lable at ScienceDirect

Marine Environmental Research 106 (2015) 103e113

Contents lists avai

Marine Environmental Research

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

Using otolith microchemistry and shape to assess the habitat value ofoil structures for reef fish

Ashley M. Fowler a, *, Peter I. Macreadie b, c, David P. Bishop d, David J. Booth a

a Centre for Environmental Sustainability, School of the Environment, University of Technology Sydney, P. O. Box 123, 2007 Broadway, NSW, Australiab Plant Functional Biology and Climate Change Cluster, School of the Environment, University of Technology Sydney, P. O. Box 123, 2007 Broadway, NSW,Australiac School of Life and Environmental Sciences, Centre of Integrative Ecology, Deakin University, VIC, Australiad Elemental Bio-imaging Facility, University of Technology Sydney, PO Box 123, Broadway, NSW, Australia

a r t i c l e i n f o

Article history:Received 1 December 2014Received in revised form11 March 2015Accepted 16 March 2015Available online 17 March 2015

Keywords:DecommissioningRigs-to-reefsArtificial reefReef fishSerranidResidency

* Corresponding author.E-mail address: ashley.fowler@uts.edu.au (A.M. Fo

http://dx.doi.org/10.1016/j.marenvres.2015.03.0070141-1136/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

Over 7500 oil and gas structures (e.g. oil platforms) are installed in offshore waters worldwide and manywill require decommissioning within the next two decades. The decision to remove such structures orturn them into reefs (i.e. ‘rigs-to-reefs’) hinges on the habitat value they provide, yet this can rarely bedetermined because the residency of mobile species is difficult to establish. Here, we test a novel solutionto this problem for reef fishes; the use of otolith (earstone) properties to identify oil structures of resi-dence. We compare the otolith microchemistry and otolith shape of a site-attached coral reef fish(Pseudanthias rubrizonatus) among four oil structures (depth 82e135 m, separated by 9.7e84.2 km) onAustralia's North West Shelf to determine if populations developed distinct otolith properties duringtheir residency. Microchemical signatures obtained from the otolith edge using laser ablation inductivelycoupled plasma mass spectrometry (LA-ICP-MS) differed among oil structures, driven by elements Sr, Baand Mn, and to a lesser extent Mg and Fe. A combination of microchemical data from the otolith edge andelliptical Fourier (shape) descriptors allowed allocation of individuals to their ‘home’ structure withmoderate accuracy (overall allocation accuracy: 63.3%, range: 45.5e78.1%), despite lower allocation ac-curacies for each otolith property in isolation (microchemistry: 47.5%, otolith shape: 45%). Site-specificmicrochemical signatures were also stable enough through time to distinguish populations during 3separate time periods, suggesting that residence histories could be recreated by targeting previousgrowth zones in the otolith. Our results indicate that reef fish can develop unique otolith propertiesduring their residency on oil structures which may be useful for assessing the habitat value of individualstructures. The approach outlined here may also be useful for determining the residency of reef fish onartificial reefs, which would assist productivity assessments of these habitats.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Over 7500 oil and gas structures (e.g. rigs, platforms; hereafter‘oil structures’) are installed in offshore waters throughout theglobe, and many of these will require decommissioning over thenext two decades (Parente et al., 2006). Decommissioning of oilstructures has historically involved complete removal, because thiswas considered best environmental practice (Jørgensen, 2012).However, we now know that oil structures provide habitat for a

wler).

wide range of marine organisms, with some structures supportingregionally-significant populations of fish and invertebrates(Macreadie et al., 2011, 2012). The apparent habitat value of oilstructures prompted the development of ‘rigs-to-reefs’ (RTR) as analternative decommissioning option to complete removal. RTR in-volves leaving all or part of an obsolete structure in the marineenvironment as an artificial reef. The practice has becomeincreasingly popular since its inception in Florida in 1979,spreading throughout southern US states and internationally toBrunei and Malaysia (Macreadie et al., 2011).

RTR decisions may be made on a case-by-case basis, becausehabitat value is likely to vary greatly among oil structures(Schroeder and Love, 2004; Fowler et al., 2014). Substantial

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113104

differences in the abundance and diversity of marine communitieshave been observed among similar structures located in the samegeographical area (Love et al., 2000; Stanley andWilson, 2000; Loveet al., 2005; Martin and Lowe, 2010; Pradella et al., 2014). Forexample, densities of mature bocaccio rockfish associated withPlatformGail in southern Californiawere found to be 8 times higherthan densities on similar platforms located within 50 km (Loveet al., 2005). In contrast, oil structures with minimal habitat valuemay not be suitable for RTR conversion, because the impacts oftheir deployment on surrounding ecosystems (e.g. scattering ofdebris) may outweigh any environmental benefit they provide (seeFowler et al., 2014).

Determining the habitat value of oil structures for mobile taxaremains a major challenge, because residency of individuals canrarely be established. Mobile reef fishes (e.g. lutjanids, carangids)are often observed in high abundance on oil structures (Seamanet al., 1989; Love et al., 2000; Gallaway et al., 2009; Pradella et al.,2014), leading to the suggestion that oil structures contributesubstantially to population biomass for such taxa (e.g. Gallawayet al., 2009). However, observed individuals may simply be tran-sient visitors rather than permanent residents. In such cases, thehabitat resources provided by an oil structure may contribute littleto population biomass (see Bohnsack, 1989 for a discussion of the'attraction vs production' debate). Tagging studies have been con-ducted for some species to assess the degree of residency on oilstructures, with high site fidelity (80e100%) observed for redsnapper (Lutjanus campechanus) on oil platforms in the northernGulf of Mexico for periods of up to 12 months (Gallaway, 1981;Peabody and Wilson, 2006). While tagging methods are useful forobtaining high-resolution data on residency, they are extremelytime-consuming, and often involve very low recapture or detectionrates (Elsdon and Gillanders, 2003). Tagging is therefore unlikely toprovide a practical tool for assessing the residency of reef fish on oilstructures, given the large number of structures approachingdecommissioning, and the numerous species requiring assessmenton each structure.

Microchemical analysis of fish otoliths (earstones) may providea useful alternative for assessing residency on oil structures. Traceelements in ambient seawater are incorporated into the calciumcarbonate matrix of the otolith, providing a record of residency(Campana, 1999). Due to variations in ambient water chemistry,concentrations of trace elements in the otolith can differ amonglocations, producing unique microchemical ‘signatures’ (Campanaet al., 2000; Gillanders and Kingsford, 2000). If such signaturesdiffer among oil structures, microchemical analysis may be used toidentify which structure an individual resided on prior to capture.Analysis of sequential growth increments within the otolith mayalso indicate the extent of time spent on a particular structure,because increments are deposited with regular periodicity (dailyand annually, Campana and Thorrold, 2001). Many studies haveused otolith microchemistry to distinguish spatially-segregatedpopulations in natural ecosystems (e.g. Campana et al., 1994;Gillanders and Kingsford, 2003; Ashford et al., 2006; Fergusonet al., 2011), yet only two have investigated the otolith micro-chemistry of fish associated with oil structures (Love et al., 2009;Nowling et al., 2011). These studies focused on distinguishing in-dividuals associated with oil structures from those associated withother types of habitat (natural or artificial reef), rather than dis-tinguishing among particular oil structures. Additionally, neitherstudy investigated the temporal consistency of microchemicalsignatures at oil structures, despite well-known inter-annual vari-ability of signatures in natural systems (Gillanders, 2002;Ruttenberg et al., 2008). Further investigation of the potential forotolith microchemistry to identify oil structures of residence forreef fish is warranted, given that microchemical analysis is likely to

be considerably less time-consuming and cheaper than taggingmethods for residency assessment.

Otolith shape may also be useful for identifying oil structuresof residence, because otolith shape is known to vary amongspatially-segregated populations (Campana and Casselman,1993). Allometric growth patterns in the otolith are affected bynumerous environmental factors, including temperature(Lombarte and Lleonart, 1993), depth (Wilson, 1985), habitat type(Vignon et al., 2008), and feeding history (Gagliano andMcCormick, 2004). If environmental conditions differ suffi-ciently among oil structures, individuals may develop otolithshapes that are unique to particular structures. Shape analysis ofotolith outlines is a rapid technique that does not require furtherpreparation of the otolith beyond extraction and cleaning.Furthermore, otolith shape may reflect population differentiationoccurring over a broader timescale than otolith microchemistry,because otolith shape is the product of allometric growth pro-cesses occurring throughout the life of the fish, while micro-chemical analyses often employ laser ablation to target smallregions of the otolith corresponding to �6 months of accretion(Ferguson et al., 2011). Otolith shape analysis may thereforeprovide a practical compliment to other methods of residencyassessment that involve sampling of individuals, such as micro-chemical analysis. The use of more than one method for popu-lation discrimination has previously been recommended (Beggand Waldman, 1999), because individuals from different pop-ulations may be distinguished with greater accuracy (Fergusonet al., 2011). The potential for otolith shape to distinguishamong fish populations residing on different oil structures hasnot previously been investigated.

Here, we assess the utility of otolith properties for identifying oilstructures of residence for coral reef fish. An opportunistic collec-tion of Pseudanthias rubrizonatus from isolated oil structures onAustralia's NorthWest Shelf (NWS) allowed us to examine whetherreef fish populations can develop distinct microchemical signaturesand otolith shapes during their residency. These populations wereuseful for such a purpose, because P. rubrizonatus is a small site-attached reef species, and it is highly likely that individualssettled onto the structures following their pelagic larval phase andremained there until they were collected (see Fowler and Booth,2012). Inter-reef excursions of small reef fishes are generallyrestricted to distances of �50 m (Frederick, 1997; Overholtzer-McLeod, 2005; Turgeon et al., 2010), and ROV inspectionsconfirmed that each structurewas located on soft sediment with nohard substrate located within at least 100 m (Fowler and Booth,2012). Furthermore, the congeneric Pseudanthias squamipinnis isknown to remain within 20 m of its ‘home’ rock or coral outcrop(Popper and Fishelson 1973). The residence history of each popu-lation in the current study can therefore be assumed, and can belinked to otolith properties. The distances between the structuresalso reflect a spatial scale relevant to oil installations (10e100 km),which is an order of magnitude smaller than the scale of moststudies attempting to distinguish fish populations using otolithproperties (Elsdon et al., 2008).

We first examine whether microchemical signatures laiddown immediately prior to capture (on the otolith edge) could beused to distinguish individuals of P. rubrizonatus residing ondifferent structures. We then compare the discrimination accu-racy achieved using edge microchemistry to that achieved usingelliptical Fourier descriptors of otolith shape, as well as the ac-curacy achieved using both datasets combined. Lastly, we assessthe consistency of site-specific microchemical signatures throughtime by examining signatures contained in previous growth re-gions of the otolith.

Table 1Sampling summary of Pseudanthias rubrizonatus from 4 well heads located onAustralia's North West Shelf (NWS). Age data were obtained from Fowler and Booth(2012).

Structurename

Deploymentyear

Depth (m) Distanceto neareststructure(km)

Numberof fish

Size range(mm, LS)

Age range(years)

WH1 1992 82 9.7 15 26e56 0e3WH2 1992 84 9.7 32 22e73 0e3WH3 1986 133 31.1 51 22e82 0e4WH4 1990 135 31.1 22 31e97 0e5

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113 105

2. Materials and methods

2.1. Study location and sample collection

Pseudanthias rubrizonatus were collected from 4 obsolete well-heads located within a 180 km2 area on the outer NWS, c. 125 kmNNE of Barrow Island (20� 470 S; 115� 240 E; Fig. 1). Well-heads(WH) are complex steel structures installed on the seabed toallow safe extraction of hydrocarbons from wells drilled into theseafloor. The well-heads in this study were originally installed be-tween 1986 and 1992, and were due for removal by the operatingcompany at the time of this study. They were separated by dis-tances ranging from 9.7 (WH1 and WH2) to 84.2 km (WH1 andWH4), and were located at depths between 82 and 135 m (Table 1).The well-heads had dimensions of 2.6 � 2.6 � 4 m (last dimensionis height) and contained numerous holes and overhangs, providingcomplex habitat for a range of reef fish species (Pradella et al.,2014). Benthic habitat in the region is characterised by fine sandand mud (Lyne et al., 2006).

Well-heads were removed from the seabed using a deck craneand sling system over 3 cruises between late June and late August,2008. WH4 was removed on the 1st cruise (June), while the other 3well-heads were removed on the 3rd cruise (August). A small(0.5� 0.5 m) section of WH3 was removed on the 2nd cruise (July);however, no individuals were caught during the removal. Removalof well-heads on the 3rd cruise was completed over a 3-day period.Following recovery of well-heads to the survey vessel, all P. rubri-zonatus individuals present inside each structure were collectedand euthanized in an ice-slurry. Collected individuals were storedat �20 �C and transferred to the laboratory, where standard length(LS) was measured to the nearest 0$1 mm, and wet weight (W) wasmeasured to the nearest 0$0001 g. Sagittal otoliths were extractedusing forceps, cleaned in Milli-Q water and stored dry.

Fig. 1. Locations of four well heads from which Pseudanthias rubrizon

2.2. Otolith preparation

Otoliths from 120 individuals were used for microchemical andshape analysis (Table 1), following subsampling at abundant sites,and exclusion of juveniles that had recruited too recently to have asufficient post-settlement region in their otolith (�21 mm LS). Thefinal 120 individuals included all individuals from WH1 and WH4,and a sub-sample of individuals from the two most abundant sites,WH2 and WH3. Sub-sampling was stratified within 10 mm LS binsto ensure all size classes were represented. Size ranges overlappedamong sites, however maximum size varied (Table 1).

The same otolith from each individual was used for bothmicrochemical and otolith shape analysis. Left otoliths were usedwherever possible to avoid potential confounding effects of differ-ences between left and right otoliths. Otoliths were first imagedwhole at 50�magnification using a stereomicroscope (LeicaMZ16)and image analysis system (Leica IM500 v. 4, Leica Microsystems)with reflected light on a black background. Otoliths were posi-tioned so that the distal surface faced the camera and the ros-tralepostrostral axis aligned with the vertical axis of the image. In

atus were collected on the North West Shelf (NWS) of Australia.

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113106

addition to shape analysis, images were used to estimate the age ofindividuals in years by counting alternating opaque and translucentincrements. Details of ageing methods and length-age structure ofthe populations are described in Fowler and Booth (2012).

After imaging, otoliths were sectioned for microchemical anal-ysis. Otoliths were mounted on glass slides using thermoplasticcement (Crystalbond™ 509), so that the rostrum protruded overthe edge of the slide. The rostrum was ground down to the pri-mordium using 15 and 3 mm lapping films in that order. The otolithwas then repositioned in the centre of the slide rostral-surface-down and the postrostrum was ground down to produce a thin(~200 mm) transverse section through the primordium. Sectionswere rinsed with Milli-Q water to remove grinding residue, soni-cated for 3 min to remove remaining surface contaminants anddried in a laminar flow cabinet. Due to the pre-ablation proceduredescribed below, further decontamination procedures were notnecessary prior to analysis.

2.3. Otolith microchemistry

Trace element concentrations in the otoliths were determinedby laser ablation inductively coupled plasma mass spectrometry(LA-ICP-MS) using a New Wave UP-213 laser ablation unit with aLarge Format Cell (Kenelec Scientific) coupled to an Agilent Tech-nologies 7500ce ICP-MS. A transect of discrete ablations was madefrom the nucleus to the edge of the otolith along the longest radiusof the section (Fig. 2). The position of the transect was standardisedamong samples to reduce potential error resulting from spatialvariation in elemental concentrations throughout the otolith ma-trix (Hamer and Jenkins, 2007). A pre-ablation of 100 mm diameterwas made at each ablation site for 2 s at 1 Hz to remove anyremaining surface contaminants. Ablations of 40 mmdiameter werethen made for 20 s at 10 Hz using maximum laser output(13.8 ± 3.5 J cm�2), following a warm-up period of 10 s and a blankacquisition period of 10 s. A ‘washout’ period of 20 s was madebetween each ablation. The ablation diameter was selected as acompromise between detection power and temporal resolution ofmicrochemical signatures. Ablations were separated by only 10 mmalong the transect to ensure at least one ablation fell within eachgrowth increment. To account for instrument drift, 3 ablations weremade on a glass reference standard (NIST 612, Pearce et al., 1997)between each sample and samples were analysed in a random or-der. Instruments were tuned prior to each run using the referencestandard to ensure that greatest sensitivity and lowest abundanceof oxide interferences were achieved.

Sixteen elements were initially assayed, including thosecommonly used in population discrimination studies (Li, Mg, Ca,Mn, Fe, Zn, Sr, Ba, Pb), as well as those likely to be present at oil andgas fields in drill spoil or drill lubricants (V, Co, Ni, Cu, Ag, Cd, U;

Fig. 2. Otolith section of P. rubrizonatus showing the transect of discrete laser ablationsextending from the nucleus to the otolith edge. Arrows indicate ablations selected foreach growth period, as well as the edge ablation.

Nowling et al., 2011). Six elements in addition to Ca were consis-tently present above limits of detection (LOD, Table 2) and thesewere analysed for all samples. The selected elements were aboveLOD in �77% of all samples, with Mg, Fe, Sr and Ba above LOD in100% of samples (Table 2). Values that fell below LOD for selectedelements were included in analyses, because LOD thresholds arearbitrarily defined, and these values are still potentially useful fordistinguishing populations (Ben-Tzvi et al., 2007). Estimates of in-strument precision (% relative standard deviation, RSD) were madefrom repeated analyses of the NIST 612 standard (Table 2).

Count-per-second data were background corrected and con-verted to concentrations (mg g�1) using the data reduction softwareGlitter (GEMOC, Macquarie University). 43Ca was used as the in-ternal standard and NIST 612 was used as the external standard.Elemental values in the sample and the NIST standard were firstnormalised to 43Ca and then converted to quantities using knownconcentrations in the NIST standard. Although the NIST standarddoes not match the otolith matrix, it has been shown to be suitablefor quantifying elemental concentrations in otoliths (McGowanet al., 2014). Glitter also calculated LODs within 99% confidenceusing the equation: LOD ¼ 2:3�

ffiffiffiffiffiffi

2Bp

, where B indicates totalcounts in the background signal period (Longerich et al., 1996).

Otolith sections were imaged following microchemical analysisto assess quality and select ablations used for statistical analyses.The ablation closest to the edge of the otolith was selected torepresent the period immediately before capture (Fig. 2). Thesewere examined to ensure they were not contaminated by theadjacent thermoplastic cement. On the few occasions where thisoccurred, data from the next closest ablation to the edge were usedto avoid potential confounding effects of edge contamination. Pre-and post-ablation images were used in combination to identifyablations located within target growth regions of the otolith fortemporal analyses (see Statistical Analysis section). Only ablationslocated completely within target growth regions were selected. Asingle ablation per growth region was used for analysis.

2.4. Otolith shape

Otolith shape was quantified using elliptical Fourier analysis(EFA). EFA deconstructs the outline of a two-dimensional objectinto a series of harmonically related ellipses that can be used incombination to approximate the object's shape (Kuhl and Giardina,1982). EFA is a more powerful method of shape description thanfast Fourier transformation (FFT), despite being used less frequentlyfor otoliths (M�erigot et al., 2007).

Forty harmonics were generated for each otolith using thesoftware SHAPE (v. 1.3, Iwata and Ukai, 2002). This software con-verts images into binary silhouettes, extracts the otolith outline,and produces normalised harmonics for use in multivariate ana-lyses. To determine how many harmonics were needed to suffi-ciently describe otolith shape, the average Fourier power spectrumwas calculated for 40 randomly-selected individuals; 10 from each

Table 2Mean limits of detection (LOD), percentage of values below LOD and estimated in-strument precision (RSD) for elements selected for microchemical analysis of P.rubrizonatus otoliths.

Element Isotope LOD (mg g�1) % < LOD RSD (%)

Li 7Li 1.837 23 5Mg 24Mg 0.676 0 11Mn 55Mn 1.001 23 3Fe 57Fe 6.706 0 9Sr 88Sr 0.097 0 3Ba 137Ba 0.553 0 5

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113 107

oil structure (Crampton, 1995). Fourier power (PFn) indicates theamount of shape information described by each harmonic and isexpressed by the equation: PFn ¼ ðA2

n þ B2n þ C2n þ D2

n�2, where An,Bn, Cn and Dn are the coefficients of the nth harmonic (Pothin et al.,2006). The cumulated power percentage (PFc) was then calculatedby summing PFn for all harmonics. The first 13 harmonics explained>99.99% of otolith shape, and additional harmonics were thereforeeliminated from further analyses. Retained harmonics provided 49Fourier shape coefficients per individual for multivariate analyses(see below); 4 coefficients from each harmonic, less the first 3 co-efficients (A1, B1 and C1) that are rendered invariant among in-dividuals due to normalisation.

2.5. Statistical analyses

Otolith microchemistry and shape data were compared amongoil structures using multivariate PERMANOVA (PERMANOVAþ, v.1.0, PRIMER-E). This procedure uses permutations to test for sig-nificant differences between groups and therefore does not assumedata normality or homogeneity of variances (Anderson et al., 2008).PERMANOVA is also preferred to other multivariate procedures(e.g. ANOSIM) when testing multi-factorial designs because itprovides a test for interaction terms.

2.5.1. Edge microchemistry and otolith shape analysesTo determine if microchemical signatures laid down immedi-

ately prior to capture differed among oil structures, trace elementconcentrations from edge ablations were compared using a one-factor PERMANOVA (Fig. 2). Data were fourth-root transformed tobalance the contribution of more abundant and less abundant el-ements to dissimilarity values, and Euclidian distances were usedto generate the dissimilarity matrix. Ls was used as a covariate,because microchemistry of otoliths in known to vary throughoutdevelopment (Beer et al., 2011). The relationship between the co-variate and the concentration of each element was inspected priorto the main analysis to ensure distributions were linear and nooutliers were present (Anderson et al., 2008). This was confirmedfollowing fourth-root transformation. Type I sum of squares wasused in the analysis so that the factor ‘site’ was fitted to the dataafter the covariate. Permutations were conducted on residualsunder a reduced model, rather than on raw data, to avoid inflatedType 1 error rates associated with covariates in multivariate ana-lyses (Anderson et al., 2008). P-values were generated using 9999permutations. The same experimental design and model settingswere used to compare Fourier shape coefficients among oil struc-tures. However, data for these analyses were normalised, ratherthan fourth-root transformed. Ls was retained as a covariate in theshape analysis because otolith shape can change throughoutdevelopment, potentially confounding comparisons among pop-ulations with different age-length structures (Campana andCasselman, 1993). A p-value < 0.05 was considered significant forall tests.

Canonical analysis of principal coordinates (CAP; PRIMER v. 6.1,PRIMER-E) was used to visualise multivariate differences inmicrochemical signatures and otolith shape between oil structures,and to determine how accurately individuals could be allocated totheir structure of origin. CAP is a constrained ordination that findsaxes which best separate groups defined a priori (Anderson et al.,2008). The procedure uses the ‘leave-one-out’ (cross-validation)method to determine the accuracy (%) with which individuals canbe allocated to their original group. In the current study, an allo-cation success of 25% would be expected by chance alone (4groups). Vector overlays were used to visualise the elementsdriving differences among groups. Vectors were based onSpearman rank correlations between the elemental concentrations

and primary CAP axes, with the size and direction of the vectorcorresponding to the strength and sign of the relationship,respectively (Anderson et al., 2008). Only correlations � ±0.3 wereincluded as vectors. Vector overlays were not produced for shapecoefficients, due to the large number of coefficients used. Combinedmicrochemical and otolith shape datawere normalised prior to CAPanalysis.

2.5.2. Temporal microchemical analysesBecause P. rubrizonatus populations could only be sampled once,

the temporal consistency of microchemical signatures was testedby examining signatures contained in previous growth regions ofthe otolith. Two separate PERMANOVA analyses were performed:1) signatures were compared among oil structures during multipleprevious growth periods for a single age cohort, and 2) signaturescontained in the ‘Age 0’ region of the otolith were compared amongmultiple cohorts at a single oil structure (WH3). The first analysistested whether structures could be consistently distinguished overmultiple time periods and avoided the potential confounding effectof age on comparisons among structures. However, this analysiscould not be used to determine if site-specific signatures remained‘the same’ across time periods, because time was potentiallyconfounded with developmental stage. This distinction wasexplored using the second analysis, where comparisons throughtime were restricted to a single developmental stage (Age 0).

The Age 1þ cohort was selected for the first temporal analysis,because this was the only cohort with a sufficient number of in-dividuals acrossmultiple structures (n¼ 7,14 and 11 forWH1,WH2and WH3, respectively). WH4 was excluded from the analysisbecause too few one-year-olds were collected from this structure. Arepeated-measures design was used to compare signatures be-tween oil structures and growth periods, with both factors treatedas fixed (Quinn and Keough, 2002). The factor ‘growth period’ hadthree levels: Post-settlement, Age 0 and Age 1. The post-settlementgrowth period was represented by the first ablation visible beyondthe settlement mark (Fig. 2). The Age 0 growth period was repre-sented by the last ablation prior to the 1st annual increment, whilethe Age 1 growth period was represented by the last ablation priorto the 2nd increment (Fig. 2). The time period between the post-settlement ablation and the Age 0 ablation may have been lessthan one year, depending on the timing of settlement relative to thedeposition of the first annual increment (see Fowler and Booth,2012). The entire period encompassed by the 3 targeted growthzones was ~1.5e2 years (see Fowler and Booth, 2012).

The second temporal analysis was restricted to WH3, becausethis was the only structurewith sufficient individuals frommultiplecohorts (n ¼ 11, 10 and 13 for Age 1þ, 2þ and 3þ cohorts, respec-tively). Elemental concentration data from the Age 0 region of theotolith (as described above) were compared among age cohortsusing a one-factor PERMANOVA. Data transformation and modelsettings for the two temporal analyses were the same as those usedfor the edge analysis (see above); however, LS was not required as acovariate, so Type III sum-of-squares was used instead of Type I.

PERMDISP procedures were run for all significant PERMANOVAresults to test for homogeneity of multivariate dispersions. PER-MANOVA is sensitive to differences in both the location anddispersion of multivariate data, so a significant PERMANOVA resultindicates either a difference in data location between treatmentgroups, or a difference in data dispersion, or both (Anderson et al.,2008). In multivariate cases where PERMDISP was significant, CAPplots were examined to assess whether a difference in data locationwas apparent in addition to a difference in dispersion. For the Sranalysis among oil structures, confidence intervals were used toassess the likelihood of a difference in data location in addition to adifference in dispersion.

Table 4Allocation success of P. rubrizonatus individuals back to their structure of originusing a) trace elemental composition on the edge of the otolith (edge signature), b)Fourier shape coefficients obtained from the otolith outline, and c) combined edgesignatures and shape coefficient data for 4 oil structures located on Australia's NorthWest Shelf (NWS).

Original structure Allocated structure

WH1 WH2 WH3 WH4 Total % Correct

a)WH1 10 3 2 0 15 66.7WH2 11 13 3 5 32 40.6WH3 2 6 23 20 51 45.1WH4 1 4 6 11 22 50.0b)WH1 7 3 3 2 15 46.7WH2 4 17 7 4 32 53.1WH3 3 13 23 12 51 45.1WH4 3 7 5 7 22 31.8c)WH1 9 4 2 0 15 60.0WH2 2 25 2 3 32 78.1WH3 4 5 32 10 51 62.7WH4 2 1 9 10 22 45.5

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113108

3. Results

3.1. Edge microchemical signatures among structures

Multi-elemental signatures on the edge of the otolith differedamong oil structures (PERMANOVA, pseudo-F3,115 ¼ 10.39,p < 0.001). Edge signatures also varied significantly with LS (PER-MANOVA, pseudo-F1,115 ¼ 42.77, p < 0.001), yet differences amongstructures were still apparent following partitioning of this effect inthe model. Pairwise comparisons indicated that edge signaturesdiffered between all structures exceptWH3 andWH4 (Table 3a). Nodifferences in multivariate dispersions of microchemical data werefound among oil structures (PERMDISP, pseudo-F3,116 ¼ 1.09,p ¼ 0.412).

Despite significant differences in edge signatures among oilstructures, overall allocation accuracy of individuals to theirstructure of origin was low (47.5%), with accuracies for individualstructures ranging from 40.6 to 66.7% (CAP, Table 4a, Fig. 3a).Misallocated individuals from WH1 were primarily allocated toWH2 (next closest structure, Fig. 1) and vice versa, while mis-allocated individuals from WH3 were primarily allocated to WH4(next closest structure, Fig. 1) and vice versa (Table 4a). CAP Axis 1was primarily responsible for separating the two western-most(deeper) structures from the two eastern-most (shallower) struc-tures, while CAPAxis 2 was primarily responsible for separating thetwo eastern structures from each other (Fig. 3a). Vector overlaysindicated that group separation on CAPAxis 1 was primarily drivenby Sr and Ba, while separation on CAP Axis 2 was driven by Sr, Ba,Mn and Mg, and to a lesser extent Fe (Fig. 3a).

Univariate analyses indicated that Sr, Ba, and Mn concentrationson the otolith edge differed among oil structures (PERMANOVA,Table 5). PERMANOVA indicated lower Sr concentrations at WH1than at all other structures (PERMANOVA, pairwise, all p � 0.006;Fig. 4a), while other structures did not differ from each other (allp � 0.397). However, multivariate data dispersions were also foundto differ among oil structures for this element (PERMDISP, pseudo-F3,116 ¼ 3.14, p¼ 0.028), indicating that the significant PERMANOVAresult may have resulted from differences in data spread, ratherthan an actual difference in elemental concentrations. Inspection of95% confidence intervals for mean Sr concentrations indicated nooverlap between intervals forWH1 and any other structure, with anupper confidence bound for WH1 of 2721 mg g�1 and lower boundsof 2741, 2813 and 2737 mg g�1 for WH2, WH3 and WH4, respec-tively. This finding supports the PERMANOVA result of lower Sr

Table 3Pairwise PERMANOVA comparisons of a) trace elemental composition on the edge ofthe otolith (edge signature) and b) Fourier shape coefficients obtained from theotolith outline between populations of P. rubrizonatus inhabiting 4 oil structureslocated on Australia's North West Shelf (NWS).

Analysis Comparison Distancebetweenstructures(km)

df Pseudo-t p

a) Edgesignatures

WH1, WH3 55.0 63 3.00 <0.001WH1, WH2 9.7 44 1.75 0.027WH1, WH4 84.2 34 2.89 <0.001WH3, WH2 45.7 80 4.15 <0.001WH3, WH4 31.1 70 1.06 0.320WH2, WH4 74.6 51 4.15 <0.001

b) Shapecoefficients

WH1, WH3 55.0 63 1.72 0.008WH1, WH2 9.7 44 1.07 0.295WH1, WH4 84.2 34 0.73 0.894WH3, WH2 45.7 80 0.92 0.518WH3, WH4 31.1 70 1.33 0.081WH2, WH4 74.6 51 1.41 0.053

concentrations in otoliths at WH1 at the p ¼ 0.05 level.Ba concentrations were higher at the two eastern-most (shal-

lower) structures (WH1 and WH2) than at the two western-moststructures (WH3 and WH4; PERMANOVA, pairwise, all p � 0.001;Fig. 4b). Mn concentrations were lower at WH2 than all otherstructures (PERMANOVA, pairwise, all p � 0.007; Fig. 4c), whileother structures did not differ from each other (all p � 0.057).Multivariate dispersions did not differ among structures for eitherBa (PERMDISP, pseudo-F3,116 ¼ 2.21, p ¼ 0.118) or Mn (PERMDISP,pseudo-F3,116 ¼ 0.56, p ¼ 0.671). Concentrations of Sr, Ba, Mn, Mgand Fe also varied significantly with body length (PERMANOVA,covariate, Table 5).

3.2. Otolith shape, and combined microchemical and shape data,among structures

PERMANOVA indicated that otolith shape, as described byelliptical Fourier coefficients, differed among oil structures(pseudo-F3,115 ¼ 1.70 p ¼ 0.032). However, pairwise comparisonsonly identified differences between WH1 andWH3, with WH2 andWH4 approaching significance (Table 3b). Although multivariatedata dispersions also differed between WH1 and WH3 (PERMDISP,pseudo-t ¼ 2.94, df ¼ 64, p ¼ 0.006), the CAP plot indicated a dif-ference in the location of the two data groups with only minor dataoverlap (Fig. 3b), supporting a difference in otolith shape betweenWH1 and WH3. Otolith shape also varied significantly with LS(PERMANOVA, pseudo-F1,115 ¼ 103.36, p < 0.001).

Overall allocation accuracy using otolith shapewas similar to thatfor edge signatures (45%, CAP, Fig. 3b); however, accuracies for in-dividual structures differed considerably (Table 4b). Allocation ac-curacy forWH1 was lower than that achieved using edge signatures(46.7% compared to 66.7%), as was the accuracy for WH4 (31.8%compared to 50%, Table 4a andb). Accuracy forWH2was higher thanthat achieved using edge signatures (53.1% compared to 40.6%),while accuracy for WH3 was the same between the two methods(45.1%, Table 4a andb). No clear pattern ofmisallocationwas evident,with misallocated individuals spread more evenly across otherstructures than misallocations using edge signatures (Table 4b).

Combining microchemistry and shape data improved allocationsuccess, with an overall accuracy of 63.3% achieved (CAP, Fig. 3c,Table 4c). Allocation accuracies for individual structures rangedfrom 78.1% at WH2 to 45.5% at WH4 (Table 4c).

Fig. 3. CAP plots for a) trace elemental composition on the edge of the otolith (edgesignature), b) Fourier shape coefficients obtained from the otolith outline, and c)combined edge signatures and shape coefficient data for P. rubrizonatus inhabiting 4 oilstructures located on Australia's North West Shelf (NWS). Vector overlays in Plot a)indicate correlations of individual elements with primary axes. Note different axisscales on each plot.

Table 5Univariate PERMANOVA comparisons of elemental concentrations on the edge ofthe otolith of P. rubrizonatus among 4 oil structures located on Australia's NorthWestShelf (NWS). Body length (LS, mm) was used as a covariate in all analyses.

Element Model term df Pseudo-F p

Sr Oil structure 3 4.85 0.004Length 1 37.66 <0.001

Ba Oil structure 3 33.81 <0.001Length 1 51.61 <0.001

Mn Oil structure 3 9.45 <0.001Length 1 123.02 <0.001

Mg Oil structure 3 2.34 0.078Length 1 37.98 <0.001

Li Oil structure 3 1.53 0.223Length 1 0.04 0.838

Fe Oil structure 3 2.61 0.056Length 1 5.49 0.021

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113 109

3.3. Temporal consistency of microchemical signatures

3.3.1. Comparison among oil structures across growth periodsMicrochemical signatures of the Age 1þ cohort differed among

oil structures (PERMANOVA, pseudo-F2,87 ¼ 4.72, p < 0.001), andthese differences were consistent across the 3 growth periods(PERMANOVA, structure � growth period, pseudo-F4,87 ¼ 1.02,p ¼ 0.429). Signatures also differed among growth periods (PER-MANOVA, pseudo-F2,87 ¼ 5.97, p < 0.001), yet not enough toobscure differences between structures. Pairwise comparisonsindicated that signatures differed between WH1 and WH3(pseudo-t ¼ 2.20, df ¼ 48, p ¼ 0.004), and WH2 andWH3 (pseudo-t ¼ 2.73, df ¼ 69, p < 0.001), but not between WH1 and WH2(pseudo-t ¼ 1.18, df ¼ 57, p ¼ 0.226). Signatures differed between

all growth periods (all p � 0.049).Multivariate data dispersions did not differ among oil structures

(PERMDISP, pseudo-F2,93 ¼ 1.00, p ¼ 0.402), but did differ amonggrowth periods (PERMDISP, pseudo-F2,93 ¼ 5.44, p ¼ 0.007). Pair-wise tests indicated that dispersions differed among the post-settlement period and both the Age 0 (PERMDISP, pseudo-t ¼ 2.22, df ¼ 64, p ¼ 0.035) and Age 1 (PERMDISP, pseudo-t ¼ 3.61,df ¼ 64, p ¼ <0.001) periods, indicating that significant PERMA-NOVA results between these periods may have resulted from dif-ferences in data spread, rather than an actual difference inmicrochemical signatures. Inspection of the CAP plot indicated thatthe post-settlement signature likely differed to that of the Age 1period, but not to that of the Age 0 period, given the relative dataoverlap of these groups (Fig. 5).

3.3.2. Comparison among age cohorts at a single structureMicrochemical signatures in the Age 0 region of the otolith

differed among sequential age cohorts at WH3 (PERMANOVA,pseudo-F2,31 ¼ 2.30, p ¼ 0.026). Pairwise comparisons indicatedthat signatures of the Age 3þ cohort differed from both the Age 1þ(pseudo-t ¼ 1.69, df ¼ 22, p ¼ 0.038) and Age 2þ cohorts (pseudo-t ¼ 1.56, df ¼ 21, p ¼ 0.030), but Age 1þ and Age 2þ cohorts did notdiffer from each other (pseudo-t ¼ 1.30, df ¼ 19, p ¼ 0.162).

4. Discussion

Determining the habitat value of oil structures for reef fish isessential for making ecologically-sound decommissioning de-cisions. Structures which support large resident populations arelikely to be more valuable as habitat than structures which supportsmall or transient populations, and would therefore be more suit-able for decommissioning options which involve leaving all or partof the structure in place (i.e. ‘rigs-to-reefs’, RTR, Schroeder andLove, 2004; Fowler et al., 2014). Despite this realisation, thehabitat value of oil structures for reef fish can rarely be determined,because the mobility of many species makes it difficult to confirmresidence of individuals on particular structures. The current studyprovides the first step in assessing the utility of otolith propertiesfor distinguishing structures of residence for reef fish. Using pop-ulations of a small coral reef fish (Pseudanthias rubrizonatus) withknown residence histories, we have shown that reef fish inhabitingoil structures can develop distinct otolith properties over a smallspatial scale (10e100 km). Further, we found that a combination ofmicrochemical data from the otolith edge and elliptical Fourier(shape) descriptors can be used to allocate individuals to theirstructure of residence with moderate success. We also determinedthat site-specific microchemical signatures were stable enough

Fig. 4. Mean (±SE) concentrations of a) Sr, b) Ba, c) Mn, d) Mg, e) Li and f) Fe on the edge of the otolith of P. rubrizonatus inhabiting 4 oil structures located on Australia's North WestShelf (NWS). Structures that share a lower case letter were not significantly different from each other (p � 0.05) according to a pairwise PERMANOVA test. Letters are omitted forMg, Li and Fe because no significant differences were found between structures for these elements.

Fig. 5. CAP plot of otolith microchemistry during 3 growth periods for P. rubrizonatusinhabiting oil structures on Australia's North West Shelf (NWS).

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113110

through time to allow distinction of populations during 3 separatetime periods spanning ~1.5e2 years. Our results suggest that otolithproperties may be useful for determining the residency of reeffishes on oil structures, and may therefore assist selection of themost ecologically-sound decommissioning option for obsoletestructures. However, the low allocation accuracies achieved usingany single discriminationmethod highlight the importance of usinga combination of methods to identify ‘home’ structures, as well asthe need for further research to determine the applicability of thecurrent approach to other species and locations.

For otolith properties to provide a useful tool for identifying oil

structures of residence, they must vary over spatial scales that arecomparable to the distances between structures. Oil structures areseparated by 100s of m through to 10s of km, depending on theregion and type of structure considered (see Sammarco et al., 2004;Page et al., 2008; Fowler and Booth, 2012; Thorpe, 2012). Variationin otolith properties over a small spatial scale is therefore requiredto achieve population differentiation. Numerous studies havedetected differences in the otolith microchemistry of reef fish over10s of km (Chittaro et al., 2005; Dorval et al., 2005; Ruttenberget al., 2008), and even 100s of m (Gillanders and Kingsford, 2000;Kingsford and Gillanders, 2000), indicating the potential formicrochemistry to vary over scales necessary for population dif-ferentiation on oil structures. However, most investigations to datehave been conducted in nearshore environments, either withinestuaries, or on shallow coastal reefs. Microchemical variation isgenerally lower in offshore environments than nearshore envi-ronments, due tomore subtle gradients in salinity, temperature andwater chemistry (Thorrold et al., 2007). It was therefore unclearprior to this study whether adequate scales of population differ-entiation could be achieved in deeper offshore locations, wheremany oil structures are located. We were able to detect differencesin the otolith microchemistry of P. rubrizonatus between oil struc-tures separated by 9.7e84.2 km on the outer North West Shelf(NWS) of Australia (~130 km from the mainland), indicating thepotential for otolith microchemistry to differentiate structures ofresidence for reef fish in offshore areas. And although the allocationaccuracies achieved using microchemical data alone are lower thanmost studies that examined bays and estuaries (>80% accuracy, seeGillanders, 2005), they are comparable to those of numerous

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113 111

studies conducted on the open coast over a similar spatial scale(~30e70% accuracy; e.g. Patterson et al., 2004; Kingsford et al.,2009; Fairclough et al., 2011). While low in an absolute sense,such allocation accuracies may still be useful for rapid assessmentsof residency on oil structures when few alternative methods exist.Technological advancements are also likely to increase the preci-sion of LA-ICP-MS systems in the future, which will likely increasethe range of elements that can be use to characterise site-specificsignatures and improve allocation accuracies (Thorrold et a.2007). Given the unique nature of the current study, it is unclearwhether the scale of differentiation, or the allocation accuraciesachieved, will be similar for other species in other locations.

The temporal consistency of microchemical signatures willgreatly affect their utility for identifying oil structures of residencefor reef fish. If signatures vary through time, then individualscollected at different times cannot be reliably allocated to theirstructure of origin. While this concept is usually considered in thecontext of identifying natal origins of reef fish (Gillanders, 2002;Thorrold et al., 2007), it is also of concern when attempting toidentify primary areas of residence for adults, because all in-dividuals in a study would need to be collected within the sametime period. This is particularly limiting for investigations of resi-dency on oil structures, because the logistical challenges associatedwith offshore sampling often prevent collection of all specimens atone time. We were able to assess the temporal consistency ofmicrochemical signatures of P. rubrizonatus by targeting previousgrowth zones in the otolith, under the assumption that individualsremained resident at their site of capture. Despite reduced samplesizes necessitated by targeting of specific age cohorts, we foundconsistent differences among 3 oil structures across 3 time periods,suggesting that within-site variability at offshore structures may below enough to allocate individuals to their structure of residence,irrespective of the time of capture. The finding also indicates it maybe possible to reconstruct residence histories of individuals on oilstructures by obtaining microchemical signatures throughout theotolith. This may allow detection of movements between struc-tures, as well as allow determination of the extent of time spent onparticular structures. We offer these conclusions tentatively, how-ever, because we were only able to examine temporal variation in asingle cohort (Age 1þ), and the total time period encompassed bythe 3 growth zones examined was only ~1.5e2 years (Fowler andBooth, 2012). Additionally, we detected significant differences inmicrochemical signatures between successive cohorts at a singlesite (WH3), indicating that site-specific signatures vary throughtime to some degree. Further research is therefore required todetermine the long-term stability of microchemical signatures atoffshore structures, and how this will influence distinction of in-dividuals collected across multiple time-periods.

The utility of otolith shape for distinguishing oil structures ofresidence for reef fish is less clear than otolith microchemistry, atleast in isolation. Despite similar overall allocation accuracy to thatachieved with microchemical signatures, elliptical Fourier de-scriptors for P. rubrizonatus were only found to differ significantlybetween a single pair of structures separated by 55 km (WH1 andWH3, Table 3), indicating minimal variation in otolith shape overthe spatial scales necessary for distinguishing oil structures ofresidence. Elliptical Fourier descriptors also resulted in the lowestallocation accuracy for any single site (WH4: 31.8%, Table 4). Thefew previous studies comparing otolith microchemistry and otolithshape for population discrimination also found reduced utility ofotolith shape, with elliptical Fourier descriptors resulting in poorerallocation accuracy than microchemical signatures for the samesample of individuals (Longmore et al., 2010; Ferguson et al., 2011).Reduced spatial variation in otolith shape relative to otolithmicrochemistry may reflect the difference in temporal resolution

between the two techniques, or the different processes influencingotolith shape and microchemistry, or a combination of both.Microchemical signatures obtained from the otolith edge usinglaser ablation represent a discrete time period, often �6 months,while otolith shape is the product of allometric growth processesoccurring throughout the entire life of an individual (Fergusonet al., 2011). If growth processes at each oil structure in the cur-rent study varied through time, any differences in otolith shapeamong structures may have been ‘smoothed out’ by the integrationof slower and faster growth patterns through time, ultimatelyreducing spatial variation in shape. Alternatively, the processesaffecting otolith shape may have varied less across the study areathan processes affecting otolith microchemistry. Unlike otolithmicrochemistry, otolith shape is believed to be primarily influencedby growth rate (Campana and Casselman,1993), and annual growthrates of the populations used the current study were similar ac-cording to a previous otolith-based investigation (Fowler andBooth, 2012).

Despite reduced variation in otolith shape among oil structuresin the current study, the combination of shape data with micro-chemical data improved allocation success of P. rubrizonatus in-dividuals to ‘home’ structures. This result suggests that acombination of population discriminators are likely to be moresuccessful for identifying oil structures of residence than any singlediscriminator in isolation; a scenario which has previously beenrecognised for detection of sub-stocks in fished species (Begg andWaldman, 1999; Ferguson et al., 2011). Otolith shape is an under-used tool for population discrimination, yet should be considered infuture investigations of residency on oil structures, because itprovides a measure of population separation over longer time-scales than most other methods (e.g. edge microchemistry,tagging), and data can be obtained relatively easily and cheaply.Otolith shape is therefore likely to provide a useful complement toany investigation of residency that involves the collection of spec-imens. Additionally, otolith shape varies considerably among spe-cies. So while it was not a powerful discriminator for P. rubrizonatuspopulations in the current study, the technique may prove moreuseful for other species inhabiting oil structures. Further research isrequired to determine the spatial scales over which otolith shapevaries in offshore environments in order to identify scenarios inwhich otolith shape will be a useful population discriminator.

Differences in otolith microchemistry and shape between the oilstructures did not simply reflect geographical separation, suggest-ing that large-scale gradation in environmental properties acrossthe study area was not the sole driver of variation in otolith prop-erties. Concentrations of common trace elements in the otoliths ofmarine fishes (e.g. Sr and Ba) are strongly influenced by ambienttemperature and salinity (Elsdon and Gillanders, 2003; Elsdonet al., 2008). Otolith shape is also influenced by temperaturethrough its effect on growth rate (Campana and Casselman, 1993).We therefore expected that otolith properties would be moresimilar between populations located closer together than pop-ulations located further apart, due to large-scale variation in watertemperature and salinity. Similarly, we expected that populationsfrom the two deeper structures (WH3 andWH4,133e135m)wouldbe more similar to each other than populations from the twoshallower structures (WH1 andWH2, 82e84m), and vice versa, dueto potential differences in water temperature and salinity betweendepths. These hypotheses were supported by greater misallocationof individuals to neighbouring structures when using micro-chemical signatures from the otolith edge. However, some of thegreatest differences in the concentration of individual elementswere found between WH1 and WH2, the two closest structures(9.7 km apart, Fig. 3). Furthermore, differences in otolith shapewere found between structures separated by intermediate

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113112

distances (WH1 and WH3), but not between structures furthestapart. Misallocation of individuals based on shape data also showedno clear pattern with regard to the geographical separation ofstructures. These results suggest that otolith properties wereinfluenced by factors operating over small (10e100 km) spatialscales, and indicate the potential for development of site-specificotolith properties on oil structures in the offshore environment.

Further research is required to determine the factors driving theotolith microchemistry of fish inhabiting oil structures. Micro-chemical signatures on oil structures may differ to other habitats,due to the high concentrations of less common elements (e.g. Ni,Cu, Cd) which may be present in sediments surrounding thestructure, even after cessation of drilling activity (Breuer et al.,2004). Individual structures may develop signatures that differ toadjacent structures, due to the potentially unique combination ofelements found at specific sites, or the different types of drillinglubricant used during the production phase. Nowling et al. (2011)found ‘drilling elements’ useful in discriminating between redsnapper (L. campechanus) inhabiting oil platforms and artificialreefs in the northern Gulf of Mexico, and we therefore expectedthat such elements might be prominent in the otolith micro-chemistry of P. rubrizonatus inhabiting oil structures on the NorthWest Shelf. However, we did not detect any of these elements inconcentrations above limits of detection, despite adequate limits ofdetection being achieved for a suite of other elements. Discrimi-nation between oil structures in the current study was driven byelements that have commonly been useful for discriminating fishpopulations in natural habitats (i.e. Sr, Ba, Mn and Mg). Drillingelements may not be widely useful for identifying oil structures ofresidence for reef fish; however, considerable investigations ofother locations and species will be required to test the generality ofthis conclusion. Individuals of P. rubrizonatus in the current studymay simply not have come into contact with sufficient quantities ofdrill spoil, perhaps due to planktivorous feeding behaviour higherup in the water column.

The fact that discrimination among oil structures in the currentstudy was achieved using common trace elements (i.e. not ‘drilling’elements) indicates the approach may be useful for assessing thehabitat value of artificial reefs in general. Despite their popularity insome regions, there is little evidence that artificial reefs actuallyincrease the biomass of reef fish, as opposed to simply aggregatingindividuals (Bohnsack, 1989; Brickhill et al., 2005). Otolith micro-chemistry may allow individuals to be linked to particular artificialreefs, which would assist estimates of productivity for these habi-tats. If microchemical signatures are stable through time, orsequential sampling is undertaken, potential movements of in-dividuals between artificial reefs and surrounding habitats may bedetectable, allowing estimates of biomass export from artificialreefs.

5. Conclusion

To assist in assessing the habitat value of oil structures, otolithproperties must be able to identify structures of residence for reeffishes of ecological and commercial value. The current study hasprovided the first step toward this goal by demonstrating thatotolith microchemistry and shape could be used to allocate in-dividuals of a small reef fish to their ‘home’ structures with mod-erate accuracy over a spatial scale relevant to oil installations(10e100 km). However, P. rubrizonatus is a strongly site-attachedspecies and the individuals used in the current study most likelysettled on their home structures and remained there until capture.Many species considered to hold ecological or commercial valuewill be considerably more mobile than P. rubrizonatus and may notreside on a single oil structure for their entire lives. The spatial and

temporal scales over which reasonable allocation accuracies can beachieved for such species may be considerably different to thosefound here. Future investigations should test the method outlinedin the current study on larger more mobile reef species, in combi-nation with tagging methods to ascertain residence histories forindividuals.

Acknowledgements

Wewould like to thank Bevan Yiu for his assistance with otolithpreparation and analysis.

References

Anderson, M.J., Gorley, R.N., Clarke, K.R., 2008. PERMANOVAþ for PRIMER: Guide toSoftware and Statistical Methods. PRIMER-E, Plymouth.

Ashford, J.R., Arkhipkin, A.I., Jones, C.M., 2006. Can the chemistry of otolith nucleidetermine population structure of Patagonian toothfish Dissostichus ele-ginoides? J. Fish. Biol. 69, 708e721.

Beer, N.A., Wing, S.R., Swearer, S.E., 2011. Otolith elemental evidence for spatialstructuring in a temperate reef fish population. Mar. Ecol. Prog. Ser. 442,217e227.

Begg, G.A., Waldman, J.R., 1999. An holistic approach to fish stock identification.Fish. Res. 43, 35e44.

Ben-Tzvi, O., Abelson, A., Gaines, S.D., Sheehy, M.S., Paradis, G.L., Kiflawi, M., 2007.The inclusion of sub-detection limit LA-ICPMS data, in the analysis of otolithmicrochemistry, by use of a palindrome sequence analysis (PaSA). Limnol.Oceanogr. Methods 5, 97e105.

Bohnsack, J.A., 1989. Are high densities of fishes at artificial reefs the result ofhabitat limitation or behavioral preference? Bull. Mar. Sci. 44, 631e645.

Breuer, E., Stevenson, A.G., Howe, J.A., Carroll, J., Shimmield, G.B., 2004. Drill cuttingaccumulations in the Northern and Central North Sea: a review of environ-mental interactions and chemical fate. Mar. Pollut. Bull. 48, 12e25.

Brickhill, M.J., Lee, S.Y., Connolly, R.M., 2005. Fishes associated with artificial reefs:attributing changes to attraction or production using novel approaches. J. Fish.Biol. 67, 53e71.

Campana, S.E., 1999. Chemistry and composition of fish otoliths: pathways, mech-anisms and applications. Mar. Ecol. Prog. Ser. 188, 263e297.

Campana, S.E., Casselman, J.M., 1993. Stock discrimination using otolith shape-analysis. Can. J. Fish. Aquat. Sci. 50, 1062e1083.

Campana, S.E., Thorrold, S.R., 2001. Otoliths, increments, and elements: keys to acomprehensive understanding of fish populations? Can. J. Fish. Aquat. Sci. 58,30e38.

Campana, S.E., Fowler, A.J., Jones, C.M., 1994. Otolith elemental fingerprinting forstock identification of Atlantic cod (Gadus morhua) using laser ablation ICPMS.Can. J. Fish. Aquat. Sci. 51, 1942e1950.

Campana, S.E., Chouinard, G.A., Hanson, J.M., Fr�echet, A., Brattey, J., 2000. Otolithelemental fingerprints as biological tracers of fish stocks. Fish. Res. 343e357.

Chittaro, P.M., Usseglio, P., Fryer, B.J., Sale, P.F., 2005. Using otolith microchemistry ofHaemulon flavolineatum (French grunt) to characterize mangroves and coralreefs throughout Turneffe Atoll, Belize: difficulties at small spatial scales. Es-tuaries 28, 373e381.

Crampton, J.S., 1995. Elliptic fourier shape analysis of fossil bivalves: some practicalconsiderations. Lethaia 28, 179e186.

Dorval, E., Jones, C.M., Hannigan, R., van Montfrans, J., 2005. Can otolith chemistrybe used for identifying essential seagrass habitats for juvenile spotted seatrout,Cynoscion nebulous, in Chesapeake Bay? Mar. Freshw. Res. 56, 645e653.

Elsdon, T.S., Gillanders, B.M., 2003. Reconstructing migratory patterns of fish basedon environmental influences on otolith chemistry. Rev. Fish. Biol. Fish. 13,219e235.

Elsdon, T.S., Wells, B.K., Campana, S.E., Gillanders, B.M., Jones, C.M., Limburg, K.E.,Secor, D.H., Thorrold, S.R., Walther, B.D., 2008. Otolith chemistry to describemovements and life-history parameters of fishes: hypotheses, assumptions,limitations and inferences. Oceanogr. Mar. Biol. Annu. Rev. 46, 297e330.

Fairclough, D.V., Edmonds, J.S., Lenanton, R.C.J., Jackson, G., Keay, I.S., Crisafulli, B.M.,Newman, S.J., 2011. Rapid and cost-effective assessment of connectivity amongassemblages of Choerodon rubescens (Labridae), using laser ablation ICP-MS ofsagittal otoliths. J. Exp. Mar. Biol. Ecol. 403, 46e53.

Ferguson, G.J., Ward, T.M., Gillanders, B.M., 2011. Otolith shape and elementalcomposition: complementary tools for stock discrimination of mulloway(Argyrosomus japonicus) in southern Australia. Fish. Res. 110, 75e83.

Fowler, A.M., Booth, D.J., 2012. Evidence of sustained populations of a small reef fishon artificial structures; does depth affect production on artificial reefs? J. Fish.Biol. 80, 613e629.

Fowler, A.M., Macreadie, P.I., Jones, D.O.B., Booth, D.J., 2014. A multi-criteria decisionapproach to decommissioning of offshore oil and gas infrastructure. Ocean.Coast Manage 87, 20e29.

Frederick, J.L., 1997. Post-settlement movement of coral reef fishes and bias insurvival estimates. Mar. Ecol. Prog. Ser. 150, 65e74.

Gagliano, M., McCormick, M.I., 2004. Feeding history influences otolith shape in

A.M. Fowler et al. / Marine Environmental Research 106 (2015) 103e113 113

tropical fish. Mar. Ecol. Prog. Ser. 278, 291e296.Gallaway, B.J., 1981. An Ecosystem Analysis of Oil and Gas Development on the

Texas-Louisiana Continental Shelf. U.S. Fish and Wildlife Service, Washington,DC, p. 89.

Gallaway, B.J., Szedlmayer, S.T., Gazey, W.J., 2009. A life history review for redsnapper in the Gulf of Mexico with an evaluation of the importance of offshorepetroleum platforms and other artificial reefs. Rev. Fish. Sci. 17, 48e67.

Gillanders, B.M., 2002. Temporal and spatial variability in elemental composition ofotoliths: implications for determining stock identity and connectivity of pop-ulations. Can. J. Fish. Aquat. Sci. 59, 669e679.

Gillanders, B.M., 2005. Using elemental chemistry of fish otoliths to determineconnectivity between estuarine and coastal habitats. Estuar. Coast Shelf Sci. 64,47e57.

Gillanders, B.M., Kingsford, M.J., 2000. Elemental fingerprints of otoliths of fish maydistinguish estuarine 'nursery' habitats. Mar. Ecol. Prog. Ser. 20, 273e286.

Gillanders, B.M., Kingsford, M.J., 2003. Spatial variation in elemental composition ofotoliths of three species of fish (family Sparidae). Estuar. Coast Shelf Sci. 57,1049e1064.

Hamer, P.A., Jenkins, G.P., 2007. Comparison of spatial variation in otolith chemistryof two fish species and relationships with water chemistry and otolith growth.J. Fish. Biol. 71, 1035e1055.

Iwata, H., Ukai, Y., 2002. SHAPE: a computer program package for quantitativeevaluation of biological shapes based on elliptic fourier descriptors. J. Hered. 93,384e385.

Jørgensen, D., 2012. OSPAR's exclusion of rigs-to-reefs in the North Sea. Ocean.Coast Manage 58, 57e61.

Kingsford, M.J., Gillanders, B.M., 2000. Variation in concentrations of trace elementsin otoliths and eye lenses of a temperate reef fish, Parma microlepis, as afunction of depth, spatial scale, and age. Mar. Biol. 137, 403e414.

Kingsford, M.J., Hughes, J.M., Patterson, H.M., 2009. Otolith chemistry of the non-dispersing reef fish Acanthochromis polyacanthus: cross-shelf patterns fromthe central great Barrier Reef. Mar. Ecol. Prog. Ser. 377, 279e288.

Kuhl, F.P., Giardina, C.R., 1982. Elliptic fourier features of a closed contour. Comput.Graph. Image Process. 18.

Lombarte, A., Lleonart, J., 1993. Otolith size changes related with body growth,habitat depth and temperature. Environ. Biol. Fish. 37, 297e306.

Longerich, H.P., Jackson, S.E., Gunther, D., 1996. Laser ablation inductively coupledplasma mass spectrometric transient signal data acquisition and analyte con-centration calculation. J. Anal. At. Spectrom. 11, 899e904.

Longmore, C., Fogarty, K., Neat, F., Brophy, D., Trueman, C., Milton, A., Mariani, S.,2010. A comparison of otolith microchemistry and otolith shape analysis for thestudy of spatial variation in a deep-sea teleost, Coryphaenoides rupestris. Envi-ron. Biol. Fish. 89, 591e605.

Love, M.S., Caselle, J.E., Snook, L., 2000. Fish assemblages around seven oil platformsin the Santa Barbara channel area. Fish. Bull. 98, 96e117.

Love, M.S., Schroeder, D.M., Lenarz, W., 2005. Distribution of bocaccio (Sebastespaucispinis) and cowcod (Sebastes levis) around oil platforms and natural out-crops off California with implications for larval production. Bull. Mar. Sci. 77,397e408.

Love, M.S., Nishimoto, M., Saiki, M., 2009. Reproductive ecology and body burden ofresident fish prior to decommissioning. St. Barbara 78.

Lyne, V., Fuller, M., Last, P., Martin, M., Scott, R., 2006. Ecosystem characterisation ofAustralia's North west shelf North west shelf joint environmental managementstudy. CSIRO Mar. Atmos. Res. 72.

Macreadie, P.I., Fowler, A.M., Booth, D.J., 2011. Rigs-to-reefs: will the deep seabenefit from artificial habitat? Front. Ecol. Environ. 9, 455e461.

Macreadie, P.I., Fowler, A.M., Booth, D.J., 2012. Rigs-to-reefs policy: can sciencetrump public sentiment? Front. Ecol. Environ. 10, 179e180.

Martin, C.B., Lowe, C.G., 2010. Assemblage structure of fish at offshore petroleumplatforms on the San Pedro shelf of Southern California. Mar. Coast. Fish. Dyn.Manag. Ecosyst. Sci. 2, 180e194.

McGowan, N., Fowler, A.M., Parkinson, K., Bishop, D.P., Ganio, K., Doble, P.A.,Booth, D.J., Hare, D.J., 2014. Beyond the transect: an alternative microchemicalimaging method for fine scale analysis of trace elements in fish otoliths duringearly life. Sci. Total Environ. 494, 177e186.

M�erigot, B., Letourneur, Y., Lecomte-Finiger, R., 2007. Characterization of local

populations of the common sole Solea solea (Pisces, Soleidae) in the NWMediterranean through otolith morphometrics and shape analysis. Mar. Biol.151, 997e1008.

Nowling, L., Gauldie, R.W., Cowan Jr., J.H., De Carlo, E., 2011. Successful discrimi-nation using otolith microchemistry among samples of red snapper Lutjanuscampechanus from artificial reefs and samples of L. campechanus taken fromnearby oil and gas platforms. Open Fish Sci. J. 4, 1e9.

Overholtzer-McLeod, K.L., 2005. Post-settlement emigration affects mortality esti-mates for two Bahamian wrasses. Coral Reefs 24, 283e291.

Page, H.M., Culver, C.S., Dugan, J.E., Mardian, B., 2008. Oceanographic gradients andpatterns in invertebrate assemblages on offshore oil platforms. ICES J. Mar. Sci.65, 851e861.

Parente, V., Ferreira, D., Moutinho dos Santos, E., Luczynski, E., 2006. Offshoredecommissioning issues: deductibility and transferability. Energy Policy 34,1992e2001.

Patterson, H.M., Kingsford, M.J., McCulloch, M.T., 2004. Elemental signatures ofPomacentrus coelestis otoliths at multiple spatial scales on the great BarrierReef, Australia. Mar. Ecol. Prog. Ser. 270, 229e239.

Peabody, M.B., Wilson, C.A., 2006. Fidelity of Red Snapper (Lutjanus campechanus) toPetroleum Platforms and Artificial Reefs in the Northern Gulf of Mexico. U.S.Department of the Interior, Minerals Management Service, New Orleans, p. 64.

Pearce, N.J.G., Perkins, W.T., Westgate, J.A., Gorton, M.P., Jackson, S.E., Neal, C.R.,Chenery, S.P., 1997. A compilation of new and published major and traceelement data for NIST SRM 610 and NIST SRM 612 glass reference materials.Geostand. Newsl. 21, 115e144.

Popper, D., Fishelson, L., 1973. Ecology and behavior of Anthias squamipinnis (Pe-ters, 1855) (Anthiidae, Teleostei) in the coral habitat of Eilat (Red Sea). J. Exp.Zool. 184, 409e424.

Pothin, K., Gonzalez-Salas, C., Chabanet, P., Lecomte-Finiger, R., 2006. Distinctionbetween Mulloidichthys flavolineatus juveniles from Reunion Island andMauritius Island (south-west Indian Ocean) based on otolith morphometrics.J. Fish. Biol. 69, 38e53.

Pradella, N., Fowler, A.M., Booth, D.J., Macreadie, P.I., 2014. Fish assemblages asso-ciated with oil industry structures on the continental shelf of north-westernAustralia. J. Fish. Biol. 84, 247e255.

Quinn, G.G.P., Keough, M.J., 2002. Experimental Design and Data Analysis for Bi-ologists. Cambridge University Press, Melbourne.

Ruttenberg, B.I., Hamilton, S.L., Warner, R.R., 2008. Spatial and temporal variation inthe natal otolith chemistry of a Hawaiian reef fish: prospects for measuringpopulation connectivity. Can. J. Fish. Aquat. Sci. 65, 1181e1192.

Sammarco, P.W., Atchison, A.D., Boland, G.S., 2004. Expansion of coral communitieswithin the Northern Gulf of Mexico via offshore oil and gas platforms. Mar. Ecol.Prog. Ser. 280, 129e143.

Schroeder, D.M., Love, M.S., 2004. Ecological and political issues surroundingdecommissioning of offshore oil facilities in the Southern California Bight.Ocean. Coast Manage 47, 21e48.

Seaman Jr., W., Lindberg, W.J., Gilbert, C.R., Frazer, T.K., 1989. Fish habitat providedby obsolete petroleum platforms off southern Florida. Bull. Mar. Sci. 44,1014e1022.

Stanley, D.R., Wilson, C.A., 2000. Variation in the density and species composition offishes associated with three petroleum platforms using dual beam hydro-acoustics. Fish. Res. 47, 161e172.

Thorpe, S.A., 2012. On the biological connectivity of oil and gas platforms in theNorth Sea. Mar. Pollut. Bull. 64, 2770e2781.

Thorrold, S.R., Zacherl, D.C., Levin, L.A., 2007. Population connectivity and larvaldispersal: using geochemical signatures in calcified structures. Oceanography20, 80e89.

Turgeon, K., Robillard, A., G�regoire, J., Duclos, V., Kramer, D.L., 2010. Functionalconnectivity from a reef fish perspective: behavioral tactics for moving in afragmented landscape. Ecology 91, 3332e3342.

Vignon, M., Morat, F., Galzin, R., Sasal, P., 2008. Evidence for spatial limitation of thebluestripe snapper Lutjanus kasmira in French Polynesia from parasite andotolith shape analysis. J. Fish. Biol. 73, 2305e2320.

Wilson Jr., R.R., 1985. Depth-related changes in sagitta morphology in six macrouridfishes of the Pacific and Atlantic Oceans. Copeia 1985, 1011e1017.