1. Introduction

Anthropogenic litter in the environment is causing undeniable harm (Beaumont et al., 2019; Campbell et al., 2019; Cole et al., 2015; Duncan et al., 2021; Lamb et al., 2018; Vegter et al., 2014). Mitigation strategies include direct removal (e.g. beach clean-ups; Zielinski et al. (2019)), and source reduction strategies including targeted legislation such as bans on individual types of packaging or products (e.g. Muposhi et al., 2022), or recycling incentives, such as material buy-back and container deposit schemes. Container deposit schemes offer a small financial incentive to return a container so that it can be recycled (White et al., 2001). These schemes work in increasing recycling rates (Hopewell et al., 2009; Kosior and Crescenzi, 2020) and reducing litter as a local source (Willis et al., 2022) but there is little published about their effectiveness in reducing observed local debris loads on beaches after implementation (Schuyler et al., 2018).

The accumulation of debris observed on beaches is dictated by natural processes such as supply from the local ocean (Turrell, 2018), the residence time on the beach (Kako et al., 2010), and storms or weather events out of the ordinary (Chubarenko and Stepanova, 2017), but also through anthropogenic processes such as littering behaviour (Campbell et al., 2014) and beach cleaning activities removing litter from beaches (Zielinski et al., 2019). Previous work by Schuyler et al. (2018), suggests that container deposit legislation reduces the amount of observed litter in local systems but focuses on the difference between states with and without container deposit legislation, at a much larger spatial scale than the current work. Willis et al. (2022) used systematic sampling at 100 km intervals around Australia at two-time points to assess the temporal changes in beached litter over that time. They found on average a reduction in litter present on Australian beaches, with a large range and variability across beaches. Here, we use a focused study area over a much longer time frame than either of these studies, with increased sample frequency to further explore litter dynamics in response to management intervention – in this case, the introduction of a container deposit scheme.

Community science data often comes from passionate volunteers who give up their free time because they want to see change or to have an impact on something they care deeply about, this is the case with much of the marine debris data available (e.g. Ambrose et al., 2019; Bauer-Civiello et al., 2018; Gacutan et al., 2023; Nelms et al., 2017). However, due to the nature of volunteer efforts, these schemes rarely have a strategized sampling method through space and time. While it is common that within a beach, methods have been standardised (e.g. the Marine Debris Initiative led by Tangaroa Blue), which beaches they visit and when is often dictated by weather conditions, volunteer availability, and cost. For many of these schemes, collecting the maximum amount of debris is the goal with the data being a useful by-product. However, these schemes have been running in some cases for over a decade, collecting large amounts of information about when and where debris accumulates. In this study, we aim to use this wealth of imperfect data to assess the effectiveness of a container deposit scheme in reducing litter loads in the local environment.

2. Methods

In this study, we will explore a case of the container deposit scheme (CDS) implemented in Queensland, Australia in November 2019. The scheme includes aluminium cans, glass bottles, PET and HDPE, liquid paper, and steel beverage containers between 150 ml and 3 l. The reported uptake of the scheme in Queensland has been good with an estimated 63 % of all eligible containers returned in the 2021–2022 financial year, and a total of 6.2 billion containers diverted from other waste streams since the start of the scheme (Container Exchange, QLD, 2022).

The local management area used for this case study is the Whitsunday region of the Great Barrier Reef World Heritage Area (Fig. 1). The resident population in the local management area is about 37,000 (2021 Census), the region is one of the key tourism areas of the Great Barrier Reef with domestic and international tourist spend approximately 3.5 million nights in the Whitsundays per year (Tourism Research Australia). The region receives two distinct weather seasons, 1) southeast trade windseason from approximately April to September with strong winds from the southeast (Fig. 1C) and 2) the more wind-variable northerly, monsoon wind season that occurs from about October to March (Fig. 1D). Beached litter data for the Whitsundays region was primarily collected and recorded by a local Non-Government Organisation, Eco Barge Clean Seas Inc. During each clean up event, volunteers collected all the debris that can be found from the water up into the vegetation of the back beach. The debris is loaded onto the barge and returned to shore where it is sorted using the Australian Marine Debris Initiative (AMDI) method cataloguing. The AMDI database stores counts of debris items in categories based on the material that the debris item is made of and its use (https://www.tangaroablue.org/resources/how-to-manual.html) and has extensive coverage of the Australian Coastline.

1. Download : Download high-res image (674KB)
2. Download : Download full-size image

Fig. 1. Map of the study area including A) clean-up locations coloured by the total number of visits across the dataset, B) the location of the Whitsundays in the Australian context, C) wind rose showing speed and frequency of wind received at Bowen Airport (Bureau of Meteorology station number 033327, 30 min observations, 2015–2018) during the Southeast Trade Wind season (approximately April to September), D) wind rose showing speed and frequency of wind received at Bowen Airport during the Monsoon season (approximately October to March). E) and F) show the current speed and direction recorded by daily satellite altimetry (IMOS gridded sea level anomaly product delayed mode, 1993–2020 extracted from coordinates 149.5,−20.5) for the Southeast Trade Wind and Monsoon Wind season, respectively.

Since Eco Barge Clean Seas began operations in 2009, the amount of debris removed by the Eco Barge team has been 1,991,153 items total and 34,088 total Plastic Drink bottles (debris category: “Plastic drink bottles (water, juice, milk, soft drink)”) alone from the environment (as of June 2023). However, by the nature of the community science data set, sites, visits, and number of volunteers are inconsistent through the years. This is often tied to inconsistent funding for a program such as this, and the non-systematic nature of site selection by volunteer programs. Therefore, standardisation of the data was required. The data were truncated to begin in 2012 as the first few years of the program had too few visits to meaningfully contribute to long-term trends, where there were only 2 visits per year to inconsistent sites in the years 2009–2011. There were no locations that were visited consecutively across all years, therefore we assume that the sites visited in any one year provide a representative sample of the debris load in the Whitsunday region. As community science organisation is run by locals with local knowledge, they go to the sites that accumulate most litter and are accessible on any given sampling day, making this assumption robust at the temporal scale of a year, and spatial scale of the Whitsunday region.

To assess the temporal trends in the long term dataset, we used a multiple before after control impact design (MBACI) (Keough and Mapstone, 1997). We ran a generalized linear mixed effect model (GLMs; lme4 package (Bates et al., 2015), in R programming language; (R Core Team, 2021)) with Analysis of Variance (ANOVA) to indicate significance for explanatory variables and interaction (car package (Fox and Weisberg, 2019)), using the effects package (Fox, 2003), to demonstrate significance. Our model includes Treatment which is the debris covered by the scheme or not which is two levels (CDS and non-CDS), Impact which is the implementation of the CDS with two levels pre-implementation (2012–2019) and post-implementation (2020−2022). We included random effects of site and year in the model to account for spatially explicit differences and the number of visits per site per year as an offset to account for different visitation rates at each site (for summary table, please see supplementary material). We also assume that when volunteers visit a site that the beach is totally cleared of litter regardless of number of volunteers. We justified this assumption by conducting a correlation analysis between the number of volunteers and the observed debris load for each collection event and found no significant correlation (Pearson’s r = 0.0067; p = 0.8905; see S2 in supplementary material).

To assess the impact of the container deposit scheme (CDS), we split the data into the categories that contain items collected by the scheme (“Plastic drink bottles (water, juice, milk, soft drink)” and “Aluminium cans”) and for other common litter items, which became the Treatment of the MBACI analysis. There are considerations when using legacy data, for example the plastic drink bottles category contains bottles that are not currently collected by the scheme and so our results may be diluted by those items. However, this will provide a conservative estimate of change. There are many categories that appear in the data only once, or very infrequently, therefore we selected only categories that appear in the dataset in at least 50 clean ups. We also removed one clean up event that was in response to a boat running ashore that we think did not reflect the normal debris load on the beach. We note that, in previous work, a container-to-lid ratio has been used to assess the effect of container deposit legislation at a national scale (Schuyler et al., 2018), however the community science data does not separate lids from containers that would be covered by the scheme and other lids; therefore we could not replicate that analysis. R code and data used for these analyses are available in an online repository linked with this article (doi:https://doi.org/10.5061/dryad.qjq2bvqn8).

3. Results

Clean-up events (n = 239) were spread across 90 sites (Fig. 1) in the 11-year study period (mean sites visited per year = 21.7, s.d. = 7.4), and events occurred in all months of each year. There is data from across the region, with site receiving uneven numbers of visits (median visits per year = 1, range 1–5).

The number of debris items collected per visit on Whitsunday beaches for both items covered by the container deposit scheme and those not covered was variable between years (Fig. 2). For CDS debris, there was significantly less debris after the implementation of the scheme (Fig. 2a; Table 1) with a mean of 120.4 (sd 78.4) containers collected per visit before the implementation compared with 77.5 (sd 48.2) after the implementation. The same trend did not exist for non-CDS debris (Fig. 2b; Table 1) with mean items per visit 5162.6 before and 6001.9 after (sd 4046.7, 3475.4 respectively). According to the multiple before after control impact analysis the implementation of the container deposit scheme had a significant effect on the amount of CDS debris present on the Whitsundays region of the Great Barrier Reef Marine Park, seen by the significant interaction between treatment and impact (Table 1) i.e. before and after implementation of the scheme and the debris covered by the scheme and the debris not covered (Fig. 3).

1. Download : Download high-res image (363KB)
2. Download : Download full-size image

Fig. 2. Weighted debris collected per visit across all sites each year. Weighted debris is calculated as the total count of debris collected per site in a year divided by the number of times the location was visited. The dashed line indicated the initiation of the container deposit scheme at the end of 2019. n = 239 clean up events A) only considering the debris covered by the CDS, B) the common debris items not covered by the CDS. Please note the difference in scale of the y-axis.

Table 1. Analysis of deviance table (Type II Wald chisquare tests).

Variable	Z Statistic (Chisq)	Degrees of freedom	P-value
Treatment (CDS)	554,633.729	1	<0.001
Impact (pre and post implementation)	0.485	1	0.4861
Treatment ∗ Impact	3051.834	1	<0.001

1. Download : Download high-res image (122KB)
2. Download : Download full-size image

Fig. 3. Effects plot showing the predicted debris counts based on the generalized linear model. Star added to denote significant difference as per analysis of devience.

4. Discussion

COVID-19 impacted the abundance and types of litter in the environment, in most cases there was in increase, especially in instances of personal protective equipment found in the environment (Abedin et al., 2022; Nghiem et al., 2021; Peng et al., 2021; Torres and De-la-Torre, 2021). However, our results indicate there was a reduction in container debris, likely due to increase recycling participation (Schuyler et al., 2018) and compounded by lower levels of littering during “lockdowns” and voluntary periods of reduced socialisation and activities.

The Whitsunday region likely receives litter from local littering, however, with the dramatic drop in both debris classes observed in 2022, when the region received higher than pervious tourist visitation (Tourism Research Australia, 2023), suggests the source of debris on the outer islands is more likely to be from remote sources transported by ocean currents. Buoyant objects transported by the wind (Critchell and Lambrechts, 2016) will likely be transported from the populated areas in the south by the southeast trade winds (Fig. 1). Debris in the Pacific Ocean and Coral Sea could also be a source of debris to the region, however the mechanism of transport is less clear. The fall in observed CDS debris from 2020 to 2022, not seen in the other debris, suggests an impact from the container deposit scheme. This was likely coupled with some lock-down effect, where debris may have been reduced due to restricted movements of people in populated South East Queensland, reducing littering in general in 2020. The dramatic decrease in both classes of debris in 2022, could be a lag effect of the lockdowns in Southeast Queensland, however, data from 2023 and beyond will be necessary to confirm this.

Gacutan et al. (2023) used community science effort to identify drivers of accumulation on beaches, they undertook uniform sampling across Queensland at 17 beaches across three years. Their findings provided unique insights into drivers of accumulation, however, they found no effect of lock downs to the total litter abundance on beaches. They note that the sampling locations were often remote, which may reduce the effect of additional PPE debris and reduction of litter that we observe here. Our data is complemented by Sherow et al. (2023), who found that COVID-19 related lockdowns reduced the abundance of larger plastic items found in storm drains in Melbourne, Australia. Strom drains provide an opportunity to sample very close the potential litter sources, with a short delay between the littering event and the sampling event compared to beach sampling, as we have in this study. The increased time, and distance, between littering and sampling could allow litter to become diffused into the environment making signals of changes in source volume more difficult observe. However, the clear signal observed in this study shows that local action to reduce the amount of plastic lost from the waste management system can have a direct impact on coastal environments. This confirms the finding by Schuyler et al. (2018), who compared states with and without container deposit schemes to assess the effectiveness of the strategy. Here we show the success of a newly implemented scheme in a single region showing that time lag between implementation and evidence in the environment can be short. With schemes about to begin elsewhere, for example in Victoria, Australia, this is a promising opportunity to confirm this with strategic before and after sampling.

Our work shows that legacy clean-up data can be used to effectively see changes brought about by policy changes. It is likely that the effect is diluted due to the broad categories used then quantifying debris, however the signal is present, which should encourage further targeted action towards litter classes at local, state, and national scales. Our statistical analysis was made possible by the effort of the volunteers on the ground in the Whitsunday region. Many of the beaches in the region are relatively small with moderate debris loads which allows volunteers to collect much of the debris (see effort volunteer correlation analysis Sup Mat.), in other regions where beaches are much longer, or debris loads much higher, this may not be the case. We suggest groups attempting to replicate our work conduct their own analysis of effort and observed loads to check that the assumption holds in their region. If effort needs to be considered, we suggest observed debris load be scaled by number of volunteers and time spent on the beach to make the data comparable across beaches and clean up events. The work of Eco Barge Inc. and their dedicated volunteers is not the only example of sustained effort towards quantifying ocean litter, we believe that an approach such as the one presented here could be used on other legacy datasets to detect changes in litter abundance through time. Key decisions around which classes to include should be undertaken so that the analysis is capturing the appropriate data.

Our ability to monitor the success or failure of policy implementation, and schemes such as container deposit schemes, depend on the sampling design and effort. The uptake of the AMDI sampling and cataloguing protocols by Australian and international community science organisations is a great step forward in the ability to use the data collected to answer scientific questions. However, to increase the ability to detect changes in baseline abundance of debris in the environment we recommend that community science organisations visit the same locations at the same time of year. If possible visiting locations multiple times a year will allow for more robust and straightforward comparisons through time. Volunteer effort is such a valuable resource and as data collection efforts continue, we will gain further information on these promising trends.

References