counting coral reef fishes: interaction between fish life-history traits and transect design
Post on 05-Sep-2016
Embed Size (px)
Article history:Received 14 July 2009Received in revised form 26 February 2010Accepted 2 March 2010
Journal of Experimental Marine Biology and Ecology 387 (2010) 1523
Contents lists available at ScienceDirect
Journal of Experimental Ma
.e l1. Introduction
Understanding patterns of abundance, structure and diversity ofpopulations and communities is a major goal in ecology and requiressampling procedures which can produce spatial and temporalcomparable samples (e.g. Caughley, 1977). With the emergence ofmacro-ecology and at a time of global environmental threats thisrequirement is becoming an increasing concern as the data availablefor large-scale analyses and global assessments generally come from avariety of sources and involve a range of sampling methods.Understanding the characteristics of each census method and howdemographic estimates differ amongst methods is a rst but crucial
one of the most widely used methods to survey animals (seeSupplementary material I). Visual censuses are used for insects, birds,reptiles, mammals, invertebrates, and shes, notably in coral reefswhere underwater visual census (UVC) are the main methods used forthe long term monitoring of emblematic ecosystems such as the GreatBarrier Reef (see Australian Institute of Marine Sciences Long TermMonitoringProgramHalford andThompson, 1994), the assessmentoflarge-scale sheries such as the Pacic island reef sheries (seePROCFISH program of the South Pacic Commission) or for theworldwide monitoring of coral reefs (see the Global Coral ReefMonitoring Network GCRMN). On reefs (which encompass coralreefs butmore andmore temperate and coldwater reefs are sampled bystep to assess and correct the potential biasesinformation from different sources in large in
Visual censuses are non-destructive, lowmentable in monitoring programs, and as suc
Corresponding author.E-mail address: email@example.com (M. Ku
0022-0981/$ see front matter 2010 Elsevier B.V. Aldoi:10.1016/j.jembe.2010.03.003 2010 Elsevier B.V. All rights reserved.Density estimateFishTransectUnderwater visual censusUnderwater visual censuses are the most commonly used methods to estimate the density of reef shpopulations and assemblages. One basic assumption is that the observer will always detect sh in the sameway from one sampling unit to the next, implying that, on average, the spatial distribution pattern of shabundance or occurrence remains the same from one transect to the next (H0). The present work tested H0using data from 730 transects covering two regions (New Caledonia and French Polynesia), 604 species and504000 sh. Within transect variations in reef sh abundance and occurrence were studied according to sitefactors (region, reef type), life-history traits (adult size, home range, schooling behaviour, color, pattern,swimming speed, level in the water column, inquisitiveness, crypticity), and observations characteristics(distance of observation, size of the observed shes, number of shes within an observation, observeridentity). Two general trends were detected: 1 at the start of transects, both sh occurrence andabundance were higher than the values expected under H0; 2 a similar trend was also observed at the endof transects, but at a much lower magnitude. These effects were observed with varying degrees of magnitudefor all regions, reef types and observers, varied signicantly according to three life-history traits (size, homerange, and behaviour), but were not inuenced by species richness or abundance. These results indicate thatdatasets gathered from transects of various lengths cannot be pooled without correction. They also shed lighton some of the known differences between transects and point counts.involved when mergingternational databases.-cost, and easily imple-h have been and still are
UVC) these meththe technique prcounting sh witand Brown, 1996Supplementary mpoint countswagiven radius (BoSamoilys and Calbicki).
l rights reserved.a r t i c l e i n f o a b s t r a c tCounting coral reef shes: Interaction betransect design
Michel Kulbicki a,, Nathaniel Cornuet b, Laurent ViglGrard Moutham b, Pascale Chabanet d
a IRD, Universit de Perpignan, 52, Avenue Paul Alduy, 66860, Perpignan, Franceb IRD, B.P. A5, 98845, Nouma, New Caledoniac LIVE, University of New Caledonia, B.P. R4, 98851, Nouma Cedex, New Caledoniad IRD, B.P. 172, 97492, Ste Clotilde Cedex, La Runion, France
j ourna l homepage: wwween sh life-history traits and
a b, Laurent Wantiez c,
rine Biology and Ecology
sev ie r.com/ locate / jembeods were initiated by Brock (1954). Transect counts,oposed by Brock, is still the most used and consists ofhin a corridor of a given length and width (see Cappoand Edgar et al., 2004 for a quick review see alsoaterial I). The other type of visual census in use is
here the observer will census in a circle or half-circle ofhnsack and Bannerot, 1986;Watson andQuinn, 1997;rlos, 2000; McNeil et al., 2008). There have been
numerous articles describing the biases of these techniques and manysolutions have been proposed to correct them (e.g. Harmelin-Vivienet al., 1985; Jennings and Polunin, 1995; Watson et al., 1995; Cheal andThompson, 1997; Kulbicki, 1998; Kulbicki and Sarramgna, 1999;Samoilys and Carlos, 2000; McNeil et al., 2008; see also Supplementarymaterial I). A basic assumption of all these methods is that the observerwill record sh in the same way from one sampling unit to the next.This implies that, on average, an observer should detect sh with thesame accuracy at the start, middle and end of a transect (H0). In otherwords, this means that the same proportion of sh present should beobserved on any portion of the transect and not that the density of thesh should remain constant within the transect. If H0 is not respected,then combining results from different transect types within an analysisor using sub-units of a transect as pseudo-replicates would lead to
16 M. Kulbicki et al. / Journal of Experimental Marine Biology and Ecology 387 (2010) 1523biased results unless a correcting method is applied.It is likely that because many organisms react to the presence of an
observer, their detection will not in fact remain homogeneous withina transect (H1). In particular, organisms attracted to an observer willprobably be found in higher densities at the start of transects, whereasspecies which are frightened by an observer may increase in densityalong the transect, as they will often return as they gradually get usedto the presence of the observer. Thus factors such as transect lengthand width are likely to affect diversity and density estimates. Theamplitude of such phenomena has not been investigated to ourknowledge. In particular, H0 has never been tested formally and theeffect of transect length on demographic estimates has never beenevaluated. Yet, from the known behaviour of coral reef shes(Kulbicki, 1998) it is expected that animal movements could besufciently important as to induce important biases in diversity anddensity estimates from visual census techniques, and these will needto be addressed prior to using data from different sources for large-scale and/or global assessments.
The present article uses data from a large database on coral reefshes to test whether animal detectability remains, on average,constant within transects (H0). Specically, we investigate how reefsh observations are distributed along transects and examine anydeparture from H0 according to intrinsic (e.g. species, sh size,behaviour) and extrinsic (e.g. reef type, region, observer) factors.Finally, we discuss our results in order to explain some of the reporteddifferences amongst the most frequently used visual census methodsand highlight the consequences that these may have for large-scaleassessments of ecosystem status and functioning.
2. Materials and methods
2.1. Collection methods
The study encompasses 730 line transects performed in two regionsof the South Pacic Ocean, NewCaledonia and the Tuamotu archipelago(French Polynesia) (Table 1). These regions are characterized bydifferent levels of regional diversity, being higher in New Caledonia
Table 1Number of transects performed by each diver (nos. 1, 2, 5, and 7) according to habitatand region.
Habitat Diver no. Region Total
New Caledonia Tuamotu
Barrier 1 115 1152 65 117 1825 54 115 1697 115 115
Total barrier 119 462 581Fringing 2 84 84
5 65 65Total fringing 149 149Total 268 462 730than in the Tuamotu (Kulbicki, 2007), with 729 and 391 easilyobservable coral reef sh species respectively. The inner slopes ofbarrier reefs were sampled in both regions, and fringing reefs weresampled only in New Caledonia. Counts were performed by four diversin the Tuamotu (divers 1, 2, 5, and 7) but only two of those (divers 2 and5) collected data inNewCaledonia. All diverswere highly accomplishedreef sh counters, each with more than 10 years experience in UVC.
Line transects were performed as described by Labrosse et al.(2001). Briey, sh were recorded along a 50 m tape by two divers,one on each side of the tape. The tape was divided into 5 sections of10 m each. For each sh sighting the observer identied the speciesand recorded the size (TL in cm) and number of shes (school size) aswell as the perpendicular distance of the sh from the transect line.For schools, the average size and distance to both the nearest andfurthest shes of the group were recorded. Distances were recordedby 1 m class till 5 m, 2 m class from 6 to 10 m and 5 m class beyond10 m. Depth, substrate composition (ne sediment, gravel, debris,small and large blocks, rock, coral), algae and live coral were recordedfor each section by the MSA method (Clua et al., 2006). Transectsperformed in New Caledonia were set parallel to the reef slope(const