Vous êtes sur la page 1sur 11

Marine Pollution Bulletin 135 (2018) 862–872

Contents lists available at ScienceDirect

Marine Pollution Bulletin

journal homepage: www.elsevier.com/locate/marpolbul

Type and quantity of coastal debris pollution in Taiwan: A 12-year T

nationwide assessment using citizen science data

Bruno A. Walthera, , Alexander Kunzb, Chieh-Shen Huc
Department of Biological Sciences, National Sun Yat-sen University, Gushan District, Kaohsiung City 804, Taiwan
National Taiwan University, Department of Geosciences, No.1, Sec.4, Roosevelt Road, Taipei 10617, Taiwan
Society of Wilderness (SOW), No.204, Zhao'an St., Zhongzheng Dist., Taipei City 100, Taiwan


Keywords: Man-made coastal debris pollution is a growing concern for Taiwan. In 2004, Taiwanese environmental orga-
Beach cleaning nizations led by the “Society of Wilderness” began gathering data on 19 categories of debris items collected
Citizen science during cleanup events. We present our analysis of the resulting 12-year dataset collated from 541 events held
Coastal conservation between October 2004 and December 2016. In total, 904,302 items weighing 131,358.3 kg were collected, and
Marine anthropogenic litter
63.6% and 27.2% of items were made of either plastic or plastic mixed with other materials, respectively. The
Marine macro-debris
Plastic pollution
five most commonly recorded debris categories were plastic shopping bags, plastic bottle caps, disposable ta-
blewares, fishing equipment, and plastic drinking straws. We estimated that during the 12-year period on
average between 3.7 and 7.9 million items weighing 560–1110 metric tons polluted Taiwan's coastline. We offer
recommendations for improving the quality of data collected during Taiwan's cleanup events and report some
policy changes due partly to previous reports of this dataset.

1. Introduction 2014; Koelmans et al., 2014; Seltenreich, 2015; Thompson et al., 2009;
Vethaak and Leslie, 2016; Wilcox et al., 2015; Wilcox et al., 2016). As
Plastic pollution is a rapidly worsening environmental problem in plastic objects fragment, they break into ever smaller pieces, which
terrestrial habitats (Thompson et al., 2009) but even more so in have been classified as macroplastics, mesoplastics, microplastics, and
oceanic, coastal, and riverine habitats (Barboza and Gimenez, 2015; nanoplastics (Laglbauer et al., 2014; Lee et al., 2013).
Bergmann et al., 2015; Gross, 2015; Wilcox et al., 2015). This problem Possible impacts on human health are: (1) accidents (see above); (2)
is rapidly growing because global plastic production and waste gen- the direct ingestion of microplastic particles via food, mostly seafood,
eration have been growing exponentially, having reached approxi- and the possible resulting internal injury (Dehaut et al., 2016; Rochman
mately 335 million metric tons (MT) in 2016 (Geyer et al., 2017; et al., 2015); (2) the indirect contamination of air, food, and water with
PlasticsEurope, 2017). Of these, ~4.8–12.7 million MT are estimated to unhealthy substances leached from the plastics (Talsness et al., 2009;
enter the oceans annually (Jambeck et al., 2015). This estimate is much Thompson et al., 2009); and the possibility of microplastics serving as
greater than the reported global mass of floating plastic debris in the pathogen vectors (Vethaak and Leslie, 2016).
world's oceans which was estimated at 0.25 million MT (Eriksen et al., Concerns about plastic pollution should be especially pertinent to
2014). It is thus unknown what happens to the vast amount of plastic the Taiwanese public because (1) man-made debris pollution (which is
debris entering the oceans. predominantly macroplastic pollution) has reached pervasive and cat-
Once in the environment, plastic objects and fragments (1) cause the astrophic proportions along Taiwan's coastline (Kunz et al., 2016; Kuo
injury and death of animals through entanglement and ingestion, (2) and Huang, 2014; Liu et al., 2013; Society of Wilderness, 2014; this
damage and endanger shipping vessels, (3) visually and structurally study) and (2) a relatively high proportion of people's diet comes from
damage oceanic, coastal, and riverine environments, including a ne- seafood (Fisheries Agency, 2007–2012; National Library, 2017). How-
gative effect on tourism and the economic costs of continuous cleanup ever, despite the urgency and magnitude of the problem, only three
operations, (4) spread invasive species, and (5) eventually break down scholarly articles have been published so far, one about microplastic
to microplastics which can enter the food chain directly or contaminate (Kunz et al., 2016) and two about macroplastic pollution. Liu et al.
it via the leaching of chemical ingredients (Gregory, 2009; Jang et al., (2013) surveyed four beaches on one Taiwanese island from August

Corresponding author.
E-mail addresses: bawalther2009@gmail.com (B.A. Walther), a.kunz96@yahoo.com (A. Kunz), jason@wilderness.tw (C.-S. Hu).

Received 18 April 2018; Received in revised form 10 August 2018; Accepted 10 August 2018
0025-326X/ © 2018 Elsevier Ltd. All rights reserved.
B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872

Fig. 1. Locations of 541 cleanup events conducted between October 2004 and December 2016 in Taiwan. Note that one event took place in the Matsu Islands, two
events each on Xiaoliuqiu [=Lamay] Island and Orchid Island, and 34 events on the Penghu Islands. Events on islands were excluded in some analyses (see main
text). The map was made with ArcGIS version 10.1.

2009 to October 2011 and documented the types and proportions of (3) document the types, numbers, and weights of debris items. Since
debris types (of which 78.3% was macroplastics). Kuo and Huang there was no government-led or ENGO-led monitoring program, the
(2014) surveyed six sites in northern Taiwan from June 2012 to May SOW decided in 2008 to adopt the International Coastal Cleanup (ICC)
2013 and documented the types, proportions, categories, and sources of sampling scheme which had been developed by the Ocean Conservancy
debris types (of which 85.5% was macroplastics). Therefore, these two (2018a) in the 1980s (Wikipedia, 2018a). Another reason to begin
studies had a relatively limited spatiotemporal scope. cleanup events was that the Taiwanese public was beginning to em-
To highlight this emerging problem of coastal pollution, Taiwanese brace environmental activities, e.g., Earth Day, and that there was
environmental non-governmental organizations (ENGOs) began edu- therefore a demand for environmental activities. Consequently, the
cation campaigns and coastal cleanup events in the 2000s. One ENGO SOW studied the sampling scheme and debris categories developed by
at the forefront of this effort has been the “Society of Wilderness” the ICC and adopted the methodology to Taiwan's unique situation (see
(SOW, www.sow.org.tw) which began to organize coastal cleanup Materials and methods for more details).
events in 2008 in order to (1) decrease coastal pollution, (2) involve From the beginning, one of the aims of collecting these data was to
and educate Taiwan's public and media (cleanocean.sow.org.tw), and be able to put pressure on the government to consider source reduction

B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872

Table 1 (Cigliano et al., 2015). All reviews concluded that citizen science is an
19 categories of coastal debris which were used to identify each debris item effective approach to conservation and education as well as the gen-
during 546 cleanup events (see main text for details and www.sow.org.tw/ eration of large and valuable datasets which increase the available in-
blog/iccdataforms and Fig. S29 for Chinese-language instructions for cleanup formation on marine litter sources, distribution, and ecological impacts.
events and definitions and illustrations of each debris category).
In order to gain a nationwide and long-term understanding of
Category Taiwan's coastal pollution problem, we analyzed the 12-year dataset on
coastal debris pollution collated by the SOW which includes 541
Plastic materials only
cleanup events conducted between October 2004 and December 2016.
1. Polyethylene terephthalate (PET) bottles, (category not used between March
2006 and the end of 2010) This dataset only contains macroplastic items (defined as > 25 mm, Lee
2. Plastic bottle caps separated from the bottle et al., 2013), while smaller fragments were ignored. This is a unique
3. Other plastic (non-PET) bottles, plastic food containers and plastic trays, all used dataset because it is the only long-term dataset and the only dataset
for edible contents
which covers the entire nation's coastline. Specifically, we (1) analyzed
4. Plastic bottles (e.g., motor oil, shampoo, detergents, etc.), plastic containers
(e.g., plastic trays) and plastic cases (e.g., CDs), all for non-food contents
the types, numbers, and relative percentages of different kind of coastal
5. Plastic shopping bags with handles debris, (2) investigated which variables determine the number and
6. Plastic food wrappers without handles (e.g., to hold fries), and other plastic weight of collected debris items, and (3) estimated the total number
packaging bags (category used from 2015 onwards, before was in category and weight of debris items found around the entire coastline of Taiwan.
“plastic shopping bags with handles”)
We also make some recommendations about how to improve the data
7. Plastic drinking straws
8. Fishing nets and ropes only, usually used for commercial fishing: count definition collection by Taiwanese volunteers.
somewhat unclear, because some volunteers counted each item separately, while
other volunteers counted one ball with different pieces of nets and ropes tangled 2. Materials and methods
together as one item
9. Fishery and aquaculture buoys, floats, and boat bumpers usually used for
commercial fishing and excluding nets and ropes (see 8 above): count definition
2.1. Study area
somewhat unclear, because some volunteers will enter fragmented foam buoys as
one item, while other volunteers counted each fragment separately Taiwan is a nation state in East Asia consisting of one large island
10. Plastic toothbrushes (category used mostly from 2015 onwards, before was not (Taiwan Island) and several much smaller islands and island groups
(Fig. 1). Taiwan Island covers latitudes 22°–25°18′N and longitudes
Plastic mixed with other materials
11. Beverage cups for hot and cold drinks (different kinds, some all styrofoam, 120°27′E–122°E with a maximum elevation of 3952 m. It can be
some all plastic and some paper-plastic mixtures) (category used from 2011 roughly divided into an almost flat western plain, which has been
onwards, before was not used) highly modified by humans, and the mountainous areas in central and
12. Disposable tablewares (e.g., utensils, lunch boxes, dishes, chopsticks, cutlery - eastern Taiwan which comprise almost 65% of the island and are much
materials plastic, paper, bamboo, wood)
13. Laminated and usually cuboid-shaped drink and juice boxes, e.g., Tetra Paks
less developed. The length of the coastline of Taiwan Island is
14. Pole fishing gear, obviously used for recreational fishing (thus excluding the 1338.517 km (Construction and Planning Agency, 2017). We also ob-
items under 8–9 above): lines, hooks, small floating devices, lures, fluorescent tained the length of the coastlines for some of the other smaller islands,
rods, and related packaging – materials are mixed, e.g., metal hooks but plastic but nevertheless only used the length of Taiwan Island's coastline in our
lines; count definition somewhat unclear, because some volunteers counted each
analysis because it was not always clear on which of the smaller islands
item separately, while other volunteers counted one ball with different gears
tangled together as one item the cleanup events took place.
15. Cigarette butts
16. Plastic syringes and metal needles 2.2. Cleanup events
17. Cigarette lighters
Non-plastic materials, but could have plastic labels
18. Tin and aluminium cans
The ICC has been organized by the Ocean Conservancy (www.
19. Glass bottles oceanicsociety.org) since 1986 (Ocean Conservancy, 2016). In October
2004, the Taiwan-based Kuroshio Ocean Education Foundation (www.
kuroshio.org.tw) organized the first coastal debris monitoring and
policies by documenting what types of debris items accumulate on cleanup event in Taiwan. The SOW joined these events in 2008 and has
Taiwan's coasts. For example, by documenting the overwhelming pre- been collating the data since then. In addition, the SOW created a
sence of single-use plastics in the collected debris, SOW was able to put website (cleanocean.sow.org.tw/result.php) in 2015 where volunteers
pressure on the government to regulate them; this has recently brought can download the instructions for a cleanup event, including data-
some successes (see Discussion). Another advantage of using citizen sheets. The data can then be directly entered into the SOW website by
volunteers was the fact that a nationwide coverage of Taiwan's coastline the volunteers.
was possible, which is not possible for small teams of scientists (Kuo During each cleanup event, volunteers were split into several col-
and Huang, 2014; Liu et al., 2013). lection teams if necessary, and each collection team then collected the
The use of citizen science to tackle this problem was thus a con- human-made debris along a coastal strip. The length of this strip was
scious choice on the part of the SOW. By its very nature, citizen science recorded, but not the width which was variable because of the different
involves and thereby educates a large number of people. For example, coastal topographies. Assigned members of each collection team re-
citizen science is now used on large spatial and temporal scales to corded the numbers for each debris category (see below) on pre-printed
monitor biodiversity (e.g., Dickinson et al., 2010; Hochachka et al., datasheets (Fig. S29, Text S1). All data were then entered into the da-
2012; Hong et al., 2018; Sullivan et al., 2014). The use of citizen sci- tabase by the third author.
entists to monitor marine pollution has also grown exponentially over For each cleanup event recorded between October 2004 and
the last 30 years and was recently reviewed in Thiel et al. (2014), December 2016, data for most or all of the following variables were
Cigliano et al. (2015) and Hidalgo-Ruz and Thiel (2015). While there is collected and entered into the database (for various reasons, data for
a diversity of approaches and goals to citizen science, most studies fo- some variables were not recorded during some events, cf. Tables 1 and
cused on the distribution and composition of coastal marine debris over 2):
local, regional, and international scales (Hidalgo-Ruz and Thiel, 2015).
There are generally six conservation-related outcomes that citizen sci- 1. Date and location of the event, including the length of the cleaned
ence projects can potentially support: policy, education, community coastal strip and its geographic coordinates. We used this informa-
capacity building, site management, species management, and research tion to generate a variable ‘time’ which is the integer number of days

B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872

Table 2
Median, mean, standard deviation and range (minimum-maximum) of 22 variables collected during 541 cleanup events (cf. Materials and methods and Table 1). The
percentages were calculated from the total number of recorded items (namely 904,302 items).
Coastal debris category (n) Median Mean ± st. dev. Minimum–maximum Sum (%)

Number of volunteers (498)a 48 123.8 ± 211.9 1–1660 61,639

Weight of debris in kg (435)a 90 302.0 ± 594.4 1–4500 131,358.3
Length of coastal strip in m (459)a 300 660.1 ± 923.2 9–7100 303,002
Plastic material items
PET bottles (486)a 24 106.0 ± 246.2 0–2654 51,540 (5.7%)
Plastic bottle caps (541) 59 193.5 ± 338.9 0–2839 104,679 (11.6%)
Non-PET bottles (541) 33 104.8 ± 204.1 0–2088 56,678 (6.3%)
Non-food plastic bottles etc. (541) 4 26.3 ± 60.3 0–545 14,251 (1.6%)
Plastic shopping bags (541) 59 256.0 ± 698.2 0–8026 138,474 (15.3%)
Plastic food wrappers (305)a 10 42.9 ± 106.9 0–1377 13,091 (1.4%)
Plastic drinking straws (541) 34 146.8 ± 327.2 0–3212 79,395 (8.8%)
Fishing nets and ropes (541) 13 54.5 ± 115.8 0–1044 29,506 (3.3%)
Buoys, floats, boat bumpers (541) 20 159.0 ± 441.8 0–5625 86,032 (9.5%)
Plastic toothbrushes (335)a 1 5.0 ± 11.7 0–86 1663 (0.2%)
Plastic mixed materials items
Beverage cups (445)a 23 86.9 ± 160.1 0–1142 38,689 (4.3%)
Disposable tablewares (541) 32 176.8 ± 608.7 0–9802 95,647 (10.6%)
Laminated drink boxes (541) 6 21.9 ± 41.4 0–453 11,295 (1.2%)
Pole fishing gear (541) 9 34.0 ± 64.0 0–510 18,393 (2.0%)
Cigarette butts (541) 12 108.0 ± 322.8 0–3301 58,443 (6.5%)
Plastic syringes, metal needles (541) 1 7.1 ± 18.0 0–174 3863 (0.4%)
Cigarette lighters (541) 10 37.0 ± 75.8 0–727 20,010 (2.2%)
Non-plastic material items
Tin and aluminium cans (541) 7 23.5 ± 52.5 0–521 12,728 (1.4%)
Glass bottles (541) 34 129.3 ± 301.7 0–3984 69,925 (7.7%)
Plastic material items (541) 408 1063.4 ± 1698.8 0–11,548 575,309 (63.6%)
Plastic mixed materials items (541) 150 455.3 ± 959.7 0–10,755 246,340 (27.2%)
Non-plastic material items (541) 47 152.8 ± 328.6 0–4135 82,653 (9.1%)
Total number of items (541) 698 1671.5 ± 2621.2 14–16,884 904,302 (100.0%)

These variables were not collected during part of the study period (cf. Table 1).

which have passed since the first collection date which is 1 October event, we estimate that about 80–90% of all visible debris had been
2004. removed. The reasons for not reaching 100% were usually an in-
2. The number of volunteers which participated. sufficient number of volunteers and/or that the remaining debris
3. The entire collected debris was weighed to the nearest kilogram, was too small, large, or dangerous to handle. Since plastic debris
including non-plastic materials, using a mechanical platform scale. disintegrates into smaller and smaller fragments, down to micro-
After each cleanup, the collected debris was either recycled in a plastics (≤5 mm) (Kunz et al., 2016), the volunteers could not
nearby governmental recycling center (about 20% of the total collect all of these small fragments, but concentrated on collecting
debris) or picked up by a garbage truck which transported it to an all the larger items which were still whole or largely whole (as de-
incinerator or a landfill. fined above).
4. The number of items which were collected and removed from the
coastal strip was recorded for 19 categories of coastal debris To provide a visual impression of some of the coastline types around
(Table 1). The SOW developed these 19 categories, originally using Taiwan and the typical levels of coastal debris pollution as well as the
the 45 categories developed by the ICC sampling scheme but then cleanup events, we provide a series of illustrative photos (Figs. S1–S23).
eliminating categories which were not pertinent to Taiwan (e.g.,
bullet shells), and also further reducing the number to make the
2.3. Statistical analyses
categories easier to use for Taiwanese people. Over time, a few ca-
tegories (1, 6, 10, and 11) were added because of additional data
All of our variables were very right-skewed (namely, time, number
needs, e.g., category 11 was added in 2011 because free, single-use
of volunteers, length of coastal strip or strip length, number of debris
beverage cups became an important issue in Taiwan. These 19 ca-
items, weight of debris items, and the 19 categories of coastal debris).
tegories have worked very well because they are easy to understand
Therefore, all variables were Box-Cox-transformed (Krebs, 1999) to
and follow. Types of coastal debris not listed in Table 1 (e.g.,
achieve a normal distribution (tested by Shapiro-Wilk test, all variables
wooden furniture, rubber tires, shoes, motorcycle helmets, pens,
had p > 0.05 after transformation). These transformations allowed us
etc.) were not recorded but were weighed and removed unless it
to use simple and multiple linear regressions in SPSS version 19 for
proved impossible (e.g., the item was too big or dangerous). Items
testing relationships between variables. We used both forward and
were only counted if the item (e.g., bottle, buoy, cup, food tray, etc.)
backward selection of variables to see if results differed depending on
was at least 50% intact. The exact definition translated from Chinese
the selection method. However, results were always the same regardless
is: “If more than 50% of the original volume of an item is remaining,
of the selection method.
the item is classified as a whole item, and if less than 50% of the
In one case, we tested a correlation with the Spearman-rank test
original volume is present, the item is classified as a fragment.” The
because the data were not normally distributed.
main reason for this definition was to avoid counting obviously
fragmented debris (e.g., foam buoys, styrofoam boxes, etc.) as sev-
eral items. Thus, fragments were disregarded in the total count of 2.4. Number and weight of debris items
items but were included in the total weight. After each cleanup
We used two methods for calculating the number and weight of

B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872

debris items per kilometer, using the pooled data for the entire period Z = 2.32, rho = 0.67, p = 0.02; Fig. 2). The number of volunteers of
from October 2004 to December 2016. Therefore, the calculations de- the 541 cleanup events varied from 1 to 1660 people (Table 2). This
scribed below give us an estimate of the average state of pollution for large variation of volunteers explains the large variation among events
the entire 12-year period without regard to obvious spatiotemporal in the number of items of coastal debris recorded (14 to 16,884 items),
variation (see Discussion). the weight of the debris collected (1 to 4500 kg), and the length of
coastal strip cleaned (9 to 7100 m) (see also Section 3.3 below).
2.4.1. Calculation method 1
The direct way of estimating the number and weight of items per km 3.2. Types and numbers of coastal debris
of coastline is to simply divide the total number and the total weight of
items by the combined length of the coastal strips over which they were Table 2 summarizes the numbers and ranges of the 19 debris cate-
collected. However, the problem with this calculation method was that gories (Table 1) collected during the 541 cleanup events, as well as the
strip length was a worse explanatory variable than the number of vo- number of volunteers, the weight of the debris, and the length of coastal
lunteers for both predicting the number and weight of items (see strips which were cleaned.
Results). Considering the mean number of items and their percentage of the
total, the five most commonly recorded debris categories were plastic
2.4.2. Calculation method 2 shopping bags, plastic bottle caps, disposable tablewares, fishing
Therefore, we also used a method based on our regression analyses equipment, and plastic drinking straws. Looking at the median, the top
with the number of volunteers as the independent variable (Figs. 3–5) five debris categories were plastic shopping bags, plastic bottle caps,
for the following reasons: (1) the regression analyses clearly showed plastic drinking straws, glass bottles, and non-PET bottles. Considering
higher correlation coefficients and higher p-values for the number of the mean and median number of items, the five least commonly re-
volunteers than for the length of coastal strip (see Results), and (2) the corded debris categories were toothbrushes, syringes and needles, la-
influence of outliers was decreased in the regression analyses because of minated drink boxes, tin and aluminium cans, and non-food plastic
the Box-Cox transformations. Indeed, the residuals around the linear bottles, containers, and cases. Looking at the percentage of the total, the
equations shown in Figs. 3–5 are all normally distribution (tested by bottom five debris categories were toothbrushes, syringes and needles,
Shapiro-Wilk test) which demonstrates that outliers do not unduly in- laminated drink boxes, plastic food wrappers, and tin and aluminium
fluence the estimation of the linear regression equations. We therefore cans.
assert that the correlations displayed in Figs. 3–5 are the best descrip- Six categories which are predominantly used to carry drinks (PET
tions of the signal in our data as they “contain most of the information” bottles, plastic bottle caps, non-PET bottles, plastic drinking straws,
hidden within the data (sensu Breiman, 2001). We then used the re- beverage cups, laminated drink boxes) combined for a total of 47.0% of
gression equations of Figs. 3–5 to calculate what we consider to be more all items. Two categories which are predominantly used to carry foods
reliable estimates of the number and weight of debris items per kilo- (plastic food wrappers, disposable tablewares) combined for 12.0% of
meter. all items. Plastic bags accounted for another 15.3%, and three cate-
gories for fishing equipment combined for 14.8% of all items. Two
3. Results categories related to smoking (cigarette butts, cigarette lighters) com-
bined for 8.7% of all items. The remaining < 10% of items belonged to
3.1. Cleanup events four categories: glass bottles, cans, syringes, and toothbrushes.
For every debris category, the median is below the mean (Table 2)
The 541 cleanup events were distributed all around Taiwan Island's which emphasizes the right-skewed data distributions of all the vari-
coastline, with concentrations in the north and southwest; 39 events ables. In other words, most events collected few items of each category,
took place on other islands (Fig. 1). The number of events increased and only very few events collected many items of a debris category.
significantly from 2004 to 2016 (Spearman-rank test, n = 13 years, In total, the 61,639 volunteers recorded 904,302 items of coastal










2004 2005 2006 2008 2009 2010 2011 2012 2013 2014 2015 2016

Fig. 2. Frequency diagram of the number of cleanup events (n = 541) from 2004 to 2016.

B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872


Total weight of debris 2.5



-0.5 0 0.5 1 1.5 2 2.5 3 3.5

Number of volunteers
Fig. 3. Scatterplot of the number of volunteers versus the total weight of the debris items (see statistics in Table 3). The relationship is described by
Y = 0.828 ∗ X + 0.528. Note that the Box-Cox transformation means that the actual numerical values do not correspond with the recorded numbers (cf. Fig. S24).

debris weighing 131,358.3 kg. The mean and median number of items 3.4. Number and weight of debris items
collected per event were 1671.5 and 698, respectively (weighing 302
and 90 kg, respectively). Of these items, 63.6% were made of plastic 3.4.1. Calculation method 1
and a further 27.2% were made of plastic mixed with other materials. Using only the 421 events which occurred on Taiwan Island and for
which we have data for both the number of items and strip length, we
calculated 780,682 items/285.941 km = 2730 items/km. Using only
3.3. Explanatory variables the 378 events which occurred on Taiwan Island and for which we have
data for both the weight of all items and strip length, we calculated
The need for the Box-Cox transformation of our variables was im- 115,502 kg/276.831 km = 417.2 kg/km. When we multiplied these
mediately apparent when we plotted the raw data (Figs. S24–S28) be- numbers with Taiwan's coastline length of 1338.517 km, we calculated
cause these scatterplots show the non-normal distribution and hetero- estimates of 3,654,151 items and 558,468 kg (approximately 3.7 mil-
scedasticity of our data which would lead to the undue influence of lion items and 560 MT).
outliers on the calculation of the regression equations. After Box-Cox
transformations of all of our variables, these problems were amelio-
3.4.2. Calculation method 2
rated (Figs. 3–5).
We calculated the number and weight of items per km of coastline
The ‘number of volunteers’ is the variable which best explains the
using the three correlations depicted in Figs. 3–5. The Box-Cox trans-
variation of the other 25 variables, explaining between 20% and 61% of
formation for strip length was a log(x) transformation, so we inserted
the variation, with all relationships being positive and highly sig-
the y-value of log(1000) = 3 for a 1000 m stretch of coastline into the
nificant (Table 3). The other variable which could be expected to ex-
equation given in Fig. 5. Solving this equation for x yielded an x-value
plain variation in the other 24 variables listed in Table 3 is the ‘length
of 2.889 (which is therefore the number of participants expected for a
of coastal strip’ which was cleaned. Indeed, the ‘length of coastal strip’
1000 m long coastal strip using the relationship depicted in Fig. 5).
correlates positively and significantly with these 24 variables (r2-value
Inserting this x-value of 2.889 into the equations given in Figs. 3 and 4
0.01–0.24, p-value always < 0.0001, see also Figs. 5 and S26–S28).
yielded y-values of 2.920 and 3.774, respectively. Translating these y-
However, when we entered ‘number of volunteers’ and ‘length of
values back into real numbers [the Box-Cox transformations of the
coastal strip’ together as independent variables into a multiple linear
number and total weight of debris items were also log(x) transforma-
regression against the remaining 23 variables, the additional variation
tions] yielded estimates of 5937 items/km and 831 kg/km. When we
explained by ‘length of coastal strip’ decreased to 0–1%, and the p-
multiplied these numbers with Taiwan's coastline length, we calculated
values were all > 0.05. We therefore concluded that the ‘length of
estimates of 7,946,775 items and 1,112,308 kg (approximately 7.9
coastal strip’ had relatively little influence on the recorded numbers or
million items and 1110 MT).
weight of debris items, while the ‘number of volunteers’ was con-
sistently and positively correlated with the recorded numbers and
weight of debris items (Figs. 3–4). 4. Discussion
The variable ‘time’ was not significantly correlated with any of the
25 variables in Table 3. The variable ‘time’ was also not significantly Since it affects the entire Discussion below, the first issue to discuss
correlated with the percentages calculated for the 23 variables listed in is the fact that we are dealing with somewhat ‘messy’ data in this study
rows 3–25 in Table 3 (e.g., the percentage of PET bottles calculated as because the data were produced by citizen volunteers, not trained sci-
the number of PET bottles over the total number of items). entists. In order to ensure data quality, studies involving citizen sci-
entists should include (1) preparation of easy and straightforward
protocols, (2) volunteer training, (3) supervision by professional

B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872


Total number of items




-0.5 0 0.5 1 1.5 2 2.5 3 3.5

Number of volunteers
Fig. 4. Scatterplot of the number of volunteers versus the total number of debris items (see statistics in Table 3). The relationship is described by
Y = 0.749 ∗ X + 1.610. Note that the Box-Cox transformation means that the actual numerical values do not correspond with the recorded numbers (cf. Fig. S25).

participation, and (4) validation of data and samples (Hidalgo-Ruz and Velde et al., 2017). A Chilean study which recounted microplastic
Thiel, 2015). Our cleanup events only included the first two: (1) there particles collected by schoolchildren in the laboratory concluded that
were easy instructions on the header of every survey sheet (Fig. S29, the schoolchildren were able to follow the instructions and generate
Text S1); (2) most of the organized cleanup events included volunteer reliable data (Hidalgo-Ruz and Thiel, 2013). On the one hand, these
training, although some of the smaller or individual events did not; two studies give some confidence into data generated by citizen sci-
points (3) and (4) were not included, which are therefore included in entists; on the other hand, the conclusions of these studies cannot be
our recommendations below about possible improvements which could transferred one-to-one to Taiwan's situation which is obviously dif-
be made to the data collection during Taiwan's cleanup events. An ferent.
Australian study which compared the marine debris data collected by Despite efforts to improve our data quality, the lack of (3) super-
researchers and citizen scientists concluded that overall the citizen vision and (4) validation means that our data likely contain some un-
science data were of equivalent quality to the researcher data (van der known noise (i.e., random variation) and bias (Walther and Moore,


Length of coastal strip



-0.5 0 0.5 1 1.5 2 2.5 3 3.5

Number of volunteers
Fig. 5. Scatterplot of the number of volunteers versus the length of coastal strip (see statistics in Table 3). The relationship is described by Y = 0.404 ∗ X + 1.833.
Note that the Box-Cox transformation means that the actual numerical values do not correspond with the recorded numbers (cf. Fig. S26).

B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872

Table 3 were produced for the convenience of carrying drinks, and another 12%
Linear regression of the independent variable ‘number of volunteers’ against 25 for the convenience of carrying foods. Thus, about 60% of the recorded
dependent variables. We report the sample size (N), F-value, the r2-value and debris items originated from the single-use food and drinks packaging
the standard coefficient, while the p-value was < 0.0001 for all analyses. industry. Given that many plastic bags (which made up 15% of all
Sample sizes were variable because some variables were not always recorded
items) are also used for the single-use of carrying food and drink, the
(Table 2). Only Box-Cox transformed variables were used.
percentage was even higher than 60%. The three categories of fishing
Dependent variable N F-value r2-Value st. coeff. equipment accounted for another 14.8% of the entire debris. Given that
fishing equipment is also one of the main causes of entanglement of
Weight of debris 435 544.7 0.56 0.75
Length of coastal strip 459 145.0 0.24 0.49
marine animals (Gregory, 2009; Wilcox et al., 2016), concerted efforts
PET bottles 446 222.1 0.33 0.58 should be made to drastically reduce the discarding and loss of fishing
Plastic bottle caps 498 243.5 0.33 0.57 equipment. For example, Kuo and Huang (2014) suggested to en-
Non-PET bottles 498 334.7 0.40 0.64 courage fishermen to recycle fishing gear and to turn that gear into
Non-food plastic bottles 498 133.3 0.21 0.46
energy sources. The most important non-plastic debris categories were
Plastic shopping bags 498 363.4 0.42 0.65
Plastic food wrappers 305 84.0 0.22 0.47 glass bottles, tin cans, and aluminium cans. However, all container
Plastic drinking straws 498 281.5 0.36 0.60 items receive payments when recycled in Taiwan, so there is already a
Fishing equipment 498 248.7 0.33 0.58 financial incentive for people to collect and return them. Nevertheless,
Fishing nets and ropes 498 251.5 0.34 0.58 too many people still discard them, so perhaps deposit schemes should
Plastic toothbrushes 313 77.7 0.20 0.45
Beverage cups 445 287.7 0.39 0.63
be introduced for these items, as well as for certain plastic items, e.g.,
Disposable tablewares 498 304.3 0.38 0.62 PET bottles, to further increase the financial incentive for recycling.
Laminated drink boxes 498 164.9 0.25 0.50 We were not able to detect any temporal trends in debris categories,
Pole fishing gear 498 149.4 0.23 0.48 e.g., such as an increasing or decreasing percentage of PET bottles over
Cigarette butts 498 137.7 0.22 0.47
time. While there were some changes over time (unpublished results),
Plastic syringes, metal needles 498 224.8 0.31 0.56
Cigarette lighters 498 273.3 0.36 0.60 these were not statistically significant.
Tin and aluminium cans 498 268.3 0.35 0.59 Although almost all of the debris categories have proven negative
Glass bottles 498 471.7 0.49 0.70 consequences for environmental and human health (Vethaak and Leslie,
Plastic material items 498 637.2 0.56 0.75 2016), the fact that 3863 plastic syringes and metal needles were found
Plastic mixed materials items 498 471.3 0.49 0.70
Non-plastic material items 498 513.1 0.51 0.71
is especially worrisome, given the possibility of serious injury and in-
Total number of items 498 785.9 0.61 0.78 fection (Cheung, 2018; Sheavly and Register, 2007; Williams et al.,
2005). Circumstantial evidence (labels of glass containers, location
downstream from large animal farms, the disposal of syringes used in
2005). However, the advantage of using citizen scientists is the possi- human health is strictly regulated in Taiwan) points towards the fact
bility to gather large amounts of data, as for this study. Another po- that most or all of these syringes were used for veterinary purposes.
tential bias may lie in the fact that Taiwan's coastline is composed of Very few previous publications (e.g., Bartram and Rees, 2000; Himans,
sandy beaches, rocky coasts, and reef coasts (covering approximately 2013; Laglbauer et al., 2014) have specifically noted syringes in their
50%, 44%, and 6% of the coastline, cf. Fig. 1 in Yang et al., 2010). findings, but there are many additional reports on the internet (e.g.,
Unfortunately, coastal type was not recorded, so we cannot exclude Tong, 2017; Wikipedia, 2018b). Clearly, there is a need for further
some bias based on preferential sampling of sandy beaches, for ex- regulation and enforcement.
ample. Since sandy and rocky coasts are found mostly on Taiwan's The numbers in Table 2 are likely minimum estimates for the fol-
western and eastern coastlines, respectively (Yang et al., 2010), our lowing reasons. 1. As pointed out in the Materials and methods, items
Fig. 1 suggests that there may have been some bias towards cleaning which were less 50% intact were not recorded. For example, for every
sandy beaches on Taiwan's western coast (which is also where more piece of macroplastic removed, there were hundreds if not thousands of
people live). pieces of mesoplastic and microplastic which remained (Figs. S18–S20)
We therefore recommend that our numerical results should not be (Kunz et al., 2016). Previous studies demonstrated that a considerable
seen as absolute and precise, but as generally indicative of the overall amount of plastic debris is not collected because it is overlooked (Lavers
situation (see also our Discussion of orders of magnitudes below). For et al., 2016) or buried in the sediment (Lavers and Bond, 2017). 2. Rare
example, the three debris categories plastic shopping bags, plastic types of coastal debris were not recorded, and large or dangerous items
bottle caps, and disposable tablewares were the only ones with > 10% were not removed. 3. Some debris items which were neither large nor
of the total number of debris items (Table 2). We would contend that dangerous could not be removed because they were partially covered
these item categories were therefore among the most abundant, but we by sand or rocks, making them too difficult to retrieve (Figs. S16–S17).
would not be confident in stating that there were significantly more With all these considerations in mind, the numbers are nevertheless
plastic bags than plastic bottle caps, even though a t-test comparing truly staggering. Eriksen et al.'s (2014) global analysis of oceanic plastic
their numbers yielded a p-value of 0.02. For this reason, we did not use pollution places Taiwan into the red zone of those areas with the
inferential statistics in Table 2. In other words, the numbers should be highest weight density of pollution with plastic pieces > 200 mm. Our
seen as indicative of an overall ranking of the debris categories. average estimates for the 12-year period bear this out. Our minimum
Despite these considerations, it is immediately apparent that the estimate (calculation 1) was 2730 items/km and 417.2 kg/km, resulting
vast amount of coastal debris was either made entirely of plastic or in a total of approximately 3.7 million items weighing 560 MT for
made of plastic mixed with other materials (about 90% of the total, Taiwan's entire shoreline. Our higher and probably more reliable esti-
Table 2), and that most of these items were made for single-use. The mate (calculation 2) was 5937 items/km and 831 kg/km, resulting in a
predominance of plastic items among the man-made coastal debris is total of approximately 7.9 million items weighing 1110 MT for Taiwan's
thus in complete agreement with previous studies (e.g., Barnes et al., entire shoreline. Reassuringly, our higher estimate of 831 kg/km is re-
2009; Coe and Rogers, 1997; Derraik, 2002; Schmuck et al., 2017; markably similar to the estimate of 740 kg/km when 3250 sites cov-
UNEP, 2005), including previous studies in Taiwan (Kuo and Huang, ering about 1% of Japan's coastline were surveyed in 2006
2014; Liu et al., 2013). However, this is the first study in Taiwan to (Anonymous, 2007).
show the predominance of plastic items for such a large spatiotemporal To further check the reliability of our estimates, we compared our
scale. results with the results from an independently collected dataset which
Our results also clearly show that almost half of the disposed items the Environmental Protection Administration (EPA) of Taiwan has

B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872

collated since 2009 (Table S1). Excluding the unusually high value of remained about 400–800 kg/km and 3000–6000 items/km of debris on
2009 due to typhoon Morakot, the mean weight of coastal debris re- Taiwan's coastline during our 12-year period. Future studies should also
moved annually during the period 2010–2017 was 6160.6 MT (mean of focus on documenting the spatiotemporal changes in pollution and
column 4 in Table S1). The mean total length of coastline cleaned an- reasons thereof.
nually during the same time period was 15,681.3 km (mean of column 2 Given the uncertainties discussed above, we make seven re-
in Table S1), which is about 9.3 times the entire length of Taiwan's commendations how to improve the data gathering and verification
coastline. In other words, the entire coastline was cleaned nine times during cleanup events.
each year on average. Dividing 6160.6 MT by 9.3 yields 662.4 MT
which is slightly above our estimate based on calculation method 1 but 1. As mentioned above, data gathering can be improved by profes-
substantially lower than our estimate based on calculation method 2. sional supervision and the validation of data and samples (Hidalgo-
The EPA data includes one category (bamboo and wood, which is Ruz and Thiel, 2015).
mostly driftwood) which was not recorded in our cleanup events (Text 2. Additional variables should be collected: length and width of the
S1) and one indeterminate category (Table S1). Clearly, our dataset is coastal strip should be measured accurately; geographic coordinates
not the same but similar to the EPA dataset. Therefore, it is again re- of both ends of the cleaned coastal strip; any information of source
assuring that our estimate and the EPA estimate are similar in magni- (e.g., photos of product labels); any information on whether regular
tude. professional cleaning is taking place by asking local hotel and tourist
The highest ever recorded density of macroplastic debris was operators; an estimate of human accessibility of the site (Schmuck
239.40 items/m2 on the uninhabited Henderson Island (Lavers and et al., 2017); and coastal type (such as sandy beach, rocky shore,
Bond, 2017). On Caribbean islands, the mean density of macro debris harbor, etc.). Given the different percentages of the three coastal
was 6.34 items/m2 (Schmuck et al., 2017) while it was 1.80 items/m2 types in Taiwan (Yang et al., 2010), we also recommend that efforts
on Chilean beaches (Bravo et al., 2009), 1.25 items/m2 on Slovenian are made to sample representative numbers of sites in each coastal
beaches (Laglbauer et al., 2014), 1.00 items/m2 on South Korean bea- type, and that sites are distributed more evenly along Taiwan's
ches (Lee et al., 2013), 0.90 items/m2 on South Taiwanese beaches (Liu coastline because there were too few cleanup events along the less
et al., 2013), 0.20 items/m2 on Mediterranean beaches (Munari et al., accessible rocky eastern coastline (Fig. 1).
2016), and 0.19/m2 on North Taiwanese beaches (Kuo and Huang, 3. During each organized cleanup event, a subgroup of interested vo-
2014). Since our estimates were not measured in items/m2 but in lunteers should be trained to sample a smaller subsection (e.g.,
items/m, we cannot compare our estimates of ~3–6 items/m directly 10 m2 divided into small squares or transects, see, e.g., Fig. 2 in
with these other estimates. Using Google Earth, the four sandy beaches Opfer et al., 2012) very intensively and systematically, perhaps even
examined in Kunz et al. (2016) have a mean width of ~50 m which collecting mesoplastics and microplastics, and collecting up to a
yields an estimate of 3–6 items/m/50 m = 0.06–0.12 items/m2. Since depth of 10 cm. Such an intensively sampled subsection would yield
many rocky beaches around Taiwan are much narrower, the mean much more complete and reliable data on debris numbers and
width which was cleaned during our 541 cleanup events was probably weight per m2 (and not per m as in this study). This subsection
below 50 m, but there is no way of knowing for sure since only length sampling should be accompanied by at least professional super-
was recorded. Nevertheless, it is again reassuring that our estimates are vision (and perhaps also data validation, see above).
similar in magnitude to already published estimates. 4. Given that debris also blows into the dunes and other areas adjacent
The comparisons of our results with the results from Japan's to the coast (Fig. S21), some collection should also take place in
coastline (Anonymous, 2007), from the EPA dataset, and from the eight these areas to estimate the levels of pollution spreading from the
scientific studies all demonstrate that the estimates which we calculated coastal areas. Such sampling should adhere to the same sampling
using citizen-science data compare reasonably well with the numbers protocol as the coastal sampling, unless modifications are needed
from these other independent sources. Given that our numbers corre- (e.g., thorny vegetation does not allow sampling across a continuous
spond well with other sources and are not ‘off the charts’ gives us higher area).
confidence in the validity of our results, despite the disadvantages of 5. More efforts should be made to communicate the results of the
citizen-science data which we also acknowledge. cleanup events in a timely fashion back to the volunteers, the public,
A final consideration has to be that the spatiotemporal distribution and the media (Eastman et al., 2014). While the present study is
of coastal debris is highly dynamic. First, wind, waves, and water such an effort to communicate the results, we recommend to pro-
currents constantly move debris items around, e.g., blowing them into duce a yearly report which can be disseminated to all volunteers, the
the dunes and even further inland (Fig. S21). Storms, and especially public, and the media (e.g., such as Ocean Conservancy, 2018b).
typhoons, deposit vast amounts of debris onto coastlines (Figs. S9–S10, This would reward the volunteers with some recognition and si-
S22–S23) (Fig. 1e and f in Kunz et al., 2016) (Doong et al., 2011; Liu multaneously generate regular public attention for this important
et al., 2013). For example, Figs. S22–S23 probably show two storm issue. Furthermore, Cigliano et al. (2015) discussed how to involve
events which deposited two separate layers of debris in a coastal sedi- citizen scientists in policy change, education, community outcomes,
ment. This constant stream of deposition is counterbalanced by the and site management outcomes.
constant removal by citizens and professional collectors with almost 6. Several studies had very good results when involving school chil-
monthly cleanups on average (Table S1 and Discussion above). Fur- dren (Eastman et al., 2014; Hidalgo-Ruz and Thiel, 2013; van der
thermore, there are also citizens who regularly remove coastal debris Velde et al., 2017). Therefore, more efforts should be made to take
but who do not communicate any data to the EPA, SOW, or any other school children to the cleanup events.
ENGO (B. Kretzschmar, pers. commun. 2017). Moreover, popular bea- 7. One study conducted in the Georgia Sea developed a smartphone
ches or beaches associated with nearby hotels are cleaned profession- application to enter data, and the analysis of geo-referenced photos
ally at regular intervals to keep the beaches attractive to tourists (see has also been tried (Hidalgo-Ruz and Thiel, 2015). The Ocean
Fig. 1f and g in Kunz et al., 2016) because not cleaning beaches would Conservancy (2018a) has also developed an app called CleanSwell.
result in financial losses (Jang et al., 2014; Lamb, 2018). Not all of these Given the almost complete coverage of mobile phone ownership in
cleanup activities are necessarily reported to the EPA or SOW. Conse- Taiwan, SOW decided to provide a mobile-web for smartphone
quently, the total number and weight of debris items along Taiwan's users, and the development of a suitable mobile phone application
coastlines would be much higher if it was not for the constant removal should be considered.
by either citizen volunteers or professional cleanup operations (e.g.,
Strong, 2017). However, even with these constant removals, there In summary, despite some obvious shortcomings of the present

B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872

dataset, it is the best dataset at present to obtain estimates of the types References
and quantity of the coastal debris pollution in Taiwan in the absence of
a professionally conducted research program. We believe that adopting Anonymous, 2007. Nationwide Survey of the Drifted Litter Japan. Japanese Ministry of
our above recommendations can realistically be achieved and would Land, Infrastructure, Transport and Tourism (in Japanese).
Barboza, L.G.A., Gimenez, B.C.G., 2015. Microplastics in the marine environment: current
make the citizen science data even more valuable. Other advantages are trends and future perspectives. Mar. Pollut. Bull. 97, 5–12.
low costs for large amounts of data, large spatial coverage, and the Barnes, D.K.A., Galgani, F., Thompson, R.C., Barlaz, M., 2009. Accumulation and frag-
educational value (Cigliano et al., 2015; Nelms et al., 2017). mentation of plastic debris in global environments. Philos. Trans. R. Soc. Lond. B 364,
Recently, the absence of a professional program and the growing Bartram, J., Rees, G., 2000. Monitoring Bathing Waters: A Practical Guide to the Design
realization of the urgency and magnitude of the plastic pollution pro- and Implementation of Assessments and Monitoring Programmes. Taylor & Francis,
blem (Walther, 2015) has triggered an effort by Taiwan's EPA in col- New York, New York, USA.
Bergmann, M., Gutow, L., Klages, M., 2015. Marine anthropogenic litter. Springer
laboration with eight ENGOs (including Greenpeace Taiwan, Hiin Openhttp://link.springer.com/book/10.1007/978-3-319-16510-3.
Studio, Kuroshio Ocean Education Foundation, Tse-Xin Organic Agri- Bravo, M., Gallardo, M.d.l.Á., Luna-Jorquera, G., Núñez, P., Vásquez, N., Thiel, M., 2009.
culture Foundation, Foundation of Pescadores Citizens, SOW, Taiwan Anthropogenic debris on beaches in the SE Pacific (Chile): results from a national
survey supported by volunteers. Mar. Pollut. Bull. 58, 1718–1726.
Environmental Information Association, and Wild at Heart Legal De-
Breiman, L., 2001. Statistical modeling: the two cultures. Stat. Sci. 16, 199–231.
fense Association) to launch a large-scale quantitative beach debris Cheung, K., 2018. In pictures: environmentalist seeks medical attention after collecting
monitoring program in 2018 as well as to educate the public and reduce dozens of syringes washed up on Lamma beach. Hong Kong Free Presshttps://www.
the use of plastic products that cause marine pollution (EPA, 2018; hongkongfp.com/2018/04/17/pictures-environmentalist-seeks-medical-attention-
collecting-dozens-syringes-washed-lamma-beach/ (17 April).
Yang, 2018). Using a preliminary analysis of the dataset used in this Cigliano, J.A., Meyer, R., Ballard, H.L., Freitag, A., Phillips, T.B., Wasser, A., 2015.
study, SOW published reports about coastal pollution (Taiwan Making marine and coastal citizen science matter. Ocean Coast. Manag. 115, 77–87.
Macroview Television Service Online, 2016; Society of Wilderness, Coe, J.M., Rogers, D.B., 1997. Marine Debris: Sources, Impacts, and Solutions. Springer-
Verlag, New York, New York, USA.
2014, 2016) which received widespread media coverage and were in- Construction and Planning Agency, 2017. The length of the natural and artificial coastline
strumental in raising public attention and springing the government of Taiwan (R.O.C.). Ministry of the Interior, Taipei, Taiwanhttp://www.cpami.gov.
into action. Indeed, the government has now decided to implement a tw/chinese/index.php?option=com_content&view=article&id=14251&
timeline of banning certain single-use plastic items. By 2020, plastic Dehaut, A., Cassone, A.-L., Frère, L., Hermabessiere, L., Himber, C., Rinnert, E., Rivière,
shopping bags, plastic drinking straws, beverage cups, and disposable G., Lambert, C., Soudant, P., Huvet, A., Duflos, G., Paul-Pont, I., 2016. Microplastics
tablewares will be partially banned, with the intention of introducing in seafood: benchmark protocol for their extraction and characterization. Environ.
Pollut. 215, 223–233.
payment costs in 2025 and likely complete bans in 2030 (EPA, 2018; Derraik, J.G.B., 2002. The pollution of the marine environment by plastic debris: a re-
Leary, 2018; Lin, 2018). Taiwan's society annually consumes an esti- view. Mar. Pollut. Bull. 44, 842–852.
mated 60,000 MT of disposable tableware, 18 billion plastic bags, 4.5 Dickinson, J.L., Zuckerberg, B., Bonter, D.N., 2010. Citizen science as an ecological re-
search tool: challenges and benefits. Annu. Rev. Ecol. Evol. Syst. 41, 149–172.
billion PET bottles, 3 billion plastic straws, and 5–10 billion beverage
Doong, D.-J., Chuang, H.-C., Shieh, C.-L., Hu, J.-H., 2011. Quantity, distribution, and
cups. Therefore, any source reduction policy which reduces these impacts of coastal driftwood triggered by a typhoon. Mar. Pollut. Bull. 62,
numbers substantially would have a positive impact on ecosystems, 1446–1454.
tourism, and the environmental sustainability of Taiwan's society. Eastman, L., Hidalgo-Ruz, V., Macaya-Caquilpán, V., Nuñez, P., Thiel, M., 2014. The
potential for young citizen scientist projects: a case study of Chilean schoolchildren
collecting data on marine litter. J. Integr. Coast. Zone Manag. 14, 569–579.
5. Conclusions EPA, 2018. Marine debris action plan for Taiwan. Environmental Protection
Administration, Taipei, Taiwanhttps://enews.epa.gov.tw/enews/fact_Newsdetail.
asp?InputTime=1070213135101 (in Chinese).
Our study is the first long-term and nationwide study of the type and Eriksen, M., Lebreton, L.C.M., Carson, H.S., Thiel, M., Moore, C.J., Borerro, J.C., Galgani,
quantity of coastal debris pollution in Taiwan. It is also the first one to F., Ryan, P.G., Reisser, J., 2014. Plastic pollution in the world's oceans: more than 5
exclusively use data generated by citizen volunteers. We used these trillion plastic pieces weighing over 250,000 tons afloat at sea. PLoS ONE 9, e111913.
Fisheries Agency, 2007–2012. Taiwan fisheries yearbook. Fisheries Agency, Council of
data to document the most abundant types of coastal debris as well as to Agriculture, Executive Yuan, Taipei, Taiwanhttps://www.fa.gov.tw/en/
estimate the approximate quantity of debris found around Taiwan PublicationsYearbook/.
during our 12-year data collection period. Based on our experiences and Geyer, R., Jambeck, J.R., Law, K.L., 2017. Production, use, and fate of all plastics ever
made. Sci. Adv. 3, e1700782.
results, we make seven recommendations how to improve the data
Gregory, M.R., 2009. Environmental implications of plastic debris in marine settings-
gathering and verification. To improve the catastrophic conditions entanglement, ingestion, smothering, hangers-on, hitch-hiking and alien invasions.
around Taiwan's coasts, Liu et al. (2013) and Kuo and Huang (2014) Philos. Trans. R. Soc. Lond. B 364, 2013–2025.
Gross, M., 2015. Oceans of plastic waste. Curr. Biol. 25, R93–R96.
furthermore called for (1) increased cleanup frequencies, (2) more
Hidalgo-Ruz, V., Thiel, M., 2013. Distribution and abundance of small plastic debris on
source-reduction policies, (3) more environmental education and out- beaches in the SE Pacific (Chile): a study supported by a citizen science project. Mar.
reach, (4) better recycling and waste management efforts, and (5) more Environ. Res. 87-88, 12–18.
collaborative monitoring and research efforts. Although marine debris Hidalgo-Ruz, V., Thiel, M., 2015. The contribution of citizen scientists to the monitoring
of marine litter. In: Bergmann, M., Gutow, L., Klages, M. (Eds.), Marine
pollution is a global issue, its adverse effects impact at both the local Anthropogenic Litter. Springer Open, pp. 429–447. http://link.springer.com/book/
and the global level. Therefore, partnerships with local governments, 10.1007/978-3-319-16510-3.
communities, and industry may also help to reduce marine debris. Himans, I.P., 2013. Assessment of Marine Debris and Water Quality Along the Accra-
Tema Coastline of Ghana (MPhil thesis). University of Ghana, Accra, Ghana. http://
Acknowledgements %20Himans_Assessment%20of%20Marine%20Debris%20and%20Water%20Quality
We kindly thank the thousands of volunteers who helped us to Hochachka, W.M., Fink, D., Hutchinson, R.A., Sheldon, D., Wong, W.-K., Kelling, S., 2012.
collect the data analyzed in this study. We specifically thank Yueh-mei Data-intensive science applied to broad-scale citizen science. Trends Ecol. Evol. 27,
Liu and the volunteers of the SOW and the other ENGOs from the 130–137.
Hong, S.-Y., Lin, H.-S., Walther, B.A., Shie, J.-E., Sun, Y.-H., 2018. Recent avian poi-
Taiwan Ocean Cleanup Alliance (TOCA) for collecting data. We thank
sonings suggest a secondary poisoning crisis of black kites during the 1980s in
Bodo Kretzschmar for information and Yu-Wen Emily Dai for transla- Taiwan. J. Raptor Res. 52, 326–337.
tions. We thank three reviewers for helpful and insightful comments. Jambeck, J.R., Geyer, R., Wilcox, C., Siegler, T.R., Perryman, M., Andrady, A., Narayan,
R., Law, K.L., 2015. Plastic waste inputs from land into the ocean. Science 347,
Appendix A. Supplementary data Jang, Y.C., Hong, S., Lee, J., Lee, M.J., Shim, W.J., 2014. Estimation of lost tourism
revenue in Geoje Island from the 2011 marine debris pollution event in South Korea.
Supplementary data to this article can be found online at https:// Mar. Pollut. Bull. 81, 49–54.
Koelmans, A.A., Besseling, E., Foekema, E.M., 2014. Leaching of plastic additives to

B.A. Walther et al. Marine Pollution Bulletin 135 (2018) 862–872

marine organisms. Environ. Pollut. 187, 49–54. Sheavly, S.B., Register, K.M., 2007. Marine debris & plastics: environmental concerns,
Krebs, C.J., 1999. Ecological Methodology, 2nd ed. Benjamin/Cummings, Menlo Park, sources, impacts and solutions. J. Polym. Environ. 15, 301–305.
California, USA. Society of Wilderness, 2014. Reduce plastic and avoid the ‘age of plastics’. Society of
Kunz, A., Walther, B.A., Löwemark, L., Lee, Y.-C., 2016. Distribution and quantity of Wilderness, Taipei, Taiwanhttps://www.sow.org.tw/info/news/20141120/3229 (in
microplastic on sandy beaches along the northern coast of Taiwan. Mar. Pollut. Bull. Chinese).
111, 126–135. Society of Wilderness, 2016. 2016 marine debris data: local PET bottle recycling system is
Kuo, F.-J., Huang, H.-W., 2014. Strategy for mitigation of marine debris: analysis of effective as outlying islands have highest number of foreign-source pollution. Society
sources and composition of marine debris in northern Taiwan. Mar. Pollut. Bull. 83, of Wilderness, Taipei, Taiwanhttps://www.sow.org.tw/info/news/20161221/5141
70–78. (in Chinese).
Laglbauer, B.J.L., Franco-Santos, R.M., Andreu-Cazenave, M., Brunelli, L., Papadatou, M., Strong, M., 2017. Taiwan environmentalists clear 10,000 plastic bottles from Penghu
Palatinus, A., Grego, M., Deprez, T., 2014. Macrodebris and microplastics from beaches. Taiwan (16 December).
beaches in Slovenia. Mar. Pollut. Bull. 89, 356–366. Sullivan, B.L., Aycrigg, J.L., Barry, J.H., Bonney, R.E., Bruns, N., Cooper, C.B., Damoulas,
Lamb, K., 2018. Bali hopes to regain paradise island status with mass cleanup. The T., Dhondt, A.A., Dietterich, T., Farnsworth, A., Fink, D., Fitzpatrick, J.W., Fredericks,
Guardianhttps://www.theguardian.com/world/2018/feb/23/bali-hopes-to-regain- T., Gerbracht, J., Gomes, C., Hochachka, W.M., Iliff, M.J., Lagoze, C., La Sorte, F.A.,
paradise-island-status-with-mass-cleanup (23 February). Merrifield, M., Morris, W., Phillips, T.B., Reynolds, M., Rodewald, A.D., Rosenberg,
Lavers, J.L., Bond, A.L., 2017. Exceptional and rapid accumulation of anthropogenic K.V., Trautmann, N.M., Wiggins, A., Winkler, D.W., Wong, W.-K., Wood, C.L., Yu, J.,
debris on one of the world's most remote and pristine islands. Proc. Natl. Acad. Sci. U. Kelling, S., 2014. The eBird enterprise: an integrated approach to development and
S. A. https://doi.org/10.1073/pnas.1619818114. application of citizen science. Biol. Conserv. 169, 31–40.
Lavers, J.L., Oppel, S., Bond, A.L., 2016. Factors influencing the detection of beach plastic Taiwan Macroview Television Service Online, 2016. Society of Wilderness, scholar draw
debris. Mar. Environ. Res. 119, 245–251. attention to issue of marine debris. https://www.youtube.com/watch?v=
Leary, K., 2018. Taiwan has committed to banning plastic items by 2030. World MP0ONpmes_o (23 December (in Chinese)).
Economic Forumhttps://www.weforum.org/agenda/2018/02/taiwan-commits-to- Talsness, C.E., Andrade, A.J.M., Kuriyama, S.N., Taylor, J.A., vom Saal, F.S., 2009.
banning-plastic-items-by-2030 (26 February). Components of plastic: experimental studies in animals and relevance for human
Lee, J., Hong, S., Song, Y.K., Hong, S.H., Jang, Y.C., Jang, M., Heo, N.W., Han, G.M., Lee, health. Philos. Trans. R. Soc. Lond. B 364, 2079–2096.
M.J., Kang, D., Shim, W.J., 2013. Relationships among the abundances of plastic Thiel, M., Penna-Díaz, M.A., Luna-Jorquera, G., Sala, S., Sellanes, J., Stotz, W., 2014.
debris in different size classes on beaches in South Korea. Mar. Pollut. Bull. 77, Citizen scientists and marine research: volunteer participants, their contributions and
349–354. projection for the future. Oceanogr. Mar. Biol. Annu. Rev. 52, 257–314.
Lin, C.-n., 2018. EPA sets timetable to ban plastic use. Taipei Times. http://www. Thompson, R.C., Moore, C.J., vom Saal, F.S., Swan, S.H., 2009. Plastics, the environment
taipeitimes.com/News/front/archives/2018/02/14/2003687612 (14 February). and human health: current consensus and future trends. Philos. Trans. R. Soc. Lond. B
Liu, T.-K., Wang, M.-W., Chen, P., 2013. Influence of waste management policy on the 364, 2153–2166.
characteristics of beach litter in Kaohsiung, Taiwan. Mar. Pollut. Bull. 72, 99–106. Tong, E., 2017. 529 syringes: activists submit newest and largest batch of medical waste
Munari, C., Corbau, C., Simeoni, U., Mistri, M., 2016. Marine litter on Mediterranean found on beach to gov't. Hong Kong Free Presshttps://www.hongkongfp.com/2017/
shores: analysis of composition, spatial distribution and sources in north-western 06/30/529-syringes-activists-submit-newest-largest-batch-medical-waste-found-
Adriatic beaches. Waste Manag. 49, 483–490. beach-govt/ (30 June).
National Library, 2017. National food consumption database. http://tnfcds.cmu.edu.tw/ UNEP, 2005. Marine litter: an analytical overview. United Nations Environment
index.php?action=index. Programme, Nairobi, Kenyahttp://wedocs.unep.org/handle/20.500.11822/8348.
Nelms, S.E., Coombes, C., Foster, L.C., Galloway, T.S., Godley, B.J., Lindeque, P.K., Witt, van der Velde, T., Milton, D.A., Lawson, T., Wilcox, C., Lansdell, M., Davis, G., Perkins, G.,
M.J., 2017. Marine anthropogenic litter on British beaches: a 10-year nationwide Hardesty, B.D., 2017. Comparison of marine debris data collected by researchers and
assessment using citizen science data. Sci. Total Environ. 579, 1399–1409. citizen scientists: is citizen science data worth the effort? Biol. Conserv. 208,
Ocean Conservancy, 2016. 30th anniversary international coastal cleanup. Ocean 127–138.
Conservancy, Washington, D.C., USAhttp://www.oceanconservancy.org/our-work/ Vethaak, A.D., Leslie, H.A., 2016. Plastic debris is a human health issue. Environ. Sci.
marine-debris/2016-data-release/2016-data-release-1.pdf. Technol. 50, 6825–6826.
Ocean Conservancy, 2018a. Fighting for trash free seas. Ocean Conservancy, Washington, Walther, B., 2015. Nation engulfed by plastic tsunami. Taipei Timeshttp://www.
D.C., USAhttps://oceanconservancy.org/trash-free-seas/international-coastal- taipeitimes.com/News/editorials/archives/2015/01/09/2003608789 (9 January).
cleanup/cleanswell/. Walther, B.A., Moore, J.L., 2005. The definitions of bias, precision, and accuracy, and
Ocean Conservancy, 2018b. Fighting for trash free seas. Cleanup reports. Ocean their use in testing the performance of species richness estimators, with a literature
Conservancy, Washington, D.C., USAhttps://oceanconservancy.org/trash-free-seas/ review of estimator performance. Ecography 28, 815–829.
international-coastal-cleanup/annual-data-release/. Wikipedia, 2018a. Ocean conservancy. https://en.wikipedia.org/wiki/Ocean_
Opfer, S., Arthur, C., Lippiatt, P., 2012. NOAA marine debris shoreline survey field guide. Conservancy.
National Oceanic and Atmospheric Administration, Maryland, USAhttps:// Wikipedia, 2018b. Syringe tide. https://en.wikipedia.org/wiki/Syringe_Tide.
marinedebris.noaa.gov/sites/default/files/ShorelineFieldGuide2012.pdf. Wilcox, C., van Sebille, E., Hardesty, B.D., 2015. Threat of plastic pollution to seabirds is
PlasticsEurope, 2017. Plastics – the facts 2017: an analysis of European latest plastics global, pervasive, and increasing. Proc. Natl. Acad. Sci. U. S. A. 112, 11899–11904.
production, demand and waste data. PlasticsEurope, Wemmel, Belgiumhttp://www. Wilcox, C., Mallos, N.J., Leonard, G.H., Rodriguez, A., Hardesty, B.D., 2016. Using expert
plasticseurope.org/application/files/5715/1717/4180/Plastics_the_facts_2017_ elicitation to estimate the impacts of plastic pollution on marine wildlife. Mar. Policy
FINAL_for_website_one_page.pdf. 65, 107–114.
Rochman, C.M., Tahir, A., Williams, S.L., Baxa, D.V., Lam, R., Miller, J.T., Teh, F.-C., Williams, A.T., Gregory, M., Tudor, D.T., 2005. Marine debris – onshore, offshore, sea-
Werorilangi, S., Teh, S.J., 2015. Anthropogenic debris in seafood: plastic debris and floor litter. In: Schwartz, M.L. (Ed.), Encyclopedia of Coastal Science. Springer, The
fibers from textiles in fish and bivalves sold for human consumption. Sci. Rep. 5, Netherlands, pp. 623–628.
14340. Yang, C.-h., 2018. Forging alliances for plastic-free oceans. Taipei Timeshttp://www.
Schmuck, A.M., Lavers, J.L., Stuckenbrock, S., Sharp, P.B., Bond, A.L., 2017. Geophysical taipeitimes.com/News/editorials/archives/2018/03/02/2003688511 (2 March).
features influence the accumulation of beach debris on Caribbean islands. Mar. Yang, R.-Y., Wu, Y.-C., Hwung, H.-H., Liou, J.-Y., Shugan, I.V., 2010. Current counter-
Pollut. Bull. 121, 45–51. measure of beach erosion control and its application in Taiwan. Ocean Coast. Manag.
Seltenreich, N., 2015. New link in the food chain? Marine plastic pollution and seafood 53, 552–561.
safety. Environ. Health Perspect. 123, A34–A41.