Beach litter sourcing in the Bristol channel and Wales, U.K

BEACH LITTER SOURCING IN THE BRISTOL CHANNEL ANDWALES, U.K.

A. T. WILLIAMS1, D. T. TUDOR2 and P. RANDERSON3∗1 School of Applied Sciences, University of Glamorgan, Pontypridd, Wales, U.K.; 2 Air Terra Water

Ltd, Beach Road, Porthcawl, Wales, U.K.; 3 School of Biosciences, Cardiff University, Cardiff,Wales, U.K.

(∗ author for correspondence, e-mail: [email protected], Fax: 443 482285)

(Received 23 October 2001; accepted 15 July 2002)

Abstract. Principal component analysis (PCA) and cluster analysis were carried out on beach litterfound primarily along the Bristol Channel, U.K., together with beaches located around the Princip-ality of Wales. Both techniques indicated three beach survey sites as outliers, with distinctive beachlitter profiles. PCA distinguished between riverine, sewage related, fishing and shipping sourceditems, but did not produce a coherent grouping for beach user litter. This was attributed to a differencein modes of transport. Beaches from mid and north Wales, which had small litter amounts, werenot differentiated from the central grouping. The western edge of the Channel was more heavilyinfluenced by shipping and fishing inputs than was the eastern portion. Cluster analysis showed adiscrete group of beaches located mainly on the southern side of the Bristol Channel.

Keywords: beach litter, cluster analysis, litter sourcing, principal component analysis, Wales

1. Introduction

Determining the source of litter found on beaches is often proclaimed as the prin-cipal aim of many monitoring and survey programmes. However, the effectivenessof such schemes to attribute litter accurately to a source is in some doubt. At presentthere is no explicit or widely used methodology that facilitates the sourcing ofbeach litter. Research studies, e.g. on small beach areas Gabrielides et al. (1991)and Thornton and Jackson (1998), have assigned sources of beach litter for a par-ticular location, but often these are based on assumptions or educated estimationsthrough local knowledge. Whilst such knowledge of the locality is very importantin order to assist sourcing, the methods used in the attribution process are oftenunclear and do not seem to be systematic or theory based. There appears to be atheoretical vacuum with respect to litter sourcing. To have any realistic hope ofpreventing, or at least abating, the beach litter problem it is essential to ascertainits source and to establish a robust methodology to facilitate this.

Most litter surveys conducted on beaches simply enumerate and categorise litteraccording to material composition, i.e. plastic, metal, glass etc. (e.g. Corbin andSingh, 1993; Frost and Cullen, 1997). Such a breakdown is useful in establishingthe effectiveness of legislation such as MARPOL Annex V (1973–1978), but this

Water, Air, and Soil Pollution 143: 387–408, 2003.© 2003 Kluwer Academic Publishers. Printed in the Netherlands.

388 A. T. WILLIAMS ET AL.

approach provides no information regarding potential sources. Although it is avery difficult and often imprecise task, sourcing, along with education, is perhapsthe prime weapon in the fight against this type of pollution. If a source can beestablished, then those perpetrating the pollution can be targeted and hopefullymeasures taken to address and subsequently prevent the problem. Beach managersand port/harbour authorities can use information gained regarding the sources ofbeach litter to formulate plans and actions with regard to prevention measures. Thisis perhaps an idealistic aim, in that there are certain beach locations that possesssuch a mix of litter with several potential sources that any specific attributions maybe extremely difficult.

Beach litter can be categorised into two broad source groups, which can befurther split to enable a more useful and accurate attribution. Sea-based sources oflitter include all types of sea-going vessel as well as offshore installations (Earll etal., 1999). Land-based sources incorporate litter left by beach users (Golik andGertner, 1992), litter entering the sea via rivers or municipal drainage systems(Williams et al., 2000), and litter directly deposited at or near the beach (Nash,1992). A third broad category can also be considered, namely truly pelagic litter.This litter will have spent lengthy periods afloat with distant sources, whether thesewere land or sea based (Gregory, 1998). The problem with considering this finalcategory is that the litter would be difficult to distinguish from litter emanatingfrom the other two prime sources, especially as originally such items would havederived from one or other of these.

2. Linking Items to Source

Occasionally, litter can be very directly linked to a specific source, on other occa-sions the sources can be numerous, with little indication or information availableon litter items to allow easy attribution. Items such as water or soft drink bottlesrarely contain any labeling when they are found on a beach: these items are usedby beach visitors, those at sea, and can also arrive at a beach via rivers. Thereare sometimes clues to source on the surface of debris items; for example, marinegrowths or oiling could point towards sea input, although items that have beenat sea a long time, but originated from a land source, could also be tainted withthese. Indeed, colonisation can be very quick and plastic substrates need not havedrifted very far before they become heavily coated with bryozoans and barnacles(M. Gregory, pers. comm.).

Containers carry a wealth of information on their surface or on labels. Anylabeling present, fully intact containers, and recent sell by dates imprinted on thecontainer will help identify beach user discards. However a caveat must be re-cognised, burial or some time spent at sea may mask the original source of theseitems, and may lead some to consider these as being from a sea borne source.Other labeling, such as milk containers with specific local markings, help to pin-

BEACH LITTER SOURCING IN THE U.K. 389

point sources, plastic shopping bags may also have local addresses, which aid theattribution process. Foreign items are widely accepted as indicating a shippingsource; this is because items either come from foreign vessels, or U.K. vessels,which have purchased foreign goods abroad. A number of foreign containers (e.g.milk, detergents) have been found on Pembrokeshire beaches, illustrating the useof the area by Spanish fishing vessels.

3. Attribution Process

Attributing a source to litter found on beaches is very complex. Trying to establish asource from an amalgam of debris is not an easy task and it is important to considerseveral factors and often to make informed assumptions.

• IdentificationThis is probably the most crucial factor. Without correct and robust identifica-tion of a litter item no link to source can be made. There is evidence of peoplemis-identifying items of litter, particularly those that are potentially hazardousor of sewage derived origin (Williams et al., 1999; Tudor and Williams (inpress)). Aids to help identify items of debris have been developed both in theU.S. and U.K. (CMC, 1993; Earll et al., 2000). These guides consist either ofphotographs or sketches of individual pieces of litter, with descriptions of theirattributes and function. The use of such visual aids is an important step forwardin the process of sourcing, and the pooling of information into a photographicresource can help both layman and ‘experts’ alike.

• FunctionOnce an item has been identified it is essential to know its use. For example,containers on beaches are often found to have been deliberately cut or splitin half. These containers have been used for a secondary purpose, i.e. they arebeing used as bailers in boats, or as a receptacle for oil changes or to hold paint.Similarly, containers or tyres will often be found with rope attached, indicatinga secondary use. The function of the item will therefore link back to the source,in this example to a shipping/sea based source. It is simply not enough to recordsuch an item as a ‘container’. The data gathering process is vital where thesource of litter is the ultimate aim of any monitoring programme.

• QuantitySome items are particularly rare (e.g. syringes), and the presence of a solit-ary item is not sufficient to point to a particular source dominating the litteron a beach. Items found in large quantities (e.g. cotton bud sticks, pieces offishing net) will point either to deliberate dumping, an accidental spill, or aregular input. The quantity of litter items found on a beach must be taken intoconsideration where sourcing is concerned.


4. Attribution to a Specific Source

These three factors – identity, function and quantity – contribute to the attribu-tion process. Even with this information, ascribing a precise source can still beperplexing. Some attempts have been made to apportion litter to specific types ofshipping vessels (e.g. Whiting, 1998; Earll et al., 1999), but this is an impreciseprocess that is open to large errors. There are though some distinctions that can bemade between shipping vessels, for example merchant shipping may have largerquantities or sizes of household and food goods on board than those of smallerfishing vessels. They may also carry duty free spirits and packaging items frommore distant parts of the world (Dixon, 1995). Ferry and cruise liners often carryproducts on board that bear their logo, which will obviously aid sourcing if suchitems are found on beaches.

5. Associations

There are certain items that are ‘indicators’ of a particular source. Such itemsinclude cotton bud sticks as an indicator of sewage/river source, or fragments ofnetting as an indication of a sea/fishing source. Other items, such as plastic drinkbottles, can have numerous sources. It is the association of items that is important,for example, if these bottles are found in conjunction with many items of fishingdebris or shipping waste, it can be assumed that a proportion of these beveragecontainers are from a shipping source. On the other hand, if these same bottles areon a beach with no fishing or shipping debris, but with large amounts of sewageand domestic containers, then this would point to debris being of a predominantlyland-based nature (Earll et al., 1999). However, trying to establish what proportionof litter has come from each source is a more difficult proposition. There is oftena huge diversity of litter items on beaches, and there is also an enormous diversitywithin litter item groupings. For example, beverage and food containers can varyenormously in their size, shape, and colour. This large range of diversity withinitem groups may help point to a particular source.

Often sourcing attribution is based heavily on assumptions (Shiber and Barrales-Rienda, 1991). Much of the attribution process is simply common sense applied tothe items encountered. If there are large amounts of broken lobster pots or floatsfor lines, then a significant source of litter will be fishing vessels. The difficultyarises in trying to apportion litter to specific sources; it is the mix of litter togetherwith associations between items that is important.


Figure 1. Location of beaches studied (Refer to Table IV for site names).


6. The Study Area (Figure 1)

The Bristol Channel is a large-scale estuary that has the second highest tidal range16.4 m in the world. The southern shore is essentially an agricultural rural areadrained only by a few small rivers e.g. the Axe and is an area that experiencesa lot of summer tourism. The northern shore encompasses the heavily populated,industrial area of South Wales to the east, and a central portion which merges intothe rural tourist/agricultural region of west Wales. Little shipping is found in theeast, but fishing and shipping occurs in considerable quantities to the west. Severalrivers e.g. the Taff, Ogmore, Neath, bring much litter into the system. Mid Wales isessentially rural with a scattering of small coastal towns e.g. Aberporth, Harlech,that serve the tourist industry. The north Wales coastline has several large-scaletourist centers e.g. Rhyl, Llandudno, that cater for the mass market.

7. Methodologies

7.1. METHOD FOR RECORDING BEACH LITTER

Litter items encountered on beaches covered in this article were recorded with asmuch detail as possible, making a note of all printed and additional information.The size of survey site followed the EA/NALG (2000) methodology i.e. recordingall litter within a 100 m transect parallel to the beach. Site selection was based ona uniform geographic spread of beaches mainly along the Bristol Channel coast,U.K. and not because they were known to be heavily polluted, or in an area re-cognised for suffering from specific types of pollution e.g. open coasts or areasnear shipping lanes. In addition, several beaches from other parts of Wales wereincluded (Figure 1).

7.2. METHODOLOGIES USED IN SOURCING BEACH LITTER

There are a number of methodologies that attempt to attribute litter items to aspecific source. The procedures used and the merits of each method have beenconsidered along with some proposals for improvements. Each method is brieflyconsidered in the light of the aims of this article and survey area, but there havebeen few studies that have illustrated a systematic and robust method for beachlitter sourcing. The exceptions are:

• Percentage Allocation (e.g. Earll et al., 1999). The ideas and methods set out byEarll et al. (1999), and further developed by Earll et al. (2000) in formulating amethodology for the identification of ship derived litter. In essence it attributesa percentage score to the likelihood of a litter item coming from a particularsource, e.g. a wooden pallet could not come from a toilet and down a sewagepipe, so its probability score would be very small.


• Attribution by Litter Type (e.g. Marine Conservation Society – BeachwatchReports – MCS, 2000). The method employed by the Marine ConservationSociety (MCS) Beachwatch study is to assign each litter item to a specificsource. The main weakness of the Beachwatch approach is the attributionprocess. It is carried out away from the beach and from the items themselvessimply by attributing all records (from a form) of a particular item, as recordedby the volunteer, to the particular source (Earll et al., 1999). This method ofattributing litter certainly has merits and the use of lists of items linked tosources serves as a useful database of information. However, the prescriptiveand rigid nature of this method is perhaps not ideal for the purposes of thisstudy.

• Sourcing with the use of container information (e.g. Tidy Britain Group (TBG)– Dixon, 1995). This method is based on a national ocean focused vessel-source litter assessment study which seeks to identify any major differencesin the composition and quantities of beach litter from paired observations,approximately 10 years apart, at 185 sampling units situated around the U.K.coastline (Dixon, 1995; Earll et al., 1999). It was specifically designed to as-sess whether MARPOL Annex V (1973/1978) was working, and the methoddoes meet this criterion. The TBG method has focused almost exclusively oncontainers for the sourcing of litter. It was felt that the great diversity of itemsfound on Bristol Channel beaches meant that a more holistic sourcing methodwas necessary, with all litter items included in sourcing attempts.

• Use of Indicator Items (e.g. Ribic, 1998). This method is similar to that em-ployed by Beachwatch (MCS, 2000) in that lists of items are considered foreach source, the difference being that only specific indicators are consideredand only these are recorded at each beach survey. This scheme was developedto give an indication of changing litter amounts over time, rather than estab-lishing sources. However, the lists of litter items arranged into source groupscould be utilised for sourcing.

• Matrix Scoring Method (e.g. Whiting, 1998). This method attempts to propor-tion a percentage allocation of each debris item to each source to produce anoverall percentage allocation figure. Each litter item encountered is assigneda probability, and subsequent score, of the item originating from a particularsource. Litter items are cross tabulated with potential sources and the scoresbased on several factors: namely; markings and labeling of items, type ofdebris, distance to each source, amount of activity of each source within theregion, seasonal wind and current patterns (Whiting, 1998). Although back-ground knowledge and an understanding of the vagaries of marine debris isneeded before utilising this system, a certain amount of subjectivity is used inapportioning a likelihood score to each item of litter.

• Multivariate Analysis. This technique has not previously been employed inattempts at sourcing beach litter. It has the following aims and the first two


neatly describe the use of this form of analysis for a study of litter sourcing:

1. Searching for possible causal relationships between distribution and envir-onmental factors.

2. Searching for pattern or structure in a set of data (Tudor et al., 2002).3. Describing or summarising the data efficiently to reduce the data matrix to

a more manageable form (Gauch, 1982; Randerson, 1993).

Careful consideration of the merits of all the above methods was made, especiallywith regard to the appropriateness of implementation on beaches in the study areaof the Bristol Channel. All have their strengths and weaknesses, but some wereeither not easily applicable or are not pertinent for the area of study and the use ofmultivariate analysis was considered to be most appropriate.

7.3. MULTIVARIATE ANALYSIS

This uses an inductive, non-experimental approach to generate rather than test hy-potheses. In relation to litter sources it was hoped that relationships between certainlitter groups would be realised, along with associations between beach locationand the types and abundance of litter found. Multivariate analysis methods followone of two strategies, either Ordination (e.g. principal component analysis, factoranalysis, discriminant analysis), or Clustering (e.g. cluster analysis) or hybrids ofthese. Two methods of multivariate analysis were utilised in an effort to ascertainpatterns amongst beach litter items and survey sites, namely principal componentanalysis and cluster analysis.

7.3.1. Principal Component Analysis (PCA)This is a method of ordination widely used in many fields, in which axes or com-ponents are successively extracted from a matrix of similarities. In PCA all in-dividuals contribute equally to the components, avoiding dominance by outliers.Another advantage is that simultaneous ordinations for both individuals and at-tributes, such as beach sites and litter types, can be obtained by a single analysis.Ordination allows each individual, either a site or litter type, to be placed on one ormore constructed axes so that its geometrical position relative to its fellows reflectsits similarity to them (Randerson, 1993). The rationale for using this powerfulpattern recognition tool was to identify factors that accounted for variations withinthe data set. Plots are produced which enable visual interpretation to take place.

Mathematically, PCA involves eigen analysis of a symmetric matrix of simil-arities to produce a series of eigen values and their corresponding eigen vectors(Marshall and Elliott, 1998). There are as many eigen values as there are rows(or columns) in the matrix and conceptually they can be considered to measurethe strength (relative length) of an axis. Each eigen value has an associated eigen


vector. An eigen value gives the length of an axis; the eigen vector determines itsorientation in space.

7.3.2. Cluster AnalysisCluster analysis is a multivariate analysis technique and not as much a typical stat-istical test as it is a collection of different algorithms that put objects into clusters.The clusters formed with this family of methods should be highly internally homo-genous (members are similar to one another) and highly externally heterogeneous(members are not like members of other clusters). Unlike many other statisticalprocedures, cluster analysis methods are mostly used when there are no prior hy-potheses, but where research is still in an exploratory phase (Backer, 1994). Inessence, cluster analysis finds the most significant solution possible. Group mem-bers will share certain properties in common and it is hoped that the resultantclassification will provide insight into the structure of the data. A dendrogram(‘tree-like’ diagram) is produced, which summarises the process of clustering.Similar cases are joined by links whose position in the diagram is determined bythe level of similarity between the cases (Aldenderfer and Blashfield, 1984).

8. Results and Discussion

The standard EA/NALG (2000) protocol states that litter should be counted overa 100 m length of beach which includes litter from the backshore (dune, seawall,dry beach, etc.), to the current litter line, i.e. omitting the stretch of beach runningseawards from this point. This is borne out by Table I where, on average, onlyca. 1% of litter is found below the current strandline.

Principal Component Analysis is a novel and statistically robust method ofattempting to establish sources of debris, in this case on beaches from the BristolChannel, U.K. This method has been attempted on litter studies of rivers (Simmonsand Williams, 1997), but no such wide scale sourcing study has been attemp-ted for beach litter. The use of PCA avoids any subjectivity in attributing litteritems to a source. Placing litter items into functional groupings or classes was avery important task before analysis could begin. The litter was classed initiallyinto 12 categories where functionality was common (Table II). For example, thedomestic/household debris category consisted of items that included toiletries, de-tergents, cigarette lighters, etc. Similarly, drink related debris included items suchas beverage containers, straws, milk containers, bottle tops.

Pilot analyses were carried out using both the covariances and correlations asmeasures of similarity between the 12 broad functional categories of litter itemsrecorded at 22 beach survey sites. The covariance measure (non-standardised) re-tains differences in variance between litter categories, allowing those which occurin large abundance to exert large weightings on the beach survey sites where theyoccur. Alternatively, using the correlation coefficient standardises the variances


TAB

LE

I

Per

cent

age

ofde

bris

atva

riou

sar

eas

ofth

ebe

ach

%M

ineh

ead

San

dB

ayD

unst

erP

utsb

orou

ghW

oolo

com

beW

estw

ard

Lyn

mou

thW

esto

nB

erro

wH

artl

and

Com

beH

o!M

ain

Qua

yM

arti

nN

=42

N=

167

N=

34N

=66

N=

63N

=32

0N

=79

N=

576

N=

1603

N=

138

N=

412

(a)

Sou

thsh

ore

ofB

rist

olC

hann

el

Abo

vehi

ghes

t0

03

520

00

00.

563

0st

rand

line

Hig

hest

stra

ndli

ne0

88.5

06

1.5

860

1265

2656

Zon

ebe

twee

n0

00

01.

513

.50

195

00

stra

ndli

nes

Cur

rent

stra

ndli

ne10

011

9742

970.

510

069

27.5

843

Bel

owcu

rren

t0

0.5

00

00

00

23

1st

rand

line

%A

berd

ovey

Tow

ynB

arm

outh

Har

lech

Pw

helli

Lla

ndud

noR

hyl

Fre

shw

ater

Ang

leT

resi

lian

Wes

tB

ayN

=17

N=

63N

=44

N=

40N

=36

N=

68N

=54

N=

480

N=

258

N=

215

(b)

Wal

esC

oast

line

Abo

vehi

ghes

t5

40

03

18

211

35st

rand

line

Hig

hest

stra

ndli

ne18

143

6550

8519

3556

40Z

one

betw

een

15

135

80

414

1016

stra

ndli

nes

Cur

rent

stra

ndli

ne75

7582

3034

1367

2520

8B

elow

curr

ent

12

20

51

24

31

stra

ndli

ne


TABLE II

Broad litter categories used in pilot analyses

Sewage related debris (SRD)

Shipping/fishing related debris (FRD)

Unidentifiable fragments (UPF)

Drink related debris (DRD)

Food related debris (FOOD)

Domestic/household related debris (HOUS)

DIY/maintenance related debris (DIY)

Packaging items (PACK)

Miscellaneous items (MISC)

Gross litter (GROS)

Potentially harmful (HARM)

Animal faeces (FAE)

of all litter categories such that all are given equal weightings, so items whichoccur simply with large numerical abundance and non-uniform distribution, do notunduly influence the analysis.

It was apparent from the covariance analysis (non-standardised) of the broadlitter categories, that sewage related debris (SRD), shipping/fishing debris (FRD),and unidentifiable plastic fragments (UPF) were given heavy weightings on oneor more of principal components (PCs) 1, 2 and 3, reflecting their relatively largeabundance and distinctive distributions. The remaining litter categories clusteredaround the zero points of all 3 PCs, indicating their low overall occurrence oruniform distribution. Similarly, most of the beaches were found to cluster aroundthe zero points of PCs 1–3, with the exception of a few distinctive beach locationscharacterized by large amounts of the few litter categories referred to above. Verylittle insight was gained from this analysis, which is dominated by the most abund-ant litter categories, and by the sites at which they occur. A central cluster containedsites with either small litter abundance figures (e.g. Aberdyfi), or the presence onlyof items which tend to be ubiquitous (e.g. sweet wrappers, plastic fragments).

PCA of correlations of the broad litter categories (standardised) produced verydifferent results, with fewer litter categories clustered around the origin of the PCs(zero point). ‘Household’ and ‘gross’ litter categories were both strongly weightedon component 2, and ‘fishing related debris’ and ‘DIY’ were separated from othercategories. Of the beach sites, Berrow, Hartland Quay and Freshwater West wereclear outliers. Both Hartland Quay and Freshwater West are at the extremity of thestudy area, on the outer Bristol Channel, and litter at these two beaches comprisedmany fishing and shipping items. At Berrow, numerous cotton bud sticks werepresent (n = 711) in the 100 m stretch of beach.


TABLE III

Key to litter items in Figures 4 and 6

PC code Litter items PC code Litter items

1 Soft drink bottle container 24 Cloth, shoe

2 Aluminium can – beer or soft drink 25 Party popper

3 Milk container 26 Pen

4 Toiletry container – e.g. toothpaste, tooth-brush, shampoo, deodorant

27 Syringe

5 Food containers – e.g. margarine, mayonnaise 28 Balloon

6 Take away food container/plasticcups/wooden forks-plastic spoon

29 Children’s toys

7 Detergent container 30 Tangles of netting

8 Cotton Bud Stick 31 4 pack holder

9 Sewage derived items (condoms, femininesanitary products)

32 Polyurethane

10 Netting/line 33 DIY/maintenance

containers (e.g. diesel

injector cleaner, bucket)

11 Other fishing items (e.g. lobster pot, fish box,etc.)

34 Toilet freshener

12 Shipping general (e.g. tyre with rope, fender,buoy)

35 Flower pot

13 Unidentifiable fragments 36 Wood

14 Sweet wrappers, drinking straw, lollipopsticks, soft drink cartons

37 Balloon

15 Packing strap 38 Piping/ducting

16 Polystyrene 39 25 L oil drum

17 Cigarette lighter 40 5–10 L oil containers

18 Cigarette stubs 41 bait bag

19 Beverage bottle top, tamper proof ring 42 plastic sheet

20 Plastic bag 43 glass bottle

21 Secondary use container 44 paper

22 Land based items: e.g. hub cap, traffic cone,car products, shopping trolley, road works

45 light bulb

23 Shotgun cartridge

An inherent problem with these initial attempts at litter sourcing using mul-tivariate analysis was the broad categories of items used. For example, ‘drinkrelated debris’ covered litter such as plastic drink bottles, drinking straws, milkcontainers, etc. Whilst all these items can reasonably be described as ‘drink re-


TABLE IV

Key to litter survey sites in Figures 1, 2, 3, 5, 7, 8 and 9, with survey dates

PC code Location PC code Location

1 Sand Bay 20/07/2000 25 Dunster Beach 21/03/2000

2 Sand Bay 22/03/2000 26 Minehead 21/03/2000

3 Aberdyfi 23/08/2000 27 Westward Ho! 21/03/2000

4 Towyn 23/08/2000 28 Brean 21/03/2000

5 Barmouth 23/08/2000 29 Weston 21/03/2000

6 Harlech 24/08/2000 30 Berrow 21/03/2000

7 Pwllheli 24/08/2000 31 Hartland Quay 22/03/2000

8 Broadhaven 06/11/2000 32 Combe Martin 22/03/2000

9 Tenby North 06/11/2000 33 Freshwater West 12/09/1999

10 Tenby South 06/11/2000 34 Angle 12/09/1999

11 Nolton 06/11/2000 35 Blue Anchor 06/08/2000

12 Mwnt 06/11/2000 36 Ilfracombe 08/08/2000

13 Poppit Sands 06/11/2000 37 Merthyr Mawr 26/01/1998

14 Wisemans Bridge 06/11/2000 38 Tresilian 20/12/1998

15 Pendine Sands 06/11/2000 39 Tresilian 04/01/1999

16 Croyde 10/09/2000 40 Tresilian 17/01/1999

17 Putsborough 10/09/2000 41 Tresilian 03/02/1999

18 Putsborough 22/03/2000 42 Tresilian 21/02/1999

19 Woolocombe 10/09/2000 43 Tresilian 08/03/1999

20 Woolocombe 22/03/2000 44 Merthyr Mawr 01/04/1998

21 Lynmouth 20/09/2000 45 River Ogmore 01/04/1998

22 Lynmouth 21/03/2000

23 Blue Anchor 20/09/2000

24 Blue Anchor 21/03/2000

lated’, the potential sources could be quite different. For example, it was assumedthat drinking straws would most likely have originated from a beach user source,whereas milk containers probably derived from a sea borne source. Combiningitems from potentially differing sources was likely to hinder interpretation of anypatterns which may appear on principal component axes. Hence data on litter fromthe full 45 beach surveys were re-classified by sub-dividing into more specificand less prescriptive categories and minor groups such as sewage derived items(Table III). Six of the beaches where surveys had been carried out on different dayswere included more than once (e.g. Tresilian Bay and Merthyr Mawr (Table IV)).

PCA results (using correlation coefficient) for these expanded categories oflitter items and for the 45 beach surveys are given in Figures 2 and 3 showing


Figure 2. Principal component analysis of beach survey sites using specific litter item classification– Principal components 1 and 2 (standardised data). For key to beach sites see Table IV.

Figure 3. Principal component analysis of beach survey sites using specific litter item classification– Principal components 2 and 3 (standardised data). For key to beach sites see Table IV.


Figure 4. Principal component analysis of litter items (excluding three outliers) using specific litteritem classification – Principal components 1 and 2 (standardised data). Key: R = River source; Sh =Shipping source; F = Fishing source; B = Beach user source; S = Sewage related debris source. Forkey to litter items see Table III.

the distributions of beach survey sites on the first three PCs. As in the pilot studythree beaches were found to be clear outliers on each of the three components.Berrow (30; Figure 2), and Hartland Quay (31; Figure 3) were again separatedfrom the main group of beaches but, using the specific litter items, in this analysis,Freshwater West (33) was replaced as an outlier on PC1 by a Merthyr Mawr beachsurvey (37; Figure 2). These three distinctive beach sites each contain exceptionallylarge numbers of particular litter items which, despite standardising the variables,acquire high loadings on the respective components (e.g. 711 cotton bud sticksfound at Berrow; 96 sewage derived items at Merthyr Mawr). Excluding themfrom subsequent analyses would enable greater discrimination of beaches in themain cluster and could elucidate relationships between litter sources.

When the three beaches (30, 31 and 37) were excluded, source groupings be-came apparent as PCA picked out patterns (relationships) among the litter items(Figure 4). Fishing sources (F), shipping sources (Sh), sewage related sources (S)and to some extent river sources (R) were identifiable as relatively discrete groups.Beach user sources, although separated in Figure 4 from other litter types (land/drywaste; e.g. 14 and 18) did not form a coherent group. These two items occupiedopposite poles of PC1, indicating contrasting beach locations. This highlights theproblem of distinguishing such ‘beach user’ items (sweet wrappers, plastic drinkbottles) from riverine or ocean based sources. The mode of transport rather than


Figure 5. Principal component analysis of beach survey sites (excluding three outliers) using specificlitter item classification – Principal components 1 and 2 (standardised data). For key to beach sitessee Table IV.

the original source of the litter types may determine their beach distribution. Hencesmall sewage related items do not group with the other ‘land’ based items such astraffic cones or shopping trolleys as they could also come from direct outfall inputsto the sea.

Figure 4 produced a number of distinct source groups of litter items. The up-per left segment has an association of what can be defined as river derived items(group R). The riverine classification is difficult to justify, with certain items (suchas traffic cones, shopping trolleys, hub caps) having an almost certain land/riversource, whereas others such as DIY/maintenance containers or plastic bags areless definite. This riverine grouping does correlate with the survey sites (Figure 5),comprising the River Ogmore (45), and Merthyr Mawr beach (44), the latter beingitself situated at the mouth of the Ogmore, which is known to be heavily influencedby riverine debris (Simmons and Williams, 1997).

Items of shipping waste also formed a discrete group (Sh, Figure 4). Itemsgrouped together were rope, fender buoys, milk containers, food containers (mar-garine tubs etc., but not ‘take away’ containers), and secondary use containers(e.g. bailers, oil change containers). Items of fishing debris were also groupedtogether (F, Figure 4), with netting/line found in conjunction with lobster pots, fishboxes, packing straps, plastic sheeting, and manufactured wood. The shipping andfishing groups are separated by sewage related items (i.e. sanitary towels, tamponapplicator etc.), as well as cotton bud sticks (Q-tips) and ‘toilet cleansers’.


As indicated above, the beach user source category is less well defined. Itemsthat were expected to originate from this source did not group together (B, Fig-ure 4). Sweet wrappers (14) were close to the sewage related group, others weredispersed. Cigarette stubs (18) were found to be separate from the other litter items,occupying the positive pole of component 1 (18; Figure 4). Although this item isalmost certainly from beach users, it may have different movement patterns to theother items of litter commonly left by beach users such as children’s toys or take-away food wrappers/packaging. Whereas cigarette stubs tend to stay where theywere dropped, other ‘beach user’ items may equally have come down rivers (sweetwrappers) or from the sea (drink containers). Fishing, shipping and sewage all haveitems within their classifications that are almost certain to come from a particularsource, in contrast to items left by beach users, and to some extent those of riverorigin.

Freshwater West (33; Figure 5) is separate from the main cluster of beachsites and its orientation is correlated with the fishing source group from Figure 4.Similarly, the beaches Sand Bay and Brean (1 and 28, Figure 5) are at a similarorientation with respect to PCs 1 and 2 to the sewage source group in Figure 4. Thiscorrespondence of location on the PCs between particular beaches (the componentscores of the PCA) and the litter items (the eigenvectors of the PCA), arises eitherbecause the particular beach contains large numbers of those items, e.g. sewagerelated items at Sand Bay (1) and Brean (28), or that they have predominantly litteritems from a single source, e.g. shipping/fishing debris at Freshwater West (33).Alternatively sites may be positioned because they lack litter items which occupyother regions of the PC plane, e.g. Tresilian (39, 40), lacking sewage derived items.

Examining the distribution of litter items and beaches in the plane of principalcomponents 2 and 3 (Figures 6 and 7), helps to draw out other source groupings thatwere not apparent on components 1 versus 2. Sewage related sources now separatestrongly on PC3 from the fishing and shipping sources (Figure 6). Sweet wrappers(14) remain grouped with sewage items. Large oil containers (39) are now groupedclosely with the shipping debris. River sourced items are more loosely grouped onPC3. Items such as detergents or toiletry containers often end up on beaches withshipping inputs (Hartland Quay, Freshwater West) as well as on beaches with riversources (Merthyr Mawr). Such multi-source items are difficult to source; they areperhaps best categorized with reference to other items with which they are grouped.

Freshwater West (33, Figure 7) is again distinct, being located at the poles ofboth PC2 and 3, away from the main cluster, and in opposition to Sand Bay (1),Brean (28) and Weston-Super-Mare (29). These three beaches are geographicallyvery close. They show obvious differences in litter composition in comparison toother beaches in the main cluster, having large amounts of sewage related items,probably arising from the River Parrett. Freshwater West (33) also has a differentlitter profile to many of the other beaches. Merthyr Mawr (44) and the River Og-more (45), having very similar litter composition, are grouped together on PC3, asthey were on both principal components 1 and 2 (Figures 5 and 7).


Figure 6. Principal component analysis of litter items (excluding three outliers) using specific litteritem classification – Principal components 1 and 2 (standardised data). Key: R = River source; Sh =Shipping source; F = Fishing source; B = Beach user source; S = Sewage related debris source. Forkey to litter items see Table III.

Figure 7. Principal component analysis of beach survey sites (excluding three outliers) using specificlitter item classification – Principal components 1 and 2 (standardised data). For key to survey sites,see Table IV.


Figure 8. Cluster analysis – Specific litter classification categories (non-standardised variables). Forkey to beach sites see Table IV.

Cluster analysis was performed, both with and without standardization of thedata, using the squared Euclidean distance similarity coefficient, with Ward’s me-thod of clustering (Figures 8 and 9). In this approach, it is not necessary to ex-clude numerically extreme beaches, e.g. Berrow (30) and Merthyr Mawr (37),from cluster analyses, as outliers of this sort do not affect the clusters producedfor the remaining sites, as they do in PCA. In both analyses, these two beacheswere split off prior to formation of the main clusters (Figures 8 and 9). Clusteranalysis without standardizing the data (Figure 8) identified a distinct group ofsites, Sand Bay, Brean and Weston (1, 28, 29) as in PCA (Figure 7). Other beachescomprised a relatively homogeneous group, with the exception of a sub-group ofthree beaches, Westward Ho!, Combe Martin and Freshwater West (27, 32, 33),all located in the outer Bristol channel. When the data are standardized, the influ-ence of the most numerous litter items is reduced. This analysis identifies HartlandQuay (31, Figure 9) as an outlier, in addition to beaches 30 and 37, again similarto PCA (Figure 7). A large sub-group appears comprising mostly beaches on thesouthern shore of the Bristol Channel, which indicates a difference in litter profilesbetween these beaches and those of the northern shore and the inner Channel. Nodifferences were found between litter obtained from beaches of the northern andsouthern shores of the inner Bristol Channel.


Figure 9. Cluster analysis – Specific litter classification categories (standardised variables). For keyto beach sites see Table IV.

9. Conclusions

Principal component analysis successfully distinguished riverine, sewage related,fishing, and shipping sourced items, but did not produce a coherent grouping forbeach user litter. This may be due to the different modes of transport of the types oflitter thought to originate from beach users. Beaches of mid and north Wales werenot differentiated from the central group of beaches located around the origin ofall three components, probably due to the small amounts of litter found. Berrow,Hartland Quay and Freshwater West, before their exclusion from the PCA, had verydifferent litter profiles to other Bristol Channel beaches. The western segment ofthe Channel is influenced more by shipping/fishing inputs than the eastern. Clusteranalysis grouped the southern outer Bristol Channel beaches, but no distinctionwas made between beaches on the two sides of the inner Bristol Channel.

References

Aldenderfer, M. S. and Blashfield, R. K.: 1984, Cluster Analysis, Sage, London, 88 pp.Backer, E.: 1994, Computer-assisted Reasoning in Cluster Analysis, Pearson Education, 300 pp.CMC: 1993, Pocket Guide to Marine Debris, Center for Marine Conservation, Washington, DC,

U.S.A., 35 pp.Corbin, C. J. and Singh, J. G.: 1993, ‘Marine debris contamination of beaches in St. Lucia and

Dominica’, Marine Pollut. Bull. 26(6), 325–328.


Dixon, T. R.: 1995, Temporal Trend Assessments of the Sources, Quantities and Types of LitterOccurring on the Shores of the United Kingdom: Introduction and Methods with Results fromPaired Observations 8 and 11 Years Apart on 63 Sampling Units in Mainland Scotland and theWestern Isles, Marine Litter Research Programme, Stage 7, The Tidy Britain Group, Wigan,U.K., 84 pp.

EA/NALG (Environment Agency and The National Aquatic Litter Group): 2000, Assessment of Aes-thetic Quality of Coastal and Bathing Beaches. Monitoring Protocol and Classification Scheme,May 2000, 15 pp.

Earll, R. C., Moore, J., Williams, A. T. and Tudor, D. T.: 1999, ‘The Measurement of Oily Wasteand Garbage Disposed of Into the Marine Environment by Shipping’, A Report to the Maritimeand Coastguard Agency, Prepared by Coastal Management for Sustainability, Candle Cottage,Kempley, Glos, U.K., 75 pp.

Earll, R. C., Williams, A. T and Tudor, D. T.: 2000, Pilot Project to Establish Methodologies andGuidelines to Identify Marine Litter from Shipping, Maritime and Coastguard Agency ResearchProject, No. 470, 137 pp.

Frost, A. and Cullen, M.: 1997, ‘Marine debris on northern New South Wales beaches (Australia):Sources and the role of beach usage’, Marine Pollut. Bull. 34(5), 348–352.

Gabrielides, G. P., Golik, A., Loizides, L., Marino, M. G., Bingel, F. and Torregrossa, M. V.: 1991,‘Man-made garbage pollution on the Mediterranean coastline’, Marine Pollut. Bull. 23, 437–441.

Gauch, H. G.: 1982. Multivariate Analysis in Community Ecology, Cambridge University Press,Cambridge, 298 pp.

Golik, A. and Gertner, Y.: 1992, ‘Litter on the Israeli coastline’, Marine Environ. Resear. 33, 1–15.Gregory, M. R.: 1998, ‘Pelagic Plastics and Other Synthetic Marine Debris – A Chronic Problem’,

in C. Wallace, B. Weeler and S. Buchan (eds), Marine Ecosystem Management: Obligations andOpportunities. Proceedings of the February 1998 Sea Views Conference; ECO (Environment andConservation Organisations of New Zealand), Wellington, N.Z., pp. 128–135.

Marshall, S. and Elliott, M.: 1998, ‘Environmental influences on the fish assemblage of the Humberestuary, U.K.’, Estuar. Coast. Shelf Sci. 46, 175–184.

MARPOL: 1973/1978, ‘International Convention for Prevention of Pollution of the Sea from Ships’,International Maritime Organisation Convention, 1973.

MCS (Marine Conservation Society): 2000, ‘Beachwatch ’99’, Nationwide Beach Clean and SurveyReport, 60 pp.

Nash, A.: 1992, ‘Impacts of marine debris on subsistence fishermen – An exploratory study’, MarinePollut. Bull. 24(3), 150–156.

Randerson, P. F.: 1993, ‘Ordination’, Chapter 5, in J. C. Fry (ed.), Biological Data Analysis, IRLPress, pp. 173–217.

Ribic, C. A.: 1998, ‘Use of indicator items to monitor marine debris on a New Jersey beach from1991–1996’, Marine Pollut. Bull. 36(11), 887–891.

Shiber, J. G. and Barrales-Rienda, J. M.: 1991, ‘Plastic pellets, tar, and megalitter on Beirut beaches,1977–1988’, Environ. Pollut. 71, 17–30.

Simmons, S. L. and Williams, A. T.: 1997, ‘Qualitative versus Quantitative Litter Data Analysis’, inE. Özhan (ed.), Proceedings of the Third International Conference on the Mediterranean CoastalEnvironment, MEDCOAST ’97, pp. 397–406.

Thornton, L. and Jackson, N. L.: 1998, ‘Spatial and temporal variations in debris accumulation andcomposition on an estuarine shoreline, Cliffwood Beach, New Jersey, U.S.A.’, Marine Pollut.Bull. 36(9), 705–711.

Tudor, D. T. and Williams, A. T.: 2002, ‘Public perception and opinion of visible beach aestheticpollution: The utilisation of photography’, J. Coastal Resear. (in press).

Tudor, D. T., Williams, A. T., Randerson, P. F., Ergin, A. and Earll, R. E.: 2002, ‘The use of mul-tivariate statistical techniques to establish beach debris pollution sources’, J. Coastal Resear. (inpress).


Whiting, S. D.: 1998, ‘Types and sources of marine debris in Fog Bay, Northern Australia’, MarinePollut. Bull. 36(11), 904–910.

Williams, A. T., Morgan, R. and Tudor, D.: 2000, ‘Beach Litter from Land-based Sources’, Inter-Coast Network, Coastal Resources Center, University of Rhode Island, U.S.A., 36, Spring 2000,25 pp.

Williams, A. T., Pond, K., Tudor, D. T., Jansen, H. and Liu, H. B.: 1999, ‘The Robustness ofLitter Transect Data Collection by Different Survey Groups’, in E. Özhan (ed.), Proceedingsof the MEDCOAST ’99 – EMECS ’99 Joint Conference: Land Ocean Interactions – ManagingCoastal Ecosystems, 9–13 November 1999, Antalya, Turkey. MEDCOAST, Middle East TechnicalUniversity, Ankara, Turkey, pp. 715–725.

Beach litter sourcing in the Bristol channel and Wales, U.K

Documents

Transcript of Beach litter sourcing in the Bristol channel and Wales, U.K