A methodology for quantifying and characterizing litter from trash capture devices (TCDs) to measure impact and inform upstream solutions

Prior to deploying a TCD, it is critical to determine the best type of trap, the best location, and the timing of deployment. First, it is important to understand spatial and temporal patterns of local contamination—which can inform the type of device, the location for its deployment, and the times of the year that the device should be deployed. It is best to strategically place a device where it can safely make the most impact, which may depend on major pathways for litter and accumulation hotspots. It also may depend on seasonal variability as well as any potential to pose negative effects on the ecosystem. To inform the most efficient and successful trash trapping programs, we developed a method for a visual audit, which quantifies and characterizes local anthropogenic debris across various locations.

Prior to conducting a visual audit, steps should be taken to plan and prepare. First, identify the times when an audit should be conducted relevant to when a trap may be deployed. For example, we do not keep our TCDs deployed in Winter in Toronto; thus, audits do not need to occur during winter months. Next, identify multiple potential sites for a TCD to audit. This may be based on expert or local stakeholder knowledge of where plastic pollution is common. A preliminary assessment can be conducted to aid in this process and make sure sites are accessible.

For our visual audit, 12 sites (see Fig. 1) were selected based on accessibility and previous reports published by a local non-profit organization. These reports identified areas along the Toronto Waterfront (Fig. 1) with high litter accumulation (Swim Drink Fish 2018). The visual audit covered nearly all the Central Toronto Waterfront from approximately The Waterfront Neighbourhood Centre (43°38′08.4″N, 79°23′49.0″W) to Sugar Beach (43°38′33.8″N, 79°22′02.1″W) (Fig. 1). Each site labelled on Fig. 1 was audited by walking the entire perimeter of each slip, identifying any anthropogenic debris as described below.

Our visual audit entails recording the type and amount of large anthropogenic debris (defined as anything larger than a bottle cap) visually identified. A semiquantitative estimate of small anthropogenic debris is also recorded (defined as anything smaller than a bottle cap). We also record general information on the weather and water conditions, which provide insight into temporal patterns of litter accumulation depending on local conditions. Specifically, we recorded wind, cloud coverage, swell, and whether a wet event occurred in the past 24 h (defined as precipitation over 10 mm). We aimed to conduct audits weekly between May and October (n = 2–8 audits per site) to coincide with a typical season for a TCD in Toronto. However, due to COVID-19, the audits only took place from August to October.

The audit is conducted by walking along a specified route (based on the entire perimeter of the chosen site) near the edge of the water to get the best possible view of the floating anthropogenic debris. To standardize collection among sites and across weeks, anthropogenic debris (which included plastic debris and non-plastic debris) was quantified by counting the plastic litter floating and characterized from a standing position and not bent over. If a piece of anthropogenic debris of interest needed to be closely examined, we would bend down to get a closer look but would not characterize or quantify any other debris while doing so. At each site, we first quantified and characterized debris by product type (i.e., plastic straw, single-use plastic bottle, takeout container) that was larger than a bottle cap. If we encountered a large accumulation of floating debris that was impractical to count (i.e., >50 pieces), a picture was taken, and the location of the photo was recorded on the datasheet with a mark on the map. Each photo was then analyzed later. The photos were examined from left to right and from top to bottom. Each item in the image was quantified and characterized, and if some items could not be identified, they were recorded as unidentifiable. These data were then added to the tally taken visually in situ. If anything unusual or noteworthy was identified during the audit, such as a large accumulation of organic matter or biota, this was recorded in the notes section of the datasheet. To semiquantitatively assess the amount of small anthropogenic debris at each site, four images were used as a guide to match which image best represents what is observed in the environment (Fig. 2). For the full protocol and datasheets, visit our website (https://oceanconservancy.org/wp-content/uploads/2021/03/Visual-Audit-Protocol.pdf; link is also given in Supplementary material).

Three Seabins were installed in the Outer Harbour Marina in Toronto in 2020 (43°38′34.6″N, 79°19′30.8″W) (Fig. 1). The location of these Seabins were chosen as a pilot to develop the methodology described below and to gauge the amount of maintenance needed to run a Seabin prior to choosing locations across the Toronto Waterfront. Two different methodologies were developed to quantify and characterize litter trapped by TCDs, using Seabins as a case study. The first is a “simple waste characterization,” intended to collect daily data in a simple and accessible manner that can be carried out by anyone maintaining the TCD. The second is a “detailed waste characterization” to collect data 5–10 times per year (aiming to capture variability by spreading out the data collection throughout each TCD season or year), where the content of each Seabin is meticulously sorted and characterized by material type. A season can be defined as that that the TCD is deployed annually. In Toronto, our TCD season typically occurs between May and October, which coincides with the ice-free season. Combined, these two methodologies provide a quantitative measure of the weight and count of anthropogenic debris diverted and information regarding what is collected to inform source-reduction upstream.

To test the effectiveness of our washing/extraction method, the organic material was kept to the side after having been rinsed three times and was sorted through after the initial analysis. We quantified and categorized all small anthropogenic debris in the washed leftover organic material. To test our subsampling technique, we quantified the three remaining piles left on the 2 mm sieve to compare their totals against the extrapolated totals from the original subsampled pile.

To determine whether there were significant differences in the amount of litter among each of the 12 visual audit sites, we ran a permutational multivariate analysis of variance (ANOVA) (PERMANOVA; adonis function in vegan R package, version 3.5.0, 999 permutations) due to unequal sample sizes. To determine the relationship between the mass of organic material and the mass of anthropogenic litter and quantity of small anthropogenic litter in the Seabins, we ran simple linear regressions. Regressions included the weight of the full Seabin on the x-axis and either the total weight of anthropogenic debris, total count of anthropogenic debris, and total weight of small anthropogenic debris or the total count of small anthropogenic debris on the y-axis. To explore the relationship between organic material and small plastic litter, we ran a linear regression with the weight of the plant material on the x-axis and the count of small anthropogenic debris on the y-axis. All analyses were performed with R (version 3.5.0) or Microsoft Excel (version 16.54). Statistical significance was considered at p < 0.05.

Twelve potential sites along the harbour were audited from May to October of 2020 between two and eight times. In future years, we suggest auditing each site the same number of times, which proved difficult for us during COVID-19. Each site had variable amounts of floating litter—with average accumulations of 10–276 anthropogenic litter items per site (Table S2; Fig. 1). The smallest average accumulation was observed at Harbour Square West Park and the highest at Peter Street Basin.

Overall, a PERMANOVA showed a significant difference among sites (PERMANOVA, p = 0.001). The sites on the west end of the waterfront tended to accumulate more anthropogenic debris, while sites to the east were all below 200 pieces. This difference appears to be mainly driven by the high counts of litter found at the Peter Street Basin. Across all sites and visits, more than 100 different types of anthropogenic debris were observed and recorded. The most common type of anthropogenic debris was a plastic bottle cap, and all types within the top 10 are identified as single-use plastics (Table 1).

Our semiquantitative assessment of the small debris concluded that 75% of sites had “little” small debris (Fig. 3). In contrast, we observed “no,” “moderate,” and “large” amounts of small debris at relatively the same frequency of occurrence across all sites. “No” small debris was mainly reported at Harbour Square Park West, and “large” small debris was reported most frequently at Peter Street Basin (Fig. 3).

While conducting the audits, part of the metadata collected was wind direction and conditions and whether a rain event had occurred in the past 24 h (>10 mm of rain). Anecdotally, we observed that an easterly wind was most common, and no trends were detected when accounting for precipitation. Collecting data at least once per week throughout an entire TCD season is recommended to capture more representative data. See Supplementary data for a detailed dataset.

The mass of the full catch bag (with organic and anthropogenic debris) was compared to two different metrics of anthropogenic debris: the count and the mass (kg). We found a significant positive relationship between the total mass of the contents within the bin and the total mass of anthropogenic debris extracted from the bin (large and small) (adjusted R² = 0.4; p = 0.007, 95% CI [0.002, 0.009 per kg of anthropogenic debris]); Fig. 5A). When looking just at the small size fraction, the relationship was still significant (adjusted R² = 0.2; p = 0.025; 95% CI [0.00007, 0.0009 per kg of anthropogenic debris] Fig. 5B), although the relationship was not as strong. In addition, we found a significant positive relationship between the total mass of the bin and the total count of anthropogenic debris (adjusted R² = 0.68; p = 0.0002; 95% CI [24.8, 58,7]; Fig. 5C), with the pattern driven by the count of small debris (adjusted R² = 0.63; p = 0.0004; 95% CI [19.8, 110]; Fig. 5D). To explore this further and assess what might be driving this pattern, we compared the mass of the organic material (aquatic macrophytes) to the total number of small debris in the Seabin. We found a similar, and significant, positive relationship between the mass of the plants and the count of small debris (adjusted R² = 0.68, p = 0.0002; 95% CI [20.9, 49.1]; Fig. S5).

Using the regression models (Fig. 5), we were able to estimate the total mass and count of anthropogenic debris diverted using the information from the simple protocol (i.e., mass of Seabin). Over the season, we estimate that we diverted a total of 13.1 kg (95% CI [0.002, 0.009 per kg]) of anthropogenic debris. This equates to about 4.35 kg of anthropogenic debris per bin. For large debris, this is an estimated 11.7 kg diverted in total or 3.9 kg per bin. For small debris, we estimate that we diverted 84 854 pieces, with an average of 28 284 pieces per bin.

Prior to selecting a TCD for a body of water, it is important to assess the area. The visual audit was created to be an effective tool for determining whether a TCD would be useful and have the most impact. Our visual audit aligns with a framework proposed by Helinski et al. (2021) to aid in choosing an appropriate location and type of TCD. Our protocol is meant to inform locations that enable effective plastic debris capture, that are accessible, and that do not present risks to the ecosystem. In our case study, it is clear that anthropogenic debris tends to accumulate more in specific sites (Fig. 1), providing valuable information on where devices should be placed to remove debris from our harbour and prevent it from travelling further into Lake Ontario. For example, if we could only place one TCD in our region, we would place it at Peter Street Basin according to our results. Visual audits can also be used to assess any potential adverse ecological impacts. If a visual audit determines there is not a lot of plastic litter and the costs of the device and program will outweigh the benefits, or that there is potential for ecological impact, a TCD should not be deployed.

A visual audit also allows for the identification of the type of device that will be most effective and suitable for their area of interest, preventing wasted time and resources that might result from purchasing the wrong device. Again, this recommendation aligns with the proposed framework by Helinski et al. (2021), which is to examine the type of site where anthropogenic debris tends to accumulate. For example, high amounts of debris in our region were found in slips—making types of TCDs that can be deployed on docks most effective.

Through this case study, we were able to design two protocols that are easy to use and accessible. Our goal is for them to be harmonized across a global network of local stakeholders effectively, diverting plastic debris from our environment with diverse types of TCDs. Our simple protocol uses low-cost and accessible materials, takes very little time to implement, and uses a free data collection app. The detailed protocol provides data on the amount of different types of debris to inform upstream solutions. This protocol takes more time (1–3 h, depending on the quantity of litter) but uses accessible materials (that can be purchased from a local hardware store) and is easy to follow. To make this protocol more accessible, we only recommend groups use it 5–10 times per year. See protocols for the list of materials and instructions on how to put together your supply kit.

Our data collection methods also inform the effectiveness of TCDs. The top three items collected by the Seabins were small anthropogenic debris categories: hard fragments, film, and pellets. This finding is interesting, as most TCDs, including the Seabin, were not designed to target microplastics (Schmaltz et al. 2020). Schmaltz et al. 2020 report that 81.5% of plastic pollution collection technologies focus only on macroplastics. Our findings suggest that existing TCDs can be a valuable mitigation strategy for microplastics, which is important as researchers have shown the negative effects of microplastics on freshwater organisms (Alimba and Faggio 2019; De Felice et al. 2019; Guimarães et al. 2021). However, it should also be noted that with the ability to capture microplastics, there is also the risk to capture and unintentionally trap biota (Falk-Andersson et al. 2020). During our TCD season, we experienced the unintentional collection of some biota such as juvenile fish, phytoplankton, and a juvenile turtle (the turtle was alive and safely released). In our case, these occurrences were rare. But, whenever biota is seen in a TCD, we suggest noting down what was found and the quantity. These records can be used to inform future strategies to reduce effects, such as moving TCDs to other locations or taking them out during certain times of year when risk is high.

Moreover, data collection can inform upstream solutions. Over the 19-week period, our Seabins diverted approximately 11 000 pre-production pellets, suggesting a focus on the industrial emission of pellets. This finding supports previous research suggesting upstream industrial facilities as a source of plastic contamination in the region (Corcoran et al. 2015; Ballent et al. 2016). Combined, this indicates a need for policies to prevent pellet loss at production sites and during storage and transport. TCDs can be a best management practice at the point of manufacturing to facilitate reduced pellet loss to the environment.

In addition to pellets, our Seabins collected over 40 different categories of anthropogenic debris, both large and small. Large films were one of the top 10 items collected (Table 1) and are often associated with degraded food packaging or agriculture mulch films (Ocean Conservancy 2020; Huang et al. 2020). This suggests local littering of food packaging and a potential need for more local garbage bins, educational campaigns, or bans on single-use plastic items. Cigarette butts were another top 10 items collected and are a major source of litter in many urban areas (Araujo and Costa 2019; Ocean Conservancy 2020). Research has shown that although many citizens recognize cigarette butts as litter, over 75% of smokers litter them after use (Patel et al. 2012; Rath et al. 2012). Educational campaigns, increased amounts of ash trays for proper disposal, and degradable filters could be upstream solutions. Here, we show that data can inform upstream solutions—allowing TCDs to be effective for both cleanup and prevention.

Our assessments suggest we have developed effective and efficient protocols to quantify and characterize the litter found in TCDs. Previous discussions with other groups who had a TCD suggested they were taking 6–7 h to characterize their catch. Our methodology speeds up the process with a simple extraction and subsampling. Of course, because we subsample, we may miss some of the detail of what is captured in a device. However, through validating our extraction methodology, we found it has a decent recovery rate of 60.6%, and that the data are representative to the diversity of debris collected. In addition, the subsampling method has a very low error rate when compared to the manual count. Our protocol provides enough accuracy and is time-efficient, which lessens the risk of community scientists losing motivation (Rambonnet et al. 2019).

Because small debris gets entangled in aquatic plants and algae, our extraction procedure did not capture all pieces of debris. The anthropogenic debris most unaccounted for were small anthropogenic debris: film and hard fragments. This may be due to the transparency of film as it is often hard to see when present on the plant material. Moreover, the morphology of some of the hard fragments and film may make them more adherent to the plant material and more likely to avoid separation even with the force of a high-pressured hose. Although this may present itself as a limitation, it brings light to the relationship between microplastics and aquatic macrophytes, which has been understudied yet stated to have an important role in the behaviour and fate of microplastics in aquatic systems (Kalcikova 2020).

To further explore this relationship, we ran regressions with the mass of plant material and count of small anthropogenic debris. The strong positive and significant correlation (adjusted R²= 0.68, p = 0.0002, Fig. S5) we observed suggests that aquatic macrophytes are a natural aggregator of anthropogenic debris, specifically microplastics that may otherwise be lost by TCDs due to a larger mesh size or other mechanisms for trapping trash that excludes small particles. The mechanism of aggregation by the plant material is unknown, but research suggests it may be driven by electrostatic interactions between plant cells and microplastics, leaf morphology, and the presence of periphyton (Kalcikova 2020). The interactions between microplastics and seagrass have been studied and shown to be potential vectors of microplastics through encrusting microplastics via epibionts. This allows them to naturally trap plastics when they create balls made of natural aggregates of vegetal fibres (Goss et al. 2018; Sanchez-Vidal et al. 2021). This highlights the importance for further work to understand the potential of aquatic macrophytes and their mechanisms acting as a natural TCD. Seasonality may be a significant variable in aiding the amount of plastics and microplastics caught by a TCD, as the density of plant material changes seasonally.

Finally, our regression models can be used to extrapolate the mass and count of anthropogenic debris collected over a 24 h period. The models include total mass of both large and small anthropogenic debris (kg) and count of small anthropogenic debris. Users can choose how to report their data, but we suggest reporting the counts of the small anthropogenic debris as mass does not convey the magnitude of debris being removed from the devices. We acknowledge the variability of our regressions; however, with the expansion of reporting through our network, we believe this will increase the predictability and accuracy of the models. Reporting both measurements also allows for diversified uses of the data. For example, mass is necessary when computing plastic flows and budgets, while monitoring plastics in various environments is reported by the number observed (Smith and Turrell 2021). Lastly, as Seabins are targeted to marinas that are usually located in coastal harbours or shallow embayment with similar vegetation, we recognize our model is likely less applicable to other traps and locations and plan to collect more diverse data in the coming years to strengthen the model.

To prevent the business-as-usual scenario for plastic pollution in 2030 predicted by Borelle et al. (2020), diverse initiatives must be implemented quickly. Along with reduced plastic production, better waste management, and beach cleanups, TCDs should be included as an additional solution. TCDs directly remove litter from our waterways 24 h a day and 7 days a week—robots never sleep.

To put TCDs to the task on a larger scale, we, the University of Toronto Trash Team (https://uofttrashteam.ca/), have created the International Trash Trap Network (ITTN; trashtrapnetwork.org) in collaboration with Ocean Conservancy. The goal of this network is to bring together independent stakeholder groups using any type of trash capture technology for global data collection, litter diversion, and impact. Within the network, our goal is that each group follow the simple data protocol and the detailed protocol to ensure standardized data collection. Eventually, these data can be collected using the CleanSwell app and linked to the open-access International Coastal Cleanup database.

At present, our protocols have only been tested with the Seabin and LittaTrap. We aim to collaborate and test our protocols with different devices and locations to ensure our subsampling is still representative of what is caught by each device. We intend to adapt our protocols for different devices as the network grows, and as new devices are developed. We also plan to increase the data that goes into our modeling so it is more globally representative.

The importance and necessity of open-access global databases have been highlighted at international plastic pollution conferences and in the peer-reviewed literature (Jambeck and Johnsen 2015; Kandziora et al. 2019). Various applications for data collection relevant to plastic pollution exist and allow people to collect and share cleanup and litter data, such as through Clean Swell, Marine Debris Tracker, and OpenLitterMap. Data from these global databases can inform policy, e.g., data from the Clean Swell app helped inform US adoption of the International Convention for the Prevention of Pollution from Ships (MARPOL 73/78Annex V) and the passage of the Microbead-Free Waters Act of 2015 (Ocean Conservancy 2011, 2016). By collecting data, we can also quantify the global impact of cleanup, inform source-reduction, and better understand baseline levels of contamination and how contamination changes over time.