1. Introduction
With a yearly production of more than 40,000, the Mediterranean aquaculture of the Ž
. sea bream and the European sea bass
Dicentrarchus labrax has lately undergone a very rapid development. At the basis of such development stand both the availability of
reliable rearing techniques and the effort devoted to seed production. Although since the Ž
. beginning, many authors e.g., Barahona-Fernandez, 1982; Chatain, 1994 emphasised
the problem of anomalies in hatchery-produced juveniles, skeletal anomalies are still considered an important problem in such activities. They may cause drastic reductions of
Ž the product’s quality in terms of growth, survival and external morphology Boglione et
. al., 1994; Cataudella et al., 1995; Koumoundouros et al., 1997 . This is also true in
restocking programs, where many authors emphasised the need to obtain individuals Ž
. with wild-like performances Howell and Baynes, 1993; Mesa, 1994; Olla et al., 1994 .
In such contexts, the study of the morphological characteristics of reared fish at the skeletal level and at a higher level of integration, such as the external morphology, is
becoming a pivotal topic.
Ž The use of geometric morphometrics Bookstein, 1991; Rohlf and Marcus, 1993;
. Marcus et al., 1996 is proposed in this paper in combination with internal anatomical
analysis as a tool in the assessment of the phenotypical quality of reared fish. The effect of different conditions of larval and postlarval rearing on the external
Ž .
morphology shape and internal anatomical characters was studied in two samples of Ž
. juvenile and adult D. labrax Teleostea, Moronidae .
Differences in shape between two samples were analysed with geometric morphomet- rics. On the same specimens, internal anatomical data were collected from X-rays. In
this way, it was possible to correlate fish shape with internal anatomical characteristics. After different larval and postlarval rearings, sea bass were introduced into floating
cages: juveniles were sampled before the introduction into these facilities; adults were sampled after 15 months of common rearing in floating cages. Such design was defined
to test whether differences in external morphology increased, decreased or remained stable during the trial.
If, today, fish shape represents an important component within the large number of Ž
. factors that affect the market value, the proposed approach may evidence and visualise
abnormalities in the external morphology: these can be associated with particular patterns of skeletal anomalies and substantiated with rigorous statistical inference. If this
is true, the combined tool could be considered as part of the approach in the evaluation of fish quality.
2. Materials and methods
Ž Specimens of European sea bass from a French intensive commercial hatchery FRA,
. Ž
initial mean standard length 55 mm and an Italian intensive commercial hatchery ITA, .
Ž .
initial mean standard length 50 mm were analysed Table 1 . Specimens were reared in floating cages 9 = 9 = 5 m
3
, 405 m
3
. Initial rearing densities were 0.32 kgrm
3
for FRA
Table 1 Ž
. Samples sizes, mean standard length and mean centroid sizes with standard deviations SD for juvenile and
adult European sea basses for each rearing group, ITA and FRA Centroid size, see text.
Ž .
Numbers of observations Standard length mm
SD Centroid size
SD ITA juveniles
43 49.93
2.9 64.62
4.118 FRA juveniles
45 55.49
3.91 71.9
5.231 ITA adults
47 192.38
16.389 279.52
18.151 FRA adults
40 210.93
15.868 256.21
17.792
and 0.13 kgrm
3
for ITA. The first sample was collected after 7 months from hatching, when individuals were introduced in the floating cages. The second sample was
Ž .
collected 22 months after hatching 15 months after the introduction in floating cages . Sea bass were fed twice a day with commercial pellets, in quantities depending on the
weight growth of individuals and water temperature. Survival rate was 85. Final food Ž
. Ž
. conversion indexes were 4.38 FRA and 4.61 ITA . Final rearing densities were 11.8
Ž .
3
Ž .
FRA and 8.3 kgrm ITA .
2.1. Geometric morphometrics: a brief introduction to the method The development of the techniques which constitute the discipline known as geomet-
ric morphometrics is quite recent and the tool is still not extensively applied. Few Ž
applications related to fish quality assessment are reported in the literature Cataudella et .
al., 1995; Loy et al., 1995, 1999 . Because of the complex computational steps involved in this analysis, a brief introduction on the method is given in Section 2.1.
The power of the technique lies in the possibility to visualise shape differences Ž
. through series of deformation grids known as splines , to decompose such differences
into a uniform component, which describes stretching, compression or shearing of the entire landmark configuration, and a non-uniform, localised, component. Co-variation
among portions or entire regions of the body is taken into account by the model. The numerical output can be analysed with traditional multivariate statistics.
At the core of the technique lies the thin-plate spline function; details on the analyses Ž
. Ž
. and algorithms may be found in Bookstein 1991, 1996a, b , Rohlf 1993 , Rohlf and
Ž .
Ž .
Ž .
Marcus 1993 , Rohlf et al. 1996 , Marcus et al. 1996 .The technique is based on the detection of homologous landmark points, in the form of x, y and eventually, z
Ž coordinates. The homology of the landmarks can be purely operative from a geometri-
. cal correspondence of landmarks on the objects to biologically homologous structures
according to the purposes of the study. Each individual is then described by a landmark configuration. Once the landmark configurations are collected for each specimen, a
series of algorithms is applied. The first step involves a superimposition of all landmark
Ž .
configurations translation and rotation based on a generalised least square method Ž
. GLS; Rohlf and Slice, 1990 . From the superimposed configuration, a consensus
Ž .
mean configuration is obtained and used as reference. All configurations are scaled
Ž according to a size measure known as centroid size the square root of the sum of
. squared distances from each landmark to the specimen’s centroid . Centroid size is
removed from the analysis of shape and can be analysed independently. Residuals after the superimposition can be visualised at each landmark location to explore deviations
Ž from the consensus configuration growth trends, within group shape variability; Walker,
. 1993 . On the other hand, the residuals and the consensus can be modeled with the
thin-plate spline function. The thin-plate spline algorithm uses the residuals of the superimposition to compute a new set of variables. The matrix containing this new set of
variables is called the weight matrix. Each row of the weight matrix represents an individual, each column the new variables. Thanks to the properties of the thin-plate
Ž .
spline function Bookstein, 1991 , the new variables can be analysed with traditional multivariate techniques and shape change or shape differences can be visualised as
Ž .
deformation grids splines . The first two columns of the weight matrix represent Ž
. localised shape features local deformations or non-uniform shape components ; the last
Ž two columns represent the uniform shape components U , shearing and U , stretching,
1 2
. along the major axis of the fish . Thus, the model offers the possibility to decompose the
Ž .
Ž .
shape change into uniform stretching and shearing and non-uniform localised compo- nents, which can be analysed, visualised and interpreted separately, adding to the rigour
of the statistical analyses, the quality and immediateness of pictorial representation. No information contained in the original landmark configurations is lost at this step except
Ž .
the one about size centroid size, which can be analysed independently , translation and Ž
. rotation which have no biological meaning .
Fig. 1. Landmarks collected on the European sea bass: top, juveniles; bottom, adults.
2.2. Geometric morphometrics: analyses Table 1 summarises sample sizes. Standard length was measured with a digital
caliper. The images of each specimen were recorded with a Hi8 camcorder. Images were Ž
. digitised using the image analysis system, Quantimet 970 Cambridge Instruments : 12
Ž .
landmarks were collected on juveniles and 17 on adult specimens Fig. 1 . Ž
Shape differences were computed for each age class with a MANOVA F values .
Ž approximated by Wilk’s l of the weight matrix, and U and U
computed with the
1 2
. software, TpsRelww . The scores of canonical variate analysis computed on the shape
Ž .
variables weight matrix were plotted as a histogram and were used as the independent variables in a regression analysis. Splines relative to the extreme values of the canonical
Ž .
variate scores were computed and plotted software TpsRegrw . These splines represent major shape differences between groups.
In order to verify if shape differences between FRA and ITA increased or decreased during growth, squared Mahalanobis distances were computed between the mean value
of each group on canonical axes. To compare results at the different ages, the same
Table 2 Classes of anomalies with legends as observed in the European sea bass
Region Ž
. Ø
A, cephalic 1st–2nd vertebra Ž
. Ø
B, pre-haemal 3rd–10th Ž
. Ø
C, haemal 11th–21st Ž
. Ø
D, caudal 22nd–24th Anomalies of the vertebral axis
Type 1: lordosis Type 2: kiphosis
Vertebral anomalies Type 3: fusion of the vertebral body
Type 4: deformation of the vertebral body Type 5: malformed neural arch andror neurapophisys
Type 6: malformed haemal arch andror hemapophisys Type B7: malformed pleural cost
Fins Ø
F, anal Ø
H, dorsal spines Ø
L, soft dorsal rays Fins anomalies
Ž .
Type 7: abnormal ray deformed, absent, fused, extranumerary Anomalies of the splanchnocranium
Type I4: elongated dental Type I5: reduced dental
Type I6: others Type I7: anomalies of the operculum
number of landmarks is needed for each class. Thus, only 12 of the 17 landmarks collected on adults were used. Morphological variability for each group at each age class
was measured as generalised variance. Ž
. The morphometric softwares cited TpsRelww and TpsRegrw were implemented by
F.J. Rohlf and are available at http:rrlife. bio.sunysb.edurmorph. 2.3. Internal anatomical analysis
After landmark data collection, all individuals were fixed in buffered formaldehyde Ž
. Ž
. 10 and X-rayed 4 minr5 mAr80 kW . Internal anatomical anomalies were then
Ž observed and classified according to Table 2. The binary matrix of anomalies trans-
. formed in frequency data was analysed with correspondence analysis. On the corre-
spondence analysis plot, only the classes of anomalies were visualised. Specimens scattered on the plot according to their anomalies’ patterns. Individual scores were used
as the independent variables in the regression analysis described in Section 2.4.
2.4. Geometric morphometrics combined with internal anatomical analysis The individual scores on each correspondence axis were used as the independent
Ž .
variables in a regression on the weight matrix software TpsRegrw . Individuals scored differently on correspondence axes according to particular anatomical patterns. This
Ž .
allowed for the visualisation of splines characteristic shapes relative to particular cadres of anomalies.
3. Results