Trends in larval densities, spawning timing, and site quality of capelin (Mallotus villosus) across bays and habitats in Newfoundland, Canada

Abstract

Capelin is a short-lived forage fish species that underwent a population collapse in 1991 on the Newfoundland and Labrador Shelf, Canada (Northwest Atlantic Fisheries Organization Divisions 2J3KL) and has not recovered. The collapsed stock is characterized by delayed spawning timing and low recruitment. As year-class strength is set early in life, long-term larval monitoring at an intertidal spawning site in Trinity Bay (TB) is used as a proxy for stock-wide recruitment. Capelin spawn at both intertidal and subtidal habitats; however, there is minimal larval information from bays with a high incidence of subtidal spawning. We aimed to (i) compare inter-annual trends in temperatures, timing of spawning, and annual larval densities between two northeastern bays [TB and Notre Dame Bay (NDB)] and (ii) compare habitat quality metrics and larval densities between subtidal and intertidal sites in NDB. The date of first spawning in TB and NDB was positively correlated, suggesting that years with delayed spawning are experienced shelf-wide. While larval density was lower in NDB compared to TB, inter-annual trends were similar. In NDB, larval densities at both intertidal and subtidal spawning habitats were similar within a year, but temperatures and the proportion of dead eggs were lower at subtidal sites. These habitat quality metrics, however, were not related to annual larval density. These findings improve our knowledge on sources of capelin productivity, including the potential importance of subtidal spawning to recruitment dynamics.

Introduction

Animals are adapted to select the timing and location of reproduction based on environmental conditions that maximize offspring survival and, thus, fitness. In particular, offspring survival in fish has been directly related to habitat conditions, such as temperature (Houde, 1989; Pepin, 1991) and salinity (Lee et al., 1981). Spawning habitat characteristics are especially relevant for survival in fish species with adherent eggs because their spawning site becomes the egg incubation site and, thus, directly determines the conditions offspring will experience in early life stages when mortality is the highest (Dahlberg, 1979). As the timing of spawning and site-specific temperature determine the timing of hatch, these factors can determine whether larvae emerge into favourable conditions (match–mismatch hypothesis; Cushing, 1974, 1990), thereby influencing larval survival.

Capelin is a small, short-lived (3–6 years) forage fish species that serves a critical role in marine ecosystems, transferring energy from lower trophic levels to many upper trophic level predators (Carscadden and Vilhjálmsson, 2002). Similar to other forage fish, capelin populations are characterized by “boom–bust” cycles where they naturally fluctuate between high and low abundances (Chavez et al., 2003; Anstead et al., 2021). Population collapses in the early 1990s were observed in all three commercially fished capelin stocks in waters around Iceland, in the Barents Sea, and on the Newfoundland and Labrador Continental Shelf (Carscadden et al., 2013). While there was subsequent recovery of the Iceland (Carscadden et al., 2013) and the Barents Sea (Gjøsæter et al., 2009) capelin stocks, there has been no documented recovery for the Newfoundland capelin stock [Northwest Atlantic Fisheries Organization (NAFO) Divisions 2J3KL], which has been in a low productivity phase for 30 years (Buren et al., 2019; Murphy et al., 2021). Since the collapse, persistent changes in Newfoundland capelin biology and behaviour have been observed, including delayed timing of spawning (by ∼3 weeks), faster immature growth and an increase in proportion of fish maturing at age 2 instead of at ages 3+, and decreased body condition for older ages (Nakashima, 1996; Carscadden et al., 1997; DFO, 2018). In particular, the delayed timing of spawning has been associated with lower stock productivity (Murphy et al., 2021) potentially due to the mismatch between larval emergence and peak larval prey availability as well as the reduced opportunity to gain critical mass prior to overwintering (Vilhjálmsson, 2002).

Throughout their circumpolar range, capelin populations in the North Pacific and Northwest Atlantic predominately spawn adherent eggs at intertidal sites (Templeman, 1948; Frank and Leggett, 1981a; Stergiou, 1989), while capelin in the Northeast Atlantic predominately spawn subtidally (Vilhjálmsson, 1994). The choice of intertidal or subtidal spawning habitat appears to be facultative in the Northwest Atlantic, as evidenced by the lack of genetic differentiation between individuals spawning within each habitat (Penton et al., 2014; Kenchington et al., 2015), and temperature-based shifts in spawning habitat use within years at the individual level (Davoren, 2013) and among years at the population level (Crook et al., 2017). Due to these temperature-based habitat shifts, it is expected that intertidal spawning sites will fall outside the preferred temperature range throughout the majority of the spawning season as climate change continues to warm the ocean, even if capelin spawning shifts back to the pre-1991 norm of mid- to late June (Davoren, 2013; Crook et al., 2017; Murphy et al., 2021). Indeed, fishers observed an increase in use of subtidal spawning habitats by capelin from 1994 to 1997 (Nakashima and Clark, 1999). With the continued low productivity of the Newfoundland capelin stock, understanding the contribution of specific habitats and bays to stock productivity, which is currently unknown, is important to identify areas of high productivity for stock conservation.

Larval densities at the Bellevue Beach (BB) index site in Trinity Bay (TB), Newfoundland (NAFO Division 3L) at 1–2 weeks post-hatch are a good predictor of age-2 capelin biomass in spring acoustic surveys 2 years later (Murphy et al., 2018), demonstrating that year-class strength is likely determined very early in life. Furthermore, the BB larval index is one of three parameters used in a forecast model that predicts capelin biomass in the spring acoustic survey for two management cycles (Lewis et al., 2019). The BB larval index is based on sampling at a site that consists of a large beach and two shallow (12–14 m) subtidal spawning sites (Figure 1). Although earlier research indicated that BB is an index beach on the northeast coast (Nakashima, 1996), the BB larval index may not represent coastwide trends in larval densities, especially in areas with a high incidence of subtidal spawning. Notre Dame Bay (NDB) is the only other bay in Newfoundland where subtidal spawning has been documented and monitored (Davoren et al., 2008). It is also unclear whether larval densities differ between intertidal and subtidal spawning habitats as habitat-specific larval densities have not been quantified.

For capelin, high-quality spawning habitat is characterized by suitable sediment grain size (2–25 mm) and temperature (2–12°C), as evidenced by high egg densities and high hatching success (Templeman, 1948; Carscadden et al., 1989; Nakashima and Wheeler, 2002; Penton and Davoren, 2013). One study concluded that subtidal spawning sites in coastal Newfoundland likely did not contribute to recruitment, based on a high proportion of dead and abnormal eggs (Nakashima and Wheeler, 2002). In contrast, other studies showed lower proportions of dead/abnormal eggs at subtidal relative to intertidal sites (Penton et al., 2012) and that recently hatched larvae from subtidal sites were in good condition (Penton and Davoren, 2008). While intertidal spawning sites tend to be warmer than subtidal sites, resulting in quicker embryo development, the wider temperature fluctuations and low water circulation (i.e. decreased oxygen and waste removal) may be responsible for higher proportions of dead eggs at these sites (Penton et al., 2012). Longer embryo development at subtidal sites, however, may increase the risk of egg predation by winter flounder (Pseudopleuronectes americanus; Frank and Leggett, 1984), juvenile herring (Clupea harengus; Hallfredsson and Pedersen, 2009), crabs (Mikkelsen and Pedersen, 2017), and cannibalism by adult capelin (Slotte et al., 2006; Bone and Davoren, 2018), which may result in decreased overall productivity of subtidal sites relative to intertidal sites. Therefore, although intertidal and subtidal spawning sites have similar egg densities (Penton et al., 2012), these habitats may differ in larval densities.

The Newfoundland capelin stock has been in a critical zone since 1991 (except in 2013 and 2014; DFO 2023). The importance of the early life history stage to stock dynamics is well researched in the stock area (Frank and Leggett, 1981a; Leggett et al., 1984; Dalley et al., 2002; Nakashima and Mowbray, 2014; Murphy et al., 2018; Lewis et al., 2019), but the majority of this research is centred in Division 3L and at one spawning habitat type (i.e. intertidal). To identify areas of high productivity to help understand stock dynamics, the aims of this study are to (i) compare inshore temperatures, timing of intertidal spawning, and annual larval densities in two northeastern Newfoundland bays (TB in NAFO Division 3L and NDB in NAFO Division 3K) and (ii) compare intertidal and subtidal habitats within NDB over multiple years using a variety of habitat quality metrics (i.e. temperature, egg density, proportion of dead eggs, and larval densities). We hypothesize that (i) interannual variations in inshore temperatures, timing of spawning, and larval densities are consistent between the two northeastern bays, which would explain why the larval index from TB is correlated with the lagged biomass of age-2 capelin offshore and (ii) habitat quality metrics are lower at subtidal relative to intertidal habitats as a result of longer incubation times and higher egg predation. Quantifying larval densities across spawning habitats and bays will help to evaluate which spawning habitat and bay may be contributing more to capelin stock productivity, which will provide important information on stock dynamics for the assessment of the stock.

Methods

Study design

Research conducted was in adherence with the Canadian Council of Animal Care (Protocols: F04-031, F08-022, F12-020, F16-017, and F20-017). Intertidal and subtidal spawning habitats were monitored in two northeastern bays: NDB (NAFO Division 3K) and TB (NAFO Division 3L) (Figure 1). Both bays have been subject to long-term monitoring programmes conducted independently by Fisheries and Oceans Canada (TB, e.g. Nakashima, 1996; DFO, 2018; Murphy et al., 2018) and the University of Manitoba (NDB, e.g. Davoren et al., 2008; Crook et al., 2017), resulting in a >20-year dataset (2004–2021) of annual timing of spawning and temperature in both bays. In NDB, spatially distinct intertidal (n = 5; site names: Shalloway, Anchor Brook, Lumsden, Mussel Shells, and Capelin Cove) and subtidal spawning sites (n = 4; 15–40 m; site names: Gull I, II, and III and Turr; Figure 1) were monitored annually since 2004. Site-specific emergent larval densities, temperature, egg densities, and the proportion of dead eggs were collected at these sites in NDB over two shorter time series (2004–2006 and 2018–2021). In TB, the BB area has been monitored for emergent larval densities since 2001 (Figure 1; Nakashima, 1996; Nakashima and Taggart, 2002; DFO, 2021), where surface plankton tows are conducted at five fixed sites in the BB nearshore area (Nakashima and Mowbray, 2014). The BB area is composed of one large spawning beach, a few smaller spawning beaches, and two subtidal spawning sites (Nakashima and Wheeler, 2002). As the large beach is persistently used annually, this is the only intertidal site monitored within TB. The subtidal sites are spatially distinct but nearby (<4 km) the spawning beaches and are in ∼12 m (site 61) and ∼14 m (site 65) water depths (Figure 1), which preclude site-specific emergent larval densities. While all sites in both bays are similar in sediment composition (0.5–25 mm; Nakashima and Wheeler, 2002; Penton and Davoren, 2012), more intertidal and subtidal sites are monitored each year in NDB compared to TB due to variable use of habitat and sites by capelin in NDB (Crook et al., 2017).

Bay comparison (objective 1)

Temperature

Temperature loggers (Minilog-ii-T-351099 Vemco thermographs, Innovasea, Bainbridge Island, WA, USA; Hobo Water Temp Pro V2, Hoskin Scientific, Burnaby, Canada; Star Oddi DST CTD, Gardabaer, Iceland) were moored ∼1 m from the seabed at regularly monitored spawning sites in NDB and TB. Moorings were deployed prior to spawning in June (BB) or early July (NDB) and were removed mid-August (NDB) or September (BB) once larval emergence was complete.

Temperature data were summarized at one regularly used intertidal site in each bay [BB in TB, Lumsden (site 3; Figure 1) in NDB] during July in two ways. First, to compare the temperature between bays during the period when capelin typically arrived inshore and were selecting a spawning habitat (i.e. early July), hourly temperature recordings at each site were averaged across each day and then the daily mean was averaged across the first 7-d period in early July when there were consistent daily recordings across years (i.e. 4–10 July). Second, to compare intertidal habitat suitability throughout July between bays, mean daily temperatures at each intertidal site in each bay were averaged across years for each day in July (4–31 July).

Timing of spawning

Timing of spawning was defined as the date of first spawning in each year. For NDB and TB, the first date of spawning was defined as the first day when capelin eggs were found adhered to the sediment at BB in TB and any regularly monitored intertidal site in NDB. In both bays, sediment at intertidal sites was examined visually for the presence of eggs by hand every 1–2 d (NDB: n = 5 sites; TB: n = 1 site). The timing of intertidal spawning was compared between NDB and TB using the ordinal date of first spawning determined at BB in TB and at the first intertidal site that spawning occurred in NDB.

In NDB, timing of spawning was also determined at subtidal sites to ensure an appropriate timing of subsequent egg and larval sampling (see below). To do this, sediments at subtidal sites were examined for the presence of eggs every 3–5 d (n = 4 sites). Three replicate sediment samples were collected on each day using a 30 cm² ponar grab system (Wildco, Yulee, FL, USA), following Crook et al. (2017). As sampling could not be completed as frequently at the subtidal relative to the intertidal spawning sites due to ship time limitations and inclement weather, developmental stages were quantified to estimate the date of first spawning (i.e. development stages I + II indicate spawning within the last 48 h; Fridgeirsson, 1976).

In NDB, the total number of monitoring days per year varied between 15 and 27 days for intertidal sites and 9 and 21 days for subtidal sites, and monitoring occurred between 17 June and 30 August (Table 1). In TB, the annual number of monitoring days was consistently longer (28–31 days) than in NDB, with monitoring occurring between 17 June and 1 September.

Larval density

In NDB, larval tows in each year were conducted at two sites with the highest egg densities (one site per habitat type). As intertidal and subtidal spawning sites were separated by ∼4–40 km, we considered these to be site-specific larval densities. In support, larval densities captured during tows at each site peaked on the estimated hatch date (Penton and Davoren, 2008) and larvae were small in size (3–6 mm), suggesting recent hatching (Penton and Davoren, 2008). Site-specific sampling occurred every 48 h until larvae in the samples became scarce or the field season ended. At intertidal sites, two tows were conducted ∼0.7–1.3 km parallel to the shore of the beach. At subtidal sites, three replicate tows were conducted in a triangle shape around the site in most years, but only a single tow was conducted in 2 years (2004 and 2006) due to ship time limitations.

In TB, five fixed sites in the BB area were sampled every 48 h, depending on the weather, until larvae in the samples became scarce. The five fixed sampling sites were nearby the main intertidal spawning site and close to two subtidal spawning sites (Figure 1). Due to the close proximity of intertidal and subtidal spawning sites (<4 km), larval densities from the BB area are from multiple sites/habitats combined (Nakashima and Mowbray, 2014).

Larval sampling in both bays started on the median hatch date, which was estimated by combining timing of spawning and published temperature-dependent development rates (Frank and Leggett, 1981a; Penton et al., 2012). Larval tows in both bays were conducted using methods described in Penton and Davoren (2008). In brief, surface tows were conducted by horizontally towing a pair of 0.44 m² cone nets (0.28 mm mesh) deployed on each side of the boat for 10 min at a speed of ∼4 km h⁻¹ (Penton and Davoren, 2008; Nakashima and Mowbray, 2014). One net did not have a codend but was necessary to evenly distribute the weight of the nets on the small 6–13 m vessels. A flowmeter (model 2030R, General Oceanics Inc., Miami, FL, USA) was affixed to the mouth of the net with the codend. Time and coordinates were recorded at the start and end of each larval tow. Following each tow, one net was washed down with filtered seawater to wash the contents into the codend. The codend contents were placed in a 0.27 mm mesh sieve to remove the excess seawater before being funnelled into a 1 L glass jar and preserved immediately with 95% ethanol (Fisher Scientific, Whitby, Canada) or buffered in 5% formalin (Fisher Scientific, Whitby, Canada) and seawater. To determine the volume (V) of water filtered through the larval tow net, the distance (d) travelled along with the radius of the net (r) were input into the equation

Emergent capelin larvae were enumerated by either counting all the larvae when the sample contained <500 individuals or counting a sub-sample using a beaker sub-sampling technique (Van Guelpen et al., 1982). Annual larval density (per m³; Supplementary Table S1) was calculated using the trapezoidal integration method (Nakashima and Mowbray, 2014):

where N is the estimated annual larvae abundance per m³, t is the ordinal date, n is the number of sampling days, and X (t) is the number of larvae per m³ averaged from the replicate larval tows conducted on day t. In TB, all five larval tows per day were averaged to determine daily mean BB larval densities whereas in NDB replicate tows were averaged for each site to determine daily mean site-specific larval densities.

Although spawning occurred at an intertidal site in NDB in 2004–2005, different gear was used to sample these larvae (i.e. plankton net towed at waist height along the beach; Nakashima and Slaney, 1999) and was not comparable with larval surface tows. Therefore, these larval densities were not included in this study. For 2019 only, when spawning and larval sampling occurred at both an intertidal and subtidal spawning site in NDB, the annual larval density for NDB was averaged across the two sites as they were similar.

Habitat comparison (objective 2)

As the same habitat quality metrics were not consistently collected in TB at both intertidal and subtidal spawning sites, Objective 2 was investigated using data collected in NDB only. Larval density data collected in NDB at intertidal and subtidal spawning sites were used in Objective 2 to investigate the differences in densities from each spawning habitat type and to relate to habitat quality metrics (i.e. temperature, egg densities, and % dead eggs).

Temperature

Hourly temperature recordings using data loggers (see above) at all regularly monitored intertidal (n = 5) and subtidal spawning sites (n = 4) in NDB were used to calculate a daily mean temperature. A pooled average of temperature during the incubation period at each site per year was determined by averaging the mean daily temperatures starting on the first day of spawning (i.e. the first presence of eggs at the site) and ending on the estimated median day of hatch for first spawning. During 2004, mean temperature during the incubation period at the intertidal site was determined by averaging a single daily water temperature measurement taken at waist height during site visits (Penton et al., 2012), which were found to be similar to the daily averages of logger-based hourly recordings in subsequent years (2005, 2006; Penton et al., 2012).

Egg densities and proportions of dead eggs

Following standard methods (see Penton et al., 2012), egg/sediment core samples were collected from one intertidal and one subtidal spawning site in NDB (Table 1). At the intertidal site, three replicate core samples were collected using a 6.5-cm-internal-diameter steel corer at low, mid, and high tides 7 d after the first day of spawning. At the subtidal spawning site, three replicate core samples were collected, one from each of three bottom grab samples (0.3 m² Ponar Grab System). Each grab sample was carefully placed in a tray to preserve the layers and water was allowed to drain off before sampling using a 6.35 cm internal diameter corer. All core samples were preserved in 10% formalin solution buffered with sodium borate (Frank and Leggett, 1981b). In addition to core samples, three replicate samples of ∼100 eggs were collected, one from each of three bottom grab samples, and preserved in Stockard’s solution (% volume: 4 glacial acetic acid, Acros Organics, NJ, USA; 5 formaldehyde, 6 glycerin, and 85 water, Ricca Chemical Company, Arlington, TX, USA) every 48 h from intertidal spawning sites and every 3–5 d from subtidal spawning sites until eggs became scarce (i.e. majority of eggs hatched; Table 1).

Intertidal and subtidal core samples were placed in a 250 µm sieve and rinsed gently with freshwater to remove traces of formalin preservative before full submersion in 2% KOH solution (Sigma-Aldrich, Burlington, NJ, USA) for 24–36 h until adhesive eggs were fully detached from the sediment. Eggs were decanted and stored in 95% ethanol for a minimum of 24 h to harden. Egg density was estimated by subsampling using an alternative plankton splitter, where samples were divided until a subsample of ∼200–300 eggs was achieved and enumerated. The subsample count was extrapolated to estimate the total number of eggs in each core sample by multiplying by two to the nth power, where n was the number of splits performed. Total egg estimates were then divided by the volume of the sediment collected to determine egg densities (i.e. eggs/m³). Egg densities from intertidal sites were reported for low, mid, and high tides, while replicate samples from subtidal spawning sites were averaged and reported as mean ± SE (Table 1).

Egg samples preserved in Stockard’s solution were processed by examining a random sample of ∼50 eggs total from the three replicate scintillation vials to determine the number of dead eggs (i.e. cloudy and opaque eggs) and the number of eggs in different developmental stages, based on Frank and Leggett (1981a), which modified Fridgeirsson (1976). The proportion of dead eggs in each sample was determined along with the presence of eggs in early developmental stages (stages I + II), representing recent spawning, or eggs fertilized in the past 24–48 h. The proportion of monitoring days with these early-stage eggs was used as a proxy for the extent of spawning at intertidal and subtidal spawning sites.

Statistical analysis

For both objectives, Spearman’s rank correlations were used to explore relationships between variables where either a causal relationship was not being explored (objective 1) or data were not normally distributed and, thus, a non-parametric test was warranted (objective 2). As Spearman’s rank correlation assesses the rank of two variables and, thus, does not assume linearity, lines are not displayed on figures (Quinn and Keough, 2002). For all tests comparing means among groups, normality was assessed visually using histograms and quantile–quantile plots of the response variables and Shapiro–Wilk’s goodness of fit test. Egg density and larval density were log transformed to meet the assumptions of parametric tests. All proportion data (i.e. % dead eggs, proportion of days with early-stage eggs) were arcsine square root transformed to better meet the assumptions of parametric statistics prior to analysis (Quinn and Keough, 2002 ; Ahrens et al., 2017). All statistical analyses were performed in JMP Pro (v. 16), where α = 0.05 and all means are reported as ± SE.

For the bay comparison (objective 1), Spearman’s rank correlations were used to explore whether the rank order of temperature at intertidal spawning sites in early July (4–10 July), timing of first spawning, and annual larval density showed similar inter-annual variations between TB and NDB. To match the shorter data series collected in NDB with TB, the number of years included in each analysis varied among variables: temperature (2004–2021), intertidal spawning timing (2004–2021), and emergent larval density (2004–2006; 2018–2021). Two sample t-tests were used to explore differences between NDB and TB in the intertidal temperatures during July (4–31 July), timing of intertidal spawning, and annual larval density.

For the habitat comparison (objective 2), the influence of spawning habitat (intertidal and subtidal) on habitat quality metrics (i.e. temperature, egg density, and % dead eggs) at sites within NDB was examined using random effects analysis of variance (ANOVAs), where spawning habitat was a fixed factor and year was a random factor to control for inter-annual variation. As only four of the six years used in this analysis had both intertidal and subtidal spawning, and habitat quality metrics were only sampled within both habitats in three years, interactions between year and habitat could not be explored. Additionally, larval data were only collected from both spawning habitats in 1 year, so differences in annual larval density from the two spawning habitats were discussed but could not be statistically compared. To compare the extent of spawning in each habitat during each year, we used a two-sample t-test to compare the proportion of monitoring days with early-stage eggs at intertidal and subtidal habitats in NDB. Spearman’s rank correlations were also used to explore correlations between habitat quality metrics and annual larval density with data collected across 6 years (i.e. 2004–2006 and 2018–2021) and both habitats (i.e. intertidal and subtidal).

Results

Bay comparison (objective 1)

The dates of first intertidal spawning ranged from 3 July (2005) to 5 August (2017) in NDB and 13 June (2006) to 18 July (2009) in TB. Spawning occurred significantly earlier in TB than NDB (t_31.9= −6.35, p-value < 0.0001), whereby spawning occurred first in TB in all years and began 19 ± 2 d (range: 6–33 d) earlier than in NDB. Average daily intertidal temperatures in the month of July (4–31 July) for the years 2004–2021 were significantly higher in NDB (11.98°C ± 0.20) relative to TB (5.29°C ± 0.11; t_43.0=-−28.3, p-value < 0.0001; Figure 2).

The mean intertidal temperatures during early July (4–10 July) in NDB and TB were not significantly correlated (Spearman’s Rho ρ = 0.26, p-value = 0.122; Figure 3), indicating that the inshore coastal temperatures encountered by spawning capelin did not show similar inter-annual variation in these two bays. By contrast, the ordinal date of first spawning at BB in TB and the first intertidal site used in NDB were significantly correlated (ρ = 0.60, p-value = 0.009; Figure 4), indicating that delayed spawning in one bay coincided with delayed spawning in the other bay. Although higher annual larval density in one bay tended to coincide with higher density in the other bay, annual larval densities were not significantly correlated between bays (ρ = 0.80, p-value = 0.07; Figure 5). The mean annual larval density in TB (974.2 ± 392.9 m³) was significantly higher than NDB (55.9 ± 32.9 m³; t_8.5 = 4.06, p-value = 0.01).

Habitat comparison (objective 2)

In NDB, spawning occurred at both intertidal and subtidal habitats in four of the six years (i.e. 2004, 2005, 2018, and 2019; Table 1). In years when capelin used both habitats, intertidal spawning generally occurred 7–30 d before spawning at subtidal spawning habitats; however, in 2019, spawning occurred at the subtidal site 2 d prior to intertidal spawning. In 2018, spawning at the subtidal site occurred late in the season (9 August), and estimated hatching date was past the end of the field season. Therefore, the subtidal spawning site was not sampled during 2018. In 2020, there was no spawning at any monitored spawning sites; instead, capelin appeared to spawn subtidally just off a monitored intertidal site, as evidenced by eggs attached to detached algae that washed on the beach. The proportion of days with early-stage eggs (i.e. stages I + II), which we used as a proxy for the extent of site use, did not differ between intertidal (38.2 ± 6.8%) and subtidal spawning sites (51.1 ± 6.1%; t_6.4 = 1.41, p-value = 0.10).

Mean egg incubation temperatures were lower at subtidal relative to intertidal sites (F_1,136.1 = 69.01, p < 0.0001; Figure 6a), but year explained a lot (31.5%) of the variability in the temperature dataset. Egg densities differed significantly across habitats, where subtidal sites had significantly higher egg densities relative to intertidal sites (F_1,19.7 = 15.54, p = 0.0008; Figure 6b), but year again explained a large proportion of the variability (87.4%) in this dataset. There were significantly higher proportions of dead eggs at intertidal relative to subtidal sites (F_1,64.0 = 10.39, p = 0.002; Figure 6c), but unlike the other habitat quality metrics examined (i.e. temperature and egg density), there was no habitat-specific interannual variability in the proportion of dead eggs. Annual larval density was estimated to be higher from the subtidal site in 2019 (14.0 m³) and 2021 (28.2 m³) compared to the intertidal spawning site in 2018 (1.5 m³; Supplementary Table S1). In 2019, the only year when larvae were sampled at both intertidal and subtidal spawning habitats, annual larval densities were similar from the two spawning habitats (intertidal: 9.6 m³; subtidal: 14.0 m³; Supplementary Table S1; Figure 7). In 2019, larval emergence occurred first from the intertidal habitat and cooler temperatures at the subtidal site resulted in a greater number of incubation days (16–24 d) relative to the intertidal site (14–22 d). There were no significant correlations between annual larval density in NDB and egg density (ρ = −0.23, p-value = 0.87; Figure 8) or any of the other habitat quality metrics (i.e. temperature: ρ = –0.17, p-value = 0.70; proportion of dead eggs: ρ = −0.64, p-value = 0.11). Correlations were also explored among all habitat quality metrics but none were significant (data not shown).

Discussion

When comparing inter-annual trends between bays (objective 1), low and high larval density years tended to be similar in both TB and NDB, suggesting that the long-term annual monitoring in TB reflects the trends in larval densities for other northeastern bays in Newfoundland. This finding supports the continued use of the BB larval index in TB as a predictor of year-class strength for the 2J3KL capelin stock (Murphy et al., 2018; Lewis et al., 2019; DFO, 2021). Despite similar annual trends, larval densities from TB were an order of magnitude higher than NDB. This large difference in larval densities might, in part, reflect the site-specific nature of larval sampling in NDB. Indeed, there is greater geographic separation between intertidal and coastal subtidal spawning sites in NDB, whereas there are multiple intertidal and subtidal spawning sites located in the BB sampling area, resulting in BB larval density estimates incorporating larvae from numerous sites. Higher larval densities in BB compared to NDB spawning habitats may also indicate that spawning habitats in TB are higher quality, and, consequently, result in higher egg and larval survival relative to NDB. In support, although site-specific depths and sediment type were similar between bays, the mean daily July temperatures at BB from 2004 to 2021 were within the optimal range for high hatching success of capelin embryos (4–7°C; Penton and Davoren, 2012), unlike the July intertidal temperatures in NDB, which were much higher later in the month. In another study, the trends in late-stage capelin larval densities sampled in August were not always in sync between TB and the northern bays (White Bay and NDB), with similar larval densities in TB and northern bays in one year and higher larval densities in TB in another year (Shikon et al., 2019). The similar inter-annual trends in density of early-stage larvae between bays (this study) but different density trends in late-stage larvae among bays (Shikon et al., 2019) suggest that there are additional mortality processes occurring after larvae drift away from spawning sites. Therefore, the positive relationship between recently hatched larval densities (i.e. 1–2 weeks post-hatch) in TB and the 2-year lagged biomass of age-2 capelin surveyed offshore by Fisheries and Oceans Canada’s (DFO) spring acoustic survey (Murphy et al., 2018) suggests either that emergent larval densities from TB are large enough to influence shelf-wide recruitment or that the trend in annual larval density in TB typically mirrors patterns observed throughout northeastern Newfoundland, as supported by our study.

Capelin spawning has been delayed on average 3 weeks since the collapse of the stock in 1991 (Buren et al., 2019). This delayed spawning has been suggested as one of the drivers of the continued low productivity state of this stock (Murphy et al., 2021). Inter-annual trends in timing of first spawning were consistent between TB and NDB, but this coherence did not appear to result from coastal temperatures encountered upon arrival due to a lack of consistent temperature trends between bays. Instead, the predictors of spawning timing appear to be broadscale environmental variables like the North Atlantic Oscillation and the Newfoundland and Labrador Climate Index that affect capelin maturation, condition, feeding, and migration while offshore (Murphy et al., 2021). Indeed, previous studies have reported that variation in spawning timing likely results from interactions among length of maturing fish, spring sea ice conditions, and temperature and food availability (i.e. spring bloom) during gonadal development in offshore overwintering areas (Carscadden et al., 1997; Davoren et al., 2012; Buren et al., 2014; Murphy et al., 2021). The timing of spawning was also consistently earlier in TB relative to NDB, which is supported by DFO’s citizen science capelin beach spawning diary programme (Murphy et al., 2021; Murphy, 2022) and citizen scientist reports on ecapelin.ca (Johnson and Davoren, 2021); however, intertidal temperatures during early July when capelin typically arrive inshore were within the optimal range for spawning and offspring survival in both bays. Therefore, consistently earlier timing of spawning in TB does not appear to be related to warmer inshore conditions, but rather may be more related to other factors, such as migration distance and the subsequent timing of arrival in a bay. In support, mature capelin tagged in southerly bays on the east coast of Newfoundland during June–July have been shown to move against the Labrador Current into more northerly bays (Nakashima, 1992). Temperatures at inshore spawning sites instead seem to be more related to habitat selection, with fish shifting among suitable spawning sites based on site-specific temperature both at the individual level (Davoren, 2013) and population level (Crook et al., 2017). As delayed spawning has been observed for the past 30 years and no large year classes have been produced during this period, likely due to a mismatch between larvae and favourable conditions and less time to gain critical mass to survive overwintering (Murphy et al., 2021), continued delayed spawning that is consistent across coastal regions is expected to continue to have negative consequences for capelin recruitment, with reduced productivity from all northeastern bays and spawning habitats.

Capelin typically do not spawn at sites that fall outside the optimal temperature range for offspring survival (i.e. 2–12°C; Davoren, 2013; Crook et al., 2017), presumably to maximize fitness. The mean daily temperatures at intertidal spawning sites in NDB after 20 July, however, were frequently higher than the optimal range (i.e. >12°C), while those at the main intertidal spawning site in TB were consistently within the optimal range throughout July, possibly explaining the consistent use of BB in TB (DFO, 2021). In 2020, capelin arrived in NDB around 30 July, as indicated by the opening of the commercial fishery, but spawning did not occur at any of the monitored intertidal and subtidal spawning sites. Instead, nearly a week later, capelin eggs adhered to algae washed up onto a monitored intertidal site (Anchor Brook, site 2, Figure 1). Due to the late arrival of capelin, temperatures at intertidal spawning sites were higher than optimal, forcing the use of alternate “off-beach” spawning sites (i.e. spawning at a subtidal site contiguous with an intertidal site). The reason for the lack of use of monitored subtidal spawning sites is unknown as the temperatures were <12°C (10.1 ± 0.81°C) during the period when capelin were presumably moving northerly along the coast past those sites (24–30 July) from a southerly pre-spawning staging area (Davoren et al., 2006; Davoren, 2013). However, depending on when the sites were visited, temperatures might have fallen outside of the optimal temperature range as the temperatures at these monitored sites varied between 3.6 and 13.2°C during this week. To avoid overripening (i.e. higher egg mortality and reduced fertilization with increasing days past ovulation; de Gaudemar and Beall, 1998) during their transit back to subtidal sites, they may have simply spawned in suitable nearby habitat. The use of algae as a suitable spawning habitat by capelin rather than sediment has been observed in Placentia Bay, Newfoundland, which suggests greater flexibility in spawning habitat characteristics than previously thought (Bliss and Davoren, 2021).

When comparing habitat quality metrics between spawning habitats (objective 2), we found similar larval densities in the one year in NDB where emergent larvae were sampled from both intertidal and subtidal spawning habitats (2019), with the annual density from the subtidal site likely being a conservative estimate as sampling ceased before larvae became scarce. Furthermore, spawning and larval emergence occurred at subtidal habitats in NDB in most years of this study, and there was higher annual larval density from subtidal spawning sites in all of the examined years (2004–2006, 2018–2019, and 2021) compared to the years in which intertidal spawning occurred and larval density was determined (2018 and 2019). This is contrary to a previous study in TB that found no larval emergence from subtidal spawning sites, which suggested that the contribution of larvae from this habitat type to recruitment was negligible (Nakashima and Wheeler, 2002). There were also significant differences found in all metrics used as proxies for site-specific habitat quality between intertidal and subtidal habitats in NDB, where subtidal sites had higher egg densities, lower temperature, and lower % dead eggs. All of these findings do not support our prediction that the intertidal habitat represents higher habitat quality relative to the subtidal habitat, and suggest that subtidal spawning habitat is an important source of recruits. This result aligns with spawning habitat selection in other regions, as subtidal spawning is the primary mode of reproduction for capelin in the Northeast Atlantic (Sætre and Gjøsæter, 1975; Thors, 1981; Olafsdottir and Rose, 2013), where this habitat is critical for stock productivity (Carscadden et al., 2013). Understanding the contributions of subtidal spawning habitat to capelin recruitment in Newfoundland is critical as temperature shifts due to climate change are expected to increase the reliance of capelin on subtidal spawning habitat as intertidal habitat falls outside the preferred temperature range during the majority of the spawning season, especially in the more northern bays where spawning occurs later in the summer (i.e. July/August).

None of the habitat quality metrics examined in NDB (temperature, egg density, and % dead eggs) were related to larval density. Other studies have also found no relationship between egg densities and emergent larvae from intertidal sites (BB) following the capelin population collapse (Nakashima and Slaney, 1999; DFO, 2008). These findings support the lack of a stock–recruitment relationship for capelin (Carscadden et al., 2000; Buren et al., 2014; Murphy et al., 2018) and suggest that egg densities and our other metrics are not important in driving mortality processes between the incubation and immediately post-hatch periods (Houde, 1987, 1989). Instead, other processes occurring during this early life stage may explain the observed patterns in larval densities. For instance, egg quality rather than egg quantity may be important in determining hatching success and subsequent larval survival. Larger female fish produce more eggs of larger size (Penton and Davoren, 2013) and larger larvae at hatch (Chambers and Leggett, 1996). However, with capelin maturing at age 2+ rather than age 3+ in the post-collapse period, the spawning population is overall smaller in length compared to the 1980s (Nakashima, 1996; Carscadden et al., 2001; Buren et al., 2019), and since the fishery is primarily a roe fishery, larger females are targeted. The impact of the fishery on egg quality, larval survival, and recruitment is unknown, but the removal of large females before they spawn would be expected to have some impact on egg quality. Additionally, we did not measure egg predation, which may explain the lack of correlation between egg and larval densities at subtidal and intertidal sites. Finally, larval predator and prey densities immediately post-hatch might be important determinants of post-hatch larval densities, as was found previously for capelin (Frank and Leggett, 1982; Leggett et al., 1984). As none of the examined habitat quality metrics appeared to be strong drivers of post-hatch larval density in NDB, we recommend that future studies target the importance of bottom-up and top-down drivers of post-hatch capelin larval survival upon emergence from spawning sites.

Conclusions

Understanding the drivers of recruitment variability is a fundamental aim of fisheries science. Marine fish are characterized by high fecundity and high mortality in the pelagic larval stage, so even small changes in mortality rates in the first weeks and months of life can result in orders of magnitude differences in juvenile recruitment (Houde, 1987). For Newfoundland capelin, the accessibility of one of its spawning habitats (i.e. intertidal) has allowed for decades of extensive data collections on spawning timing, habitat use, and emergent larval densities (Frank and Leggett, 1981a,b; Nakashima and Slaney, 1999; Davoren et al., 2008; Murphy et al., 2018, 2021), which has led to important discoveries on drivers of recruitment variability (i.e. onshore wind events, prey and predator densities; Frank and Leggett, 1982; Leggett et al., 1984). However, after the collapse of the stock, changes in spawning timing and habitat use have resulted in known relationships between recruitment and the environment breaking down (Murphy et al., 2018; 2021) and low recruitment (DFO, 2021). By comparing habitat suitability, spawning timing, and larval densities across spawning habitats and bays, this study has improved our knowledge on sources of larval productivity (i.e. both intertidal and subtidal sites) and identified potential mechanisms for continued low stock productivity (e.g. unsuitably high temperatures, later spawning, and low larval densities). Consistent stock-wide trends in timing of spawning and larval densities support previous findings that capelin phenology is driven by shelf-wide biophysical drivers (Carscadden et al., 1997; Buren et al., 2014; Murphy et al., 2021).

As the climate changes, it can be expected that relationships between fish recruitment and the environment will change. For the Newfoundland capelin stock, as the climate warms, a further increase in subtidal spawning will likely occur as beaches become too warm (>12°C) for spawning earlier in the season; however, an increase in subtidal spawning after the stock collapse in the early 1990s is associated with continued low recruitment in this stock, perhaps due to longer incubation times and delayed emergence timing compared to intertidal sites. For the Iceland–Jan Mayen–Greenland capelin stock, collapses in the 2000s and late 2010s were associated with poor recruitment due to a climate-related shift in migration and distribution that occurred in 2002–2003 associated with warming seas around Iceland (Jensen et al., 2021). Negative relationships between fish recruitment and long-term environmental changes have also been shown in other fish species (e.g. cod, haddock, and herring) in the Northeast Atlantic (Brunel and Boucher, 2007), and similar climate forcing in the North/Northwest Atlantic (Greene et al., 2013; Bryndum-Buchholz et al., 2020). As forage fish play a critical linking role in marine food webs, understanding environmental factors that may impact recruitment during early life when mortality is the highest is an important step in advancing ecosystem-based management practices (Boldt et al., 2022).

Acknowledgement

We are indebted to the captain and crew of Lady Easton III for their assistance with fieldwork and sample collection in Notre Dame Bay, especially L. Easton, P. Goodyear, J. Chaulk, B. Abbott, and A. Chaulk. Thanks also to B. Squires, P. Lundrigan, D. Drover, and the DFO team for sample collection in Trinity Bay.

Author contributions

AT: investigation, formal analysis, visualization, writing—original draft

HM: conceptualization, methodology, resources, writing—review & editing, supervision, funding acquisition

PP: validation, investigation, writing—review & editing

GD: conceptualization, resources, writing—review & editing, supervision, funding acquisition

Competing interests

The authors declare that there are no competing or conflicting interests.

Funding

Funding for the research in Notre Dame Bay, Newfoundland, was provided by Natural Sciences and Engineering Research Council of Canada Discovery Grants [grant numbers 283212-07, 06290-2014, and 2019-06290] and an NSERC Strategic Grant [W. A. Montevecchi, Principal Investigator (PI), GKD, and three others (co-PIs)], along with annual University of Manitoba Faculty of Science Fieldwork Support Program grants (2012–2020) and a Coastal Restoration Fund grant (jointly funded by World Wildlife Fund—Canada and Fisheries and Oceans Canada) to GKD. Ship time funding was provided by NSERC Ship Time Grants [grant numbers 372594-09, 387231-10, 403438-2011, 419835-2012, 436824-2013, 453216-2014, 470195-2015, 486208-2016, 501154-2017, 55517-2018, and 486208-2019] to GKD, the NSERC Strategic Grant (2004–2006), and an International Polar Year grant (2007–2008). Fisheries and Oceans Canada funded all work in Trinity Bay, Newfoundland.

Data availability

The datasets generated for this study are available on reasonable request to the corresponding author.

References