Ichthyoplankton transport from the African coast to the Canary Islands

Journal of Marine Systems 87 (2011) 109–122

Contents lists available at ScienceDirect

Journal of Marine Systems j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j m a r s y s

Ichthyoplankton transport from the African coast to the Canary Islands Timothée Brochier a,b,⁎, Evan Mason b, Marta Moyano b, Amina Berraho c, Francois Colas d, Pablo Sangrà b, Santiago Hernández-León b, Omar Ettahiri c, Christophe Lett e a

CNRS, UMR LOCEAN, Unité Mixte de Recherche 7159 CNRS/IRD/Université Pierre et Marie Curie/MNHN, Institut Pierre Simon Laplace, Boîte 100-4, Place Jussieu, 75252 Paris Cedex 05, France Facultad de Ciencias del Mar, Universidad de Las Palmas de Gran Canaria, Canary Islands, Las Palmas de Gran Canaria, 35017, Spain Institut National de Recherche Halieutique, 2 Rue Tiznit, Casablanca, Morocco d Institute of Geophysics and Planetary Physics (IGPP), University of California Los Angeles, Los Angeles, CA, USA e Institut de Recherche pour le Développement, UMI IRD 209 UMMISCO, Centre de Recherche Halieutique Méditerranéenne et Tropicale, Avenue Jean Monnet, BP 171, 34203 Sète Cedex, France b c

a r t i c l e

i n f o

Article history: Received 11 August 2010 Received in revised form 17 February 2011 Accepted 25 February 2011 Available online 13 April 2011 Keywords: Canary Current Hydrodynamic model Lagrangian model Upwelling filaments Transport Gran Canaria Ichthyoplankton Pelagic fish

a b s t r a c t The Canary Upwelling System (CUS), a major eastern boundary upwelling system, sustains large cross-border fisheries of small pelagic fish, which poses the question of stock connectivity. Studies suggest that ichthyoplankton transport from the northwest African coast to the Canary Islands (CI) is facilitated by coastalupwelling associated filaments. Here we analyze connections between larval supply to the CI and sardine and anchovy populations that spawn over the continental shelf. For both species, ichthyoplankton observations (1) at the shelf and (2) near the island of Gran Canaria (GC) are used. Predictions of ichthyoplankton transport to GC are obtained from the Ichthyop Lagrangian transport model, which is forced by a high-resolution hydrodynamic model (ROMS) that reproduces the regional circulation. Results show that upwelling filaments play an important role in the transport of larvae to GC. However, (1) filaments are not the only mechanism, and (2) filament presence does not necessarily imply larval transport. Anchovy and sardine larval presence at GC appears to be independent of the respective adult spawning seasonality. Combining of observed and modeled data does not succeed in reproducing the observed larval patterns at GC. Various hypotheses are proposed to explain this discrepancy in larval transport to GC. © 2011 Elsevier B.V. All rights reserved.

1. Introduction The Canary Upwelling System (CUS), located along the northwest coast of Africa, is one of the four major eastern boundary upwelling systems of the world ocean and supports an economically important fishery (Barton, 1998; Arístegui et al., 2009). The Canary Current (CanC) is the major regional current, transporting ~3 Sv southwards with significant seasonal variations (Fig. 1; Stramma and Siedler, 1988; Machín et al., 2006; Mason et al., in press). Upwelling, maintained by the northeasterly Trade winds, takes place over the continental shelf throughout the year but tends to be less intense north of the Canary Island (CI) archipelago at ~ 28°N (Wooster et al., 1976; Mittelstaedt, 1991). Associated with the upwelling is the equatorward Canary Upwelling Current that transports relatively cool upwelled water, and is confined to the near-shelf region (Pelegrí et al., 2005). The Trade winds and the CanC are strongly perturbed by the CI archipelago, which is comprised of seven large islands located between 100 and 500 km from the African coast. The CI act as a barrier to the current, producing an extensive mesoscale eddy field ⁎ Corresponding author at: CNRS, UMR LOCEAN, Unité Mixte de Recherche 7159 CNRS/IRD/Université Pierre et Marie Curie/MNHN, Institut Pierre Simon Laplace, Boîte 100-4, Place Jussieu, 75252 Paris Cedex 05, France. E-mail address: [email protected] (T. Brochier). 0924-7963/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jmarsys.2011.02.025

downstream of the archipelago (Arístegui et al., 1994; Tejera et al., 2002; Sangrà et al., 2009). These eddies interact with coastally generated upwelling filaments, promoting offshore transport of ichthyoplankton and other biological material from the African neritic zone, which may reach the Canary archipelago (Arístegui et al., 1997; Rodríguez et al., 1999, 2004, 2009; Hernández-León et al., 2007). As in all major upwelling regions, primary production is enhanced by nutrient enrichment, and large populations of small pelagic fish dominate the ecosystem biomass (Cury et al., 2000). These fish sustain important local and international fisheries. Since upwelling centers that serve as habitats to small pelagic fish extend across the boundaries of neighboring countries, stock identification and connectivity become relevant issues for fisheries management. Assessing the impact and consequences of connectivity of small pelagic fish populations within northwest African waters has been addressed by several CI research groups (University of Las Palmas de Gran Canaria, University of La Laguna, Canary Institute of Marine Science, and Spanish Oceanographic Institute) over the last two decades (Rodríguez et al., 2009, and references therein). Sardine (Sardina pilchardus) and anchovy (Engraulis encrasicolus) have been the key species for this research due to their commercial and ecological importance in the area. Both species, found in abundance along the northwest African coast, spawn in surface waters (upper 60 m; Rodríguez et al., 2006) over the shelf, such that their larvae are highly susceptible to offshore advection by

110

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

Fig. 1. Schematic map showing the ROMS model domain (red box) situated within the Canary Basin. The principle currents of the eastern subtropical gyre are shown in yellow: Azores Current — AzC; Canary Current — CanC. The 3 virtual larvae release zones for the Lagrangian experiments are indicated along the northwest African coast between 24° and 29°N. The target zone at Gran Canaria (marked GC) in the Canary Islands is also shown. The adjacent island of Tenerife is marked Te. The topography is taken from the GEBCO database (Hunter and Macnab, 2003), and isobaths are plotted in black at 200, 1000 and 3000 m.

Ekman transport associated with the upwelling. Field observations suggest that offshore larval transport for both species within upwelling filaments between Capes Juby and Bojador (Fig. 1) is widespread during summer (Rodríguez et al., 1999, 2004), the peak of the upwelling season. However, filament-related transport has also been observed in winter (Bécognée et al., 2009), such that larvae are likely to reach the CI throughout the year. As a consequence, recruitment and catches of small pelagic fish at the easternmost islands of the Canary archipelago may be linked to external larval supplies that came from the nearby African continent. Two recent projects led by the University of Las Palmas de Gran Canaria, PELAGIC (2000–2001) and CONAFRICA (2005–2007), have focused on evaluating the potential impact of this external larval supply on clupeoid larval populations at Gran Canaria (GC), an island situated at the center of the Canary archipelago (Fig. 1). These projects involved, respectively, fortnightly and weekly collections of fish larvae at the shelfbreak off GC, as well as hydrographic data. Both monitoring programs suggest that the presence of sardine larvae at GC coincides with the arrival of upwelling filaments at the island (Bécognée et al., 2006; Moyano et al., 2009; Moyano and Hernández-León, 2011). Sardine reproduce throughout the year off northwest Africa, but spawning peaks in winter/early spring between Capes Ghir and Bojador (Fig. 1). South of Cape Bojador, there is little difference in rates of spawning between summer and winter (Ettahiri et al., 2003; Berraho, 2007). At the CI, sardine are thought to reproduce only in winter (November–March) (Méndez-Villamil et al., 1997); however, results from the CONAFRICA monitoring suggest that sardine may not reproduce near GC, at least during the sampled period. This assumption is based on two facts: (1) neither sardine eggs nor early larvae were found in the samples around GC from 2005 to 2007; and (2) biometric studies at GC (Herrera Rivero et al., 2008) and Tenerife (Santamaría et al., 2008) found only immature fish after a two-year sampling (2005–2006), aside from a few individuals N19 cm (the size

of mass maturation) in April 2005. This apparent contradiction between Méndez-Villamil et al. (1997) and later observations may be related to the collapse of the sardine stock observed in 1996–1997 (Machu et al., 2009). Moyano (2009) suggested that the Canary sardine stock may not have recovered after this collapse, possibly because of competition with round sardinella (Sardinella aurita, tropicalization hypothesis, Brito et al., 2005). In any case, larval transport of African sardine larvae to GC is evident at least for those found in spring/summer off GC, when water temperatures around the island are at or above the upper thermal limit (22 °C) for this species to spawn in the area (Berraho, 2007). Anchovy spawning occurs year-round off the central African coast (24° − 32°N), peaking in summer and secondly in spring (Berraho, 2007). Although the spawning period of anchovy at GC is unknown, there is evidence that anchovy do reproduce at GC, since eggs were found every year in May and June during the 2005–2007 surveys (Moyano, 2009). Considering the spawning thermal range estimated for this species for the northwest African coast (15.5°–21.4 °C; Berraho, 2007), it is likely that anchovy reproduce at GC over the entire year, peaking in late spring–early summer. However, the presence of anchovy larvae off GC largely coincides with filaments reaching the island (Bécognée et al., 2006; Moyano, 2009), which suggests that offshore larval transport within these structures may significantly contribute to the local anchovy population. A previous modeling study has already stressed the importance of larval transport to the Canary archipelago (Brochier et al., 2008), but did not describe in detail the underlying associated mechanisms. In addition, Brochier et al. (2008) performed simulations at the relatively large scale of the whole Canary archipelago. However, observations show that the island of GC constitutes the western limit of the upwelling filaments that frequently extend offshore from the African coast (Arístegui and Montero, 2005). In the present study, an ichthyoplankton transport model (Ichthyop) forced by high-

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

111

Fig. 2. (a) Daily 4-km MODIS sea surface temperature (SST). 3-day averages of (b) SST and (c) sea surface salinity (SSS) from the 5.5-km solution of the ROMS ocean model used in this study.

resolution hydrodynamic model outputs is used to make a detailed study of the mechanism of fish larvae supply to GC. The purpose of this study is to investigate in detail the properties of this filament transport mechanism in order to improve our understanding of the link between continental fish abundance and larval supply at the CI. Therefore, we address the question of the geographical origin of neritic larvae that reach GC and the frequency, seasonality and duration of this transport. The main hypotheses addressed are: (1) episodic negative temperature and salinity anomalies in the surface waters around GC, indicators of the arrival of coastal upwelling filaments, may also be indices of ichthyoplankton transport to the island; and (2) the seasonality of anchovy and sardine larvae at GC depends on adult spawning seasonality over the continental shelf. We test these hypotheses by comparing transport model results with recent field observations of the seasonal abundance of sardine and anchovy ichthyoplankton around GC (Moyano, 2009) and spawning patterns over the African continental shelf (Berraho, 2007). 2. Material and methods 2.1. Oceanic model configuration Ocean simulations were carried out using the 3D Regional Ocean Modeling System (ROMS). ROMS is a free-surface, sigma-coordinate, primitive-equation ocean model, which employs the Boussinesq and hydrostatic approximations within a split-explicit advection scheme (Shchepetkin and McWilliams, 2005, 2009). In order to reach the spatial resolution required to accurately reproduce the mesoscale, we used an offline-nesting approach (Mason et al., 2010), where a coarseresolution, large-domain solution (the parent) is used to supply lateral boundary information (i.e., the prognostic variables: temperature, salinity, sea surface height, and barotropic and baroclinic velocity) to force a higher-resolution, small-domain solution (the child). In the present configuration the child domain is shown embedded (or nested) within the parent domain (not shown) by the red outline in Fig. 1. The parent domain is focused on the CUS, but also includes western Iberia to the north and the Azores to the west. The grid has a horizontal resolution of 10 km. A monthly climatological forcing regime is applied at the lateral boundaries (prognostic variables) and at the surface (wind stress, heat fluxes and precipitation), prepared using ROMSTOOLS (Penven et al., 2008). The parent was run for 10 years, with the first 4 years discarded as spin-up. This solution is an early version of the configuration which is described and validated by Mason (2009) and Mason et al. (in press). The solution has a realistic

seasonal cycle, and levels and distributions of mesoscale variability that are comparable with observations from altimetry. Maxima of mesoscale energy correspond to the Azores Current and to the lee region of the CI archipelago. The generation of intrinsic mesoscale variability is a characteristic feature of climatologically forced (quasi-equilibrium) simulations of this type. Similar ROMS solutions have been produced for the California Current System (Marchesiello et al., 2003) and the Peru– Chile region (Penven et al., 2005), where the primary aims of both studies was a rigorous evaluation of the seasonal dynamics and variability of the respective regions. However, climatological solutions lack realistic interannual variability, and synoptic forcing, and so may not be the right choice for certain problems (e.g., simulations of particular years, or of El Niño). As the present study aims to investigate the origins and dynamics of larval transport to GC, a climatological approach is an appropriate choice. The broad large- and mesoscale structure (major currents and eddies) resolved by the parent is transmitted to the child at the open boundaries (Fig. 1). The child domain has a resolution of 5.5-km with 32 vertical levels. The CI lie at the center of the domain, with the African coast extending north and south. The major capes that influence the hydrodynamic circulation in the region, namely Capes Ghir (~ 30.7°N), Draa (~ 28.7°N), Juby (~ 27.9°N) and Bojador (~26.1°N), and also the Dakhla peninsula (~23.7°N), are present. The child solution was run for 6 years in correspondence with the available parent data. Child boundary forcing files are prepared by interpolating the prognostic variables from the parent to the child boundaries using the roms2roms procedure outlined by Mason et al. (2010). The child surface forcing is identical to that of the parent. Fig. 2 shows fields of sea surface temperature (SST) and sea surface salinity (SSS). In Fig. 2a, a 4-km MODIS SST image (dated 4 September 2007) reveals a band of cool coastal surface water that corresponds to the coastal upwelling. The clear signal of a filament extends towards GC, its origin at the shelf between Capes Juby and Bojador. A smaller filament is seen off Cape Juby to the north. Fig. 2b and c shows respective SST and SSS 3-day mean snapshots from the child solution centered at 14 September of model year 2.1 A filament-pair between the capes in the ROMS SST is similar to that in the MODIS image. The larger of the filaments is entrained around a large cyclonic eddy, a frequent phenomenon in this region (Navarro-Pérez and Barton, 1998). The signature of the filament is also visible in the model SSS as a negative anomaly.

1

Note that ROMS model year 2 corresponds to Ichthyop model year 1.

112

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

Discrepancies in Fig. 2a and b between the observed and modeled SST, such as the offshore extent of cooler waters in the model and the magnitude of the large-scale SST, are explained by the climatological forcing, such that the model is not expected to simulate local conditions at discrete time periods. These figures serve to demonstrate the intense mesoscale activity that takes place in the upwelling region and around the archipelago; we see the model's capacity to reproduce realistic mesoscale structures over the study region, with the generation of fronts, filaments and eddies, that are comparable with observations.

2.2. Ichthyoplankton transport model and simulations The ichthyoplankton transport model Ichthyop (Lett et al., 2008), forced by 3-day-averaged velocity outputs from years 2 to 5 of the previously generated ROMS child configuration, is used to track the trajectories of virtual ichthyoplanktonic individuals. The transport model interpolates the velocity fields both temporally (with a time step of 2 h) and spatially. In order to study the frequency of ichthyoplankton transport to GC from different geographical areas, we defined three release zones along the African continental shelf (Fig. 1). The three areas are delimited by the African coast and the 200-m isobath. Latitudinal limits correspond to Capes Draa, Juby, Bojador and the Dakhla peninsula (see Section 2.1 for corresponding latitudes). Both anchovy and sardine eggs have been observed in the three release areas (Berraho, 2007). A previous modeling study has shown that the bulk of the ichthyoplankton transport to the CI originates from within these three areas (Brochier et al., 2008). The target area within which we count the arrival of simulated ichthyoplankton is defined by the 1000-m isobath around GC, which corresponds to an offshore extension of ~5–10 km (Fig. 1). We assume that larvae transported into this area may be able to swim toward the island. The island shelf region, where food is abundant, provides shelter and sustenance to the larvae. In contrast, larvae transported into oceanic waters are likely to die of starvation or predation. In general, the mechanisms of animal migration include the use of magnetic, celestial, and olfactory cues. Several experiments have shown that for some species, including clupeids (Sardinops neopilchardus), larvae use ambient sound as a navigation cue to orient their swimming towards reefs (Tolimieri et al., 2000; Leis et al., 2002; Leis and Lockett, 2005). This mechanism may be used by sardine and anchovy larvae to reach the Canary archipelago when the currents bring them sufficiently close to the islands, although this is a matter of ongoing research (Montgomery et al., 2006). Every 3 days during the four simulation years, a total of 5000 randomly distributed individuals were released from within the three release areas. This large number was chosen in order to reduce any significant variability in the number of virtual larvae reaching GC between each repeat of the experiment. As the areas have different sizes, this resulted in 34% of the total being released in zone 1 (1700 every 3 days), 11% in zone 2 (550 every 3 days) and 55% in zone 3 (2750 every 3 days), on average. Individuals were released at depths between the surface and 50 m, and tracked for 30 days within the virtual environment. Individuals were considered to be neutrally buoyant particles, and rapidly dispersed over the mixed layer following release. Although such behavior is unrealistic for larvae, this neutral scenario was chosen for simplicity. The potential impacts of vertical migrations were discussed in the light of sensibility tests done in a previous study (Brochier et al., 2008). At each time step, the number of individuals reaching the target area around GC was recorded. An upper time limit, or “lifetime”, of 30 days was imposed upon each individual, serving as a proxy for the duration of the planktonic larval stage of sardine and anchovy (Santos et al., 2007). The position of every simulated individual was saved every three days.

2.3. Ichthyoplankton abundance and distribution along the African coast The ichthyoplankton data along the African coast were synthesized from the Moroccan INRH (Institut National de Recherche Halieutique) biannual monitoring program from 1994 to 1999, and completed with four annual surveys from 2003 to 2006. Between 1994 and 1999, sardine and anchovy ichthyoplankton abundance and distribution were assessed over 11 oceanographic cruises in the Moroccan Atlantic waters, one each in winter and summer except during winter 1996 (Berraho, 2007). The subsequent monitoring cruises took place in October 2003, July 2004, November 2005 and July 2006. 5 cruises took place in the month of July (3 in zone 1), 2 cruises in August (4 in zone 1), 2 cruises in January (all zones) and 1 for the remaining months, except May, September and November when there were no cruises. The cruises in November 2005 and July 2006 coincided with the ichthyoplankton observations performed around GC (Section 2.4), and were used for direct comparison of the two time series. Sampling station locations were variable, extending from the coast to the 1500-m isobath. Ichthyoplankton samples were collected using a small-size Bongo net with a 20-cm mouth diameter and 417-μm mesh size, equipped with flow-meters to measure the volume of filtered water. Sampling was carried out using oblique hauls from the surface to a maximum of 100 m depth. The samples were preserved in 5% formalin.

2.4. Ichthyoplankton abundance around Gran Canaria During the PELAGIC project, fortnightly ichthyoplankton samplings were performed during daylight on the eastern and southern flanks of GC from July 2000 to June 2001. The six sampling stations were located over the 100-m isobath and were 10 nm apart. Fish larvae were collected every fortnight with double WP2 net (July– October 2000) and with Bongo net (November 2000–June 2001) tows down to 90-m depth at speeds of 2–3 knots. All fish larvae were sorted and clupeoid larvae were identified. The same stations were later sampled from January 2005 to June 2007 during the CONAFRICA project. Ichthyoplankton were sampled weekly at these stations using oblique Bongo net tows similarly as for PELAGIC. Nets were fitted with 200-μm mesh and with a flow-meter (General Oceanics) to measure the volume of filtered water. Clupeoid eggs were identified and all fish larvae were sorted and identified down to the lowest taxonomic level. See Bécognée et al. (2006) and Moyano (2009) for further information on the sampling and analysis procedures.

2.5. Analysis of the results To analyze and interpret our results we compared the simulated ichthyoplankton transport patterns with (1) the corresponding salinity and temperature fields of the hydrodynamic model, as these fields contain the signature of the upwelling filaments, and (2) field observations of ichthyoplankton abundance and distribution around GC. In order to improve the detection of upwelling filaments, we used minimum values of temperature and salinity at 30-m depth, from the coast to the 1000-m isobath; i.e. the target area in which individuals are considered to reach GC (see Section 2.2). The 30-m level was chosen in order to sample within the mixed layer and mitigate the effects of surface heat fluxes that potentially reduce upwelling filament signatures. The use of minimums (rather than averages) within the target area improved the detection of upwelling filaments approaching the eastern flank of the island. The anomalies were obtained by removing a 3-month running mean. Finally, observed egg distributions were used to build a climatology of sardine and anchovy spawning seasonality along the African coast.

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

113

b

1800 1500

600

1200 900

300

600 300

0

J FMAM J J A SOND J FMAM J J A SOND J FMAM J J A SOND J FMAM J J A SOND J Year 1 Year 2 Year 3 Year 4

0.2

2

Salinity Anomaly

0.15

1.5

0.1

1

0.05

0.5

0

0

−0.05

−0.5

−0.1

−1

−0.15 −0.2

0

No. virtual larvae (from zone 1)

900

2100

From zone 1 From zone 2 From zone 3

Salinity anomaly

Temperature anomaly

J FMAM J J A SOND J FMAM J J A SOND J FMAM J J A SOND J FMAM J J A SOND J Year 1 Year 2 Year 3 Year 4

−1.5

Temperature Anomaly

No. virtual larvae (from zone 2 & 3)

a

−2

Fig. 3. (a) 4-year time series of the numbers of simulated larvae (Ichthyop) from the 3 release zones reaching Gran Canaria. Note that a different scale is used for zone 1. (b) Corresponding time series of the subsurface (30 m) hydrodynamic model (ROMS) salinity and temperature anomalies around Gran Canaria.

3. Results 3.1. Modeled ichthyoplankton transport (Ichthyop) In this section we describe the transport patterns of virtual ichthyoplankton to GC predicted by the model for a theoretical homogeneous spawning over the continental shelf (within the three release areas). The 4-year time series of numbers of virtual larvae reaching the GC area shows large seasonal and interannual variability (Fig. 3a). The correspondence between the number of virtual larvae reaching GC and temperature and salinity anomalies around GC is reported in Table 1. Water temperature and salinity are expected to display negative anomalies in the presence of upwelling filaments. The majority (67%) of the virtual larvae reach GC during a period of negative salinity anomaly (R = −0.33, P b 0.001), but only 50% during a negative temperature anomaly (R = −0.11, P b 0.05). The correlation with the temperature anomaly was much lower, except for individuals released in zone 3, which usually reached GC at times of positive temperature anomalies (R = 0.255, P b 0.001). Higher corre-

lations are found between monthly virtual larvae arrivals and monthly standard deviations of salinity and temperature anomalies (respectively R = 0.59, P b 0.001 and R = 0.52, P b 0.001). This correlation shows that virtual larvae reached GC when intense temperature and salinity anomalies occurred near GC. For individuals released in zone 3 this correlation is weaker for salinity (R = 0.31, P b 0.05) and non significant for temperature. Monthly mean maps of virtual larvae concentrations from year 1 of the Ichthyop simulation presented in Fig. 4 indicate that mesoscale activity is stronger in spring and summer, when upwelling filament activity is at its peak (Wooster et al., 1976; Mittelstaedt, 1991). The concentration maps suggest three main pathways for offshore transport that originate near to Cape Juby, Cape Bojador and the Dakhla peninsula; these pathways are particularly visible in month 5. The capes are well known locations for the generation of upwelling filaments (Barton et al., 1998, 2004; Pelegrí et al., 2005), however less is known about the influence of the Dakhla peninsula, where the shelf is relatively wide, upon the mesoscale circulation. Examination of the model SST near to Dakhla along the 6 years of the ROMS simulation shows cooler waters covering the shelf, whilst offshore there are

Table 1 Correspondence of virtual larval arrivals at Gran Canaria with salinity and temperature anomalies.

Mean salinity anomaly upon arrival at GC Proportion of individuals arriving at GC during a negative salinity anomaly Correlation with salinity anomaly Correlation with monthly standard deviation of salinity anomaly Mean temperature anomaly upon arrival at GC Proportion of individuals arriving at GC during a negative temperature anomaly Correlation with temperature anomaly Correlation with monthly standard deviation of temperature anomaly NS not significant. ⁎ P b 0.05. ⁎⁎ P b 0.01. ⁎⁎⁎ P b 0.001.

Release zone 1

Release zone 2

Release zone 3

All release zones

− 0.01 70% − 0.333⁎⁎⁎ 0.51⁎⁎⁎

− 0.01 62% − 0.19⁎⁎⁎ 0.45⁎⁎

− 0.007 57% NS 0.31⁎

− 0.01 67% − 0.33⁎⁎⁎ 0.59⁎⁎⁎

− 0.06 °C 57% − 0.2⁎⁎⁎ 0.56⁎⁎⁎

− 0.03 °C 50% NS 0.28⁎

0.29 °C 6% 0.25⁎⁎⁎ NS

− 0.03 °C 50% − 0.11⁎ 0.52⁎⁎⁎

114

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

Fig. 4. Monthly virtual larvae concentrations from year 1 of the ichthyoplankton transport simulation. We note that year 1 is an arbitrary choice, transport patterns for years 2–4 were generally similar to those shown here.

frequent occurrences of cold fronts, not unlike the MODIS scenario in Fig. 2. The concentration maps in Fig. 4, coupled with examination of corresponding model SST (not shown), show two distinct scenarios that result in virtual larvae reaching GC: (1) filaments arriving from Cape Bojador (e.g., months 1, 7 and 8), and (2) filaments arriving from Cape Juby (e.g., months 6 and 10). Although offshore-propagating structures are frequently seen to originate from the shelf region

immediately north of Dakhla (release zone 3; e.g., months 4, 5 and 6), these only turn northwards on one occasion (month 9), enabling virtual larvae to reach GC within the 30-day transport limit. A monthly climatology of the numbers of virtual larvae reaching GC (Fig. 5) shows a period of increased individual arrivals extending from May through December. Within this period a major peak occurs between June and August, with larvae arriving mainly from zones 1

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

1200

From zone 1 From zone 2 From zone 3

500

1000

400

800

300

600

200

400

100

200

0

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

No. virtual larvae (from zone 1)

No. virtual larvae (from zone 2 & 3)

600

0

Fig. 5. Monthly climatology of numbers of virtual larvae reaching Gran Canaria. Note the use of a different scale for zone 1.

and 2. A second peak is seen for these two areas from October to December. Interestingly, both peaks are preceded by periods where individuals arriving at GC mainly come from zone 3, in May and September. Fig. 6 shows the minimum travel duration from the release areas to GC to be 12 days for individuals released in zones 1 and 2 (June–July and November–December), and 15 days for zone 3 (September). From January to May, individuals from zone 1 travel for at least 27 days before reaching GC, and in April no individuals arrive at GC during their 30-day lifetime (Fig. 6). No individuals released in zone 3 reach GC from November to March. The size of each release area varies, thus the corresponding number of individuals released in each zone was different (see Section 2). Figs. 3 and 5 present the number of virtual larvae reaching GC according to their release zone. Note, however, that the relative proportions of individuals released in each area that were finally transported to GC ranges between 0 and 12.1% (zone 1), 11.8% (zone 2) and 3.5% (zone 3). 3.2. Observed ichthyoplankton abundance and distribution along the African coast Seasonal averages of egg distributions sampled during the Moroccan surveys confirm previous observations of sardine and anchovy spawning distributions on the African continental shelf (Furnestin and Furnestin, 1959; Berraho, 2007). For both species, eggs were found all year round over the entire shelf, but there were significant differences between summer and winter distributions of eggs and larvae north of Cape Juby, where sardine spawned more in winter and anchovy more in summer (Fig. 7). Fig. 8 shows monthly averaged egg and larval densities for anchovy and sardine, observed

Virtual larval age (days)

30

From zone 1 From zone 2 From zone 3

25 20 15 10 5 0 Jan

Feb Mar

Apr May Jun

Jul

Aug Sep

Oct

Nov Dec

Fig. 6. Climatological minimum ages of virtual larvae arriving at Gran Canaria, corresponding to travel time from the coastal release areas (zones 1, 2 and 3) to the target area around Gran Canaria.

115

within the three spawning zones defined for the Ichthyop simulations (Fig. 1). We now detail the situations observed in November 2005 and July 2006 in order to later compare with the larval sampling made at the same period at GC. In 2005 and 2006 the NAO index was negative at −1.34 and −0.20 (Hurrell, 2011), respectively, which usually means less intense upwelling activity over the Canary upwelling region (Machu et al., 2009). In terms of acoustic biomass estimations, both 2005 and 2006 can be considered “normal” years (El Ayoubi et al., 2005; Strømme et al., 2006). Mean egg densities (eggs 10 m− 2) calculated from the transect data at these periods are reported in Table 2. In November 2005, sardine eggs were found in similar quantities in the three release zones 1, 2 and 3. In the same period, sardine larvae were more abundant than eggs in the three zones. Anchovy eggs were as abundant as sardine in zone 1 (although larvae were less abundant), scarce in zone 2, and absent from zone 3. In July 2006, only zones 2 and 3 were sampled. Sardine eggs and larvae were scarce in zone 2 but numerous in zone 3. Anchovy eggs were a little bit more abundant than sardine eggs in zone 2, while anchovy larvae were a lot more abundant. Anchovy eggs were absent from zone 3 and larvae were scarce. For both species, eggs were located over the shelf (above the 200-m isobath), but larvae presented an extended distribution between the 200- and 1200-m isobaths (the upper limit of sampling for the two cruises), mainly in zones 2 and 3 (9 stations out of 25 were beyond the continental shelf). 3.3. Observed ichthyoplankton abundance around Gran Canaria During the PELAGIC and CONAFRICA surveys, environmental conditions followed the typical annual cycle at the CI, as detailed in Bécognée et al. (2006) and Moyano and Hernández-León (2011), respectively. Temperature ranged between 18° and 23 °C and salinity between 36.4 and 36.95. The initiation of the late winter bloom occurred when surface temperature dropped below ∼ 19 °C. At this point, mixing of the water column and consequent pumping of nutrients into the upper layers began, promoting an enhancement in production. The climatology of sardine and anchovy larvae around GC showed seasonal and interannual variability, but without any clear pattern (Fig. 9). Both species were found in all seasons, with highly variable abundances. Anchovy presence at GC was intermittent, larvae were not found in 6 months of the year (January, April, June, July, September and October). The highest larval densities were observed in November and August. Sardine larvae were absent only in June, September, October and November. The highest sardine larval density occurred in August. 3.4. Integration of model predictions and field data In this section, a monthly climatology of observed sardine and anchovy spawning, averaged within each release area (Fig. 1), is used to modulate the “weight” of virtual individuals released. This allows us to integrate the relative spatial (among the three zones) and seasonal variability of spawning intensity for sardine and anchovy (Fig. 8). Egg and larval distribution data were used separately in order to double-check the patterns, as data were scarce. Larvae collected over the shelf were usually recently hatched, and incubation time is typically 2–3 days, so their abundance is still a good proxy for spawning activity. The result is a prediction for a monthly climatology of larval transport to GC from the continental shelf (Fig. 10). For sardine larvae, the main arrival was predicted for the months of October and December (no data in September and November), the second period of arrivals was from June to August, and finally there was a small arrival predicted in February. The predictions for sardine egg distributions were similar to the larvae, except for October and December. For anchovy larvae, the main arrival at GC was predicted for June to August, although in July the prediction based on egg data was very low. However some larvae were predicted to reach GC in December, and some in October and April. No arrival was predicted in winter.

116

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

Fig. 7. (a) Summer and (b) winter maps of egg and larvae densities for sardine (left) and anchovy (right). Note that the egg positions are offset 4° to the west.

4. Discussion 4.1. Correlation of ichthyoplankton abundance around Gran Canaria with temperature and salinity anomalies In the central CCS, upwelling filaments are thought to be the main mechanism for ichthyoplankton transport from the continental shelf to the CI (Arístegui et al., 2009). Indeed, observations of sardine larvae around GC do usually correspond to low salinity anomalies, which were more frequent between May and December during the CONAFRICA sampling period (e.g., May, August and December 2005, and August 2006). The negative salinity anomalies are coincident with upwelling filaments reaching the eastern or southern shores of GC (as detected by remote sensing; Moyano, 2009), filaments that originate between the region just north of Cape Juby and Cape Bojador (Ichthyop zones 1 and 2). Our numerical results confirm this pattern, since we found a significant correlation between salinity anomalies and virtual larvae arriving at GC (Table 1). However the correlation was very weak for individuals released south of Cape Bojador (zone 3). Virtual larvae arrivals from this area were rare compared to those from zones 1 and 2. The arrivals from zone 3 were correlated with positive temperature anomalies and corresponded to longer travel times (Table 1; Fig. 6). Therefore larvae transported in upwelling filaments originating south of Cape Bojador may experience more

rapid growth (due to relatively higher temperatures; Rodríguez et al., 1999) over a longer period, than larvae coming from other areas of the continental shelf, and hence are significantly larger upon arrival at GC. Our results also suggest that (1) larvae being transported from the shelf to GC might not always imply a drop in salinity (for example November and December of year 3; Fig. 3) and, conversely (2) negative salinity anomalies do not always correspond with larvae arrivals (for example November of year 2; Fig. 3). Investigating these situations further, we find that the first scenario is associated with episodic events of strong anticyclonic circulation around GC that are associated with a northwesterly CanC incident flow (Mason, 2009) which typically occurs in summer in the model solution. Fig. 11 shows such an event: velocity vectors in white are overplotted onto the SSS in September of model year 4; trajectories of virtual larvae released over the continental shelf that reach the GC target area are plotted in red. The southeastward flow north of GC is advecting salty water, which compensates for the negative salinity anomaly associated with the coastal water that is responsible for the larval transport. However, although virtual larvae cross the 1000-m isobath they stay relatively far offshore of GC, and some are subsequently transported toward the south of Tenerife by the anticyclonic circulation around GC. Ontogenetic swimming behavior (absent from the Ichthyop dynamics) directed onshore may allow these larvae to remain near the island. Furthermore, it is worth mentioning that some 30-m salinity samples that were known (through SST remote

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

a − Sardine eggs

1500

1000

500

0

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

c − Anchovy eggs

600 400 200 0

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

d − Anchovy larvae

1000

1500

1000

500

0

800

Zone 1 Zone 2 Zone 3 no data

Density (N 10m−2)

Density (N 10m−2)

2000

b − Sardine larvae

1000

Density (N 10m−2)

Density (N 10m−2)

2000

800 600 400 200 0

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

117

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

−2

Fig. 8. Mean egg and larval density (N 10 m ) for sardine (a and b) and anchovy (c and d) measured over the continental shelf in zones 1, 2 and 3 (see Fig. 1). The data were collected over the years 1994–1999 and 2003–2006. Note that the y-axes are different for eggs and larvae. Extreme values are not shown, these are: (a) February (zone 1: 21,857) and October (zone 3: 34,977); (b) December (zone 2: 19,579); (c) April (zone 1: 40,812).

Table 2 Mean ichthyoplankton abundance (individuals per 10 m− 2) along the Moroccan coast observed in November 2005 and July 2006. “–” means no data. Species

Sardine

Anchovy

Release area

Eggs

November 2005

July 2006

Zone Zone Zone Zone Zone Zone

472 338 225 492 50 0

– 21 422 – 80 0

1 2 3 1 2 3

Larvae

798 1958 942 77 24 0

Eggs

Larvae

– 28 387 – 539 225

upwelling filaments could occur according to the lunar cycle (Bécognée et al., 2006), which will inevitably happen within a 30day period. Such long-lived upwelling filaments may either come from south of Cape Bojador, as observed by Barton et al. (2004), or from the north when a branch of the permanent filament off Cape Ghir drifts southward, a phenomenon not described in the literature but periodically seen in the ROMS solution. Finally, the correlation of virtual larvae arrivals around GC with rapid variations of temperature and salinity anomalies is consistent with the hypothesis that mesoscale eddy- and filament-like structures are responsible for these cross-shore transports. Indeed, intense mesoscale activity also corresponds to a higher variability of temperature and salinity fields in a given area. 4.2. Upwelling filament seasonality, larval surveys at GC and coastal spawning patterns The four-year climatology of ichthyoplankton transport model outputs shows a strong seasonal pattern of upwelling filament activity 0.012

Anchovy Sardine

0.01

No. larvae m−3

sensing) to have been collected from within upwelling filaments did not register as negative salinity anomalies (Moyano, 2009). Although there are no published observations of the GC anticyclonic circulation phenomenon, Navarro-Pérez and Barton (2001) used tide gauge data to show that in summer there is a tendency for northward (southward) flow west (east) of Tenerife, giving some observational support to the model results. It is expected that as high-resolution models at the CI continue to be developed, observational programs will be implemented to validate the model results and to further investigate the potential biophysical impacts of these flows. The second case, when salinity drops were detected without virtual larvae arrivals at GC, occurs when filaments older than 30 days (the value we used for the duration of the planktonic larval stage) reached GC. This scenario was also observed in the data, when significant drops in salinity did not coincide with either sardine or anchovy larval presence near GC (Moyano, 2009). Two reasons might explain the absence of larvae in the sampling at GC during the arrival of an “old filament”. The first reason proposed is that (1) a larva transported for longer than 30 days within an upwelling filament may have grown enough to attain significant swimming ability enabling it to escape from the sampling net. Another explanation (2) could be that larvae were preyed upon whilst being transported within the filament. Indeed, a periodically higher predation rate within the

0.008 0.006 0.004 0.002 0

Jan Feb Mar Apr May Jun

Jul

Aug Sep Oct Nov Dec

Fig. 9. Averages of different time series (2000–2001 and 2005–2007) of sardine and anchovy larvae abundance around Gran Canaria.

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

No. larvae reaching GC

No. larvae reaching GC

118

4

x 105

a − Predicted sardine larval abundance at Gran Canaria

Egg data Larvae data No data

3 2 1 0

4

Jan

Feb

x 105

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

b − Predicted anchovy larval abundance at Gran Canaria

Egg data Larvae data No data

3 2 1 0

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Fig. 10. Predicted monthly climatology of (a) sardine and (b) anchovy larval arrival at Gran Canaria, obtained by applying the model transport climatology to Moroccan egg (left) and larval (right) density observations. For ease of visualization, extreme values are hidden. The extreme values are in (a) October (egg data: 60,230), and in (b) July (larvae data: 79,321) and August (egg data: 46,531).

that drives virtual larvae transport from their spawning areas towards GC (Fig. 5). Although the model predicts that this transport may occur intermittently all year round, it is much more frequent and intense in summer, with a secondary maximum in fall. This is in line with oceanographic cruise observations around GC (e.g., Barton et al., 1998; Rodríguez et al., 2001) or in the coastal transition zone (García-Muñoz et al., 2004; Yebra et al., 2004; Rodríguez et al., 2009) where filaments are more frequently found in summer. The secondary peak in transport, predicted by the model in November–December, corresponds with the period of upwelling relaxation. The model also showed that periods of maximum intensity and frequency of

transport from the African coast to GC correspond to a shorter travel time (Fig. 6). 4.3. Age of larvae collected at GC versus model prediction of transport time We may expect to find younger larvae at GC during summer because of the faster transport from spawning areas suggested by our results. The size of larvae collected at GC ranges from 5.5 mm to 13.4 mm for sardine and from 6.4 mm to 12.2 mm for anchovy (Moyano, 2009), but the number of individuals was too small to

Fig. 11. Snapshot of surface salinity and virtual larvae trajectories (in red) from September of model year 4. The trajectories correspond to the full month of advection. Black dashed lines show the 1000-m isobath.

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

observe a significant seasonality of the size spectrum. In the following paragraphs, the age of the larvae, depending on growth rate, is discussed and compared with the model prediction of transport time. In the literature, sardine larvae have a faster growth rate (S. pilchardus, 0.41–0.57 mm day− 1, larvae caught in ∼15.5 °C water; Ré, 1984) than anchovy (E. encrasicolus, 0.25–0.41 mm day− 1, larvae caught in water ranging from 15.5° to 19.5 °C; Ré, 1996). Applying these growth rates, considering a size at hatching of 3 mm, the age of larvae collected near GC may range from 7 to 21 days for sardine and from 12 to 30 days for anchovy. The model predicts a minimum travel time of 12 days from spawning areas to GC in summer, which is consistent with the estimated age of anchovy larvae collected at GC. However, there is a difference of five days between the estimated age of the younger sardine larvae (7 days) and the minimum travel time predicted by the model (12 days). This inconsistency may be due to difference in local larval growth with values obtained from the literature. Furthermore, the shrinkage due to the preservation technique (formaline; Theilacker, 1980) used to calculate the age may cause an underestimate of larval size. It is worth mentioning that the smallest sardine larvae (∼5.5–6 mm) were collected off GC in August 2005, outside of the assumed spawning period for the species (winter) and in the presence of an upwelling filament reaching the island. Therefore it is very likely that, despite their small size, these sardine larvae were transported from the African coast to the island. Differences in travel time between anchovy and sardine may be caused by specific vertical swimming behavior. Indeed, Rodríguez et al. (2006) found differences in the vertical distribution of sardine and anchovy larvae. Their in situ observations showed that sardine larvae usually performed significant diurnal vertical migrations (amplitudes of ∼ 15 m), whilst anchovy larvae vertical migration was not significant (∼ 4 m). Furthermore, the daytime mean depth of the larvae showed a more superficial distribution for sardine (40.9 m) than for anchovy (46.9 m), which implies that during the night sardine larvae may be much more superficial than anchovy larvae (∼26 m vs. ∼ 43 m depth), significantly increasing their transport from the continental shelf to the CI, as Brochier et al.'s (2008) numerical study suggested. Temperature seasonal variability may also affect larval growth differently for anchovy and sardine. Indeed, Takasuka et al. (2007) found dome-shaped relationships between growth rate and temperature for both Japanese anchovy (Engraulis japonicus; optimal growth rate at 22.0 °C) and Japanese sardine (Sardinops melanostictus; optimal growth rate at 16.2 °C). Around the CI the typical seasonal variation of the mixed layer temperature is 18 °C in winter and up to 23 °C in summer, and observed optimal spawning temperature ranges at the Moroccan coast are 18°–20.5 °C for E. encrasicolus and 16° − 17 °C for S. pilchardus (Berraho, 2007). Thus, if Takasuka et al.'s (2007) result is general as they suggest, we might expect a faster growth rate in winter for sardine and in summer for anchovy (consistent with laboratory growth experiments on E. encrasicolus; Aldanondo et al., 2008). As the model predicts a shorter transport time in summer than in winter, sardine larvae should arrive at GC with a larger size in winter (long journey and high growth rate) than summer (short journey and low growth rate). For anchovy, the seasonal effects of temperature and transport time are the opposite, making it more difficult to draw any conclusion. A possible explanation for the disparity of the estimated age of larvae collected at GC and the travel time predicted by the Ichthyop model concerns grid resolution, the surface forcing of the ROMS hydrodynamic model (monthly averaged wind forcing), and the 3day averaging of the outputs. Mason (2009) showed that mesoscale energy in a ROMS solution is increased when using higher horizontal grid resolution and/or frequency of the surface wind forcing. For a horizontal grid resolution of 1 km, Mason (2009) also showed a clear emergence of intense submesoscale structures (eddies, fronts and filaments) which may have a significant effect on material concen-

119

trations and transport (Capet et al., 2008; McWilliams et al., 2009). The present ROMS resolution (5.5 km and 3-day averages) may therefore be a limitation to our approach as part of the eddy kinetic energy is missing at scales that are potentially important for upwelling filament kinematics. Hence we can expect that the present model configuration may not fully reproduce the vigorous filaments found in nature. However, this is unlikely to affect the seasonality of the mesoscale processes. 4.4. Mean spawning patterns versus transport seasonality We now discuss the seasonality of the number of virtual larvae transported from the African coast to GC in the light of the averaged observations of larval presence around GC, and the spawning seasonality over the continental shelf. The hypothesis is that seasonal patterns in spawning biomass along the African coast, and the likelihood of transport from the coast to GC, should be reflected in ichthyoplankton arrival at GC. However, despite strong seasonal differences in transport predicted by our model (Fig. 5), we noticed that there was no clear pattern in the pooled observations of larval presence around GC (Fig. 9). The spawning pattern was different for sardine and anchovy. Sardine mainly spawn during winter all along the shelf while anchovy spawn during summer in the region north of Cape Juby (Fig. 7). This pattern is in line with previous observations (Furnestin and Furnestin, 1959; Berraho, 2007). Both species spawn mainly over the inner part of the continental shelf, which is consistent with previous observations in this area (Rodríguez et al., 2004). However, as adult anchovy abundance is much lower than sardine's, during summer anchovy spawning intensity is actually only slightly more intense than that of sardine. The coupling of transport model predictions with pooled field observations of sardine and anchovy egg and larval densities in the three release areas clearly shows peaks of predicted anchovy and sardine larvae arrival at GC, in October and December for sardine, and in July and August for anchovy (Fig. 10). Summer is the main spawning period for anchovy along the African coast, and also the period when upwelling filaments reach GC most frequently. Furthermore, the model shows that these filaments originate mainly from zones 1 and 2, between Capes Draa and Juby (Fig. 5), which are the main anchovy spawning areas (Berraho, 2007; Brochier et al., 2008; Fig. 7). Thus, the combination of spawning and transport climatology predicts a strong anchovy larval arrival at GC in July and August (Fig. 10). This is not fully supported by the observations, as during summer anchovy larvae were found only in August (Fig. 9). An explanation for poor anchovy larval arrival at GC in summer despite the significant predicted transport might be that strong wind stress, occurring especially in June and July (Wooster et al., 1976), may increase larval mortality because of the high turbulence of the water column (Lasker, 1978). Alternatively, in November, the weakening of the wind opens an “optimal environmental window” (Cury and Roy, 1989), i.e., a suitable level of turbulence and upwelling intensity for larval survival. The strong anchovy larval presence at GC in November may be a consequence of the concordance of this window with a second peak of transport from zone 1, the main spawning zone for anchovy, to GC. In winter, the model predicts very low anchovy larval transport to GC, because of infrequent upwelling filaments and low spawning activity over the continental shelf (Fig. 10b). This result contrasts with the field observations as anchovy larvae were found in February and March. However, the seasonal pattern of anchovy larvae found at GC may only partially reflect the seasonality of transport because they also reproduce locally, as explained in Section 1. In contrast to anchovy, sardine do not appear to reproduce at GC, at least in recent years. Sardine spawn throughout the year over the continental shelf, with a maximum in winter (Fig. 7; Furnestin and Furnestin, 1959). However, because upwelling filaments are scarce in winter, our ichthyoplankton transport model predicts very little

120

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

transport to GC during this season. Combining these contrasting patterns, the model predicts sardine larval arrival at GC predominantly in October and December. This is not supported by the larval monitoring around GC, since no sardine larvae were found from September to November, and only a few were found in December. The optimal transport conditions from June to August compensated for limited spawning during these months, so that moderate sardine larvae transport is predicted by the model (Fig. 10a). Sardine spawn all year round between Cape Bojador and Dakhla (zone 3 in Fig. 8; Furnestin and Furnestin, 1959; Berraho, 2007). According to our model, transport from zone 3 to GC occurs mainly in May–June and September–October (Fig. 5). These episodes, limited to the beginning and the end of the upwelling season, may correspond with northward flow events through the eastern Canary Islands, as described by Hernández-Guerra et al. (2002) and Fraile-Nuez et al. (2010). As transport from zone 3 to GC was predicted to be longer than from other zones, especially in July, larvae coming from zone 3 may either have acquired sufficient swimming ability to escape the sampling net, or suffer high predation during transport (as explained in Section 4.1).

towards GC, intermittent patterns may be expected due to the intrinsic variability of these structures. Furthermore, despite the observed seasonality of anchovy and sardine spawning, monthly intermittency of spawning may also frustrate the formation of seasonal patterns of larvae presence at GC. Another contributing factor may be variability of predation in association with the lunar cycle (Bécognée et al., 2006; Hernández-León, 2008), which modulates larval mortality during transport to GC. In combination, these three modes of variability (i.e., filaments, spawning, and predation) may induce a random (chaotic) element to larval transport success, resulting in weak seasonality of neritic larval abundance around GC. However, a longer observation time series, including monitoring of both GC larval abundance and coastal spawning is necessary to confirm this result. Finally, anchovy spawning intensity (for the period studied) over the continental shelf may generally be too weak to supply larvae to GC through upwelling filament transport.

4.5. Spawning over the shelf versus larval presence at Gran Canaria: November 2005 and July 2006

It would be interesting to study in more detail the size spectrum of larvae both at GC and near to the continent, to improve our understanding of ichthyoplankton dynamics during transport. The obtained data could be compared using a similar modeling approach to that applied here, but with the inclusion of a bioenergetic larval growth model for each species (e.g., Urtizberea et al., 2008). Dynamical food field distributions can be obtained from a biogeochemical model, as previously applied in the Canary region by Machu et al. (2009). However, such bioenergetic growth models for Lagrangian individual based models are still under development at the present time. Also, it would be informative to use higher horizontal resolutions and/or synoptic atmospheric forcing in future numerical studies, as we suggest these were a possible limitation to the present numerical approach. Finally, it would be interesting to perform genetic studies on juveniles and adults in order to determine whether individuals transported to GC and that recruit at GC remain in Canary waters or return to the upwelling area to reproduce.

The results from the ichthyoplankton transport model suggest that the main source for larval supply at GC during summer and autumn is the continental shelf between Capes Draa and Juby (zone 1) and, to a lesser extent, between Capes Juby and Bojador (zone 2). Moreover, transport to GC from zones 1 and 2 is on average maximum in July, with a second peak in November (Fig. 5). However, two sets of observations demonstrate that spawning over the African coast does not always lead to larval detection at GC, even in the presence of an upwelling filament. Moderate anchovy spawning was observed over the continental shelf in November 2005 (zones 1 and 2) and July 2006 (zone 2; Table 10), but no anchovy larvae were found at GC during subsequent weeks despite the presence, in both months, of an upwelling filament identified in satellite imagery (Moyano, 2009). On the other hand, for both dates sardine spawning was globally more intense than that of anchovy, with an extended spatial distribution (eggs and larvae also found in zone 3; Table 2), and a large quantity of sardine larvae found at GC (Moyano, 2009). We see two possible explanations for the absence of anchovy larvae on both dates: (1) anchovy spawning intensity was too weak, and larvae arriving at GC were too sparse to be found; and (2) as anchovy spawning occurred mostly before the survey in November 2005 and July 2006 (many more larvae were collected than eggs; Table 2), larvae might have reached a sufficient size to swim and avoid the sampling net at GC, or to vertically migrate in order to avoid offshore transport and hence mature in nurseries at the African coast. 5. Conclusions The hydrodynamic model supports field observations of a correlation between negative salinity anomalies and larval transport to GC, but it also shows that this relation is not a systematic rule. Indeed, uncorrelated events may occur, either in the case of an anticyclonic circulation around GC (ichthyoplankton transport, but no salinity drop), or in the case of an “old filament” (salinity drop without ichthyoplankton transport). We therefore conclude that negative salinity anomalies are useful indicators of ichthyoplankton transport from the African continental shelf to GC, but caution should be exercised to ensure that neither of the above two conditions are present. Despite seasonality in the offshore transport regime and in spawning, there was no clear seasonal pattern to observed anchovy or sardine larvae presence at GC. A first explanation is that, as filaments and eddies are largely responsible for larvae transport

6. Perspectives

Acknowledgments The authors thank Philippe Verley for his valuable support with the Ichthyop software, and Pierrick Bécognée for providing part of the data used in this paper. Evan Mason was supported by the Spanish Government through projects MOC2 (CTM2008-06438-C02-01) and RODA (CTM2004-06842). Marta Moyano was supported by an FPU grant (AP 2005-4742). Field data collection for this study was funded by the projects Pelagic (EU-CICYT 1FD97-1084) and CONAFRICA (CICYT CTM2004-02319). Timothée Brochier was successively supported by EUR-OCEANS, a European Network of Excellence co-funded by the European Commission (6th Framework Programme, Contract No. 511106), by the PEPS project led by Vincent Echevin, and finally by the AMPHORE project led by Raymond Laë. Gaspard Bertrand's proofreading was useful, thanks for that. Finally, the authors express their gratitude to three anonymous referees for their helpful comments. References Aldanondo, N., Cotano, U., Etxebeste, E., Irigoien, X., Álvarez, P., Martinez de Murguía, A., Herrero, D.L., 2008. Validation of daily increments deposition in the otoliths of European anchovy larvae (Engraulis encrasicolus L.) reared under different temperature conditions. Fish. Res. 93 (3), 257–264. Arístegui, J., Montero, M.F., 2005. Temporal and spatial changes in plankton respiration and biomass in the Canary Islands region: the effect of mesoscale variability. J. Mar. Syst. 54 (1–4), 65–82. Arístegui, J., Sangrà, P., Hernández-León, S., Cantón, M., Hernández-Guerra, A., Kerling, J.L., 1994. Island-induced eddies in the Canary Islands. Deep-Sea Res. 49 (10), 1087–1101.

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122 Arístegui, J., Tett, P., Hernández-Guerra, A., Basterretxea, G., Montero, M.F., Wild, K., Sangrà, P., Hernández-León, S., Cantón, M., García-Braun, J.A., Pacheco, M., Barton, E.D., 1997. The influence of island-generated eddies on chlorophyll distribution: a study of mesoscale variation around Gran Canaria. Deep-Sea Res. 44 (1), 71–96. Arístegui, J., Barton, E.D., Álvarez-Salgado, X.A., Santos, A.M.P., Figueiras, F.G., Kifani, S., Hernández-León, S., Mason, E., Machú, E., 2009. Sub-regional ecosystem variability in the Canary Current upwelling. Prog. Oceanogr. 83 (1–4), 33–48. Barton, E.D., 1998. The Sea. The Global Coastal Ocean: Regional Studies and Syntheses. : Ch. Eastern boundary of the North Atlantic: Northwest Africa and Iberia coastal segment. John Wiley & Sons, New York, pp. 633–658. Barton, E.D., Arístegui, J., Tett, P., Canton, M., Garcia-Braun, J., Hernández-León, S., Nykjaer, L., Almeida, C., Almunia, J., Ballesteros, S., Basterretxea, G., Escánez, J., García-Weill, L., Hernández-Guerra, A., López-Laatzen, F., Molina, R., Montero, M.F., Navarro-Pérez, E., Rodríguez, J.M., van Lenning, K., Vélez, H., Wild, K., 1998. The transition zone of the Canary Current upwelling region. Prog. Oceanogr. 41 (4), 455–504. Barton, E.D., Arístegui, J., Tett, P., Navarro-Pérez, E., 2004. Variability in the Canary Islands area of filament–eddy exchanges. Prog. Oceanogr. 62 (2–4), 71–94. Bécognée, P., Almeida, C., Barrera, A., Hernández-Guerra, A., Hernández-León, S., 2006. Annual cycle of clupeiform larvae around Gran Canaria Island, Canary Islands. Fish. Oceanogr. 15 (4), 293–300. Bécognée, P., Moyano, M., Almeida, C., Rodríguez, J.M., Fraile-Nuez, E., HernándezGuerra, A., Hernández-León, S., 2009. Mesoscale distribution of clupeoid larvae in an upwelling filament trapped by a quasi-permanent cyclonic eddy off Northwest Africa. Deep-Sea Res. 56 (3), 330–343. Berraho, A., 2007. Relations spatialisées entre milieu et ichthyoplancton des petits pélagiques de la côte Atlantique marocaine (zones centrale et sud). Ph.D. thesis, Université Mohamed V, Agdal, Rabat, Morocco, 306pp. Brito, A., Falcón, J.M., Herrera, R., 2005. Sobre la tropicalización reciente de la ictiofauna litoral de las islas canarias y su relación con cambios ambientales y actividades antrópicas. Vieraea 33, 515–525. Brochier, T., Ramzi, A., Lett, C., Machu, E., Berraho, A., Fréon, P., Hernández-León, S., 2008. Modelling sardine and anchovy ichthyoplankton transport in the Canary Current System. J. Plankton Res. 30 (10), 1133–1146. Capet, X.J., McWilliams, J.C., Molemaker, M.J., Shchepetkin, A.F., 2008. Mesoscale to submesoscale transition in the California Current System. Part I: flow structure, eddy flux, and observational tests. J. Phys. Oceanogr. 38 (1), 29–43. Cury, P., Roy, C., 1989. Optimal environmental window and pelagic fish recruitment success in upwelling areas. Can. J. Fish. Aquat. Sci. 46 (4), 670–680. Cury, P., Bakun, A., Crawford, R.J.M., Jarre, A., Quiñones, R.A., Shannon, L.J., Verheye, H.M., 2000. Small pelagics in upwelling systems: patterns of interaction and structural changes in “wasp–waist” ecosystems. ICES J. Mar. Sci. 57 (3), 603–618. El Ayoubi, S., Chbani, C., Charouki, N., 2005. Rapport de la campagne du N/R Al Amir Moulay Abdallah, entre Cap Boujdor et Cap Cantin. Tech. rep., Institut National de Recherche Halieutique, Casablanca, Morocco. Ettahiri, O., Berraho, A.M., Vidy, G., Ramdani, M., Do Chi, T., 2003. Observation on the spawning of Sardina and Sardinella off the south Moroccan Atlantic coast (21–26°N). Fish. Res. 60 (2–3), 207–222. Fraile-Nuez, E., Machín, F., Vélez-Belchí, P., Lóez-Laatzen, F., Borges, R., Benítez-Barrios, V., Hernández-Guerra, A., 2010. Nine years of mass transport data in the eastern boundary of the North Atlantic Subtropical Gyre. J. Geophys. Res. 115. Furnestin, J., Furnestin, M.-L., 1959. La reproduction de la sardine et de l'anchois des côtes atlantiques du maroc (saisons et aires de ponte). Rev. Trav. Inst. Peches Marit. 23 (1), 81–104. García-Muñoz, M., Arístegui, J., Montero, M.F., Barton, E.D., 2004. Distribution and transport of organic matter along a filament–eddy system in the Canaries–NW Africa coastal transition zone region. Prog. Oceanogr. 62 (2–4), 115–129. Hernández-Guerra, A., Machín, F., Antoranz, A., Cisneros-Aguirre, J., Gordo, C., MarreroDíaz, A., Martínez, A., Ratsimandresy, A.W., Rodríguez-Santana, A., Sangrà, P., López-Laatzen, F., Parrilla, P., Pelegrí, J.L., 2002. Temporal variability of mass transport in the Canary Current. Deep-Sea Res. 49 (17), 3415–3426. Hernández-León, S., 2008. Natural variability of fisheries and lunar illumination: a hypothesis. Fish Fish. 9 (2), 138–154. Hernández-León, S., Gómez, M., Arístegui, J., 2007. Mesozooplankton in the Canary Current System: the coastal–ocean transition zone. Prog. Oceanogr. 74, 397–421. Herrera Rivero, I., Barrera Luján, A., Hernández-León, S., 2008. Length–weight relationships and reproduction of small and medium pelagic fish species in the Canary Island waters. June 2–6 Eastern Boundary Upwelling Ecosystems Symposium: Integrative and Comparative Approaches. Las Palmas de Gran Canaria, Spain, poster. Hunter, P., Macnab, R., 2003. The GEBCO Digital Atlas published by the British Oceanographic Data Centre on behalf of IOC and IHO. North Atlantic region. Hurrell, J., 2011. Annual Station Based NAO IndexURL http://www.cgd.ucar.edu/cas/ jhurrell/indices.data.html2011. Lasker, R., 1978. The relations between oceanographic conditions and larval anchovy food in the California Current: identification of factors contributing to recruitment failure. Rapp. p. v. Reun. ICES 173, 212–230. Leis, J.M., Lockett, M.M., 2005. Localization of reef sounds by settlement-stage larvae of coral-reef fishes (Pomacentridae). Bull. Mar. Sci. 76 (3), 715–724. Leis, J.M., Carson-Ewart, B., Cato, D., 2002. Sound detection in situ by the larvae of a coral-reef damselfish. Mar. Ecol. Prog. 232, 259–268. Lett, C., Verley, P., Mullon, C., Parada, C., Brochier, T., Penven, P., Blanke, B., 2008. A Lagrangian tool for modelling ichthyoplankton dynamics. Environ. Model. Softw. 23 (9), 1210–1214. Machín, F., Hernández-Guerra, A., Pelegrí, J.L., 2006. Mass fluxes in the Canary Basin. Prog. Oceanogr. 70 (2–4), 416–447.

121

Machu, E., Ettahiri, O., Kifani, S., Benazzouz, A., Makaoui, A., Demarcq, H., 2009. Modeling the environment of Sardina pilchardus over the Saharan Bank to investigate the collapse of the stock in 1997. Fish. Oceanogr. 18, 287–300. Marchesiello, P., McWilliams, J.C., Shchepetkin, A.F., 2003. Equilibrium structure and dynamics of the California Current System. J. Phys. Oceanogr. 33 (4), 753–783. Mason, E., 2009. High-resolution modelling of the Canary Basin oceanic circulation. Ph. D. thesis, Universidad de Las Palmas de Gran Canaria, 235 pp. Mason, E., Colas, F., Molemaker, J., Shchepetkin, A.F., Troupin, C., McWilliams, J.C., Sangrà, P., (in press). Seasonal variability of the Canary Current: a numerical study. J. Geophys. Res. doi:10.1029/2010JC006665. Mason, E., Molemaker, J., Colas, F., Shchepetkin, A.F., McWilliams, J.C., Sangrà, P., 2010. Procedures for offline grid nesting in regional ocean models. Ocean Modell. 35 (1–2), 1–15. McWilliams, J.C., Colas, F., Molemaker, M.J., 2009. Cold filamentary intensification and oceanic surface convergence lines. Geophys. Res. Lett. 36 (18). Méndez-Villamil, M., Lorenzo Nespereira, J.M., González Pajuelo, J.M., Soto Aguilera, R., 1997. Periodo reproductor y madurez sexual de la sardina Sardina pilchardus (Walbaum, 1792) en aguas de Gran Canaria (Islas Canarias). Bol. Inst. Esp. Oceanogr. 13 (1–2), 47–55. Mittelstaedt, E., 1991. The ocean boundary along the northwest African coast: circulation and oceanographic properties at the sea surface. Prog. Oceanogr. 26 (4), 307–355. Montgomery, J.C., Jeffs, A., Simpson, S.D., Meekan, M., Tindle, C., 2006. Sound as an orientation cue for the pelagic larvae of reef fishes and decapod crustaceans. Adv. Mar. Biol. 51, 143–196. Moyano, M., 2009. Temporal and spatial distributions of the ichthyoplankton in the Canary Islands. Ph.D. thesis, Universidad de Las Palmas de Gran Canaria, 293 pp. Moyano, M., Hernández-León, S., 2011. Intra- and interannual variability in the larval fish assemblage off Gran Canaria (Canary Islands) over 2005 to 2007. Mar. Biol. 158, 257–273. Moyano, M., Rodríguez, J.M., Hernández-León, S., 2009. Larval fish abundance and distribution during the late winter bloom off Gran Canaria Island, Canary Islands. Fish. Oceanogr. 18 (1), 51–61. Navarro-Pérez, E., Barton, E.D., 1998. The physical structure of an upwelling filament off the north-west African coast during August 1993. S. Afr. J. Mar. Sci. 19, 61–73. Navarro-Pérez, E., Barton, E.D., 2001. Seasonal and interannual variability of the Canary Current. Sci. Mar. 65, 205–213 supplement 1. Pelegrí, J.L., Arístegui, J., Cana, L., González-Dávila, M., Hernández-Guerra, A., Hernández-León, S., Marrero-Díaz, A., Montero, M.F., Sangrà, P., Santana-Casiano, M., 2005. Coupling between the open ocean and the coastal upwelling region off northwest Africa: water recirculation and offshore pumping of organic matter. J. Mar. Syst. 54 (1–4), 3–37. Penven, P., Echevin, V., Pasapera, J., Colas, F., Tam, J., 2005. Average circulation, seasonal cycle, and mesoscale dynamics of the Peru Current System: a modeling approach. J. Geophys. Res. 110. Penven, P., Marchesiello, P., Debreu, L., Lefevre, J., 2008. Software tools for pre- and post-processing of oceanic regional simulations. Environ. Model. Softw. 23 (5), 660–662. Ré, P., 1984. Evidence of daily and hourly growth in Pilchard larvae based on otolith growth increments, Sardina pilchardus (Walbaum, 1792). Cybium 8 (1), 33–38. Ré, P., 1996. Anchovy spawning in Mira estuary (southwestern Portugal). Sci. Mar. 60 (2), 141–153. Rodríguez, J.M., Hernández-León, S., Barton, E.D., 1999. Mesoscale distribution of fish larvae in relation to an upwelling filament off northwest africa. Deep-Sea Res. 46 (11), 1969–1984. Rodríguez, J.M., Barton, E.D., Eve, L., Hernández-León, S., 2001. Mesozooplankton and ichthyoplankton distribution around Gran Canaria, an oceanic island in the NE Atlantic. Deep-Sea Res. 48 (10), 2161–2183. Rodríguez, J.M., Barton, E.D., Hernández-León, S., Arístegui, J., 2004. The influence of mesoscale physical processes on the larval fish community in the Canaries CTZ, in summer. Prog. Oceanogr. 62 (2–4), 171–188. Rodríguez, J.M., Hernández-León, S., Barton, E.D., 2006. Vertical distribution of fish larvae in the Canaries–African coastal transition zone in summer. Mar. Biol. 149 (4), 885–897. Rodríguez, J.M., Moyano, M., Hernández-León, S., 2009. The ichthyoplankton assemblage of the Canaries–African Coastal Transition Zone: a review. Prog. Oceanogr. 83 (1–4), 314–321. Sangrà, P., Pascual, A., Rodríguez-Santana, Á., Machín, F., Mason, E., McWilliams, J.C., Pelegrí, J.L., Dong, C., Rubio, A., Arístegui, J., Marrero-Díaz, Á., Hernández-Guerra, A., Martínez-Marrero, A., Auladell, M., 2009. The Canary Eddy Corridor: a major pathway for long-lived eddies in the subtropical North Atlantic. Deep-Sea Res. 56, 2100–2114. Santamaría, M., González, J., López Abellán, L.J., Barrera, A., Balguerías, E., 2008. Substitution of sardine (Sardina pilchardus) for round sardinella (Sardinella aurita) in the Canary Island waters. June 2–6 Eastern Boundary Upwelling Ecosystems Symposium: Integrative and Comparative Approaches. Las Palmas, Spain, poster. Santos, A.M.P., Chícharo, A., Dos Santosa, A., Moita, T., Oliveira, P.B., Peliz, A., Ré, P., 2007. Physical–biological interactions in the life history of small pelagic fish in the Western Iberia Upwelling Ecosystem. Prog. Oceanogr. 74 (2–3), 192–209. Shchepetkin, A.F., McWilliams, J.C., 2005. The Regional Oceanic Modeling System (ROMS): a split-explicit, free-surface, topography-following-coordinate oceanic model. Ocean Modell. 9 (4), 347–404. Shchepetkin, A.F., McWilliams, J.C., 2009. Correction and commentary for “Ocean forecasting in terrain-following coordinates: formulation and skill assessment of the regional ocean modeling system” by Haidvogel et al., J. Comp. Phys. 227,

122

T. Brochier et al. / Journal of Marine Systems 87 (2011) 109–122

pp. 3595–3624. J. Comput. Phys. 228 (24), 8985–9000 J. Comput. Phys. 228 (24), 8985–9000. Stramma, L., Siedler, G., 1988. Seasonal changes in the North Atlantic subtropical gyre. J. Geophys. Res. 93 (C7), 8111–8118. Strømme, T., Charouki, N., Alvheim, O., 2006. Survey of the pelagic fish resources off North West Africa. Part III, Cape Blanc — Cape Sim, Morocco. 19 November–16 December 2006. Cruise reports “Dr. Fridjof Nansen”. Tech. rep., Institute of Marine Research, Bergen, Norway, nORAD/FAO Project — GCP/INT/730/NOR. Takasuka, A., Oozeki, Y., Aoki, I., 2007. Optimal growth temperature hypothesis: why do anchovy flourish and sardine collapse or vice versa under the same ocean regime? Can. J. Fish. Aquat. Sci. 64 (5), 768–776. Tejera, A., García-Weil, L., Heywood, K., Cantón-Garbín, M., 2002. Observations of oceanic mesoscale features and variability in the Canary Islands area from ERS-1 altimeter data, satellite infrared imagery and hydrographic measurements. Int. J. Remote Sens. 23 (22), 4897–4916.

Theilacker, G.H., 1980. Changes in body measurements of larval northern anchovy, Engraulis mordax, and other fishes due to handling and preservation. Fish. Bull. 78 (3), 685–692. Tolimieri, N., Jeffs, A., Montgomery, J.C., 2000. Ambient sound as a cue for navigation by the pelagic larvae of reef fishes. Mar. Ecol. Prog. Ser. 207, 219–224. Urtizberea, A., Fiksen, Ø., Folkvord, A., Irigoien, X., 2008. Modelling growth of larval anchovies including diel feeding patterns, temperature and body size. J. Plankton Res. 30 (12), 1369–1383. Wooster, W.S., Bakun, A., McLain, D.R., 1976. The seasonal upwelling cycle along the eastern boundary of the North Atlantic. J. Mar. Res. 34, 131–140. Yebra, L., Hernández-León, S., Almeida, C., Bécognée, P., Rodríguez, J.M., 2004. The effect of upwelling filaments and island-induced eddies on indices of feeding, respiration and growth in copepods. Prog. Oceanogr. 62 (2–4), 151–169.