FISHERIES OCEANOGRAPHY Fish. Oceanogr.

16:1, 16–30, 2007

Modelling potential spawning habitat of sardine (Sardina

pilchardus) and anchovy (Engraulis encrasicolus) in the Bay
of Biscay

BENJAMIN PLANQUE,1,* EDWIGE BELLIER1 spawning are well reproduced by the model, indicating
AND PASCAL LAZURE2 that hydrographic changes may explain a large fraction
1
IFREMER, Département Ecologie et Modèles pour of spawning spatial dynamics. Such models may prove
l’Halieutique, rue de l’ı̂le d’Yeu, BP21105, 44311 Nantes Cedex useful in the context of forecasting potential impacts
3, France of future environmental changes on sardine and an-
2
IFREMER, Laboratoire de Physique Hydrodynamique et chovy reproductive strategy in the north-east Atlantic.
Sédimentaire, BP 70, 29280 Plouzané, France
Key words: anchovy, Bay of Biscay, environmental
effects, generalized additive models, potential
ABSTRACT spawning habitat, sardine

Large amplitude variations in recruitment of small

pelagic fish result from interactions between a fluctu- INTRODUCTION
ating environment and population dynamics processes
such as spawning. The spatial extent and location of Defining the spawning habitats
spawning, which is critical to the fate of eggs and Although small pelagic fish spawning habitats have
larvae, can vary strongly from year to year, as a result been the central focus of a large number of studies, a
of changing population structure and environmental clear definition of spawning habitats is often lacking,
conditions. Spawning habitat can be divided into and different authors have used the same terminology
‘potential spawning habitat’, defined as habitat where for concepts that can radically differ in terms of eco-
the hydrographic conditions are suitable for spawning, logical understanding. The most common and often
‘realized spawning habitat’, defined as habitat where implicit definition of spawning habitat is ‘geographical
spawning actually occurs, and ‘successful spawning location where eggs are found’ (see e.g. Lynn, 2003)
habitat’, defined as habitat from where successful which can be directly derived from egg surveys. An-
recruitment has resulted. Using biological data col- other definition which is also often not explicitly
lected during the period 2000–2004, as well as stated is ‘set of environmental conditions that are
hydrographic data, we investigate the role of envi- suitable for spawning’ (see e.g. van der Lingen et al.,
ronmental parameters in controlling the potential 2001). The use of the term ‘spawning habitat’ for these
spawning habitat of anchovy and sardine in the Bay of two definitions often depends on the state of
Biscay. Anchovy potential spawning habitat appears advancement of the observation (field sampling) and
to be primarily related to bottom temperature followed understanding of the relationships between spawning
by surface temperature and mixed-layer depth, whilst adults and their environment. The terminology may
surface and bottom salinity appear to play a lesser role. also be used for the two definitions in the same pub-
The possible influence of hydrographic factors on the lication. The related terminology ‘suitable spawning
spawning habitat of sardine seems less clear than for habitat’ which often refers to the fate of eggs could be
anchovy. Modelled relationships between anchovy stated ‘region where the conditions are suitable for
and sardine spawning are used to predict potential eggs and larvae survival’ (see e.g. Bakun, 1996;
spawning habitat from hydrodynamical simulations. Agostini and Bakun, 2002) and is related to the
The results show that the seasonal patterns in habitat suitability of Fretwell (1972): ‘the average
potential contribution from that habitat to the gene
pool of succeeding generations of the species’. The set
*Correspondence. e-mail: benjamin.planque@ifremer.fr of hypotheses, the physical and biological processes
Received 18 October 2004 involved, as well as the means of investigations
Revised version accepted 29 August 2005 necessary to study spawning habitat will vary greatly

16 doi:10.1111/j.1365-2419.2006.00411.x  2006 The Authors.

 Modelling potential spawning habitat of sardine and anchovy 17

depending on the underlying definition of what is the that data collected over a wide range of environmental
‘spawning habitat’. conditions are collated. These data have to include
To clarify the nature of the spawning habitat information on areas and seasons when spawning does
studied, we have separated spawning habitat into three not/never occur. Experimental results on the biologi-
distinct components. Each habitat component has cal and physiological characteristics of a given species
specific characteristics which may depend upon the can also be used to define its potential spawning
abiotic and biotic environment, population structure habitat. Potential spawning habitat can still be defined
and observation techniques. The methods available to for areas where a population has collapsed and no
study each habitat component may also be specific. A spawning actually occurs. Potential spawning habitat
schematic representation of the three habitat compo- has some similarities with the ‘basin’ model of Mac-
nents is given in Fig. 1. This distinction between Call (1990). However, the potential spawning habitat
habitat components was proposed during the Global depends on spawning, but not on the contribution to
Ocean Ecosystem Dynamics – Small Pelagics and population growth rate, i.e. subsequent recruitment, as
Climate Change (GLOBEC-SPACC) workshop on is the case in MacCall’s basin model. Recruitment-
spawning habitat and assessment of small pelagic fish related issues will be considered in the ‘successful
(Castro et al., 2005). spawning habitat’ below.
Potential spawning habitat Realized spawning habitat
‘Habitat where the environmental conditions are ‘Habitat where spawning actually occurs’. The realized
suitable for spawning’. The potential spawning habitat spawning habitat is defined by the region where fish
may be seen as the largest envelope of spawning actually spawn in a given year at a given time. It is
habitat. Based on environmental characteristics, it bounded by the potential spawning habitat. Factors
defines the set of conditions that a given species may that affect the location and extent of realized spawn-
find suitable for spawning. Spatial and seasonal extent ing habitat are primarily related to adult population
of potential spawning habitat will be primarily affected size and structure as well as density-dependent pro-
by variability in climate and the environment. It can cesses. The extent to which fish will use the potential
not be observed directly from individual regional field spawning habitat will depend upon the number of
cruises which are more suited to study ‘realized mature fish, their age, migration, mating behaviour
spawning habitat’ (see below). To define potential and other population traits, as well as possible inter-
spawning habitat it is necessary to observe spawning in action with other populations in the same areas
all possible suitable environmental conditions. In (predators, competitors and preys). Density-dependent
practice, defining potential spawning habitat requires habitat selection (DDHS; MacCall, 1990) will be
primarily related to realized spawning habitat. The
realized spawning habitat is the one generally observed
Figure 1. A schematic representation of the three spawning
during egg surveys. It is expected that realized
habitat components: potential, realized and successful.
Potential habitat is primarily influenced by prespawning
spawning habitat will display larger year-to-year fluc-
environmental conditions which affect adults, realized hab- tuations than will potential spawning habitat. Note
itat is primarily dependent on population size and demo- that comparing directly observed realized habitat with
graphic structure and successful habitat varies primarily as a environmentally based potential habitat may in most
function of post-spawning environmental conditions which cases not make sense, as there is no reason a priori that
will affect egg and larval survival. a given population will fully occupy the available
potential habitat in a given year.
Not suitable for spawning
Successful spawning habitat
‘Habitat where fish have spawned and from where
geographical coordinates

successful recruitment has resulted’. The successful

Spawning, Suitable,
but no contribution
to recruitment
but no spawning spawning habitat is defined by the fate of eggs spawned
into it. It is bounded by the realized spawning habitat
Spawning,
but no contribution
to recruitment
(and consequently by the potential spawning habitat).
spawning habitat type
Factors that affect realized spawning habitat are factors
potential
Spawning,
that will affect eggs, larvae and juveniles after
realised produces most recruits spawning has taken place. The successful spawning
successful
habitat can not be observed at the time of spawning
geographical coordinates but only after young fish enter the population as

 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.

18 B. Planque et al.

recruits. It has to be inferred from information on the thermal range is between 14 and 18C. In other
juveniles and young adults. The successful spawning regions the thermal constraints may be slightly dif-
habitat can also be studied by numerical modelling ferent; for example, van der Lingen et al. (2001) re-
(and in particular individual-based models (IBM)) by ports thermal range of 17.4–21.1C in the Benguela
following cohorts or individuals and their environ- region. In his world-wide analysis of anchovy distri-
ment during their early life stages. bution, Reid (1966) argues that the genus Engraulis is
The present study focuses on possible ways to define found from estuaries to high-salinity waters, so that it
and model anchovy and sardine ‘potential spawning is associated with coastal areas rather than with water
habitat’ in the Bay of Biscay. The issues related to of a given salinity range. Whilst river plumes (i.e. low
realized and successful habitat will be briefly discussed. salinity) appear to be recurrent preferential areas for
spawning, anchovy also spawns in other areas such as
Anchovy and sardine spawning habitats in the Bay of slope water eddies or the shelf break which are char-
Biscay acterized by high salinity throughout the water col-
The study of anchovy and sardine populations in the umn. Although there is a converging set of evidence
Bay of Biscay goes back at least to the early 20th that increase in river runoff is associated with increase
century (Fage, 1911; Furnestin, 1945), but it is only in in anchovy recruitment (see Lloret et al., 2004 and
the early 1960s that systematic sampling of sardine and references therein), the association between anchovy
anchovy eggs and larvae over most of the Bay of Biscay spawning and salinity is less obvious than for tem-
continental shelf was undertaken (Arbault and Bou- perature and the influence of salinity on anchovy
tin, 1967). These surveys were carried out and used by spawning distribution is still a matter of debate. Stra-
Arbault and Lacroix (1971, 1977) to produce the first tification, retention and plankton production have
description of sardine and anchovy spawning in the been proposed as other controlling factors for anchovy
area. The geographical occupation of sardine and an- spawning (or spawning success) in the Bay of Biscay
chovy extends beyond the Bay of Biscay, with sardine (Motos et al., 1996). However, there has yet been no
ranging from the western African coast (Furnestin and quantitative assessment of the link between these
Furnestin, 1959; Ettahiri et al., 2003) to southern factors and anchovy spawning.
Norway (Parrish et al., 1989) and anchovy ranging Finally, spawning habitat appears to depend (i) on
from western Africa (5N) to the northern North Sea seasonal timing, with adults migrating north and west
(Reid, 1966). However, the two populations appear in as seasons progress (Uriarte et al., 1996) and (ii) on
substantial quantities in the Bay of Biscay and both adult population size with spawning habitat extent
spawn in the area. The spatial and seasonal extent of increasing with adult population size (Motos et al.,
spawning in the Bay of Biscay has been reviewed and 1996). This latter effect is directly related to ‘realized’,
re-analysed in Bellier et al. (2006). They have shown rather than ‘potential’ spawning habitat.
that for the two species, spawning areas can be divided Sardine
into recurrent (or refuge) sites where spawning is ob- In the Bay of Biscay, sardine spawning occurs within a
served every year and optional sites where the prob- thermal window of 12.5–15C according to Sola et al.
ability of spawning varies greatly from year to year. In (1990) and of 10–16C according to Arbault and
addition, it appears that the average spawning region Lacroix (1977). The spatial distribution of sardine is
has moved from the 1960–70s to the present period generally more widespread and fragmented than that
2000–2004, possibly as a result of environmental of anchovy. Optimal temperature for spawning sardine
changes. can vary greatly between regions and between studies.
Anchovy In the North Pacific the thermal range for Sardinops
Anchovy spawning occurs preferentially close to the sagax is often given as about 13.5–17C (Tibby, 1937;
coast and sometimes at the shelf break or in oceanic Ahlstrom, 1965; Parrish et al., 1989; Lluch-Belda
slope water eddies (Motos et al., 1996). Spawning et al., 1991), although Hammann et al. (1998) report a
season extends from March to August with a maxi- much warmer temperature range of 16.9–20.8C and
mum intensity between May and June. Intensity of conversely Lynn (2003) reports spawning occurring in
spawning appears to be constrained by thermal envi- colder temperature of 12–13C off southern and cen-
ronment: Arbault and Lacroix (1977) have reported tral California. In the Benguela upwelling system, the
anchovy spawning within a thermal window of 14– range of temperature for spawning sardine (S. sagax) is
20C, Sola et al. (1990) have found a thermal window bimodal, with a major peak at 15.5–17.5C and a
of 16.5–19C, whilst according to Motos et al. (1996) secondary peak between 18.7 and 20.5C (van der

 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.

 Modelling potential spawning habitat of sardine and anchovy 19

Lingen et al., 2001). In south Pacific waters, Ward and Figure 2. Spatial location of hydrological stations (open
Staunton-Smith (2002) report spawning temperature circles) and CUFES sampling midpoints (dots) during the
range of 14–23C. Off the Moroccan coast (north west cruise PELGAS2004. The location of the Loire, Gironde and
Africa) Sardina pilchardus spawning is observed within Adour river mouths are indicated as well as Brittany and ‘Les
Landes’ regions. Bathymetry for 20, 50, 100, 200, 500 and
the temperature range 16–18.5C (Furnestin and
1000 m isobaths is also indicated.
Furnestin, 1959; Ettahiri et al., 2003). The seasonality
of spawning appears to vary with latitude as a result of
latitudinal gradients in sea surface temperature regimes
(Stratoudakis et al., 2004). The impact of salinity has France
Britta
been little described for sardine populations. ny

Objectives of the study

Loire
In the present study, we attempt to define the set of
hydrographic conditions that are suitable for spawning
of sardine and anchovy populations, i.e. to define the
potential spawning habitat of sardine and anchovy in
the Bay of Biscay. To do so, we aggregate data on fish
egg and hydrology collected during five spring surveys Gironde
(2000–2004) in the Bay of Biscay. The presence of
eggs (modelled as a probability of presence as in Bellier

Les landes
et al., 2006) is then related to hydrographic conditions
by generalized additive modelling. Our intention is to
test whether our observations during the period 2000–
2004 confirm previous conclusions on the possible
influence of salinity and temperature on sardine and Adour
anchovy spawning habitat. In addition, we investigate Spain
the possible role of other hydrographic parameters
related to water column stratification, on the potential
spawning habitat of the two species. Finally, we show
how this information can be used to represent poten- May to 24th June in 2003 and 27th April to 24th May
tial spawning habitat on the basis of simulated in 2004. As this is the period during which thermal
hydrographic fields, as a first step towards predicting stratification is established and river runoff diminishes,
spawning habitat changes under possible climate sce- slight changes in the timing of cruises can have a large
narios. impact on the hydrographic conditions encountered.
The cruise track showing the location of Con-
tinuous Underway Fish Egg Sampler (CUFES) samples
DATA AND METHOD
and hydrographic stations in 2004 is shown in Fig. 2.
The PELGAS cruises 2000–2004 Cruise tracks for 2000, 2001, 2002 and 2003 are sim-
Since 2000, large scale cruises covering most of the Bay ilar (although not strictly identical) to the one in
of Biscay continental shelf along the French coast have 2004.
been carried out during spring. These cruises, called
Collection of egg samples – data processing
PELGAS (Pelagic Gascogne), are primarily designed
for the acoustic assessment of small pelagic fish stocks Continuous fish egg sampling was performed using a
in the area. However, a number of additional data are CUFES (Checkley et al., 1997), mounted outboard of
collected, which include fish egg and larvae sampling, the R/V Thalassa. The CUFES continuously pumps
hydrology, phytoplankton and zooplankton sampling, sea water at 3 m depth at a rate of about 500 L min)1
sea mammal and sea bird observations. In the current (8.3 · 10)3 m3 s)1). The eggs are concentrated into a
study, we have used fish egg and hydrographic data. small volume of water and samples are collected every
Because of operational and logistical constraints, 20 min. A sample approximately corresponds to 10 m3
the cruises have taken place at different dates every of filtered sea water whilst the ship has covered a
year: 17th April to 14th May in 2000, 28th April to distance of about 3 nautical miles (5.5 km). The exact
4th June in 2001, 10th May to 5th June in 2002, 30th pump flow rate and duration of sampling are recorded.

 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.

20 B. Planque et al.

After collection, the eggs are identified to species level (2006). Egg concentrations were transformed to pres-
for sardine and anchovy and are counted. The results ence/absence binary data, before being interpolated on
are standardized to egg concentration i.e. ‘number of a regular grid of 1/8th degree by ordinary point kriging
eggs per 10 m3 of filtered sea water’. The distribution (Cressie, 1993). The cartography of the six hydrogra-
of egg presence in the Bay of Biscay derived from phic parameters was also performed by kriging on the
CUFES sampling is limited to the subsurface layer same regular grid of 1/8th degree. Kriged data for fish
where water is pumped (3 m). However, the vertical eggs and hydrographic parameters were used as input
distribution of fish eggs in the water column is rarely to the Generalized Additive Model fitting.
uniform or random (Coombs et al., 1985; Stenevik
et al., 2001; Boyra et al., 2003) so that CUFES esti- Potential habitat defined from Generalized Additive Model
mates may not truly reflect egg abundance in the fits
water. To limit the discrepancies between true and Individual predictor fits
observed egg abundance, the analysis was restricted to Potential spawning habitat is defined as the set of
egg presence/absence as in most cases, when eggs were hydrographic conditions within which egg presence
sampled with a vertical net (WP2, 200 microns mesh- can be observed. Prior to modelling potential spawn-
size, bottom to surface haul), they were also found in ing habitats, the kriged data (i.e. interpolated proba-
CUFES samples. We envisaged the use of numerical bilities of egg presence and interpolated hydrographic
models to simulate the vertical distribution of fish eggs predictors) from all years have been pooled in a single
in the water column. However, there are major diffi- table. Generalized additive models (GAMs, Hastie
culties in following such an approach because (i) and Tibshirani, 1990) constitute a practical method
current models are designed to simulate egg abundance for fitting smoothed curves to sets of data with single
rather than egg presence, (ii) different sets of models or multiple predictor and single response variable. In a
based on steady state solutions (Sundby, 1983; Boyra first step, we have used GAM models with single
et al., 2003) or dynamic solutions (Westgard, 1989) do predictors to identify the relationships between indi-
not provide identical results and (iii) recent results vidual hydrographic predictors and the probability of
suggest that egg density may vary between years egg presence. The selection of the GAMs smoothing
(Petitgas et al., 2004) but data on egg density of sar- predictors was done following the method proposed by
dine and anchovy were not available for the period of Wood and Augustin (2002), using the ‘mgcv’ library in
study. As a consequence the modelling of egg vertical the R statistical software (R Development Core Team,
distribution was not undertaken and the results of this 2004). The output of the GAMs are smoothed fits for
study only refer to subsurface egg presence/absence. each hydrographic predictors. The individual models
can not be tested for significance using the P-values
Collection of hydrographic data – data processing provided by ‘mgcv’ library as the true number of de-
Hydrographic profiles (temperature, salinity and den- grees of freedom is unknown, and probably much
sity) were conducted at fixed stations at night using a smaller than the one used to compute the P-value
conductivity, temperature and depth (CTD) probe. The because of strong spatial autocorrelation in the data.
number of stations has varied slightly between years. However, each fit can be analysed with regards to the
From each CTD profile, six parameters were derived: level of deviance explained (0–100%; the higher the
surface and bottom temperature, surface and bottom better), the Generalized Cross Validation score (GCV;
salinity, potential energy deficit (PED) and mixed-layer the lower the better) and the confidence region for the
depth (MLD). PED, which is a measure of vertical smooth (which should not include zero throughout the
density stratification was calculated as in Allain et al. range of the predictor). The hydrographic predictors
(2001). MLD was estimated using a two-layers model as can be ranked according to the above criteria, so that
in Planque et al. (2004). When vertical profiles excee- the best model can be selected.
ded 100 m depth, only the first 100 m were retained for Multiple predictor fits
the analysis, as some external sensors mounted on the The GAMs allow for fitting a single response variable
probe did not always allow for deeper sampling. (here, the probability of egg presence) to multiple
predictors (here, the hydrographic predictors). On the
Cartography of fish egg and hydrographic data
basis of predictor ranking performed above, we have
Fish egg and hydrographic data were interpolated on constructed series of GAMs of increasing complexity
the same spatial grid to allow direct comparison of the to model the probability of egg presence for sardine
two types of data. The cartography of the probability and anchovy. The ‘best’ models for sardine and an-
of egg presence was performed as in Bellier et al.
 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.
 Modelling potential spawning habitat of sardine and anchovy 21

chovy are again selected on the basis of GCV and RESULTS

level of deviance explained. These models form the
Egg distribution
basis of the potential habitat prediction (see below).
The distribution of anchovy eggs is generally confined
Prediction of potential spawning habitat from to the southern part of the Bay of Biscay continental
hydrodynamical simulations shelf. There is a large degree of interannual variability
Using the GAMs (constructed from kriged egg distri- in anchovy egg distribution. For example, the spatial
butions and hydrographic situations) it is possible to distribution in 2003 is patchy and spreads over most of
predict the probability of egg presence from a new set the shelf whilst in 2004 it is much narrower and
of hydrographic predictors. We have made such an concentrated along the southern coast of Les Landes
attempt using hydrographic results for hydrodynamical and the Adour, Gironde and Loire estuaries (Fig. 3).
simulations. The simulations are generated by Distribution of sardine eggs can extend throughout the
MARS3D, a 3D hydrodynamical model covering the area covered by the survey, i.e. the Bay of Biscay
Bay of Biscay continental shelf. The model has a 5 km continental shelf. Eggs can be found at the southern
horizontal resolution and 10 vertical layers in sigma and northern bounds and in coastal as well as shelf
coordinates. A detailed description of the model break areas. As in the case of anchovy, there is a high
properties is given in Lazure and Jégou (1998) and degree of interannual variability of the location and
Planque et al. (2004). extent of sardine egg presence.

Figure 3. Spatial distribution of sardine (top) and anchovy (bottom) in 2000, 2001, 2002, 2003 and 2004. Grey scale is
proportional to the probability of presence (>1 egg m)3) of eggs.
2000 2001 2002 2003 2004

48 48 48 48 48

47 47 47 47 47

46 46 46 46 46
Anchovy

45 45 45 45 45

44 44 44 44 44

43 43 43 43 43

5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1

2000 2001 2002 2003 2004

48 48 48 48 48

47 47 47 47 47

46 46 46 46 46
Sardine

45 45 45 45 45

44 44 44 44 44

43 43 43 43 43

5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1

Probability of egg presence

0 0.5 1

 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.

22 B. Planque et al.

Hydrology solar heating in 2003). The six hydrographic param-

The distribution of sea surface temperature (SST, eters chosen provide complementary information, and
Fig. 4a) reflects the state of thermal stratification although they can be well structured in space (spatial
during the cruise. In 2000 and 2004 warmer temper- autocorrelation), they are not strongly correlated with
atures were recorded in the northern part of the shelf each other.
as this area was visited at the end of the cruise, once
Relationships between egg presence and individual
stratification had taken place. In 2001, the sampling
hydrographic predictors
plan was reversed, with the southern part of the shelf
visited during the second part of the cruise. This ex- Temperature
plains the apparent strong temperature gradient ob- The link between temperature and the presence of
served around 47N. In 2003 the cruise took place anchovy eggs is clearly visible (Fig. 5a,c left) with
after the onset of spring thermal stratification and all spawning mostly occurring in waters of 14.5–19C at
SST display high values. Bottom temperatures display the surface and 12–15C at the bottom. The upper
less interannual variability. The characteristic features temperature limits can not be clearly defined from the
are the presence of the ‘cold pool’, a cold (<12C) GAM plot as the coefficient does not decrease towards
bottom body of water (Vincent and Kurc, 1969; Puillat a value significantly lower than zero. The lower bot-
et al., 2004) visible in the northern part of the shelf in tom temperature limit is evident, with a sharp decline
2000, 2002 and 2004, and warm coastal water strips in in coefficient values below 12C. The influence of
2001–2004 (Fig. 4b). The distribution of surface temperature on the potential habitat of sardine is also
salinity (Fig. 4c) is related to the Loire and Gironde visible (Fig. 5a,c right). There appears to be a thermal
river outflows which have greatly varied during the preference around 12.5–15C at the surface. For bot-
5 yr. A maximum runoff in 2001 is visible with low tom temperature the response is bimodal with a first
salinity extending far out on the shelf and a low mode around 11.75C and positive coefficients values
salinity water lens located in front of the Gironde beyond 12.5C.
estuary. Bottom salinity is not directly related to sur- Salinity
face salinity and generally follows a the bathymetry Anchovy eggs are found preferentially in waters with
towards the shelf break (Fig. 4d). The MLD (Fig. 4e) low surface salinity (<34 Fig. 5d left) with almost a
is also conditioned by local bathymetry with greater monotonic decline towards high salinity values. The
MLD in deeper waters. However, the geographical influence of bottom salinity is less obvious although
extent and location of maximum MLD along the shelf there is a decline in the GAM function with
break varies between years, as seen for example in increasing salinity (Fig. 5b left). A small peak is ob-
2001 (maximum MLD around 47N) and in 2000 and served for high salinity (around 35.5) probably related
2004 (maximum MLD around 45N). The PED re- to spawning observed in the region of the shelf break.
flects the combined thermal and haline stratification For sardine the effect of surface salinity appears
and is greatly variable between cruises (Fig. 4f). In bimodal with ranges of <33.5 and >35.25 (Fig. 5d
2000, the high PED in the north reflects thermal right), this latter value corresponding to open ocean
stratification towards the end of the cruise. In 2001 the conditions. The bottom salinity situation (Fig. 5b
northern stratified lens reflects thermal stratification in right) is contrasted with a single mode around 34.25–
addition to the presence of low salinity water 35.25 and no secondary peak for high bottom salinity
extending off the coast. The southern stratified lens is values.
mostly the result of very low surface salinity in front of Stratification
the Gironde estuary. In 2002, the averaged PED values Anchovy spawning is related to mixing depth, with
reflect the average surface salinity and temperature greater probability of egg presence in shallow mixed-
distributions. In 2003 the late timing of the cruise layer conditions (Fig. 5e left). Eggs are generally found
combined with abnormally warm temperatures resul- in very low numbers in waters with MLD >50 m. This
ted in strong thermal stratification. Similarly, in 2004 is combined with a preference for either weakly stra-
the PED is parallel to the distribution of surface tem- tified (PED <30 J) or highly stratified waters (PED
peratures but with much lower values. >110 J; Fig. 5f left). The combination of shallow MLD
In summary, the five cruises have taken place in and high stratification corresponds to river plume
contrasting hydrographic situations as a result of waters with strong and shallow haline stratification.
changes in individual cruise timing and meteorological The high values of the GAM for unstratified waters
conditions (e.g. high precipitation in 2001, strong

 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.

 Modelling potential spawning habitat of sardine and anchovy 23

Figure 4. Spatial distribution of (a) sea surface temperature averaged over the first 5 m, (b) sea bottom temperature averaged
over the last 5 m, (c) sea surface salinity averaged over the first 5 m, (d) sea bottom salinity averaged over the last 5 m, (e)
mixed-layer depth and (f) potential energy deficit in 2000, 2001, 2002, 2003 and 2004. When bottom depth exceeds 100 m,
hydrological parameters are calculated from the first 100 m. Black dots indicate location of CTD stations.
2000 2001 2002 2003 2004

48 48 48 48 48 a

47 47 47 47 47

46 46 46 46 46

45 45 45 45 45
Surface
44 44 44 44 44
Temperature

43 43 43 43 43

5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 10 15 20
2000 2001 2002 2003 2004

48 48 48 48 48 b

47 47 47 47 47

46 46 46 46 46

45 45 45 45 45 Bottom
44 44 44 44 44
Temperature

43 43 43 43 43

5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 10 15
2000 2001 2002 2003 2004

48 48 48 48 48 c

47 47 47 47 47

46 46 46 46 46

45 45 45 45 45 Surface
44 44 44 44 44
Salinity

43 43 43 43 43

5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 30 35
2000 2001 2002 2003 2004

48 48 48 48 48 d

47 47 47 47 47

46 46 46 46 46

45 45 45 45 45
Bottom
44 44 44 44 44 Salinity

43 43 43 43 43

5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 32 34 36

 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.

24 B. Planque et al.

Figure 4. (continued).
2000 2001 2002 2003 2004
e
48 48 48 48 48

47 47 47 47 47

46 46 46 46 46

45 45 45 45 45
Mixed-layer
44 44 44 44 44 Depth

43 43 43 43 43

5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 0 50 100
2000 2001 2002 2003 2004
f
48 48 48 48 48

47 47 47 47 47

46 46 46 46 46
Potential
45 45 45 45 45
Energy
44 44 44 44 44 Deficit

43 43 43 43 43

5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 5 4 3 2 1 0 100 200

corresponds to tidally mixed coastal waters. The Multiple predictor GAMs

situation for sardine is almost reversed, with probab- The GAMs with increasing number of predictors (1–
ility of egg presence decreasing as stratification 6) have been constructed. For each level, all possible
increases (Fig. 5f right) and eggs found throughout the arrangements of predictors were tested and only the
range of mixing depths with higher presence proba- best combination (i.e. with the lowest GCV score) was
bilities for either deep or very shallow mixed layers retained. The results of these models are presented in
(Fig. 5e right). Table 2. For anchovy, the quality of the models
Ranking of hydrographic predictors increases up to five predictors, although little is gained
Aside from the shape of the relationship between a beyond the three-predictors model which includes
given predictor and egg presence, it is possible to rank bottom temperature, surface salinity and surface tem-
the hydrographic predictors according to the percent- perature. For sardine, there is a regular increase in
age of deviance they can explain and the GCV scores model quality (both in terms of GCV and deviance) as
of the GAM models (Table 1). For anchovy, the pre- the number of predictors increases and the best model
dictor that can best explain egg presence is bottom includes all hydrographic predictors. The overall per-
temperature, with the highest deviance (35%) and centage of deviance explained is always lower than for
lowest GCV score (0.508). Following predictors are anchovy, a result consistent with previous observa-
surface temperature and mixed-layer depth and to a tions on single predictor models.
lesser extent surface salinity. Bottom salinity and PED
seem to have a more marginal link to egg presence, Model simulations
with <10% of deviance explained. The situation is not Using hydrodynamical simulations for year 1999, we
as clear for sardine, with all hydrographic predictors have extracted the six hydrographic descriptors and
having similar values of both percentage of deviance used them for the prediction of egg presence probab-
explained and GCV scores. The generally lower values ility. For anchovy, the model with three predictors was
of deviance explained for sardine compared with an- retained, whereas for sardine all six predictors were
chovy suggests that hydrographic influence on sardine used, for the reasons given above. Predictions were
spawning areas is more limited than for anchovy. restricted for areas shallower than 1000 m. When
 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.
 Modelling potential spawning habitat of sardine and anchovy 25

Figure 5. Coefficients of the Generalized Additive Models (GAMs) for sardine and anchovy against bottom temperature (a),
bottom salinity (b), surface temperature (c), surface salinity (d), mixed-layer depth (e) and potential energy deficit (f). Black
thick lines indicate the value of GAMs coefficient, dotted line are the confidence intervals at P ¼ 0.05, and the horizontal line
indicates zero level.

Anchovy Sardine

4
4

a b a b
2

2
0

0
-2

-2
-4

-4
11 12 13 14 15 34.0 34.5 35.0 35.5 11 12 13 14 15 34.0 34.5 35.0 35.5
Bottom Temperature (°C) Bottom Salinity Bottom Temperature (°C) Bottom Salinity
4

4
c d c d
2

2
0

0
-2

-2
-4

-4

12 14 16 18 32.0 33.5 35.0 12 14 16 18 32.0 33.5 35.0

Surface Temperature (°C) Surface Salinity Surface Temperature (°C) Surface Salinity
4

e f e f
2

2
0

0
-2

-2
-4

-4

20 40 60 80 0 50 150 20 40 60 80 0 50 150
Mixed-layer depth (m) Potential Energy deficit (J) Mixed-layer depth (m) Potential Energy deficit (J)

modelled hydrographic predictors were outside the of the Bay of Biscay up to the Gironde estuary. The
range of observed values, no prediction for egg pres- potential habitat extends further offshore towards the
ence probability was estimated. north. Along the coast to the north of the Gironde
The modelled potential habitats for sardine and estuary, the modelled hydrographic conditions are
anchovy during spring 1999 are presented in Fig. 6 outside the range of the observations used to fit the
for four dates: 12 March, 6 April, 30 April and 24 GAMs and it is therefore not possible to predict the
May. potential spawning habitat for this region. At the end
At the end of winter (12 March) the modelled of April, the southern part of the Bay seems less
potential spawning habitat of anchovy is almost absent favourable for spawning whilst to the north of the
from the Bay of Biscay continental shelf, apart from a Gironde the potential habitat extends further towards
small coastal region north of Spain and west of 3W. the middle of the shelf. It is also noticeable that
In early spring potential spawning habitat extends potentially suitable spawning habitats are seen along
further east and along the French coast from the south the southern coast of Brittany. In late May the

 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.

26 B. Planque et al.

Table 1. Single predictor GAM fits for

Anchovy Sardine anchovy and sardine. For each predictor,
% GCV % GCV the percentage of deviance and the
Parameter deviance score deviance score Generalized Cross Validation score is
given.
BT (bottom temperature) 35.3 0.508 10.0 0.516
BS (bottom salinity) 6.7 0.648 13.1 0.483
ST (surface temperature) 16.5 0.614 10.9 0.520
SS (surface salinity) 10.2 0.633 8.5 0.517
MLD (mixed-layer depth) 15.2 0.602 10.8 0.506
PED (potential energy deficit) 7.0 0.635 13.7 0.494

Table 2. Multiple predictors GAM fits

Model % deviance GCV score for anchovy and sardine. The models
Anchovy retained for spawning habitat prediction
1: BT 35.3 0.508 are highlighted in grey.
2: BT + SS 39.2 0.484
3: BT + SS + ST 45.3 0.452
4: BT + SS + ST + MLD 46.6 0.443
5: BT + SS + ST + MLD + BS 47.0 0.441
6: BT + SS + ST + MLD + BS + PED 48.8 0.443
Sardine
1: BS 13.1 0.482
2: BS + BT 19.2 0.464
3: BS + SS + PED 30.5 0.419
4: BS + SS + PED + ST 33.6 0.407
5: BS + SS + ST + BT + MLD 38.4 0.392
6: BS + SS + ST + BT + MLD + PED 41.0 0.378

BT, bottom temperature; SS, surface salinity; ST, surface temperature; BS, bottom
salinity; MLD, mixed-layer depth; PED, potential energy deficit.

hydrographic conditions appear favourable over a wide

DISCUSSION
area south of 46N and more restricted towards the
coast in the north. There are small potentially The objective of the present work was to define the
favourable locations along the shelf break from 45 to hydrographic conditions which are potentially suitable
46N. The coastal band along the northern coast of for spawning of anchovy and sardine in the Bay of
Spain also seems potentially suitable for spawning. Biscay. As stated in the definition of ‘potential
The modelled potential spawning habitat of sardine spawning habitat’, this would have required data col-
generally covers larger areas and is more patchy than lection over a wide range of environmental conditions,
that of anchovy. In late winter, most of the coast along including areas and seasons when spawning does not/
northern Spain appears favourable for spawning, as never occur. As the field cruises are primarily designed
well as the southern part of the Bay of Biscay up to to monitor anchovy and sardine populations, they are
45N, excluding a narrow coastal strip. Hydrographic centred on the area and timing of spawning of the two
conditions also appear favourable in patchy regions species and therefore include only a fraction of the
over the central shelf. By early spring the potential non-spawning sites and are limited in the range of
spawning habitat has extended over a broader area but hydrographic situations encountered. Extension of the
maintains a similar geographical structure. At the end spatial and seasonal coverage of the cruises would
of April, most of the continental shelf is suitable for improve the general methodology, although this is
sardine spawning. By late May, potential spawning often hardly achievable because of field sampling cost.
habitat is restricted towards shallower waters, but is Alternatively, inclusion of results from other cruises
also present along the shelf break. which are conducted around the Iberian Peninsula, in
 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.
 Modelling potential spawning habitat of sardine and anchovy 27

Figure 6. Modelled potential spawning habitat of anchovy (top) and sardine (bottom) in 1999. Grey scale is proportional to egg
presence probability. Probabilities are derived from modelled hydrological fields and GAMs in Fig. 5.

12/3/1999 6/4/1999 30/4/1999 24/5/1999

48 48 48 48

47 47 47 47

46 46 46 46
Anchovy

45 45 45 45

44 44 44 44

43 43 43 43
4 3 2 1 4 3 2 1 4 3 2 1 4 3 2 1

48 48 48 48

47 47 47 47

46 46 46 46
Sardine

45 45 45 45

44 44 44 44

43 43 43 43
4 3 2 1 4 3 2 1 4 3 2 1 4 3 2 1

Probability of egg presence

0 0.2 0.4 0.6 0.8 1

the Celtic Sea, Irish Sea and Channel would certainly that the vertical distribution of anchovy and sardine
improve the results obtained and provide a more ro- eggs is variable but in most cases a fraction of the eggs
bust description of European anchovy and sardine is always present in the first 5 m (one notable excep-
potential spawning habitats. The current description tion is the absence of sardine eggs at the surface in one
of potential habitat is limited to observations realized haul in the western Channel in Coombs et al., 1985).
in the subsurface layer and could be improved if the It therefore appears to be a reasonable assumption that
vertical distribution of eggs was taken into consid- if eggs are present in the water column they will be
eration. This could be achieved by using vertical egg found at 3 m depth where the CUFES operates, and
distribution models such as those of Sundby (1983) or the presence/absence index used in the present study
Westgard (1989), which have been adapted to the Bay can be considered as a good proxy for presence or
of Biscay respectively by Boyra et al. (2003) and Pet- absence of eggs in the water column.
itgas et al. (2006). Previous studies (Coombs et al., Additional hydrographic or environmental varia-
1985; Olivar et al., 2001; Boyra et al., 2003) suggest bles may also be considered, although it is practical to

 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.

28 B. Planque et al.

constrain the list of parameters to a small number of Brittany, down to the latitude of the Gironde estuary,
variables that are regularly measured during field centred over the 100-m depth zone. Its location
cruises and which provide a thorough description of corresponds to the zone of weak tidal stirring, be-
the water column structure. The simulations of tween areas of stronger vertical mixing along the
potential spawning habitat rely heavily on the capa- coast and along shelf break (Le Fèvre, 1986). It is
bility of the hydrodynamical model to mimic realisti- difficult to specify a spawning thermal range from the
cally hydrographic conditions. It is expected that results in Fig. 5, but the optimal surface temperature
undergoing developments of the existing model will for spawning appears to be around 17C, a value
provide improved capabilities for habitat hind-, now- consistent with previous findings obtained with
or fore-casting. independent data in the same region (Arbault and
Another variable which can potentially influence Lacroix, 1977; Motos et al., 1996). The moderate role
the spatial extent of spawning is the biomass of adult of salinity in the Bay of Biscay is consistent with the
fish spawning. During the spring cruises, sardine and findings of Reid (1966) at the scale of the global
anchovy adult biomass are estimated by means of ocean. Aside from temperature effects, high egg
acoustics at the time when fish eggs are sampled with abundance seems to prevail in either coastal well
the CUFES, and this allows for direct comparison mixed areas (i.e. shallow mixed-layer and low strati-
between biomass and egg distribution at the time of fication energy) or highly stratified river plumes (i.e.
sampling. The relationship between observed biomass shallow mixed-layer depth, high stratification and low
and mean probability of egg presence for anchovy is surface salinity; Fig. 5).
almost null (r2 ¼ 0.02, N ¼ 5). For sardine there is a The possible influence of hydrographic factors on
positive relationship (r2 ¼ 0.6, N ¼ 4) but it results the spawning habitat of sardine seems less clear than
from 1 yr of observation only, in 2001, when the for anchovy (Tables 1 and 2). This result is consistent
estimated biomass was low and the mean probability with current knowledge of sardine behaviour which
of egg presence was also low. As a result, it is not includes swimming large distances (Parrish et al.,
possible to conclude that year-to-year variations in 1981), having a more fragmented spatial distribution
realized spawning habitat are directly linked to adult (Barange and Hampton, 1997; Curtis, 2004) and
fish biomass, at least for the period studied in the having a greater tolerance to a range of environmental
present work (although it may be true over longer conditions than anchovy (Bakun and Broad, 2003).
periods of time as suggested in Bellier et al., 2006). All hydrographic factors appear to have a similar
The present analysis shows that potential spawn- degree of influence on the spawning distribution of
ing habitat of sardine and anchovy in the Bay of sardine. Sardine appears to have a greater tolerance
Biscay can be at least partially modelled using than anchovy for low bottom temperature (Fig. 5)
hydrographic predictors. Surprisingly, bottom tem- although it also appears to avoid cold pool waters. The
perature appears to be the best predictor for anchovy range of optimal surface temperature is also shifted
potential spawning habitat, followed by mixed-layer towards lower values (12–15C) which is consistent
depth, and surface temperature whilst surface and with the results of Arbault and Lacroix (1977).
bottom salinity appear to play a lesser role (Table 1). Sardine eggs can be found in coastal waters (i.e.
To our knowledge the importance of bottom tem- shallow mixing depth and low stratification energy),
perature has not been reported before, probably but contrary to anchovy they are also found in areas of
because it has rarely been measured in similar studies deep mixing and low stratification energy which cor-
(many studies use satellite imagery or subsurface respond to thermal rather than haline stratification in
hydrographic measures and are therefore restricted to early spring. Sardines spawn over much of the area
surface temperature). It is likely that anchovy are not covered by the survey, and it is likely that the envi-
directly dependent on temperature close to the bot- ronmental conditions are not contrasted enough to
tom, but since temperature displays little gradient exert a clear control on spawning within the region.
below the thermocline, bottom temperature appears Such control may be observed at a larger scale, over
as a good proxy for the conditions between the the spatial and seasonal range of sardine spawning, as
bottom and the thermocline. The lower limit of suggested by Stratoudakis et al. (2004).
bottom temperature is about 12C, a value that The spatial distribution patterns generated from
corresponds to a cold bottom body of water found in hydrodynamic simulations (Fig. 6) provide a first
the Bay of Biscay and known as the ‘cold pool’ attempt to predict spawning habitats from environ-
(Vincent and Kurc, 1969; Puillat et al., 2004). This mental information only. The generated patterns are
structure is known to be present from southern for potential spawning habitat rather than realized
 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.
 Modelling potential spawning habitat of sardine and anchovy 29

spawning habitat, so that it is not suggested that an- REFERENCES

chovy and sardine have actually spawned in all the Agostini, V.N. and Bakun, A. (2002) ‘Ocean triads’ in the
habitat, but rather that the modelled habitats were Mediterranean Sea: physical mechanisms potentially struc-
suitable for spawning. turing reproductive habitat suitability (with example appli-
The modelled succession of spatial patterns for cation to European anchovy, Engraulis encrasicolus). Fish.
anchovy potential spawning habitat in 1999 is in Oceanogr. 11:129–142.
Ahlstrom, E.H. (1965) A review of the effects of the environ-
agreement with field observations, as described in ment of the Pacific sardine. ICNAF Spec. Publ. 6:53–76.
Motos et al. (1996), with early spawning taking place Allain, G., Petitgas, P. and Lazure, P. (2001) The influence of
along the coast in the southern part of the Bay, meso-scale ocean processes on anchovy (Engraulis encrasico-
followed by increasing spawning off the Gironde lus) recruitment in the Bay of Biscay estimated with a three-
estuary and a gradual displacement towards northern dimensional hydrodynamic model. Fish. Oceanogr. 10:151–
163.
coastal latitudes to the South of Brittany. The late Arbault, S. and Boutin, N. (1967) Oeufs et larves de poissons
spawning at the shelf break north of 45N is not well téléostéens dans le golfe de Gascogne en 1964. Cons. Int.
reproduced, although there are areas with high Explor. Mer. Comité Plancton, no L: 11.
probability of egg presence but with very restricted Arbault, S. and Lacroix, N. (1971) Aires de ponte de la sardine,
spatial extension. The modelled succession of spatial du sprat et de l’anchois dans le golfe de Gascogne et sur le
plateau Celtique. Résultats de 6 années d’études. Rev. Trav.
patterns for sardine is also consistent with existing Inst. Pêches Marit. 35:35–56.
field observations, with potential spawning habitat Arbault, S. and Lacroix, N. (1977) Oeufs et larves de clupeides
showing a more fragmented distribution and covering et engraulides dans le golfe de Gascogne (1969–1973). Dis-
a larger area than for anchovy and progressively tribution des frayères. Relations entre les facteurs du milieu
extending northward as the season progresses. As for et la reproduction. Rev. Trav. Inst. Pêches Marit. 41:227–
254.
anchovy, the shelf break spawning sites are repro- Bakun, A. (1996) Patterns in the Ocean. Ocean Processes and
duced by the model, but to a very spatially restricted Marine Population Dynamics. La Jolla: California Sea Grant
extent, at the edge of the model geographical College System, 323 pp.
boundaries. Bakun, A. and Broad, K. (2003) Environmental ‘loopholes’ and
fish population dynamics: comparative pattern recognition
with focus on El Niño effects in the Pacific. Fish. Oceanogr.
CONCLUSION 12:458–473.
Barange, M. and Hampton, I. (1997) Spatial structure of co-
Overall, the simulations suggest that previously ob- occurring anchovy and sardine populations from acoustic
served seasonal patterns in spatial distribution of data: implications for survey design. Fish. Oceanogr. 6:94–
spawning can be explained to a large extent by sea- 108.
Bellier, E., Planque, B. and Petitgas, P. (2006) Historical fluc-
sonal changes in hydrographic conditions. Such tuations in spawning location of anchovy (Engraulis encra-
conditions can be seen as providing the ‘hydrographic sicolus) and sardine (Sardina pilchardus) in the Bay of Biscay
envelope’ of the spawning species, a notion which during 1967–1973 and 2000–2004. Fish. Oceanogr.
can be related to the more general ‘bioclimate doi:10.1111/j.1365-2419.2006.00410.x.
envelope’ or ‘climate space’ (see Pearson and Daw- Boyra, G., Rueda, L., Coombs, S.H. et al. (2003) Modelling the
vertical distribution of eggs of anchovy (Engraulis encrasico-
son, 2003 and references therein). This provides hope lus) and sardine (Sardina pilchardus). Fish. Oceanogr. 12:381–
that realistic climate-driven simulations can lead to 395.
realistic assessment of hydrographic impact on an- Castro, L., Fréon, P., van der Lingen, C.D. and Uriarte, A.
chovy and to a lesser extent sardine potential (2005) Report of the SPACC Meeting on Small Pelagic Fish
spawning habitat spatial and temporal extent. How- Spawning Habitat Dynamics and the Daily Egg Production
Method (DEPM). Concepción, Chile GLOBEC report 22:
ever, one must not forget that predicting potential 107 pp.
spawning habitat is neither predicting realized Checkley, D.M.J., Ortner, P.B., Cummings, S.R. and Settle, L.R.
spawning nor predicting spawning success which (1997) A continuous underway fish eggs sampler. Fish.
depends upon adult population structure, biological Oceanogr. 6:58–73.
conditions in the ocean (e.g. predators and prey) as Coombs, S.H., Fosh, C.A. and Keen, M.A. (1985) The buoy-
ancy and vertical distribution of eggs of sprat (Sprattus
well as environmental hydrodynamical conditions sprattus) and pilchard (Sardina pilchardus). J. Mar. Biolog.
after spawning. The proposed methodology provides a Assoc. UK 65:461–474.
means for assessing changes in potential spawning Cressie, N.A.C. (1993) Statistics for Spatial Data. Revised Edi-
habitat in the context of predicted climate variability tion. New York: Wiley Inter-science, 900 pp.
and change. Whether or not these habitats will be Curtis, K.A. (2004) Fine scale spatial pattern of pacific sardine
(Sardinops sagax) and northern anchovy (Engraulis mordax)
used and will contribute to population growth re- eggs. Fish. Oceanogr. 13:239–254.
mains an open question.
 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.
30 B. Planque et al.

Ettahiri, O., Berraho, A., Vidy, G., Ramdani, M. and Do Chi, T. Parrish, R.H., Serra, R. and Grant, W.S. (1989) The monotypic
(2003) Observation on the spawning of Sardina and Sardi- sardines, Sardina and Sardinops: their taxonomy, distribution,
nella off the south Moroccan Atlantic coast (21–26N). Fish. stock structure and zoogeography. Can. J. Fish. Aquat. Sci.
Res. 60:207–222. 46:2019–2036.
Fage, L. (1911) Recherches sur la biologie de l’anchois (Engraulis Pearson, R.G. and Dawson, T.P. (2003) Predicting the impacts
encrassicholus Linné) - races - âge - migrations. Ann. Inst. of climate change on the distribution of species: are biocli-
Océanogr. 2:45. mate envelope models useful? Global Ecol. Biogeogr. 12:361–
Fretwell, S. (1972) Populations in a Seasonal Environment. New 372.
Jersey: Princeton University Press, 217 pp. Petitgas, P., Magri, S. and Lazure, P. (in press) One-dimensional
Furnestin, J. (1945) Contribution à l’étude biologique de la biophysical modelling of fish egg vertical distributions in
sardine atlantique (Sardina pilchardus WALBAUM). Rev. shelf seas . Fish. Oceanogr. 15:413–428.
Trav. Off. Pêches Marit. 13:221–386. Planque, B., Lazure, P. and Jegou, A.M. (2004) Detecting
Furnestin, J. and Furnestin, M.L. (1959) La reproduction de la hydrological landscapes over the Bay of Biscay continental
sardine et de l’anchois des côtes Atlantiques du Maroc shelf in spring. Clim. Res. 28:41–52.
(saisons et aires de pontes). Rev. Trav. Inst. Pêches Marit. Puillat, I., Lazure, P., Jégou, A.M., Lampert, L. and Miller, P.
23:79–104. (2004) Hydrographical variability on the French continental
Hammann, M.G., Nevarez-Martinez, M.O. and Green-Ruiz, Y. shelf in the Bay of Biscay during the 1990’s. Cont. Shelf Res.
(1998) Sapwning habitat of the Pacific sardine (Sardinops 24:1143–1164.
sagax) in the Gulf of California: egg and larval distribution R Development Core Team (2004) A Language and Environment
1956–1957 and 1971–1991. CalCOFI Rep. 39:169–179. for Statistical Computing. Vienna, Austria: R Foundation for
Hastie, T.J. and Tibshirani, R.J. (1990) Generalized Additive Statistical Computing, ISBN 3-900051-00-3, URL http://
Models. London: Chapman and Hall, 335 pp. www.R-project.org
Lazure, P. and Jégou, A.-M. (1998) 3D modelling of seasonal Reid, J.L. (1966) Oceanic environment of the genus Engraulis
evolution of Loire and Gironde plumes on Biscay bay con- around the world. CalCOFI Rep. 11:29–33.
tinental shelf. Oceanol. Acta 21:165–177. Sola, A., Motos, L., Franco, C. and Lago de Lanzos, A. (1990)
Le Fèvre, J. (1986) Aspects of the biology of frontal systems. Seasonal occurrence of pelagic fish eggs and larvae in the
Adv. Mar. Biol. 23:163–299. Cantabrian Sea (VIIIc) and Galicia (IXa) from 1987 to
van der Lingen, C.D., Hutchings, L., Merkle, D., van der 1989. ICES CM 1990/H:25:14.
Westhuizen, J.-J. and Nelson, J. (2001) Comparative Stenevik, E.K., Sundby, S. and Cloete, R. (2001) Influence of
spawning habitats of anchovy (Engraulis capensis) and sardine buoyancy and vertical distribution of sardine Sardinops sagax
(Sardinops sagax) in the southern Benguela upwelling eggs and larvae on their transport in the northern Benguela
ecosystem. In: Spatial Processes and Management of Marine ecosystem. S. Afr. J. mar. Sci. 23:85–97.
Populations. G.H. Kruse, N. Bez, A. Booth, M.W. Dorn, Stratoudakis, Y., Coombs, S.H., Halliday, N.C. et al. (2004)
S. Hills, R.N. Lipcius, D. Pelletier, C. Roy, S.J. Smith and Sardine (Sardina pilchardus) spawning season in the North
D. Witherell (eds). Fairbanks Alaska: University of Alaska East Atlantic and relationships with sea surface temperature.
Sea Grant College Program, pp. 185–209. ICES CM 2004/Q:19:19.
Lloret, J., Palomera, I., Salat, J. and Sole, I. (2004) Impact of Sundby, S. (1983) A one-dimensional model for the vertical
freshwater input and wind on landings of anchovy (Engraulis distribution of pelagic fish eggs in the mixed layer. Deep Sea
encrasicolus) and sardine (Sardina pilchardus) in shelf waters Res. Part I Oceanogr. Res. Pap. 30:645–661.
surrounding the Ebre (Ebro) River delta (north-western Tibby, R.B. (1937) The relation between surface water tem-
Mediterranean). Fish. Oceanogr. 13:102–110. perature and the distribution of spawn of the California
Lluch-Belda, D., Lluch-Cota, D.B., Hernandez-Vasquez, S., Sa- sardine (Sardinops caerulea). Cal. Fish. Game 23:132–137.
lina-Zavala, C.A. and Schwartzlose, R.A. (1991) Sardine Uriarte, A., Prouzet, P. and Villamor, B. (1996) Bay of Biscay
and anchovy spawning as related to temperature and up- and Ibero Atlantic anchovy populations and their fisheries.
welling in the California Current system. CalCOFI Rep. Sci. Mar. 60 (Suppl. 2): 237–255.
32:105–111. Vincent, A. and Kurc, G. (1969) Hydrologie, variations sai-
Lynn, R.J. (2003) Variability in the spawning habitat of Pacific sonnières de la situation thermique du Golfe de Gascogne en
sardine (Sardinops sagax) off southern and central California. 1967. Rev. Trav. Inst. Pêches Marit. 33:79–96.
Fish. Oceanogr. 12:541–553. Ward, T. and Staunton-Smith, J. (2002) Comparison of the
MacCall, A.D. (1990) Dynamic Geography of Marine Fish Popu- spawning patterns and fisheries biology of the sardine, Sar-
lations. Washington: University of Washington Press, 153 pp. dinops sagax, in temperate South Australia and sub-tropical
Motos, L., Uriarte, A. and Valencia, V. (1996) The spawning southern Queensland. Fish. Res. 56:37–49.
environment of the Bay of Biscay anchovy (Engraulis encra- Westgard, T. (1989) Two models for the vertical distribution of
sicolus L.). Sci. Mar. 60 (Suppl. 2): 117–140. pelagic fish eggs in the turbulent upper layer of the ocean.
Olivar, M.P., Salat, J. and Palomera, I. (2001) Comparative Rapp. P.-v. Réun. Cons. int. Explor. Mer. 191:195–200.
study of spatial distribution patterns of the early stages of Wood, S.N. and Augustin, N.H. (2002) GAMs with integrated
anchovy and pilchard in the NW Mediterranean Sea. Mar. model selection using penalized regression splines and
Ecol. Prog. Ser. 217:111–120. applications to environmental modelling. Ecol. Model.
Parrish, R.H., Nelson, C.S. and Bakun, A. (1981) Transport 157:157–177.
mechanisms and reproductive success of fishes in the Cali-
fornia Current. Biol. Oceanogr. 1:175–203.

 2006 The Authors, Fish. Oceanogr., 16:1, 16–30.