Transport and accumulation of plastic litter in submarine canyons—The role of gravity flows

Marine plastic pollution is a pressing global challenge because of its deleterious impact on the environment and its ecosystems, on human health, and on social economy (U.N. Environment, 2019). It has been estimated that ∼4.8–12.7 million tons of plastics enter the oceans each year (Jambeck et al., 2015), and 70% of this amount would reach the seafloor (Pham et al., 2014). Compared to waste floating on the sea surface, benthic litter has received considerably less attention, primarily due to the difficulty of access. Limited studies have nevertheless shown that benthic litter is ubiquitous in the ocean (Tekman et al., 2017; Chiba et al., 2018).

Submarine canyons not only serve as conduits delivering vast amounts of sediments, nutrients, and pollutants into the deep sea, but also as sinks (permanent and/or transient) of large volumes of such materials (Fildani, 2017). Litter density in submarine canyons has been reported to be 2–3 times higher than that on adjacent open shelves and slopes (Pham et al., 2014; Cau et al., 2017; Kane et al., 2020). The mechanisms by which benthic litter is dispersed in submarine canyons are, however, largely unknown. It has been suggested that certain underflows, including internal tides (Schlining et al., 2013; van den Beld et al., 2017) and various types of gravity flows (Wei et al., 2012; Tubau et al., 2015; Daly et al., 2018; Kane and Clare, 2019; Pierdomenico et al., 2019; Pohl et al., 2020), might be responsible.

Here, we report the results of manned submersible dives that found focused accumulations of benthic litter in a submarine canyon of the South China Sea (SCS). We investigated the transport and depositional processes that formed these accumulations by documenting litter distribution together with grain-size and morphodynamic analyses.

Located in the northwestern SCS, the studied submarine canyon (18.10–18.64°N, 111.86–112.10°E) is headed at 350 m water depth on the outer shelf, ∼150 km from the nearest coast, and it terminates at ∼2350 m depth on the continental slope. At its distal end, it merges with the Xisha Trough, a large submarine canyon in the SCS (Fig. 1A). The shelf-indented canyon consists of a 20-km-long, relatively gentle upper reach, followed by an 8-km-long, steep and rugged middle reach, and a 32-km-long flat lower reach. The middle reach of the canyon, the focus of this study, consists of a long stepped chute leading into two successive large scours at the chute toe (Fig. 1B; Fig. S1 in the Supplemental Material¹).

Circulation on the northern SCS margin is characterized by three layers of western boundary currents (WBCs), i.e., the upper layer cyclonic (<500–1000 m water depth), the intermediate anticyclonic, and the deep cyclonic (>2000 m water depth) WBCs (Wang et al., 2011; Zhou et al., 2017; Zhu et al., 2019). Shelf-parallel currents also occur, including the Guangdong Coastal Current in the nearshore area, and the SCS Warm Current (SCSWC) mainly between the 200 and 400 m isobaths in the outer shelf to shelf-break zone (Fig. 1A; Guan and Fang, 2006). In addition, the margin is subject to frequent typhoons and strong internal waves (Alford et al., 2015).

We conducted seven manned submersible dives in May 2018 (dive 78), July 2019 (dives 159 and 172–173), and March 2020 (dives 227–229; Figs. 1–3). The initial purpose of these dives was to investigate the step-like morphology in the canyon’s middle reach. However, we encountered unexpectedly large litter accumulations in our first dive (Fig. 2; Fig. S1; Peng et al., 2019). These accumulations became the focus of the present study.

We used visual observations and video footage collected during the dives to investigate the occurrence of the benthic litter, locate the litter piles, and measure their sizes. The video footage was clear enough to allow us to identify litter items larger than 10 cm in size. The scattered litter items were individually counted along the dive tracks on each photographic snapshot, with density denoted by number of litter items per snapshot (nlps, Fig. 3). Areas of high density (≥4 nlps) in litter items were delineated (Figs. 2A, 2B, and 3). The length, width, height, and orientation of the litter piles were measured (Table S1). These estimates may have an error of up to 20%–30% as evaluated by repeat observations of individual litter piles. One 27-cm-long piston core was sampled at 1 cm intervals for grain-size analysis using a laser particle-size analyzer to understand the nature of the bottom sediment (Fig. 1C). Details of the background and methods are provided in the Supplemental Material.

Benthic litter in the canyon occurred either as scattered items or in piles (Figs. 2 and 3). Statistics from video footage shot along the canyon thalweg revealed that the scattered litter items were dominantly plastics (∼89%), and up to 88% of these plastics were distributed within the scours (Fig. S1). Litter density varied from 0 to >50 nlps (Fig. 3; Fig. S1). Notably, the highest litter densities were not found in the deepest portions, or centers, of the scours; instead, they were found on the west and south (downstream) sides of the scours (Figs. 2A, 2B, and 3). The west and south boundaries of the high-density litter areas were topographically higher than their east and north counterparts by 1.8–16.0 m and 6.4–12.6 m, respectively, although there was one exception at transverse profile T4 (Fig. 3). In general, litter density decreased laterally away from the scour centers along the transverse and longitudinal profiles (Fig. 3). Litter items near large boulders and topographic obstacles were noted to occur generally on their upstream sides (Figs. 2C and 2I–2K).

We delineated 71 litter piles in this study, all of which were distributed in the scours (Figs. 2A and 2B). The litter piles averaged 2–61 m in length, 0.5–8 m in width, and 0.1–1.2 m in height, with a maximum width and height of up to 12–15 m and 5–6 m, respectively (Table S1). The largest litter pile, consisting of three segments, measured 61 m in length and ∼241 m³ in volume (Figs. 2G–2H and 2L–2M). The litter piles were mostly northwest- to northeast-oriented, parallel or at an acute angle to the adjacent canyon thalweg. The remaining smaller litter piles (<5 m³ in volume), and those occurring upstream of obstacles, were roughly east-west–oriented, transverse to the valley (Figs. 2A–2B and 2I–2K). The majority of the litter piles were distributed in clusters on the up-valley dipping slopes downstream of the scour centers (Figs. 2A and 2B). Comparison between repeated dive observations showed that considerable changes in the configuration and components of litter piles had taken place (Figs. 2I and 2J).

Grain-size analysis showed that the bottom sediment consisted of clayey silt with sandy silt interbeds. The clayey silt was characterized by unimodal grain-size distributions with mode at 6–7Φ (15.6–7.8 μm), while a secondary mode at 2–4Φ (250–62.5 μm) was recorded for the sandy silt interbeds (Fig. 1C).

The heterogeneous but focused distribution of the litter indicates that it has experienced reworking in the canyon. Litter items that accumulated on the upstream side of obstacles suggest that they were transported by down-valley gravity-driven flows. The occurrence of poorly sorted sediment piles, including large blocks sometimes mixed with plastic items (Figs. 2E and 2F), may signal the occurrence of more cohesive flows (such as debris flows). The plastic particles and fragments appeared to be incompletely mixed with seabed sediment (Figs. 2 and 3), which may be a result of kinetic sieving (Middleton, 1970). The latter could potentially bring up relatively light and large plastic fragments on the top layer of the flow deposits. The unimodal grain-size distribution of the silt-dominated sediment is typical for turbidites (Visher, 1969) and different from hemipelagic deposits, which are often polymodal, being composed of a variety of sources (McCave et al., 1995; Holz et al., 2004). The bimodality of sandy interbeds is characteristic of many basal turbidite layers (Reynolds, 1987). Therefore, the transport and deposition of sediment and litter in the canyon are probably linked to turbidity currents, with lesser influence by submarine landslides.

The litter accumulations occurred largely at the immediate downstream end of the scour centers, which may be associated with changes in the morphodynamic conditions. Currents traversing the steep upstream slopes of the scours accelerate toward the central depression, most likely undergoing a hydraulic jump and decelerating against the upstream-facing slopes. The two scours at the chute toe are interpreted as two cyclic-step bed forms formed by alternating Froude supercritical and subcritical flows (Figs. 4A and 4B; Fildani et al., 2006; Cartigny et al., 2011). The decelerated post-jump subcritical flows may favor the deposition of litter items on the up-valley dipping slopes downstream of the scour centers (Fig. 4; Carvajal et al., 2017). Depending on the velocity of the decelerated currents, litter items could be either deposited further downstream by more powerful currents or laid down directly at scour centers by weaker currents (Figs. 2A and 2B). Large boulders and topographic features act as obstacles to litter items carried in turbidity currents, akin to log jams in fluvial systems.

Preferential distribution of the litter items to the west side of the scours (Figs. 2A, 2B, and 3) might be associated with the Coriolis effect, which would deflect the bulk of the currents to the right (looking downstream; Komar, 1969). The possible involvement of the intermediate WBC, active in the same depth range, is excluded considering that it flows eastward, contrary to the deviation (Fig. 1A).

Similar to plant materials in floods, the low-density, platy plastic waste tends to be transported in the upper layer of a gravity flow (Kane and Clare, 2019). Nevertheless, we cannot rule out the possibility that large platy plastic items with a particle size of 50–100 cm or more might be transported as traction loads in the decelerated post-jump currents, as indicated by the imbricated plastic items observed in some larger litter piles (Fig. 2D). This phenomenon may be explained by the settling velocity: Compared to the silt-dominated sediment in the canyon, plastics may have a much larger size and possibly an increased density due to the densification processes (e.g., mineralization, biofilms, and aggregates of sediments; Kane and Clare, 2019); larger plastic items might therefore have hydraulic equivalence with relatively finer mineral grains and be carried in the same parts of flows and be prone to tractional transport (Fig. 4).

We rule out the possibility of hyperpycnal flows in view of the headless nature of the canyon. Submarine landslides may constitute an important trigger of gravity flows as discussed above (Fig. 4A). Internal wave–induced return flows may be important in reorganizing the litter, as the SCS has the world’s largest known internal waves (Pomar et al., 2012; Alford et al., 2015). Being indented into the shelf, the canyon may obtain at its head a stable supply of litter and sediment from the shelf-parallel SCSWC and various shelf currents, especially the return density flows associated with typhoons (Fig. 4A). Litter and sediment stored in the canyon head may episodically fail and cascade down slope, resulting in the formation of gravity flows (Pohl et al., 2020). These flows should occur frequently, as inferred from the changing configuration and litter components of the same litter piles by time-lapse dive observations. Nevertheless, the possibility cannot be ruled out that litter items were delivered and deposited by other processes, including cross-shelf transport, up-valley internal tidal flows, slope-parallel WBCs, and direct settling from the water column, although most of litter items deposited by these processes may be finally reworked by gravity flows, as is suggested by the seabed sediment consisting primarily of turbidites. Finally, the litter accumulations in the scours would either be successively buried by newly deposited sediments or be further transported to the deep sea by large gravity flows.

While the larger-scale dispersal and accumulation patterns of deep-sea plastic waste remain unclear, our study suggests that the benthic litter accumulated in the scours of the submarine canyon studied here was most likely transported and deposited by gravity flows. Although the hypothesis of gravity flow–controlled litter dispersion has been raised by several authors, concrete evidence has been lacking; here, we present supporting evidence from mapping the distributions of both the scattered litter items and the litter piles, combined with grain-size and morphodynamic analyses. Our findings suggest that the headless canyon is subject to frequent turbidity currents and receives litter and sediment delivered by along-shelf currents. These plastic particles and fragments of different sizes are transported down the canyon by turbidity currents, as demonstrated by the grain-size distributions of associated sediment and the preferential distribution of litter piles. This plastic may be reworked on the bed into accumulations that preferentially occur upstream of obstacles and on the upstream-facing flanks of scours. Macroplastic litter, as a specific type of anthropogenic sediment, may be used to reconstruct modern underflows due its distinct characteristics (color, shape, density) that natural sediments do not possess. Finally, canyons may be a staging point for plastics that will ultimately be delivered to the deep ocean basins, where they will interact with delicate ecosystems, highlighting the need for mitigation of plastic waste dispersal into the natural environment.

We thank Pinxian Wang, Kang Ding, and Zhimin Jian for leading the expeditions, and the crew of the R/V Tansuoyihao, the pilot team of the manned submersible Shenhaiyongshi, and the onboard diving scientists for their technical support during the expeditions. We thank Andrea Fildani, Ian Kane, and an anonymous reviewer for their constructive comments, and Ian Kane for language polishing. This work was funded by the National Natural Science Foundation of China (grants 91028003, 41676029, and 41876049) and the National Key Research and Development Program of China (grant 2016YFC030490).