lable at ScienceDirect

Environmental Pollution 245 (2019) 545e557

Contents lists avai

Environmental Pollution

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

Plastic Pirates sample litter at rivers in Germany e Riverside litter andlitter sources estimated by schoolchildren*

Tim Kiessling a, b, *, Katrin Knickmeier b, Katrin Kruse b, Dennis Brennecke b,Alice Nauendorf b, Martin Thiel a, c, d

a Facultad de Ciencias del Mar, Universidad Cat�olica del Norte, Coquimbo, Chileb Kieler Forschungswerkstatt, University of Kiel and Leibniz Institute for Science and Mathematics Education (IPN), Kiel, Germanyc Millennium Nucleus Ecology and Sustainable Management of Oceanic Islands (ESMOI), Coquimbo, Chiled Centro de Estudios Avanzados en Zonas �Aridas (CEAZA), Coquimbo, Chile

a r t i c l e i n f o

Article history:Received 8 May 2018Received in revised form6 November 2018Accepted 8 November 2018Available online 13 November 2018

Keywords:Plastic litterMacrolitterFreshwaterSource identificationCitizen science

* This paper has been recommended for acceptance* Corresponding author. Kieler Forschungswerkstatt

24118 Kiel, Germany.E-mail addresses: kiessling@ipn.uni-kiel.de (T. Kie

kiel.de (K. Knickmeier), thiel@ucn.cl (M. Thiel).

https://doi.org/10.1016/j.envpol.2018.11.0250269-7491/© 2018 Elsevier Ltd. All rights reserved.

a b s t r a c t

Rivers are an important source of marine anthropogenic litter, but the particular origins of riverine litteritself have not been well established. Here we used a citizen science approach where schoolchildrenexamined litter at riversides and identified possible sources at over 250 sampling spots along large andsmall rivers in Germany, during autumn 2016 and spring 2017. Litter densities have an overall median of0.14, interquartile range 0e0.57 items m�2 and an overall average (±standard deviation) of 0.54 ± 1.20litter items m�2. Litter quantities differed only little by sampling year. The principal litter types foundwere plastics and cigarette butts (31% and 20%, respectively), followed by glass, paper, and metal items,indicating recreational visitors as the principal litter source. At many sites (85%), accumulations of litter,consisting principally of cigarettes and food packaging, have been found. At almost all sampling sites(89%), litter potentially hazardous to human health has been observed, including broken glass, sharpmetal objects, used personal hygiene articles and items containing chemicals. In the search for littersources, the schoolchildren identified mainly people who use the rivers as recreational areas (in contrastto residents living in the vicinity, illegal dumping, or the river itself depositing litter from upstreamsources). These results indicate the urgent need for better education and policy measures in order toprotect riparian environments and reduce input of riverine litter to the marine environment.

© 2018 Elsevier Ltd. All rights reserved.

1. Introduction

Marine pollution by anthropogenic litter (especially plastics) hasreceived much scientific and public attention and its impacts arewell-documented (Thompson et al., 2009; Kühn et al., 2015;Newman et al., 2015). Although much litter originates from sea-based sources, like fishing and aquaculture, most marine litterhas land-based sources (GESAMP, 2010; Andrady, 2011). It wasestimated that up to 12.7 million tons of plastic waste entered themarine environment in 2010 from coastal sources alone, and thatthis number is growing (Jambeck et al., 2015). Additionally, rivers

by Prof. Dr. Klaus Kümmerer., Am Botanischen Garten 16i,

ssling), kknickmeier@uv.uni-

have been suggested as important transport routes for litter to themarine environment from populations living farther inland (e.g.Willoughby, 1986; Tudor and Williams, 2004; Shimizu et al., 2008;Laglbauer et al., 2014; Rangel-Buitrago et al., 2017). Two recentglobal studies estimate that up to 2.8 million tons of plastic wasteenter the ocean annually from rivers (Lebreton et al., 2017; Schmidtet al., 2017). According to those models, the most polluting riversare located in Asia and Africa, accounting for about 70% of globallitter input to the coastal and marine environment. However, howthe litter gets into the rivers in the first place is not well known.

Potential sources of litter pollution in rivers are manifold:sewage outlets from wastewater treatment plants are known tocontain microplastics (for example microbeads used in cosmeticproducts or clothing fibres; Browne et al., 2011; McCormick et al.,2014; Dris et al., 2015) but sewage waters can also liberate largersanitary items (Williams and Simmons, 1999; Morritt et al., 2014).Litter floating in the river or present at a riverside can also originatefrom recreational activity in the vicinity (Gasperi et al., 2014;

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557546

McCormick and Hoellein, 2016; Carpenter and Wolverton, 2017) orfrom areas of high urban activity (e.g. commercial districts;Armitage, 2007; Carson et al., 2013). Larger litter accumulations orhousehold items are often the result of people intentionallydepositing litter at riversides (Williams and Simmons, 1997a; 1999;Rech et al., 2015; McCormick and Hoellein, 2016). Littering ofhousehold items by residents without access to waste collectioninfrastructure along with domestic sewage release is anothersource of riverine litter (Franz and Freitas, 2011; Di and Wang,2017).

Most studies on river litter pollution focus on the river as apathway for buoyant (and therefore mainly plastic) debris, andultimately quantify the amount of items that are already on themove within a river (e.g. Moore et al., 2011; Gasperi et al., 2014;Lechner et al., 2014; Morritt et al., 2014; Mani et al., 2015).Anthropogenic litter on the shores or in the vicinity of rivers hasbeen studied less frequently (but see Williams and Simmons,1997b, 1999; Rech et al., 2014, 2015; McCormick and Hoellein,2016), although it might be one of the principal sources of litterflowing towards the sea. For example, heavy rainfall, subsequentflooding and strong wind (Moore et al., 2011; Carson et al., 2013;Veerasingam et al., 2016) can mobilize litter deposited on theriverside.

These events are often seasonal and the occurrence of litterappears to fluctuate throughout the year, for example coincidingwith the monsoon season in regions where many of the mostpolluting rivers are located (Lebreton et al., 2017). The season is alsolikely to influence the number of visitors to the river environmentas hypothesized by McCormick and Hoellein (2016) to explainhigher litter quantities during summer. Visitors can also (deliber-ately or unintentionally) displace litter items from riversides intothe river. It is therefore important to identify primary sources ofriverine litter pollution and to investigate the effect of seasonalchange on litter densities.

In recent years, the investigation of coastal and freshwater litterpollution has been supported by data originating from citizen sci-ence project, which allows data collection over large areas and timespans (e.g. Hoellein et al. 2015; Rech et al., 2015; Nelms et al., 2017;Hidalgo-Ruz et al., 2018), given appropriate data quality mecha-nisms (Zettler et al., 2017).

The present study used a citizen science approach to examineriverside litter pollution (i.e. litter items that are not yet or notanymore located within a river) in Germany. Large rivers as well asmany small streams extend throughout Germany, often convergingto large river systems (Fig. 1). With over 7000 km of navigablewaterways (WSV, 2017), and almost all major cities (including thepopulous Rhine-Main industrial complex) being located at or closeto streams, rivers and the riparian environment in Germany playalso an important ecological, economic, and recreational role. Thepresent study with its citizen science approach allowed us to (i)estimate the quantity of litter at rivers of various sizes and identifythe material composition of litter, (ii) evaluate whether larger ac-cumulations of litter or objects potentially hazardous to humanhealth are present at riversides, and (iii) determine the sources oflitter.

2. Materials and Methods

2.1. Citizen science approach

This study is part of an extensive citizen science project (“Plas-tikpiraten”, Plastic Pirates) that examined different aspects of litterpollution at rivers in Germany. The sampling methodology wasadapted from Rech et al. (2015) and developed by the Kieler For-schungswerkstatt in Germany (http://www.forschungs-werkstatt.

de/) and the Científicos de la Basura in Chile (http://www.cientificosdelabasura.cl/). Schools and other youth organizationsfrom Germany were invited to investigate litter contamination at ariverside of their choice. A project booklet with sampling in-structions for each participant (Supplement S1), and a workbookwith information about litter and the riverine and marine envi-ronment for each supervisor were available at no cost. One desig-nated supervisor, the teacher or leader of youth organization,served as our contact to organize shipping of material, obtainingdata, and answering questions in case of ambiguity. The intendedage of participants was 10e16 years, although younger and olderstudents participated.

2.2. Sampling dates, number of participants and study area

The sampling was conducted during boreal autumn 2016 (16thof September to 18th of November) and spring 2017 (4th of May to17th of July). In total more than 5500 students from over 340different schools and organizations from all 16 German adminis-trative regions participated. Sometimes multiple classes from oneschool participated, leading to a total of 408 project groups thatconducted the sampling or parts thereof. Participants were free tochoose any river, regardless of size or location. Smaller groups wereformed to investigate different aspects of litter pollution in theriverine environment. There was no obligation to investigate allaspects; each school and organization chose the activities accord-ing to their capacity and motivation (i.e. number of participants,available time, and interests of participants).

All sampling sites were grouped according to river system, i.e. amain river and all tributary streams and rivers, or (if the riversystem was small or investigated by few groups) according to thesea they finally flow into. In total nearly all major river systems andmany smaller rivers in Germany have been investigated (Fig. 1): theRhine, Weser and Elbe river systems draining into the North Sea,smaller rivers flowing into the Baltic Sea, and the Danube riversystem draining into the Black Sea. Other rivers flowing into theNorth Sea, which do not belong to any of the above-mentionedsystems, have also been grouped for analysis. The Rhine river sys-temwas sampled most frequently, followed by the Elbe, Weser andDanube system. About half of the groups went to rivers less than20m wide (an estimation of river width was submitted by theparticipants, and if necessary corrected based on measurementswith the ruler tool in Google Earth 7.31.4507).

2.3. Sampling of riverside litter

Up to three transects were established perpendicular to the rivercourse. Each transect consisted of three sampling stations, one ineach predefined zone: the river edge (0e5m distance to river,assumed to have regular contact with it), the river bank (5e15mdistance to river, irregular contact with water of river during floodevents), and the river crest (15m or more distance to river, not incontact with the river; Fig. 2A). At each station a circle with a radiusof 1.5m (~7.1m2) was established with a stick, string and pebbles.This method was based on the study by Rech et al. (2015) that usedcircles instead of quadrats because circles were easier to establishon complex substrata and/or sites with abundant vegetation cover.Participants sampled all litter within the circles and classified itaccording to type: paper, cigarettes, plastic, metal, glass, food left-overs, and other items. This classification followed categories pre-viously used in litter studies with schoolchildren (e.g. Rech et al.,2015; Hidalgo-Ruz et al., 2018). More detailed lists with morespecific categories of litter items were not used to not increase thecomplexity of this activity. A photo of the litter from each stationwas taken on a white background (Fig. 3A) and sent, together with

Fig. 1. Map of major rivers and sampling spots of the Plastic Pirates in 2016 and 2017 in Germany. The colour of the dots represent the different river systems (or seas the smallerrivers flow into). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557 547

the collected data, to the coordinating laboratory (Kieler For-schungswerkstatt). Afterwards the litter was disposed into litterbins.

2.4. Sampling of litter accumulations and identification ofdangerous materials

A roughly rectangle-shaped area was established with ameasuring tape at the riverside and coordinates of the cornerpoints were recorded with a GPS device or smartphone. The longerside was established parallel, the shorter side perpendicular to theriver course. The size of the rectangle varied according to theavailable space, but covered at least 1000m2, in which litter accu-mulations were surveyed. An “accumulation” was counted as suchif three or more litter itemswere located not more than 50 cm apartfrom each other (Fig. 2B). The participants had to count and takepictures of small (3e10 litter items, Fig. 3B), medium-sized (11e25items, Fig. 3C), and large litter accumulations (more than 25 items,Fig. 3D). The surface area a litter accumulation occupied was not

measured by participants as this had led to complications in aprevious study (Rech et al., 2015). In 2016 we had asked partici-pants to identify the main material the litter accumulations con-sisted of, additionally to the size. This frequently led to confusionandmany groups counted single litter objects as accumulations. Forthe sampling in 2017, only the size of a litter accumulation wasregistered.

Participants also identified any of the following items poten-tially dangerous to human health in the same area used to surveylitter accumulations: broken glass, sharp metal objects, used per-sonal hygiene articles, decomposing food leftovers (which couldattract disease-carrying animals or harm small children uponaccidental ingestion), and litter items containing chemicals (e.g.aerosol cans, batteries, paint containers). Potentially dangerouslitter items were not quantified but only absence or presence wasnoted. If litter was collected during this part of the investigation(this decision was left to the supervisor) it was disposed into litterbins.

Fig. 2. (A) Sampling method for riverside litter: three transects were established, spanning three different zones, according to the distance to the river. (B) Identification of litteraccumulations according to the numbers of litter items encountered at one spot: at least three items had to be located close to each other to count as a litter accumulation. All itemsless than 50 cm apart were part of the same accumulation. If items were more than 50 cm apart they counted as a new accumulation (or as no accumulation, if less than three itemswere present). Images can be reused under Creative Commons license Attribution 4.0 International (CC-BY 4.0).

Fig. 3. (A) Some of the riverside litter found by students from the school Reichswald-Gymnasium, having identified a total of 383 litter items at their sampling site at the Mohrbach(part of the Rhine river system). The majority of litter encountered by this group were broken pieces of glass. (B) A small litter accumulation consisting of cigarette butts, found bystudents from the Realschule Maria Stern at the Eger-W€ornitz river, belonging to the Danube river system. (C) A medium-sized litter accumulation encountered by students from theGymnasium Fabritianum at the Rhine, consisting of plastic cutlery and wet wipes. (D) A large litter accumulation representing leftovers from a barbecue (fireplace, plastic bottles,plates and bags, amongst other litter). Also found by students from the Gymnasium Fabritianum at the Rhine. Photos can be reused under Creative Commons license Attribution 4.0International (CC-BY 4.0).

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557548

2.5. Inference of litter sources

Participants explored the surroundings of their sampling site,and, using several criteria (the use-type of encountered items, thesize of litter objects or accumulations, and the location of litter atthe riverside), they inferred which of the following sources are

likely contributing to local litter pollution: residents, visitors(people using the river environment as a recreational area), peopledumping litter illegally, industry, ship/boat traffic (including smallleisure crafts as well as large river barges), and the river itself(washing up litter from upstream sources). Each of the sources hadto be ranked on a five-point-scale: possible responses to the

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557 549

research question “Is the respective source contributing to locallitter pollution?” were “Yes”, “Likely”, “Possibly”, “Unlikely”, and“No”. For analysis the first two responses were merged to representa positive response to the question and the latter two responseswere merged to represent a negative response (the answer“Possibly” was not taken into account).

2.6. Stepwise verification of submitted citizen science data

To verify data it was first checked whether vital informationabout the sampling was available (e.g. date and place). Then, datafor each individual group were assessed in multiple steps andaccepted if criteria were passed (Fig. 4, Supplement S2-1). For

Fig. 4. Data verification flowchart for riverside litter, litter accumulations, the identification o(originating from one sampling circle) was analysed individually (located within dotted boxalso Supplement S2 for further information.

riverside litter and litter accumulations, the photos of the respec-tive litter findings were compared to the data submitted by par-ticipants. Only data from groups that provided at least onephotograph were accepted (unless no litter was found). If no photowas provided, or if the litter was misidentified or misclassified, thedataset was rejected (see Supplement S2-2 and S2-3 for a detaileddescription of the stepwise verification process and examplephotos used to evaluate data). To verify the size of the samplingarea established by participants surveying litter accumulations, weused the submitted GPS coordinates and (with the polygon tool inGoogle Earth Pro 7.31.4507) compared the value to the areameasured by the groups with the measuring tape.

Data for the identification of dangerous materials and the

f dangerous materials, and inference of litter sources. For riverside litter each data point) and subsequently considered in the context of all data from the respective group. See

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557550

inference of litter sources were accepted if submitted data sheetswere legible, complete, and unambiguous (Supplement S2-4, S2-5).To support the evaluation of litter sources by the participants, wereviewed photos of litter accumulations and assigned likely sourcesto each accumulation based mainly on the use-type of items withinan accumulation (Supplement S2-5).

2.7. Statistical analyses

For the analysis of riverside litter each sampling circle wastreated as one data point (each circle had the same area), so thatone group could contribute up to nine data points. A zero-alterednegative binomial (hurdle) model was developed (Zeileis et al.,2008; Zuur et al., 2009), testing which variables (river system,sampling year, riverside zone, or any interaction thereof) wouldexplain the differences in litter quantities encountered at thesampling sites. This is a two-step model comparing the amount ofzeroes independently from the positive values, because it isassumed that zero values can have two different causes: there maybe no litter because a litter source is absent (e.g. riverside is notaccessible or unattractive for recreational visitors or the river is toosmall for ship/boat traffic). On the other hand, no litter may befound because a particular source that was present did not causeany litter at the sampling site.

Only integer values are considered by the model, therefore thetotal litter count per circle was used for this analysis. Datasets fromrivers flowing into the North Seawere excluded (n¼ 3), as only datafrom 2016 were available, leading to missing interactions whichmade the model incomputable. Different models were created withcombinations (and interactions) of the variables “river system”

(Rhine, Weser, Elbe, Baltic Sea, Danube), “sampling year” (2016,2017), and “riverside zone” (river edge, river bank, river crest). Themodel with the lowest AIC value was accepted. Subsequently,pairwise post-hoc tests (with Tukey-HSD p-value adjustment) wereconducted for each variable and significant interaction.

Litter densities at rivers of different sizes (i.e. widths of rivers)were compared using a Kruskal-Wallis rank sum test, as variancesbetween the size categories were not homogeneous (Levene's test,df¼ 5, F-value¼ 2.94, p-value¼ 0.01). Dunn's post-hoc test (withHolm-�Sid�ak p-value adjustment) was conducted for pairwisecomparisons when significant differences were found.

For litter accumulations a one-way ANOVA was conducted foreach size class (small, medium, and large) and for the total numberof accumulations (variances were homogeneous across river sys-tems, Supplement S3-1).

For the analysis of dangerous materials, and the inference oflitter sources, Fisher's exact test of independence (two tailed,10,000 repetitions) was used to compare whether the response(presence or absence for dangerousmaterials, and likely or unlikelylitter source) was distributed differently across dangerous littertypes, and litter sources, respectively (McDonald, 2014). If a sig-nificant difference was discovered it was followed up with aBonferroni-corrected pairwise comparison.

The p-value of all tests was set at 0.05. The following packagesfrom R (version 3.4.1; R Development Core Team, 2017) were usedfor analyses: pscl 1.5.2 (Zeileis et al., 2008; Jackman, 2017), lmtest0.9e35 (Zeileis and Hothorn, 2002), lsmeans 2.27e61 (Lenth, 2016),car 2.1e6 (Fox and Weisberg, 2011), and dunn.test 1.3.5 (Dinno,2017).

3. Results

3.1. Data verification

After stepwise verification of the submitted data, 62 out of 408

project groups (15%) had to be excluded because vital informationabout the sampling was missing.

360 groups conducted the sampling of riverside litter and datafrom 179 groups (50%) were accepted and used for analysis (forexample 80 groups did not submit any photograph and their datawere therefore rejected, see Supplement S2-2 for details). Litteraccumulations were surveyed by a total of 355 groups and datafrom 66 groups (20%) were accepted. 158 groups conducted thesampling in 2016 for which data were rejected (see Materials andMethods) andmany other groups did not corroborate their findingswith photographs (Supplement S2-3). For the 46 groups that sub-mitted appropriate GPS measurements of their sampled area anaverage deviation of 6100m2 to their measured area was found.Nevertheless, we chose to consider all 66 datasets, as errors aremore likely related to faulty measurement with GPS devices orsmartphone applications (instruction on how to record coordinateswere not explicit in the workbook) than to actual errors usingmeasuring tape during sampling.

For the identification of dangerous materials datasets from 387groups were received and data from 320 groups (83%) accepted(Supplement S2-4). 314 groups submitted data for the inference oflitter sources and data from 261 groups (83%) were accepted (seeSupplement S2-5 for details).

3.2. Riverside litter

A total of 5955 litter items were quantified at riversides for anoverall median of 0.14, interquartile range (IQR) 0e0.57 items m�2

and an overall average (±standard deviation) of 0.54 ± 1.20 litteritems m�2. Litter densities ranged from 0 items per circle to amaximum of 174 items found in a circle at the river Mohrbach. Themajority of groups (91%) identified at least one litter item at theirsampling site, and almost two thirds of all sampling circles con-tained litter (Table 1). The best-fitting model considered the riversystem and sampling year (but not the riverside zone) as signifi-cant to predict encountered litter densities (Table 2). The per-centage of sampling circles with litter (Fig. 5A) as well as the countof litter items differed significantly between river system andsampling year (Fig. 5B, see Supplement S3-2 for posthoc tests) butthis pattern was inconsistent: in some river systems the densitywas higher in 2016 (e.g. the Rhine), while in others higher den-sities were observed in 2017 (e.g. the Elbe and Danube). Riversflowing into the Baltic Sea consistently had low densities. Litterquantities and percentage of sampling circles containing litter atother river systems were in between (Fig. 5, Table 1). The modelalso showed a significant difference in litter quantities per sam-pling year (Tukey-adjusted posthoc test, df¼ 1516, t-ratio¼�3.10,p-value < 0.01), yet the difference is very small: the median forboth years was the same at 0.14, IQR 0e0.57 items m�2. Litterdensities at the riverside further differed according to the size ofthe river (Kruskal-Wallis rank sum test, df¼ 5, c2¼ 67.90,p < 0.01): smaller rivers (up to 10m wide) had significantly lesslitter than wider rivers (median small rivers: 0.07, IQR: 0e0.39items m�2; median wide rivers: 0.28, IQR: 0e0.71 items m�2; seeSupplement S3-3 for all size classes).

Riverside litter consisted mainly of plastics (30.5% of all litterfound), followed by cigarette butts (20%), glass (16%), paper (13%),and metal (11.5%). Food leftovers (2%) and other items (7%)accounted for a fraction of the litter found (Fig. 6). Litter compo-sition differed between river systems, years, and rivers of differentsizes (e.g. more plastics were found at the Rhine, compared to theElbe and Danube river system, Kruskal-Wallis rank sum test, df¼ 5,c2¼ 36.53, p< 0.01), but differences were often relatively small andnot consistent.

Table 1Overview of riverside litter, litter accumulations and dangerous materials for each river system and sampling year. Only accepted datasets are represented.

Percentage of samplingsites with litter findings(number of surveyedsites)

Percentage of sampling circlescontaining litter (number ofsurveyed circles)

Median litter density, withinterquartile range per m2 insampling circles (surveyed areain m2)

Percentage of sampling sites with litteraccumulations (number of surveyedsites)

Average number of litteraccumulations of any size per1000m2 ± standard deviation(surveyed area in m2)

Percentage of sampling siteswith findings of dangerousmaterials (number of surveyedsites)

All riversystems,both years

91% (179) 61% (1564) 0.14, 0e0.57 (11,056) na na 89% (320)

All riversystems2016

90% (95) 60% (832) 0.14, 0e0.57 (5881) na na 87% (160)

All riversystems2017

93% (84) 63% (732) 0.14, 0e0.57 (5175) 85% (66) 1.52± 2.03 (418,875) 91% (160)

Rhine 2016 98% (48) 69% (426) 0.28, 0e0.85 (3011) na na 87% (75)Rhine 2017 93% (43) 60% (376) 0.14, 0e0.57 (2658) 90% (30) 1.69± 2.42 (185,420) 91% (75)Weser 2016 91% (11) 51% (87) 0.14, 0e0.42 (615) na na 88% (24)Weser 2017 92% (12) 58% (98) 0.14, 0e0.57 (693) 60% (5) 1.53± 1.88 (22,200) 89% (19)Elbe 2016 77% (17) 50% (150) 0.07, 0e0.42 (1060) na na 83% (30)Elbe 2017 92% (13) 62% (117) 0.14, 0e0.71 (827) 77% (13) 1.05± 0.92 (112,890) 94% (31)North Sea,

other2016

100% (3) 74% (27) 0.28, 0.7e0.85 (191) na na 75% (4)

North Sea,other2017

na na na na na 100% (2)

Baltic Sea2016

67% (9) 41% (81) 0.00, 0e0.28 (573) na na 87% (15)

Baltic Sea2017

100% (5) 71% (42) 0.14, 0e0.42 (297) 80% (5) 0.75± 1.43 (33,375) 89% (9)

Danube 2016 86% (7) 59% (61) 0.14, 0e0.28 (431) na na 100% (12)Danube 2017 91% (11) 73% (99) 0.28, 0e0.92 (700) 92% (13) 2.28± 1.95 (64,990) 88% (24)

T.Kiessling

etal./

Environmental

Pollution245

(2019)545

e557

551

Table 2Results of the zero-altered negative binomial (hurdle) model selection process. A *marks interactions between variables. The results of the likelihood ratio test refer tocomparison with the previous model in the table. The selected model with thelowest AIC value is marked in bold.

Model df AIC Likelihood ratio test

River system * Year * Zone 61 7069River system * Year þ Zone 25 7045 Х 2 ¼ 48.11 (df ¼ -36, p ¼ 0.085)River system * Year 21 7040 Х2¼ 2.68 (df¼ -4, p¼ 0.613)River system þ Year 13 7061 Х 2 ¼ 36.68 (df ¼ -8, p < 0.001)River system 11 7061 Х 2 ¼ 4.43 (df ¼ -2, p ¼ 0.109)

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557552

3.3. Litter accumulations

Most datasets considered for analysis of litter accumulationsmention the occurrence of small accumulations (83%). Medium-sized or even large litter accumulations had been seen lessfrequently (by 43% and 34% of the groups, respectively). In total 638

Fig. 5. Representation of data considered by the two-step hurdle model to analyse riversidprobability to encounter litter in a sampling circle is analysed (zero versus non-zero values)compared across variables. Whiskers of the boxplot represent the 1.5 interquartile range.number of outliers exceeding the scale are indicated by the number next to arrows at the tconsidered for respective part of the model.

litter accumulations were identified in an area of 418,875m2,averaging 1.5 ± 2.0 accumulations * 1000 m�2. More litter accu-mulations were spotted at riversides of the Danube river system(2.3 ± 2.0 accumulations * 1000 m�2), followed by the Rhine,Weser, and Elbe system (Table 1). Least accumulations were foundat rivers flowing into the Baltic Sea (0.8 ± 1.4 accumulations *1000 m�2). No datasets were available for litter accumulations atother rivers flowing into the North Sea. There were no significantdifferences between river systems considering the total amount oflitter accumulations or size classes of litter accumulations (seeSupplement S3-1 for ANOVA results).

For 277 photographs of litter accumulations the materialcomposition could be inferred by the coordinating laboratory.Plastic and cigarettes were the principal components in most ofthem (43% and 23%, respectively). Regarding the use-type of items,many of these accumulations contained items related to foodpackaging (44%), smoking (26%), preparing and consuming food(18%, indicated by e.g. plastic cutlery or barbecue grills), and con-sumption of alcohol (11%).

e litter in sampling circles for different river systems and sampling years: (A) first the, and subsequently, (B) if litter is found, the total amount of litter per sampling circle isOutliers are represented by dots (filled if multiple outliers have the same value). Theop of the boxplot. N ¼ Number of datasets considered, n¼ number of sampling circles

Fig. 6. Composition of riverside litter according to sampling year and river system. N¼ number of datasets considered, n¼ number of litter items found.

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557 553

3.4. Observation of dangerous materials

Nearly all groups (87% in 2016 and 91% in 2017, overall 89%)observed at least one dangerous litter item of any category at theirsampling site (Fig. 7A, Table 1). The type of dangerous materialsfound differed significantly (Fisher's test, 10,000 replicates, p-value< 0.01): broken glass was found significantly more often (at70% of the sites) than sharp metal objects and used personal hy-giene articles (at ~50% of all sites). Those latter items were alsosignificantly more frequent than decomposing food leftovers(found at 34% of sampling sites), which in turn were found signif-icantly more often than items classified as chemicals (present at21% of sampling sites; Table 3). There was little and non-significantvariation between river systems and sampling years in the per-centage of sampling sites where dangerous materials is present(Fisher's test, 10,000 replicates, p-value¼ 0.97, and Fisher's test,10,000 replicates, p-value¼ 0.38, respectively).

3.5. Inference of litter sources

Groups at almost all sampling sites (90% in 2016 and 84% in 2017,overall 87%) considered visitors who use the river as a leisure areaas a likely source for local litter pollution (Fig. 7B). This wassignificantly more often than the consideration of residents and theriver itself (depositing litter from upstream sources) as litter sour-ces (Fisher's test, 10,000 replicates, p-value< 0.01, Table 3), whichhave been identified by about a third of the groups (38% and 31%,respectively). People dumping litter illegally, ship/boat traffic andindustry have been identified less frequently as likely sources.These evaluations were consistent between years and with fewexceptions also for river systems: visitors have been identified in allriver systems and both years as the most likely source of litter, butthe order of other sources varied slightly (Supplement S3-4). Size ofthe river did not affect the frequency of mentioned sources, withthe exception of ship/boat traffic being mentioned more frequentlyas a likely source of litter for rivers wider than 100m (SupplementS3-4).

The revision of 277 photos of litter accumulations suggested

that the majority of accumulations could be attributed to visitors(88%). Fewer photos contained indication that litter accumulationsoriginated from residents (13%), were illegally deposited (11%), orcame from an industrial activity (4%), with other sources (ship/boattraffic, river itself) being of minor importance (Supplement S2-5).

4. Discussion

4.1. Citizen science approach and data collection

Many studies addressing litter pollution and its impacts in thecoastal environment have made use of data contributed by volun-teers (e.g. Gregory, 1991; Moore et al., 2009; van Franeker et al.,2011; Nelms et al., 2017; Hidalgo-Ruz et al., 2018). The citizen sci-ence approachmay have limitations (e.g. Dickinson et al., 2010), butwhen these are taken into account and adequate strategies applied(e.g. training of volunteers, simple instructions, and data verifica-tion mechanisms, Hidalgo-Ruz and Thiel, 2015), the quality of datacontributed by volunteers are able to match that of professionalscientists (Zettler et al., 2017). Citizen science is used in a variety offields, e.g. biodiversity, conservation, astronomy, meteorology, andso on (Dickinson et al., 2010), and does not only contribute scientificdata but can also increase the scientific literacy of participants andallow them to participate in decision-making processes byinforming policy makers (e.g. Thiel et al., 2014).

The present study incorporated a stepwise verification processof citizen science data. If requirements were not met, respectivedatasets were excluded from analyses to guarantee reliable andreplicable data. Here we applied the most stringent criteria,rejecting entire datasets when minor discrepancies between sub-mitted data and photos were detected (Supplement S2-2). Further,all datasets without photographic evidence were rejected, eventhough the data verification process (for datasets with photo-graphs) illustrated that relatively few data points were erroneous(about 6%). Therefore, most groups and participants were able tocorrectly identify, quantify and classify the litter encountered atriversides. A large problem were missing photographs or infor-mation, which led to rejection of datasets from over 100 groups.

Fig. 7. (A) Percentage of sampling sites with findings of respective dangerous materials in 2016 and 2017 at sampling sites across Germany (number of datasets for 2016¼ 160,2017¼160). (B) Percentage of sampling sites across all Germany in 2016 and 2017 where respective source has been identified as a likely origin of encountered litter (number ofdatasets for 2016¼134, 2017¼127).

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557554

Reasons for this were mainly logistical, including problems withuploading photos to the project website, and the time that passedbetween the reception of data and asking for missing informationby the coordinating laboratory due to the large number of partici-pating groups.

The identification of dangerous litter items in the present studyrelied on commonly known and well-identifiable objects. Similarlitter items are also classified by volunteers in widely employedbeach litter samplings (OSPAR Commission, 2010; OceanConservancy, 2018). For the inference of litter sources, the photosof litter accumulations were helpful to verify that most litter orig-inated from visitors. To further support the source identification itwould be helpful to investigate each litter item more closely, andassign possible sources via multivariate techniques or the identi-fication of indicator items (Tudor and Williams, 2004; Silva et al.,2008; Prevenios et al., 2018). A specific guide for participants ontaking pictures of litter would help implement these steps in citizenscience studies, where the collected litter items are not available tothe coordinating laboratory for posterior verification.

Litter accumulations in 2016 were frequently misquantified,most likely because participants were instructed to additionallyrecord the composition of litter accumulations. In 2017, aftersimplifying instructions, we rarely found this mistake repeated.Participants in a study by Rech et al. (2015) often misjudged thearea surveyed for litter accumulations. To avoid this we askedparticipants to actually measure the surveyed area with ameasuring tape. The additional use of GPS coordinates resulted in

many mismeasurements as participants were not properlyinstructed on how to track coordinates with a GPS device orsmartphone.

4.2. Litter quantities at rivers in Germany

Quantities of riverside litter differed between river systems: themore polluted riversides were located along the Rhine and Danubesystem, flowing through Germany's most, and second-mostpopulous region, respectively. For the coastal environment,Jambeck et al. (2015) showed a link between population densityand the amount of mismanaged plastic waste entering the sea. This,combinedwith a high per capitawaste production of over 2.1 kg perinhabitant per day in Germany (Hoornweg and Bhada-Tata, 2012),may lead to high litter burden at densely-populated sites in thepresent study. However, there are exceptions: several study sites ofthe less polluted Elbe and Weser system were located in populouscities, and therefore other factors should be considered as a pre-dictor of local litter pollution as well. Larger rivers (wider than10m) also showed higher litter densities than smaller streams. Thismay be due to better accessibility of larger rivers, larger areas thatare open for recreational use, or less possibilities to cross largerivers, which therefore may lead to aggregation of visitors at onewell accessible riverside. Larger rivers are also likely to attract morerecreational visitors.

Even though there is a significant difference in litter densitiesbetween sampling years, this difference is very small and

Table 3Results of Bonferroni-corrected pairwise comparisons for Fisher's exact test of in-dependence for dangerous materials and litter sources. The p-value of 0.05 has to bedivided by the numbers of comparisons conducted by the Bonferroni posthoc-test(McDonald, 2014), i.e. p-value of 0.005 for dangerous materials (p-value 0.05/10comparisons), and 0.003 for litter sources (p-value 0.05/15 comparisons). Significantcomparisons are marked in bold.

Comparison p-value

Dangerous materialsBroken glass e Sharp metal objects <0.005Broken glass e Used personal hygienic articles <0.005Broken glass e Decomposing food leftovers <0.005Broken glass e Chemicals <0.005Sharp metal objects e Used personal hygienic articles 0.155Sharp metal objects e Decomposing food leftovers <0.005Sharp metal objects e Chemicals <0.005Used personal hygienic articles e Decomposing food leftovers <0.005Used personal hygienic articles e Chemicals <0.005Decomposing food leftovers e Chemicals <0.005Litter sourcesVisitors e Residents <0.003Visitors e River depositing litter from upstream sources <0.003Visitors e Illegal dumping <0.003Visitors e Ship/boat traffic <0.003Visitors e Industry <0.003Residents e River depositing litter from upstream sources 1Residents e Illegal dumping <0.003Residents e Ship/boat traffic <0.003Residents e Industry <0.003River depositing litter from upstream sources e Illegal dumping 0.073River depositing litter from upstream sources e Ship/boat traffic <0.003River depositing litter from upstream sources e Industry <0.003Illegal dumping e Ship/boat traffic 0.662Illegal dumping e Industry <0.003Ship/boat traffic e Industry 0.427

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557 555

tendencies are opposing in the different river systems, so it cannotbe determined conclusively whether litter quantities are higherduring spring or autumn. A future study to investigate this wouldideally have participants sample the same site during differentseasons. In the present study the sampling effort was not balancedbetween seasons and sites due to logistical reasons (time-frame inwhich schools and organizations could participate and their abilityto access sampling sites).

The present study also showed no significant difference in litterdensities across shore zones. A higher density of litter in the upperzones of beaches has been observed in many coastal litter studies(see Hidalgo-Ruz et al., 2018 and references therein), and this hasbeen hypothesized to coincide with preferred spots used by beachvisitors. A possible explanation for the lack of a clear zonationpattern in the present study is a more heterogeneous terrain ofriver environments when compared to coastal beaches. Forexample, little space is available at riversides in cities (meaning thatlitter would be deposited within the first meters of the water'sedge), while at other sites large areas are available for public usethat may stretch beyond our designated sampling zone.

Considering the occurrence of litter accumulations anddangerous materials, overall there is no difference between riversystems and sampling year. The high probability of finding poten-tially dangerous materials in the present study illustrates that litternot only endangers wildlife but also people (for a discussion ofhazardous litter at beaches see Williams et al., 2013). A study onTasmanian beaches found that a fifth of visitors were affected bylitter at some point, mainly by lacerations caused by sharp items(Campbell et al., 2016). Sharp objects were also the most frequenthazardous item in the present study but other items could alsopresent a risk upon accidental ingestion by small children (e.g. foodleftovers or even cigarette butts, which have been mentioned as aconcern by do Sul and Costa, 2007). A quantification of these litter

items, and an evaluation of their presence in frequented areaswould facilitate a risk assessment of riverside litter to humanhealth.

On average, about two thirds of the litter encountered at theriverside has buoyant properties: paper and cigarette butts canfloat for a short time, and the latter can quickly contaminate largeamounts of water (Green et al., 2014). Most plastics are “persistentbuoyant” (Rech et al., 2014) and have a long floating lifetime. It isnot given that they reach the sea, though, as river infrastructure(e.g. dams or pillars of bridges), designated litter collection devices(such as litter booms or litter traps; Armitage, 2007; Carson et al.,2013; Gasperi et al., 2014), or vegetation at the riverside(Williams and Simmons, 1996, 1999) can trap floating litter. How-ever, smaller litter items or microplastics can more readily over-come these obstacles.

4.3. Sources of riverine litter pollution and litter mitigationstrategies

In the present study visitors to the river environment have beenidentified as the main source of litter. This is in contrast toWilliamsand Simmons (1996, 1999) and Rech et al. (2015) who investigatedriversides in Wales and Chile, respectively, and identified illegaldumping as a main source of litter. In Germany illegal dumping isprobably less likely to occur, because of a better waste infrastruc-ture, also taking care of bulky household waste (when compared topresent-day Chile and toWales two decades ago, judging from solidwaste recycling rates; Comisi�on Nacional del Medio Ambiente,2010; OECD, 2018), and possibly more effective law enforcementand prosecution. A higher chance of finding large patches of litterwas also associated with riversides that are accessible by car, andaesthetically less appealing (Williams and Simmons, 1999), aspectsthat would merit future investigations in Germany at sites wherelitter accumulations have been found.

Other sources of river litter that have been suggested in otherstudies, for example household litter by residents (Franz andFreitas, 2011; Di and Wang, 2017) or industry-related litter havebeen identified in few cases, but the latter source contributessignificantly to pollution by microplastics (Lechner et al., 2014,2015; Klein et al., 2015), which were not sampled in the presentstudy. Ship/boat traffic was not frequently identified as a likely littersource, though naturally more often so at larger rivers. Litter fromships/boat traffic is also not available for survey at riversides (unlesswashed up), and therefore this potential source is likelyunderestimated.

Visitors as a principal litter source have been identified inseveral river litter studies (e.g. Gasperi et al., 2014; Rech et al., 2015;McCormick and Hoellein, 2016; Carpenter and Wolverton, 2017),and some of those studies have further analysed characteristics ofindividual litter objects (e.g. their weight and expected motility orpurpose of use) to identify which group of visitors cause most litter.In the present study the analysis of photos of litter accumulationsindicated that consuming or preparing food and smoking are theprincipal activities causing litter from visitors to reach the riverenvironment. McCormick and Hoellein (2016) found highest litterdensities related to alcohol consumption at accessible but hiddenspaces, where vandalism was also common. Carpenter andWolverton (2017) identified visitors who consume food or discardlitter fromvehicles at a well accessible riverside with parking lot, inaddition to passing-by visitors and recreational fishers at a hikingpath (inferred from beverage bottles and fishing gear that havebeen left behind).

Litter mitigation strategies should be designed accordingly.Based on the results from the present study we recommend tofocus litter mitigation efforts on smokers and people consuming or

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557556

preparing food at riversides in Germany, although site-specificcharacteristics (access or proximity to populous areas) shouldalso be considered. In the coastal environment, community edu-cation but also prosecution and fees are accepted by many beachvisitors to discourage littering (Santos et al., 2005; Eastman et al.,2013). In the riverine environment, picking up litter in the pres-ence of visitors has also been shown to reduce litter quantities(Wagstaff and Wilson, 1998; Cingolani et al., 2016). Involvingschoolchildren in practical, short-lived activities (such as the pre-sent project) seems also important to create awareness and engagethem in litter-reducing activities (Hartley et al., 2015). Finally, agovernmental strategy to reduce the production of single-useplastic products (as is currently proposed in the European Unionfor certain items; European Commission, 2018) would ensure thatfewer products typically consumed by riverside visitors would bepackaged in persistent and environmentally harmful material(Rochman et al., 2013).

5. Conclusions and Outlook

Most studies addressing environmental litter pollution treatrivers primarily as a source or litter pathway, contributing to ma-rine pollution. Impacts of litter in the riverine environment,quantities of litter at the riverside (with the potential of becomingmarine debris), and sources of riverine litter have been rarelyinvestigated. The present study found that certain litter at manyriversides (e.g. sharp and toxic items) present a danger to humanhealth, and that most litter on German riversides is produced byrecreational visitors. Future studies should consider which char-acteristics of sampling sites influence litter densities and distribu-tion (e.g. distance to populous areas or accessibility of riverside),how litter quantities at riversides relate to litter located within ariver, and how riverine litter impacts wildlife and people, similar toinvestigations conducted at beaches and the coastline.

Acknowledgements

This research would not have been possible without theenthusiastic support of all the students, teachers and volunteercollaborators e their help is greatly appreciated (see SupplementS4)! We thank Iv�an Hinojosa, Marcelo Rivadeneira, and LennartSchreiber for guidance with the statistical analyses. Lukas Pott andMagdalena Gatta-Rosemary helped with entering data and lookingthrough photos. Florian Druckenthaner, Katharina Kummer, andDaniel Henkel from the Kommunikationsbüro Wissenschaftsjahr,and Doris Knoblauch and LindaMederake from the Ecologic Institutfacilitated the creation and shipping of educational material andsampling instructions, the creation of the project webpage, andcommunication with the contact person of each group. Severalother members of the Kieler Forschungswerkstatt, especiallyHenrike Bratz, helped to realize this project and provided logisticalsupport. We are grateful to two anonymous reviewers who pro-vided valuable feedback and helped to substantially improve thismanuscript. This project received funding and logistical supportfrom the German Federal Ministry of Education and Science (BMBF)as part of the ‘Science Year 2016*17 Seas and Oceans’, funding fromthe Lighthouse Foundation in Germany, and logistic support fromUniversidad Cat�olica del Norte (UCN), the Millennium NucleusEcology and Sustainable Management of Oceanic Islands (ESMOI),the Ecologic Institut, the Cluster of Excellence “Future Ocean” of theUniversity of Kiel, the Leibniz Institute for Science andMathematicsEducation (IPN), and the Ministry of Education and VocationalTraining of Schleswig-Holstein.

Appendix A. Supplementary data

Supplementary data to this article can be found online athttps://doi.org/10.1016/j.envpol.2018.11.025.

References

Andrady, A.L., 2011. Microplastics in the marine environment. Mar. Pollut. Bull. 62,1596e1605.

Armitage, N., 2007. The reduction of urban litter in the stormwater drains of SouthAfrica. Urban Water J. 4, 151e172.

Browne, M.A., Crump, P., Niven, S.J., Teuten, E., Tonkin, A., Galloway, T.,Thompson, R., 2011. Accumulation of microplastic on shorelines woldwide:sources and sinks. Environ. Sci. Technol. 45, 9175e9179.

Campbell, M.L., Slavin, C., Grage, A., Kinslow, A., 2016. Human health impacts fromlitter on beaches and associated perceptions: a case study of ‘clean’ Tasmanianbeaches. Ocean Coast Manag. 126, 22e30.

Carpenter, E., Wolverton, S., 2017. Plastic litter in streams: the behavioral archae-ology of a pervasive environmental problem. Appl. Geogr. 84, 93e101.

Carson, H.S., Lamson, M.R., Nakashima, D., Toloumu, D., Hafner, J., Maximenko, N.,McDermid, K.J., 2013. Tracking the sources and sinks of local marine debris inHawai‘i. Mar. Environ. Res. 84, 76e83.

Cingolani, A.M., Barber�a, I., Renison, D., Barri, F.R., 2016. Can persuasive anddemonstrative messages to visitors reduce littering in river beaches? WasteManag. 58, 34e40.

Comisi�on Nacional del Medio Ambiente, 2010. Primer Reporte Del Manejo DeResiduos Solidos En Chile, A~no 2010. Gobierno de Chile, Comisi�on Nacional delMedio Ambiente.

Di, M., Wang, J., 2017. Microplastics in surface waters and sediments of the threegorges reservoir, China. Sci. Total Environ. 616e617, 1620e1627.

Dickinson, J.L., Zuckerberg, B., Bonter, D.N., 2010. Citizen science as an ecologicalresearch tool: challenges and benefits. Annu. Rev. Ecol. Evol. Syst. 41, 149e172.

Dinno, A., 2017. Dunn.test: Dunn's test of multiple comparisons using rank sums. Rpackage version 1.3.5. https://CRAN.R-project.org/package¼dunn.test.(Accessed 26 October 2018).

do Sul, J.A.I., Costa, M.F., 2007. Marine debris review for Latin America and the widerCaribbean region: from the 1970s until now, and where do we go from here?Mar. Pollut. Bull. 54, 1087e1104.

Dris, R., Gasperi, J., Rocher, V., Saad, M., Renault, N., Tassin, B., 2015. Microplasticcontamination in an urban area: a case study in Greater Paris. Environ. Chem.12, 592e599.

Eastman, L.B., Nú~nez, P., Crettier, B., Thiel, M., 2013. Identification of self-reporteduser behavior, education level, and preferences to reduce littering on beachese a survey from the SE Pacific. Ocean Coast Manag. 78, 18e24.

European Commission, 2018. Single-use plastics: new EU rules to reduce marinelitter. Press release. http://europa.eu/rapid/press-release_IP-18-3927_en.htm.(Accessed 26 October 2018).

Fox, J., Weisberg, S., 2011. An R Companion to Applied Regression, second ed. Sage,Thousand Oaks, CA.

Franz, B., Freitas, M.A.V., 2011. Generation and impacts of floating litter on urbancanals and rivers: rio de Janeiro megacity case study. WIT Trans. Ecol. Environ.167, 321e332.

Gasperi, J., Dris, R., Bonin, T., Rocher, V., Tassin, B., 2014. Assessment of floatingplastic debris in surface water along the Seine River. Environ. Pollut. 195,163e166.

GESAMP, 2010. Proceedings of the GESAMP international workshop on plasticparticles as a vector in transporting persistent, bio-accumulating and toxicsubstances in the oceans. GESAMP Rep. Stud. 82, 68pp.

Green, A.L.R., Putschew, A., Nehls, T., 2014. Littered cigarette butts as a source ofnicotine in urban waters. J. Hydrol. 519, 3466e3474.

Gregory, M.R., 1991. The hazards of persistent marine pollution: drift plastics andconservation islands. J. Roy. Soc. N. Z. 21, 83e100.

Hartley, B.L., Thompson, R.C., Pahl, S., 2015. Marine litter education boosts children'sunderstanding and self-reported actions. Mar. Pollut. Bull. 90, 209e217.

Hidalgo-Ruz, V., Thiel, M., 2015. The contribution of citizen scientists to themonitoring of marine litter. In: Bergmann, M., Gutow, L., Klages, M. (Eds.),Marine Anthropogenic Litter. Springer, Berlin.

Hidalgo-Ruz, V., Honorato-Zimmer, D., Gatta-Rosemary, M., Nu~nez, P., Hinojosa, I.A.,Thiel, M., 2018. Spatio-temporal variation of anthropogenic marine debris onChilean beaches. Mar. Pollut. Bull. 126, 516e524.

Hoellein, T.J., Westhoven, M., Lyandres, O., Cross, J., 2015. Abundance and envi-ronmental drivers of anthropogenic litter on 5 Lake Michigan beaches: a studyfacilitated by citizen science data collection. J. Great Lake. Res. 41, 78e86.

Hoornweg, D., Bhada-Tata, P., 2012. What a Waste: a Global Review of Solid WasteManagement. Urban Development Series. The World Bank.

Jackman, S., 2017. Pscl: Classes and Methods for R Developed in the Political ScienceComputational Laboratory. United States Studies Centre, University of Sydney,Sydney, Australia. R package version 1.5.2. URL. https://github.com/atahk/pscl/.(Accessed 26 October 2018).

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, 768e771.

Klein, S., Worch, E., Knepper, T.P., 2015. Occurrence and spatial distribution of

T. Kiessling et al. / Environmental Pollution 245 (2019) 545e557 557

microplastics in river shore sediments of the Rhine-Main area in Germany.Environ. Sci. Technol. 49, 6070e6076.

Kühn, S., Bravo Rebolledo, E.L., van Franeker, J.A., 2015. Deleterious effects of litteron marine life. In: Bergmann, M., Gutow, L., Klages, M. (Eds.), Marine Anthro-pogenic Litter. Springer, Berlin.

Laglbauer, B.J., Franco-Santos, R.M., Andreu-Cazenave, M., Brunelli, L.,Papadatou, M., Palatinus, A., Grego, M., Deprez, T., 2014. Macrodebris andmicroplastics from beaches in Slovenia. Mar. Pollut. Bull. 89, 356e366.

Lebreton, L.C.M., van der Zwet, J., Damsteeg, J., Slat, B., Andrady, A., Reisser, J., 2017.River plastic emissions to the world's oceans. Nat. Commun. 8, 15611.

Lechner, A., Keckeis, H., Lumesberger-Loisl, F., Zens, B., Krusch, R., Tritthart, M.,Glas, M., Schludermann, E., 2014. The Danube so colourful: a potpourri of plasticlitter outnumbers fish larvae in Europe's second largest river. Environ. Pollut.188, 177e181.

Lechner, A., Ramler, D., 2015. The discharge of certain amounts of industrialmicroplastic from a production plant into the river Danube is permitted by theAustrian legislation. Environ. Pollut. 200, 159e160.

Lenth, R.V., 2016. Least-Squares means: the R package lsmeans. J. Stat. Software 69,1e33.

Mani, T., Hauk, A., Walter, U., Burkhardt-Holm, P., 2015. Microplastics profile alongthe Rhine river. Sci Rep-UK 5, 17988.

McCormick, A., Hoellein, T.J., Mason, S.A., Schluep, J., Kelly, J.J., 2014. Microplastic isan abundant and distinct microbial habitat in an urban river. Environ. Sci.Technol. 48, 11863e11871.

McCormick, A.R., Hoellein, T.J., 2016. Anthropogenic litter is abundant, diverse, andmobile in urban rivers: insights from cross ecosystem analyses using ecosystemand community ecology tools. Limnol. Oceanogr. 61, 1718e1734.

McDonald, J.H., 2014. Handbook of Biological Statistics, third ed. http://www.biostathandbook.com/fishers.html. (Accessed 26 October 2018).

Moore, E., Lyday, S., Roletto, J., Litle, K., Parrish, J.K., Nevins, H., Harvey, J.,Mortenson, J., Greig, D., Piazza, M., Hermance, A., Lee, D., Adams, D., Allen, S.,Kell, S., 2009. Entanglements of marine mammals and seabirds in central Cal-ifornia and the north-west coast of the United States 2001e2005. Mar. Pollut.Bull. 58, 1045e1051.

Moore, C.J., Lattin, G.L., Zellers, A.F., 2011. Quantity and type of plastic debris flowingfrom two urban rivers to coastal waters and beaches of Southern California.Journal of Integrated Coastal Zone Management 11, 65e73.

Morritt, D., Stefanoudis, P.V., Pearce, D., Crimmen, O.A., Clark, P.F., 2014. Plastic inthe Thames: a river runs through it. Mar. Pollut. Bull. 78, 196e200.

Nelms, S.E., Coombes, C., Foster, L.C., Galloway, T.S., Godley, B.J., Lindeque, P.K.,Witt, M.J., 2017. Marine anthropogenic litter on British beaches: a 10-yearnationwide assessment using citizen science data. Sci. Total Environ. 579,1399e1409.

Newman, S., Watkins, E., Farmer, A., ten Brink, P., Schweitzer, J.P., 2015. The eco-nomics of marine litter. In: Bergmann, M., Gutow, L., Klages, M. (Eds.), MarineAnthropogenic Litter. Springer, Berlin.

Ocean Conservancy, 2018. International coastal cleanup data card. http://act.oceanconservancy.org/site/DocServer/ICC_Eng_DataCardFINAL.pdf?docID¼4221. (Accessed 26 October 2018).

OECD, 2018. OECD.Stats municipal waste, generation and treatment, % recycling.https://stats.oecd.org/Index.aspx?DataSetCode¼MUNW. (Accessed 26 October2018).

OSPAR Commission, 2010. Guideline for Monitoring Marine Litter on the Beaches inthe OSPAR Maritime Area, first ed. https://www.ospar.org/ospar-data/10-02e_beachlitter%20guideline_english%20only.pdf. (Accessed 26 October 2018).

Prevenios, M., Zeri, C., Tsangaris, C., Liubartseva, S., Fakiris, E., Papatheodorou, G.,2018. Beach litter dynamics on Mediterranean coasts: distinguishing sourcesand pathways. Mar. Pollut. Bull. 129, 448e457.

R Development Core Team, 2017. R: a Language and Environment for StatisticalComputing. R Foundation for Statistical Computing, Vienna, Austria, 3-900051-07-0. http://www.R-project.org/. (Accessed 26 October 2018).

Rangel-Buitrago, N., Williams, A., Anfuso, G., Arias, M., Gracia, A., 2017. Magnitudes,sources, and management of beach litter along the Atlantico departmentcoastline, Caribbean coast of Colombia. Ocean Coast Manag. 138, 142e157.

Rech, S., Macaya-Caquilp�an, V., Pantoja, J.F., Rivadeneira, M.M., Jofre Madariaga, D.,

Thiel, M., 2014. Rivers as sources of marine litter e a study from the SE Pacific.Mar. Pollut. Bull. 82, 66e75.

Rech, S., Macaya-Caquilp�an, V., Pantoja, J., Rivadeneira, M.M., KroegerCampod�onico, C., Thiel, M., 2015. Sampling of riverine litter with citizen sci-entists e findings and recommendations. Environ. Monit. Assess. 187, 335.

Rochman, C.M., Browne, M.A., Halpern, B.S., Hentschel, B.T., Hoh, E.,Karapanagioti, H.K., Rios-Mendoza, L.M., Takada, H., Teh, S., Thompson, R.C.,2013. Policy: classify plastic waste as hazardous. Nature 494, 169e171.

Santos, I.R., Friedrich, A.C., Wallner-Kersanach, M., Fillmann, G., 2005. Influence ofsocio-economic characteristics of beach users on litter generation. Ocean CoastManag. 48, 742e752.

Schmidt, C., Krauth, T., Wagner, S., 2017. Export of plastic debris by rivers into thesea. Environ. Sci. Technol. 51, 12246e12253.

Shimizu, T., Nakai, J., Nakajima, K., Kozai, N., Takahashi, G., Matsumoto, M., Kikui, J.,2008. Seasonal variations in coastal debris on Awaji Island, Japan. Mar. Pollut.Bull. 57, 182e186.

Silva, J.S., Barbosa, S.C., Costa, M.F., 2008. Flag items as a tool for monitoring solidwastes from users on urban beaches. J. Coast Res. 24, 890e898.

Thiel, M., Penna-Díaz, M.A., Luna-Jorquera, G., Salas, S., Sellanes, J., Stotz, W., 2014.Citizen scientists and marine research: volunteer participants, their contribu-tions, and projection for the future. Oceanogr. Mar. Biol. Annu. Rev. 52,257e314.

Thompson, R.C., Moore, C.J., vom Saal, F.S., Swan, S.H., 2009. Plastics, the environ-ment and human health: current consensus and future trends. Philos T Roy SocB 364, 2153e2166.

Tudor, D.T., Williams, A., 2004. Development of a ‘matrix scoring technique’todetermine litter sources at a bristol channel beach. J. Coast Conserv. 10, 119e127.

van Franeker, J.A., Blaize, C., Danielsen, J., Fairclough, K., Gollan, J., Guse, N.,Hansen, P., Heubeck, M., Jensen, J., le Guillou, G., Olsen, B., Olsen, K., Pedersen, J.,Stienen, E.W.M., Turner, D.M., 2011. Monitoring plastic ingestion by thenorthern fulmar Fulmarus glacialis in the North Sea. Environ. Pollut. 159,2609e2615.

Veerasingam, S., Mugilarasan, M., Venkatachalapathy, R., Vethamony, P., 2016. In-fluence of 2015 flood on the distribution and occurrence of microplastic pelletsalong the Chennai coast, India. Mar. Pollut. Bull. 109, 196e204.

Wagstaff, M.C., Wilson, B.E., 1988. The evaluation of litter behavior modification in aRiver environment. J. Environ. Educ. 20, 39e44.

Williams, A.T., Simmons, S.L., 1996. The degradation of plastic litter in rivers: im-plications for beaches. J. Coast Conserv. 2, 63e77.

Williams, A.T., Simmons, S.L., 1997a. Estuarine litter at the river/beach interface inthe Bristol Channel, United Kingdom. J. Coast Res. 13, 1159e1165.

Williams, A.T., Simmons, S.L., 1997b. Movement patterns of riverine litter. Water,Air, Soil Pollut. 98, 119e139.

Williams, A.T., Simmons, S.L., 1999. Sources of riverine litter: the river taff, southWales, UK. Water, Air, Soil Pollut. 112, 197e216.

Williams, A.T., Pond, K., Ergin, A., Cullis, M.J., 2013. The hazards of beach litter. In:Finkl, C.W. (Ed.), Coastal Hazards. Springer Science and Business Media.

Willoughby, N.G., 1986. Man-made litter on the shores of the thousand island ar-chipelago, Java. Mar. Pollut. Bull. 17, 224e228.

WSV (Wasserstraßen- und Schifffahrtsverwaltung des Bundes), 2017. L€angen derbundeswasserstraßen. https://www.wsv.de/wasserstrassen/gliederung_bundeswasserstrassen/Laengen_der_Bundeswasserstrassen/index.html.(Accessed 26 October 2018).

Zeileis, A., Hothorn, T., 2002. Diagnostic checking in regression relationships.R. News 2, 7e10. https://CRAN.R-project.org/doc/Rnews/. (Accessed 26 October2018).

Zeileis, A., Kleiber, C., Jackman, S., 2008. Regression models for count data in R.J. Stat. Software 27, 1e25.

Zettler, E.R., Takada, H., Monteleone, B., Mallos, N., Eriksen, M., Amaral-Zettler, L.A.,2017. Incorporating citizen science to study plastics in the environment. AnalMethods-UK 9, 1392e1403.

Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A., Smith, G.M., 2009. Mixed effectsmodels and extensions in ecology with R. In: Gail, M., Krickeberg, K.,Samet, J.M., Tsiatis, A., Wong, W. (Eds.), Statistics for Biology and Health.Springer Science and Business Media.