Basin-scale sources and pathways of microplastic that ends up in the Galápagos Archipelago

. The Galápagos Archipelago and Marine Reserve lies 1,000 km off the coast of Ecuador and is among the world’s most iconic wildlife refuges. However, plastic litter is now found even in this remote island archipelago. Prior to this study, the sources of this plastic litter on Galápagos coastlines were unidentiﬁed. Local sources are widely expected to be small, given the limited population and environmentally-conscious tourism industry. Here, we show that remote sources of plastic pollution are also fairly localized and limited to nearby ﬁshing regions and South and Central American coastlines, in particular Northern 5 Peru and Southern Ecuador. Using virtual ﬂoating plastic particles transported in high-resolution ocean surface currents, we analysed the plastic origin and fate using pathways and connectivity between the Galápagos region and the coastlines and known ﬁsheries locations around the East Paciﬁc Ocean. We also analysed how incorporation of wave-driven currents (Stokes drift) affects these pathways and connectivity. We found that only virtual particles that enter the ocean from Peru, Ecuador and (when waves are not taken into account) Colombia can reach the Galápagos. It takes these particles a few months to travel from 10 their coastal sources on the American continent to the Galápagos region. The connectivity does not seem to vary substantially between El Niño and La Niña years. Identifying these sources and the timing and patterns of the transport can be useful for identifying integrated management opportunities to reduce plastic pollution from reaching the Galápagos Archipelago.

concentration of small floating plastic is 10 million times higher in the subtropical accumulation regions than in the Southern Ocean (van Sebille et al., 2015). Because of deep upwelling of water in the Southern Ocean and Ekman drift towards the subtropical gyres (Rintoul and Naveira Garabato, 2013), there is a net transport of floating plastic away from the region (Onink et al., 2019). The same is true for regions on the Equator, such as the Galápagos Archipelago, where upwelling and surface divergence mean that the surface flow is predominantly directed away from the Equator (Law et al., 2014). 5 The Galápagos Archipelago and Marine Reserve are among the world's most valued and most iconic ecosystems. Its special qualities were first noticed when Charles Darwin visited the archipelago in 1835. They were later recognised in the islands being granted the first UNESCO World Heritage status for natural value in 1978, with the marine reserve following the archipelago itself onto the UNESCO World Heritage List two decades later. However, even this remote archipelago is not as pristine as one would hope (Mestanza et al., 2019). So, despite the archipelago being in a region of ocean surface divergence (Fiedler et al.,10 1991) with relatively low expected plastic concentrations, the blight of plastic pollution has now also arrived in Galápagos.
There, it has unquantified but likely significant impacts on the unique ecosystem as well as on the sustainability of the tourism industry which supports the economy of the Galápagos locally, and Ecuador more broadly.
Management and mitigation of the plastic problem in the Galápagos Archipelago requires understanding the scale and sources of the pollution. While some of the plastic found on coastlines and in the marine reserve may originate from the 15 islands themselves, including tourism, there is a widespread view, based on information from coastal clean up efforts (Galápagos National Park, unpublished data), that much of the plastic found in the Galápagos comes from mainland America, from continental Asia, and from fisheries in the Pacific Ocean.
Here, we investigated the pathways of floating microplastic between the Galápagos Islands and coastlines and known fisheries locations around the Pacific. There is some observational data on pathways into the Galápagos region, from satellite-

25
To augment the GDP drifter observations, we employ state-of-the-art numerical models. We used a combination of the fineresolution NEMO global hydrodynamic model for ocean surface currents (Madec, 2008), the WaveWatch III model for waves (Tolman, 2009), and the Parcels v2.0 Lagrangian particle tracking toolbox (Lange and van Sebille, 2017; Delandmeter and van Sebille, 2019). We compared these with the trajectories of floating drifters in the real ocean.
There is still a debate in the physical oceanography community to what extent wave-induced currents -so-called Stokes drift 30 (Stokes, 1847) -has an impact on the transport of plastic (Lebreton et al., 2018;Onink et al., 2019). Therefore, we analyzed the particle pathways both with and without this effect of waves. Stokes drift is the net drift velocity in the direction of wave propagation experienced by a particle floating at the free surface of a water wave (see van den Bremer and Breivik, 2018, for a recent review). Its magnitude is generally much smaller than that of the surface currents (e.g. Figure 1 of Onink et al., 2019), but because Stokes drift has large spatial coherence patterns its long-term effect on particle transport can be significant (Fraser et al., 2018).
Finally, we also describe how the modelling performed here can work alongside other methodologies, to demonstrate the benefits of multidisciplinary approaches to helping resolve the problem of marine plastic pollution.

5
We performed six experiments in three scenarios: one scenario where we tracked the origin of particles, by computing particles that end up near the Galápagos in backward time; one scenario where we tracked the fate of particles that were released from the west coast of the Americas in forward time; and one scenario where we tracked the fate of particles that were released at know fishing locations in forward time. In all three scenarios, we simulated the transports by ocean surface currents only and by the combination of surface currents and waves. As the NEMO model data is available on 8km resolution, we focused only 10 on the basin-scale transports, and leave transports within and between the different islands of the Galápagos Archipelago for future work.
We used the two-dimensional surface flow fields from the NEMO hydrodynamic model, simulation ORCA0083-N006, which has a global coverage at 1/12 • resolution (nominally 8km around the Equator) (Madec, 2008). The NEMO data is available from January 2000 to December 2010, on 5-day temporal resolution. As Qin et al. (2014) showed that time-averaging 15 errors are small for temporal resolutions shorter than 9 days in a 1/10 • spatial resolution, this 5-day temporal resolution is sufficient.
For the Stokes currents, we used the WaveWatch III data based on CFSR winds (Tolman, 2009), which has a global coverage at 1/2 • resolution (nominally 55km around the Equator). The WaveWatch III data is also available from January 2000 to December 2010, on 3-hour temporal resolution. 20 We advected Lagrangian particles using the Parcels v2.0 toolbox (Lange and van Sebille, 2017; Delandmeter and van Sebille, 2019) in either only the NEMO surface flow fields (hereafter referred to as the 'currents' simulations), or the combined NEMO surface flow and WaveWatchIII Stokes drift fields (hereafter referred to as the 'currents+waves' simulations). Parcels v2.0 has inbuilt support for advection of particles on multiple different Fields using SummedField objects, so that the velocities at each location are interpolated and then summed at each RK4 substep (see also Delandmeter and van Sebille, 2019), and 25 the currents+wave simulations were performed using that feature. The particles represented microplastic that are sufficiently buoyant to not mix too deep in the mixed layer (Onink et al., 2019). We used a Runge-Kutta 4 integration scheme with a time-step of one hour. We stored the location of each particle on a daily (24 hours) resolution. All scripts that were used to run the simulations are available at https://github.com/OceanParcels/GalapagosBasinPlastic.
On each set of fields, we performed three different simulations based on three scenarios. In the 'Origin from Galápagos' 30 scenario, we released 154 particles every 10 days in a box between [91.8 • W -89 • W, 1.4 • S -0.7 • N] (the red box in Figure 2), on a 0.2 • x0.2 • grid, for a total of 61,908 particles. We integrated these particles back in time for a maximum length of 10 years, or until the first day available in the NEMO dataset. Redoing all the analyses below with only half of the particles does not affect the results and conclusions, giving us confidence that we released sufficient particles.
In the 'Fate from the South American coastline' scenario, we released one particle each 0.5 • between 38 • S and 31 • N every 5 days, for a total of 120,450 particles. Again, using only half of the particles in our analysis did not change the results and conclusions drawn below. For each latitude, we picked the easternmost longitude that is still in the Pacific Ocean, so that the 5 release points traced the coastline of the Americas. We then integrated our particles forward in time for a maximum of 5 years, or until the last day available in the NEMO dataset. We identified those particles that crossed the box at [91.8 • W -89 • W, 1.4 • S -0.7 • N], the same box as the release for the 'Origin from Galápagos' simulation, and defined these to be passing through the Galápagos Archipelago region.
In the 'Fate from regional fisheries' scenario, we released particles according to the distribution of total fishing effort, as 10 mapped by Global Fishing Watch (Kroodsma et al., 2018), in a region around the Galápagos (Figure 2). We selected only these locations where there was at least 24 hours of fishing activity between 1 January 2012 and 31 December 2016. As these dates did not overlap with the available NEMO surface flow data from 2000 to 2010, we repeatedly released one particle each month -weighted to the number of fishing hours -at each of the 3,885 locations in Figure 2, for a total of 520,590 particles. We then integrated these particles forward in time for a maximum of 5 years, or until the last day available in the NEMO dataset. 15 We used the same definition of passing through the Galápagos Archipelago region as in the 'Fate from the South American coastline' simulations above.

Results
In the 'Origin from Galápagos' scenario, most particle trajectories were confined to the Eastern Tropical Pacific Ocean, the South American coastline, and the Antarctic Circumpolar Current (Figure 3). In the currents+waves run, some particles even 20 arrived in the Galápagos region that originated from the Indian Ocean (Maes et al., 2018;van der Mheen et al., 2019). However, none of the almost 65,000 particles came from the North or South Pacific accumulation zones (Kubota, 1994;Martinez et al., 2009;Eriksen et al., 2013;van Sebille et al., 2015) or from even close to mainland Asia. While some particles in the currents-only simulation originated from the very southern part of California, most particles originated from much farther south. Interestingly, the inclusion of Stokes drift meant that particles were much more dispersed through the Southern Ocean, 25 in agreement with recent simulations of Kelp in that region (Fraser et al., 2018).
In the 'Fate from the South American coastline' scenario, most particles released from the American coastline ended up in either the North Pacific or South Pacific accumulation zones within the five years that they were advected for ( Figure 4). Some particles even ended up in the Indian Ocean, having passed through the Indonesian Throughflow (e.g. van Sebille et al., 2014).
There was a local minimum in the density of particle trajectories on the Equator, especially west of the Galápagos, which 30 agrees with the GDP drifters (lower panel of Figure 1). Compared to the currents-only simulation, the convergence zones were more spread-out and reached farther westward in the currents+waves simulation. The accumulation zones were also smaller and had lower maxima in the currents+waves simulation, partly because the waves constantly push particles eastward onto the shore, so that they had less chance of reaching the open ocean. Indeed, the narrow strip of very high concentrations seen along the South America coastline in the lower panel of Figure 4 confirms that one effect of the eastward Stokes drift induced by the waves was to contain the particles close to their release locations.
The fraction of particles that reached the Galápagos region, starting from the western American coast, is shown in Figure 5.
Only very few of the particles released south of 16 • S or north of 3 • N reached the Galápagos, and even for the regions between 5 16 • S and 3 • N the fraction of particles arriving in the Galápagos region is never higher than 25%. There was a clear difference between the two flow simulations: in the currents+waves simulation (blue line in Fig 5) the particles that reached the Galápagos came almost exclusively from Peru, while in the currents-only simulation there was also a significant fraction of virtual particles from Ecuador, Colombia, Costa Rica and even farther north.
In both 'Fate from the South American coastline' simulations, less than 1% of the particles from the Chilean coast arrived The travel time from the west coast of Americas to the Galápagos was typically a few months ( Figure 6). In the cur-rents+waves simulation, almost all particles that reached the Galápagos did so within 3 months (100 days; blue bars in Figure 6). In the currents-only simulation, there was a much longer tail, reaching travel times up to 5 years (yellow bars). Note 20 however, that none of the simulations here take sinking of particles into account, which can be expected to be more likely for longer times at sea (Kooi et al., 2017;Koelmans et al., 2017). Furthermore, longer residence times in the ocean will also likely lead to more fragmentation, but this is also not taken into account because the time scales involved are very poorly constrained from observations (Cózar et al., 2014).
An analysis of the particles reaching the Galápagos from mainland America for each year showed that there was little 25 impact of El Niños and La Niñas on the transport of particles from the American coastline to the Galápagos region ( Figure 7). However, it should be noted here that, because in the currents-only simulation a significant fraction of particles take multiple years to arrive in the Galápagos region, a large part of the downward trend in the left panel in Figure 7 is due to particles having a probability to reach the Galápagos that decreases with time for the last six years of the simulation.
The 'Fate from regional fisheries' scenarios revealed that the probability for particles starting in most of the known fishing 30 locations around the Galápagos to end up on the Galápagos was very small (Figure 8). The total, fishing-hour-weighted fraction of particles that ended up in the Galápagos box was less than 1% for both the currents and currents+wave simulations. Probabilities higher than 5% were only found in fishing locations north and east of the Galápagos in the currents-only simulation, and along the Ecuadorian and Peruvian coastline in the currents+waves simulation, which was in agreement with the results from the other two scenarios described before. 35 We have analysed the pathways of virtual particles representing floating microplastics in two sets of simulations; with currents only and with both currents and waves. It is clear that the inclusion of waves had a major effect on the transport of this plastic, and that especially connections to the Northern Hemisphere are reduced due to the effect of waves. The 'Origin from Galápagos' scenario ( Figure 3) revealed that it is extremely unlikely for plastic from anywhere but a relatively local region in 5 the Eastern Tropical Pacific, the coastline of South America, and the Southern Ocean to arrive into the Galápagos region.
It is important to note that the virtual particles in these simulations represent highly idealised plastic only. We did not consider beaching, degradation, sinking nor ingestion of plastic. We also did not consider what happens within the Galápagos region.
The simulations agreed well with the trajectories of the GDP drifters (Fig 1). While 40 drifters is not sufficient to do robust statistical comparison (e.g. van Sebille et al., 2009), the patterns of the drifters show similar patterns as the distributions of the 10 virtual particles, especially for the 'Fate from the South American coastline' wind+currents simulation. Since these drifters have mostly lost their drogues by the time they reach the Western Tropical Pacific Ocean (blue lines in Figure 1), it is indeed expected that waves play a role in the dispersion of the satellite-tracked drifters.
The differences between the currents only and currents+wind simulations thus demonstrates the importance of the inclusion of wind effects on the transport of microplastics (Lebreton et al., 2018;Fraser et al., 2018;Onink et al., 2019). These wind-15 driven Stokes currents, however, are not routinely incorporated into numerical hydrodynamic models, and in fact are not even well-observed. This may change, however, if the European Space Agency's SKIM concept mission to directly measure surface currents from space is launched (Ardhuin et al., 2018). The research presented here highlights again how important it is to observe Stokes drift on a global scale for the simulation of floating debris.
This project forms part of a wider multi-disciplinary programme involving scholars and research teams in marine biology, 20 ecotoxicology, environmental psychology and archaeology. Working collaboratively, and in partnership with local communities, this collaborative effort is expected to develop a better understanding of the causes and consequences of marine plastic pollution in Galápagos than existed previously. Given the understanding of oceanographic currents, the degree of management and policy instruments available, and iconic status of Galápagos, the archipelago is well, even uniquely positioned to provide a demonstration of how a marine reserve can manage and reverse its marine plastic burden. The hope is also that the processes, 25 methodologies, management tools and partnerships established in Galápagos can be extended to other places around the world.
Understanding how currents and waves carry plastic from points of deposition ('taps') to places of accumulation ('sinks') is vital. By combining this understanding with the results of other approaches can bring additional insight. For example, an archaeological methodology being trialled in Galápagos uses 'object biographies' or 'life stories' to create narratives around individual items collected from beaches in the archipelago (Schofield, 2018;Schofield et al., in press) to help understand how 30 they got there.
Fieldwork conducted in May and November 2018 involved collecting a representative sample of plastic items from a beach on San Cristobal. These items were then examined in a series of 'Science to Solutions' workshops involving academics and members of the local community, with the aim of building narratives around the coded and visual information each object contains. The coded information typically includes details of place and date of origin, and the original content (of containers), while visual inspection can betray length of exposure, for example through signs of bleaching and colonisation by marine life.
Preliminary results from the workshops can be compared to the results of the analyses reported here. Most plastic objects found on the beaches were of west-coast South American origin, with many bearing Peruvian and Ecuadorian labels, in agreement with the modelling here. In terms of the objects with Asian labels recorded on the beaches, the results are less clear. It 5 is suspected these objects had not been in the sea for long when they 'landed' in Galápagos, as all are very fresh. This latter observation accords with the results from the finding in this study that items released in Asia would not reach Galápagos. From the object biography workshops, the suggestion instead was that these items were coming from nearby fishing boats originating in SE Asia. This conclusion however is hard to reconcile with the results of the oceanographic modelling, that only a very small percentage of plastics from areas known to be popular fishing grounds would reach the archipelago. Working collaboratively, these very different disciplines and methodologies therefore illustrate both the benefits and some of the challenges of cross-disciplinary and cross-sector partnership to help understand (if not resolve) the challenge of marine plastic pollution.

Currents from Americas
Currents+waves from Americas 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7 10 8  Figure 6. Histogram of the time in days required for particles to travel from the west coast of America to the Galápagos region, for particles carried by currents only (yellow bars) and for particles carried by the currents and waves (blue bars). Most particles arrive within 3-4 months, although there is a significant tail all the way to 5 years for the simulation with currents only.   Left panel shows percentages for the currents-only simulation, right the percentages for the currents+wave simulation. Floating particles from most of these locations have a zero probability of ending up near the Galàpagos within 5 years (grey circles), but there are extensive regions of non-zero probabilities (coloured circles) near the Peruvian and Ecuadorian coast.