Content uploaded by Cornelius Hammer
Author content
All content in this area was uploaded by Cornelius Hammer
Content may be subject to copyright.
Reviews in Fish Biology and Fisheries 13: 27–61, 2003.
© 2003 Kluwer Academic Publishers. Printed in the Netherlands.
27
Growth and reproduction of horse mackerel, Trachurus trachurus
(carangidae)
P. Abaunza
1
, L. Gordo
2
, C. Karlou-Riga
3
,A.Murta
4
, A.T.G.W. Eltink
5
,M.T.Garc
´
ıa
Santamar
´
ıa
6
, C. Zimmermann
7
,C.Hammer
8
,P.Lucio
9
,S.A.Iversen
10
,J.Molloy
11
& E. Gallo
1
1
Instituto Español de Oceanograf´ıa, Apdo. 240, 39080 Santander, Spain (E-mail: pablo.abaunza@st.ieo.es);
2
Facultade de Ciências da Universidade de Lisboa, Bloco C-2, Campo grande, 1749-016 Lisboa, Portugal;
3
Ministry of Agriculture, Fisheries laboratory, Karaoli and Demetriou 15, 18531 Piraeus, Greece;
4
Instituto
de Investigaçao das Pescas e do Mar, Av. Brasilia, 1400 Lisboa, Portugal;
5
Netherlands Institute for Fisheries
Research, P.O. Box 68, 1970 AB IJmuiden, The Netherlands;
6
Instituto Español de Oceanograf´ıa, Apdo. 1373,
38080 Santa Cruz de Tenerife, Spain;
7
Bundesforschungsanstalt für Fischerei, Institut für Seefischerei, Palmaille
9, D-22767 Hamburg, Germany;
8
Bundesforschungsanstalt für Fischerei, Institut für Ostseefischerei, An der
Jägerbäk 2, D-18069 Rostock, Germany;
9
AZTI, Isla de Txatxarramendi s/n, 48395 Sukarrieta, Bizkaia, Spain;
10
Institute of Marine Research, P.O. Box 1870 Nordnes, N-5817 Bergen, Norway;
11
The Fisheries Research Centre
of the Marine Institute, Abbottstown, Dublin 15, Ireland
Accepted 28 May 2003
Contents
Abstract page 27
Introduction 28
Material and methods 30
Horse mackerel growth 30
Aging fish: A historical development of age interpretation criteria in the Northeast Atlantic
Annuli formation
Measurements of growth
Factors affecting growth
Horse mackerel reproduction 43
Timing of reproduction
Fecundity
Determinate or indeterminate spawner
Conclusions 55
Acknowledgements 56
References 57
Key words: age reading, fecundity, growth, reproductive biology, spawning frequency, Trachurus trachurus
Abstract
There is a broad knowledge of the growth and reproduction of Trachurus trachurus, although important
gaps still exist. Horse mackerel are a long-lived species, reaching up to 40 years of age. They have isometric
growth, although the alometric parameter b may vary throughout the year and in relation to latitude. Growth to age
3 is rapid compared to slower growth later in life. Phenomena of density dependent growth have been observed
in the northeast Atlantic. Horse mackerel are an asynchronous species. The following stages of atresia have been
validated in horse mackerel: alpha, beta, and delta. The transition in females from the spawning state to post
spawning is very fast. The spawning fraction in horse mackerel is estimated to be between 8.3% and 20.9%. Horse
mackerel have a long spawning season (up to 8 months), which varies according to geography. Length at first
28
maturity is between 16 and 25 cm, most commonly around 21 cm. Males mature at a slightly smaller length
than females. The age at first maturity for females has been estimated to range from 2 to 4 years, depending on
the geography. Batch fecundity has been estimated to range between 172–209 oocytes per gram-female-weight.
Female spawning lasts between 65 and 94 days. An individual female can release from 5 to 16 batches during
the spawning period. The estimated potential annual relative fecundity ranges from 1040 to 3280 oocytes per
gram-female-weight.
Introduction
Horse mackerels (Trachuridae, Carangidae) are
widely distributed in the sea and support large fish-
eries. The total commercial catch of the genus Trach-
urus increased from about 1 mill t in 1960 to more than
6.5 mill t in 1995, and then decreased to 2.5 mill t in
1999. Since the early 1970s, the vast majority of these
catches (20–75%) were composed of the Chilean jack
mackerel (Trachurus murphy Nichols, 1920), caught
in the southeastern Pacific Ocean. In recent years, the
most northern representative of the trachurid family,
T. trachurus (Linnaeus, 1758) has ranked second in
catches (2000: 275,000 t, >10% of total catch; FAO,
2000).
In the northeast Atlantic Ocean and adjacent areas,
horse mackerel commonly occurs on the continental
shelf: from the West African Cape Verde Islands
(Crawford, 1987), northwards to the Norwegian Sea
and North Sea (Knijn et al., 1993), including Iceland,
as well as in the Mediterranean Sea (Fischer, 1973;
Tsangridis and Filippousis, 1991) and Black Sea
(Shiganova, 1998). In the study area, the Central and
Northeast Atlantic area, four Trachurus species are
found with slightly different but overlapping distri-
bution areas: T. trachurus, T. mediterraneus (Stein-
dachner, 1868), T. picturatus (Bowdich, 1825), and
T. trecae (Cadenat, 1949; Smith-Vanith, 1986; Suda
et al., 1995). The most important of these species in
terms of catch is T. trachurus. In the central eastern
Atlantic, three stocks of T. trachurus are considered,
being the most important for the fishery the Saharo-
Mauritanian stock (Maxim, 1995). There is currently
no clear separation between horse mackerel stocks in
the Mediterranean and, for practical reasons, Caddy
(1998) suggested the separation of stocks based on
areas defined from commercial landing data.
In the northeast Atlantic, three stocks of T. trach-
urus have been defined for management and assess-
ment purposes and it is assumed that these form dis-
tinct spawning populations: the southern horse mac-
kerel around the Iberian peninsula; the western horse
mackerel in the Norwegian Sea, northern North Sea,
west and south of the British Isles, western English
Channel, and west of France; and the North Sea
horse mackerel, mostly restricted to the central and
southern parts of the North Sea, and eastern English
Channel. According to Rückert et al. (2002), the latter
stock is divided into a distinct northern and southern
component.
Horse mackerel are a migratory species and the
distribution of the western and North Sea stocks
overlap partly during over-wintering in the Eng-
lish Channel (Macer, 1977). Their spawning and
feeding areas seem to be separated (Eltink, 1992;
ICES, 1999a). Approaches to separating stocks have
included morphometry (Lourdes Marecos, 1986;
Murta, 2000) and genetics (Borges et al., 1993a).
Currently, attempts are being made to develop a
sound scientific basis for stock separation by simultan-
eously applying a variety of different methods, such
as genetic markers and biological tags (Abaunza, IEO
Santander, Spain, pers. comm.). The southern and
western horse mackerel stocks are assessed annually
by the International Council for the Exploration of
the Sea (ICES). Age-structured models are applied
in the assessments, and information from commer-
cial catches and egg and demersal surveys are used
to obtain biomass and number-at-age estimates (e.g.,
ICES, 1996, 1997). The stock size depends on (a) the
rate of exploitation and natural mortality, (b) recruit-
ment, and (c) individual growth rate. Recently, ICES
estimated the spawning biomass of the southern stock
to be in the range of 250,000–500,000 t (Porteiro et
al., 1993; ICES, 2001). Based on egg surveys, the
size of the North Sea stock was estimated from 1988-
1991 to be just over 200,000 t (Eltink, 1990, 1991a,
1992). Rückert et al. (2002) distinguished two distinct
components (the northern and the southern) in the
North Sea stock. They estimated the biomass of the
southern component for the period 1991–1997 to be
in the range of 57,000–248,000 t. These calculations
were based on a ratio estimate that related catch-per-
unit-effort (CPUE) data from surveys of horse mack-
29
erel to the CPUE data of a reference species from the
same surveys (Rückert et al., 2002). The western horse
mackerel is the largest of the three stocks. Stock size
has fluctuated throughout the past two decades due
to considerable differences in recruitment and varying
levels of exploitation. The stock reached its maximum
in 1988 at nearly 2.9 mill t and is expected to reach
0.9 mill t in 2000 (ICES, 2002a,b). Over the past 15
years, the stock was dominated by the extraordinarily
strong 1982-year class, which could be traced until
very recently. Accordingly, fishing mortality on this
stock has increased drastically since 1988 and stock
size has declined.
Both the stock and the catches of the Saharo-
Mauritanian horse mackerel have decreased since
1984 due to fishing pressure and weak recruitment
(Maxim, 1995). From a maximum spawning stock
biomass of 1.7 mill t in 1980, it declined to 700,000 t
in 1992. Based on an analysis of the historical series
of catches, Fiorentini et al. (1997) classified the horse
mackerel fishery in the western Mediterranean as an
“intermittent” fishery, while catches in the eastern
Mediterranean show a linear increasing trend (rising
fishery).
Horse mackerel is a schooling species, caught in
the northeast Atlantic mainly with pelagic trawls and
purse seines, often close to the sea floor. Horse mack-
erel schools often mix with mackerel schools (Van
Marlen, 2000). Observations from Wardle et al. (1996)
and Herrmann and Enders (2000) suggest that, unlike
mackerel, horse mackerel are neutrally buoyant, and
can remain inactive with only gentle pectoral move-
ments and no continuous swimming motion, at least
under captive conditions. Enders (1998) confirmed,
on the basis of oxygen consumption measurements,
that horse mackerel are good swimmers adapted to
swimming at a low, but very constant, speed.
Shelf attachment is a predominant distributional
pattern of western and southern horse mackerel stocks
(e.g., Dornheim, 1987; Macer, 1977; Dornheim and
Kerstan, 1985; Eaton, 1989; Dornheim, 1993; Porteiro
et al., 1993). However, for the North Sea stock, a
distribution in the southern and eastern coastal areas
of the central and southern North Sea was already
described by Olsen (1883).
Horse mackerel have distinct areas for spawning,
feeding, and over-wintering. Migration is probably
driven by water temperature and availability of prey
organisms; in autumn, at a temperature below about
10
◦
C, T. trachurus retreat from feeding areas in the
Norwegian Sea and the North Sea, and migrate to
the over-wintering areas further south, predominantly
in the western English Channel (for the North Sea
Stock; Lockwood and Johnson, 1977; Macer, 1974,
1977), some areas around the British Isles on the
continental slope (Macer, 1977), and the eastern Bay
of Biscay shelf edge (for the western stock; Eaton,
1983). In winter they form dense schools in the
deeper waters, while in spring they become far more
dispersed (Polonsky, 1965) and migrate northwards
with increasing water temperature (e.g., Chuksin and
Nazarov, 1989). The North Sea stock appears in
April at the southern Dutch and English coasts (Meek,
1916), and reaches the western Jutland coast and
southern Norwegian coast by August. On occasion,
larger individuals of the western stock reach as far
north as Trondheim in Jul.–Aug. Other parts of this
stock feed in areas west of Ireland or at the Bay
of Biscay continental slopes. The southern stock has
large overlaps between spawning and feeding areas.
The Saharo-Mauritanian stock migrates annually
between 15
◦
N(Jan)and26
◦
N (Cape Bojador; May).
It can be found at Cape Timiris (19
◦
N) in November
(Maxim, 1995).
Apparently, the water temperature of 8
◦
Cisthe
lower limit for horse mackerel, which they avoid
during over-wintering (Polonsky, 1965), and they stop
feeding at water temperatures below 9
◦
C (Herrmann,
Institute of Hydrobiology, Hamburg, Germany, pers.
comm.). In the southern areas, the required salin-
ities seem to be above 35.2 ppm (Polonsky, 1965)
and Lozano Cabo (1952) gave optimal water temper-
atures of 19–23
◦
C, thus higher temperatures seem
to be avoided. Horse mackerel seem to digest their
food relatively quickly, as compared to cod and whit-
ing (Temming and Herrmann, 2001a). This, how-
ever, depends not only on temperature, but also
on prey type and its energy content (Temming
and Herrmann, 2001b). Food intake is strongly
temperature-dependent, with a doubling of the intake
rate at 13
◦
C compared with 10.7
◦
C (Temming and
Herrmann, 2001a).
Fat and energy content of adult horse mackerel
is lowest during and after spawning in late spring
and summertime, and highest in autumn (Sahrhage,
1970; Lucio, AZTI Sukarrieta, Spain, pers. comm.).
In the North Sea, in August and September body
energy content rises rapidly in horse mackerel, but
feeding ceases as soon as the water temperature drops
below 10
◦
C (Herrmann, Institute of Hydrobiology,
Hamburg, Germany, pers. comm.). At 8–9
◦
C, the fish
stop feeding completely and leave the area to over-
30
winter. In spring, before spawning, only limited fat
reserves are left in the gut and muscle (Leloup and
Gilis, 1964; Polonsky, 1965).
Horse mackerel are a fairly long-lived species,
reaching a maximum age of well over 30 years
(Eltink and Kuiter, 1989). Therefore, an occasional
strong year class can lead to high abundance of
horse mackerel. In the case of the western horse
mackerel, the extraordinary strong 1982 year-class
created a substantial fishery in the northern areas,
which continued for more than a decade. As a result,
T. trachurus became an important commercial species
in the 1980s and 90s and is now one of the three
most important pelagic species in the European fish
industry. Another reason for this is the decrease of the
availability of resources due to over-fishing and the
subsequent increase of value of other species.
The extraordinary 1982 year class posed a meth-
odological problem with regard to the assessment and
management of this species. It was not able to produce
a comparably strong year-class itself. The recruitment
to this stock seems to be more dependent on environ-
mental factors than on the size of the parental stock.
Knowledge of the recruitment processes is, however,
critical for predicting stock development. For this
reason, it is imperative to approach the question of
which factors contribute to horse mackerel recruit-
ment. Moreover, there is great plasticity in the growth
of horse mackerel. As a result, it is possible for indi-
viduals of only a few years of age to reach sizes that
are attained by other individuals or populations only
many years later. Consequently, it is difficult to estab-
lish a reliable relationship between fish length and
age.
The decoupling of stock size and recruitment,
and the variability of growth raise important ques-
tions about how these factors determine the structure
and the condition of the stocks. This paper aims to
summarize current knowledge on growth and repro-
duction parameters, crucial for any sensible assess-
ment and management of the stock. It also explores
– for the first time for this fish species – any evidence
of density dependent growth.
Material and methods
The documents used for this review were compiled
through bibliographic searches and analysis of biblio-
graphic resources for each of the fisheries institutions
involved in this work. This produced a great deal of
information otherwise difficult to access, published
in the form of internal reports, doctoral theses, and
other kinds of gray literature, such as communications
to congresses and working documents to different
meetings, etc.
In this review, we analyzed the databases presented
to ICES for the assessment of horse mackerel stocks
in the northeast Atlantic, to study the growth of horse
mackerel, particularly to investigate possible occur-
rences of density dependent growth. This specific
analysis was carried out on western horse mack-
erel. Assessments of horse mackerel have used age-
structured models. The age composition of the catches
is obtained by otolith reading, following criteria estab-
lished during a number of otolith reading workshops
(ICES, 1991a, 1999a). Age data for the western stock
was collected by the Netherlands, Norway, Ireland,
Germany, Spain, and the United Kingdom, covering
approximately 51% of the total catch in 1999 and 56%
in 2000. The international mean length-at-age for the
western stock was estimated by ad hoc procedures that
combine the age and catch-data taken from the same
area and in the same period of the year.
To study the relationship between geographical
factors (latitude) and the alometric parameter of
the length-weight relationship obtained by different
authors, a simple linear regression was applied and
the adjusted coefficient of multiple determination (R
2
)
calculated.
Horse mackerel growth
Aging fish: A historical development of age
interpretation criteria in the Northeast Atlantic
To age horse mackerel (T. trachurus) in the North-
east Atlantic, a variety of methods for processing
different calcified structures as well as different age-
reading methods have been used. Letaconnoux (1951),
Ramalho and Pinto (1956), and Barraca (1963) did not
find satisfactory results when using scales, otoliths,
or other calcified structures, and restricted their work
on growth to an analysis of length distributions (e.g.,
Petersen, 1892). Most other authors (Baidalinov and
Staroselskaya, 1964; Polonsky, 1969; Sahrhage, 1970)
have attempted to use the unburnt otoliths, but have
commented on the difficulty of interpreting the ring
structure. Macer (1968) started to burn the trans-
verse sections of the otolith according to the method
of Møller Christensen (1964). This clarified the ring
31
structure considerably. Sahrhage (1970) stated that
further investigations were necessary to solve the
question of whether or not each translucent zone
should be counted as an annulus, and he mentioned
migration as a likely factor in causing secondary ring
structures.
Macer (1977) investigated a variety of structures
including bones, scales, and otoliths to assess their
suitability for age determination. Only otoliths showed
a consistently clear ring structure, particularly when
the ring structure was clarified by a treatment of
breaking, polishing, and burning. Macer validated
the agings for the first four years of life by using
the length-frequency method and demonstrated the
occurrence of one translucent ring per annual growth
zone. He could not prove whether the large number
of rings observed (up to 35) in burnt otoliths of four
year and older fish were formed annually. The inter-
pretation of one translucent ring per annual growth
zone did conflict, at that time, with the work of
previous authors (Baidalinov and Staroselskaya, 1964;
Polonsky, 1969), who reported no difficulties in using
whole otoliths. They determined a maximum age of
nine and 10 years in their Bay of Biscay samples
respectively.
Trouvery (1977) mainly used scales, but she also
tried the burning technique on a limited number of
transverse sections of otoliths. This caused problems
in the interpretation, because there were less rings in
the scales compared to the otoliths. Lourdes Marecos
et al. (1978) determined the age by counting the trans-
lucent zones of whole otoliths. They regarded the
burning technique to be of possible use to solve the
differences in age estimation between readers. Eaton
(1983) used both broken/burnt and whole otoliths for
age estimation. He stated that it is possible to age
horse mackerel without difficulty up to at least seven
years old, and, thereafter, it becomes increasingly
difficult, but he found that horse mackerel are at least
as long-lived as mackerel, 15 years or more.
In the early 1980s, the different laboratories pro-
viding age data to the ICES working groups deter-
mined a wide range of growth parameters, indicating
that different age reading methods might have been
applied. To detect these differences, an ICES recom-
mended (ICES, 1984) otolith exchange was conducted
(Eltink, 1985). The evaluation of this exercise demon-
strated that the determined age varied by a factor of
two among different readers, mostly depending on the
aging method. It was concluded that consistent inter-
pretation of horse mackerel otoliths was achievable up
to five years of age. To improve the method of aging
fish older than five years, a workshop was organized
in 1987. However, this serious problem could not be
solved during the workshop and it was concluded that
a validation study would be needed to indicate which
age reading method would be correct.
Macer (1977), Kerstan (1985), and Junquera et al.
(1988) tried validation of age reading methods. Unfor-
tunately, Macer (1977) did not succeed in an indirect
validation by following year classes in successive
years because the time series was too short and the
differences in year class strength were not sufficiently
evident. Kerstan (1985) stated that scanning electron
microscopy and light microscopy proved the rela-
tion between otolith surface structures and annual
growth zones in whole otoliths. However, this cannot
be regarded as validation because the identification
of these annual growth zones was not confirmed.
Junquera et al. (1988) tried to test whether frequency
modes might be identified to discriminate between
annuli and false rings by measuring the distances from
the center of the otolith to all translucent zones. As
many radii as translucent zones were recorded, but
whenever one of them was obviously an annulus it was
recorded separately in order to obtain the frequency
distribution of radii belonging to each age. However,
these obviously true annuli are the only annuli if a
validated age reading technique is used. This method
is therefore biased, being dependent on the input
values for the annuli. The only indirect validation was
obtained from a comparison between ages and the
length-frequencydistributions (Petersen, 1892), which
has confirmed the age determination in the first years
of life up to age four (Letaconnoux, 1951; Ramalho
and Pinto, 1956; Barraca, 1963; Polonsky, 1969;
Sahrhage, 1970; Macer, 1977).
The indirect validation method of comparing suc-
cessive year’s age compositions was chosen for the
purpose of indicating which age reading method
would be correct (Eltink and Kuiter, 1989). In prin-
ciple, only one of the two age reading methods can
follow strong/weak year classes in successive year’s
age compositions if the age reading methods differ
approximately by a factor of two. Fortunately, an
extremely strong year-class was produced in 1982,
mostly by the western stock, and entered the fishery
some years later. This provided a unique oppor-
tunity for such an indirect validation of the age
reading methods. In 1988, all age readers within the
ICES community accepted the age reading method
previously applied by the Netherlands (one translu-
32
cent ring per annual growth zone; ICES, 1988) and
subsequently horse mackerel were found to be much
older than believed so far. Horse mackerel can reach
an age of approximately 40 years and are therefore a
long-lived species.
Two more exchange programs were carried out
in 1988/89 (Borges, 1989) and 1996 (Eltink, 1997),
each followed by a workshop. The second workshop
was held in Lisbon in 1990 (ICES, 1991a), mainly
aiming at reducing the standard deviations for age of
fish collected from Portuguese waters compared to the
more northern Atlantic areas. For the third exchange
and workshop, otoliths of “known” age (broken/burnt
and whole otoliths of the 1982 year-class collected
during 1983–1995) could be used. These otoliths had a
very high probability that the originally estimated age
was corrected and were therefore treated as otoliths of
“known” or “actual” age. This provided information
on the precision and accuracy of age reading by age
group for individual readers and all readers combined.
A serious bias problem in age reading could be iden-
tified: the ages of fish from approximately age seven
onwards were underestimated, and the interpretation
of the edge of otoliths appeared to be a major diffi-
culty. The third workshop, held in Lowestoft, UK, in
1999 (ICES, 1999a) aimed at solving both problems.
While an evident improvement in the precision and
the accuracy for almost all readers could be noted,
the underestimation of the older ages (bias) could
not be significantly reduced. The latter appears to be
caused by poor visibility of the peripheral rings in
broken/burnt otoliths of older fish. To improve the
visibility of these structures and reduce the age reading
bias, the workshop repeated a recommendation of the
previous workshop to compare the sliced transverse
section technique with the traditional broken/burnt
transverse section technique.
Eltink (2001) carried out this study, again using
otoliths of “known” age from the 1982 year-class,
and additionally tested different staining methods.
The age readings of the unstained sliced transverse
sectioned otoliths appeared to be less precise and
less accurate than age readings of the broken/burnt
otoliths. Staining sliced otoliths with acidified Neutral
Red, as recommended by Richter and McDermott
(1990), improved the precision and accuracy only
slightly, while using a light wood stain, Honey
Pine Light Fast Stain (Supplier: Morrels Wood-
finishes, UK; http://www.morrels-woodfinishes.com),
provided more precise and more accurate results. This
indicated that the stain is probably more important for
improved visibility of the translucent and opaque rings
than the slicing process, and that different staining
methods should be investigated in the future.
An analysis of the effect of age reading errors on
the horse mackerel assessment (addendum to ICES,
1999a) showed that precision errors and precision
errors combined with bias have opposite effects on the
estimation of recruitment, spawning stock biomass,
and fishing mortality, and tend to compensate for
each other. The results were expressed as percentage
over- or underestimation of the mentioned parameters,
as well as population at age and selection pattern.
Figure 1 shows as an example the effect on the esti-
mates of spawning stock biomass. The precision errors
cause an overestimation of age, because younger year
classes are more abundant causing relatively more fish
to be transferred from younger year classes to older
ones than vice versa. Underestimation of the older
ages (bias) has the opposite effect because it causes
relatively old fish to be transferred to younger year
classes. A low precision (CV = 15%) combined with
bias resulted in better assessment estimates than a
high precision (CV = 5%) combined with bias. This
implied that the precision in age reading should not be
improved before the bias problem would be solved.
More recently, Waldron and Kerstan (2001) valid-
ated the age determination of horse mackerel otoliths
from whole otoliths up to age four. They compared
marginal increment (MI) widths (Kerstan, 1985,
2000), measured from annuli under a light micro-
scope, with daily increment counts obtained with
a scanning electron microscope (direct validation
method). The estimated ages (0.6–4.3 years) agreed
well for horse mackerel up to four years old. Exami-
nation of subsequent growth zones indicated that false
rings and annuli are often of a similar appearance and
that true annuli can only be identified if concurrent
measurements of growth zone widths are available.
It appears that age reading of horse mackerel is
specifically difficult because:
• Large discrepancies exist in the number of rings
between scales for whole otoliths and broken/burnt
otoliths;
• For ages 5–7 and older, the growth in both the
long and short axis of the otolith slows down with
age, occurring mainly in the lateromedial axis,
resulting in a thickening of the otolith (Macer,
1968, 1977; Geldenhuys, 1973; Nazarov, 1978);
• It was hard to accept that horse mackerel are such
a long-lived species, which can reach an age of 40
years;
33
Figure 1. Effect of age reading errors (precision and accuracy) on the estimation of spawning stock biomass (SSB) of northeast Atlantic Ocean
horse mackerel (western stock), with the deviation from the SSB value used in the assessment considering different levels of CV, in relation
with the accuracy in age readings, and the presence or amount of bias (precision in age readings). Recruitment factor is 16 (From addendum of
ICES, 1999a).
• False rings might occur not only during the
juvenile period (Fariña Pérez, 1983), but during
the adult period as well (Arruda, 1987). Otoliths
collected in Iberian waters appear to produce more
frequent false rings in between the true annuli
compared to otoliths from the more northern area,
where only occasional false rings occur during
the juvenile phase (Eltink, RIVO Ijmuiden, the
Netherlands, pers. comm.); and
• Protocols for an indirect validation of the age
reading method for the ages 5 and older were avail-
able only after the emergence of the strong 1982
year class (Eltink and Kuiter, 1989).
More information on the terminology, guidelines and
tools for age reading comparisons (including a small
program to ease the evaluation; Eltink, 2000) can be
found at the European Fish Aging Network’s home-
page (http://www.efan.no/guidelines/Guidelines.zip).
Annuli formation
As in other teleost (bony) fish, horse mackerel otoliths
are small carbonate bodies found in their inner ears
and form part of the hearing and balance system. The
inner ear consists of three sacs called the utriculus,
lagena,andsacculus, which are connected by the
semi-circular canals. There are three otoliths on each
side of the head known as the lapillus, asteriscus,and
sagitta. In horse mackerel, the asteriscus and lapillus
measure only a few millimeters, but the sagitta can
measure close to a centimeter and a half. Being the
largest of the three, the sagitta is the most widely
studied and practically the only one used for fish aging
purposes. It is often simply referred to by the term
otolith, although this term more correctly applies to
all three bodies of the inner ears.
The shape of the horse mackerel sagitta is not
constant. It varies according to the fish age, both in
the relative proportions of their different parts and in
the developing of the marginal denticulation (Lozano
Cabo, 1952). According to Härkönen (1986), the horse
34
mackerel sagitta has a shape oval to elliptical, with
an anterior end more pointed than the posterior. The
dorsal margin is irregular, but there are only a few
cuts or lobes present. The ventral margin is often
strongly, but mostly shallowly, lobate and is more
convex than the dorsal. The posterior end is highly
variable. Rostrum is big and bluntly pointed, where the
dorsal margin is smooth, and the ventral margin of the
rostrum is lobed. Antirostrum is small and rounded or
pointed with a strongly convex surface smooth inside.
Sulcus is moderately deep. Cauda is more than double
the length of the ostium. Margins of the cauda are
parallel in the central section. The posterior end is
curved to the ventral and very shallow. The ostium is
wide, funnel shaped, and slightly curved to the dorsal.
A very shallow area is present dorsal of the sulcus with
a slightly concave outside. Radial lines leading to the
margins of the sagitta can be seen here.
Inner composition of the sagitta otolith
Carbonates (CaCO
3
) are the mineral component of
otoliths. Calcium carbonate has three polymorphs;
they have the same chemical composition (CaCO
3
),
but different crystal structures: calcite, aragonite,
and vaterite. Fish otoliths have been found to be
monomineralic and most are composed of aragonite
(Degens et al., 1969). Fish otoliths can be considered
“acellular bone” – they do not contain osteocytes, but
are composed of a mineral component and an organic
matrix (Fleming, 1967; Pannella, 1980). The organic
matrix component is known as otolin. This matrix has
been found to have a high abundance of acidic amino
acids and has essentially the same chemical compos-
ition for all fish (Degens et al., 1969; Watabe et al.,
1982). It usually composes 0.2 to 10% of the otolith
by weight. The area for precipitation of carbonate is
the inner ear. This area is completely isolated from
the environment and the rest of the fish. The fluid
of the inner ear is called endolymph and it is from
this fluid that the otoliths are precipitated. One of
the first studies of otolith formation was carried out
by Degens et al. (1969). They presented a model
for otolith formation in which carboxyl groups at the
surface of the otolin provide oxygen that coordinates
with Ca
2
+ to form metal ion polyhedra. Bicarbonate
becomes linked via hydrogen bridges to the organic
template and oxygen substitution occurs between the
bicarbonate and the polyhedra, which stabilizes the
structure. The result is the formation of Ca
2
+O
9
poly-
hedra providing nucleation sites for the precipitation
of carbonate. The fact that these are Ca
2
+O
9
poly-
hedra determines the formation of aragonite instead of
calcite. Thin section examination reveals that otoliths
have a detailed microstructure consisting of bands of
opaque and translucent material at various scales. The
very central portion of the otolith is known as the
kernel, or primordium. This represents the earliest
phase of calcification. It is thought that otolith form-
ation occurs very early in the development of the fish
and that they are often the first calcified structures to
form (Campana and Neilson, 1985). The kernel, along
with the first opaque ring, constitutes the nucleus
of the otolith. Aragonite needles grow perpendicu-
larly to the growth increments. Some are truncated at
the discontinuous zones, while others are continuous
between the layers (Degens et al., 1969; Watabe et al.,
1982).
Three main phases of post-kernel growth can be
recognized (Pannella, 1980). During the first phase,
there is a fast growth rate and only rare discontinuities
are seen. Growth is nearly isotropic in all directions.
The growth rate falls during the second phase, the
width of the increments becomes smaller and interrup-
tions in deposition become more frequent. There is a
further reduction in the growth rate during the third
phase, and discontinuities and unconformities become
more evident.
As the fish develops from a larva to a juvenile,
accretion rates onto different portions of the otolith
change. This causes the otolith to develop from a
spherical form into its adult form (Lagardere et al.,
1995). Studies using 45Ca have shown that macular
cells are largely responsible for secreting Ca for this
precipitation process (Pannella, 1980; Campana and
Neilson, 1985). These cells are concentrated in certain
areas and thus, control the form that the otolith will
eventually take.
The most obvious bands are the annuli. They can
appear as opaque or translucent (hyaline) bands when
they are viewed under reflected light, but it must be
taken into account that under transmitted light their
appearance is the contrary one. These opaque and
translucent bands are thought to correlate with fast
and slow growth periods, respectively. The translu-
cent zones are dominated by organic material while
the opaque zones are dominated by carbonate (Degens
et al., 1969; Watabe et al., 1982). In some fish from
temperate areas, it is thought that opaque zones corres-
pond with summer growth, while translucent zones
form during winter. For Beckman and Wilson (1995),
the majority of fish formed opaque zones during the
spring and summer months. But, this probably does
35
not occur in all species in all their life periods, as in
the case of horse mackerel.
It is clear that carbonate precipitation and/or matrix
formation must vary cyclically in order to produce
growth increments. The formation of otolin has not
been examined in detail. Watabe et al. (1982) postu-
lated that matrix deposition is continuous although
not necessarily constant. Some work has been done
to study the deposition of carbonate and it has been
found to slow around sunrise for some fish (Watabe et
al., 1982; Campana and Neilson, 1985). A study by
Woodhead (1968) showed that although plasma Ca
2
+
levels varied seasonally in Arctic cod, this could not be
correlated with otolith increment formation, however,
the work by Pannella (1980) conflicted with these
findings.
It has more recently been recognized that otoliths
have even finer increments than these seasonal zones.
Monthly, daily, and even lunar zones have been
recognized (Campana and Neilson, 1985). Lunar
zones probably represent times of stress imposed on
some species of fish by the tidal cycle (Campana
and Neilson, 1985). The possibility of daily zones
was first examined by Pannella (1971). The factors
governing the formation of the various zones are not
clear, however, photoperiod, temperature, food avail-
ability, growth, and reproductive activity have all been
examined for their effects (Campana and Neilson,
1985; Casselman, 1987). A model for daily increment
growth by an endogenous circadian rhythm, which
can be affected by environmental variables, was put
forward by Campana and Neilson (1985). Some of the
work in this area has been contradictory and the exact
mechanisms of otolith formation are still not fully
understood. It is probably a complex combination of
environmental and physiological factors.
Most mineralized tissues undergo some kind of
reworking and resorption, but this has not been
observed in otoliths (Campana and Neilson, 1985;
Casselman, 1987). Some researchers suggest that this
is because of the otoliths important function as part
of the hearing and balance system. Pannella (1980)
has suggested that resorption does occur, however it
seems that little work has been done to support this.
In some cases, we have found horse mackerel otoliths
that appear totally or partially in a “crystallized” form
(less than 1%; Lucio, AZTI, Sukarrieta, Spain, pers.
comm.).
In horse mackerel, as in other species, although
it is relatively straightforward to determinate de visu
the type of the different bands that appear in the
inner part of an otolith, it is not always so easy to
assess macroscopically and with precision the exact
type of border founding of any otolith. The changes
from one to another type of border usually do not
occur suddenly and drastically due to the inherent
physiological processes in the fish (forward and back-
ward “ministeps” could be imagined along a short time
period). Moreover when otoliths are viewed under an
optic microscope, the fine, more extreme parts of the
otolith become very difficult to be classified with all
assurance as belonging to one or another type of band.
In spite of all these restrictions, and based on the
reading of an important number of otoliths collected
in the Bay of Biscay between 1987 and 1992 and
assessed by experienced readers, some conclusions
can be obtained in relation to the monthly border
formation in horse mackerel.
1. When sagitta otoliths of all ages of fish are con-
sidered:
− No less than about 15% of opaque and trans-
lucent bands in the borders are present in all
months;
− April, May, June, and July are the months with
lower proportions of opaque band in the border;
− Autumn and winter seasons constitute the year
period in which the highest proportion of
opaque band in the border is observed; and
− No consistent differences between males and
females were found in relation to the proportion
of type of border.
2. When only sagitta otoliths of mainly adult fish (3+
years old) are considered:
− The period April-July continues to be the time
with a lower proportion of opaque band in the
border, but above all May and June. These
months correspond to the period of the year in
which the peak spawning takes place in the Bay
of Biscay; and
− The autumn and winter period is the time
with highest proportions of opaque band in the
border. This time corresponds clearly to the
resting period of this species in the Bay of
Biscay. In winter, the proportion of opaque band
begins to decrease as time of fish maturation
advances.
3. When just sagitta otoliths of presumed imma-
ture fish (0–2 years old) are considered, some
differences can be found in relation to adult fish:
− Summer seems to be the period of the year with
higher proportion of opaque band in the border;
and
36
− Autumn, winter, and spring appear to be the
time with lower proportions of opaque band in
the border.
4. These monthly results on type of border match
fairly well with the monthly total lipids content in
the horse mackerel body from the same sea area
in 1989 (Lucio, AZTI, Sukarrieta, Spain, pers.
comm.):
− For adult fish (≥ 30 cm total length), the total
lipids reach the lowest values in May, June,
and July, and the highest values in autumn and
winter; and
− For immature fish (< 20 cm total length), the
total lipids reach the lowest values in autumn
and winter, and the highest values at the end of
spring and summer.
In the eastern Mediterranean area, the formation
of translucent bands occurs during spawning season
(Karlou-Riga, 1995).
Measurements of growth
Growth of an organism can be defined as the balance
between the energy input provided by food consump-
tion and the output represented by the losses associ-
ated with maintenance (Wootton, 1998). This equation
could be expressed in terms of an energy budget, but
most growth studies deal with the length or the weight
of the fish (Ricker, 1979). Horse mackerel are not an
exception and the majority of references dealing with
this species’ growth focus on changes in length and
weight in space and over time.
Length-weight relationships
The concept of growth implies changes in the relative
magnitudes of length and weight (Weatherley and Gill,
1987). Length or weight data combined with age
information can be used to construct growth curves.
Length data are easier to obtain than weight, and
both are easier to obtain than fish age. It is there-
fore desirable to obtain a relationship between length
and weight to convert from one variable to the other.
The most commonly used expression is the allometric
relationship:
W = a · L
b
(1)
Where W is the weight of the fish; L is the length;
and a and b are constants. The value of b is usually
determined by the logarithmic form of equation 1.
log(W ) = log(a) + b · log(L) (2)
In this equation b is the slope of the linear regression
of the logarithm of weight against the logarithm of
length.
The length-weight relationship for horse mack-
erel has been estimated from different areas covering
almost the entire range of the distribution in the north-
east Atlantic (see Table 1). However, in the Mediter-
ranean Sea the references are scarce (Carrillo, 1978;
Alegría- Hernández, 1984a; Karlou-Riga and Sinis,
1997). The mean value of the allometric parameter
b found in the different studies is 3.002 (± 0.089,
n = 18), which means that in general, horse mack-
erel grows isometrically. However, there are differ-
ences, some of them significant (i.e., Arruda, 1984),
among the areas in the b parameter estimates. A weak
trend in relation with latitude can be observed, with
a minimum at about 41
◦
N and an increase of b to
the north and south. In the northeast Atlantic Ocean
area, the value of b decreases more or less steadily
(R
2
= 0.426, n = 9) from approximately 20
◦
Nto
41
◦
N, where the two recorded cases show negative
allometric growth (the fish becomes lighter for its
length as it grows). Arruda (1984) also described such
a pattern along the Portuguese coast. From 41
◦
Nto
higher latitudes an opposite trend, although weak, can
be observed (R
2
= 0.41, n = 8), indicating positive
allometric growth. The reason for this might be the
availability of food to Atlantic horse mackerel: the
northwest of Africa (low latitudes) is one of the
most productive areas of the world because of the
intensity of the upwelling (Margalef, 1974). Along
most eastern Atlantic coasts of Europe, especially
off the Iberian Peninsula during the spring-summer
seasons, this phenomenon is also typical (Santos
et al., 2001), although weaker than off the north-
west coast of Africa. The intensity of upwelling is
reflected in the differences in primary production and
is almost linearly related to the secondary production
(zooplankton), which is the main food for horse mack-
erel (Ben-Salem, 1988; Olaso et al., 1999; Cabral
and Murta, 2002). On the other hand, the northern
European waters are characterized by the presence
of extensive shelves with areas of intermediate-high
productivity, which support rich benthic communities
(Ware, 2001). These conditions could be related with
the increase of the allometric parameter observed in
the most northern areas as compared to respective
values of b at 41
◦
N.
Until now, no differences have been observed in
the length-weight relationships between sexes (Fariña
Pérez, 1983; Lucio and Martín, 1989; Karlou-Riga
37
Table 1. Horse mackerel growth parameters from VBGF and length-weight relationship, provided by different authors. Included are
information on sampling period, area of coverage, and the aging material used
Author Sampling Area Growth Length/Weight
period Model Aging material t
0
kL
inf
relationship
ab
Letaconnoux (1951) 1943–46 Bay of Biscay Mean length-at-age Length structure
Lozano Cabo (1952) 1946–47 NW of Spain Mean length-at-age Scales 3.073
Planas and Vives 1950–52 W Mediterranean Mean length-at-age Scales 0.0043
1
3.17
1
(1953)
Anad
´
on (1960) 1955 NW of Spain Mean length-at-age Length structure 0.00816 3.023
Overko (1974) 1971–73 NW of Africa Mean length-at-age Otoliths and rays
Wengrzyn (1975) 1967–74 NW Africa Von Bertalanffy Length structure –2.32 0.13 50 0.0049 3.14
Borges et al. (1977) 1976 N Portuguese coast 2.931
Central Port. coast 2.936
S Portuguese coast 2.962
Macer (1977) 1967–70 W English Channel Mean length-at-age Otoliths
and North Sea
Trouvery (1977) 1975–76 Bay of Biscay and Von Bertalanffy Scales –0.59 0.20 44.88 0.158
6
1.83
6
Celtic platform 0.0063
7
3.08
7
Carrillo (1978) 1977–78 NW Mediterranean Von Bertalanffy Otoliths –1.016 0.22 37.66 0.0102 2.945
Nazarov (1978) 1968–77 Bay of Biscay and Von Bertalanffy Otoliths –1.347 0.205 40.0 0.00585 3.087
Celtic Sea
English Channel –1.515 0.18 39.2 0.00540 3.114
and North Sea
Portuguese coast 0.00859 2.961
Lourdes Marecos et 1976 N Portuguese coast Von Bertalanffy Otoliths –0.692 0.221 41.68
al. (1978) S Portuguese coast –1.024 0.163 51.74
Carrasco (1980) 1979 Cantabrian Sea Von Bertalanffy Otoliths –2.266 0.111 51.80 0.0145 2.812
Fariña P
´
erez (1983) 1982 NW Spain Von Bertalanffy Otoliths –0.982 0.225 40.90 0.01291 2.8545
Alegr
´
ıa Hern
´
andez 1980–81 Adriatic Sea Von Bertalanffy Otoliths –1.28 0.218 37.55 2.92
(1984a,b)
Arruda (1984) 1978–81 N Portuguese coast Von Bertalanffy Otoliths –3.86 0.119 41.05 0.0199 2.885
Center Port. coast –3.78 0.123 41.05 0.0173 2.927
S Portuguese coast –3.72 0.132 41.05 0.0135 3.005
Turner et al. (1984) 1976–83 NW Africa Von Bertalanffy Length structure 0.33 38.0
Kerstan (1985) NE Atlantic Von Bertalanffy Otoliths –0.65
3
0.223
3
41.59
3
0.0043
1
3.125
1
(Ireland and UK) –1.49
4
0.187
4
43.19
4
0.0044 3.141
–2.35
5
0.131
5
49.64
5
Junquera et al. (1988) 1984–86 NW of Spain Von Bertalanffy Otoliths –0.17 0.19 42.04
Back calculation –0.31 0.18 57.15
Lucio and Mart
´
ın 1987–88 Bay of Biscay 0.00005 3.061
(1989)
38
Table 1. Continued
Author Sampling Area Growth Length/Weight
period Model Aging material t
0
kL
inf
relationship
ab
Borges and Gordo 1988–90 Portuguese coast 0.00002 2.957
(1991)
Maxim (1995) 1972–92 NW Africa Von Bertalanffy –1.16 0.278 38.98 0.0139 2.961
Karlou-Riga and 1989–91 Gulf of Saronikos Von Bertalanffy Otoliths –0.943 0.366 30.27 0.0061 3.07
Sinis (1997) (Greece) –0.76
2
0.37
2
30.65
2
–0.83
8
0.37
8
30.67
8
ICES (1991a,b; 1985–00 ICES area Mean length-at-age Otoliths ∗∗∗
1996, 1997, 1999a,b; (Northeast Atlantic) and mean weight-at-
2001) age
1
Estimated from gutted specimens.
2
Back calculation by scale proportional hypothesis.
3
From otolith growth: calculation of mean ages at a given marginal increment and these mean ages were fitted to the mean fish lengths.
4
Fitted to mean lengths-at-age, from an ALK, in which it is considering the date of birth the 1st of January.
5
Fitted to mean lengths-at-age, from an ALK, in which it is considering the date of birth the 1st of July.
6
Specimens from 0 to 23 cm.
7
Specimens greater than 23 cm.
8
Back calculation by body proportional hypothesis.
∗
The ICES database has the historical series of mean weight at age and mean length at age in the horse mackerel catch (by area, season, and
year). With this information further estimates on growth characteristics could be obtained.
and Sinis, 1997). However, there are changes in the
parameters of the length-weight relationships within
the year. Lucio and Martín (1989) found signifi-
cant differences in these parameters between autumn-
winter and spring-summer periods in the Bay of
Biscay. In Portuguese waters, the differences are also
evident between spring and the rest of the year (Borges
and Gordo, 1991). These changes are probably related
to the different physiological stages that horse mack-
erel exhibit during the year. It appears that a negative
allometric growth or a lower value of the b para-
meter occurs during the spawning period (Lucio and
Martín, 1989; Borges and Gordo, 1991; Karlou-Riga,
1995). The allocation of energy to the reproductive
activities suggests that the fish becomes lighter for
its length. In contrast, after the spawning season the
fish can use more energy for growth. This interpret-
ation is corroborated by evidence that the highest fat
content in the flesh of T. trachurus is found when
the spawning period has already finished (Lockwood
and Johnson, 1977; Bandarra et al., 2001). Therefore,
if length-weight relationships or parameter b values
from different areas are compared then these should
be based on fish sampled during similar periods (e.g.,
equally over the whole year).
In the papers reviewed, references are made to the
relation between food and the parameter b, but rather
few between latitude and the parameter b. For future
work, fish migration should be taken into account. If
fish migrate in an age-size distribution, then migration
might be a factor that could affect the length-weight
relationship at latitude and probably also by depth.
In all reviewed references, an explicit description
of the regression model used in the length-weight rela-
tionships is lacking. Ricker (1973, 1984) recommends
the use of the geometric mean regression, as there
is variability in both the length and weight obser-
vations that is inherent to the material. The use of
an ordinary least-squares regression model, which is
by far the most used model for the estimation of
length-weight relationships in the literature, leads to
an under-estimation of b (Ricker, 1973). Fortunately,
there is not much difference between the two regres-
sion models when the range of lengths and weights
available is large (Ricker, 1979), as it occurs in the
majority of the reviewed papers.
Condition factor
The relationship between length and weight provides
an index to quantify the state of well being of fish
39
(Wootton, 1998). This index, the condition factor,
has been expressed in different ways. Properly the
condition factor is the parameter a in the equation (1):
a =
W
L
b
(3)
But usually the ponderal index K is used, which is
a modification of the same expression, in which the
value of parameter b is fixed to be 3:
K = 100 ·
W
L
3
(4)
with W in grams and L in centimeters.
Very few references deal explicitly with the
analysis of the condition factor in horse mackerel.
Lozano Cabo (1952) described the variability by
length and by sex, but no clear patterns could be
outlined. Carrillo (1978) compared the ponderal index
between entire and gutted specimens and found that,
at increasing length the weight of the viscera showed
a proportionally higher importance with respect to
the total weight of the specimens. A more complete
study was carried out by Lucio and Martín (1989) with
specimens from the southern part of the Bay of Biscay.
They described the monthly development of the condi-
tion factor for whole and for gutted fish, and noted that
the different maturity stages did not seem to influence
largely the variability in the observed ponderal factor
values. This would in principle be unexpected (i.e.,
Weatherley and Gill, 1987). The high ponderal factor
values for fish in “fully spent” stage and when the
atresia rates are high could be explained by physiolo-
gical processes (energy demands), avoiding the weight
depletion due to reproductiveeffort (Lucio and Martín,
1989; Karlou-Riga, 1995).
Ricker (1975), and Weatherley and Gill (1987)
explain the possibilities and limitations of using the
condition and ponderal factors for comparative studies
in fish populations. Safran (1992) and Jennings et al.
(2001) judged these factors to be less important than
the allometric parameter b, since these parameters are
closely correlated with b. On the other hand, Safran
(1992) also criticizes the use of K, since in fact, the
value of b equal to 3 is rarely obtained at stock or
sub-population level.
Growth functions
There are several ways of modeling the growth of fish
(Ricker, 1979). However, the Von Bertalanffy Growth
Function (VBGF) has been by far the most studied
and most used of all length-age models in fish biology
because (1) it is based on bioenergetics principles; (2)
it is useful in other fishery assessment models (e.g.,
Beverton and Holt, 1957); and (3) of its empirical
success in describing growth (Wootton, 1998; Quinn
II and Deriso, 1999). This function describes the
relationship between age and length as follows:
L
t
= L
∞
·[1 − exp(−k · (t − t
0
))] (5)
in which, using the equation to describe the mean
growth of a population of fish, L
t
is the mean length
at age t; L
∞
, k,andt
0
are parameters that determine
the shape of the growth curve. We can define L
∞
as
the asymptotic mean length; k as the rate at which the
curve approaches the asymptote; and t
0
is the age at
which mean length is zero (Francis, 1995).
Combining this equation with the power model for
the length-weight relationship (see equation 3), and
defining W
t
as the mean weight at age t,wegetthe
VBGF for body weight:
W
t
= W
∞
·[1 − exp(−k · (t − t
0
))]
b
(6)
which is like the length–growth model but adding the
allometric growth parameter b.
Almost all studies on horse mackerel growth deal
with fitting VBGF to length-age data, where the
age data are inferred either from calcified struc-
tures (otolith or scales) or from the size frequency
analysis (see Table 1). There is no application of
tag-recapture methods to study the growth in length.
The VBGF for weight-age data was also estimated by
Wengrzyn (1974), Trouvery (1977), Carrillo (1978),
Nazarov (1978), and Kerstan (1985). In the analysis of
the VBGF parameter estimates obtained by different
authors, one has to take the following factors into
consideration: sampling design, material used to infer
the age, material preparation, age-reading method and
whether it has been validated, and finally, the para-
meter estimation procedure. Therefore, it is very diffi-
cult to make consistent comparisons between different
authors. The majority of reviewed papers applied
unvalidated age reading methods. There is historically
good agreement on age readings up to age four, but not
for ages five and older, for which the age readings even
have differed up to a factor of two (see section 3.1).
This implies that the historical VBGF parameters are
expected to be unreliable, if they include the mean
length and weight data of 5 year and older fish.
Horse mackerel grows quickly, up to 3 years old,
when they are about 25 cm (Macer, 1977; Trouvery,
1977; Nazarov, 1978; Fariña Pérez, 1983; Alegría
40
Hernández, 1984b; Arruda, 1984; Kerstan, 1985;
Karlou-Riga and Sinis, 1997; present paper), whereas
later on, the increase in length continues at a slower
rate. Macer (1977) found a seasonal growth in the
young horse mackerel by analyzing the monthly
mean lengths-at-age during the year. Considering the
monthly mean weights-at-age, the juveniles presented
a steady increment during the year whereas the adults
showed seasonal fluctuations mainly due to repro-
ductive activities (Karlou-Riga and Sinis, 1997). Until
now, there has been no evidence of growth differ-
ences by sex (Overko, 1974; Fariña Pérez, 1983;
Karlou-Riga and Sinis, 1997).
There is no estimation of the VBGF parameters
leading to age interpretation on sectioned otoliths as
one translucent zone per year (i.e., maximum age is
around 35 years old in the Northeast Atlantic stocks).
The errors in reading the age of older fish will partic-
ularly affect the estimated L
∞
, and to a lesser extent
the values of k and t
0
.Thevalueofk obtained from
studies of the maximum age up to 15-years-old ranged
from 0.11 (Arruda, 1984) to 0.37 (Karlou-Riga and
Sinis, 1997), while most of the cases ranged between
0.18 and 0.22 (see Table 1). The lowest and the highest
values of k estimates might be explained by the diffi-
culties in obtaining both very small and very large
fish. This affects the estimates of t
0
and L
∞
,and
consequently of k (King, 1995). Using scales and
length frequency analysis to estimate growth para-
meters, the values obtained for k are in the range
of 0.13 (Wengrzyn, 1975) and 0.33 (Turner et al.,
1984). It is also strange to assume estimates of L
∞
to be greater than 50 cm (Wengrzyn, 1975; Lourdes
Marecos et al., 1978; Carrasco, 1980; Junquera et
al., 1988), when fish larger than 44 cm are rarely
reported from commercial and research vessel catches
(ICES, 2002a). On the other hand, negative values of
t
0
lower than –2 (Wengrzyn, 1975; Carrasco, 1980;
Arruda, 1984) may indicate problems in the sampling
process, usually in inadequate sampling of the very
young specimens.
Growth parameters obtained using back calcula-
tion (Francis, 1990) can produce contradictory results
in comparison to growth parameters estimated from
ototliths. Karlou-Riga and Sinis (1997) studied speci-
mens from the Aegean Sea and found high conformity
between the two methods, while Junquera et al. (1988)
found significant differences in the estimated growth
parameters using both.
In spite of the amount of references dealing with
horse mackerel growth, there is still a large lack of
knowledge. Since horse mackerel, most certainly, are
a long-lived fish (up to 35 years old), it should be
possible to estimate the VBGF revising all anterior
estimates. Knowing the difficulties in obtaining small
and large fish, the values of L
∞
and t
0
reported until
now often represent large extrapolations beyond the
range of the sampled data, and the estimates are unre-
liable. These growth estimates are only representative
for the length range in the samples.
Length frequency analysis (Petersen, 1892)
requires the presence of very clear cohorts with no
or very little overlapping length distributions for the
different ages. These circumstances only appear in
species with strict seasonal reproduction and a very
short life span (Hilborn and Walters, 1992). This is
not the case for horse mackerel. The use of scales
is not recommended for use in estimates of growth
parameters due to the high possibility that fish live for
many years with little or no scale growth (Beamish and
McFarlane, 1987).
There are a variety of methods for estimating the
VBGF (Quinn II and Deriso, 1999). In most papers
reviewed, the VBGF has been fitted to the observed
mean length at age, using the methods described by
Ricker (1975; i.e., Beverton’s method), where the
individual growth rate is not taken into account. In
recent papers, Junquera et al. (1988), and Karlou-Riga
and Sinis (1997) fitted the VBGF using non-linear
least squares.
Possibility of density dependent growth
Variability in mean weight-at-age and mean length-
at-age among the years could also be related with
characteristics of the different cohorts. Horse mackerel
stocks in the northeast Atlantic Ocean are charac-
terized by occasional extremely strong year-classes.
The latest strong one was the 1982 year-class, which
dominated the catches of the western stock for many
years (Figure 2). This year-class was also very strong
in the southern stock. It is likely that such an abundant
cohort will produce a density-dependent effect on
certain parameters of the population (e.g., growth,
maturity, and fecundity), as seen in other species
(e.g., Bannister, 1978; Rieman and Myers, 1992;
Rijnsdorp and Van Leeuwen, 1994). To investigate
the hypothesis of density dependence in growth, the
mean length-at-age of the year-class of 1982 in the
western stock was plotted together with those of
the year-classes from 1979 to 1988 (Figure 3). This
shows that the mean lengths-at-age of the cohorts of
1979 and 1980 were higher than those of the cohorts
41
Figure 2. Estimated numbers at age of northeast Atlantic Ocean horse mackerel (western stock) from 1982 until 2000. Data from ICES
(2002a,b).
from 1981 to 1988. The lowest mean lengths-at-age
observed were always from the 1982 cohort or from
cohorts close to it. This suggests that the exceptional
strength of the 1982 cohort influenced the growth of
fish from neighboring cohorts. Rose et al. (2001)
give an overview of density-dependence effects in
fish populations. Density-dependent growth refers to
a situation where the feeding rate of an individual is
reduced by the presence of other members of the same
population (Rose et al., 2001). In schooling species,
such as horse mackerel, this effect is probably more
pronounced between individuals of similar size. Given
the decrease in growth rate of the 1982 year-class,
prolonged competition with much younger cohorts
may have contributed to growth within age-classes.
Factors affecting growth
One of the most characteristic aspects of fish growth
is the variability observed among individuals in a
population and among populations in a species. This
variability is due to two main groups of factors:
exogenous, those related with the environment, and
endogenous, which are connected with the genotype
and physiological condition of fish (Wootton, 1998).
There are just a few references dealing specifically
with factors affecting horse mackerel growth. We
review some of them in this section.
Both the quality and quantity of food are relevant
to growth performance. In this contribution, we have
related the trends in the values observed of the allo-
metric parameter b, with the availability and compos-
ition of food (see section on length-weight relation-
ship). The same relationship was examined by Anadón
42
Figure 3. Mean length-at-age of northeast Atlantic Ocean horse mackerel (western stock) by year-class estimated from commercial samples.
The large circles indicate fish belonging to the strong 1982 year-class.
(1960), who tried to explain the differences observed
in the mean weights of horse mackerel between indi-
viduals taken from the coastal areas where the food
is abundant, and those from the open sea. In a very
detailed study on gastric evacuation in horse mack-
erel, Temming and Herrmann (2001a) stated that the
very high evacuation rates observed do not seem to
set any temperature dependent digestive constraint on
consumption or growth.
Macer (1977) studied seasonal growth in young
fish in relation to temperature, and concluded that
most of the annual growth takes place at a time
when temperatures are the highest. Kerstan (1985)
also concluded that the otolith growth of the north-
east Atlantic horse mackerel is fastest during the
summer. It seems that the horse mackerel in the north-
east Atlantic reduces their food intake during winter,
when the water is colder (see Temming and Herrmann,
2001a, and references therein). The strong temper-
ature dependent regulation of food intake through
appetite could at least partially explain these observa-
tions (Temming and Herrmann, 2001a).
Temperature greatly affects the embryonic devel-
opment and growth of fish larvae and juveniles (Pipe
and Walker, 1987; Cushing, 1995). They found that
the development of horse mackerel eggs through to
hatching occurred within the temperature range 10.5–
21.2
◦
C and the best survival rate was observed at
12.2–15.8
◦
C (Pipe and Walker, 1987).
Spatial effects also have to be taken into consid-
eration when analyzing fish growth. In the eastern
Mediterranean, the horse mackerel grow faster during
the first two years of life than in other areas of
the Mediterranean Sea, but from then onwards,
growth is reduced. Karlou-Riga and Sinis (1997)
explained this phenomenon as species adaptation to
the oligotrophic conditions of the eastern Mediter-
ranean. Although there are no references dealing
with T. trachurus, Kerstan (1995) found a depth
related trend in the growth pattern of horse mackerel
(T. capensis Castelnau, 1861) from the Agulhas Bank.
This result could be attributable to a change in habitat
selection with body length at the same age rather than
to a change in growth related to temperature (Freon
and Misund, 1999).
The influence of environmental factors on fish
growth is generally more studied than the intrinsic
control of growth, but the latter is equally important
43
(Wootton, 1998). There are no references on horse
mackerel dealing with regulatory control of growth
through the neuroendocrine system or about the
contribution of genetic factors in this phenotypic trait.
Herrmann and Enders (2000) estimated the effect
of body size on routine and standard metabolism,
concluding that the horse mackerel is not a perman-
ently swimming species, as it is commonly assumed.
Horse mackerel reproduction
Timing of reproduction
Gonadal development
Seasonal developmental changes of the gonads
(gametogenesis) in teleosts follow successive more or
less common stages (West, 1990). However, the rates
of these changes differ a lot among species. Gameto-
genesis concerns both the oogenesis (oocyte growth)
and the spermatogenesis (germ cell growth).
Oogenesis can be classified into five phases: pri-
mary oocyte growth, cortical alveolus stage, vitel-
logenesis, maturation, and ovulation (Wallace and
Selman, 1981; Bromage and Cumaranatunga, 1988),
while spermatogenesis is characterized by three
phases: spermatocytogenesis, meiosis, and spermio-
genesis (Selman and Wallace, 1986).
Oocyte growth is initiated or modified in a variety
of ways (reproductive strategy) leading to reproductive
success. The reproductive strategy is related with the
final product, e.g., the egg size and the dynamic
organization of the ovary itself. Three types of ovary
organization can occur (Wallace and Selman, 1981;
Burton, 1998):
1. Synchronous: a relatively rare type, where all
oocytes, once formed, grow and ovulate from
the ovary in union; further replenishment of one
stage by an earlier stage does not take place. Such
ovaries may be found in teleosts, which spawn but
once and then die (semelparous species);
2. Group-synchronous: where at least two oocyte
populations are developed, one with the most
advanced oocytes clearly separated and defining
a clutch (batch) and another more heterogen-
eous population consisting of smaller oocytes from
which the clutch is recruited; and
3. Asynchronous: where oocytes of all stages and
sizes are present without any dominant population.
During the spawning season, unyolked oocytes
continue to mature and are spawned in a number
(species specific) of subsequent clutches. Horse
mackerel belong to this group.
The examination of the gonad developmental seasonal
changes incorporates the adoption of a maturity stage
key (MSK), i.e., the seasonal changes that both ovaries
and testes undergo until spawning. Staging systems
incorporate physiological, biochemical, morpholo-
gical, and histological criteria (Tyler and Sumpter,
1996). A variety of MSKs adjusted to different species
have been used in the literature. However, it is more
common to apply more detailed stages for the oogen-
esis than for spermatogenesis. The reason underlying
this situation lies on the lower complexity of sper-
matozoa formation when compared with that of the
egg.
In the case of horse mackerel, assigning maturity
stages by macroscopical analysis has been the most
common approach used so far to describe gonadal
development. However, there has been no agreement
on the use of a common maturity scale. For Lucio
and Martín (1989), a simplified MSK with five stages
was used to describe the gonadal cycle, while others
used 11 stages to fulfill the same goal (Polonsky
and Tormosova, 1969). Between these two extremes,
Deniel (1989) and Abaunza et al. (1995) used an
empirical MSK with 6 stages, Komarov (1964) with
7 stages, Kerstan (1985) with 8 stages, and Macer
(1974) with 9 stages. Other authors adopted MSKs
previously described, such as Nazarov (1977), who
followed the MSK used by Polonsky and Tormosova
(1969); Arruda (1982, 1983) and Alegría-Hernández
(1994), who followed that of Macer (1974); or Karlou-
Riga and Economidis (1996, 1997) who adopted the
MSK of Macer (1974) with slight changes. The use
of so many maturity scales for the same species is
probably due to the fact that horse mackerel have been
described as asynchronous spawners by all the above-
mentioned authors. In any case, the final goal of using
a specific maturity scale is to distinguish immature
from developing, ripe, and spent females, whereas
these broad divisions are usually separated into more
stages used by the different authors.
For multiple spawners, such as horse mackerel,
macroscopical analysis of the gonads cannot provide
a correct and precise means to follow the development
of both ovaries and testes. Histological analysis has to
be included because it provides precise information on
oocyte developmental stages (West, 1990). Besides,
it is the only type of analysis that can distinguish
between immature gonads and regressing ones or
those partly spawned (Karlou-Riga and Economidis,
44
1996). Subsequently, macroscopic analysis of gonads
may lead to serious mistakes in the interpretation of
the processes that occur inside the gonads. This may
have an important role in the definition of criteria for
maturity and fecundity estimates (Gordo and Martins,
1986).
It is worthwhile to mention that very few studies
are supported by histological techniques that would
have allowed verification of changes that really occur
in both ovaries and testes. In fact, only the works by
Macer (1974), Arruda (1982, 1983), Eltink (1991b),
and Karlou-Riga and Economidis (1996) described
gonadal development of horse mackerel using histo-
logical criteria. Some of these studies show that
results from the two approaches are to some extent
similar (Karlou-Riga and Economidis, 1996), or show
good agreement (Pérez, IEO, Vigo, Spain, pers.
comm.). A table presenting horse mackerel develop-
mental changes in ovaries using either macroscopical
or microscopical criteria is given in Table 2.
However, even in the cases when immature indi-
viduals can be easily classified by the macroscop-
ical criteria, the microscopical examination of the
gonads is fundamental to detect and evaluate: (a) the
number of oocytes per batch which are ready to be
released (batch fecundity) in the present spawning
season; (b) the time elapsed since recent spawnings
by the presence of postovulatory follicles of known
age (spawning frequency); and (c) oocytes, which for
some reasons could not reach their full development
(atresia). It is noted that batch fecundity and spawning
frequency are necessary parameters for the estimation
of spawning biomass using the daily egg production
method (DEPM; Lasker, 1985; Gunderson, 1993).
Atresia
The resorption of yolked oocytes and follicles, a pro-
cedure known as atresia, characterizes the spawning
period mainly for asynchronous spawners. Bretsch-
neider and Duyvene de Wit (1947), who studied
the endocrinologic cycle of Rhodeus amarus (Bloch,
1872), Lebistes reticulatis (Peters, 1860), and Zoarces
viviparus (Linnaeus, 1758), were the first to describe
sequential atresia stages in the ovary, while they
noticed that the atretic formations coincide with
hormone production responsible for spawning capab-
ility. Many years later Hunter and Macewicz (1980,
1985a,b) estimated the duration of each atresia stage
for Engraulis mordax (Girard, 1854) and concluded
that the percentage of atretic oocytes increases at the
end of spawning period.
The study of atresia should be carried out through-
out the year because:
1. It provides histological criteria to forecast end of
spawning;
2. It is the only way to distinguish immature from
mature individuals, which is essential for the
estimation of length/age at first maturity; and
3. It provides those histological criteria needed for
the age assignment of postovulatory follicles (pof)
and thus the estimation of spawning frequency.
Very few studies have dealt with the atresia process
in horse mackerel. Macer (1972, 1974) observed a
mean percentage of 6.31% in developing late ovaries
and 3.73% in ripe ovaries for specimens collected
in the western English Channel and southern half
of the North Sea. Eltink (1991b) found that during
the spawning period 40% of those ovaries containing
yolked oocytes were in alpha (see below) stage
atresia. Karlou-Riga and Economidis (1996), who
studied the reproductive cycle of horse mackerel in
the Saronikos Gulf (Greece), validated the atresia
stages (Figures 4, 5), which seemed similar to those
described for T. symmetricus (Ayres, 1855) off Cali-
fornia (Macewicz and Hunter, 1993). Alpha (α)stage
of atresia ends when the entire oocyte is resorbed
(all cytoplasm and yolk are gone). At this stage only,
the oocyte is defined as an atretic oocyte because the
type of oocyte (i.e., yolked) undergoing atresia is still
discernible. The subsequent atretic stages (beta [β]-
delta [δ]) are steps in the resorption of the remaining
structure (follicle), which from the beta stage onwards
is defined as atretic follicle. Gamma (γ ) atresia stage
was not observed in regressing ovaries of horse mack-
erel (Karlou-Riga and Economidis, 1996). This may
have been due to the short duration of this stage or to
the fact that the follicle passes directly from the β to
the δ stage without passing through the intervening γ
stage (Hunter and Macewicz, 1985a,b).
Karlou-Riga and Economidis (1996) estimated
potential spawning of horse mackerel ovaries with
atretic oocytes. It is noted that the proportion of atretic
oocytes in the ovaries in the literature is often called
“intensity of atresia” (Priede and Watson, 1993). The
above-mentioned authors first classified the ovaries
into the following atretic states, while sequentially
they identified the presence/absence of spawning
histological characteristics (migratory-nucleus stage
oocytes, hydrated oocytes, and pof) in those ovaries
with atretic oocytes (atretic state 1 and 2):
Atretic state 0: No α atresia of yolked oocytes;
45
Table 2. Description of the reproductive stages of horse mackerel females by macroscopical and microscopical criteria
Female maturity Ovary
stage External appearance Histological appearance
Virgin Rounded translucent ovaries; less
than a quarter of length of body
cavity; no oocytes are visible
Well-spaced ovigerous folds orientated towards the center of the ovary;
oogonia and primary oocytes at both the chromatin nucleolus and perinuc-
leolus stage; oocyte size 10–60 µ
Developing virgin
or resting
Rounded orange-pink ovaries; a
quarter to a third of body cavity;
no oocytes are visible
Few spaces between ovigerous folds; few oogonia, the majority of primary
oocytes at the perinucleolus stage; oocyte size 20–150 µ
Early developing Orange ovaries; a third to a half of
body cavity; oocytes are visible
Oocytes with cytoplasmic vacuoles (lipid droplets); yolk granules first
appear in the cytoplasmic periphery while subsequently spread intern-
ally; elongated sprindlelike cells constitute the follicle layer; oocyte size
150–400 µ
Later developing Orange-yellow ovaries; two thirds
to whole of body cavity; oocytes
are visible
Yolk granules becoming larger (yolk spherules) proliferate; oil droplets
spread throughout the cytoplasm, while at the end of the stage they coalesce
and accumulate around the nucleus; zona radiata is present; granulose cells
become cuboidal; oocyte size 400–600 µ
Ripe/running Yellow ovaries fill the whole of
body cavity; hyaline oocytes visi-
ble beneath the thin oocyte mem-
brane may run from vent on slight
pressure
Yolk spherules coalesce to globules or plates; large oil droplets follow the
nucleus migration to the animal pole where the nucleus disperses its content
into the cytoplasm; oocyte size 600–800 µ. By the nucleus dispersion rapid
uptake of fluid (hydration) is taken place; zona radiata loosing its striation
becomes very thin; oocyte size 700–1200 µ
Partly spent Flaccid yellow with patches of red
ovaries; smaller than those of the
previous stage
Present post ovulatory follicles (pof); oocytes in any developing stage
including that of ripe; possible oocytes in alpha or subsequent stage atresia
Spent Small flaccid red ovaries covering
a third of body cavity
Possible pof; yolk oocytes where 50% or more are in alpha stage atresia or
no yolk oocytes but atretic follicles (beta or later stage atresia) and primary
oocytes
Atretic state 1: <50% (1 oocyte to 49%) of yolked
oocytes were in α atresia;
Atretic state 2: ≥50% of yolked oocytes were in α
atresia; and
Atretic state 3: No remaining yolked oocytes, but
β atresia or later atresia stages were present.
Karlou-Riga and Economidis (1996) finally found
that 82.9% of the ovaries classified in atretic state
1 showed evidence of spawning, whereas only 7.7%
of the ovaries classified in atretic state 2 were in
spawning (past) condition. On the contrary, 92.3%
of the ovaries classified in atretic state 2 had no
spawning characteristics. These results showed that
by that time the ovaries reach atretic state 2, they are
no longer capable of spawning, thus defining atretic
state 2 as a very good measure of the ovary resorption.
Actually, the examination of ovaries (females) clas-
sified in the four atretic states throughout the repro-
ductive period showed that atretic state 1 was the most
common atretic condition during spawning season,
while ovaries into atretic state 2 increased towards the
end of the season and were completely replaced by
those in atretic state 3 at the end of spawning. These
conclusions coincide with those found in the liter-
ature for some other species, e.g., E. mordax (Hunter
and Macewicz, 1985b), where high prevalence (high
proportion of females with atretic oocytes) and high
intensity of atresia (great number of atretic oocytes in
the ovary) forecast the end of spawning season.
However, high levels of atretic states 2 and 3 when
occurring in the middle of the spawning season do
not necessarily indicate the end of the season, but
merely the end of a spawning cycle within the season
and probably the limitation of this season. This actu-
ally occurred in a case of horse mackerel in Greek
waters (Karlou-Riga and Economidis, 1996), where
during a subsequent reproductive period high percent-
ages of ovaries classified in atretic states 2 and 3
occurred in the middle of the season, while some-
time later there followed a period classified in atretic
state 0. The transition of females from the spawning
to the post spawning condition and vice versa,avery
46
Figure 4. New postovulatory follicle (n.pof) of horse mackerel. (a) N.pof as a very convoluted formation; and (b) A part of the Figure 4a
enlarged, where the linearly arranged granulosa cells (gc), the thecal cell layer (tc), and the lumen (lu) can be seen. The photograph shows the
state 12 h after the spawning. Magnification: (a) = ×137, (b) = ×550.
rapid procedure, has been observed for other species,
such as E. mordax (Hunter and Macewicz, 1985b) and
Katsuwonus pelamis (Linnaeus, 1758) (Hunter et al.,
1986).
Atresia affects spawning frequency (Karlou-Riga
and Economidis, 1997) and fecundity (Macer, 1974).
According to Karlou-Riga and Economidis (1997),
spawning frequency is quite low towards the end
of the spawning season. Therefore, they concluded
that atresia is inversely correlated with the spawning
frequency, although we have to take into account that
in the beginning of the spawning season both atresia
and spawning frequency will probably be low. On the
other hand, atresia might be considered an important
mechanism for the adjustment of fecundity through
the resorption of oocytes in unfavorable seasons and/or
lesser resorption in more favorable seasons (Macer,
1974). It finally seems that horse mackerel should be
considered very flexible in adjusting their spawning
time to different conditions and that their spawning
cycle might be more variable than previously assumed.
Such a flexibility may be very important for the
survival of the species. No studies exist on the rela-
tionship of atresia incidence in horse mackerel and
latitude.
Spawning frequency
To calculate the number of spawned batches in
multiple spawners such as horse mackerel, informa-
tion about duration of spawning period, spawning
frequency, and the mean interval (days) between
batches is needed. The multiplication of the number
of batches and of average batch fecundity (the number
of migratory-nucleus and/or hydrated oocytes in the
47
Figure 5. Old postovulatory follicles (o.pof) of horse mackerel in three various shapes (a, b, and c). Smaller and less convoluted formations than
n.pof. Granulosa cells (gc) not linearly arranged. Thecal cell layer (tc) very thick, blood capillaries (b) visible and lumen (lu) still identifiable.
The photograph shows the state 36h after the spawning. Magnification: (a), (b), and (c) = ×137.
ovary) results in the potential annual fecundity of
the species (Macewicz and Hunter, 1993; Karlou-
Riga and Economidis, 1997). On the other hand,
spawning frequency is the inverse of spawning frac-
tion (the average percentage of females spawning
daily), which is needed for the calculation of spawning
stock biomass (Hunter and Macewicz, 1985a).
The estimation of spawning frequency is based
on histological characteristics of spawning, such as
postovulatory follicles, migratory nucleus stage, or
hydrated oocytes. In order to estimate the time elapsed
since recent spawning, one must be able to age the
postovulatory follicles using the extent of their deteri-
oration (Hunter and Macewicz, 1985b). On the other
48
hand, it is assumed that the migratory nucleus stage
for some species of the genus Trachurus lasts for
about 24 h (Eltink, 1991b; Macewicz and Hunter,
1993). Spawning fraction estimated for the so-called
western horse mackerel stock (Eltink, 1991b) at the
peak of spawning was based on migratory nucleus
stage oocytes and it was found equal to 8.3%. The
same parameter estimated for Pacific horse mack-
erel (T. symmetricus) was based on the incidence of
postovulatory follicles and of migratory-nucleus stage
or hydrated oocytes, and was found to be 20.2%
during a 18 day sampling period considered the peak
of spawning (Macewicz and Hunter, 1993). In the
Greek Seas (Karlou-Riga and Economidis, 1997),
spawning fraction of horse mackerel, based on the
same premises with those used for T. symmetricus,was
found equal to 17.1% and 20.9% for two successive
reproductive periods respectively.
In the above-mentioned work, the criteria for the
extent of postovulatory degeneration were based on
those developed for closely related species, such as
T. symmetricus, after some modifications for daylight
periods of sampling. Thus, the histological character-
istics of spawning used for the estimation of horse
mackerel spawning frequency in Greek seas, taking
into account that Trachurus (sp.) spawn at night
(Macewicz and Hunter, 1993), are summarized as
follows:
Imminent spawning within 12 h –presenceoflate
migratory-nucleus stage oocytes and/or hydrated
oocytes;
Spawned 12 h before – presence of new postovu-
latory follicles (Figure 4); and
Spawned 36 h before – presence of old postovu-
latory follicles (Figure 5).
Spawning frequency does not remain constant
throughout the reproductive period, with very high
values at the peak of spawning, it decreases towards
the end of the season (Karlou-Riga and Economidis,
1997). On the other hand, spawning frequency may
change from year to year and, since it is inversely
correlated with atresia, as already mentioned, it might
be another adaptative response to unfavorable condi-
tions occurring within the spawning season.
Gonadosomatic index
The gonadosomatic index (GSI), expressed as
W gon · 100 · W t
−1
(7)
or
W gon · 100 · (W t − W gon)
−1
(8)
in which Wgon is the weight of the gonad and Wt
is the weight of the fish specimen, has been largely
used as an indicator of changes during the gonadal
development (Table 3). In fact, the possibility of quan-
tifying the changes that occur in the gonads (due to
the increase of cells and/or their size) has been used to
overcomethe subjectivity that is inherent to the macro-
scopical assignment of maturity stages. Therefore, it is
expected that, for each specimen, the highest value of
the GSI corresponds to the peak of spawning.
However, the time period for the detection of the
changes that occur in the gonads is not simultan-
eously shown macroscopically, histologically, or by
GSI. According to Burton et al. (1997), a repro-
ductive female of Gadus morhua (Linnaeus, 1758)
expected to spawn in May–June should show histo-
logical changes by November, but these changes were
not detected macroscopically until February. In terms
of GSI, although GSI might show increasing values
between November and February, the greater increase
in the GSI values related to a pre-spawning condition
of the ovaries may only be detected some weeks later
after the onset of exogenous vitellogenesis.
In synchronous spawning species, the GSI of a
mature female usually ranges between 18 and 25,
while in mature asynchronous spawners, the GSI gen-
erally attain lower values (Tyler and Sumpter, 1996).
Generally speaking, studies involving the GSI of
females are more common than those on males, since
estimations related to atresia, spawning frequency,
or batch fecundity concern ovaries. On the other
hand, sex ratio in the case of horse mackerel during
spawning season attains an average value of about 0.5
(Eltink, 1991b; Karlou-Riga and Economidis, 1996)
and remains constant throughout the season (Karlou-
Riga and Economidis, 1996), leading to the conclusion
that no sex-based migration occurs in and out of the
area studied. Abaunza et al. (1995), in the Cantabrian
Sea (north of Spain), also found no sex-ratio differ-
ences in their study as a whole area, but an analysis on
a smaller geographical scale showed significant differ-
ences. These differences were explained by the more
limited migratory behavior of females to find appro-
priate spawning conditions. The emission of both eggs
and sperm is assumed synchronous.
Few studies exist dealing with the simultaneous
examination of the reproductive cycle of horse mack-
erel using both macroscopical and histological criteria.
According to Macer (1974), who studied the repro-
ductive biology of this species in the North Sea and
the English Channel, peak GSI occurred in June
49
Table 3. Summary of the available information on the reproductive parameters of horse mackerel in the NE Atlantic and the Mediterranean
Author Sampling years Area Gonadosomatic Length at first Age at first
Index maturity (cm) maturity
Polonsky (1969) 1963–66 North Sea and English Channel 20–24 (FL)
Macer (1974) 1968–69 North Sea and English Channel 13
∗
20–24
Sahrhage (1970) 1955–68 North Sea 18–19 1 year
Kerstan (1985) 1982–84 Irish and Celtic Seas 22.3 males
25.4 females
Letaconnoux (1951) 1943 Bay of Biscay 19–23
Lucio and Mart
´
ın (1989) 1987–88 Bay of Biscay 553
∗∗
20.1 males
20.6 females
Lozano Cabo (1952) North/NW coast of Spain 21.1
∗∗∗∗
Anad
´
on (1960) 1954–55 North/NW coast of Spain 25
Abaunza et al. (1995) 1984, 1986, 1987 North/NW coast of Spain 20.9 males
21.9 females
Barraca (1964) Lisbon 19 (FL)
Arruda (1982, 1984) 1978–81 Western coast of Portugal 8–9
∗
21–24 (SL) 2 years
Arruda (1982, 1984) 1978–81 Southern coast of Portugal 8
∗
16–19 (SL) 1 year
Borges and Gordo (1991) 1987–90 Portuguese Coast 6
∗∗∗
22.5
Gail (1954) 1951–52 North Africa 15
Sedletskaya (1971) North Africa 16–23
Andreu and Rodriguez-Roda Catalonia 15.5
(1951) (NW Mediterranean Sea)
Planas and Vives (1953) Catalonia 16
(NW Mediterranean Sea)
Alegr
´
ıa-Hern
´
andez (1994) 1986–88 Adriatic Sea 10.7
∗∗∗
Karlou-Riga and Economidis 1989–91 Aegean Sea 3.5
∗
22 2–3 years
(1996)
∗
GSI = W
gon
/(W
t
–W
gon
) ∗ 10
2
;
∗∗
GSI = W
gon
/W
t
∗ 10
4
;
∗∗∗
GSI = W
gon
/W
t
∗ 10
2
;
∗∗∗∗
smaller individual with ripe gonads; FL =
fork length; SL = standard length.
and coincided with the highest percentage of females
in maturity stage 5 (ripe ovaries), while at this
stage histologically identified hydrated oocytes were
predominant. On the other hand, during the period
from May to August where stage 5 was present, eggs
were found in the plankton as well. After June a
decrease in GSI was noticed simultaneous with the
progressive increase of females identified as partly
spent (stage 7).
In Karlou-Riga and Economidis’ (1996, 1997)
observations of horse mackerel in Greek waters, histo-
logically detected spawning starting in December and
ending in July, the peak of GSI occurred in February,
and the peak of spawning during the time period
between the end of March and beginning of April,
whereas spawning frequency remained high until May.
After February, when 52% of the mature females
were spawning, the decrease of GSI coincided with
the progressive increase of those females classified in
atretic states 2 and 3 (post spawning females), while
at the end of the season, both criteria of low GSI and
females in atretic state 3 were used.
It seems that for multiple spawners, GSI as an
index of ovarian changes during the spawning season
is of lesser importance than histological data. High
values of GSI do not show the peak of spawning but
rather a later developing stage, while GSI decrease
is due to subsequent batch releases. When only this
index is used, it is impossible to distinguish partly
spent (active females with postovulatory follicles)
from spent (inactive females with regressing ovaries);
but it is possible with high values of this index to
50
determine the period when hydrated oocytes can be
identified, which is very useful for batch fecundity
estimates. Estimates of GSI calculation over the entire
year can be used to approximate the reproductive
period (Karlou-Riga and Economidis, 1996). It should
also be noted that according to these authors absolute
values of GSI have to be taken cautiously because the
highest values changed from year-to-year, reflecting a
possible environmental effect between two successive
reproductive periods of horse mackerel.
In general, the gonadosomatic index does not
compensate completely (“normalize”) for the effects
of fish size (Davies, 1956; deVlaming et al., 1982;
West, 1990). For the same reproductive state, small
fish usually have a lower GSI than do larger fish
and this effect increases with maturation of the ovary
(Hunter and Goldberg, 1980; deVlaming et al., 1982).
In other words, ovary weight increases faster with
fish length than does somatic weight. In order to
show whether GSI could be a valid index of ovary
activity, the following assumptions should be con-
sidered according to deVlaming et al. (1982):
a. linear relationship between gonad weight and body
weight;
b. the regression of gonad weight on body weight has
a zero Y-intercept;
c. the coefficient of variation of gonad weight is
constant over the entire weight range; and
d. the linear relationship does not change with stage
of gonad development.
It seems that in the case of horse mackerel none of the
above assumptions are valid (Karlou-Riga, 1995). The
last observation is in agreement with that of Hunter
and Goldberg (1980) and deVlaming et al. (1982) for
other species as well. In addition, GSI is characterized
by a higher variance than that of gonad or body weight,
it does not follow a normal distribution, and it does not
have homogeneous variance along the gonad or body
weight range. These features have to be taken into
account, when changes in GSI are examined statistic-
ally and nonparametric approaches have to be applied.
Thus, since the relationship between gonad weight and
body weight does not appear to be linear, the use of
ovary weight/fish weight for the expression of GSI
should not be considered essential. Other techniques
like that suggested by Erickson et al. (1985) would
be worthwhile investigations. In this last reference,
instead of the classical GSI, the relative gonadal index
is used:
GSI
i
= a
i
W/S
b
(9)
where W is gonad weight, S is body size (either length
or body weight minus the gonad weight), and a
i
,bare
parameters for gonad of maturity stage i.
Seasonal timing of reproduction
According to the literature studied so far on repro-
ductive behavior of horse mackerel, this species has a
rather extensive spawning season, sometimes as long
as eight months.
In the North Sea and the English Channel, most
authors have defined the spawning season as a
4-month period. Polonsky and Tormosova (1969)
recorded spawning individuals from May to August.
Sahrhage (1970) assigned the spawning season from
May to September, while Macer (1974) placed it
to be between May and August. Eltink (1991b,
1992) estimated egg production in the North Sea in
four successive years (1988–1991) and determined
a spawning season of just over three months. The
spawning season was from May to July, but the timing
was dependent on temperature (i.e., high temper-
ature = earlier spawning).
In the Bay of Biscay, horse mackerel seem to have
a longer spawning period when compared to that in the
North Sea. Letaconnoux (1951) referred to the peak of
spawning in May–June, while Arbault and Lacroix-
Boutin (1969) mentioned that horse mackerel prefer-
entially spawned during spring. Nazarov and Dobrusin
(1977) reported that the reproduction period of horse
mackerel lasts from December to June. Finally, Lucio
and Martín (1989) mentioned that advanced mature
specimens appeared in higher frequencies between
March and August, with a peak in May–June.
The spawning season along the northern coast
of Spain seems to have extended from 1950 to
the present. Letaconnoux (1951) reported that the
main spawning season occurred between February
and April, and Anadón (1960), between February
and May. Solá et al. (1990) found eggs at sea from
February to December, with a main spawning period
from April to June. Off the west coast of Portugal,
horse mackerel has a prolonged spawning season very
similar to the one found north of Spain. Barraca
(1964), working with specimens caught off Lisbon,
defined the spawning time between January and June.
Borges et al. (1977) studied samples from three areas
off the Portuguese coast. For the northern and central
regions, they found that the spawning period was
between February and August, with a peak from
March to June, while for the southern part of the
coast, they reported two main periods: one from
51
January to June and a second from July to September.
More recently, Arruda (1982, 1984) also studied
the three different areas mentioned by Borges et al.
(1977). For the northern and central areas, he recorded
intensive spawning between February and April, while
spawning was observed until May. In the southern
coast, the spawning period was considerably longer
and extended from September to May. It was also
found that individuals spawning in September were
young, when at that time the older individuals were in
a “resting season”. Borges and Gordo (1991), working
with samples from the entire coast of Portugal, deter-
mined the spawning season to last from December to
June.
On the northwest coast of Africa, the spawning
season is also prolonged. Gail (1954), using specimens
from Morocco determined the spawning time to be
March–April. Overko (1964, 1969, 1974), studying
horse mackerel caught in the central region of the
east Atlantic Ocean between 14
◦
and 26
◦
N, found
out that spawning took place between November and
May, with a peak in November/December. Sedlet-
skaya (1970, 1971) sampled a vast area in the east
Atlantic between 13
◦
and 34
◦
N. This author observed
that individuals caught at lower latitudes spawned
mainly between October and May, while specimens
caught at higher latitudes spawn from January–May.
The majority of papers mentioned above deter-
mined the spawning season based on macroscopical
analysis of the gonads. Only Arruda (1983, 1984)
and Eltink (1991b, 1992) also used a histological
procedure to corroborate the data.
Little information exists in the Mediterranean Sea
on spawning periods. In the western part (Spanish
coast), the main spawning period for horse mackerel is
in the winter months. In fact, Andreu and Rodriguez-
Roda (1951), working with samples from Catalonia,
reported that spawning occurred during winter, while
Planas and Vives (1953) reported December-February.
In the Adriatic, Alegría-Hernández (1994) recorded
that the main spawning period took place between
March and June. More recently, Karlou-Riga and
Economidis (1996) studied horse mackerel from the
Saronikos Gulf, Greece, during a two year period
using both macro- and microscopical analysis of the
gonads. During the first year, they found that spawning
occurred between December and May, with the peak
at the end of March/beginning of April. During the
second year, spawning took place between January
and May.
As a conclusion, one may consider that in the
northeast Atlantic Ocean at higher latitudes, the
start of spawning is later while at southern latitudes
spawning covers a more extensive period. Throughout
the Mediterranean, it seems that spawning starts in
December but lasts longer in the eastern part.
Length and age at first maturity
The application of age-structured models to assess a
stock implies the knowledge of the number of indi-
viduals in the population that are immature or mature.
This information is obtained by establishing various
levels of maturity. The percentages of mature indi-
viduals per length-class (or age-group) are estimated
during the spawning season. A logistic curve is then
fitted to the data and the length (or age) where 50% of
the individuals are mature gives that of first maturity
(King, 1995).
Three assumptions underlie this procedure:
1. The samples used for the analysis have to be
representative and cover the whole length range of
the population. The number of fish of each age
group in the samples should correctly represent
the numbers of immature and mature fish in
the population. In probably most horse mackerel
stocks, immature fish are distributed closer inshore
along the juvenile areas. The relative proportion
of an age group differs between the juvenile and
the adult areas and, thus, the relative abundance
of this age group in both areas is unknown. This
kind of problem was addressed in maturity ogive
estimations for western stock (ICES, 1999b);
2. The entire population of mature individuals (indi-
viduals capable of spawning and those with char-
acteristics of post-spawning during the current
season) has to be included in the analysis. This
means that, if the separation between mature and
immature individuals is done macroscopically, a
number of individuals with early developing or
regressing gonads might be misclassified as imma-
ture, resulting in lower percentages of mature
individuals and, thus, in a biased (overestimated)
length at first maturity. On the other hand, some
fish classified as mature macroscopically might not
produce any eggs, because all vitellogenic oocytes
might be resorbed by atresia. If the objective is
estimation of the spawning stock biomass, then
the maturity ogive applied should not include these
fish, because the total egg production is converted
into a spawning stock biomass using fecundity
estimates (ICES, 1999b); and
52
3. The time period for which the maturity is studied
has to be relatively long. In the case when only
the spawning period is covered by the study,
then a number of post-spawning individuals with
regressing gonads might be considered immature,
leading to underestimated length at first maturity.
The whole reproductive period would be more
appropriate for these analyses. This means the time
period from the development of the gonads up to
the end of spawning.
In the case of horse mackerel, maturity ogives
presented in the literature are preferentially related to
length, than to age (Table 3). The reason is obvious
since it is much easier to measure length than to
estimate age. On the other hand, when age is used,
different criteria for interpretation among authors
increase the uncertainty of establishing maturity by
age-groups.
Fecundity
Definitions
Estimated annual fecundity is a very important para-
meter in the dynamics of a population. Furthermore, it
can be combined with estimates of the abundance of
eggs in the sea to estimate the spawning biomass of a
stock. Annual fecundity is, as are the majority of para-
meters in population dynamics, subjected to several
assumptions. However, the evaluation of the assump-
tions underlying annual fecundity estimates requires
first defining fecundity terms as follows (Hunter et al.,
1992):
Annual fecundity (or realized fecundity): total
number of eggs spawned by a female per year
(does not include eggs resorbed because of atresia
and includes eggs produced by de novo vitellogen-
esis);
Total fecundity: standing stock of advanced
yolked oocytes;
Potential annual fecundity: total advanced
yolked oocytes matured per year, uncorrected
for atretic losses and including eggs produced
by de novo vitellogenesis. In species with deter-
minate fecundity, the potential annual fecundity is
considered equi