Aquaculture research, tập 42, số 1, 2011

Quantitative microbial source apportionment as a toolin aiding the identification of microbial risk factors in shellfish harvesting waters: the Loch Etive case study Carl M Stapleton1, D

Quantitative microbial source apportionment as a tool

in aiding the identification of microbial risk factors in shellfish harvesting waters: the Loch Etive case study

Carl M Stapleton1, David Kay1, Shona H Magill2, Mark D Wyer1, Cheryl Davies3, John Watkins3,Chris Kay1, Adrian T McDonald4& John Crowther5

1 CREH, Catchment and Coastal Research Centre, River Basin Dynamics and Hydrology Research Group, IGES, University of Aberystwyth, Ceredigion, UK

2 Scottish Association for Marine Science, Dunsta¡nage Marine Laboratory, Oban, UK

3 CREH Analytical, Hoyland House, Leeds, UK

4 Faculty of the Environment, University of Leeds, Leeds, UK

5 CREH, University of Wales, Lampeter, Ceredigion, UK

Correspondence: D Kay, CREH, Catchment and Coastal Research Centre, River Basin Dynamics and Hydrology Research Group, IGES, University of Aberystwyth, Ceredigion SY23 3DB, UK E-mail: dave@crehkay.demon.co.uk

Abstract

Sanitary surveys of shell¢sh harvesting waters are

now a routine component of regulatory monitoring

These provide a qualitative appraisal of potential

pollutant sources impacting on shell¢sh microbial

quality The information provided by this type

of screening level appraisal is very useful, but does

not a¡ord quantitative assessment of the di¡erent

pollution sources and their complex dynamic

rela-tionships which result in a highly episodic £ux of

microbial parameters into shell¢sh harvesting

waters The potential £uxes derive from treated

sew-age and industrial e¥uents, intermittent discharges

from the sewerage system and di¡use sources of

pollution, principally from livestock farming areas,

but also from urban surface water drainage None of

these sources are routinely monitored for the faecal

indicator parameters used as compliance measures

by regulators worldwide and almost no high-£ow

information is available with which to construct

any quantitative £ux estimates to provide a credible

evidence base for the design of remediation strategies

where there is a need to improve water quality within

a harvesting area This study was conducted at Loch

Etive, near Oban, Scotland, UK It applies an

approach to Quantitative Microbial Source

Appor-tionment developed to inform management and

remediation of water quality at bathing water

loca-tions The results suggested that, in this case study

location, di¡use sources of microbial indicator isms derived from livestock farming activities incatchments draining to the loch were the dominanthigh-£ow contribution of bacterial loadings This

organ-¢nding was unexpected by local managers who hadperceived ‘environmental’ water quality as ‘high qual-ity’ in this traditionally pristine area of west Scotland.The ¢ndings led to a series of recommendations forfuture management of Scottish shell¢sh harvestingwaters directed at appropriate data acquisition,through a detailed sampling programme design, toacquire microbial £ux data from all sources, particu-larly during high-£ow event conditions It wasrecommended that such data acquisition was essen-tial to the design of any remediation strategies thatneed a credible evidence base directing appropriateinvestment in interventions designed to attenuatemicrobial £ux from either the sewerage infrastruc-ture and/or adjacent farming activities The utility ofthis study could be further enhanced through micro-bial tracer studies to establish connectivity betweenthe key hydrological inputs (both those studied hereand potential sources outside of the lower basin) andthe shell¢sh beds

Keywords: faecal indicator organisms, tive Microbial Source Apportionment (QMSA), bath-ing, shell¢sh, Water Framework Directive, CleanWater Act, source characterization

Quantita-Aquaculture Research, 2011, 42, 1^20 doi:10.1111/j.1365-2109.2010.02666.x

r 2010 The Authors

In the European Community (EC), shell¢sh growing

waters (SGW) are identi¢ed as a ‘protected area’ by

the Water Framework Directive (WFD) (see Annex

IV: Section 1(ii); Anonymous 2000) Article 4c of this

Directive places a legal responsibility on EC member

states to ensure that these protected areas achieve

compliance with ‘any standards and objectives’

de-¢ned in the WFD or its daughter directives

The principal water quality standards in

near-shore marine and estuarine waters are de¢ned in

terms of microbiological parameters, which indicate

pollution by faecal wastes from human and/or

animal sources These indicator bacteria include the

coliforms and enterococci In the case of shell¢sh

waters and shell¢sh £esh, the coliform bacteria and,

the intestinally derived component of this group, i.e

Escherichia coli, are the principal compliance

para-meters in current standard systems de¢ned in the

Shell¢sh Water Directive (Anonymous 1979, 1997,

2006a) and the Shell¢sh Hygiene Directive

(Anon-ymous 1991), now replaced by EC Regulation No

854/2004 (Anonymous 2004)

Parallel and complementary EC Directives

cover-ing coastal bathcover-ing waters in the United Kcover-ingdom

also de¢ne coliform standards (Anonymous 1976,

2006b) as well as marine water quality criteria based

on epidemiological evidence which use the

entero-cocci group of indicator bacteria (WHO 2003; Kay,

Bartram, Pruss, Ashbolt, Wyer, Fleisher, Fewtrell,

Rogers & Rees 2004; Kay, Ashbolt, Wyer, Fleisher,

Fewtrell, Rogers & Rees 2006)

Regulation of the microbial quality of shell¢sh

£esh, which is a food product, is the responsibility of

the Food Standards Agencies in the United Kingdom

This is responsible for the classi¢cation of SGW into

four categories (de¢ned as A^D) determined by E coli

standards assessed by pre-determined sampling of

shell¢sh £esh However, the environmental

regula-tors in the United Kingdom, i.e the Environment

Agency (EA) in England and Wales and the Scottish

Environmental Protection Agency in Scotland, are

responsible for the implementation of the WFD and

consequential compliance with daughter directives,

i.e they must so regulate the environment to result

in a required food hygiene criteria

The EA in England and Wales is charged to ensure

that ‘bacteriological standards should be achieved

which allow designated waters to achieve at least

cate-gory B standard under the system applied by the Food

Standard Agency to classify shell¢sh harvesting areas

for food safety purposes’ (EA 2003a) The WFD vides the mechanism for environmental regulators toachieve this statutory requirement This is outlined inArticle11of theWFD that required that Member Statesproduce a‘Programme of Measures’ (POM) designed toachieve required standards in each river basin This is

pro-to include integrated management and control of both

‘point source’ discharges [Article 11.3(g)] and ‘di¡use’pollution [Article 11.3(h)] These POMs contribute toresultant basin and sub-basin management plans re-quired under Article 13.5

The principal sources of the bacterial complianceparameters for near-shore waters used for shell¢shharvesting and/or bathing is faecal material derivedfrom human and/or animal populations In the Uni-ted Kingdom, the human input is characterized by

‘point source’ discharges of treated e¥uents fromsewage works and/or intermittent discharges fromthe sewerage system during times of high rainfallwhen combined sewage over£ows (CSOs), and/orstorm tank over£ows (SSOs), operate to allow excessvolumes of untreated sewage e¥uent, diluted by sur-face water, to exit the sewerage system once its capa-city is exceeded (Kay, Crowther, Stapleton, Wyer,Fewtrell, Edwards, Francis, McDonald, Watkins &Wilkinson 2008; Kay, Kershaw, Lee, Wyer, Watkins &Francis 2008) The predominant source of di¡use mi-crobial pollution is the livestock farming sector Live-stock pollution of streams impacting on coastalprotected areas can be associated with animals void-ing faeces directly to the land surface or into streamsand onto riparian areas (Kay, Edwards, Ferrier, Fran-cis, Kay, Rushby,Watkins, McDonald,Wyer, Crowther

& Wilkinson 2007; Kay, Crowther, Stapleton, Wyer,Fewtrell, Anthony, Bradford, Edwards, Francis, Hop-kins, Kay, McDonald, Watkins & Wilkinson 2008).Other signi¢cant sources include farmyards wherediary cattle concentrate for milking, leaving faeces

on hard standing areas, which create signi¢cant

‘dirty water’ problems as rainfall mobilizes the faecesfrom concrete surfaces (Edwards & Kay 2008).Indeed, accepted animal waste handling and dis-posal practices also contribute to the di¡use micro-bial loading through land applications and slurryspreading activities In addition, practices designed

to concentrate livestock for stock managementand husbandry, such as woodchip corrals, can pro-duce signi¢cant source areas of faecal indicator £uxfrom catchment systems (McDonald, McDonald, Kay

& Watkins 2008)

By comparison with other water quality meters such as the nutrients (principally compoundsMicrobial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

para-of nitrogen and phosphorus), pesticides, sediment

and oxygen demand, the microbial parameters have

received very little scienti¢c and/or research

atten-tion that can be used by the regulatory community

to design ‘evidence-based’ POMs at the catchment

scale to achieve water quality standards guaranteed

to ensure appropriate food hygiene standards de¢ned

by E coli concentrations in shell¢sh £esh (Kay,

McDo-nald, Stapleton,Wyer & Fewtrell 2006) This is

poten-tially signi¢cant, given the experience in the United

States, where legislation very similar to the EC WFD

is seen in the Clean Water Act (CWA; Anonymous

2002) The United States Environmental Protection

Agency (USEPA) now publishes running totals of

the reasons for non-compliance of all waters coveredunder the CWA (USEPA 2009) This is termed ‘impair-ment’ under the US legislation Figure1 shows the top

10 reasons for impairments to date under the CWA.The US equivalent for the EC POMs is a Total Maxi-mum Daily Load (TMDL) study to de¢ne remedialmeasures needed to end the impairment Figure 2shows the top 10 US TMDL pollutant groups fromstudies completed since 1995

Implementation of the CWA is some 20 years ahead

of the WFD and the signi¢cance of microbial meters is clear from Figs 1 and 2 If this experience isrepeated in the comparable developed countries of the

para-EC, then the comparative lack of scienti¢c attention to

TurbidityCause Unknown - Impaired Biota

pH/Acidity/Caustic conditions

Polychlorinated Biphyneyls (PCBs)

Organic Enrichment/Oxygen Depletion

SedimentNutrientsMetals (other than mercury)

Mercury

8855736468616251623361793860

31993055

Cause of Impairment Group Name Number of Causes of Impairment Reported

Causes of Impairment for 303(d) Listed Waters

Figure 1 Reasons for impairment in all USA waters covered by the CleanWater Act on 17 September 2009 The microbialparameters implied by the word ‘Pathogens’are coliforms and enterococci (redrawn from http://iaspub.epa.gov/waters10/attains_nation_cy.control?p_report_type=T#causes_303d)

Pollutant Group Number of TMDLs Impairment Addressed Number of Causes of

National Cumulative TMDLs by Pollutant

This chart includes TMDLs since October 1, 1995

AmmoniapH/Acidity/Caustic conditionsSalinity/Total Dissolved Solids/Chlorides/Sulfates

TemperatureOrganic Enrichment/Oxygen Depletion

SedimentNutrientsMetals (other than mercury)

Mercury

667164434275

308017741536152315071028

7177670465655107356918491572153015481079

Figure 2 Total maximum daily load investigations completed by the USEPA to 17 September 2009 under the terms of theClean Water Act since 1995 The microbial parameters implied by the word ‘Pathogens’are coliforms and enterococci (re-drawn from http://iaspub.epa.gov/waters10/attains_nation_cy.control?p_report_type=T#tmdl_by_pollutant)

Aquaculture Research, 2011, 42, 1^20 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al.

r 2010 The Authors

the microbial parameters in catchment science and

modelling to date may yet prove unfortunate

The foundation of any POM or TMDL is a

quanti¢-cation of the microbial £uxes from each contributing

source impacting, or causing impairment of, a

shell-¢sh harvesting water There are very few such

stu-dies, to date, reporting empirically derived data from

¢eld survey and water quality analyses of shell¢sh

harvesting waters However, Quantitative Microbial

Source Apportionment (QMSA) studies, at UK

bath-ing waters, have been conducted for each of the sites

outlined in Fig 3 and the approach developed has

re-ceived peer review scrutiny in a series of

interna-tional journal papers (Wyer, Kay, Dawson, Jackson,

Jones,Yeo & Whittle 1996; Wyer, Kay,

Crowther,Whit-tle, Spence, Huen, Wilson & Carbo 1998; Kay, Wyer,

Crowther, Stapleton, Wilkinson & Glass 2005; Kay,

Wyer, Crowther, Stapleton, Bradford, McDonald,

Greaves, Francis & Watkins 2005; Wither, Greaves,

Dunhill, Wyer, Stapleton, Kay, Humphrey, Watkins,

Francis, McDonald & Crowther 2005; Stapleton,

Wyer, Crowther, McDonald, Kay, Greaves, Wither,

Watkins, Francis, Humphrey & Bradford 2008)

Materials and methodsThe Loch Etive studyLoch Etive is a large sealoch on the west coast of Scot-land (Fig 4) and is divided into two main basins,upper and lower (Magill, Black, Kay, Stapleton,Kershaw, Lees, Lowther, Francis, Watkins & Davies2008) The loch has a large catchment area (approxi-mately1350 km2) with a number of freshwater inputs

to the system The largest of the inputs to the lowerbasin is the River Awe, which enters the loch to theeast of Taynuilt A hydroelectric power scheme (HEP)

is located on this river, thus regulating the £ow Theimmediate catchment area of the study has a low hu-man population density (approximately 2500 concen-trated along the shoreline) with a further 1000(approximately) within 4 km of the mouth of the loch

In common with many sealoch coastlines of the westcoast of Scotland, the catchment consists of a number

of small villages and many rural dwellings

Approximately 53% of the immediate population isserved by public network sewage facilities Accuratedetailed information could not be gathered on all pri-vate domestic sewage systems in the catchment, but

it is estimated that a large number of dwellings areserved by septic tank systems However, in a number

of areas, shoreline dwellings discharge raw sewageinto the loch, both outside and within the SGW.One wastewater treatment works (WwTW; activatedsludge secondary treatment, population equivalentdesign capacity 51400) operates within the village

of Taynuilt, serving around 60% (approximately 700)

of the local population in that area The treated e¥uentand intermittent discharges of untreated storm e¥u-ent are discharged into the River Nant downstream

of the tidal limit Three public network septic systemsoperate within the area The two largest systems dis-charge to marine outfalls outwith the main lower basin

of the loch There are CSO and Emergency Over£owswithin the loch itself, but are outside the SGW A smallnetwork in the village of Bonawe discharges raw sew-age directly into the loch within the SGW This network

is small in capacity (23 dwellings), but is in close mity to the shell¢sh production sites

proxi-Agriculture in the catchment is dominated by stock grazing for sheep and cattle In excess of 5000ewes and 500 cattle graze within the immediatecatchment Sheep numbers during the summermonths, when lambs are present, may be approxi-mately double this estimate Much of the grazingareas are in close proximity to the shoreline Live-stock census data for the agricultural parishes in the

live-Figure 3 Location of quantitative microbial source

ap-portionment studies in the United Kingdom 1996^2007

Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

catchment area indicate that sheep numbers have

decreased by approximately 10% from 2000 to 2005

Examination of available historical classi¢cation

data collected by Fisheries Research Services (1999^

2000), and Food Standards Agency Scotland (FSAS;

2000^2006) indicated variation in E coli levels

be-tween the production sites in Loch Etive lower basin,

which houses the sampling sites It should be noted

that the historical data used in this sanitary survey

were collected before the implementation of more

stringent sampling protocols by FSAS in 2007 These

protocols required sampling to be taken from ¢xed

monitoring points by designated Environmental

Health O⁄cers (who are local government employees

in the United Kingdom) Potential variations in

sam-pling points used before 2007 means that the

inter-pretation of historical data within the sanitary

survey must be viewed with caution

Study design

A QMSA study was implemented to quantify faecal

indicator organism (FIO) budgets from sewage and

riverine sources draining to the lower basin of Loch

Etive, Scotland, above the Falls of Lora road crossing

The budgets were constructed to provide an

indica-tion of the relative contribuindica-tion of 26 out of 34

identi-¢ed sewage and stream inputs This did not include

study of the fate and transport of faecal indicatorswithin the waters of the loch itself that might infer alink from any discharge to compliance monitoringpoints Other inputs, for example from avian popula-tions direct to the loch, were also not quanti¢ed

It should be noted that this study considered a mer condition when other investigations have sug-gested that agricultural sources of faecal indicators

sum-in Scottish streams are at their maximum (Rodgers,Soulsby, Hunter & Petry 2003; Kay, Aitken, Crowther,Dickson, Edwards, Francis, Hopkins, Je¡rey, Kay,McDonald, McDonald, Stapleton, Watkins, Wilkinson

& Wyer 2007) Similar previous studies have largelybeen driven by bathing water compliance considera-tions and have, thus, centred on summer £uxes.Empirical data were acquired in two ¢eld campaignsspanning the 2006 and 2007 summer periods Waterquality, stream and sewage e¥uent £ow data werecollected during 2006 while further water qualityand river stage data were collected during 2007

River and sewage discharge, stream level andrainfall

Flow data for the River Awe at the HEP barrage, proximately 5 km upstream of the sampling point(Fig 4), were supplied by the operators of the HEP,Scottish Southern Energy Flow throughout the

ap-Figure 4 Riverine and sewage sample points monitored during the study (see Table 1 for sample point details) Rain gaugeswere located at sites 202, 204, 207 and 302 Stage recording monitors were located at sites 202, 204, 207, 208, 209 and 210

Aquaculture Research, 2011, 42, 1^20 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al.

r 2010 The Authors

2006 study period was constant at 14.2 m3s 1, with

the exception of ‘freshets’ of three 12-h periods over

the weekends of 8/9, 16/17 and 22/23 July 2006,

when the £ow was 23.7 m3s 1 For the remaining

period of the ¢eld study, planned freshets were not

re-leased due to low levels in Loch Awe

Five catchment outlet sites were monitored for stage

through installation of manometric-level recorders

and open-channel £ow gauging (EA 2003b) These

data were used to derive stage-discharge relationships

with r2-values (coe⁄cient of determination) ranging

between 98.3% (site 209) and 99.2% (sites 204 and

210) Rainfall input (mm) was monitored at four

tip-ping bucket rain gauges, installed for the duration ofthe study across the catchment (Fig 4,Table 1).The hourly discharge records for £ow monitoringstations were split into two components: (i) base-£owand (ii) high-£ow event response to rainfall This wasachieved using a combination of computer pro-grammes (PASCAL) and visual inspection of individualevents by detailed hydrograph analysis) Given theregulated £ow of the River Awe, it was considered inap-propriate to separate the £ow into base £ow and high

£ow because this was unlikely to be driven by rainfall.However, to estimate a representative £ow budget forLoch Etive, discharge from the River Awe should be

Table 1 Riverine and sewage e¥uent sample sites visited during the 2006 and 2007 surveys

Catchment area (km2) River and stream sample points

Sewage effluent sample points

Rain gauge locations.

wRiver level and £ow gauging locations.

zDid not operate during 2006 sampling.

ptc, prior to con£uence; FE, ¢nal e¥uent; CSO, combined sewage over£ow.

Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

included within both the base-£ow and the high-£ow

components To achieve this, £ow in the River Awe

was considered to be ‘high £ow’ during the periods of

high £ow in the River Luachragan, this being the

lar-gest and most proximal of the gauged catchments

Discharge from catchments from which no £ow

data were available were based on catchment area

and rainfall input using 2006 rainfall data from the

nearest available CREH rain gauge The proportion of

rainfall contributing to total £ow and the base-£ow

index (proportion of total £ow contributing to

base-£ow volumes) was taken from the nearest most

simi-lar monitored catchment

Flow data for Taynuilt WwTW [inlet, CSO and

com-bined ¢nal e¥uent (FE) and CSO £ow] were available

only for part of the 2006 survey period (i.e 456 h

be-tween 7 July 2006 and 26 July 2006) from the sewerage

undertaker The FE £ow was calculated as the

di¡er-ence of the combined FE1CSO and CSO data In the

ab-sence of a time-series of £ow for the complete study

period (i.e 7 July 2006 to 10 August 2006; 34 days), it

was felt that the most robust method for estimating the

£ow would be to scale the £ow for the 28-day period

between 28 June 2006 and 26 July 2006 by a factor of

1.214 (i.e 34/28) While this method would provide a

reasonable estimation of the total £ow over the study

period, it does not provide an hourly time-series for

the period of the study after the sewage £ow monitors

were removed Consequently, while overall budget

esti-mates of £ow and FIO delivery can be made, a full

time-series of £ow and FIO delivery can only be provided up

to the end of the available WwTW £ow data

The hourly 2006 discharge record for the FE at

Taynuilt WwTW was split into base-£ow and

high-£ow event components in response to rainfall This

was achieved through detailed inspection of the £ow

record and comparison with a typical daily dry

weather £ow pattern derived using data from eight

dry days and rainfall records Flows that deviated

sig-ni¢cantly from the typical daily dry weather £ow

pat-tern corresponding with rainfall were categorized as

high £ows All discharges from the CSO at Taynuilt

WwTW were categorized as high £ows The other

sewage e¥uent sample site, a CSO in Connel, did not

have any £ow measurement equipment installed

Riverine freshwater and sewage e¥uent

sampling

Sampling was carried out between 14 July and 6

Au-gust 2006 and between 6 AuAu-gust and 14 September

2007 Samples were collected from 29 river andstream sites (16 sites during 2006 with a further 13additional sites sampled in 2007) (Fig 4,Table1) Dur-ing 2006,11sites were selected to allow budget calcu-lations, while four sites were added to the samplingregime after initial results indicated particularlyhigh base-£ow FIO concentrations at three catch-ment outlets: two were located within the AbhainnAchnacree catchment, one on the main river (site213) and one on a tributary, Allt nam Ban (site 214),both upstream of their con£uence; one site wasadded to Inion Farm stream, upstream of Inion Farm(site 215), and one was located upstream of Ardchat-tan school on the Kenmore stream (site 216).The 2007 survey included three sites sampled dur-ing 2006: Inion Farm stream (site 202), the River Nant(site 211) and the River Awe (site 212) plus a further 13new catchment outlets (Fig 4, Table 1) These newstreams were all relatively small, with the largest catch-ment area being that of Allt an t-Siomain at 2.4 km2.Four sewage sample sites were sampled duringboth the 2006 and the 2007 surveys (Table 1) Threesites were located at Taynuilt WwTW and includedthe continuous treated FE discharge (site 301), the in-termittent untreated CSO e¥uent at the works (site302) and the combined e¥uent sampled from theircommon outfall (site 303) The forth sample pointwas of the storm over£ow e¥uent from the CSO inConnel, although this did not spill during the 2006study period Samples were collected during the

2007 survey, although the lack of any £ow data cludes its inclusion in the budget estimates

pre-Three sampling runs were scheduled during eachweek of the respective study periods This wasamended in response to rainfall, in order to targetsampling during hydrograph events and obtain sam-ples from intermittent sewage e¥uent dischargesfrom storm tanks and CSOs Multiple samples wereoften collected during such events

Samples were obtained either directly into 150 mLsterile disposable plastic containers (Sureseal Se-lectTM, SLS, Nottingham, UK) using a laboratoryclamp and telescopic landing rod or by lowering aclean stainless-steel can into the £owing water/e¥u-ent If the sampling can was used, this was loweredinto the £ow three times, on the ¢rst two occasions,the can was rinsed and the water/e¥uent was dis-carded After the third collection, a sample was ob-tained by pouring into a 150 mL sterile plasticcontainer On return to the sampling vehicle, the in-side of the sampling can was immediately dried withabsorbent paper towel and then wiped clean with anAquaculture Research, 2011, 42, 1^20 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al.

r 2010 The Authors

alcohol-impregnated cloth (Azo wipeTM, Synergy

Healthcare, Swindon, UK), allowing time for the

al-cohol to evaporate on the journey to the next

sam-pling location Samples were stored in the dark

inside a cool box during transport to the laboratory

Laboratory methods, data analysis and £ux

estimation

Indicator organism enumerations [colony-forming

units (CFU) 100 mL 1] followed Standing Committee

of Analysts Blue Book methods based on membrane

¢ltration (Standing Committee of Analysts (SCA)

2000, 2002, 2006) Sample dilutions were determined

from the initial sampling run and faecal coliforms (FC)

and intestinal enterococci (EN) enumerations were

performed at two or three sample dilutions A

com-plete duplicate analysis was carried out on at least

one sample collected during each sampling run for

quality control purposes All samples were analysed

within 24 h, in accordance with the Blue Book

meth-ods (SCA 2000, 2002, 2006)

A total of 469 samples were collected during the

2006 ¢eld study period, with a further 563 samples

collected during 2007 Results from each river and

stream sampling site were classi¢ed into two

cate-gories, base £ow and high £ow, according to £ow

con-ditions For temporary stations monitored during

2006, the £ow separated discharge record was used as

a basis for this classi¢cation.Where continuous records

were unavailable, the record from the nearest

neigh-bouring site was used Samples collected during the

2007 ¢eld survey were separated using stage records

from temporary monitors, stage board readings at each

individual site and ¢eld team notes As the £ow of the

River Awe was regulated, separation of the

microbiolo-gical data into base-£ow and rainfall-impacted

high-£ow categories is not applicable Therefore, all data

from each survey period for site 212 were used to

cal-culate a geometric mean (GM) for each FIO, which was

used to characterize water quality during the periods

assigned to both base £ow and high £ow

For the purposes of statistical analyses, samples

where no organisms were detected were recorded as

the detection limit value The distribution of

micro-bial concentrations found in stream and sewage

e¥u-ent samples, taken under base-£ow and high-£ow

conditions, showed a closer approximation to

nor-mality when log10transformed All microbial

con-centration data were, therefore, log10 transformed

before statistical analysis TheSPSSstatistical

compu-ter software package (SPSS2002) was used for cal analyses Descriptive statistics were used tocharacterize the distribution of bacterial concentra-tions at each sampling location These statistics in-clude the GM, calculated as the antilog of the mean

statisti-of log10-transformed concentrations, the standarddeviation of log10-transformed concentrations, the95% con¢dence interval for the mean and the range

of values at each site The signi¢cance of di¡erencesbetween GM concentrations was examined usingStudent’s t-test to compare the means of log10-trans-formed concentrations The methodology included Le-vene’s test for homogeneity of variances to determinewhether or not to use a t-test assuming equal or un-equal variances in the two groups compared with thehypothesis of equal variance being rejected at Po0.05.All statistical tests were assessed ata 5 0.05 (i.e 95%con¢dence level or 5% signi¢cance level)

Faecal indicator organism £ux under base-£owand high-£ow conditions was estimated as the pro-duct of discharge and GM faecal indicator concentra-tion for each site, i.e the base-£ow and high-£ow GMfaecal indicator concentration for each source wasmultiplied by the appropriate base-£ow and high-

£ow discharge volume for the budget period (de¢ned

as the 816-h period between 7 July 2006 and 10 gust 2006) The total load for each budget was calcu-lated as the sum of the loads from each sourcecontributing to the lower basin of Loch Etive Di¡er-ent budget scenarios were constructed, including es-timates of the situation during the ¢eld study periodsplus a hypothetical scenario to investigate the poten-tial impact of remediation measures within catch-ments identi¢ed as contributing a disproportionate

Au-£ux of FIOs Because of the large amount of tion generated by this project, only FC results are pre-sented throughout, given that coliforms are thecompliance parameter for shell¢sh waters Wherevariations in EN results were present, these are de-scribed in the text A full description of the resultsfrom this study is available in Magill et al 2008

informa-ResultsRiver and sewage e¥uent discharge

A total volume of 4.9 107

m3was discharged ing the 2006 study period into the lower basin ofLoch Etive from the studied catchments and sewagesources (Table 2) Di¡use catchment sources (i.e therivers) accounted for virtually the entire dischargebudget (99.99%) Of this, 78% was discharged duringMicrobial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

dur-base-£ow events, although this distribution was

skewed somewhat by the regulated £ow from the River

Awe (site 212), which contributed 88% of the total £ow

discharged during the study period (4.3 107m3, of

which 83% designated as base £ow; Table 2) and

accounted for 93% of the base-£ow discharge volume

(Fig 5) During high-£ow conditions, the River Awe

ac-counted for a lower proportion of the £ow (71%), with

the other rivers contributing a larger proportion of the

freshwater/e¥uent input to the loch than they did

during base-£ow conditions (Table 2, Fig 5) The next

largest inputs were the River Esragan (site 203) and

the River Nant (site 211), both of which accounted for

approximately 2.7% of the £ow into the loch (Table 2,Fig.5).The smallest £ow was from Rubha nan Carn (site222), which discharged and estimated 3.2 103m3during the study period

Only two of the gauged rivers discharged a greaterproportion of their £ow during high-£ow conditions:Blacreen Burn (site 204; 55% during high-£owevents, over a period of 125 h ^ 15% of the study peri-od) and the River Luachragan (site 209; 54% duringhigh-£ow events, over a period of 140 h ^ 17% of thestudy period) The remaining three monitored riversdischarged a greater volume during base-£ow condi-tions (Table 2) Consequently, only those rivers whose

Table 2 Estimated discharge budget (m 3 ) and percentage contribution to Loch Etive from sampled catchments for the period 7/7/06 and 10/8/06

Discharge volume (m 3 ) % Contribution to discharge budget Base flow High flow Total flow Base flow High flow Total flow

220 Achnacloich Plant Burnz 5.86  10 3 4.51  10 3 1.04  10 4 0.01 0.01 0.02

222 Str to Rubha nan Carnz 3.20  10 3 2.46  10 3 5.66  10 3 0.01 0.01 0.01

224 Allt Tig Dhonnchaidh 6.13  10 4 5.50  10 4 1.16  10 5 0.13 0.11 0.24

Flow estimate based on:

Inion Farm rain and Blacreen Burn £ow.

wBlacreen Burn rain and £ow.

zBlacreen Burn rain and Inion Farm stream £ow.

‰Culnadalloch rain and R Luachragan £ow.

zCulnadalloch rain and Allt na h-Airde £ow.

kTaynuilt WwTW rain and R Luachragan £ow.

Inion Farm rain and £ow.

Aquaculture Research, 2011, 42, 1^20 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al.

r 2010 The Authors

£ow estimates were based on either Blacreen Burn or

the River Luachragan were estimated to discharge a

greater volume during high-£ow conditions than

during base-£ow conditions (i.e Abhainn Achnacree

(site 201), the River Esragan (site 203), Lusragan Burn

(site 206) and the River Nant (site 211; Table 2)

The majority of treated sewage from Taynuilt

WwTW (3343 m3; 74%) was discharged during

base-£ow conditions, which prevailed for 92% (754 h) of

the study period The CSO at Taynuilt WwTW spilled

an estimated volume of 80 m3over a period of 28 h

(3.4% of the study period) (Table 2) This volume of

sewage from Taynuilt WwTW input to Loch Etive

(0.01%) was insigni¢cant compared with that fromthe studied rivers

Results of FIO analysisBroadly, FC concentrations (Fig 6) were around anorder of magnitude higher than EN concentrationsduring both years’ surveys GM FIO concentrations

in river samples generally showed a statistically ni¢cant elevation, often in excess of an order of mag-nitude, during high-£ow compared with base-£owconditions (Fig 6) Similar elevations in GM faecal in-dicator concentrations during high-£ow conditions

Sewage overflows

BASE FLOW HIGH FLOW Total discharge budget showing contribution of base flow and high flow sources

Base flow and high flow discharge budgets showing contribution of each source

Discharge budget: 3.81 x 10 7 m 3 Discharge budget: 1.06 x 10 7 m 3

Figure 5 Discharge budget for Loch Etive for the study period (816 h 7 July 2006 to 10 August 2006): (a) total budget split

by major source; (b) base-£ow and high-£ow budgets (split by £ow conditions)

Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

following rainfall events have been reported in

pre-vious CREH studies (Wyer et al 1996, 1998; Wyer, O’

Neill, Kay, Crowther, Jackson & Fewtrell1997; Kay,Wyer,

Crowther, Stapleton, Bradford, et al 2005; Kay, Wyer,

Crowther, Stapleton, Wilkinson, et al 2005; Wither

et al 2005; Stapleton et al 2008) They are attributable

to a combination of increased surface runo¡, the

exten-sion of the stream network into the contributing areas

enhancing ‘connectivity’ and entrainment of

organ-isms from streambed sources, all of which increase

the numbers of organisms entering watercourses

Dur-ing hydrograph events, increased stream £ow velocities

and turbidity act to reduce the opportunities for die-o¡

(through exposure to UV light) and sedimentation

while enhancing transportation of microorganisms

Three catchments were notable for particularly

high FIO concentrations at their tidal limits: Abhainn

Achnacree (site 201), Inion Farm stream (site 202)

and Kenmore Bay stream (site 206) The high

concen-trations in Abhainn Achnacree appear to be

im-pacted by elevated levels in one of its tributaries, Alltnam Ban (site 214), while concentrations at upstreamsample sites on Inion Farm stream and KenmoreBay stream were lower than those at the tidal limits(Fig 6), suggesting that the source of pollution wasbetween the upper and the lower sample points oneach stream The base-£ow EN concentrations at thetidal limit of Kenmore Bay stream during 2006(6.1 104CFU 100 mL 1) and Inion Farm streamduring 2007 (4.4 104CFU 100 mL 1) were ofparticular note, being over an order of magnitudegreater than the next highest concentration duringeach respective sampling period (Magill et al 2008).The unusually high base-£ow GM concentrations atthe tidal limits of these catchments, given their landcover, explains the lack of a statistically signi¢cantdi¡erence between base-£ow and high-£ow FIOconcentrations shown in Fig 6

The combined GM values used for the River Awe(i.e all sample data used to calculate a GM for each

Aquaculture Research, 2011, 42, 1^20 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al.

r 2010 The Authors

year irrespective of £ow) were lower during the 2007

survey than during 2006 (Fig 6) The di¡erence

between the 2006 and 2007 EN GMs was statistically

signi¢cant, although not for FC These very low

concentrations and year-on-year changes are

prob-ably due to management and regulation of the £ow

from Loch Awe

Geometric mean concentrations of FIOs in the

Tay-nuilt WwTW secondary-treated FE showed

statisti-cally signi¢cant dilution during high-£ow eventsduring the 2006 ¢eld survey period (Fig 7) This waspossibly due the fact that the works was operating ato50% capacity and a relatively high proportion ofsurface water was entering the sewerage systemduring high-£ow conditions compared with the foulsewage content, thus diluting the in£uent to theworks However, data from the 2007 ¢eld surveyperiod showed the more commonly observed pattern

301 Taynuilt WwTW Final Effluent 302 Taynuilt WwTW CSO effluent

2007 ¢eld surveys

Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

of increased GM concentrations during high-£ow

events (Fig 7) Both the treated FE and the CSO

e¥u-ent from Taynuilt WwTW displayed FIO conce¥u-entra-

concentra-tions lower than similar e¥uents sampled by CREH

in previous studies (Kay, Crowther, Stapleton, Wyer,

Fewtrell, Edwards, Francis, McDonald et al 2008)

FIO budgets

‘Current’ Loch Etive budget

This section describes budgets comprising all the

catchment outlets sampled during the 2006 and

2007 ¢eld study periods: 24 riverine inputs, the

Taynuilt WwTW FE and the Taynuilt WwTW CSO

e¥uent It was not possible to include an estimate of

the £ux from Connel CSO as no £ow data were

avail-able for this input Two estimates of the £ux to the

lower basin were made to accommodate the di¡erent

results observed for sites sampled during both ¢eld

study periods [i.e Inion Farm stream (site 202), River

Nant (site 211), River Awe (site 212) and Taynuilt

WwTW Final and CSO e¥uents (sites 301 and 302

respectively)] These budgets are referred to below as

the ‘2006’ and ‘2007’ budgets, although it should be

noted that both of these £ux estimates used the same

£ow data, which was characterized by the £ows

measured during the 2006 study period, and that

each budget contains data from either 2006 or 2007

for the inputs sampled only during one of the two

¢eld study periods

The £ux of FC and EN organisms discharged into

the lower basin of Loch Etive from the 24 catchments

and Taynuilt WwTW estimated using 2006 data for

sites sampled during both survey periods was greater

than those using 2007 data The ‘2006 budget’

showed a slightly greater load was input under

high-£ow conditions, accounting for 53% of the FC load

(Fig 8a) and 54% of the EN load, although this

pro-portion was higher for FC in the ‘2007’ budget,

con-tributing 62% of the FC load (Fig 8b) The

proportional contribution of EN to the ‘2007 budget’

during high £ows remained at 54% despite the actual

£ux of organisms being lower (Magill et al 2008) The

almost equal proportions of base-£ow and high-£ow

FIO delivery for most of the budget estimates was

somewhat at variance with the ¢ndings of previous

budget studies of other catchments (e.g Wyer et al

1996, 1997, 1998; Crowther, Kay & Wyer 2002;

Crowther, Wyer, Bradford, Kay & Francis 2003; Kay,

Wyer, Crowther, Stapleton, Bradford, et al 2005; Kay,

Wyer, Crowther, Stapleton, Wilkinson, et al 2005;

Stapleton et al 2008), which were generally nated by high-£ow delivery of FIOs from riverinesources However, none of these previous studies in-cluded the input from an impounded and regulatedinput such as the Awe

domi-The majority (o95%) of the total FIO load inputinto the estuary during the study period was deliveredfrom di¡use catchment sources (i.e the rivers), whichaccounted for 95^98% of the FC budget (Fig 7)and 95^98% of the EN budget (Magill et al 2008).The largest contribution was derived from the RiverAwe, which accounted for up to 37% of the FC and

EN loads estimated using 2006 data, although a

low-er proportion (up to 25%) was estimated if using

2007 data This contribution was a function of themuch greater discharge of the River Awe, whichaccounted for 88% of the freshwater input into thelower basin of the loch (Table 2) However, its FIO con-centrations in this source were some of the lowestobserved in the studied catchments (Fig 6) Thissource dominated FIO delivery during base-£ow per-iods (Fig 8), accounting for up to 95% of the instanta-neous £ux of organisms, although its importancewas diminished during high-£ow events, when in-puts from other catchments were dominant (Fig 9).Relatively large contributions to the FIO load dis-charged to the lower basin of Loch Etive during thestudy period were derived from the catchments ofAbhainn Achnacree (FC: Base £ow 15^23%, high

£ow 9^10%; EN: high £ow 11^15%), Inion Farmstream using 2006 data (FC: high £ow 9%; EN: high

£ow 27%) and Kenmore Bay stream (EN: base £ow11^17%; high £ow 10^15%) (Fig 8) The proportionalcontribution of these sources to the instantaneousdelivery of organisms in the two budget estimatespeaked at 53% for FC in Abhainn Achnacree and34% for FC (54% for EN) in Inion Farm stream (Fig.9) The instantaneous proportional contribution of

FC from Kenmore Bay stream was low (Fig 9a),although the high EN concentration noted in Section3.2.2 resulted in a contribution of up to 39% for EN(Fig 9b) This illustrates the counterintuitive impor-tance of small catchment sources and episodic pollu-tion £uxes in the overall faecal indicator budget.Other relatively large contributors to the FIO loaddischarged to Loch Etive were Lusragan Burn, whichcontributes a relatively high proportion to both thebase-£ow and the high-£ow instantaneous £ux oforganisms, and the River Esragan, which is an im-portant source later during high-£ow events whensome of the more ‘£ashy’ streams have reverted tobase £ow (Fig 9)

Aquaculture Research, 2011, 42, 1^20 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al.

r 2010 The Authors

‘2006 Budget’ Total Flow ‘2007 Budget’ Total Flow

‘2006 Budget’ High Flow

2.84 x 10 organisms (53.8% of 2006 total load)

‘2007 Budget’ High Flow 2.54 x 10 organisms (62.9% of 2007 total load)

WwTW CSO: 0.2%

23.4%

2.0%

203: 0.3% 205: 0.5% 11.2%

61.9%

Base flow and high flow faecal coliform budgets showing contribution of each source

Total faecal coliform budgets showing contribution of base flow and high flow sources

203 River Esragan 209 River Luachragan Other Rivers

Figure 8 Faecal coliform budgets for Loch Etive estimated using either 2006 or 2007 data for Inion Farm stream, R Nant,

R Awe and Taynuilt WwTW ¢nal e¥uent and combined sewage over£ow: (a) total budgets split by major source; (b)

base-£ow and high-base-£ow budgets (split by base-£ow conditions)

Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

Treated sewage e¥uent from Taynuilt WwTW

ac-counted for only 3.9% of the FC budget and 5.2% of

the EN budget using the 2006 data and only 1.1%

and 2% of the respective budgets using 2007 data

where possible (Fig 8) The majority of the load in

the ‘2006 budget’ was delivered during base-£ow

conditions, when FIO concentrations in the e¥uent

were generally higher than that during high-£ow

periods However, the data from 2007 showed an

in-crease in FIO concentrations during high-£ow

events, with a greater proportion of the FE total load

in the ‘2007 budget’ discharged during high-£ow

events Although the overall contribution of the FE

was relatively small, the two alternative budgets

con-structed for the current situation showed that the

proportion of the instantaneous £ux of FIOs

repre-sented by the FE peaked at 43% during the study

per-iod (2007 budget) The CSO at Taynuilt WwTW

accounted foro1% of the total FIO input into the

loch, although its instantaneous contribution

reached as high as 9% when it is discharging (Fig 9)

Impact of reducing FIO concentrations in selected

catchments

The analysis of the empirical ¢eld data above

high-lighted three catchments with particularly high FC

and/or EN GM concentrations during either or both

base-£ow and high-£ow conditions during the 2006study period Subsequent sampling upstream of po-tential sources demonstrated that each of the catch-ments displayed lower GM concentrations, often by

an order of magnitude or more In the case of hainn Achnacree (site 201), the tributary Allt namBan (site 214) displayed high GM concentrationswhile before their con£uence, Abhainn Achnacreehad lower GM concentrations The tidal limit site onInion Farm stream (site 202) displayed GM concen-trations an order of magnitude higher than upstream

Ab-of the farm (site 215), while upstream Ab-of ArdchattanSchool Kenmore Bay stream (site 216) displayed FCand EN concentrations two to three orders of magni-tude lower than the tidal limit site (site 205) InionFarm stream (site 202) was sampled again duringthe 2007 study and the data con¢rms the input asone of poorest quality streams entering Loch Etive.The budgets presented in Section 3.3.1 have also de-monstrated that these catchments contribute a sig-ni¢cant proportion of the FC and EN loads despitetheir low contribution to the discharge budget.Clearly, therefore, there are sources of FIOs withinthese catchments that could be investigated for reme-diation to bring the quality of these inputs into linewith others included in this study

To investigate the potential impact of remediationmeasures within these three catchments, further

Figure 9 Hourly rainfall (mm) at Inion Farm (site 202), instantaneous load (organisms s 1) and proportional tion (%) of organisms to the hourly load input to Loch Etive estimated using 2006 data for Inion Farm stream, R Nant, R.Awe and Taynuilt WwTW ¢nal e¥uent and combined sewage over£ow: (a) faecal coliforms; (b) enterococci

contribu-Aquaculture Research, 2011, 42, 1^20 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al.

r 2010 The Authors

budgets were calculated for the 2006 study period

re-placing the tidal limit GM concentrations of Abhainn

Achnacree (site 201), Inion Farm stream (site 202)

and Kenmore Bay stream (site 205) with the lower

concentrations observed upstream of the potential

sources (i.e sites 213, 215 and 216; Fig 6) For this

budget, FIO concentrations in the Rivers Nant and

Awe and Taynuilt WwTW FE and CSO e¥uent (sites

211, 212, 301 and 302) were characterized by data

from the 2006 study period because data for the

up-stream sites used for the Abhainn Achnacree, Inion

Farm Stream and Kenmore Bay stream were also

col-lected during 2006 All discharge volumes and GM

concentrations in the remaining sources were

un-changed from those described for the ‘2006 budget’

described in Section 3.3.1 All comparisons below are

performed between the amended budget and the

‘2006’ budget described above

Intervention in the Abhainn Achnacree, Inion

Farm stream and Kenmore Bay stream catchments

could potentially result in a 14% reduction in the FC

load and a 31% reduction in the EN load Riverine

in-puts still dominated the FC and EN budgets, although

a reduction in the riverine £ux consequently resulted

in a slight increase in the proportional contribution

of Taynuilt WwTW FE (to 4.6% of the FC load and

7.5% of the EN) While the delivery of FC in the

ad-justed budget was dominated by high-£ow riverine

inputs, as was the case in the original ‘2006’ budget,

the greatest proportion of EN was delivered during

base-£ow conditions rather than during high-£ow

conditions (Magill et al 2008)

During base-£ow conditions, the proportional

load from Abhainn Achnacree decreased to represent

only 0.1% of the adjusted budget (Fig 10)

com-pared with 14.3% in the original ‘2006’ budget The

base-£ow EN load from Kenmore Bay stream

decreased to represent o0.1% of the adjusted

budget compared with 11.4% of the original budget

During high-£ow conditions, the proportional

con-tributions of Abhainn Achnacree, Inion Farm stream

and Kenmore Bay stream decreased, respectively,

to 2.4%, 3.0% ando0.1% of the adjusted high-£ow

FC budget (Fig 10: down from 8.7%, 8.5% and 0.4%

of the original ‘2006’ high-£ow FC budget, Fig 8)

and 2.6%, 2.6% and o0.1% of the adjusted

high-£ow EN budget (down from 10.5%, 27.4% and

10.1% of the original high-£ow EN budget; Magill

et al 2008) A consequence of the reduction in

loads from these three streams was the increase

in the proportional contribution of the other sources

(Fig 10)

Discussion and conclusion

This QMSA study investigated FIO budgets from age and riverine sources draining to the lower basin

sew-of Loch Etive above the Falls sew-of Lora, within whichshell¢sh production sites were undergoing variation

in E coli levels To better understand this variation,and identify potential contributing sources to theFIO load of the loch, concentrations for both base-

£ow and rainfall-induced high-£ow conditions

with-in the selected study catchments were ascertawith-ined.Gathering such data requires an intensive e¡ort invol-ving sampling teams located close to the study areaand ready to respond to rainfall 24 h a day, so that theepisodic conditions, often missed in routine monitor-ing programmes, are adequately characterized The re-sults of this and similar studies carried out by theauthors highlight the importance of high-£ow events

in FIO £ux estimation Over half of the FIO £ux inputinto the lower basin of Loch Etive was delivered duringhigh-£ow periods, which accounted for only 11^15%

of the 2006 study period upon which £ow volumeswere based This presents a signi¢cant problem forthe use of data derived from routine monitoring,which are systematically biased towards base-£owconditions This can lead to the erroneous appraisal

of catchment-derived faecal indicator £uxes from bothdi¡use and point discharges and could result in inap-propriate expenditure decisions

The results described above have shown that smallcatchments discharging a very small proportion ofthe total freshwater input to the loch can contributerelatively high proportions of FIOs during both base-

£ow and high-£ow conditions, exceeding those inputfrom sewage sources Sampling within some of thesecatchments showed that lower concentrations ex-isted upstream or highlighted potentially contami-nated tributaries that could be potential targets forremediation measures An assessment of the impact

of potential remediation within three relatively smallcatchments suggested that the FC £ux could be re-duced by 14% and the EN £ux could be reduced by31% Given the absence of consented point sourceswithin these catchments, measures to reduce FIO in-puts should concentrate on di¡use agriculturalsources Potential measures include the fencing ofrivers to prevent livestock access, collection andtreatment of runo¡ from agricultural farm buildingroofs and hardstanding areas, relocation of farmyardmanure heaps away from ¢eld edges where these areclose to drainage ditches, streams, etc., maintenance

of bu¡er strips along river corridors and avoidance ofMicrobial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

spreading manures and slurry during wet weather or

when the soil is saturated Additionally, attention

should be given to closing potential £ow pathways

between sources such as farm hardstandings and

watercourses provided by farm tracks etc The

e¡ec-tiveness of such remediation measures (sometimes

called best management practices) have been

investi-gated by the present team and scienti¢c partners in

Scotland (Dickson, Edwards, Je¡rey & Kay 2005; Kay,

Wyer, Crowther, Stapleton,Wilkinson, et al 2005; Kay,

Aitken et al 2007; Edwards & Kay 2008)

Neverthe-less, it is possible that septic tanks serving individual

properties may also be contributing to the load

with-in these catchments

Despite displaying some of the lowest FIO trations observed during this study, the largest con-tributor of FIOs to the lower basin of Loch Etive wasthe River Awe, due to the fact that it dominated thein£ux of freshwater (88% of the total freshwater in-put from the studied catchments) The majority ofthis load was delivered during base-£ow conditionswhen the instantaneous delivery of organisms to theloch was relatively low The impoundment and regu-lation of £ows within this river means that it does not

4.56 x 10 14 organisms (86.4% of unadjusted estimate)

2.08 x 10 organisms (45.8% of adjusted total) 2.47 x 10 organisms (54.2% of adjusted total)

BASE FLOW HIGH FLOW

WwTW FE:

High Flow: 0.1%

Sewage overflows:

High Flow: 0.6%

2006 total faecal coliform budget showing contribution of base flow and high flow sources

2006 base flow and high flow faecal coliform budgets showing contribution of each source

Figure 10 Estimated faecal indicator organism budgets for Loch Etive adjusted to re£ect water quality upstream of tential FIO sources: (a) total budgets split by major source; (b) base-£ow and high-£ow budgets (split by £ow conditions)

po-Aquaculture Research, 2011, 42, 1^20 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al.

r 2010 The Authors

respond to rainfall events like other rivers and the low

FIO concentrations were probably due to microbial

die-o¡ within Loch Awe and the relatively constant

£ow Consequently, management options for reducing

the load from this source are probably limited

Clearly, FIO source apportionment, as described

above, can only indicate the proportions of di¡erent

FIO contributions to the water body as a whole, and

further to enhance the utility of these data, it would

be prudent to establish connectivity between the key

hydrological inputs and the near-shore ¢shery

waters This would facilitate more precise targeting

of the remedial measures described above An e¡ective

method of achieving this would be to use microbial

tracers (chie£y exotic endospores or bacteriophage)

These are advantageous over other tracer media such

as dyes (e.g rhodamine WT) or arti¢cial (normally

plastic) particles due to the following: their similarity

to the microbial ‘contaminants’; their availability at

high initial tracer concentrations with extreme

sensi-tivity of detection; their appropriate survival times in

the environment; and their innocuousness with low

environmental impact Furthermore, microbial

tra-cers may also act as surrogates for other

microorgan-isms, such as human pathogenic viruses in the case

of phage (Joyce, Rueedi, Cronin, Pedley, Tellam &

Greswell 2007) and, importantly, phage tracers can

also be isolated from shell¢sh £esh (West & Wipat

1988) Microbial tracers have been used successfully

in high-dilution surface water environments such as

estuaries (Morgan, Munro, Mollowney & Linwood

1995) and the marine environment (Pike, Bufton &

Gould 1969; Drury & Wheeler 1982; Richardson,

Charlton, Currie, Ashton & Lowther 1993; Paul, Rose,

Brown, Shinn, Miller & Farrah 1995; Paul,

McLaugh-lin, Gri⁄n, Lipp, Stokes & Rose 2000) and recently

applied to bathing and shell¢sh water compliance

is-sues in Guernsey and South Wales by the authors

(Davies, Kay, Kay, McDonald, Moore, Stapleton,

Wat-kins & Wyer 2008; Wyer, Kay, Stapleton,WatWat-kins,

Da-vies & Moore 2009) Such a study might be applicable

to the key inputs identi¢ed herein or alternatively to

inputs outside the lower basin of the loch, not

in-cluded in the source apportionment study, such as

the sewage input into Dunsta¡nage Bay and the tidal

£ux from the upper basin of Loch Etive

One other potential source of FIOs not considered

within the study is from birds Estimation of such an

input is di⁄cult, even when bird count data are

avail-able, because data describing daily loads are available

only for certain species and estimates have

consider-able variability (e.g Gould & Fletcher 1978; Jones &

Obiri-Danso 1999; Fogarty, Haack, Wolcott & man 2003) Thus, an accurate quantitative represen-tation of the total avian budget will not be possibleuntil more empirical data are available While it islikely that avian inputs may be a signi¢cant contribu-tor to FIO concentrations in shell¢sh waters, remedia-tion options for this input are limited and this sourceshould, perhaps, be viewed as part of the backgroundload of the loch

Whit-AcknowledgmentsThe authors would like to thank the following peoplefor their contribution to the work reported herein:SARF: Mark James,Walter Speirs; Scottish Associationfor Marine Science: Kenny Black; Scottish Environ-ment Protection Agency: Anne Henderson; ScottishSouthern Energy: Mike Cruickshank; Scottish Water:Duncan Cambell, Martin Dunne and Stephan Walker;CREH: Paula Hopkins (logistics), Daniel Bennett, TomChibnall,Tom Hilton and Katie Whiting (¢eld team)

ReferencesAnonymous (1976) Council of the European Communities Council Directive 76/160/EEC of 8 December 1975 con- cerning the quality of bathing water O⁄cial Journal of the European Communities L31, 1^7.

Anonymous (1979) Council Directive 79/923/EEC of 30 tober 1979 on the quality required of shell¢sh waters Of-

Oc-¢cial Journal of the European Communities L284, 47^52 Anonymous (1991) Council Directive of 15 July 1991 laying down the health conditions for the production and pla- cing on the market of live bivalve molluscs (91/492/EEC) O⁄cial Journal of the European Communities L268, 1^14 Anonymous (1997) The Surface Waters (Shell¢sh) (Classi¢ca- tion) Regulations 1997 Statutory Instrument 1997 No.

1332 Queen’s Printer of Acts of Parliament Available at http://www.opsi.gov.uk/si/si1997/19971332.htm (accessed 8 September 2010).

Anonymous (2000) Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 estab- lishing a framework for Community action in the ¢eld of water policy O⁄cial Journal of the European Communities L327, 1^77.

Anonymous (2002) Federal Water Pollution Control Act [As Amended Through P.L 107-303, November 27, 2002] USE-

PA, Washington, DC, USA Available at http://www.epa gov/lawsregs/laws/cwa.html.

Anonymous (2004) Regulation (EC) 854/2004 of the opean Parliament and of the Council of 29 April 2004 lay- ing down speci¢c rules for the organisation of o⁄cial controls on products of animal origin intended for human consumption O⁄cial Journal of the European Communities L226, 83^127.

Eur-Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

Anonymous (2006a) Council Directive 2006/113/EC of 12

December 2006 on the quality required of shellsh waters

(codied version) O⁄cial Journal of the European Union

L376, 14^20.

Anonymous (2006b) Directive 2006/7/EC of The European

Parliament and of The Council of 15th February 2006

concerning the management of bathing water quality

and repealing Directive 76/160/EEC O⁄cial Journal of the

European Union L64, 37^51.

Crowther J., Kay D & Wyer M (2002) Faecal-indicator

con-centrations in waters draining lowland pastoral

catch-ments in the UK: relationships with land use and

farming practice.Water Research 36, 1725^1734.

Crowther J., Wyer M.D., Bradford M., Kay D & Francis C.A.

(2003) Modelling faecal indicator concentrations in large

rural catchments using land use and topographic data.

Environmental Microbiology 94, 962^973.

Davies C., Kay D., Kay C., McDonald A., Moore H., Stapleton

C., Watkins J & Wyer M (2008) Microbial tracer study of

selected inputs into the Guernsey coastal zone A report to

the States of Guernsey, Public Services Department.

CREH, 26pp Available at

http://www.gov.gg/ccm/cms-service/download/asset/?asset_id=11226081 (accessed 8

September 2010).

Dickson J.W., Edwards A.C., Je¡rey B & Kay D (2005)

Catch-ment Scale Appraisal of Best ManageCatch-ment Practices (BMPs)

for the Improvement of Bathing Water ^ Brighouse Bay,

Edinburgh SAC Environmental, Auchincruive, Scotland,

UK Available at http://www.scotland.gov.uk/Topics/

Environment/Water/bathingwaters/BrighouseBay.

Drury D.F & Wheeler D.C (1982) Applications of a Serratia

mar-cescens bacteriophage as a new microbial tracer of aqueous

environments Journal of Applied Bacteriology 53, 137^

142.

Edwards A.C & Kay D (2008) Farmyards an overlooked

source of highly contaminated runo¡ Journal of

Environ-mental Management 87, 551^559.

Environment Agency (EA) (2003a) Water quality consenting

guidance Consenting discharges to achieve the requirements

of the Shell¢shWaters Directive (Microbial Quality) Agency

Management System Document, Number 169_01 V2.

Environment Agency (EA) (2003b) Hydrometric Manual

En-vironment Agency, Bristol, UK.

Fogarty L.R., Haack S.K., Wolcott M.J & Whitman R.L.

(2003) Abundance and characteristics of the recreational

water quality indicator bacteria Escherichia coli and

enter-ococci in gull faeces Journal of Applied Microbiology 94,

865^878.

Gould D.J & Fletcher M.R (1978) Gull droppings and their

e¡ects on water quality.Water Research 12, 665^672.

Jones K & Obiri-Danso K (1999) Non-compliance of beaches

with the EU directives of bathing water quality: evidence

of non-point sources of pollution in Morecambe Bay

Jour-nal of Applied Microbiology 85, 101S^107S.

Joyce E., Rueedi J., Cronin A., Pedley S.,Tellam J & Greswell

R (2007) Fate and transport of phage and viruses in UK

Permo-Triassic sandstone aquifers Science report ^ SC030217/SR, Environment Agency, 81pp.

Kay D., Bartram J., Pruss A., Ashbolt N.,Wyer M.D., Fleisher J.M., Fewtrell L., Rogers A & Rees G (2004) Derivation of numerical values for the World Health Organization guidelines for recreational waters Water Research 38, 1296^1304.

Kay D., Wyer M.D., Crowther J., Stapleton C., Bradford M., McDonald A.T., Greaves J., Francis C & Watkins J (2005) Predicting faecal indicator £uxes using digital land use data in the UK’s sentinel water framework directive catch- ment: the Ribble study.Water Research 39, 655^667 Kay D., Wyer M.D., Crowther J., Stapleton C., Wilkinson J & Glass P (2005) Sustainable reduction in the £ux of micro- bial compliance parameters to coastal bathing waters by

a wetland ecosystem produced by a marine £ood defence structure.Water Research 39, 3320^3332.

Kay D., Ashbolt N., Wyer M.D., Fleisher J.M., Fewtrell L., gers A & Rees G (2006) Comment on‘Derivation of numer- ical values for the World Health Organization guidelines for recreational waters’A Reply.Water Research 40, 1921^ 1926.

Ro-Kay D., McDonald A.T., Stapleton C.M.,Wyer M.D & Fewtrell

L (2006) Europe: a challenging new framework for water quality Proceedings of the Institution of Civil Engineers, Water Management 159, 58^64.

Kay D., Aitken M., Crowther J., Dickson I., Edwards A.C., Francis C., Hopkins M., Je¡rey W., Kay C., McDonald A.T., McDonald D., Stapleton C.M., Watkins J., Wilkinson J & Wyer M (2007) Reducing £uxes of faecal indicator com- pliance parameters to bathing waters from di¡use agri- cultural sources, the Brighouse Bay study, Scotland Environmental Pollution 147, 139^149.

Kay D., Edwards A.C., Ferrier R.C., Francis C., Kay C., Rushby L., Watkins J., McDonald A.T., Wyer M., Crowther J & Wilkinson J (2007) Catchment microbial dynamics: the emergence of a research agenda Progress in Physical Geography 31, 59^76.

Kay D., Crowther J., Stapleton C.M., Wyer M.D., Fewtrell L., Anthony S.G., Bradford M., Edwards A., Francis C.A., Hopkins M., Kay C., McDonald A.T., Watkins J & Wilkin- son J (2008) Faecal indicator organism concentrations and catchment export coe⁄cients in the UK Water Research 48, 2649^2661.

Kay D., Crowther J., Stapleton C.M., Wyer M.D., Fewtrell L., Edwards A., Francis C.A., McDonald A.T., Watkins J & Wilkinson J (2008) Faecal indicator organism concentra- tions in sewage and treated e¥uents.Water Research 42, 442^454.

Kay D., Kershaw S., Lee R.,Wyer M.D.,Watkins J & Francis C (2008) Results of ¢eld investigations into the impact of in- termittent sewage discharges on the microbiological quality of wild mussels (Mytilus edulis) in a tidal estuary Water Research 42, 3033^3046.

Magill S., Black K., Kay D., Stapleton C., Kershaw S., Lees D., Lowther J., Francis C., Watkins J & Davies C (2008) Risk Aquaculture Research, 2011, 42, 1^20 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al.

r 2010 The Authors

factors in shell¢sh harvesting areas Final Project Report

prepared for Scottish Aquaculture Research Forum.

SARF013/SAMS report No 256 Scottish Association for

Marine Science, Centre for Research into Environment

and Health and CEFAS Available at http://www.sarf.

org.uk/Project Final Reports/SARF013 - RISK_FACTORS_

IN_SHELLFISH_HARVESTING_AREAS - 23June2008.pdf

(accessed 8 September 2010).

McDonald A.T., McDonald D., Kay D & Watkins J (2008)

Characteristics and signi¢cance of liquid e¥uent from

woodchip corrals in Scotland Journal of Environmental

Management 87, 582–590.

Morgan N., Munro D., Mollowney B.M & Linwood P (1995)

Sewage dispersion in an estuary with distinct ebb and

£ood channels Environment International 21, 123^130.

Paul J.H., Rose J.B., Brown J., Shinn E.A., Miller S & Farrah S.R.

(1995) Viral tracer studies indicate contamination of marine

waters by sewage disposal practices in Key Largo, Florida.

Applied and Environmental Microbiology 61, 2230^2234.

Paul J.H., McLaughlin M.R., Gri⁄n D.W., Lipp E.K., Stokes R.

& Rose J.B (2000) Rapid movement of wastewater from

on-site disposal systems into surface waters in the lower

Florida Keys Estuaries 23, 662^668.

Pike E.B., Bufton W.J & Gould D.J (1969) The use of Serratia

indica and Bacillus subtilis var niger spores for tracing

sewage dispersion in the sea Journal of Applied

Bacteriol-ogy 32, 206^216.

Richardson B.J., Charlton A.L., Currie S., Ashton P & Lowther

I.M (1993) Tracer studies using bacteriophage to predict the fate

of viruses in the marine community: preliminary assessments.

Research Report, no.54, UrbanWater Research Association

of Australia Melbourne (plus ¢gures and tables), 24pp.

Rodgers P., Soulsby C., Hunter C & Petry J (2003) Spatial and

temporal bacterial quality of a lowland agricultural

stream inn northeast Scotland Science of the Total

Envir-onment 314^316, 289^302.

Standing Committee of Analysts (SCA) (2000) The

Microbiol-ogy of Recreational and Environmental Waters Methods for

the Examination of Waters and Associated Materials

Envir-onment Agency, Bristol, UK.

Standing Committee of Analysts (SCA) (2002) The

Microbiol-ogy of DrinkingWater (2002) ^ Part 4 - Methods for the

Iso-lation and Enumeration of Coliform Bacteria and Escherichia

coli (Including E coli O157:H7), Methods for the

Examina-tion of Waters and Associated Materials Environment

Agency, Bristol, UK.

Standing Committee of Analysts (SCA) (2006) The ogy of Drinking Water (2006) ^ Part 5 ^ Isolation and Enu- meration of Enterococci by Membrane Filtration, Methods for the Examination of Waters and Associated Materials En- vironment Agency, Bristol, UK.

Microbiol-Stapleton C.M.,Wyer M.D., Crowther J., McDonald A.T., Kay D., Greaves J., Wither A., Watkins J., Francis C.A., Hum- phrey N & Bradford M (2008) Quantitative catchment pro¢ling to apportion faecal indicator organism budgets for the Ribble system, the UK’s sentinel drainage basin for Water Framework Directive research Journal of Envir- onmental Management 87, 535^550.

USEPA (2009) National Summary of Impaired Waters and TMDL Information US Environmental Protection Agency, Washington, DC, USA Available at http://iaspub epa.gov/waters10/attains_nation_cy.control?p_report_type=T, West P.A & Wipat S.V (1988) Detection of bacteriophage tracers in bivalve molluscan shell¢sh Letters in Applied Microbiology 7, 95^98.

WHO (2003) Guidelines for Safe Recreational Water ments Volume 1: Coastal and Freshwaters World Health Organisation, Geneva, Switzerland.

Environ-Wither A., Greaves J., Dunhill I., Wyer M., Stapleton C., Kay D., Humphrey N., Watkins J., Francis C., McDonald A & Crowther J (2005) Estimation of di¡use and point source microbial pollution in the Ribble catchment discharging

to bathing waters in the north west of England Water Science and Technology 51, 191^198.

Wyer M., Kay D., Dawson H.M., Jackson G.F., Jones F.,Yeo J & Whittle J (1996) Delivery of microbial indicator organisms

to coastal waters from catchment sources Water Science and Technology 33, 37^50.

Wyer M., Kay D., Crowther J.,Whittle J., Spence A., Huen V., Wilson C & Carbo P.J.N (1998) Faecal-indicator budgets for recreational coastal waters: a catchment approach Journal of the Chartered Institution of Water and Environ- mental Management 12, 414^424.

Wyer M., Kay D., Stapleton C.,Watkins J., Davies C & Moore

H (2009) An investigation into water quality at Amroth, Pembrokeshire A report to Environment Agency Wales, Centre for Research into Environment and Health, Aber- ystwyth University, 24pp.

Wyer M.D., O’Neill G., Kay D., Crowther J., Jackson G & Fewtrell

L (1997) Non-outfall sources of faecal indicator organisms a¡ecting the compliance of coastal waters with directive 76/160/EEC.Water Science and Technology 35, 151^156 Microbial risk factors in Loch Etive shell¢sh waters C M Stapleton et al Aquaculture Research, 2011, 42, 1^20

Applied epidemiology with examples

from UK aquaculture

James F Turnbull1, Iain K Berrill1, Darren M Green1, Ryan Kaye1, David Morris1,

Alexander G Murray2, Jorge del-Pozo3& Andrew Shinn1

1 Institute of Aquaculture, University of Stirling, Stirling, Scotland, UK

2 Marine Scotland, Marine Laboratory, Aberdeen, UK

3 Easter BushVeterinary Centre, Roslin, Midlothian, UK

Correspondence: J F Turnbull, Institute of Aquaculture, University of Stirling, Stirling, Scotland FK9 4LA, UK E-mail: jft1@stir.ac.uk

Abstract

This paper is a brief introduction to epidemiology and

its application to farmed ¢sh health and welfare with

examples from the United Kingdom Epidemiology

has the potential to do a great deal more than just

identify risk factors Indeed in many cases useful risk

factors cannot be identi¢ed due to the complexity of

the disease problems and the lack of resources

Epide-miological principles or analytical techniques have

been applied in animal welfare studies, and they can

reduce the cost of disease monitoring or surveillance

and disease control However, for epidemiological

studies to make a real contribution to farmed ¢sh

health and welfare it is often necessary to use

multi-disciplinary teams, obtain good data and coordinate

e¡orts on the major problems

Keywords: aquaculture, disease, epidemiology,

health, welfare

Introduction

Epidemiology is a term that describes the study of

pat-terns in populations; these can be patpat-terns of health,

disease, welfare or even crime in human populations

(Loeber & Farrington 2001) The term although

ori-ginally referring speci¢cally to human populations

is now widely used to describe the studies of

popula-tions of any species, avoiding the need for speci¢c and

clumsy terms such as epizootiology, epichthyology,

etc A population is composed of units, which are

not necessarily individual animals but can be any

group from ponds of ¢sh to farms or watersheds orcountries Other approaches to the study of diseasefocus at the individual animal level (e.g pathology,virology and immunology) or even below this on cells

or genes As will be discussed later it is frequently cessary to use combinations of techniques, whichwork at di¡erent levels The main aim of epidemiol-ogy is to improve health, productivity and welfarethrough understanding their determinants andcosts Although epidemiology can improve under-standing, like all scienti¢c disciplines, it requires anunderstanding of the farming system and involve-ment of stakeholders to turn this understanding intopractical strategies

ne-The behaviour of populations cannot necessarily

be predicted from the behaviour of its individual ponents One of the many examples of the complex orunpredictable nature of phenomena at the popula-tion level comes from a study of human in£uenza inJapan (Reichert, Sugaya, Fedson, Glezen, Simonsen &Tashiro 2001) It was found that vaccinating childrencould reduce the incidence of serious or fatal in£uen-

com-za in pensioners Although pensioners often su¡eredfrom the most severe in£uenza, the infection wascommonly spread between children who then in-fected their older relatives This conclusion could nothave been extrapolated from a study of individuals inisolation

Farmed ¢sh populations have several commoncharacteristics that de¢ne them and di¡erentiatethem from farmed terrestrial populations The num-ber of individual animals can be very large, withsmolt production units producing millions of ¢sh.Aquaculture Research, 2011, 42, 21^27 doi:10.1111/j.1365-2109.2010.02667.x

r 2010 The Authors

A single marine salmon cage would often hold up to

50 000 ¢sh, with recently introduced larger cages

holding many more In terrestrial populations

com-parable numbers are only seen in the poultry

indus-try Secondly, in many cases, populations are supplied

by water that is largely outwith the control of the

farmer, both in terms of quality and ingress of

patho-gens from wild ¢sh Thirdly, farmed ¢sh are often very

di⁄cult to observe in any detail without handling

them Not only do they live underwater, but also

nor-mal and harmless coloured water or large production

units (cages or tanks) exacerbate the problem

Epidemiology is based on relatively simple ideas,

such as comparing a¡ected and una¡ected groups to

identify the di¡erences between them These basic

concepts are now supported by complex

mathe-matics and statistics Like disease diagnosis,

epide-miology can also involve a considerable amount of

detective work Although epidemiologists are

fre-quently concerned with infectious diseases, the same

approaches can be used for non-infectious diseases,

aspects of animal welfare and many other issues at a

population level

One of the earliest and most frequently cited

his-torical examples of epidemiology is the control of an

outbreak of cholera in Soho, London in 1854 by the

Reverend John Snow (Johnson 2006) Rev Snow

col-lected information from a¡ected and una¡ected

households and discovered that the a¡ected

house-holds all used a particular water pump He had the

handle of the pump removed to prevent continued

use of this source of water It has been claimed that

this action curtailed the epidemic and in all

likeli-hood it would have done, however, Rev Snow

thought that the epidemic was already declining

be-fore he had the pump disabled In his own words:

There is no doubt that the mortality was much

diminished, as I said before, by the £ight of the

population, which commenced soon after the

outbreak; but the attacks had so far diminished

before the use of the water was stopped, that it is

impossible to decide whether the well still

con-tained the cholera poison in an active state, or

whether, from some cause, the water had

be-come free from it

Despite Rev Snow’s reservations this example

de-monstrates one of the key strengths of epidemiology;

it is not always necessary to understand everything

about the disease process to control it Rev Snow

took the handle o¡ the pump in 1854: this was 6^10

years before Louis Pasteur conducted studies to

de-monstrate that fermentation and the growth of croorganisms in nutrient broths did not proceed byspontaneous generation

mi-Epidemiology identi¢es risk factors

It is frequently said that epidemiology can be used toidentify risk factors, or factors that increase the prob-ability of a disease occurring The water pump thatwas identi¢ed as the source of the ‘cholera poison’ byRev Snow was an example of a risk factor The identi-

¢cation of the association between smoking andcardiovascular disease is another example Theincreased risk of cardiovascular disease as a result

of smoking has been recognized since the 1960s viewed by Burns 2003) However, the fact that smok-ing is still so prevalent demonstrates that justidentifying risk factors and proposing control strate-gies (i.e not smoking) is not necessarily enough to re-duce the impact of a disease

(re-Epidemiologists investigating aquatic animal eases have also identi¢ed risk factors that can beavoided to reduce the impact of diseases For example,Vagsholm, Djupvik,Willumsen,Tveit & Tangen (1994)identi¢ed slaughter waste as a means of spreadingthe potentially serious viral disease, infectious sal-mon anaemia (ISA) Murray, Smith and Stagg (2002)subsequently identi¢ed that boats transporting ¢shcan also spread ISA

dis-Despite these encouraging examples, in manycases it is di⁄cult to identify controllable risk factorsbecause resources are limited and the disease pro-blems are complex Risk factors may be masked byother factors For example, it is di⁄cult to investigatethe association between sea lice on wild and farmed

¢sh since distance between farmed and wild tions is confounded by currents that vary with thewind (Amundrud & Murray 2009)

popula-With limited resources, it might be tempting toconduct a super¢cial study and base control mea-sures on preliminary evidence However, it is neces-sary to conduct studies in su⁄cient depth tocon¢rm that apparent risk factors are really asso-ciated with the disease problem in a causal manner.Although it is useful in generating testable hypoth-eses, correlation does not necessarily imply causa-tion It is possible to identify apparent risk factorswith no direct link between the cause and the dis-ease These spurious risk factors are known as con-founders, or the association is said to beconfounded An example is the signi¢cant statisticalEpidemiology in the UK aquaculture J F Turnbull et al Aquaculture Research, 2011, 42, 21^27

association between drowning and the application of

sunblock In this case the confounding e¡ect is

ob-vious: people that swim outdoors are both more likely

to wear sunblock and drown, but the application of

sunblock in no way increases the risk of drowning

In other cases the confounding e¡ect may not be so

obvious, and it may be necessary to conduct further

analysis to distinguish genuine risk factors from

con-founding e¡ects It is therefore necessary to be

cau-tious of both preliminary ¢ndings and apparently

simple solutions H L Mencken (a writer from the

United States during the early 20th century) is

re-ported to have said ‘There is always an easy solution

to every human problem ^ neat, plausible, and

wrong’ (the original quote has been obscured by

ex-tensive repetition and adaptation) E¡ective control

strategies often require improved management,

hy-giene and biosecurity, requiring signi¢cant changes

to management practices (e.g.Wheatley, McLaughlin,

Menzies & Goodall1995) This is not a contradiction of

the earlier statements: E¡ective disease control must

be robust and is seldom simple, but it does not

neces-sarily require detailed understanding the aetiology

Any disease control strategies should be tested in

¢eld trials before they are promoted as a solution for

the problem Field trials, although very rarely used in

aquatic animal health control, have the potential to

con¢rm the e⁄cacy of any control strategy and also

estimate the true cost and bene¢ts Without ¢eld

trials it is impossible to be con¢dent that the strategy

will work in a cost e¡ective manner

Although the identi¢cation of risk factors is an

important step to controlling diseases, ruling out

risk factors can also be a valuable step in the

inves-tigation of diseases For example, a recent project

studying a condition known as rainbow trout

gas-troenteritis (RTGE) quickly demonstrated a lack of

any association between the disease and either the

feed or the source of the young ¢sh This allowed

the limited investigative resources to be redirected

into other areas, such as movement within sites,

and resulted in the development of potential control

strategies (del-Pozo, Crumlish, Ferguson & Turnbull

2009)

In terms of strategic disease control it is often

ne-cessary to know the extent of the pathogen in farmed

and wild populations Again, epidemiological

meth-ods can help not only to produce useful information

but also to optimize the cost e⁄cacy of any sampling

strategy (e.g Feist, Peeler, Gardiner, Smith &

Long-shaw 2002; Chambers, Gardiner & Peeler 2008);

Taylor, Dixon, Je¡ery, Peeler, Denham & Way 2010

Epidemiological principles and analytical techniqueshave also been applied in other areas such as aquaticanimal welfare

A study on a commercial cage site recorded data onthe welfare of over 248 000 Atlantic salmon and sug-gested that while stocking density can a¡ect the wel-fare of Atlantic salmon in production cages, it is onlyone of many factors that in£uence their welfare andstocking density on its own cannot be used to eitheraccurately control or predict welfare (Turnbull, Bell,Adams, Bron & Huntingford 2005) Similar studieshave been conducted for chickens and have producedsimilar conclusions, that the association betweenstocking density and animal welfare is not simple ordirect (Dawkins, Donnelly & Jones 2004; Jones, Don-nelly & Dawkins 2005) By demonstrating the com-plexity of the relationship between stocking densityand welfare it has been possible to reduce the pres-sure for unjusti¢ably restrictive and simplistic con-trols on stocking density and focused resources onmore e¡ective means of safeguarding ¢sh welfare

Epidemiology can reduce the cost of disease monitoring

or surveillanceThere has been a move in many parts of the worldaway from blanket sampling of large numbers offarms, towards targeted surveillance The areas ofgreatest risk are identi¢ed and e¡orts are focused ontothese areas, reducing the amount of e¡ort and thecost of surveillance Though the theory is sound, thepractical application of targeted surveillance still pre-sents many challenges One aspect of targeted surveil-lance is the study of aquatic animal contact networks.Ideally, all inter-site routes for potentially infectiouscontact should be identi¢ed (e.g transport of live ¢sh,water, fomites, etc.), but these are rarely available andtherefore the contacts in the network are restricted tothe available data (Munro & Gregory 2009)

In the very small hypothetical network sented in Fig 1, the risk ^ and potential extent ^ of

repre-an epidemic might be reduced by monitoring or cing the risk of transmission between particularpairs of sites If the risk of disease transmission be-tween sites 4 and 6 (linka) were reduced, or for thisexample eliminated, it would have no e¡ect on eitherthe maximum potential size of an epidemic (ninesites) or the mean epidemic size given a random in-dex case (nine sites) However, if the link (b) betweensites 4 and 5 were also eliminated then the maxi-mum epidemic size is reduced to ¢ve sites and themean to 4.5 sites This simple example demonstratesAquaculture Research, 2011, 42, 21^27 Epidemiology in the UK aquaculture J F Turnbull et al.

redu-r 2010 The Authors

that it is not only necessary to target the most

ob-vious link in the network, but also a mechanism must

be found to identify a series of links that can be

tar-geted to reduce the overall risk of epidemics In real

networks, even relatively simple ones, the task is

much more complex Fig 2 represents a real but

rela-tively small network of live salmonid movements inScotland during 2003 (Green, Gregory & Munro2009)

Epidemiology is valuable component ofmultidisciplinary studies

Epidemiological principles can be used to more rately estimate the consequences of diseases and pro-vide data for cost^bene¢t analysis on controlstrategies However, disease control involves a variety

accu-of people including farmers, legislators and scientists.One group that can make a useful contribution to dis-ease control is economists They have the expertise tomake more accurate estimates of the costs of diseaseand its control (e.g Butler, Radford, Riddington &Laughton 2009) E¡ective communication betweenthese groups is an essential component of controllingdiseases

For successful disease control, in most cases it

is not only necessary to engage various groups ofpeople but it is also necessary to involve various

Figure 1 A very small hypothetical network with the

cir-cles representing sites (or other epidemiological units) and

the lines representing potentially infectious contact

be-tween them In this network, infections could spread in

either direction between sites

Figure 2 Contact network for live salmonid movements between registered Scottish sites in 2003 Arrows indicate thedirection of movement

Epidemiology in the UK aquaculture J F Turnbull et al Aquaculture Research, 2011, 42, 21^27

scienti¢c disciplines Aquatic animal health

pro-blems are most e¡ectively addressed by selecting the

correct tools or techniques for the study, not looking

for studies for you favourite tools

E¡ective multidisciplinary studies are not

necessa-rily easy, and may be more expensive than single

dis-cipline studies E¡ective collaboration between

disciplines is based on personal relationships and is

most successful when the individuals involved

un-derstand and trust each other Part of the challenge

in multidisciplinary studies is to ensure e¡ective

communication and a mutual understanding Even

in relatively closely related subject areas, such as

pathogen identi¢cation and epidemiology, identical

terms can be used with di¡erent meanings Such

polysemy can easily lead to misunderstandings For

example, sensitivity has a speci¢c epidemiological

de¢nition, i.e the proportion of animals with the

dis-ease or pathogen that test positive Many people

using pathogen identi¢cation tests in a laboratory

setting use the term sensitivity to describe the limit

of detection of a test, e.g how many bacteria are

necessary for the test to produce a positive result

Between more disparate disciplines such as

epide-miology and economics the problems can be even

greater In such cases the development of a common

language can take time and e¡ort

For these reasons, e¡ective collaborative networks

require resources to develop and therefore have value

that may justify sustaining them A Speci¢c Targeted

REsearch Project (STREP) supported by the European

Union Directorate-General for Maritime A¡airs and

Fisheries (Bene¢sh) (http://www.bene ¢sh.eu)

com-bined biologists, veterinarians, economists, social

scientists and industry representatives to model the

costs of improving ¢sh welfare throughout the value

chain from the farm to the consumer Personal

ex-perience during this project demonstrated that

na-tional languages were not a signi¢cant barrier to

communications, but developing mutual

understand-ing of concepts and language related to the discipline

took a considerable amount of time and e¡ort

Another essential step in the process of controlling

disease is dialogue with the people who will

even-tually use the outputs of the research Engaging the

end users in the research process is a problem that

has been extensively debated in the context of

inter-national aid and development It is clear from this

de-bate is there is no simple formula for engaging

stakeholders in the research process (Cooke & Kothari

2001) but a dissemination phase once the research is

completed is certainly inadequate and ine¡ective

Good research improves understanding but tive practical strategies require this understanding

e¡ec-be discussed with the end users Although the tists may have the best understanding of the causes of

scien-a disescien-ase, fscien-armers scien-are best plscien-aced to identify scien-aspects

of the farming system that can be changed to reducethe impact of the disease However, the end users arenot exclusively farmers but may be retailers, govern-ment agencies, legislators, groups seeking to in£u-ence opinion and others

A study on RTGE, through the application of miology, microbiology and histopathology amongother disciplines, determined that the main risk forspreading the disease was moving ¢sh within the siteduring an outbreak (del-Pozo, Crumlish, Ferguson &Turnbull, 2010) In order to turn this into a controlstrategy it was necessary to discuss with farmerswhy ¢sh were moved, and what the potential pro-blems and costs would be if the ¢sh were not moved.This resulted in a proposed control strategy that wasdisseminated by the British Trout Association(http://www.britishtrout.co.uk), the UK trout indus-try’s main representative body and a co-funder andactive participant in the research project

epide-E¡ective disease control requires good dataBroadly, epidemiological studies can be divided intointerventional and observational methods To thesemay be added theoretical approaches such as consid-eration of contact networks as described above.Withinterventional studies, we manipulate putative riskfactors or control measures and observe the result.This may be thought of as a gold standard for scienti-

¢c experimentation, but frequently manipulation isnot possible on economic, ethical or practicalgrounds In this case, we must rely on obtaining datathrough observational studies, be they derived fromacademic study or routine industrial data collection.These may in turn be divided into retrospective stu-dies: looking backwards in time to examine associa-tions between conditions and risk factors; orprospective studies: following individuals exposed torisk factors, and awaiting disease outcomes.Acquiring good data may necessitate collection ofsamples, collecting information from farmers andothers with questionnaires, or even asking farmers

to provide data Many types of study require thecollection of samples from ¢sh in tanks or cages, onvarious farms, in various areas, and over a period oftime Sampling theory, an important component ofepidemiology, allows one to estimate the quality ofAquaculture Research, 2011, 42, 21^27 Epidemiology in the UK aquaculture J F Turnbull et al.

r 2010 The Authors

the information that will be produced and

cost-e¡ec-tive compromises between quality of data and

avail-able resources to be identi¢ed and compared

A project attempting to collect strategic

industry-wide data from trout farmers demonstrated that

although there is widespread recognition of the value

of industry wide strategic data, persuading farmers

to contribute data to a central resource is a di⁄cult

process (Development of a scheme for monitoring

sentinel farms in the UK trout industry, Scottish

Aquaculture Research Forum, project number 028)

The cost in terms of time and e¡ort, data security

and ownership can all result in reluctance to

partici-pate (unpubl obs.)

Data from a salmon company has been used to

as-sess risk factors associated with sea lice (Revie,

Get-tinby, Treasurer & Wallace 2003) and to develop

models (Revie, Robbins, Gettinby, Kelly & Treasurer

2005) Such data are now being used to assess e¡ects

of disease in collaboration between University of

Stirling and Marine Scotland It is proposed to bring

data from several companies together to improve

analysis, but this is in an early stage Much valuable

work has been done with similar data from Norway

(Aunsmo, Bruheim, Sandberg, Skjerve, Romstad &

Larssen 2008) Data on the salmon industry collected

by o⁄cial ¢sh health inspectors have been used to

de-velop a simple model of the emergence of infectious

pancreatic necrosis virus in Scotland (Murray 2006)

and Ireland (Ruane, Murray, Geoghegan & Raynard

2009) and assess potential control strategies, while

that collected to authorize ¢sh movements has been

used to derive a network of contacts as described

ear-lier (Munro & Gregory 2009)

It is often suggested that data collected previously

may be re-used or re-analysed for other purposes

Unfortunately there are a considerable number of

constraints to such re-use of data Lack of familiarity

with the original data collection process can make the

data di⁄cult to interpret or make it di⁄cult to validate

the data Also, the data may have been collected for a

speci¢c purpose and be unsuitable for other purposes,

or the data may be out of date and no longer re£ect the

current situation The possibility of re-use of data is

certainly improved if the data collection and storage

is fully described with adequate metadata and the

col-lection is based on sound sampling theory

Statisti-cally, one must be aware of the danger of post hoc

‘data dredging’ analyses on secondary data

Perform-ing multiple unplanned statistical tests, often the case

with retrospective studies, necessarily results in

mul-tiple signi¢cant tests, a proportion of which will be

false positives Also, even large data sets may be cient in information content For example, should twovariables such as pH and water hardness be highlycorrelated, it might never be possible from that dataset

de¢-to ascribe one over the other as a true risk facde¢-tor Atthis point there may be no substitute for an interven-tional study

The EU funded project Bene¢sh (http://www.bene

¢sh.eu) examined 49 existing data sets relating to thehealth, welfare and production of farmed ¢sh, toidentify potential interventions to improve ¢sh wel-fare Out of the 49 datasets only eight possible inter-ventions were identi¢ed

Coordinating e¡ortThere will never be enough resources to address allthe issues faced by a sustainable aquaculture industrybut coordination of e¡ort to achieve agreed prioritiescan help Coordination is not necessarily easy, sincevested interests may create obstacles and trust takestime and e¡ort to develop Despite these challengesthere were many examples of successful coordination

at the time of writing For example, Marine Scotlandhas funded a joint post with the University of Stirling

to encourage better communication between demics and the regulatory authority, provide MarineScotland with access to academic resources, and aca-demics at Stirling with access to data and informationthat would not otherwise be available This collabora-tion led directly to the work on networks describedearlier The production of integrated UK-wide aquaticresource data would facilitate more e¡ective policyand regulation Individuals already work together ef-fectively, but this has to be matched by active coopera-tion from agencies and governments

aca-ConclusionEpidemiology is a powerful, cost-e¡ective tool for ap-plied and strategic aquatic animal health control Agreat deal of bene¢t has been derived from funding

in this area but projects must use the right tion of tools, work with the end user, have good dataand coordinate e¡orts

combina-AcknowledgmentsThis paper would not have been possible to writewithout drawing on work conducted in collaborationwith a wide range of people We would like to extendEpidemiology in the UK aquaculture J F Turnbull et al Aquaculture Research, 2011, 42, 21^27

special thanks to Kenton Morgan and the late Chris

Baldock for their inspirational approach to

epide-miology

References

Amundrud T.L & MurrayA.G (2009) Modelling sea lice

dis-persion under varying environmental forcing in a

Scot-tish sea loch Journal of Fish Diseases 32, 27^44.

Aunsmo A., Bruheim T., Sandberg M., Skjerve E., Romstad S.

& Larssen R.B (2008) Methods for investigating patterns

of mortality and quantifying cause-speci¢c mortality in

sea-farmed Atlantic salmon Salmo salar Diseases of

Aqua-tic Organisms 81, 99^107.

Burns D.M (2003) Epidemiology of smoking-induced

cardi-ovascular disease Progress in Cardicardi-ovascular Diseases 46,

11^29.

Butler J.R.A., Radford A., Riddington G & Laughton R.

(2009) Evaluating an ecosystem service provided by

Atlantic salmon, sea trout and other ¢sh species in the

River Spay, Scotland: the economic impact of recreational

rod ¢sheries Fisheries Research 96, 259^266.

Chambers E., Gardiner R & Peeler E.J (2008) An

investiga-tion into the prevalence of Renibacterium salmoninarum

in farmed rainbow trout, Oncorhynchus mykiss

(Wal-baum), and wild ¢sh populations in selected river

catch-ments in England and Wales between 1998 and 2000.

Journal of Fish Diseases 31, 89–96.

Cooke B & Kothari U (2001) Participation:The NewTyranny?

Zed Books, London, UK.

Dawkins M.S., Donnelly C.A & Jones T.A (2004) Chicken

welfare is in£uenced more by housing conditions than

by stocking density Nature 427, 342^344.

del-Pozo J., Crumlish M., Ferguson H.W & Turnbull J.F.

(2009) A retrospective cross-sectional study on

Candida-tus arthromiCandida-tus associated rainbow trout Aquaculture

209, 22^27.

del-Pozo J., Crumlish M., Ferguson H.W., Green D.M &

Turnbull J.F (2010) A prospective longitudinal study of

‘‘Candidatus arthromitus’’ associated rainbow trout

gas-troenteritis in the UK Preventative Veterinary Medicine

94, 289–300.

Feist S.W., Peeler E.J., Gardiner R., Smith E & Longshaw M.

(2002) Proliferative kidney disease and renal

myxospori-diosis in juvenile salmonids from rivers in England and

Wales Journal of Fish Diseases 25, 451^458.

Green D.M., Gregory A & Munro L.A (2009)

Small-and large-scale network structure of live ¢sh movements

in Scotland PreventiveVeterinary Medicine 91, 261^269.

Johnson S (2006) The Ghost Map: The Story of London’s Most

Terrifying Epidemic- and How it Changed Science, Cities, and

the Modern World Riverhead Books, Penguin Group Inc.,

New York, USA.

Jones T.A., Donnelly C.A & Dawkins M.S (2005) mental and management factors a¡ecting the welfare of chickens on commercial farms in the United Kingdom and Denmark stocked at ¢ve densities Poultry Science

Environ-84, 1155–1165.

Loeber R & Farrington D.P (2001) Young children who mit crime: epidemiology, developmental origins, risk fac- tors, early interventions, and policy implications Development and Psychopathology 12,737^762.

com-Munro L.A & GregoryA (2009) Application of network lysis to farmed salmonid movement data from Scotland Journal of Fish Diseases 32, 641^644.

ana-Murray A.G (2006) A model of the emergence of infectious pancreatic necrosis virus in Scottish salmon farms 1996

to 2003 Ecological Modelling 199, 64^72.

Murray A.G., Smith R.J & Stagg R.M (2002) Shipping and the spread of infectious salmon anemia in Scottish aqua- culture Emerging Infectious Diseases 8, 1^5.

Reichert T.A., Sugaya N., Fedson D.S., GlezenW.P., Simonsen

L & Tashiro M (2001) The Japanese experience with cinating schoolchildren against in£uenza New England Journal of Medicine 344, 889^896.

vac-Revie C.W., Gettinby G., Treasurer J.W & Wallace C (2003) Identifying epidemiological factors a¡ecting sea lice Le- peophtheirus salmonis abundance on Scottish salmon farms using general linear models Diseases of Aquatic Or- ganisms 57, 85^95.

Revie C.W., Robbins C., Gettinby G., Kelly L & Treasurer J.W (2005) A mathematical model of the growth of sea lice, Lepeophtheirus salmonis, populations on farmed Atlantic salmon, Salmo salar L., in Scotland and its use in the as- sessment of treatment strategies Journal of Fish Diseases

28, 603^613.

Ruane N.M., Murray A.G., Geoghegan F & Raynard R.S (2009) Modelling the initiation and spread of Infectious Pancreatic Necrosis Virus (IPNV) in the Irish salmon farming industry: the role of inputs Ecological Modelling

220, 1369^1374.

Taylor N.G.H., Dixon P.F., Je¡ery K.R., Peeler E.J., Denham K.L & Way K.M (2010) Koi herpesvirus: distribution and prospects for control in England and Wales Journal of Fish Diseases 33, 221^230.

Turnbull J., Bell A., Adams C., Bron J & Huntingford F (2005) Stocking density and welfare of cage farmed Atlan- tic salmon: application of a multivariate analysis Aqua- culture 243, 121^132.

Vagsholm I., Djupvik H.O., Willumsen F.V., Tveit A.M & gen K (1994) Infectious Salmon Anemia (ISA) epidemiol- ogy in Norway PreventiveVeterinary Medicine19, 277^290 Wheatley S.B., McLoughlin M.F., Menzies F.D & Goodall E.A (1995) Site management factors in£uencing mortal- ity rates in Atlantic salmon (Salmo salar L.) during marine production Aquaculture 136, 195^207.

Tan-Aquaculture Research, 2011, 42, 21^27 Epidemiology in the UK aquaculture J F Turnbull et al.

r 2010 The Authors

What has been done to minimize the use of

antibacterial and antiparasitic drugs in Norwegian aquaculture?

Paul J Midtlyng1, Kari Grave2& Tor Einar Horsberg2

1 LaboratoryAnimal Unit, Norwegian School of Veterinary Science, Oslo, Norway

2 Department of Pharmacology and Toxicology, Norwegian School of Veterinary Science, Oslo, Norway

Correspondence: P J Midtlyng, Laboratory Animal Unit, Norwegian School of Veterinary Science, PO Box 8146, Dep N-0033, Oslo, way E-mail: paul.midtlyng@nvh.no

Nor-Abstract

Since1987, the use of antibacterial drugs in Norwegian

¢sh farming has been drastically reduced from about

48 tonnes to approximately 1tonne annually, varying

slightly from year to year The 649 kg active substance

prescribed in 2007 corresponds to approximately

0.03% of the produced biomass being treated once

with an antibacterial drug This is an exceptional

¢g-ure, particularly when compared with antibacterial

drug use in terrestrial animal production or humans

It is therefore highly unlikely that this limited use of

antibacterial drugs in farmed ¢sh poses a signi¢cant

risk to human medicine via the development of

anti-bacterial drug resistance Among the factors

contri-buting to this favourable situation are: (a) a unique

government^industry initiative adopted in the early

1990s to facilitate vaccination against classical

furun-culosis; (b) development of high-quality vaccines by

the pharmaceutical industry; (c) the continuing

pre-dominance of vaccination strategies for disease

con-trol among ¢sh farmers; (d) adoption of ‘all-in-all-out’

production systems, with mandatory fallowing

peri-ods between year classes; and (e) zoning and the

spa-tial re-arrangement of marine production sites to

minimize horizontal spread of infections In Norway,

targeted and £exible governance has allowed the rapid

implementation of mass vaccination as well as

zoo-sa-nitary measures The limited use, however, also tends

to limit the availability of licensed veterinary drugs in

the marketplace This creates a dilemma that should be

addressed as a strategic issue for further long-term

de-velopment of sustainable industrialized aquaculture

Keywords: antibiotic, ¢sh farming, vaccination,

chemotherapy, treatment, veterinary

The use of antibacterial drugs inNorwegian aquaculture

During the development of industrialized salmon ing in the 1980s, epizootics caused by bacterial infec-tions, such as vibriosis, cold water vibriosis andfurunculosis, caused a major increase in the use of anti-bacterial drugs in Norwegian aquaculture (Grave, En-gelstad, Solie & Hastein 1988; Grave 1991) Followingthe implementation of a number of successful mea-sures to control these diseases, the use of antibiotics de-creased rapidly by more than 90% (Grave, Markestad &Bangen 1996) Today, published ¢gures show that, de-spite the fact that the Norwegian production of farmedsalmonids has more than doubled during the last 10years, the use of antibacterial for aquaculture purposes

farm-is still at the same low level, £uctuating around

1000 kg active drug substance per year (Fig 1) Thisserves to demonstrate that the large-scale production

of farmed salmon is possible without the regular use ofantıbacterial drugs, an important consideration withrespect to the sustainability of any industrialized foodproduction

From an ecological point of view, the usage of bacterial drugs in Norwegian aquaculture only com-prises approximately 2% of the total annual usage ofantibacterial drugs in Norway (Table 1)

anti-A breakdown of the sales per farmed ¢sh speciesreveals further interesting details (Table 2) Whereas

in 2007 only 133 kg antibacterial drugs (activesubstance) were prescribed for use in Atlantic sal-mon, nearly three times this ¢gure was used forfarmed Atlantic cod (Norwegian Food Authority2009) Relating this to the harvested biomass, thiscorresponds to 0.17 g (170 mg) of antibacterial drugsper 1000 kg of harvested farmed salmon

Aquaculture Research, 2011, 42, 28^34 doi:10.1111/j.1365-2109.2010.02726.x

The amounts of farmed ¢sh that can be treated

with the amount of antibacterial drugs used in

Nor-way during 2007 represent 0.03% of the produced

biomass in metric tonnes It is therefore unlikely that

the use of antibacterial drugs in Norwegian farmed

¢sh poses a signi¢cant risk to human medicine with

respect to driving antibacterial drug resistance As

data on the usage in Norway of antibacterial drugs

for terrestrial animals are not available, a similar

es-timation is not possible for this compartment In

comparison, however, data from the Norwegian drug

prescription database show that in 2005 and 2006,

approximately 25% of the human population were

treated one or more times with an antibacterial drug

annually (Blix, Engeland, Litleskare & Rnning

2007) These ¢gures serve to illustrate that in

Nor-way, aquaculture contributes very little to the overall

usage of antimicrobials on a national scale

The use of antiparasitic drugs in

Norwegian aquaculture

Following a shift in bath treatments from

organopho-sphates to the far more potent pyrethroids, the

Norwegian annual sales of anti-sea lice treatments

have increased slowly in recent years, from mately 100 to 150 kg pure substance (Fig 2) Relatingthis usage pattern to the population at risk, thisamount of active substance corresponds to approxi-mately two anti-parasitic treatments per year(Fig 3) Stimulated by cases showing reduced sensi-tivity to the licensed medicines, prescriptions of aza-methiphos (an alternative sea lice treatment available

approxi-in Scotland) have re-emerged recently (Fig 2)

The main factors contributing to thecurrent situation

Vaccination practices

In response to a peak in antimicrobial use in 1989^

1990, the Norwegian ¢sh farming industry (via theFish Farmers Sales Organisation) and government(via the National Veterinary Institute) joined forces

to explore and promote mass vaccination against

60000

800 1000

20000

40000

200 400

Table 1 Annual sales per user segment, in kg active

substance, of antibacterial drugs in Norway during 2007

Data obtained from Grave (2008).

K Grave, unpublished estimate (calculated from National

Insti-tute of Public Health sales data for 2006).

Table 2 Use of antibacterial drugs per farmed ¢sh species

in Norwegian aquaculture in 2007

Annual harvest (tonnes)

Ratio (g per tonne harvested) Atlantic salmon 133 744 222 0.17

Aquaculture Research, 2011, 42, 28^34 Minimize the use of antibacterial and antiparasitic drugs P J Midtlyng et al.

furunculosis, the dominant cause for the use of

anti-microbial therapy In parallel projects, extraordinary

e¡orts were mobilized to gain new scienti¢c

knowl-edge about furunculosis vaccines and to disseminate

this knowledge among ¢sh farmers The rapid

suc-cess of this venture was facilitated by the

develop-ment and introduction of multivalent, oil adjuvanted

vaccine formulations, which provided hitherto

un-precedented protection against clinical disease,

last-ing through harvest (Midtlyng 1997) These e¡orts

were unconventional, in that the scienti¢c evaluation

carried out by a government institute was paralleled

by a promotional campaign,‘Stop furunculosis’,

orga-nized by the aquaculture industry association In

only 2 years, the use of oil adjuvant furunculosis

vac-cines increased from a few million doses for ¢eld

trials in 1991 to 99% coverage of the Atlantic salmon

output in 1993 (Markestad & Grave 1997) Following

this experience, there remains a continuing ence for the development of new vaccines (over newpharmaceuticals) within the aquaculture-orientedpharmaceutical industry

prefer-Creative governance

It is worth taking a closer look at how the resourcesfor these projects were mobilized, and how regula-tions governing the use of ¢sh medicines were exe-cuted to support vaccination, instead of drugtherapy Funding for the scienti¢c work was obtainedthrough a 3% levy on all national animal vaccinesales, to ¢nance an independent scienti¢c evaluation

of vaccine prototypes by the National VeterinaryInstitute In return, veterinary pharmaceuticalcompanies were o¡ered free inclusion of vaccine

Figure 2 Annual sales of drugs against salmon lice infestations (kg active substance) in Norwegian ¢sh farming 2003^

2008 Data from Anonymous (2008): Key ¢gures from the Norwegian aquaculture industry 2007

Minimize the use of antibacterial and antiparasitic drugs P J Midtlyng et al Aquaculture Research, 2011, 42, 28–34

prototypes in these trials, conditional to full

transpa-rency of the results At the same time, the veterinary

and medicinal authorities provided a time-limited

marketing authorization for inactivated furunculosis

bacterins, based on pharmaceutical quality and

safety assessment The manufacturers were then

allowed 3 years in which to submit a full

documenta-tion of the vaccine’s e⁄cacy Through this

untra-ditional way of governance, the business risks

associated with the development of new furunculosis

vaccines for ¢sh were lowered, strongly stimulating

new product development, as well as the uptake of

novel vaccination practices

General zoo-sanitary measures for disease

control

The 1988^1991 sanitary crisis of Norwegian salmon

farming also included the epidemic spread of

infec-tious salmon anaemia (ISA), in addition to endemic

furunculosis (Vgsholm, Djupvik,Willumsen,Tveit &

Tangen 1994) The combination of both disease risks

created a nearly impossible production management

situation during the winter periods, when ¢sh that

had undergone antibiotic treatment could not then

be harvested rapidly during an ISA outbreak This

di-lemma is generic in nature, and we believe that it is

experienced in a similar way today, for example in

areas of Chile, where both piscirickettsiosis and

ISA are endemic The Norwegian response to this

situation was a comprehensive implementation ofgeneral zoo-sanitary strategies to reduce, and prefer-ably prevent, horizontal disease transmission (be-tween sites and between year classes) of eitherinfection (Hstein, Hill & Winton 1999) Among themost important measures within this strategy are:

 Year class separation and fallowing;

 Re-location to increase the distance between sites;and

 Zoning and co-ordination between farmers within

a zone

Although there is a general agreement on the pal bene¢ts of yearclass separation and fallowing,discussion often arises regarding how long thefallowing period needs to be Obviously, this is due

princi-to the increased costs of periodically moving nel and equipment or periodically rendering theinvestment made into a production site economicallyinactive The aim of both fallowing and increasingthe distance between sites is to reduce the probability

person-of an on-site infection reservoir (be it in the sediments,

in vectors belonging to the local fauna, or through fected escapees) transmitting the infection to the nextinput of ¢sh or to a neighbouring population In thecase of ISA, it appears from the Norwegian experiencethat distances above ca 5 km between sites (Jarp &Karlsen 1997) and fallowing periods above 3 months(minimum 6 months after an ISA outbreak) in mostcases are e¡ective in preventing horizontal transmis-sion Empirical data are, however, lacking to verifythese parameters for furunculosis or other bacterial

x x x x x x x x x Site B

x x x x x x x x x x Site C

Aquaculture Research, 2011, 42, 28^34 Minimize the use of antibacterial and antiparasitic drugs P J Midtlyng et al.

infections A typical fallowing pattern for this

scenar-io is shown in Fig 4a As the growth characteristics

of selectively bred salmon are gradually improving,

a shorter seawater rearing period of 18^24 months

allows optimized rotation between three sites, while

maintaining continuous harvest, and increasing the

resting periods up to 12 months (Fig 4b)

Zoning and coordination between

producing units

Working to prevent horizontal and

between-year-class transmission of infectious disease, spatial

con-siderations and temporal co-ordination between

neighbouring farmers become essential In the

Norwegian situation, these aspects were resolved at

a regional level, driven and co-ordinated by the

veter-inary authorities Among these regional initiatives,

local implementations (the Sunnmre model, the

Namdalen model) gained widespread acceptance

and were consequently e¡ectively implemented by

the industry Similar bene¢ts of resolving challenges

locally and through government^stakeholder

colla-boration has since been conducted elsewhere, both

in the Faroe Islands (the Faroese model for ¢sh

disease control) and in Scotland (Area ¢sh health

management plans) In Chile, the establishment

of ‘barrios’ (neighbourhoods) has been launched

recently to co-ordinate control measures during the

current sanitary crisis

Zoning requires that groups of farming sites are

su⁄ciently separated in spatial terms to prevent

pas-sive transmission of disease agents, and this may

oc-casionally require the re-location of sites that ‘bridge’

between otherwise separated zones To further

as-sure sanitary separation, rules governing the

move-ment of live ¢sh and potentially infected equipmove-ment

need to be in place Within zones, however, priority

rests with the co-ordination between farmers

regard-ing the essential elements of ¢sh health management

and disease control, such as:

 stocking periods, which sites to stock when,

 fallowing periods for adjacent sites,

 vaccination strategies to achieve ‘herd protective

e¡ects’,

 timing of sea lice treatments, etc

Other contributing factors

Among the measures that have supported

vaccina-tion and zoo-sanitary strategies to e¡ectively control

bacterial diseases in Norwegian salmon farming isselective breeding for increased disease resistance.Disease resistance has continuously been a high-priority issue in Norwegian Atlantic salmon breed-ing programmes, providing up to 60% of the basisfor selection in AquaGen egg production stocks ofcertain years (A Storset, pers comm.) Resistance tobacterial diseases using furunculosis as the model in-fection has therefore been included in the quantita-tive family breeding index since 1995, representingfour generations of selection (Gjedrem, Salte & Gjen1991; Gjedrem & Gjen 1995) A further factor to sup-port the success of vaccination as of zoo-sanitarystrategies has been the sanitary protection of juve-nile production, which has been implementedthrough license requirements on hatcheries andsmolt production units, and through regulations ontheir operation as follows:

 water intake above the migration barrier or fection of intake freshwater;

disin- disinfection of all seawater supply;

 double disinfection of eggs (after fertilization, and

As experienced with the introduction of furunculosis

to Norway (Egidius 1987), it may take only one ductory event to establish a new infectious diseaseand consequently to change the disease panorama of

intro-an entire coastal farming area Aquaculture thus facesthe same principal situation as animal and plant pro-duction, where the legislative instruments determinehow far reaching and by which procedures society at-tempts to protect itself from epidemic hazards.Whereasvoluntary ‘Codes of Practice’are good for education andraising awareness, only mandatory measures su⁄ce tocontrol severe communicable diseases and pests It isfurther evident ^ and the current sanitary crisis in Chi-lean aquaculture may serve as a recent example ^ thatunless implemented by legislation, the necessary pro-tective measures will eventually be enforced via eco-nomic mechanisms, following severe crisis withbankruptcies, take-overs and mergers, and ¢nally there-structuring of production under a centralized busi-ness management regime

It is interesting to observe that in Norway, there-location of aquaculture sites due to acute zoo-sanitary needs has occasionally been accomplishedMinimize the use of antibacterial and antiparasitic drugs P J Midtlyng et al Aquaculture Research, 2011, 42, 28–34

within a few months It appears that the situation in

other countries, for example in Scotland or Chile, is

quite di¡erent and, in most cases, proceeds more

slowly In the Norwegian system, marine

aquacul-ture licenses represent a general right to culaquacul-ture ¢sh

in the ‘coastal commons’ Technically, a license is

limited in the total production volume, but may use

several sites, each of which may have a speci¢c

max-imum, based on estimated carrying capacity

Produc-tion licenses are retained, even if certain sites are

abandoned Although the license asset is, in one

as-pect, ¢xed to the seabed at its location, it remains

valid for ‘anchoring’ in di¡erent acceptable locations,

if the circumstances so require This legal system

allows £exibility for optimizing the spatial

arrange-ments, without risking commercial or legal positions

Although this would fall outside our veterinary ¢eld

of research, a comparative study of the consequences

of di¡erent modes of legislation for the e¡ectiveness

of disease control would be of great interest, and

should be encouraged

The dilemma of successful disease

control

As evident from Fig 1 and Table 1, industrialized

sal-mon culture faces an inherent dilemma of a strategic

nature; the more successful our vaccination and

zoo-sanitary management, the less medicines are used

for treatment and the less pro¢table becomes the

li-censing of drugs for treatment However, modern

aquaculture needs adequate medicinal products to

treat the animals even when disease outbreaks are

infrequent; for obvious animal welfare reasons, for

contingency and for the mitigation of losses to single

producers Despite the fact that Norway is one of the

world largest producers of farmed ¢sh, the number of

licensed antibacterial drugs for aquaculture remains

small (Table 3)

The lack of treatment diversity leads to

drug resistance

It is generally acknowledged that attempts to control

endemic bacterial disease through recurrent

treat-ment with drugs sooner or later will lead to the

devel-opment of drug resistance During the Norwegian

furunculosis epizootic, it took only a few years until

the ¢rst reports of strains that were resistant to the

predominant therapeutic drug class (Hie 1991), and

the same situation was experienced elsewhere

(Tsoumas, Alderman & Rodgers 1989) In Norwegianand Scottish aquaculture, resistance of sea liceagainst organophosphates developed after exclusiveuse of treatments with these compounds for about

10 years (Denholm, Devine, Horsberg, Sevatdal,Fallang, Nolan & Powell 2002), while in Chileanaquaculture, resistance to the only licensed ectopar-asitic drug for salmonids, emamectin benzoate, took

9 years to develop (Bravo, Sevatdal & Horsberg 2008).The only way to delay resistance development totherapeutic drugs is to avoid long-term and recurrentuse and to alternate between diverse treatmentoptions This implies that if a ‘one-or-two-drug’situa-tion is allowed to prevail (e.g for control of sea lice),the question is when and not whether resistance willoccur The currently limited array of sea lice treat-ments in all major salmon-producing countries thuscalls for ‘creativity in governance’ to ¢nd a sustain-able long-term solution This hitherto unresolvedchallenge goes towards the pharmaceutical suppliersand the aquaculture industry users, and to the med-icinal and food authorities of the major salmonidfarming countries

Conclusions

To summarize the Norwegian experience, it is fair tosay that the primary focus of both regulators andaquaculture industry has been on e¡ective diseasecontrol, thereby removing the need for antibacterialdrugs Although environmental aspects have played

a role, the disease control focus has clearly been thesingle strongest driving force towards minimal druguse, motivating the joint e¡orts of industry and regu-lators Secondly, there is a lesson that extraordinaryproblems and challenges require extraordinarymeasures and that ‘normal procedures’ will not besu⁄cient to resolve epidemic health problems that

Table 3 Antibacterial and antiparasitic drugs licensed for farmed ¢sh in Norway

Drug classes Active substances Antibacterial drugs Amfenicols Florfenicol

Quinolones Flumequine

Oxolonic acid Sulphonamides

and trimethoprim

Sulphadiazine1 trimethoprim Antiparasitic drugs Avermectins Emamectin benzoate

Pyrethroids Cypermethrin

Deltamethrin Aquaculture Research, 2011, 42, 28^34 Minimize the use of antibacterial and antiparasitic drugs P J Midtlyng et al.

cause the extensive use of antibacterial or

antiparasi-tic drugs And ¢nally, even though the current

situa-tion in Norway and in Scotland can be considered to

be satisfactory, the poor diversity of relevant

antibac-terial and antiparasitic drugs for ¢sh should be

addressed as a strategic issue for aquaculture

devel-opment, preferably before drug resistance problems

become acute

Acknowledgments

Thanks are due to our veterinary colleague Arne

Storset, DVM, AquaGen AS for providing information

relating to the history of selective breeding e¡orts for

bacterial diseases in their strain of Atlantic salmon

References

Anonymous (2008) Key ¢gures for the Norwegian aquaculture

industry 2007 Nkkeltall for norsk havbruksnring 2007.

Directorate of Fisheries, Bergen, Norway Available at

http://www.¢skeridir.no/statistikk/akvakultur/statistiske-publikasjoner/noekkeltall-fra-norsk-havbruksnaering

Blix H.S., Engeland A., Litleskare I & Rnning M (2007)

Age- and gender-speci¢c antibacterial prescribing in

Nor-way Journal of Antimicrobial Chemotherapy 59, 971^976.

Bravo S., Sevatdal S & Horsberg TE (2008) Sensitivity

as-sessment of Caligus rogercressey to ememectin benzoate

in Chile Aquaculture 282,7^12.

Denholm I., Devine GJ., Horsberg TE., Sevatdal S., Fallang

A., Nolan DV & Powell R (2002) Analysis and

manage-ment of resistance to chemotherapeutants in salmon

lice, Lepeophtheirus salmonis (Copepoda: Caligidae) Pest

Management Science 58, 528^536.

Egidius E (1987) Import of furunculosis to Norway with

Atlan-tic salmon smolts from Scotland Mariculture Commitee

Report no C.M 1987/F:8, International Council for the

Exploration of the Sea (ICES).

Gjedrem T & Gjen HM (1995) Gentic variation in

suscept-ibility of Atlantic salmon, Salmo salar L., to furunculosis,

BKD and cold water vibrosis Aquaculture Research 26,

129^134.

Gjedrem T., Salte R & Gjen HM (1991) Genetic variation in

susceptibility of Atlantic salmon to furunculosis

Aqua-culture 97, 1^6.

Grave K (1991) Drug utilization in veterinary medicine with special emphasis on farmed ¢sh PhD thesis, Norwegian School of Veterinary Science.

Grave K (2008) Usage in animals In: NORM/NORM-VET

2007 Usage of Antimicrobial Agents and Occurrence of Antimicrobial Resistance in Norway; ChapterV (ed by M Norstro¨m & G Skov Simonsen) ISSN: 1502–2307, pp 15–18 National Veterinary Institute and Norwegian Institute of Public Health, Oslo/Tromsø, Norway Grave K., Engelstad M., Solie NE & Hastein T (1988) Utiliza- tion of antibacterial drugs in salmon farming in Norway during 1980^1988 Aquaculture 86, 347^358.

Grave K., Markestad A & Bangen M (1996) Comparison in prescribing patterns of antibacterial drugs in salmonid farming in Norway during the periods 1980^1988 and 1989^1994 Journal of Veterinary Pharmacology and Therapy 19, 184–191.

Hstein T., Hill B & Winton JR (1999) Successful aquatic imal disease emergency programmes Revue Scienti¢que et Technique 18, 214^227.

an-Hie S (1991) [Investigations into the resistance of monas salmonicida subsp salmonicida strains] Resis- tensunderskelser av Aeromonas salmonicida subsp salmonicida stammer Norsk veterinrtidsskrift 103, 527^

Aero-528 (in Norwegian).

Jarp J & Karlsen E (1997) Infectious salmon anaemia (ISA) risk factors in sea-cultured Atlantic salmon Salmo salar Diseases of Aquatic Organisms 28,79^86.

Markestad A & Grave K (1997) Reduction of antibacterial drug use in Norwegian ¢sh farming due to vaccination Developments in Biologicals 90, 365^369.

Midtlyng P.J (1997) Vaccination against furunculosis In: Furunculosis ^ Multidiciplinary Fish Disease Research (ed.

by E.-M Bernoth, AE Ellis, PJ Midtlyng, G Olivier & PR Smith), pp 382^404 Academic Press, London, UK Norwegian Food Authority (2009) Press release February

28, 2008 Available at http://www.mattilsynet.no/¢sk/ helsemessig_trygg/forbruk_av_antibiotika_i_norsk_¢skeop pdrett_i_2008_60772 (accessed 6 March 2009).

Tsoumas A., Alderman DJ & Rodgers CJ (1989) Aeromonas salmonicida: development of resistance to 4-quinolone antimicrobials Journal of Fish Diseases 12, 493^507 Vgsholm I., Djupvik HO., Willumsen FV., Tveit AM & Tan- gen K (1994) Infectious salmon anaemia (ISA) epide- miology in Norway Preventive Veterinary Medicine 19, 277^290.

Minimize the use of antibacterial and antiparasitic drugs P J Midtlyng et al Aquaculture Research, 2011, 42, 28–34

A characterization and sensitivity analysis of the

benthic biotopes around Scottish salmon farms with a

Thomas A Wilding

The Scottish Association for Marine Science, Dunsta¡nage Marine Laboratory, Oban, Argyll, UK

Correspondence: T A Wilding, The Scottish Association for Marine Science, Dunsta¡nage Marine Laboratory, Oban, Argyll, PA37 1QA,

UK E-mail: tom.wilding@sams.ac.uk

Abstract

Man’s impact on biodiversity depends on both the type

of intervention and the nature of the receiving

envir-onment In the context of salmon farming,

environ-mental impacts occur largely through the £ux of

farm-derived organic detritus to the seabed In order

to assess these potential impacts this research aimed

(a) to identify sensitive benthic habitats, particularly

those categorized as Biodiversity Action Plan (BAP)

habitats, in relation to salmon farming; (b) to identify

and classify the nature of the benthic habitats that

oc-cur around Scottish salmon farms and assess the

overlap between sensitive habitats and salmon farms;

and (c) to conduct a pilot investigation into the

abun-dance of the sea pen Pennatula phosphorea around a

single salmon farm The benthic habitat occurring

underneath Scottish salmon farms consisted,

predo-minantly, of muds (58%) and shelly sands (27%) with

the bed-type rock, also frequently occurring in close

proximity (though not directly underneath) salmon

farms Most farms were located in water between 20

and 50 m deep These ¢ndings are commensurate

with the Scottish salmon-farming industry being

lo-cated in relatively sheltered sea lochs, in close

proxi-mity to the shore Salmon farming was considered,

on the basis of spatial overlap and habitat sensitivity,

to pose a high risk to maerl and beds of Modiolus

mod-iolus and, on the basis of a lack of information,

shel-tered muddy gravels and the megafauna associated

with the mud in deep water BAP Sea pen abundance

was highly variable but was reduced in close

proxi-mity to the salmon farm.While abundance increased

to a maximum of 10 per transect at intermediate

dis-tances, at the farm peripheries numerous transects

were found, again, to contain no sea pens This may

have been as a consequence of the protection o¡ered

by the physical presence of the salmon cages againsttrawling that occurs in the vicinity of the farm

Keywords: ¢sh farming, benthic, Biodiversity tion Plan, sea pens (Pennatula phosphorea), Scotland

Ac-IntroductionSalmon farming is a signi¢cant employer in Scotland,particularly in remote rural areas where it makes aconsiderable contribution to rural economies How-ever, salmon farming has a number of impacts onthe receiving environment and, consequently, opera-tions are only permitted following extensive consul-tation with a variety of stakeholders including thosewith statutory responsibilities, such as Scottish Nat-ural Heritage, to protect the marine environment.Sea lochs have been favoured by ¢sh farmers inScotland as they o¡er high water quality in a rela-tively sheltered environment These waters hostmany habitats that have been designated ‘Biodiver-sityAction Plan’ habitats (BAPs) under UK’s adoption

of the Convention on Biological Diversity (http://www.ukbap.org.uk/NewPriorityList.aspx) These ha-bitats include maerl (coralline algae), seagrass, the bi-valve Limaria hians, beds of the native oyster Ostreaedulis and the horse mussel Modiolus modiolus, thereef-building worm Sabellaria, serpulid reefs, andareas of sheltered muddy gravels, sublittoral sandsand gravels and muds in deep water

Salmon farming impacts stem, primarily, from therelease of particulate matter (faeces and uneatenfood) into the water column This material dispersesAquaculture Research, 2011, 42, 35^40 doi:10.1111/j.1365-2109.2010.02675.x

r 2010 The Authors

around the farm with a substantial proportion

set-tling on the seabed The organic enrichment that is

associated with the build-up of this waste material

has a profound e¡ect on the macrobenthos and is

as-sociated with a decrease in species diversity and the

superabundance of opportunistic species (Pearson &

Rosenberg 1978; Findlay,Watling & Mayer 1995; Black

1998; Cromey, Nickell & Black 2002; Read &

Fer-nandes 2003)

The impact of ¢sh farms on the surrounding

benthic infauna can be modelled using the particle

tracking/resuspension model DEPOMOD (Cromey

et al 2002; Cromey, Nickell, Black, Provost & Gri⁄ths

2002) DEPOMOD predicts the benthic carbon £ux

around salmon cages, on the basis of farm biomass,

stocking density, local hydrography and depth

(among others), and these values are converted,

using an empirically based algorithm, to a prediction

of infaunal trophic index (ITI) The ITI is based on the

relative abundance of fauna exhibiting di¡ering

feed-ing strategies with low values indicatfeed-ing a high

pro-portion of deposit/specialist feeders and high values

indicating dominance by ¢lter and suspension

fee-ders In the UK, the ITI ranges between 0 (very highly

impacted) and 59 (unimpacted) (Maurer, Nguyen,

Robertson & Gerlinger 1999) Around salmon cages,

an ITI of 30 is considered to represent the allowable

zone of e¡ects and delimits the farms

accepted-im-pact footprint Typically, the modelled ITI forms a

ser-ies of contours around a ¢sh cage that is elongated

along the axis of the main current direction, thus

modelled ITI is a better indicator of ‘distance’ from

the farm than the actual linear farm distance

Proposed new ¢sh farms or existing farms that are

expanding their operations are required to provide

DEPOMOD model output as part of the SEPA

consent-ing process [under the Water Environment

(Con-trolled Activities) (Scotland) Regulations 2005]

While the impact of ¢sh farming on benthic infauna

is relatively well understood (Black 1998), much less

is known about the impacts on larger sessile or motile

species (megafauna) There has been no research

ex-amining the relationship between the ITI predicted

by DEPOMOD and benthic megafauna

The overall aim of this (ongoing) desk study is to

provide user-friendly and informed guidance to

as-sist in the evaluation of the likely impacts of salmon

farm operations that are occurring in close proximity

to BAP habitats This research was split into three

phases:

Phase 1: to evaluate the likely sensitivities of BAP

habitats to the close proximity of salmon farms

Phase 2: to evaluate the likely spatial overlap tween salmon farms and BAP habitats and to iden-tify those BAPs at risk

be-Phase 3: to investigate the response of the sea penPennatula phosphorea to the presence of a salmonfarm (pilot study reported here)

MethodsDi¡erent techniques were required to deliver objec-tives 1^3

Habitat sensitivitiesHabitat sensitivities were determined using MarLIN’s

‘Biology and Sensitivity Key Information’ for speci¢cbiotopes associated with relevant UKBAP habitats(http://www.marlin.ac.uk/sensitivityrationale.php).Key communities from each BAP habitat wereassessed in terms of their susceptibility to thesmothering and resultant hypoxia that occurs un-der salmon farms where sensitivity is based on thebalance between intolerance (e.g direct mortality)and recoverability (over time ranging betweeno1week to425 years) Overall, BAP habitat sensitivitywas based on the most sensitive community present.MarLin also provides some spatial distribution in-formation for BAP habitats and these were used toaugment the salmon farm site characterization work(see ‘Salmon farm site characterization’ below)

Salmon farm site characterizationThe ground type and physical environment charac-terizing Scottish salmon farms were evaluated usingthe following three data layers: (1) bathymetry data(GIS shape ¢le) from SeaZone (http://www.seazone.com/index.php); (2) ground type data from the Brit-ish Geological Survey (point data) (http://www.bgs.ac.uk/); and (3) survey data (point data) fromsalmon farmers (provided in compliance with SEPAregulations and available from SEPA, n 5105; http://www.sepa.org.uk/) The compliance-survey data de-tail the sediment type at control stations that are out-side any likely farm in£uence (between 0.35 and

5 km from the farm) but which are considered to present the background conditions at the farm site(i.e pre-farming benthic characteristics)

re-The compliance-survey data were not recorded

in a methodical manner and consisted of variousConsequences of salmon farming on benthic megafauna T A Wilding Aquaculture Research, 2011, 42, 35^40

classi¢cations of the seabed (ranging from ‘sand’ to

‘muddy-sand containing shells and decaying

sea-weed’) Before collation, these reports were converted

to a common format and the following classi¢cations

were included: shelly mud, ¢ne sand, medium/coarse

sand, shelly sand, stony sand and mud (no shells)

This classi¢cation was chosen as it encompassed a

majority of the reported substratum types to the

max-imum detail.Where a single site was reported as

hav-ing more than one substratum type, all were

recorded and given an equal weight (e.g a report of

‘sand and mud’ would score 0.5 each for sand

and mud) In total, 105 farms were included in this

analysis

ArcGISTMwas used to associate each salmon farm

with the nearest point/area in the relevant data layer

and summarize the associations

Megafaunal surveys

Assessments of the megafaunal changes occurring

around the salmon farms were made using a

drop-down video camera In this preliminary study, two

transects were surveyed, both from the SW end of

the salmon farm at Dunsta¡nage (5612703.600N,

5127051.400W, WGS84) The survey vessel was tiedo¡

at the cage edge and drifted down-wind The transect

was divided into 5 m sections, the locations of which

were recorded using dGPS (WGS84) For this

preli-minary analysis, only the sea pen P phosphorea is

re-ported ArcGISTM was used to associate the start

point of each transect with the DEPOMOD-predicted

ITI However, it must be noted that the predicted ITI

relates to the farm operating at its maximum mitted biomass and that these preliminary observa-tions occurred in the ¢rst year of the productioncycle (i.e when the volume of faecal material contri-buting to the benthic carbon £ux would be less thanthat used in the model predictions)

per-ResultsHabitat sensitivitiesDi¡erent biotopes were predicted to exhibit di¡erentsensitivities to the presence of salmon farms, themain impact occurring as a consequence of the sub-stratum-smothering e¡ects of salmon faeces Thelikely sensitivities, and con¢dence in that assess-ment, are combined site characterization work (see

‘Site characterization’) and shown in Table 1

Site characterizationWater depth and seabed type are important charac-teristics in determining the likely biological commu-nity occurring under a salmon cage Approximately40% of Scottish salmon farms are located in waterthat is between 20 and 50 m deep with approxi-mately 30% being located between 50 and 100 m;very few farms are located in close proximity to water

4100 m in depth

The BGS data indicated that the most commonground type in closest proximity to salmon cageswas equally mud and rock (each 30%) Sand was thenext most common substratum (23%)

Table 1 Relative sensitivities of BAP habitats and species, their predicted spatial overlap with ¢sh-farms, con¢dence (in tion to sensitivity) and subsequent risk; where con¢dence in low risk was considered higher

Spatial overlap derived from an assessment of MarLin’s data.

wWith respect to megafauna such as sea pens.

^, a knowledge gap (at the time of writing); BAP, Biodiversity Action Plan.

Aquaculture Research, 2011, 42, 35^40 Consequences of salmon farming on benthic megafauna T A Wilding

r 2010 The Authors

Fish farm surveys (n 5105) indicated that shelly

muds (31%) and muds (with no record of shells,

27%) were the most common substratum found at

re-ference stations around salmon farms Shelly sands

(23%) were the next most common with the rest

being classi¢ed as ¢ne stony sand (totalling 19%)

The ¢ndings at this early stage of investigation

(from site characterization work) were that the BAP

habitat most commonly found in close proximity with

salmon farms was mud in deep water followed by

sub-littoral sands and gravels (Table 1) The sensitivities of

the assessed BAP habitats, and the extent of the

over-lap with salmon farming are shown in Table 1

At this stage, the BAP habitats at most risk were

considered to be maerl, seagrass, mud in deep water

(with respect to megafauna), sheltered muddy gravels

and beds of M modiolus (Fig 1)

Impacts on the sea penP phosphorea ^

preliminary ¢ndings

Sea pen abundance was negatively associated with

cage proximity with most transects (13 out of 21),

within the ITI 30 contour, containing no sea pens

giving a mean abundance waso1sea pen per

trans-ect The number of transects containing no sea pens

was much reduced (two out of 23) at an ITI of 30^50

with a mean count of 3.6 per transect At sites further

from the salmon farm (e.g those with ITI450),

abundance was highly variable with many transects

again containing no sea pens (eight out of 31) but

some containing up to 10 per transect (mean density

was 3.7 per transect)

DiscussionSpecies/BAP sensitivitiesDi¡erent BAP habitats/species will demonstrate dif-ferent sensitivities to the presence of a salmon farm.Those species typically found in higher energy envir-onments, particularly ¢lter feeders, are likely to behighly sensitive to clogging e¡ect of particulate mate-rial derived from salmon farms (Hughes, Cook &Sayer 2005) and include maerl and associated fauna(Hall-spencer, White, Gillespie, Gillham & Foggo2006; Hall-Spencer & Bamber 2007) Beds of M mod-iolus (and associated fauna) may also be sensitive tothe interstitial hypoxia associated with the accumu-lation of salmon farm detritus

Mud in deep water and sheltered muddy gravelsBAPs are considered to be only moderately sensitive

to salmon farm impacts, primarily because such

fau-na are adapted to the deposition of ¢ne material thatnaturally occurs in such habitats (e.g echiuranworms; Hughes, Ansell, Atkinson & Nickell 1993).Much less is known about the impacts of salmonfarms on benthic megafaunal (Crawford, Mitchell &Macleod 2001) Addressing this knowledge gap wasthe focus of this research (preliminary ¢ndings aregiven in ‘Impacts on the sea pen P phosphorea ^ pre-liminary ¢ndings’ below)

Site characterizationAccording to salmon farm surveys, approximatelyhalf of Scottish salmon farms (based on a sample of

105 farms) are sited on mud (often containing shells),

6050

4030

2010

Depomod predicted Infaunal Trophic Index

Figure 1 The relationship between the observed sea pen Pennatula phosphorea counts and DEPOMOD-predicted infaunaltrophic index (based on maximum biomass) around the Dunsta¡nage farm site (Loch Linnhe, Scotland)

Consequences of salmon farming on benthic megafauna T A Wilding Aquaculture Research, 2011, 42, 35^40

with the other half being sited on sands,

predomi-nantly shelly sands However, the BGS data indicate

that rock is often found in close proximity to salmon

farms (although not necessarily directly

under-neath) The presence of shells and shell fragments in

a majority of samples from salmon farm surveys, and

the close proximity of rock (BGS data) is

commensu-rate with the general picture of farms being sited in

sheltered, productive sea loch environments in close

proximity to the shore This type of environment is

associated with steep environmental gradients, not

least in bathymetry and closely associated changes

in substratum type The impact of salmon farms on

this type of environment is likely to be complex, with

greater impact occurring in shallower environments

The would apply particularly to the smothering of

hard-substratum-based organisms, including

photo-synthetic algae in the photic zone (Airoldi 2003;

Schiel, Wood, Dunmore & Taylor 2006) growing on

the sides of sea lochs

The evidence suggests that the mud in deep water

BAP is that habitat mostly commonly occurring under

salmon farms This BAP hosts UK priority species such

as the sea pen Funiculina quadrangularis Consequently,

Phase 3 of this research has focused on the impacts of

salmon farms on the megabenthos associated with

deep water muddy habitats and these preliminary

re-sults focus on one member of that community, the sea

pen P phosphorea

Impacts on the sea penP phosphorea ^

preliminary ¢ndings

The sea pen P phosphorea is commonly found in Loch

Linnhe, including around the Dunsta¡nage salmon

farm site (pers obs.) The DEPOMOD-predicted ITI

re-ferred to here is based on the farm operating at its

maximum permitted biomass and the gradient in

ITI should be interpreted as a relative farm distance

that takes into account local hydrography With this

caveat in mind, there was a clear pattern of

increas-ing abundance of sea pens with increasincreas-ing ITI, the

mean abundance of sea pens being much lower (o1

per transect) at an ITIo30 compared with an

ITI430 (where the mean was approximately 3.6 per

transect) At an ITI of450, the number of transects

containing no sea pens increased and the sea pen

count was highly variable This variability may

be due to the patchy nature of the sea pen

distribu-tion and/or because in Loch Linnhe, ¢sh cages may

o¡er considerable physical protection to the benthic

environment from trawling for the Norway prawnNephrops norvegicus At the Dunsta¡nage site, an ITI

of450 corresponds to a cage edge distance of proximately 50 m and the decrease in sea pen abun-dance observed might be attributable to an increase

ap-in towed-gear ¢shap-ing activity occurrap-ing there

ConclusionsSeveral BAP habitats, including maerl, seagrass, Sa-bellaria reefs and beds of the bivalves M modiolus,

L hians and O edulis, are likely to be highly sensitive

to salmon farms The extent of the overlap betweensalmon farms and these habitats needs further inves-tigation but it appears that most of the salmon-farm-ing industry is located over the BAP ‘mud in deepwater’ This habitat hosts species about which little

is known in terms of their sensitivity to salmonfarms Our preliminary observations indicate that atleast one member of the megafaunal community, thesea pen P phosphorea, is negatively associated withsalmon farm proximity However, this impact did notappear to extend further than 50 m and there wasevidence that the physical structure of the ¢sh cagesmay o¡er protection to sea pen populations

ReferencesAiroldi L (2003) The e¡ects of sedimentation on rocky coast as- semblages Oceanography and Marine Biology 41, 161^236 Black K.D (1998) The environmental interactions asso- ciated with ¢sh culture In: Biology of Farmed Fish (ed.

by K.D Black & A.D Pickering), pp 284–326 Sheffield Academic Press, Sheffield, UK.

Crawford C.M., Mitchell I.M & Macleod C.K.A (2001) Video assessment of environmental impacts of salmon farms ICES Journal of Marine Science 58, 445^452.

Cromey C.J., Nickell T.D & Black K.D (2002) DEPOMOD ^ modelling the deposition and biological e¡ects of waste solids from marine cage farms Aquaculture 214, 211^239 Cromey C.J., Nickell T.D., Black K.D., Provost P.G & Gri⁄ths C.R (2002) Validation of a ¢sh farm waste resuspension model by use of a particulate tracer discharged from a point source in a coastal environment Estuaries 25, 916^929 Findlay R.H.,Watling L & Mayer L.M (1995) Environmental impact of salmon net-pen culture on marine benthic communities in Maine ^ a cases study Estuaries 18, 145^179.

Hall-Spencer J & Bamber R (2007) E¡ects of salmon ing on benthic Crustacea Ciencias Marinas 33, 353^366 Hall-Spencer J.,White N., Gillespie E., Gillham K & Foggo A (2006) Impact of ¢sh farms on maerl beds in strongly tidal areas Marine Ecology-Progress Series 326, 1^9.

farm-Aquaculture Research, 2011, 42, 35^40 Consequences of salmon farming on benthic megafauna T A Wilding

r 2010 The Authors