Early life stages of fish experience high mortality rates, associated with a range of potential drivers, that lead to high losses prior to recruitment to the stock (“recruitment-processes”). Hjort’s (1914) seminal work emphasizes the need for first-feeding larvae to match prey immediately after the yolk sac is absorbed (“critical period”) and being transported to favorable nursery grounds (“aberrant drift”). This has since been further developed by utilizing new observational data and methods (Cushing, 1990; Houde, 2008; Ottersen et al., 2014). After about 100 years of research on various recruitment processes there is today an understanding that a multitude of processes on various spatiotemporal scales may impact recruitment in fish and that the relative importance of these processes varies in both time and space (Gallego et al., 2012; Hare, 2014).

The typical recruitment pattern is many poor to average years with only a few highly successful years that contribute significantly to the sustainability of fish stocks (Bogstad et al., 2016). The stocks produce vast amounts of eggs, much more than what can survive to recruitment. The high egg production often leads to recruitment being more dependent on the influence of environmental conditions on the survival of young fish as compared to the size of the adult population. Several environmental factors typically play an important role in the recruitment process, including spatiotemporally varying temperatures, dispersal, light conditions, turbulence, prey availability, and predation (Ottersen et al., 2014).

The Northeast Arctic (NEA) stock of Atlantic cod (Gadus morhua) is a commercially exploited fish species with a typical recruitment pattern of sporadically occurring years of very high recruitment. The NEA cod stock has been subjected to moderate fishing pressure in recent years that combined with strong recruitment in 2004 and 2005 has resulted in a high spawning stock biomass and increased age diversity (Kjesbu et al., 2014). Still, only a tiny fraction of cod eggs survives to reach recruitment (Bogstad et al., 2016)—on average 6 per 1 million eggs. The highest mortality rates occur during the egg and larval stages (Houde, 2008). Recent estimates indicate that NEA cod larvae mortality rates range from 0.06 to 0.08 day^–1 (Langangen et al., 2014a), which is on the low end of estimates previously reported by Sundby et al. (1989) of 0.08–0.15 day^–1. In the context of overall mortality from eggs to recruits, the largest inter-annual variability is thought to occur at the 0-group stage (developmental stage at settlement around the age of 6 months) (Langangen et al., 2014a). Considerable variability in natural mortality also occurs between juvenile stages and recruitment that may affect recruitment strength (Bogstad et al., 2016). On the other hand, density-dependent processes (e.g., competition for food and cannibalism) dampen the high variability in natural mortality rates, reducing the inter-annual variation in survivors from egg production through the early life stages until recruitment (Ottersen et al., 2014).

Sampling sufficient data at sea to obtain direct estimates of natural mortality of planktonic life stages is challenging because cohorts are dispersed by ocean currents and are therefore very difficult to track and sample repeatedly. Alternative and complimentary approaches for estimating mortality rates are therefore needed (e.g., Langangen et al., 2014b). By utilizing experimental and in situ data to parametrize coupled biophysical models of various complexity (e.g., Siddon et al., 2013; Daewel et al., 2015) one may investigate the relative importance of trophodynamics and variations in physical forcing at various scales (Peck and Hufnagel, 2012). For example, models combining ocean physics with nutrient-phytoplankton-zooplankton-detritus (NPZD) dynamics may quantify availability of suitable prey for larval fish. Combined with models of fish eggs and larvae, typically represented by individual-based lagrangian models (IBMs) of early life stage (ELS) (Daewel et al., 2011, 2015), the combined models allow assessment of bottom-up processes affecting growth and survival. The main prey items have pelagic free drift and exhibit diurnal vertical migration and can therefore be explicitly incorporated into the models. Predator presence is more difficult to address because adult migration behavior is less predictable than simple pelagic drift and because of the multitude of potential predators (but see end-to-end models e.g., Rose et al., 2015; Akimova et al., 2019; Daewel et al., 2019).

The adult NEA cod migrate from feeding grounds in the Barents Sea upstream along the Norwegian Coast to spawn during March through April with a peak in early April at multiple well-known spawning grounds (SGs, Sundby and Nakken, 2008; see Figure 1). NEA cod is a batch spawner that spawn pelagic eggs during a 2-month period from early March until late April. Pelagic eggs drift with the ocean currents toward nursery grounds in the Barents Sea and hatch after 2–3 weeks depending on temperature (Ottersen et al., 2014). Newly hatched larvae need sufficient food to cover metabolic costs and energy requirements for growth, which are additionally dependent on ambient temperature and larval size (Folkvord, 2005). The Norwegian continental shelf is a typical spring-bloom system where light conditions and structuring of the water column trigger phytoplankton and zooplankton production (Stenevik et al., 2007; Vikebø et al., 2019). NEA cod larvae and early juveniles are found to mainly feed upon various stages of Calanus finmarchicus from hatching in mid-March to mid-May until after metamorphosis around June (Sundby, 1995). The extended spawning time (ST) is believed to be a strategy to increase the chance of offspring to match the C. finmarchicus under a highly variable bloom phenology (Cushing, 1990; Solemdal, 1997). However, early spawning hampers larval visual feeding because of inadequate light conditions and low prey abundance.

Quantification of spatiotemporal variability in survival is relevant for integrated ecosystem assessments designed to consider impacts of human activities on the resilience in fish stocks (Langangen et al., 2017). A fish stock with wide-spread spawning locations and times is considered to more easily buffer changing environmental conditions (Hixon et al., 2014; but see Stige et al., 2017) and human activities, such as accidental oil spills (Carroll et al., 2018; Stige et al., 2018). In this context, an important question to be addressed is whether the portion of offspring successfully recruiting is systematically narrowed through match-mismatch leading to a reduction in the intrinsic buffer capacity to varying environmental conditions.

Here, we quantify how the match-mismatch between NEA larval cod and their main prey, C. finmarchicus, are affected by the location and timing of spawning by combining a coupled ocean- and NPZD-model with an IBM ELS for three contrasting years 1995, 2001, and 2002. We analyze the model output to address our hypothesis of a systematic lack of sufficient and suitable prey available for larvae resulting from early and late spawning, and that this lack of overlap is more pronounced at SGs to the south and north of the main SGs in the Lofoten.

We investigate individual growth in NEA cod larvae and pelagic juveniles as a function of SG and ST by providing stage-specific prey availability to an IBM of NEA cod ELS from the coupled open source SINMOD ocean and NPZD model thoroughly described in Alver et al. (2016). All data underlying this article will be shared on reasonable request to the corresponding author. We analyze 1995, 2001, and 2002 as these years represent contrasting conditions for NEA cod ELS development as compared to the average across 1983–2009 (Bogstad et al., 2016). In short, 1995 and 2002 had medium numbers of NEA cod eggs produced (about 102 and 103 trillion vs. a mean of 107 trillion), while 2001 only had about half of that. Subsequently, 1995 provided about 234 billion juveniles at the 0-group stage which is almost three times the long-term average of 85 billion, 2002 was close to the long-term average, and 2001 provided almost none.

SINMOD C. finmarchicus nauplii and copepodites are homogeneously distributed in the upper 40 and 50 m during dispersal, respectively, but only written to file as numbers per square meter surface for all 13 stages (eggs, nauplii I-VI and copepodites I-VI). When using this model output as prey for larvae, we therefore divide these numbers by 10 assuming that prey typically distribute across a 10 m water column and that larvae are able to stay within reach of their prey. We quantify which of the larvae that experience sufficient and suitable prey abundances to allow at least 90% food-unlimited growth during at least 90% of the time as a function of SG and ST. The reason why we choose to emphasize this is for two particular reasons; (i) past literature argues that the chance of survival through the early life stages is dependent on rapid growth (Houde, 1987), (ii) Folkvord (2005) reports that observed cod larvae at sea typically grow close to their size- and temperature-dependent potential, indicating that individuals unable to feed sufficiently to sustain such growth likely experience high mortality. Furthermore, we quantify ambient temperatures, larval length by the end of June and the number of days to reach 18 mm (same length used for evaluating fitness in Vikebø et al. (2007) as a function of SG and ST. The choice of 18 mm is because a major part of the individuals actually reaches this size by mid-summer, while at the same time acknowledging that mortality rates in general decrease with size. Furthermore, metamorphosis occur at lengths between 12 and 25–30 mm (Folkvord, 1997). Note that Daewel et al. (2011) used a size of 20 mm for a similar type of analysis. Finally, we investigate which areas and ages larvae grow below their growth potential and what temperature and prey conditions this occurs under. Note, that annual estimates of C. finmarchicus in the open ocean are based on annual survey measurements while abundance at the Norwegian shelf depends on cross-shelf transport (Opdal and Vikebø, 2015) or overwintering in the fjords and is reported to drop substantially around 2005 (Toresen et al., 2019). In general, the years in question here are considered medium to warm as compared to a long-term variability (Skagseth et al., 2015).

By running a coupled ocean model, a NPZD-model, and an IBM of NEA cod ELS, we systematically investigate food limitations and growth performance for individual cod larvae drifting along the Norwegian coast for three selected years. The years represent contrasting conditions, but more years are required to generalize our findings. The cod ELS were initiated as pelagic eggs at well-known spawning grounds and drift toward nursery areas in the Barents Sea for the years 1995, 2001, and 2002. The model results confirm the expected seasonal development of C. finmarchicus into larger size fractions but at decreasing concentrations that drift north along the coast. The model indicates only minor delay in C. finmarchicus stage development across latitudes. Furthermore, the duration of stages is longer than the inter-annual variation in onset of each stage, indicating consistent feeding conditions across years. However, the latter may be an underestimate of the variability in phenology, especially since the C. finmarchicus egg production is reported to be tightly connected to the timing of phytoplankton spring bloom (e.g., Melle and Skjoldal, 1998) and the onset of spring bloom is shown to vary by up to a month along the Norwegian coast (Vikebø et al., 2019). More in situ measurements are needed to investigate this (Alver et al., 2016).

Across these years, we find that larval growth is less food-limited for spawning occurring from Lofoten and farther north for the first half of the spawning season and much less so in 1995 and 2001 than in 2002. But if rapid growth is crucial for survival (Houde, 2008) then our model results indicate that both early and late spawned larvae are performing more poorly; early spawned larvae because of low temperatures and late spawned larvae because of less suitable prey. Stige et al. (2019) in fact show that early life size in NEA cod is a better predictor for recruitment than abundance, underlining the importance of fast growth. Though larval size at summer reflects the age of the individual, the time available for predators to truncate the lower size-specter of the larvae is shorter for larvae spawned in the middle of the spawning period, but progressively later with increasing latitude, shifting from late March in the south to mid-April in the north.

Notably, the temperatures and thereby the growth potential is larger farther south but limited by food. Hence, additional energy spent during spawning migration is not rewarded for the offspring in stronger growth if compared to the Lofoten. Sundby and Nakken (2008) show that spawning shifts southwards only in cold periods. We have only tested years during a relatively warm period. But if prey conditions were similar during a cold period then a spawning migration to the more southern SGs could result in stronger growth than SGs farther north because prey requirements would be proportionally lower. That said, SG4 stands out with particularly low temperatures and growth rates. Strand et al. (2019) have previously shown that plankton dispersal is particularly challenging in Vestfjorden where SG 4 is located. This is due to a combination of semi-permanent eddies, tidal pumping through narrow straits, steep surrounding mountains guiding winds and freshwater runoff making it difficult to replicate transport out of the fjord.

Daewel et al. (2011, 2015) show that North Sea cod living at the southern limit of Atlantic cod with ambient temperatures close to its thermal tolerance match poorly with prey during warm years because of rapid egg and larval development and subsequent elevated food requirements while prey phenology is less affected by temperature changes and is therefore not available this early in the season. This is not the case for NEA cod according to the results here, as the model indicates that prey become available already during the first half of March, well in time to support newly hatched larvae also resulting from the earliest part of the spawning period. Contrary, in warm years late spawned cod lack food both because of higher demands under elevated temperatures and because they are partly outgrown by their prey. Hence, our results show that larval ambient temperatures increase across the years investigated here, but that NEA cod ELS are not able to utilize this through enhanced growth. As suitable prey availability does not increase proportionally to elevated temperatures a higher fraction of the larvae grows below their potential in 2002 as compared to the other 2 years.

Across years we see that areas of strong currents are less able to sustain larval growth according to the temperature potential. These areas include the core of the NCC, the shelf edge from the Lofoten and northwards, and occasionally flow around the grand bank structure of Haltenbanken. The prey-limited larvae growth occurs consistently 1–1.5 months after hatching. This shows that C. finmarchicus egg production has ceased and either these stages are washed out of the area or they have grown to older stages. The relatively high-water temperatures to the south will cause this transition through stages to be quicker than farther north. Results here indicate that there is sufficient prey in the waters of these larvae but not in sizes suitable for their gape size. Hence, as we know that prey-limited growth is more frequent to southern and late spawned larvae, the prey must have outgrown the larvae and is no longer available.

The study demonstrates how bottom-up processes can result in variable growth conditions of larvae, likely affecting survival, due to a mismatch in time with suitable prey resources. If survival is rather heterogeneous as compared to homogeneous, then we can expect a more pronounced impact from changes in the environment. This could be either from natural variability in ambient conditions or anthropogenic impacts that negatively affect ELS growth or survival in parts of the habitat. The current well-known spawning behavior of NEA cod (Sundby and Nakken, 2008) suggests that only low-frequency variation in environmental conditions affect spawning intensity, e.g., large-scale latitudinal shift in spawning with decadal variations in temperature. A wide temporal and spatial distribution in spawning could therefore serve the purpose of buffering high-frequency environmental variability. However, we are unable to generalize our findings further toward conclusions on the benefits of having a wide temporal and spatial distribution of spawning activities to buffer natural variability given that our results are for a limited number of years. Note though that Stige et al. (2017) have already shown that a high mean age and stock size in NEA cod result in a high abundance and wide distribution of eggs, but that this does not prove to buffer environmental variability and thereby securing strong recruitment.

The results must be interpreted with care as the models have limited resolution, especially in the spatial dimension, and simplifications of processes and ecosystem structure. These simplifications are due to lack of understanding, limitations in computational power and data to parameterize models. Predation on fish ELS might be important in areas where predators naturally aggregate, e.g., sea bird colonies (Sandvik et al., 2016). There are currently no density-dependent effects in fish ELS due to e.g., competition for C. finmarchicus included in the model used here. Especially at times and areas where prey become limited, here shown to be for early and late spawned individuals and in the areas of strong currents, this may result in even worse feeding conditions for the larvae. Initial conditions for eggs are based on a limited set of survey samples and though we believe this provides representative egg distributions it still introduces uncertainties. Also, we are aware that ocean models typically have challenges representing upper ocean density structures and mixing and that this affect plankton dispersal (Strand et al., 2019). Finally, we have added a diurnal vertical migration but no horizontal swimming behavior of the NEA cod ELS. Fiksen et al. (2007) have shown how dispersal may be affected by larval and juvenile behavior. However, while NEA cod larvae and juvenile diurnal behavior seems to be a feature in the ELS (e.g., Kristiansen et al., 2014), horizontal swimming in a specific direction long enough to significantly affect dispersal is yet not observed.

This work contributes to the further development of model systems for the exploration of fish population responses in connection with the impacts and outcomes of natural processes, anthropogenic stressors and their cumulative impacts (e.g., oil-spills, Langangen et al., 2017; Carroll et al., 2018). From a management perspective, the findings of the present work highlight the need for further process oriented as well as empirically based investigations of known NEA cod spawning areas, that combined with model hypothesis testing, can lead to a better understanding of why certain periods and places along the Norwegian continental shelf seem to be particularly favorable for pelagic drift stages. It is also of critical importance to focus on the role and influence of climate warming toward changing our current understanding of the patterns and processes affecting NEA cod spawning.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

FV, OB, CE, and HF were responsible for formulating the hypothesis. FV, OB, CE, HF, and JJ prepared and performed most of the numerical calculations. FV, OB, CE, HF, JC, and ØL were responsible for synthesizing the literature. All authors have contributed to the discussions of hypothesis, methods and results, contributed to the work and approved it for publication.

OB was employed by the company SINTEF Ocean AS. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.