Latitudinally distinct stocks of Atlantic cod face fundamentally different biophysical challenges under on-going climate change

The reproductive success of marine ectotherms is especially vulnerable in warming oceans due to alterations in adult physiology, as well as embryonic and larval survival prospects. These vital responses may, however, differ considerably across the species' geographical distribution. Here we investigated the life history, focusing on reproductive ecology, of three spatially distant populations (stocks) of Atlantic cod ( Gadus morhua , Gadidae) (50– 80° N), in the Irish/Celtic Seas-English Channel Complex, North and Barents Seas, under past and projected climate. First, experimental tracking of spawning behaviour evidenced that the ovulation cycle is highly distressed at ≥9.6 ( ± 0.25)°C ( T up ). This knife-edge threshold resulted in erratic spawning frequencies, whereas vitellogenin sequestration remained unaffected, indicating endocrine rather than aerobic scope constraints. Cod in the Celtic Sea-English Channel are, therefore, expected to show critical stock depensation over the next decades as spawning grounds warm above T up , with Irish Sea cod subsequently at risk. Second, in the relatively cooler North Sea, the northward retraction of Calanus finmarchicus (Calanidae) and Para-Pseudocalanus spp. (Clausocalanidae) (1958– 2017) limit cod larvae feeding opportunities, particularly in the southernmost subarea. However, the contrasting increase in Calanus helgolandicus (Calanidae) does not counteract this negative effect, likely because cod larvae hatch ahead of its abundance peaks. Overfishing again comes as a twin effect. Third, in the still relatively cold Barents Sea, the sus-tainably harvested cod benefit from improved food conditions in the recent ice-free polar region but at the energetic cost of lengthier and faster spawning migrations. Consequently, under climate change local stocks are stressed by different mechanistic


| INTRODUC TI ON
The iconic Atlantic cod (Gadus morhua, Gadidae) (Jakobsson et al., 1994;Kurlansky, 2011) is managerially split into approximately twenty major stocks across the North Atlantic (Brander, 2005;Marteinsdottir & Rose, 2019).Local cod stocks are, therefore, subject to highly different living conditions at both sides of the North Atlantic (Brander, 2005;Sundby, 2000) and will be even more so under on-going anthropogenic climate change (IPCC, 2021) (Figure 1).Earlier synopses of this generally well-studied gadoid have provided important insights about the potential consequences of changing climate on its biology and harvestable surplus production; such integration efforts might help with deciphering local responses to climate change as these can be clearly stock dependent (Brander, 2019;Kjesbu et al., 2022).For instance, the impact of key environmental factors, such as temperature (T), on stock productivity is known to vary substantially across ocean basins (Drinkwater, 2005;Planque & Frédou, 1999).
Moreover, both the spawning cycle and early life-history survival of this species have in recent expert scorings been listed as particularly sensitive to warming waters (Kjesbu et al., 2022).Hence, these sensitivity attributes were presently hypothesized as 'the weak link in the chain'.To specifically highlight their local modes of action and consequences, we undertook an in-depth mechanistic investigation, encouraged by access to extremely rich data sources, including for the associated biophysical drivers.
Generally, warming directly impacts teleost physiology (Little et al., 2021), where the most sensitive stages, corresponding to the narrowest thermal windows of tolerance, refer to spawners and their embryos (Dahlke et al., 2020;Pörtner & Farrell, 2008).Indeed, reproductive failure at spawning accelerates at higher-than-optimal temperatures in several fish species, manifested as a shift in or even an inhibition of the spawning season and decline in reproductive success (Alix et al., 2020;Miranda et al., 2013).To describe and understand the fundamental mechanisms involved in the reproductive response (spawning performance) of adults facing variations in temperature, the oxygen-and capacity-limited thermal tolerance (OCLTT) theory offers an explanation (Pörtner et al., 2008(Pörtner et al., , 2017)).
The key principle within the OCLTT theory is that the scope for aerobic performance of a given physiological rate (or speed) follows a bell-shaped curve as a function of body temperature (Pörtner et al., 2008(Pörtner et al., , 2017)), in ectotherms reflecting ambient T (Levesque & Marshall, 2021).This thermal performance curve might be less clear at the molecular level (Pörtner et al., 2017).Anyhow, a performance optima (T opt ) is a mandatory criterion of the OCLTT, implying a lower tolerance outside this window, represented by lower (T lp ) and upper pejus T (T up ), and, ultimately, critical Ts with near zero performance (Pörtner et al., 2017;Pörtner & Farrell, 2008).However, early lifehistory stages can also be negatively influenced by ocean warming in indirect ways, even at T < < T up , where alternations in zooplankton (larval prey) biography stand out as one of the prime candidates for study (Beaugrand et al., 2003).Besides, the distribution of mid-and high-latitude species is typically displaced polewards during global warming (Morley et al., 2018;Poloczanska et al., 2013).Therefore, although T up is a sensible target for causal studies, the chains of ecophysiological reactions might be far more complicated.Furthermore, general principles developed for a species might be less relevant at the local population level, either because the detrimental effect attributed to climate change is much stronger, or, at other end of the scale, not yet detectable.
We targeted three Atlantic cod stocks experiencing close to the maximum range in spawning T in the Northeast Atlantic, that is cod located in the Irish/Celtic Seas-English Channel Complex (6-9°C), North (5-7°C) and Barents Seas (4-6°C), though conspecifics on the other side of the Atlantic, for example Northern cod, may spawn near 0°C (Brander, 2005).This spectrum of spatially diverging environmental conditions was used to contextualize the likely future performance of spawners and early life-history stages that govern local cod recruitment dynamics and, ultimately, their future productivity (Kjesbu et al., 2022;Payne et al., 2021) and management (Gullestad et al., 2020;Payne et al., 2021).Within the framework of on-going (Figure 1) and projected climate change (as specified below), our objectives were to (i) precisely and accurately define temperatures at which reproductive failure takes place in Atlantic cod and explore the underlying physiological modes of action, with particular reference to reproductive investment (oocyte growth) and egg release coinciding with natural 'warm waves' (i.e., seeing relatively extreme environmental temperatures over a limited period of time) as detected in experimental facilities (Figure 2); (ii) identify cod stocks likely to be subject to such detrimental thermal impacts and (iii) determine other specific factors that may limit cod future reproductive success in this marine realm, placing special emphasize on fine-scale changes in zooplankton biography but also any heightened migration costs.To do so, this article critically revisits earlier concepts and findings, supported by unique experimental and field time series data.Field females, collected off northern Norway (69-71° N) over twenty years (1986,1988,1989,(1999)(2000)(2001)(2003)(2004)(2005)(2006), were measured for total length (TL, in cm), whole-body weight (W, in g) and whole ovary weight (OW, in g).All of the presently considered specimens were Barents Sea cod (N = 796; 50 ≤ TL ≤ 135 cm) according to otolith zonal patterns (Rollefsen, 1934).Their potential fecundity (F P ) was determined in the laboratory from buffered formaldehydepreserved subsamples, for the first three study years by the gravimetric method (Kjesbu et al., 1998) and later on by advanced image analysis, that is the auto-diametric method (Thorsen & Kjesbu, 2001).

| MATERIAL S AND ME THODS
For exploratory consistency, 30 randomly selected individuals were analysed for F P per sampling year (Figure 3a; Table 2, Equation 1).The matching equation for W versus TL was given (Table 2, Equation 2).
Time of initiation of the reproductive cycle (vitellogenesis), that is TA B L E 2 Equations used to estimate reproductive investment in Barents Sea cod.Each equation-with its number, parameters and mathematical expression-is placed under either the pre-spawning (field), spawning (experimental) or model simulation study

EQ. Parameters Expression
Field females-Landed data on pre-spawning cod 1 Potential, pre-spawning fecundity (F P , in million) vs. total length (TL, in cm) and The corresponding daily increase in oocyte dry weight (∆VDW P , in μg day −1 ) and The corresponding figure for all oocytes prior to spawning (∆VDW P , overall , in g day −1 ) ∆VDW P , overall = ∆VDW P × F P Experimental females-tank data on spawning cod 8 Pre-spawning, total egg dry weight (TEDW P , in g), where F R (in million) is realized fecundity The corresponding complete investment (TEDW P + S , in g) and The corresponding investment attributed to spawning only (TEDW S , in g) and  LC > 250 μm, was set to typically begin on 8 October, that is including a latency period after autumn equinox (Kjesbu et al., 2010); cod starts maturing gonads for the upcoming spawning season when the night gets longer than the day (Davie et al., 2007;Kjesbu et al., 2010;Woodhead & Woodhead, 1965).Following estimation of the corresponding total length of this temperature-dependent, maturity (vitellogenic) period (VP, in days) (Kjesbu et al., 2010) (Table 2, Equation 5), the pre-spawning daily increase in vitellogenic oocyte dry weight, ∆VDW P (in μg day −1 ) could be indicated (Table 2, Equation 6).This equation implied that VDW MOD at LC = 250 μm, that is prior to any vitellogenin uptake, was considered negligible, indicated by extrapolation.Finally, the daily increase in dry weight for the entire vitellogenic mode, ∆VDW P , overall (in g day −1 ) could be established ( respectively.The whole catch was in each case transported to IMR facilities in Bergen, Norway, all specimens measured and PIT-tagged (using as always anaesthesia during all handling), and placed in two outdoor feeding tanks, each 30 m 3 (8-9°C; Dan-Ex 1562 dry pellets (www.biomar.com)).Their TL ranged from 50-60 cm, that is foreseen to become first-time (recruit) spawners in the following season (Marshall et al., 2006).As the reproductive cycle advanced nearer spawning, the sex was determined by stripping, or, if no egg or milt release could be detected, by inserting a soft, 2-mm thick plastic tube in the gonadal duct to remove a small piece of gonad tissue by suction ('catheterization') to be inspected microscopically (Kjesbu, Kryvi, & Norberg, 1996) but also, in the case of females, to predict time of start of spawning from the measured LC (Kjesbu et al., 2010).
In effect, 'cod pairs' (pair = one female and male) (Brawn, 1962) were in January/February randomly selected by length-stratified sampling from the larger group of individuals in the feeding (storage) tanks and moved into each of 10 separate, 20 m 3 indoor spawning chambers in a 15-m diameter circular tank (8-10°C, simulated light cycle at 60° N) (Kjesbu, Solemdal, et al., 1996).Ovarian catheterization, typically three times per individual, was applied to track oocyte development (Kjesbu, Kryvi, & Norberg, 1996;Thorsen & Kjesbu, 2001).After spawning, each individual was processed, considering in the further data handling only those females (N = 24) confirmed by dissection to show no remaining developing oocytes in their ovarian tissue (Appendix S1, Table S1).On the route to contrast ∆VDW S , overall with ∆VDW P , overall , a further list of algorithms was constructed ( prediction bands are shown (95% confidence intervals were too tight to be illustrated).Note that a few females apparently were in final oocyte maturation (FOM) (LC > 800 μm) (Kjesbu, Kryvi, et al., 1996).VP refers to the (pre-spawning) vitellogenic period.The vertical bar denotes the separation between VP and FOM Total length (cm) the length of the spawning period (SP, in days), (iv) ∆VDW S, overall (in g day −1 ) (Table 2, Equation 11).As above, these estimates were built on observed data sets (Appendix S1, Table S1).
In the further preparation for the below standardized model simulation, the three-season experiment was used to provide the number of batches spawned per female (N B ) as a function of the number of eggs spawned (realized fecundity; F R , in millions) (Table 2,   Equation 12).F R fitted better (r 2 = .592,p < .001) as explanatory variable than W (r 2 = .438,p < .001) or TL (r 2 = .388,p < .001).
The typical batch fecundity (F B , in thousand) was thereafter set as a function of F P and N B (Table 2, Equation 13).This expression required that F P could be taken as a proxy for F R , which seemed reasonable (Figure 3a).The varying number of eggs per batch was accounted for in the estimation of the overall, weighted mean egg diameter (WEDIAM) and dry weight (WEDW) (Kjesbu, Solemdal, et al., 1996;Ouellet et al., 2001).WEDIAM and WEDW were highly related (r 2 = .955,p < .001).However, in contrast with expectations (Chambers & Leggett, 1996;Marteinsdottir & Steinarsson, 1998), the present dataset (Appendix S1, Table S1) provided no evidence for a relationship between WEDW and TL or W, or combinations of these body metrics (p ≥ .365).We, therefore, decided to use the power function curve established for Norwegian Coastal cod South (hereafter Coastal cod South) (Appendix S1, List of Abbreviations and Terms) (Kjesbu, Solemdal, et al., 1996), supported by that this curve (Table 2, Equation 14) safely fell within the presently established 95% confidence band for Barents Sea cod.
To contrast ∆VDW P , overall and ∆VDW S , overall under, as far as possible, identical preconditions in the following calculation exercise, named for clarity the model simulation (Appendix S1, Table S2), TL was set to range from 50 to 105 cm, T held at 9°C and pre-spawning LC set at 800 μm.Consequently, LC = 800 μm functioned as a starting point for the subsequent oocyte growth during spawning ('single cohort') (Appendix S1, Table S2).As an alternative, oocyte growth was also set to start as early as at MOD ('averaged cohort'), copying in all other aspects the former calculations and, therefore, not dealt with further here (Table 2).However, an accurate calculation of the rate of incorporation of oocytic dry matter during spawning required first information on the spawning frequency (S f , in h −1 ) at T = 9°C, using in this respect the Q 10 law (Kjesbu, 1989;Schmidt-Nielsen, 1983) (Table 2, Equation 15), to thereafter provide the resulting spawning interval (SI, in h) (Table 2, Equation 16).Following these parametrizations, the typical increase in single oocyte dry weight within each SI, that is ∆VDW S (in μg day −1 ), could be given (Table 2, Equation 17).As VDW LC in Equation 17 refers to LC = 800 μm, ∆VDW S should mimic the subsequent uptake during final oocyte maturation (FOM) (Kjesbu, Kryvi, & Norberg, 1996) (Figure 3b).Finally, the resulting cohort (batch) investment, ∆VDW S, overall (in g day −1 ), was given by multiplying with F B (in thousand) (Table 2, Equation 18).

| Spawning frequency
Since the spawning frequency (S f ) of the present species is evidently influenced by ambient temperature (T), that is higher at warmer Ts (Kjesbu, 1989), this reproductive trait was selected to pinpoint any related optimal (T opt ) and upper pejus temperature (T up ) during the act of spawning (see Introduction).
The analysed 19 experimental females (Table 1) showed at least a few fertilized eggs per batch.This further selection out of the above-studied 24 females (Table 1; Appendix S1, Table S1) was necessary to be able technically to collate complete observational series of S f per female and season as a function of T.
The length of the spawning interval (SI, in h)-the inverse value of S f (see above)-formed the basis for this calculation exercise.
SI was given from detailed egg developmental rates at the actual T (Kjesbu, 1989;Kjesbu, Solemdal, et al., 1996); T was measured daily just below the sea surface, where the eggs in each spawning chamber (see above) were floating.In some cases T was interpolated, as specified in Appendix S1 (Figure S2).This approximation was feasible due to the high stability in T from one day to the next (but not over the whole season as such; Figure 2), except for in 2002 in connection with the pulsed production of relatively warmer water (warm wave), requiring closer tracking of the dayto-day changes in T (Appendix S1, Figure S2).Data on Coastal cod South (Kjesbu, 1989)-13 females in indoor tanks and 2 females in cooler, outdoor tanks-were added to this S f and body metrics dataset (Table 1).To further increase the analytic breadth, averaged SI per female-and thereby averaged S f per female-were also extracted for Gulf of St Lawrence (GStL) cod, held at 2 (N = 17) to 4-5°C (N = 21) (Ouellet et al., 2001) (Table 1).
To locate any T up from the complete, seasonal series on S f as function of T for each of the above-mentioned 19 experimental females of Barents Sea cod (Appendix S1, Table S1), they were firstly ranked by increasing max.T. Any trends were further illustrated by clarifying which of these individuals experienced common temperature conditions (N = 11) and which ones experienced a warm wave (N = 8).The specific location of T up was thereafter given by studying even more detailed tracking data, focusing on any abrupt irregularities in S f versus T for each female (Appendix S1, Figure S2).This analysis also consulted information on aquaculture (broodstock) cod showing a statistically significant drop in the overall fertilization rate for T > 9.6 (van der Meeren & Ivannikov, 2006), and finally, supplemented with dst-based information (Appendix S1, List of Abbreviations and Terms) on spawning T from several cod stocks in the Northeast Atlantic.So, the adopted range (±0.25°C) in T up resulted from carefully evaluating S f dynamics vs. max T or T and literature information but also considering any likely uncertainty in the actual ambient T in the tank, though excluding any measurement error in T (given from an oceanographic thermometer).This aggregated analysis on 19 females was followed by in-depth tracking of MEDW across the season for five females experiencing different maximum temperatures (Table 1), that is max.T reaching either 9.1 (N = 2) or 10.2-10.3°C(N = 3) (Appendix S1, Table S1).

| Embryos
The latter category referred to those females, which showed an extended overlap between spawning activity and the appearance of the warm wave (7, 8 and 8 out of 17, 21 and 13 egg batches, respectively).To measure MEDW, a total of 50 eggs per batch were flushed with distilled water and completely dried prior to any weighing (Kjesbu, 1989;Kjesbu, Solemdal, et al., 1996).The associated batch fecundity was displayed (Appendix S1, Figure S3), given by entering the mean egg diameter (based on 50 eggs) in a packing density formula and multiplying with the total volume of eggs in that particular batch (Kjesbu, 1989).

| Climate impacts on cod spawning ground persistence
Observed and modelled environmental temperatures (Table 1) were used to indicate the consequences of exceeding the upper pejus temperature (T up ).The associated uncertainty in the resulting T up (see Section 2.1.2) was implemented in a sensitivity analysis when assessing the implications for the likely fate of southern spawning grounds, that is in waters of likely, immediate interest.This research was supported by an in-depth review on 'Observed shifts in cod spawning grounds in the Northeast Atlantic' (Appendix S1).

| Fate of southern cod spawning grounds
The Regional Ocean Model System (ROMS) (Shchepetkin & McWilliams, 2005) was first run as a hindcast simulation for the recent decades with atmospheric forcing from the Coordinated Ocean Research Experiment CORE2 reanalysis 1958-2007 (Large & Yeager, 2009) and lateral boundary conditions from the Simple Ocean Data Assimilation (SODA) dataset (Carton & Giese, 2008) (Table 1).
Thereafter, the model was initialized and forced with output from the coupled Norwegian Earth System model NorESM1-M (Bentsen et al., 2013) for the Representative Concentration Pathway RCP4.5 scenario from 2006 to 2070 (Table 1).The model domain covers the North Atlantic-Arctic Ocean with an average resolution of about 10 km (i.e., AA10km grid, where AA = Atlantic-Arctic) in the Nordic and Barents Seas.Further details of the model setup and corresponding model evaluation are described in recent literature (Sandø et al., 2021).
Based on timeseries of observed and simulated temperature variability, two periods, 1965-70 and 2003-08, were chosen to represent cold and warm periods in recent decades, respectively (Sandø et al., 2020(Sandø et al., , 2021)).Spawning T for cod were taken from the bottom layer in the case where the depth is less than 100 m, and at 100 m where the bathymetry is even deeper.The projection of the 2060 s was based on a downscaling of one ensemble member from the NorESM-1 M RCP4.5 to be compared to the beginning of the same simulation, that is the 2010s, to minimize model biases.Within this scenario, suitable locations of cod spawning grounds in the southern distribution area were investigated with regard to T up .

Sea cod
This section firstly considered the whole North Sea and thereafter focused on southern, northwestern and northeastern subareas (further specified below) to acknowledge the likely existence of subpopulations of North Sea cod (Holmes et al., 2014;Wright et al., 2018Wright et al., , 2021) ) to investigate links between copepod (cod larval prey) biography and cod recruitment dynamics at varying spatial (and temporal) scales (Table 1).

| Variation in copepod abundance in the North Sea
The Continuous Plankton Recorder (CPR) datasets at https://doi.The abundances of the above-mentioned copepod categories were considered per North Sea subarea to be related to the performance (as specified below) of local populations of North Sea cod.So, for the copepod spp.time series to be directly comparable, we adopted the ICES division into subarea Southern (S), Northwestern (NW) and Viking4a (V4a, the adjacent Viking 20 in the Skagerrak was omitted) (ICES, 2017).

| Spawning stock biomass and recruits in the North Sea and subareas
An individual North Sea cod was set to reach sexual maturity at age 3 years (Nash et al., 2010).Hence, spawning stock biomass (SSB) forward lagged by 3 years (y + 3), SSB_yp3, was used as proxy for recruitment strength and related to the above-mentioned four category series on copepods (y).This correlation matrix was thereafter broadened by replacing SSB_yp3 with Recruitment (y + 1), that is Recruitment_y1.These cod stock metrics were extracted from ICES (ICES, 2015).Likewise, all current statements on stock-specific status refer to ICES (www.ices.dk).Next, SST was successively re- imr.no) (Figure 1c) was added for further insight (see above).This Hanstholm-Aberdeen-Utsira Transect is covered several times per year but to match as far as possible the spawning time of North Sea cod (January-April) (Brander, 2005), we selected T data from the month of March.

Sea cod
This exercise sought to quantify to which extent the degree the migration dynamics of Barents Sea cod has changed over recent decades, as measured by the directional displacement distance (D D ) and speed (U D ), split by body size category.

| Centre of gravity
These data derived from the 'Joint Norwegian-Russian Winter Survey' (February-March, 1981-2017) and the 'Joint Norwegian-Russian Ecosystem Survey' (August-September, 2004-2017), both surveys are detailed elsewhere (Jakobsen et al., 1997;Michalsen et al., 2011) (Table 1).The currently estimated centre of distribution (gravity) (CoG) averaged the number of cod per square nautical mile from standardized bottom trawl catches (Table 1).Latitude and longitude for each observation (catch) were converted into Cartesian coordinates.These computations in MATLAB R2016b resulted in the weighted average of CoG for small (30-59 cm in TL) and large cod (70-99 cm in TL), assumed to be sexually immature and mature, respectively.Individuals in the 60-69 cm size class were left out in order to keep these two maturation categories separate (Nash et al., 2010).
Only CTD stations taken in the vicinity of CoG were considered (Table 1), specified as ±0.5° in a latitudinal direction and ± 1° in a longitudinal direction.This grid design reflected the larger span in longitude (20-50° E) than latitude (71-77.5°N).The habitat depth was set at 100-150 m (Godø & Michalsen, 2000).Between 1981 and 1994 the CTD recorded every 5 m, from 1995 onwards every 1 m.The resulting number of CTD stations (1-17) per grid were, with few exceptions, higher during winter, typically 5, than during summer, typically 2-3.The grand mean across CTD stations within each grid was labelled as in situ (ambient) T (±SD).To compare with measures of environmental temperature, T in the Kola Transect (hereafter Kola T) (Figure 1b) in February, March, August and September were considered (Table 1).As winter CoG were 5-year smoothed, the same type of calculation exercise was applied on the four monthly resolved Kola T data sets (leaving the last 4 years unsmoothed).

| Displacement distance and directional speed
To better understand variation in D D and U D of large and small specimens over time, Kola T (see above), ice extent, SFA (suitable feeding area) and TSB (total stock biomass) were considered as 'drivers' (see the various time windows below) (Table 1).As the first step in this investigation, the ice extent was downloaded for the month of September (referring to the lowest values) (https://www.mosj.no/en/clima te/ocean/ sea-ice-exten t-baren ts-sea-fram-strait.html), the SFA series (bottom T > 0°C, 72-80° N, 20-50° E) (Kjesbu et al., 2014) updated with more recent figures (Randi B. Ingvaldsen, IMR), and the TSB series extracted from a recent assessment report (ICES, 2020).D D referred to the distance between the CoG in summer and the corresponding CoG in the following winter.For the larger cod this route reflected spawning migration (Sundby & Nakken, 2008).The distance between these two points was transferred to kilometres, accounting for Earth curvature (https://www.geeksforge eks.org/

| Additional data handling and statistical analyses
With reference to Section 2.1 (cod physiology) and Section 2.4 (Barents Sea cod migration), Systat 13 was used as the primary statistical software, that is eventually also SigmaPlot 14 in connection with publishable graphics (Inpixon, 2022).So, the given linear or power function regressions (Table 2) were from these packages.
S f and MFERT, being proportions, underwent arcsine transformation (Sokal & Rohlf, 1981)  With reference to Section 2.3.1 (North Sea copepod abundance) and Section 2.3.2 (North Sea copepod abundance in subareas), all visualization and analysis of the CPR data were performed using R (R Core Team, 2019) and various libraries.For the ocean-basin decadal visualization, all unique samples were assigned to the correct ICES statistical rectangle (www.ices.dk)based on recorded longitude and latitude for each sample using the 'mapplots' library (Gerritsen, 2018).To avoid analytic bias caused by single or few observations, only ICES grid cells (0.5° Latitude × 1.0 Longitude) with 5 or more observations per decade were used.All abundance data were ln +1 transformed.To ensure all years and months within one decade were weighted equally, independent of variations in temporal and spatial sampling effort, the mean abundance of each copepod category was first calculated based on grid cell, month, year and decade, then on grid cell, year and decade and finally on grid cell and decade.This aggregated dataset was then further used as input to perform surface interpolation ('automap' library) (Hiemstra et al., 2009).Each interpolation was limited to the area confined by drawing a polygon around the outmost valid observations, which resulted in different decadal spatial distributions.
In the subsequent subarea examination (i.e. by subarea S, NW and V4a)-to ensure observations being temporally and spatially comparable over the whole time series-only data occurring within unique ICES grid cells that contained at least 2 observations in the period April-August, as well as being sampled over at least 20 years, were considered.
With reference to Section 2.3.3 (North Sea cod recruitment), illustrations and correlation statistics (Spearman) were consistently made with the ggscatter function from the R ggpubr library (Kassambara, 2020).The R function 'loess' was used to fit subarea specific time series curves.

| Vitellogenin uptake
The power function (with b as exponent) for field-caught, experimental Barents Sea cod revealed a strong, positive effect of TL on the overall rate of oocytic dry weight (≈vitellogenin) uptake during spawning (∆VDW S , overall ) (r 2 = .60,p < .001,b = 4.82 (SE = ± 0.90)) (Figure 4a).An allometric effect was also noticed for pre-spawning fecundity as such (b = 3.59 (SE = ± 0.09)) (Figure 3a).∆VDW P , overall , supporting ovary growth prior to spawning (cf.pre-spawning reference curve), approximately equalled the modelled investment underlying the production of an egg batch (cf.single cohort) from final oocyte maturation (FOM) onwards (Figure 4a).However, apparently, two oocyte cohorts develop in succession (cf.averaged cohort) (Figure 4a).Thus, the next oocyte cohort quickly replaces the one entering FOM to be thereafter ovulated and released as eggs.Our modelled ∆VDW S , overall in large females (TL > 70 cm) were obvious underestimates (Figure 4a), despite assuming a positive effect of body size on egg dry weight (Table 2, Equation 14).

| Spawning frequency
Although the steepness of the ascending spawning frequency (S f ) curve was modest within increasing T (≈4 < T < ≈9°C), the variation in spawning interval length (SI) (p < .01)-andthereby in S f (Figure 4b; Figure 5, Appendix S1, Figure S1)-suddenly turned markedly higher above this thermal window (Figure 6a,b), though not statistically departing from the general SI level as such (p = .31)(Figure 6c).Arbitrarily changing Q 10 from 2 to 3 hardly had any impact on modelled S f (Figure 4b).No clear indications of differences in S f existed between the two cod stocks presently addressed, Barents Sea cod and Coastal cod South (Figure 4b).The inclusion of averaged S f for Gulf of St Lawrence (GStL) cod fell into this depiction at 2°C but only marginally at 4-5°C, but neither S f (p = .21)nor pre-spawning body condition changed significantly (p = .27)between these two temperature regimes (Figure 4b), consulting tabled data (Ouellet et al., 2001).As for GStL cod (p > .05),mean S f for Barents Sea cod were unaffected by pre-spawning body size (TL: p = .46,W: p = .32,8.5 < T mean < 9.1°C, N = 11).In the latter case, the more detailed picture showed that T ≥ 9.6 (±0.25)°C, even brief in nature, caused the normally highly rhythmic ovulatory cycle to turn erratic and remain so even if T thereafter decreased (Figure 5; Appendix S1, Figure S2).This adopted uncertainty in T up (Figure 5) was included in the subsequent field-related sensitivity analysis in Section 3.2.1.

| Embryos
The phenomenon of warm waves (Figure 2) did not statistically influence the subsequent mean fertilization rate of single egg batches (MFERT) (p = .53),though MFERT generally fell by a few percent points (Figure 6d).The corresponding, generally longer spawning interval (SI) (Figure 6a) lowered MFERT (p ≤ .03) in two out of eight cases; six females exhibited no such relationship (p ≥ .32)(Figure 6e).
There was no sign of that a warm wave affected mean egg dry weight (Figure 4c) or batch fecundity (Appendix S1, Figure S3).

| Fate of southern cod spawning grounds
Over the period from the cold 1960s to the warm 2000s, cod egg concentrations in the Irish/Celtic Sea-English Channel Complex undoubtedly declined, though dense aggregations were still seen farther north in the Irish Sea (Figure 7a,b; Appendix S1, Observed shifts in cod spawning grounds in the Northeast Atlantic).The southernmost spawning grounds seemingly dissipated in conjunction with the northward expansion of the 9.6°C isotherm (T up , see above) (Figure 7a,b).The inclusion of an uncertainty of ±0.25°C in the assessment of this threshold (Figure 5), resulted in a broadening of this boundary zone (Figure 7a,b).However, this sensitivity test showed markedly larger bands west of Ireland, that is an area lacking cod spawning grounds (Figure 7a,b).Our following projection pointed to extirpation of the spawning grounds in the Celtic Sea (and the adjacent Bristol Channel) and English Channel within a few decades (Figure 7a-c).The existence of spawning grounds in the Irish Sea was also apparently at risk (Figure 7a-c).

Sea cod
Except for farthest south in the North Sea, T in this ocean basin will most likely remain below T up at cod spawning grounds in the next halfcentury (Figure 7c).Historically, cod spawning appeared throughout most of the North Sea (Appendix S1, Observed shifts in cod spawning grounds in the Northeast Atlantic) but is today increasingly confined to the Northeast subarea, that is in ICES Viking4a (Figure 7b).Although an insignificant trend was seen for Para-Pseudocalanus spp.versus.SST (Figure 8c), its presence in the North Sea was F I G U R E 4 Cod reproductive physiology and spawning performance.(a) Observed and modelled typical daily rate of oocytic dry weight (≈vitellogenin) uptake (∆VDW overall ) at 9°C of Barents Sea cod in relation to total length.The field-experimental data curve (mean ± 95% confidence interval) was standardized by maturity phase (leading cohort = 800 μm) at initiation of spawning.In the model simulation (Q 10 = 2) three calculation paths were followed: pre-spawning, single and averaged oocyte cohort production, assuming in the latter two cases that the mean egg dry weight of a single batch increases with body size; (b) spawning frequency (S f ) regressed on the corresponding temperature (T) of experimental Barents Sea cod and coastal cod south, categorized by T regime-'warm wave', 'common' or 'wide'-where 'warm wave' (see main text) being further subdivided into 'at and after this event'.The earlier published Q 10 = 2 (Kjesbu, 1989) was supplemented by Q 10 = 3 to model the subsequent response in S f .Published mean S f data for Gulf of St Lawrence (GStL) cod (Ouellet et al., 2001) at 2 and 4-5°C were added for further contrast; and (c) trend in single batch mean egg dry weight during the spawning period (from first to last egg batch) of five experimental Barents Sea cod, where three of them encountered incidences of T ≥ 9.6°C (Appendix S1, Table S1) showing a more-or-less steady increase (Figure 1c).S3).Para-Pseudocalanus spp.and SSB_yp3 (p = .01)related positively but no such likely effect could be seen for Recruitment_yp1 (p = .07); Appendix S1, Table S3).North Sea cod SSB_yp3 largely mirrored C. finmarchicus abundance; our extension (1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) of the earlier 39-year time series  by Beaugrand et al. (2003) (Appendix S1, Table S4) showed a joint switch to increasing trends in the late 1990s (Figure 10a), though all these more recent data points were far below earlier values, that is located near the origin of this regression (Figure 10b).

| North Sea subareas
There was a clear variability in copepod abundance across North Sea subareas (statistical rectangles)  (Figure 11a,b).
The presence of C. finmarchicus in the Northwestern (NW) subarea markedly declined from 1958 but with indication of a partial recovery since the mid-1990s (Figure 11b).C. spp.(CI-CIV) and Para-Pseudocalanus spp.exhibited very much the same pattern as C. finmarchicus (Figure 11b).In the Southern (S) subarea a more persistent declining trend was noticed for each of these three copepod categories but likely flattening out in the case of C. finmarchicus since the 2000s (Figure 11b).The process of data quality assurance led to instances of omitted grid cells, particularly in Viking4a (V4a) subarea (Figure 11a).However, C. finmarchicus also in this subarea underwent a decline in abundance but with recent signs of recovery, as in the NW subarea (Figure 11b).Here as well, C. spp.(CI-CIV) largely reflected C. finmarchicus patterns while Para-Pseudocalanus spp.after a decline-showed indication of a slight recovery but then stagnation from the 1980 s (Figure 11b).C. helgolandicus in NW and V4a subareas is obviously on a rise whereas exhibiting a rather stable level in S subarea (Figure 11b).Finally, except for Para-Pseudocalanus spp., the abundance as such within each copepod category noticeably increased from S, via NW to V4a subarea (Figure 11b).However, these separate clines referred to different orders of magnitude, with Para-Pseudocalanus spp.most numerous (Figure 9; Figure 11b).
The consulted IBTS indices  clarified that the important role of the S subarea in upholding North Sea cod productivity is diminishing (Figure 11c).S5).For the NW and V4a subareas, these types of correlations were highly mixed in nature (Appendix S1, Table S5), accompanied by rather stable Recruitment_yp1_NW and Recruitment_yp1_V4a but recently increasing SSB_yp3_NW and SSB_ yp3_V4a (Figure 11c).

Experiment -Barents Sea cod
Ranked by increasing maximum temperature (°C)  However, Kola T inadequately reflected the encountered hydrographical conditions as such; ambient T during winter was significantly higher, except for small cod in recent times (Appendix S1, Figure S4a), and conversely, significantly lower during summer (Appendix S1, Figure S4b).
Most centre of gravities (CoGs) fell within a southwest-northeast 'route' in the Barents Sea, with markedly higher variation in summer than in winter (Figure 12b).Large and small cod swapped latitudinal position between summer and winter (Figure 12b).However, large specimens could also migrate farthest north during summer (Figure 12b).Our correlation coefficient flowchart spoke for an indirect effect of ocean warming (Kola T), cascading-in combination with the reduced ice extent-to an enlarged suitable feeding area (SFA) followed by an improved total stock biomass (TSB) and then to a pronounced increase in both Barents Sea cod displacement distance (D D ) and directional speed (U D ) (Figure 12c).On the contrary, the direct effect of Kola T explained only about 4% and <1% of the deviance in D D in small and large cod, respectively (Figure 12d).
For the latter group of individuals-the subsequent spawners-TSB explained 75% of the deviance in D D (Figure 12e) and 65% for U D (Figure 12f).

| DISCUSS ION
We believe that the approach and methodology used in this study on Atlantic cod might serve as a template for future studies exploring thermal limitations of teleost reproduction.Our investigation on the life history and reproductive ecology of the cold-temperate G. morhua reveals that local persistence under climate change is constrained by (1) knife-edge ovulation requirements, (2) altered larval prey biogeography and (3) energetically costlier spawning migrations, illustrated by the Irish/Celtic Seas-English Channel Complex, North Sea and Barents Sea scenario, respectively.Hence, each of these three addressed causal mechanisms is highly habitat specific.
Obviously, a series of other stressors are involved too, such as harvest rate (Hutchings, 2000) and lower trophic level productivity due to strengthened stratification (at least in the southern domain addressed here) (IPCC, 2021), but the performance of the currently addressed traits are fundamental for continued stock persistence.
So, this article complements related studies that either address consequences of one single stressor, as seen in meta-analyses (Dahlke et al., 2020;Free et al., 2019), or, on the other end of the scale, undertakes expert scorings of multiple stressors for both data-poor and data-rich stocks (Hare et al., 2016;Kjesbu et al., 2022).The current line of research asked for access to vast, in-depth biophysical information, where Atlantic cod was the logical choice as study subject.
To circumvent the overriding investigative problem attributed to 'extremely scarce' reproductive data for teleosts in general, and especially in view of ocean warming (Dahlke et al., 2020), we addressed relevant, basic reproductive physiology aspects of Atlantic cod as an example, given that this is a relatively well-studied species.
The resulting experimentally observed, spawning frequency (S f ) data clearly did not fit into the conventional thermal performance curve (Levesque & Marshall, 2021).The key deviation was the presence of both high and low S f for a given individual past a rather moderate T, here defined as T up , corresponding to 9.6 (±0.25)°C.Rightly so, there might be 'cells of water' in each of the 20-m 3 spawning chamber differing in T from the presently recorded surface T implying that the ambient T-leading to T up -might be partly biased, though it should F I G U R E 6 Aggregated information on variation in spawning performance for captive Barents Sea cod, grouped by ambient temperature regime, either 8.2 < T < 9.3, or 8.9 < T < 10.2°C: be kept in mind that cod swim upwards in the water column during courtship behaviour (Brawn, 1962).Much more important-and validated by field observations-is that the existence of a knife-edge threshold during spawning is strongly supported by data storage tag (dst) information (Appendix S1, List of Abbreviations and Terms) from a series of North Atlantic cod stocks (Righton et al., 2010).
However, the illustrated data in Righton et al. (2010) do not allow for precisely pinpointing any T up , except for that this threshold is found somewhere between 8 and 10°C.Nonetheless, the Gulf of St Lawrence cod (Ouellet et al., 2001) showed indications of offset, lower S f values already at 4-5°C, which probably relates to this stock being adapted to spawn at extraordinarily low temperatures, that is around 0°C (Brander, 2005).So, there is seemingly a degree of plasticity at lower temperatures in acclimatized stocks but whether this might happen in warmed waters is yet uncertain.The notion that any optimal temperature (T opt )-mathematically defined as dS f /dT = 0-was non-existent questions the predictive value of the oxygen-and capacity-limited thermal tolerance (OCLTT) theory (Pörtner et al., 2017;Pörtner & Farrell, 2008) in an transient setting such as ovulation, in line with related physiological arguments (Jutfelt et al., 2018).This lack of T opt should explain why spawning ground temperatures across cod stocks in the North Atlantic are not constant (Brander, 2005).Moreover, the seemingly inflexibility in T up , coupled with that erratic S f persisted after T decreased, challenges any practise of extrapolating generally accepted principles within aerobic scope theory to this specific part of reproductive physiology.Yet, it is unknown if this observed unstable S f pattern down to presently 8.9°C is maintained if T continues to decrease.
Although the proximate physiological cause behind this likely non-adaptive T up remains unknown, the 'weak link' is likely found within the extensive list of potential endocrine or enzymatic factors obstructing ovulation at T ≥ T up (Alix et al., 2020).According to aquaculture cod broodstock studies the overall fertilization rate in tanks significantly drops at T > 9.6°C (van der Meeren & Ivannikov, 2006).
We were unable to demonstrate this negative effect statistically on single egg batches (MFERT); our experiment focused on individuals with at least some fertilized eggs in each batch in order to be able to calculate S f from the T-dependent egg development rate.However, the circumstance that MFERT often did not noticeably fall following exceedingly long spawning intervals (low S f ) indicates that these advanced oocytes still remained in their supporting follicles instead of being ovulated and retained in the ovarian lumen under accumulating shortage of oxygen, cf. the 'egg-bound syndrome' (Hansen et al., 2016).The resulting egg size, represented by MEDW, is in cod determined by a sharp surge in vitellogenin during final oocyte maturation (Kjesbu, Kryvi, & Norberg, 1996).Warm waves (Figure 2) apparently did not disturb this route of vitellogenic sequestration (Alix et al., 2020;Kjesbu et al., 1991;Servili et al., 2020;Tyler & Sumpter, 1996).Likewise, no effect could be detected on batch fecundity, which seems reasonable as the fecundity of G. morhua is principally settled ahead of spawning (Kjesbu et al., 1998).One should, however, expect that the overall rate of vitellogenin uptake (∆VDW overall ) would fall if a persistently high T is introduced prior to spawning due to the foreseen reduction in 17β-estradiol in several species (Alix et al., 2020).Anyhow, our formulations confirmed that the teleost ovulatory cycle is highly complex in nature (Charitonidou et al., 2022); in cod probably at least two oocyte cohorts co-develop.
The further process of validation of ∆VDW overall showed, however, clear examples of underestimates for larger individuals, suggesting that a much bigger share of the vitellogenic raw material is stored in muscle and liver in larger than smaller females (Jansen et al., 2021;Kjesbu et al., 1991).In other words, ascribed to the usual pattern of (hyper)allometric reproductive investment in teleosts (Barneche et al., 2018).Hence, collectively, gonad size as such does not fully reflect realized reproductive investment, nor any associated, overall energetic strain.
We contend-irrespective of the future development of the zooplankton (cod larval prey) community in the Irish/Celtic Seas-English Channel Complex (Pitois & Fox, 2006)-that the projected increase in T in this area exceeding T up (RCP4.5, 2006-2070) will seriously disturb Long-term relationships (1963Long-term relationships ( -2017) ) between sea surface temperature (SST) and Calanus finmarchicus, C. helgolandicus, Para-Pseudocalanus spp.and cod spawning stock biomass (SSB) in the North Sea.(a) Calanus finmarchicus abundance vs. SST; (b) C. helgolandicus abundance vs. SST; (c) Para-Pseudocalanus spp. vs. SST; and (d) North Sea cod SSB vs. SST.SSB is 3-year lagged.A linear regression line with 95% confidence intervals and the associated explanatory power are provided for each panel.The split into two subperiods refers to previous (Beaugrand et al., 2003) and currently extended analyses.local cod egg release.However, any healthy cod eggs advected into these waters might possibly survive at even higher T (Geffen et al., 2006).It seems reasonable to believe that the documented T up is imprinted in the genes of G. morhua (Righton et al., 2010).
Furthermore, this knife-edge threshold at 9.6 (±0.25)°C-supported by the study of van der Meeren and Ivannikov (2006) at the group level-tells that any coarsely rounding off to the nearest integer will only vaguely define future locations of suitable spawning grounds, despite that bottom temperature clines might be reasonably sharp, as documented at or near the present southern spawning grounds.Our findings in these respects are considered novel, moving beyond descriptive narratives or physiological extrapolations to real mechanistic observations at the individual level followed by climate projections to address the ultimate consequences for local spawning grounds.
In the relatively cooler North Sea, the present 19-year extension of the earlier investigation by Beaugrand et al. (2003) on North Sea cod SSB_yp3 (spawning stock biomass forward lagged by 3 years) vs.
C. finmarchicus abundance illuminated the existence of a real link (r 2 = .43)in predator-prey dynamics.Despite this, for C. finmarchicus, brief and thereby unaccounted pulsed advection from the deep Norwegian Sea to the North Sea shelf might be an issue to study further (Heath et al., 1999;Sundby, 2000).Events of high fishing mortality-explored via complementary Recruitment_yp1 (recruitment at year +1) analyses-did not jeopardize this finding, though the dominant problem with discards and overfishing persists, as further south in the study region (www.ices.dk).The combination of subareaspecific CPR and IBTS datasets (NW = Northwestern, S = Southern, VPa = Viking4a) evidenced local differences in copepod abundance and in SSB_yp3_NW, SSB_yp3_S and SSB_yp3_VPa, as well as in Recruitment_yp1_NW, Recruitment_yp1_S and Recruitment_yp1_VPa, with the S subarea doing poorly in recent times.However, C. helgolandicus is on the rise but this trend is, as illustrated, unfavourable  for cod productivity (Beaugrand et al., 2003;Sundby, 2000); C. helgolandicus generally peaks in abundance past the cod larval feeding season (Beaugrand & Ibanez, 2004;Bonnet et al., 2005).Another relevant point in this connection is also that the long-term aver- Also, although Para-Pseudocalanus spp.appeared in our tests to be statistically uninfluenced by North Sea SST, these small copepods are evidently displaced northwards in this ocean basin, as seen for C. finmarchicus (Beaugrand et al., 2002;Sundby, 2000).Hence, we realized that the existence of pronounced collinearity between T and copepod category abundance generally hindered any true statistical insight in their individual role as predictors.Furthermore, at least the poor relationship between Recruitment_yp1_V4a and copepod categories may be due to a portion of cod recruits being in the adjacent Skagerrak (Wright et al., 2018).Nonetheless, the North Sea cod in the S subarea are certainly more at risk under on-going climate change than in the NW and V4a subareas but there are contributory factors to this scenario, such as locally altered fishing pressure (Engelhard et al., 2014).However, no direct evidence exists for related changes in adult migrations as such (Neat et al., 2014).
The IBTS was launched a few years after the Gadoid Outburst in the cold 1960-1970s and the concurrent event of a historic peak in C.
finmarchicus abundance (Cushing, 1984).Thus, the relative role of the S subarea was probably even more prominent pre-IBTS when these warm water masses cooled due to natural long-term climate oscillations (Gullestad et al., 2020) In any case, infrequent incidences of highly successful recruitment play a paramount role for any subsequent rebuilding of stock size (Hutchings, 2000).
In the relatively cold Barents Sea, we addressed cod displacement distance (D D ) and speed (U D )-a clearly understudied topic of special relevance today-as neither Barents Sea cod spawning temperature nor distribution of C. finmarchicus have yet changed substantially (Gullestad et al., 2020;Sandø et al., 2020;Sundby, 2000).This research action was inspired by previous analyses on coastal landing data indicating that the geographical distance between cod spawning grounds displaced southwards in a cold period and northwards in a warm period may sum up to a distance of 500 km (Langangen et al., 2019;Sundby & Nakken, 2008).We found that the variation in the centre of gravity (CoG) in the Barents Sea related to ice extent (Ingvaldsen & Gjøsaeter, 2013;IPCC, 2021).Also, the switch in latitudinal position of large and small cod between summer and winter is explained by the fact that large cod migrate out of the Barents Sea in late winter to spawn on the coast (Sundby & Nakken, 2008).
This said, the environmental temperature, represented by Kola T was an inadequate reflection of ambient T. So, mechanistically speaking, the indirect effect of ocean warming on D D and U D was not only totally dominating but also functioned as a cascading effect, seen by a reduced ice extent, an enlarged suitable feeding area (SFA), a larger total stock biomass (TSB) and then a pronounced increase in both D D and U D .However, in line with the general concept of lower  et al., 2021;Nøttestad et al., 1999;Schmidt-Nielsen, 1983), a statistically significant relationship between either D D or U D and TSB only applied for large cod.Because of this major contribution of TSB to D D and U D , an increase in TSB from, for instance, 1.5-4.5 million t, will increase D D and U D by a factor of 1.64 and 1.72, respectively.These accelerated migration costs (Jørgensen & Fiksen, 2006;Nøttestad et al., 1999) might negatively affect reproductive investment, such as fecundity (dos Santos Schmidt et al., 2017;McBride et al., 2015).This consideration also relates to that spawning migratory cod do  abundances (1958-2017) and the corresponding information on North Sea cod recruitment and spawning stock biomass (SSB) .(a) North Sea ICES statistical rectangles with grid cells, grouped into the southern, northwestern and Viking 4a subareas; (b) trends in abundance (mean ± SE) of the four copepod categories; and (c) recruitment-at-age-1-year and SSB indices for North Sea cod.Only CPR data (April-August) regularly sampled within ICES grid cells were considered.For North Sea cod, the data originate from the IBTS.not feed to any significant extent (Kjesbu et al., 1991).Another factor to consider is that the proximate player in question is TSB, a density-dependent factor, which is intimately linked to the adopted stock management regime, on top of the warmer climate (Kjesbu et al., 2014).Conversely, examples of extremely long Barents Sea cod migration southwards in historic cold periods (Sundby & Nakken, 2008) may well relate to other causal mechanisms, such as better recruitment in relatively warmer waters (Ellertsen et al., 1989;Planque & Frédou, 1999) translating to a gradual strengthening of the southernmost spawning grounds along the Norwegian coast (Sundby & Nakken, 2008).Nevertheless, the location of the ice edge and thereby of the polar-front-associated Barents Sea capelin (Mallotus villosus, Osmeridae) (Ingvaldsen & Gjøsaeter, 2013)-a key cod food item (Kjesbu et al., 2014)-is certainly an issue (Kjesbu et al., 2022).All aspects considered, this high-latitude example illustrates that any further increased migration distance and speed to coastal spawning grounds may turn energetically supercritical under the current extreme northward displacement of feeding areas.

F I G U R E 1
Long-term trends in global and regional temperature.(a) Global sea surface temperature (SST) anomalies; (b) regional transect temperature in the Barents Sea (0Change in temperature from 1987 to 2002 in the sea-water inlet at the experimental facility in Bergen, Norway.A 'warm wave' (see main text) was detected in 2002, briefly transient in 2001, whereas undetectable in 2000.The years 1987, 1991 and 1997 were added for further comparison.All temperature measurements were taken near the surface within the same tank (spawning chamber 8 (Kjesbu, 1989)).Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 2.1 | Cod reproductive physiology 2.1.1 | Vitellogenin uptake This analysis was designed to specifically test if the reproductive investment (and thereby the oxygen demand) becomes heightened during spawning, that is when the gonad peaks in size, especially in larger specimens (Pörtner & Farrell, 2008).Individual reproductive investment was represented by the daily, overall, dry weight-based uptake of yolk (vitellogenin) (∆VDW overall ).The established series of equations (

To
experimentally clarify to what extent ocean warming impacts early life-history stages, represented by embryos, MFERT (mean fertilization rate per batch, based on 100 eggs) and the associated MEDW (mean egg dry weight, see below) were linked to the encountered, 14672979, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License ambient T during single spawning intervals (Appendix S1, Figure S2).
org/10.17031/ 1758 ('Calanus') and https://doi.org/10.17031/1829 ('Small copepods')-provided by the CPR Survey at the Marine Biological Association of the United Kingdom, Plymouth, UK (https://www.cprsurvey.org/)-containedstandardized abundances on Calanus finmarchicus CV-CVI, C. spp.stage CI-CIV, C. helgolandicus and Para-Pseudocalanus spp.The datasets were confined within the geographical area of 49.5 to 62° N and − 4 to 10° E and spanned the period from 1958 to 2017 for the months of April to August.The CPR survey is undertaken by ferry boats and merchant ships ('shipsof-opportunity') that tow a plankton recorder in a fixed transect of the upper layer with a temporal resolution of about 1 month.Around 3 m 3 of water is sampled over the course of 10 nautical miles, which constitutes one unique CPR sample.Further details on the CPR survey are found in the published literature (Richardson et al., 2006), while the visualizations and analyses of the current study are specified in Section 2.5.14672979, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 2.3.2 | Variation in copepod abundance in North Sea subareas lated with the abundance of C. finmarchicus, C. helgolandicus, Para-Pseudocalanus spp., as well as with North Sea cod SSB.North Sea SST per month (1958-2020) was downloaded from NOAA, USA (Huang et al., 2017) and then calculating the annual average.North Sea transect temperature (0-100 m) (1958-2008) at 57° N and 59°20′ (www. 14672979, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License progr am-dista nce-two-point s-earth/).The time in question (149-198 days) was given from logbooks, followed by calculation of U D (m sec −1 ).
prior to parametric testing (two-sample Student t-test).The resulting probabilities were rechecked by running the non-parametric Kolmogorov-Smirnov test in parallel, finding irrelevant differences in p-values.The standard significance level at p = .05was Bonferroni adjusted in cases of multiple Pearson correlation tests (Figure 12c).ArcGIS Desktop 10.8.1 was used for geographical positioning of CoG.
C. finmarchicus and C. helgolandicus evidently responded oppositely to environmental temperature, represented by SST (Figure 8a,b).Their decadal and multidecadal shifts in distributional abundance were also distinctly different (Figure 9a,b).
(a) mean spawning temperature; (b) standard deviation (SD) of spawning interval; (c) mean spawning interval; (d) mean fertilization rate per batch; and (e) mean fertilization rate as a function of spawning interval.Panels (a) to (d) display 10th (whisker), 25th (box), 75th (box) and 90th percentile (whisker), grand mean (thick line) (box) and grand median (thin line) (box) and outliers (symbol).Same material and definition of common and warm wave as in Figure 5 Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

F
I G U R E 7 Observed shifts in cod spawning grounds in the Irish/Celtic Seas-English Channel complex and North Sea in the cold 1960s and warm 2000s and the projected southern border for suitable cod spawning in this region in the 2060s.(a) Downscaled hindcast simulations (ROMS-CORE2) of bottom temperature (March) in the cold 1960s and (b) warm 2000s and (c) forecast simulations (ROMS-NorESM-RCP4.5) to 2060, with the 9.6 ± 0.25°C (thick and thin red line, respectively) isotherm (T up ) annotated in all three panels.BC = Bristol Channel, EC = English Channel, IS = Irish Sea, NW = northwestern subarea, S = southern subarea and V4a = Viking 4a (northeastern) subarea.(a), (b) and (c) are provided with 1°C-resolved isotherms (thin black lines) and the 500-m isobath (thick black line).The specific location of spawning grounds (yellow colour) in (a) and (b) was the result of dedicated reviews (Appendix S1, Observed shifts in cod spawning grounds in the Northeast Atlantic)., 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License , 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

F
I G U R E 9 Mean decadal distribution of Calanus finmarchicus, C. helgolandicus and Para-Pseudocalanus spp.abundance between April and August in the North Sea (1958-2017).Contour lines show the various isolines.All these ln-transformed data were extracted from the CPR data base.Calanus finmarchicus, C. helgolandicus and Para-Pseudocalanus spp.
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License age abundance of C. helgolandicus is, as demonstrated, about one order of magnitude lower than for C. finmarchicus.This part of our fine-resolved analysis supplements a previous investigation limited to the southern (and central) North Sea (1974-2011)(Nicolas et al., 2014).Although the NW and VPa subareas give a more positive outlook in terms of copepod abundance, it is worth recognizing that generally little of the deviance (< ≈15%) is explained when split by subarea; the whole-ocean analysis generally showed significantly better explanatory power.As an example, in terms of Recruitment_yp1_S versus.copepod abundance, the current rank-in a declining order-goes from C. finmarchicus (positive influence) via Para-Pseudocalanus spp.(positive influence) to C. helgolandicus (negative influence), whereas Nicolas et al. (2014) showed an inverse pattern from C. helgolandicus via Para-Pseudocalanus to C. finmarchicus.

F
The long-term relationships between Calanus finmarchicus abundance and spawning stock biomass (SSB) of cod in the North Sea.(a) Observed trends, where the dotted line shows the last study year (1998) byBeaugrand et al. (2003); and (b) the corresponding linear relationship (mean ± 95% confidence interval).C. finmarchicus abundance refers to April-August, extracted from the CPR data base.SSB is 3-year lagged.Regression line: Y = 50.4+ 82.56 × X. , 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License migration costs per unit body mass when body size increases (Jansen
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Our research has shown that understanding and predicting (or projecting) population (stock) persistence under anthropogenic climate change should be based on revealing the key, local regulatory mechanisms.This viewpoint is further underlined in this article by concentrating on the life-history trait-based 'weak link' (critical sensitivity attribute)-using Atlantic cod as a template specieswithin each of the three Northeast Atlantic regions (50-80° N) currently addressed.Such an approach might, however, be criticized in cases where a series of sensitivity attributes in combination turn critical rather than each singularly, a common argument in climate F I G U R E 1 2 Seasonal migration dynamics of Barents Sea cod.(a) Consulted environmental drivers in the migration study(2004- 2017/2018), in addition to Kola transect temperature (Kola T) (Figure1b).The plotted time series for SFA (suitable feeding area), TSB (total stock biomass) and ice extent were for illustrative purposes extended back to 1981; (b) annual changes in the centre of gravity of Barents Sea cod during winter and summer by total length (TL) category.Line: Kola transect.The insert shows the surveyed area, in wintertime depending on ice extension(Ingvaldsen & Gjøsaeter, 2013).Bathymetry, white: 0-200; grey: 200-400; blue: 400-600 m; (c) correlation coefficient flowchart between drivers and displacement length (D D ) and speed (U D ) for small (30 ≤ TL ≤ 59 cm) and large cod (70 ≤ TL ≤ 99 cm).Kola T is given as the mean value in April to September.Parenthesis: An insignificant relationship; (d) and (e) D D regressed on Kola T and TSB for small and large Barents Sea cod, respectively.Upper regression line in (e): Y = 286 + 0.089 × X; and (f) U D regressed on TSB for large cod.Regression line: Y = 1.74•10 −2 + 6.48•10 −6 × X. Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/faf.12728by University Of Aberdeen, Wiley Online Library on [15/05/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License impact assessments using expert scorings.Thus, our endeavour should be considered as an alternative, where the strength lies in thorough, quantitative investigations of key traits and their drivers resulting in findings of clear importance for future cod stock productivity (Figure13).These novel results were made possible by running dedicated experiments, consulting extensive data bases and accessing and building upon a wealth of knowledge gathered by generations of researchers within different branches of marine science.

Table 2
)-given from the observed material on field and experimental females (Table 1)-therefore, ultimately aimed at accurately quantifying and contrasting ∆VDW P, overall (P = pre-(Table 2, Equation 11).Furthermore, the latter estimate functioned as ground-truthing (validation) in the subsequent model simulation (Table 2, Equation 18) on standardized reproductive and body size data sets.