The successful persistence of a fish population is ultimately dependent on life cycle and history closure, i.e., connectivity between each life stage, spawning adults, eggs, larvae, juveniles on nurseries and recruitment of juveniles back to the adult population (Petitgas et al., 2013). In many fish populations, adults will migrate “up stream” of suitable juvenile nursery areas so that eggs and/or larvae will be delivered to the nurseries in a suitable timeframe (Parrish et al., 1981). This window is often dictated by the development time of the early life history stages and the conditioning of a nursery area prior to the arrival of the juveniles. The drift of the eggs and/or larvae is often very influential on the success of supply of juveniles to nurseries and thereby linked to the recruitment success of the population (van der Veer et al., 2015). Life cycle closure can be characterized by the “migration triangle” (Harden-Jones, 1968), however, this framework does not have a provision for “retention of larvae” as presented by Iles and Sinclair (1982). The first allows the offspring to drift with currents away from spawning grounds toward their nurseries. The second retains the offspring in the vicinity of the spawning grounds, at least for a while, and so provide an imprinting mechanism for the return to specific spawning areas (philopatry, natal fidelity) (Sinclair, 1988). Both dispersal mechanisms are assumed to benefit larval survival depending on the physical features of the location in question (Houde, 1989).

The origin of larval drift as a key process in population dynamics of fish was firstly presented by Hjort (1926). In 1900, Johan Hjort observed early cod (Gadus morhua L.) larvae drifting in currents over hundred kilometers away from spawning grounds. With the resulting strong year-classes Hjort assumed that this larval relocation is essential for elevated recruitment. He also described that an aberrant larval drift into unsuitable nurseries may increase larval mortality rates. In contrast, instead of larval drift, retention seems to be the key when larvae instead remain at their spawning areas (Bakun, 1996). Larval retention is defined as the contrast to larval drift since larvae are kept in a certain location (Cowen et al., 2003). However, the application of the word retention is vague because it is used in several contexts. The term has been used to mean the larvae remain in one particular area (Swearer et al., 1999). In other ecosystems like important upwelling ecosystems (García-Reyes et al., 2015), larval retention is regarded as a process keeping the larvae on the shelf even though larvae can be transported alongshore (Churchill et al., 2011). Alternatively, larvae can also be retained in mesoscale eddies, even though non-stationary eddies may transport larvae into other locations (Condie and Condie, 2016).

Along the Norwegian coast, especially on the wider bank areas on the continental shelf, stationary and mobile mesoscale eddies form (Sætre, 1999). The stationary eddies are assumed to play a fundamental role in the successful recruitment of Norwegian spring-spawning (NSS) herring (Clupea harengus L.). Here larvae are suggested to be retained together with their prey on the banks (Sætre et al., 2002). This principle forms the basis of the hypothesis that retention of larval herring is dominant over the first few months of larval life (Sinclair and Power, 2015). However, there are opposing observations and hypotheses. For example, in years where early larval stages of NSS herring are exposed to a strong northward displacement this may result in strong years-classes, as was described for the period 1959–1965 (Dragesund, 1970). In other years during this period, where larvae were retained and concentrated in specific areas, only weak year-classes occurred. The northward advection theory is further supported by recruitment success being positively linked to early spawning within a year (Husebø et al., 2009; Vikebø et al., 2010; Slotte et al., 2019) as well as to the intensity of the Norwegian Coastal Current (NCC) (Skagseth et al., 2015).

Over the last three decades, the use of biophysical modeling to identify drift patterns of marine larvae has increased more than fivefold (Swearer et al., 2019). Biophysical modeling allows particle tracking that provides information about larval dispersal which previously had to be the subject of complex and challenging field studies. Tracking NSS herring larvae in drift simulations indicated that most larvae drift was within the NCC, resulting in their arrival on their presumed nurseries in the Barents Sea plausible (Vikebø et al., 2010). However, parametrisations in drift models involve assumptions of presumed positions and times of the year a larva may have occurred in the field. To counteract this shortage, survey observations can provide a meaningful add-in to refine drift modeling outcomes (Espeland et al., 2015; Kvile et al., 2017).

The main objective of this work was to generate larval drift indices from survey observations that can be adapted to similar types of drift studies worldwide. In addition, as there is obviously some debate as to whether retention or drift results in a strong year class, further investigation is required. This clarification is especially important for oceanic, high-latitude species where the spawning is closely linked to the pulsed production of suitable prey (early copepod stages) (Sundby et al., 2016). We use NSS herring as model species and by combining information from field surveys within a biophysical model we simulated back- and forward drift of NSS herring larvae between 1993 and 2016. This period covers some successful recruitment years, but especially after 2004, where a strong environmental change was observed, there were 10 years where recruitment was particularly low (Tiedemann et al., 2020). From these simulations, larval drift indices were estimated and related to a recruitment index used in a Bayesian framework to link drift strength and direction with differing levels of recruitment (Churchill et al., 2011). We address three main questions: Which dispersal pattern dominates during the drift phase of NSS herring larvae in the years between 1993 and 2016? Which dispersal patterns can be linked to the different levels of recruitment? Do and how do the larval drift indices relate with NSS herring recruitment success?

To answer these questions field-based data were combined within advanced drift simulations. The simulations utilized a hydrodynamic dataset which allowed both back- and forward drift simulation to identify larval drift patterns for the years 1993–2016. Every effort was made to base the study on recommended practices for modeling physical–biological interactions (North et al., 2009). Any constraints in the setup of the drift simulations are further addressed in the discussion.

Successful recruitment underlies several episodic and subtle mortality events in early life stages of fish (Houde, 1989). These events can involve very high mortality rates where even small perturbations in environmental conditions can have tremendous effects on the survival of the resulting brood (Houde, 1987). Larval dispersal is important as it can involve advection to unfavorable environments where larvae may encounter high predation risks and/or low food supply (Hjort, 1926). On the other hand, drift along routes with optimal environmental conditions may enhance larval survival. Knowing about the larval drift patterns that may result in various years of recruitment success is hence an important endeavor to better understand population dynamics and thereby enhance our abilities to manage harvested stocks. Consequently, in short term projections the strength of the Norwegian Costal Current (NCC) should help defining periods with lowered or raised recruitment potential (Skagseth et al., 2015). Such projections can be refined with dedicated survey observations and the generation of larval drift indices.

Increasing larval survival rates is dependent on ambient environmental conditions. Here, the adults determine where and when they spawn and thereby ultimately define the conditions their offspring will grow up in Parrish et al. (1981) and Slotte and Fiksen (2000). The spatial limitation, however, is that herring spawning is limited to suitable substrata with appropriate bottom currents (Runnström, 1941; Maravelias et al., 2000). In most of the years, recruitment is low for many herring stocks in the Atlantic and a few exceptional years allow stocks to increase significantly (Trochta et al., 2020). A few weeks in early spring after mature NSS herring have released eggs to the sea floor, larvae will inevitably occur under conditions in which they encounter variable chances of survival. As main transport route, the flow regime regulated by the NCC allows larvae to drift either northwards (Vikebø et al., 2010; Stenevik et al., 2012) or encounter stationary eddies which retain larvae on their way toward the Barents Sea (Sætre et al., 2002). When the NCC is strong, larvae can be carried quickly from spawning grounds northwards (Vikebø et al., 2010). An examination of NSS herring year-class strength and the physical conditions in the NCC (e.g., temperature anomalies, current speed and prevailing wind conditions) revealed that a strong NCC coincides with years in which NSS herring had exceptionally high recruitment success (Skagseth et al., 2015). The findings support the positive relationship between the larval northward advection (strength of the NCC) and elevated recruitment, lending support to Hjort’s conclusion that a larval relocation of hundreds of kilometers forms the basis for recruitment success in high-latitudinal waters (Hjort, 1926). This view is also in line with Dragesund’s field observations that rapid transport of NSS herring larvae northwards promotes recruitment (Dragesund, 1970). Therefore, it seems that the frequency and the total mean surplus northward drift of larvae control the level of recruitment. This statement is in contradiction of the theory that larval retention in the first months of life may generally be crucial for herring population replenishment (Sinclair and Power, 2015). Especially with the output of the drift simulations in which only one year accounted for stronger retention then northward relocation it seems that the northward advection dominates for early NSS herring larvae. Additionally, the three years in which a stronger southward drift component was observed signify only weak recruitment years.

The NCC is obviously the major transport path for NSS herring larvae to transit from their spawning grounds along the west coast of Norway to their main nursery area in the Barents Sea, but why would the speed of transport be of importance? As temperature rises in early spring the Norwegian Sea becomes more and more accessible for piscivorous feeding during the spring bloom. Juvenile saithe (Pollachius virens), a potential predator of NSS herring larvae, starts schooling at the Norwegian coast in late spring (Husebø et al., 2009). If this schooling coincides with the presence of NSS herring larvae it may lead to an elevated predation on the larvae. To compensate for any possible top-down control by saithe or any other species like Northeast Atlantic mackerel (Scomber scombrus) (Skaret et al., 2015) or Atlantic salmon (Salmo salar) (Utne et al., 2020) it would therefore be beneficial for larvae to escape high predation risks through a rapid transport away from predator aggregations. In this regard, a combination of early spawning by NSS herring and rapid northward transport could combine to enhance larval survival. The earlier drift would enhance the chances of escaping predation but could be disadvantageous due to a lower potential food supply in early spring (Husebø et al., 2009; Slotte et al., 2019). However, the drift speed only seems important in the years before 2004. Most of the years after 2004, coincide with predominant strong northward drift components that failed to produce strong recruitment. Since 2004 regimes seem to have changed in the study area where larvae are susceptible to a fast temperature shift during early development (Tiedemann et al., 2020). These altered environmental conditions may also have affected the timing of plankton blooming along the drift route, which normally is delayed toward the north along the Norwegian coast, and resulted in a mis-match with available prey for the larvae (Vikebø et al., 2012). In fact, general shifts in seasonality of Calanus species have been demonstrated over the study period in the Norwegian Sea area, in both coastal and oceanic waters (Strand et al., 2020). Another explanation is that there has been a more general decline of zooplankton production along the entire drift route of the NSS herring larvae as suggested by Toresen et al. (2019). Support for this hypothesis is found in a study looking at the production of Calanus species along the Svinøy section crossing the most important NSS herring spawning grounds, suggesting a 50% decrease for the most oceanic stations, and as much as an 81% decrease for the stations closer to the coast over the period 2000–2011 (Dupont et al., 2017). On the other hand, one cannot exclude the potential of increased predation especially by Northeast Atlantic mackerel that has had a significant northward expansion (Nøttestad et al., 2016), which may overlap and prey directly on herring larvae in the late drift period (Skaret et al., 2015; Allen et al., 2021).

By investigating not only the role of the zonal transport of larvae the model outputs indicate that when less larvae are advected inshore it coincides with heightened recruitment. Nursery grounds in fjords along the Norwegian coast are assumed minor compared to the main nursery grounds in the Barents Sea (Dragesund, 1970; Holst and Slotte, 1998). A drift along and not toward the coast would therefore allow a further transport within the main routes of the NCC. If larvae, on the other hand, would be transported coastwards it may increase chances of larval retention at the coastline or inside the fjords. The drift simulations in this regard could indicate that only a few larvae that were actually collected during the surveys would have been transported beyond the continental shelf that may be low in productivity and may negatively affect larval survival (Norcross and Shaw, 1984). However, the offshore larval transport is area specific. For instance, in areas where productive water masses are mainly observed along regions affected by coastal upwelling provide necessary feeding grounds for offspring (Allain et al., 2007; Tiedemann and Brehmer, 2017). It is hence obvious that in these areas larval retention is essential for survival. Such regions can be observed elsewhere in which fish larvae are suggested to make use of the flow regime to retain themselves onshore (Deschepper et al., 2020; Xing et al., 2020). In the Norwegian Sea, however, the flow regime of the NCC prohibits a strong transport off the continental shelf. Only some currents and eddies that form to the west of Lofoten seem to allow a larval offshore transport well away from shore, but whether this mechanism affects NSS herring recruitment to a significant extent still needs to be investigated.

A valuable add-in to the drift simulations was the actual position, the time, the total length as proxy for age and the larval abundance information from the surveys. With these data, it was possible to produce indices based on real observed spatial-temporal data and indicated years of enhanced larval retention or northward advection. However, even though combining the field observations that provide precision in spatial-temporal dimensions the simulations are marked simplifications of the real environment. The strongest bias may come from the resolution of the hydrographical model used here by probably underestimating small scale processes such as meanders and smaller scale eddies (North et al., 2009; Soufflet et al., 2016). The coarse grid was a choice that could cover the time series examined, the suitability within Ichthyop, accessibility and computational cost. However, drift on scales less than twofold of resolution is cropped that leads to particle trajectories more regular than it would be in real life (Vikebø et al., 2011). This setting has certainly increased the underestimation of the particle spread and probably the overestimation of the branches of the NCC. These limitations would also explain the end positions of the backward simulations that should roughly represent the origin. Besides that most of the origins covered typical spawning areas especially off Møre and the Halten Bank a small fraction of the simulated positions were unrealistic, like the ones at depths >250 m. Thereby, the simulation provides a rough estimate of the larval vertical migration behavior. Further to this, the constant assumed growth rate maximizes uncertainties of the realized drift period until the larvae were sampled in the field. For instance, we used a growth rate of 0.2 mm based on Doyle (1977). However, in other studies a daily growth rate of 0.3–0.6 mm was observed for different larval sizes (Stenevik et al., 1996), whereas in another drift simulation a 0.5 mm daily growth rate was adopted (Vikebø et al., 2011). Hence, the destination as well as origin that we considered in this study should only be noted approximations. Another source of bias may be based on running the drift simulations without region specific mortality rates. In different areas of the study area, larval mortality is most likely variable that would change the frequency of observed drift. For a more complex model it would be therefore necessary to implement spatially resolved mortality rates probably based on predator presence and food availability. A new pilot study was accomplished to use a molecular approach to define predation pressure in the field on NSS herring larvae (Allen et al., 2021). Such work may allow to estimate region specific mortality rates to refine the larval drift model outcomes.

The complexity of fish recruitment variability makes it difficult to assemble all factors that determine recruitment (Houde, 1987). Beside the larval drift there are many other factors that may influence NSS herring recruitment (Garcia et al., 2020). It was beyond the scope of this investigation to test all possible controlling factors; we rather focused on an in-depth investigation into one factor that is a major driver. Investigating recruitment success requires a long time-series that would allow for complex models. Here, the 24 years (24 data points for recruitment success) limits the number of co-variables that can be used in a model, which led us to invoke the “One in ten rule” for linear regression type analyses (e.g., number of observations >10 times the degrees of freedom of the co-variables). However, by developing larval drift indices fitted on survey observations it was possible to link the rapid northward advection of larvae to heightened recruitment years between 1993 and 2016. A strong southward drift component is linked to weak recruitment years. The larval drift indices worked especially well with the years before 2005, since then the changes in the environment (e.g., zooplankton availability and temperature) seemed to have mainly driven the recruitment success (Toresen et al., 2019; Tiedemann et al., 2020). The outcome of this work is intercessional toward larval drift studies in which survey data should be used to refine and calibrate drift simulations. This approach is possible for all planktonic taxa enabling a more precise view on the real drift in planktonic organisms.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

MT, RN, OK, and AS conceived the study. RN and ES collected the data. MT processed and analyzed the data (mapping, drift simulation, and modeling). OK and AS contributed with project resources. MT wrote the manuscript with contributions from RN, OK, AS, and ES. All authors contributed to the article and approved the submitted version.

The 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.