacement of 1,215 km. D2 washed ashore 39 days after deployment not far away from the beaching location of D1 and covered a distance of 1,369 km. Drifters 3 (D3) and 4 (D4) were deployed in the tidal inlet between the islands Spiekeroog and Langeoog on the 14th of March 2017 (Figures 6c,d). However, both drifters were transported to the east by stable westerly winds during the first days after the deployment. The trajectories were split up in front of the Elbe estuary 5 days after the deployment. D3 was trapped in a tidal eddy in front of the North Frisian coastline for almost 12 days before the drifter stranded in the North Frisian Wadden sea (Supplementary Video 1). The drifter covered a distance of 897 km in a period of 22 days. D4 washed ashore in the Wadden Sea in front of the island Scharhörn during flood tide on the 19th of March 2017 (Supplementary Video 1 and Supplementary Figure 1). The drifter was flushed back into the sea on the 20th of March at high tide and strong southwesterly winds (15 ms⁻¹), passed the Elbe estuary and was transported to the northwest and stranded 28 days after deployment on the island Sylt (Supplementary Video 1). The drifter covered a distance of 935 km on its journey. It was the only drifter that stranded and refloated.

The trajectories of the drifters from the October deployment are shown in Figure 7. Drifter 5 (D5) and Drifter 6 (D6) were deployed as a pair on the 8th of October 2017, southwest of the island Heligoland (Figures 7a,b). The trend of the pathways of D5 and D6 are very similar in the beginning. Due to strong winds from the west, the drifters moved in 5 days to the northern part of the Elbe estuary. From there, they were transported northward along the North Frisian coast. The drifters meandered for 7 days close to the North Frisian Islands and were flushed several times with the tidal currents into the tidal inlets before they stranded on the island Amrum (Supplementary Video 2). Both drifters D5 and D6 beached 15 days after their release. D5 covered a distance of 654 km, whereas D6 traveled a distance of 628 km in the German Bight. Drifter 7 (D7) was deployed north of the island Spiekeroog on the 13th of October 2017 (Figure 7c). Due to strong southwesterly winds, the drifter moved in the northeastern direction. A week after the deployment, D7 passed the island Heligoland and was transported northwards by strong winds from the south. Drifter D7 stranded on the island Sylt 15 days after release and covered a distance of 463 km on its journey. All drifters discussed in this study were immediately recovered by tourists on beaches or in dry falling areas in the Wadden sea after the experiments.

To illustrate observed and derived parameters, the zonal (U_d) and meridional (V_d) drifter velocities, estimated tidal velocities (zonal U_tide, meridional V_tide) and residual velocities (zonal U_res, meridional V_res) of D1 and D2 are presented in Figure 8. Values for U_d for D1 range from −1.08 ms⁻¹ to 1.19 ms⁻¹ and the zonal velocity values for D2 vary from −0.86 ms⁻¹ to 1.28 ms⁻¹. The tidal oscillation of the drift velocities is pronounced in the time series data and spring-neap tidal cycles are significant. The dominance of the M₂ signal can be identified in the zonal velocities and is more prominent than in the meridional velocities. The tidal oscillations are independent of calm or stormy weather events. The amplitude of the meridional velocity components is lower in comparison to the zonal components. Low-pass-filtered residual currents have a zonal velocity amplitude of −0.29 ms⁻¹ to 0.30 ms⁻¹ for D1 and −0.23 ms⁻¹ to 0.41 ms⁻¹ for D2. The meridional velocity amplitudes of the drifters are smaller in comparison to the zonal velocities, which range from −0.77 ms⁻¹ to 0.59 ms⁻¹ for D1 and −0.48 ms⁻¹ to 0.66 ms⁻¹ for D2. The low-pass-filtered residual meridional velocities vary from −0.23 ms⁻¹ to 0.24 ms⁻¹ for D1 and from −0.27 ms⁻¹ to 0.33 ms⁻¹ for D2.

Figure 9 provides an overview of the drifter velocities, the tidal velocities and the residual velocities for the October deployment (D5 and D6). U_d shows similar magnitudes in comparison to D1 and D2. V_d is significantly higher in comparison to D1 and D2 with values in a range from −1.12 ms⁻¹ to 1.01 ms⁻¹ for D5 and −0.91 ms⁻¹ to 0.96 ms⁻¹ for D6. The highest values in both velocity components were observed in the tidal inlet in front of the North Frisian coast. The derived residual velocities range from −0.15 ms⁻¹ to 0.31 ms⁻¹ for D5 and −0.15 ms⁻¹ to 0.25 ms⁻¹ for D6 regarding the zonal component. V_res varies from −0.19 ms⁻¹ to 0.22 ms⁻¹ for D5 and from −0.20 ms⁻¹ to 0.18 ms⁻¹ for D6. The influence of the tidal oscillation is significant for the drifter velocities. The amplitudes of the velocity components are significantly higher for the spring tide period, which occurred on the 19th of October 2017.

Statistically significant positive linear correlations (R² > 0.72, p < 0.05) between bin-averaged wind velocities and bin-averaged residual drifter velocities for all 7 drifters were observed (Figure 10). The correlation between bin-averaged winds and residual drifter velocities has a strong positive association (Figure 10A, R² = 0.72, N = 98, RMSE = 4.5 cm s⁻¹ for the zonal component; Figure 10B, R² = 0.86, N = 98, RMSE = 3.6 cm s⁻¹ for the meridional component).

The state-of-the-art low-cost drifter design presented here is primarily intended for use in shallow waters but can be easily adapted to all the above-mentioned scientific questions for the open ocean environment. The drag area ratio can be adjusted by changing the vanes and regulating the draft of the housing. Due to the low draft, the drifter can operate in estuaries and in surf zones to measure rip currents. The possibility to attach a subsurface drogue to a ring screw at the bottom of the housing enables the drifter to provide valuable information on the large-scale ocean circulation for deep-water surveys.

Positioning (Figure 2a) and data transmission (Figure 2b) analysis have shown that the drifter observations provide accurate geo-location and high temporal data that are necessary to resolve sub-mesoscale processes. The positions were transmitted at an average interval of 12.4 min with very minimal data gaps. A major disadvantage is the fact that no other sensors can be connected directly to the SPOT Trace®. The drifter deployments in October suggest that the drifters can operate and provide reliable data during stormy weather conditions with wind speeds up to 24.3 ms⁻¹ and resulting high waves. Due to the robust drifter design, it is therefore also possible to observe transport patterns of floating objects during storm events.

The drifter design is capable of representing beaching and refloating comparable to floating marine litter of a certain size. The beaching and refloating processes are very challenging issues which are poorly represented in most marine litter transport models (Critchell et al., 2015). Beaching of floating objects strongly depends on the buoyancy of the objects (Yoon et al., 2010), the structure of the coastline, winds, waves, tides and the bathymetry (Critchell et al., 2015).

The trajectories of D1 and D3 show a beaching scenario in the same coastal area under different wind conditions. D1 meandered in coastal waters for 12 days due to very low wind speeds of 1–3 ms⁻¹, whereas D3 washed ashore near the stranding location of D1 but passed through the coastal area in 2 days due to westerly wind speeds of approximately 10 ms⁻¹. Refloating of beached objects due to tides and winds is shown in the trajectory of D4, which was the only drifter that beached and refloated, because the other drifters were immediately recovered by tourists. The drifter washed ashore and was resuspended into the ocean a day later by strong southwesterly winds and tides. The dispersion of D3 and D4 is caused by the stranding and refloating of D4. Refloating of objects in a complex tidally influenced shallow water area represents an important process regarding the transport pattern and stranding locations of FML and needs to be taken into account in particle tracking models.

The trajectories of D3, D5, and D6 indicate that floating items tend to accumulate in sub-mesoscale tidal recirculation cells under low wind conditions. The formation of recirculation cells due to calm wind conditions in the eastern part of the German Bight (Dippner, 1993) can lead to significantly longer residence times of FML along the North Frisian coastline. The residence times can be calculated by computing the characteristic time of a drifter in a specified spatial domain (Choukroun et al., 2010). Dippner (1998) also discussed the formation of three mesoscale eddies between the salinity front of the river plumes of the Elbe and Weser and the North Frisian coast during westerly wind conditions caused by the interaction of wind stress and the gradient of the bottom topography. This eddy regime can result in a disintegration of the northward transport between the salinity front and the coast and can lead to accumulation and longer residence times of FML in this area. The residence times of D3, D5, and D6 in the coastal area in front of the North Frisian shoreline support these assumptions. Observations by Thiel et al. (2011) have shown a higher abundance of flotsam in the coastal area, which is influenced by estuarine frontal systems and by intermittent eddies, than in areas farther offshore, which is also in a good agreement with the residence times of D3, D5, and D6. Gutow et al. (2018) observed also very high densities of FML along the North Frisian coastline with values above 100 items km⁻², which fits well with our findings.

Furthermore, several studies reported on the accumulation of floating objects at frontal systems (Skov and Prins, 2001; Acha et al., 2003; Pichel et al., 2007; Thiel et al., 2011; Gutow et al., 2018), which is an important factor regarding the transport pattern of FML and needs to be focused on in further investigations. If floating objects remain for longer periods in frontal areas, the objects can lose their buoyancy due to marine fouling and sink to the ground or are pushed with winds to the shore (Hinojosa et al., 2011). In addition, it has been reported that bottom litter might accumulate in areas with high sediment transport and low circulation rates (Galgani et al., 2000), which could also be an important factor regarding the accumulation of marine litter in the German Bight due to high sedimentation rates in the North Frisian and East Frisian areas (Lettmann et al., 2009; Zeiler et al., 2014).

The complex interaction of tides, winds and the coastline structure in the coastal area is demonstrated by the trajectories of D5 and D6. The drift patterns of both drifters are significantly influenced by strong tidal jets in the tidal inlets in front of the North Frisian coast. The drifters were flushed several times into the tidal inlets with the flood tide and floated back in front of the islands with the ebb tide, which gives rise to the assumption that floating objects can also be trapped for longer time periods in the vicinity of tidal inlets.

The drifter velocities are dominated by tides, wind induced slippage and surface currents (Figures 4, 10). The tidal components are clearly pronounced in the zonal velocities whereas the tidal oscillations in the meridional component seem to be more irregular (Figures 8, 9). Similar results were also found in high frequency radar data (HF-radar) and model results for the German Bight, which show that the tidal ellipses are very pronounced in the zonal direction (Barth et al., 2010; Port et al., 2011). The tidal analysis shows that the most dominant tidal constituents in the surface velocity field are the principal lunar semi-diurnal tide M₂ and the solar semi-diurnal tide S₂. These findings are consistent with numerical simulations and HF-radar observations (Port et al., 2011). Magnitudes of zonal and meridional velocities agree with HF-radar-derived velocities in this area (Port et al., 2011; Stanev et al., 2015). The residual velocities are significantly correlated with the local wind fields (Figure 10). The fit parameters and the correlation coefficients show a stronger wind influence on the meridional component, which is in good agreement with results shown by Port et al. (2011). The wind-induced drifter movement is driven by the direct wind slippage (Section Wind Slip), wind stress on the surface current and the Stokes drift. The variance observed in the zonal velocities at high wind speeds could be caused by increasing crosswind leeway components which need to be taken into account at high wind speeds (Breivik et al., 2011). Furthermore, island effects have a major impact on velocity fields in the coastal area and need also to be taken into account. The findings of this study indicate, that tidal jet currents have an enormous impact on the sediment transport in the coastal zone (Staneva et al., 2009). Mantovanelli et al. (2012) have shown, that strong jet currents which are generated by the bathymetry in the coastal area and tidal channels between islands, have significant influence on the transport of particles in the coastal area. This agrees well with the observations of this study, which also show the highest drifter velocities for both components in tidal inlets generated by tidal jets (Figure 9, and Supplementary Video 2) which seem to overshadow the wind effects in the coastal area. These findings are also in good agreement with recent studies by Callies et al. (2017) that compared trajectories of six surface drifters with results of offline particle simulations in the inner German Bight.

Overall, the analysis of the drifter velocities shows that the transport of floating objects at the ocean surface is strongly affected by the local wind field, especially for areas far away from the coast. The wind data is taken from a single observational station and is not representative for the whole region of the study, but due to its central location in the German Bight it can give a good estimate of the local wind field. Laxague et al. (2018) have demonstrated, that the transport of floating objects at the surface is predominantly forced by wind- and wave-induced motions and that the currents at the surface (0.001–0.5 m) are significantly stronger compared to the surface currents averaged over the top 5 m of the water column. Most particle tracking models poorly represent these wind- and wave-induced motions at the surface or do not have adequate resolution to describe the transport processes of floating items at the surface accurately (Laxague et al., 2018). The drifter design presented in this study is targeted at the motion processes of floating items in the surface layer (0.5 m) and takes into account wind- and wave-induced motions in this small vertical column having an enormous impact on the transport pattern of FML. These results could be useful for the parametrization of wind- and wave- induced motions of surface currents in particle tracking models, which has been recently demonstrated by Stanev et al. (Forthcoming). Since the residence time of FML strongly depends on the buoyancy ratio of floating objects (Yoon et al., 2010), experiments with different drag area ratios can be very helpful for the validation of particle tracking models. The presented drifter setting with a drag area ratio of 25.6 represents macroscopic litter that floats at the sea surface with a low buoyancy. SVP drifters with drag area ratios around 40 represent typically the current of the first 20 meters of the surface layer. The movement of macroscopic litter takes place in the upper 0.5 m of the ocean and is strongly influenced by wind- and wave-induced forces (Laxague et al., 2018). Lagrangian drifter experiments with different drag area ratios can provide fundamental insights into the dynamics of FML of different shapes at the ocean's surface.

The combination of wind-induced motion and nearshore processes like tidal jets, the formations of sub-mesoscale recirculation cells and the interaction with a complex shoreline have a great impact on the pathways, accumulation, residence times and beaching of floating objects in the German Bight. The investigation of these processes is fundamental for understanding the beaching behavior of FML at the shoreline and to implement such processes in high resolution particle tracking models (Neumann et al., 2014; Schulz and Matthies, 2014; Schulz et al., 2015b).

The drifter design presented in this study was primarily developed for the observation of the transport of floating objects in shallow waters, with an adjustable drag area ratio which is 25.6 for the presented configuration. The drifter is compact, robust, accurate, affordable, simple to manufacture and is easily adaptable to address open ocean and coastal water questions related to transport pattern of pollutants and current circulation studies. The trajectories of these drifters generate information on the surface current velocity field and complex transport patterns from the coastal area to the open sea. Here we present findings from seven surface drifters deployed in the German Bight, a region with a substantial lack of Lagrangian observations. Our drifter-derived surface current velocity data and resolved transport patterns in the coastal area were consistent with hydrodynamical and particle tracking simulations in the German Bight. The direct wind slip of the drifter was determined to be 0.27% of the wind speed, and the overall wind induced motion to the drifter due to waves, direct wind slip and wind stress at the water surface was about 1% of the wind speed.

The drifter trajectories observed in this study highlight some of the constraints in simulating the transport of FML in the open ocean and even more in coastal zones. The processes presented in this study need to be taken into account to better represent the transport of FML in numerical simulations in a more realistic approach, which will be done in further work with a nested high-resolution particle tracking model. Ultimately, the typical movement of particles in water is a Lagrangian problem, which needs to be better understood with the aid of accurate in situ Lagrangian observations.

JM and TB contributed to the data collection and the data analysis. JM drafted the manuscript and all other authors contributed to the critical revision of the article and provided important advice for the improvement of the manuscript. All authors were involved in the conception of the experiments.