Simulating the Effects of Temperature and Food Availability on True Soles (Solea spp.) Early-Life History Traits: A Tool for Understanding Fish Recruitment in Future Climate Change Scenarios

Abstract

Research on recruitment variability has gained momentum in the last years, undoubtedly due to the many unknowns related to climate change impacts. Knowledge about recruitment—the process of small, young fish transitioning to an older, larger life stage—timing and success is especially important for commercial fish species, as it allows predicting the availability of fish and adapting fishing practices for its sustainable exploitation. Predicting tools for determining the combined effect of temperature rise and food quality and quantity reduction (two expected outcomes of climate change) on early-life history traits of fish larvae are valuable for anticipating and adjusting fishing pressure and policy. Here we use a previously published and validated dynamic energy budget (DEB) model for the common sole (Solea solea) and adapt and use the same DEB model for the Senegalese sole (S. senegalensis) to predict the effects of temperature and food availability on Solea spp. early life-history traits. We create seven simulation scenarios, recreating RCP 4.5 and 8.5 Intergovernmental Panel on Climate Change (IPCC) scenarios and including a reduction in food availability. Our results show that temperature and food availability both affect the age at metamorphosis, which is advanced in all scenarios that include a temperature rise and delayed when food is limited. Age at puberty was also affected by the temperature increase but portrayed a more complex response that is dependent on the spawning (batch) period. We discuss the implications of our results in a climate change context.

1. Introduction

Research on recruitment variability—here referred to as the process of small, young fish transitioning to an older, larger life stage—has gained momentum in the last years [1,2,3], undoubtedly due to the many unknowns related to climate change (CC) impacts [4,5]. For example, temperature-driven reduction in primary production in the oceans is altering the recycling cycles of nutrients [6], which impacts the diversity and nutritious quality of zooplankton communities and consequently supports a smaller biomass of higher trophic levels, the overall result being warmer and nutrient poorer oceans [7]. The impacts of temperature-induced changes in phenology—studying the timing of recurring biological events and how these are influenced by climate [8]—are hard to predict and demonstrate, but phenological adaptations, i.e., shifts in the timing of seasonal events, have been reported for many planktonic and nektonic marine species, including fish [9,10,11,12,13].

Early transitional life stages of fish have strict environmental requirements for differentiation, metamorphosis, settlement, and growth. The environmental conditions can be used to determine healthy development and survival rates, which are very variable and often very low [14]. When the environmental requirements are not met, there are consequences at later life stages that could compromise the fitness of individuals and, thus, the health of populations. Compared with any other fish life stage, larvae will have the highest potential for growth, the highest weight-specific metabolic rates, and the highest sensitivity to environmental fluctuations and stressors [15,16]. As most processes determining the recruitment strength and the spatial distribution of fish populations occur during the planktonic stage of fishes, evaluation of ichthyoplankton remains a key element for fisheries assessment. Further, the survival of fish larvae directly influences future abundances of adult fish stocks, resulting in important interannual fluctuations in fish stock biomasses.

Understanding recruitment requires coupling physical and biological processes. The dominant recruitment mechanisms affecting fish larvae’s growth and survival are environmental factors, such as temperature and salinity, prey (feed) availability, predation pressure, and “the bigger and faster” relationship, which proposes that a bigger fish will be faster to evade predators and thus have a higher survival probability [17]. All of these mechanisms may act together over the entire period—from egg to the pre-recruit juvenile and are essential determinants of the year-class strength, i.e., the number of fish spawned or hatched in a given year and their respective recruitment success. Simultaneously to these environmental factors, growth and development are affected by the physiological characteristics of the individuals. Mechanistic bioenergetic models like Dynamic Energy Budget (DEB) models combine information on an individual’s physiology and ecology with information on environmental factors [18,19]. For example, growth is tightly linked to temperature, which in DEB models governs other physiological rates such as feeding, maintenance, and development [20,21]. In this sense, DEB models are a perfect tool to study the impact of environmental variability on, for example, the size and age at metamorphosis and puberty, which are relevant early life-history traits that affect fish recruitment.

Here we simulate the effect of future CC scenarios on fishes’ early life-history traits, particularly focusing on larval growth and age at metamorphosis in two flatfish species, the common sole, Solea solea, and the Senegalese sole, S. senegalensis. Both flatfish species are aquatic resources of high commercial value, appreciated in Europe due to their flesh quality [22,23,24].

Soles have a complex life cycle with progressive lifestyles and ontogenetic habitat shifts. S. solea and S. senegalensis are both “r strategist” species and batch spawners, and as such, they are characterized by strong fecundity. In both species, temperature is the principal environmental cue triggering the spawning of hundreds of thousands of eggs with low survival rates [25,26]. Fertilized eggs and larvae are both pelagic, which enables wide dispersion but also exposes these early life stages to (variable) environmental factors, notably temperature and food abundance. Recently hatched larvae drift towards estuaries and coastal areas—their juvenile environment—while they undergo metamorphosis. Major ontogenic, morphological, physiological, and behavioral changes occur during metamorphosis [27] and are accompanied by a change of habitat and lifestyle (from pelagic to benthic). The duration of metamorphosis varies depending on temperature and food availability. In S. solea, the transition from hatching until complete metamorphosis —the complete larval stage— can last between 15 and 26 days, at 19 °C and 16 °C, respectively [28], whereas the Senegalese sole usually takes about 8 days to complete metamorphosis at 20 °C, although with a high variability depending on environmental and feed availability [29,30]. Fully metamorphosed juveniles change from a pelagic to a benthic lifestyle—a process known as settlement—colonizing estuaries and bays that act as nursery areas [31,32]. The residence time in the nursery areas depends on fish size and development status and is around two to four years. Once fish reach maturity, adults migrate offshore (a few or up to hundreds of kilometers, depending on location) for spawning. They eventually return closer to the shore after the reproductive season to reach feeding grounds.

Environmental fluctuations strongly influence the spawning time, survival rates, egg and larval life history, the larval growth, development and dispersion process, and the transition from a pelagic to a benthic lifestyle [33]. For understanding the potential implications that (environmentally and thus CC-influenced) larval growth and development have for Solea spp. settlement and recruitment, we apply the species-specific DEB models to study the expected effects under future CC scenarios. Both species have three distinct life stages that characterize their development: a post-hatch pelagic larval phase, a juvenile benthic phase, and an adult benthic phase, so we decided to focus on two distinct life stage transitions: metamorphosis and puberty. We generate seven different temperature and food availability scenarios and study their effect on the growth and development of the early-life stages of these two flatfish species. We focus on larval stages as their sensitivity to CC might be further increased due to their high (mass-specific) metabolic rate, lower energy reserves, and the fact that they are less capable of migrating toward a suitable habitat [14], but we run our simulations until puberty is attained. Our hypothesis is that DEB model simulations will help identify the main effects of CC-induced changes in Solea spp. early-life-history traits when changes in food availability and temperature are forced into the models. Further, the effect of forcing variables should differ between spawning batches, which will have experienced different temperature exposure.

2. Materials and Methods

2.1. DEB Theory and Models

Dynamic Energy Budget (DEB) theory [18,20,21] focuses on energy—and mass—pathways in a (dynamic) environment, i.e., on energy and nutrient acquisition and subsequent utilization by an individual throughout its life cycle. DEB-based models describe the energy budgets of individual organisms, quantifying metabolism and the interaction of processes that constitute it, including feeding, maintenance, development or maturation, growth, reproduction, and aging. Temperature and food availability are often the main forcing variables in the model, and thus the way these environmental factors modify metabolism is considered explicitly. Further, dynamic energy budgets follow the changes of these fluxes during an organism’s complete life cycle, translating the individual’s physiological functions into a reduced number of differential equations and a dozen or so primary parameters. Most parameters are species-specific (see next section on model parameterization), but the rules for energy pathways are universal: energy is assimilated from food into energy reserves and then is mobilized (used) to fuel all metabolic processes, with maintenance having priority over growth and development (maturation) or reproduction. Life stage transitions, such as metamorphosis and puberty, are explicit maturity thresholds. Here, ‘maturity’ is one of the three main model state variables; the other two are ‘energy reserve’ and ‘structure’. Combined, the three state variables completely describe a (juvenile) individual: maturity tracks its development or maturation, and reserve and structure its size. After attaining puberty, an additional state variable, ‘reproductive buffer’, can be used to track the size and model reproduction of adults. For the complete model description, including a list of state variables and corresponding equations, please see Appendix A.

2.2. Accounting for Temperature and Food Availability

Temperature and food availability are the two main forcing variables in the model, effectively translating the environmental conditions into physiological responses of studied individuals. For the conversion between food availability and perceived food (scaled functional response, f), we use the Holling type II functional response:

f = X/X_K + X,

(1)

where X is the food density and X_K is the half-saturation constant for that type of food. Because f is a saturating function, it can vary between 0 when no food is available (food deprivation), and 1 food is abundant (feeding ad libitum).

The correction for the effect of (water) temperature is done using the Arrhenius equation (Equation (1.2) in [20]) and is assumed to be the same for all metabolic rates (Section 1.2 in [20]). The rate at a given temperature T is then obtained as:

k(T) = c(T) k(T_ref),

(2)

where c(T) is the correction factor for a certain temperature T, and k(T_ref) is the rate at the reference temperature (T_ref = 293.15 K or 20 °C by convention [20]). Where appropriate, we used an extended 5- parameter temperature correction (pp. 21–22 in [20], [34]) to capture the fact that S. senegalensis might be experiencing suboptimally cold and S. solea suboptimally warm temperatures during simulations. The full temperature correction function is presented in Appendix A and is available as a pre-coded function tempcorr.m in the freely available Matlab package DEBtool_M [35], which was used for parameter estimation.

2.3. Model Parameterization for S. solea and S. senegalensis

The life cycle and ontogeny of both sole species (S. solea and S. senegalensis) can be described by the abj-type model, which includes a metabolic acceleration between birth and metamorphosis and accounts for the metamorphosis of flatfish species in their early life cycle [36,37]. The metabolic metamorphosis is assumed to coincide with the physical metamorphosis (the process of eye migration and head remodeling), with “age and size at metamorphosis” (in the DEB model) corresponding to the properties of the fish at the end of physical metamorphosis, i.e., head remodeling.

We decided to use the same abj-type models for S. solea and S. senegalensis. The models share the same assumptions and model setup, but they differ in species-specific parameter values. For parameterization of the S. solea model, we focused on the study by Mounier et al. [38] and on the latest available Add-my-pet entry (Teal, AmP Solea solea, version 2015/08/28). Both of these versions of the S. solea models were parameterized using the AmP tool setup [39], so we merged the scripts and complemented the data with data on larval growth and development at several temperatures [40,41]. We then re-estimated parameters to obtain a comprehensive parameter set able to describe all data. For S. senegalensis, a DEB model was parameterized using data on life history (age and size at life events) and other data, such as growth curves at a known temperature, from published and unpublished sources. All parameter estimation was performed in Matlab using AmPtool and DEBtool [35,39,42]. The fits of model predictions to data used for model parameterization and all DEB parameter estimates for S. solea and S. senegalensis are presented in Appendix B. The complete code is freely available in the AmP collection by searching for the species’ name.

2.4. Models Validation

We validated the models using independent datasets on larval growth (from birth to metamorphosis) reared under controlled conditions of temperature and food availability (data from [43,44]). Briefly, datasets consist of regular length and dry weight measurements of larvae reared under laboratory conditions using polystyrene microplates with a cover as a housing system. This type of housing allowed for monitoring larvae growth, development, and survival at the individual level during the entire experiment, which lasted for seven (S. solea) or four (S. senegalensis) weeks. The experiment included two factors: rearing temperature (16 °C and 20 °C in S. solea and 17 °C, 20 °C, and 23 °C in S. senegalensis) and feeding regimes (three levels: high, medium, and low food, approximated by different feeding frequencies). For S. solea, the data on larvae length is available for every week of the experiment (7 measurements starting at age 8 dph). In contrast, for dry weight, four measurements are available, three corresponding to the first week and one for the last week. Biometric data on S. senegalensis were collected at three different times before placing the larvae in microplates (12 dph), at intermediate age (22 dph), and at the end of the experiment (32 dph).

For model validation, physiological (species-specific) parameters were fixed for both species, while the dataset-specific scaled food availability (f) was manually adjusted until a satisfactory fit between model predictions and data was obtained. The temperature was explicitly taken into account by correcting all rates to the corresponding temperature (Equation (2)).

2.5. Simulation Setup

All simulations were set in a hypothetical location based on the temperature characteristics of the west coast of France and Portugal, as both species can be found—at different abundances—in this geographical extension [26,45]. To mimic the natural life-history events and study potential sources of inter-individual variability, we simulated early, middle, and late spawning for each fish species: for S. solea, we used early January, early March, and early May, and for S. senegalensis, early May, mid-June, and early August corresponding to the observed spawning periods in this geographic area [45]. Spawning occurs on the continental shelf and is followed by “migration” of the eggs and then larvae towards estuaries which serve as nurseries [3,5,46,47]. We assumed the migration to last 30 days, approximately corresponding to completing 75% of larval development at the time larvae reach the target nurseries (Table S1 and methods in [3]). Larvae were assumed to feed during the migration but with lower efficacy, resulting in a slight reduction of the overall scaled functional response: f during migration was assumed to be 75% of the f experienced by fish outside of the migration period. We assumed that the migration itself does not incur any additional energetic costs.

In all simulations, food availability after the migration was assumed constant, and initial energy in an egg was dependent on the mother’s condition at the time of egg-laying (maternal effect [20]). Simulation duration depended on the species and the simulated scenario, as the simulations lasted until individuals were predicted to reach puberty: for S. solea, this was approximately 2.5 years, and for S. senegalensis, between 7 and 10 years. As points of interest, we made note of the predicted age and size at (i) hatching, (ii) birth (mouth opening), (iii) end of metamorphosis (complete larval development), and (iv) puberty (sexual maturation). We also tracked the growth (in total length) as a function of time, with food and temperature as forcing variables. Results are presented until complete maturation because we assume fish then leave the estuaries and go to the coastal sea to reproduce. All simulations were performed using MATLAB R2017b.

We evaluated the effect of temperature and the effect of food availability. In total, seven scenarios were simulated (

Table 1. List of simulated scenarios (S1–S7) designed to reflect potential changes in temperature (warming of 1.8 and 3.7 degrees—RCP 4.5 and RCP 8.5, respectively, IPCC, 2014), and decrease and increase in food availability, both relative to the baseline, i.e., current conditions. Current (baseline) temperature is assumed to reflect the average seasonal fluctuations in typical coastal sea and estuaries inhabited by soles (see subsection ‘Temperature scenarios’ of Methods), and the food availability (X) is assumed to result in scaled functional response f = 0.85 in both species. In each scenario, there are three spawning events: early, middle, and late, the timing of which depends on the species.

Food (X)/Temperature (T)	Current Food (bX): Baseline	Low Food (−X): 0.5 × Baseline	High Food (+X): 1.5 × Baseline
Current temperature (bT): baseline	S1 (bTbX)	S2 (bT−X)	S3 (bT+X)
RCP 4.5 (+T): bT + 1.8 degree increase	S4 (+TbX)	S6 (+T−X)	/
RCP 8.5 (++T): bT + 3.7 degree increase	S5 (++TbX)	S7 (++T−X)	/

), with three spawning events (early-mid-late) in each: S1 (bXbT) baseline scenario of current temperature conditions and current estimated food availability; S2 (bT−X) and S3(bT+X): scenarios with baseline temperature and a 50% decrease (bT−X) or a 50% increase (bT+X) in food availability (X, see Equation (1)); S4 (+TbX) and S5 (++TbX): scenario with a 1.8 degree (+TbX) and scenario with a 3.7-degree temperature increase (++TbX) from baseline temperature (RCP 4.5 and RCP 8.5 respectively, (Intergovernmental Panel on Climate Change IPCC, 2014), with current food availability; S6(+T−X) and S7(++T−X): scenarios with a 1.8 degree and a 3.7-degree temperature increase, respectively, and a 50% decrease in food availability.

2.6. Temperature

The temperature baseline for our hypothetical location was designed as a sine function with the minimum temperature at Julian day 45 in mid-January, and the maximum temperature at Julian day 227.5, in mid-August. The coastline baseline temperature (T_sea) ranged from 10 °C to 20 °C, based on average temperature ranges in several locations on the west coast of France and Portugal (seatemperature.org). The estuary baseline temperature (T_estu) ranged from 10 °C to 24 °C, based on the average minimum and maximum temperatures in the principal estuaries of the English Channel, the Bay of Biscay, and the Portuguese coast, including Seine, Gironde, and Tagus [48,49,50,51,52]. The simplified temperature functions are:

T_sea(t) = 5 sin f₀((2π(t − 136.25))/365) + 15

T_estu(t) = 7 sin f₀((2π(t − 136.25))/365) + 17

To mimic the fact that larvae are spawned on the continental shelf and then are transported towards estuaries, we considered it essential to include a gradual shift from coastal to estuary temperatures. Therefore, we computed a final temperature function based on the two previous sinusoidal functions for the sea and for the estuary, which we used to describe the evolution of temperature as the forcing variable of our simulations. The resulting temperature was determined by elapsed time as a proportion of the total migration duration (t_mig = 30 days [3,5,53]): as time progresses, this function gives less weight to sea temperature and more weight to estuary temperature:

T(t) = (1 − dt/t_mig) × T_sea(t) + (dt/t_mig) × T_estu(t)

2.7. Food Availability

For the baseline food conditions, a constant food level equivalent to a functional response of f = 0.85 was selected based on similar values estimated for other species in the wild [54,55,56,57]. The food abundance (X) corresponding to f = 0.85 was calculated using Equation (1). For the scenarios targeted at evaluating the effect of food availability on the growth and development of fish, the decrease by 50% (relative to the baseline) resulted in a functional response of f = 0.74, whereas the increase by 50% resulted in f = 0.9.

3. Results

3.1. Models Parameterization and Validation

The models nicely captured individual traits along the whole life-cycle of both sole species, with model predictions generally fitting the data well (see Appendix B for model fits and

Table 2. Summary of DEB parameter estimates for the common sole (S. solea) and the Senegalese sole (S. senegalensis) that were used for the simulations. Two parameters (v and {p_Am}) in the abj-model applied here increase their value between birth and metamorphosis (see Appendix A), so the final value is provided in brackets. The final parameter value is calculated by multiplying the initial value by s_M = 2.7942 (S. solea) or by s_M = 3.3472 (S. senegalensis), s_M calculated for f = 1. Arrhenius parameters for the 5-parameter temperature correction (T_L, T_H, T_AL, T_L) were set manually. For a full list of parameters, please see Table A3 and Table A5 in Appendix B.

Parameter	Symbol	S. solea	S. senegalensis	Unit
Arrhenius temperature	TA	7980	6528	K
Maximum surface area specific assimilation rate	{pAm}	282.195 (788.5219)	95.61 (320.05)	J/d·cm2;
Energy conductance	v	0.06828 (0.1908)	0.0697 (0.2332)	cm/d
Allocation fraction to soma (kappa)	κ	0.782	0.8117	-
Volume specific somatic maintenance	[pM]	31.22	22.83	J/d·cm3
Specific cost for structure	[EG]	5188	5230	J/cm3
Maturity at hatching	EHh	0.1809	0.0552	J
Maturity at birth	EHb	0.3417	0.1671	J
Maturity at end of metamorphosis	EHj	7.49	6.309	J
Maturity at puberty	EHp	1.964 × 105	1.258 × 106	J
Shape coefficient for larvae	δMe	0.1578	0.221	-
Shape coefficient post-metamorphosis	δM	0.199	0.2235	-
Arrhenius temperature for high temperature extreme	TAH	20,000	20,000	K
High temperature extreme	TH	294.15 (21 °C)	301.15 (28 °C)	K
Arrhenius temperature for low temperature extreme	TAL	18,000	18,000	K
Low temperature extreme	TL	280.15 (7 °C)	283.15 (10 °C)

for a summary of the model parameters used within this study). The relatively low values for the mean relative error (MRE), symmetric mean average error (SMAE), and the symmetric mean squared error (SMSE) in both species parametrization supported the visual inspection of the model fits: S. solea MRE = 0.102, SMAE = 0.109, SMSE = 0.023; S. senegalensis MRE = 0.552, SMAE = 0.136, SMSE = 0.057.

The maturity levels at hatching, birth, and metamorphosis are estimated to be higher for S. solea than for S. senegalensis (Table 2). This can explain the observed faster development and maturation of S. senegalensis when data for the same temperatures are compared [41,58]. However, the maturity level at puberty is estimated to be higher for S. senegalensis, implying that S. solea starts allocating energy toward reproduction sooner.

Other primary parameters imply a somewhat higher assimilation potential ($\left\{{\dot{p}}_{Am}\right\}$) for S. solea, coupled with a higher specific maintenance cost ($\left[{\dot{p}}_M\right]$). With a similar specific cost for structure ([E_G]) and allocation fraction to the soma (κ), we assume that the growth rates of the two species would not drastically differ, especially after final parameter values have been attained (post-acceleration). A notable difference is in the energy conductance ($\dot{v}$), which starts at a similar value at the larval phase (birth) but undergoes a more drastic acceleration (by a factor s_M = 3.34) in S. senegalensis. This, combined with the lower maturity thresholds mentioned earlier, contributes to the faster life cycle of the Senegalese sole and would result in faster growth compared with that of the common sole.

A higher Arrhenius temperature estimated for S. solea could be linked to the fact that parameterization of S. solea included a larger proportion of larval data, compared with S. senegalensis¸a point to which we come back in the discussion. Interestingly, the larval shape coefficient is similar to that for metamorphosized S. senegalensis, implying a negligible change in (metabolically relevant) shape; this could be a computational artifact.

3.2. Model Validation

DEB model growth predictions were able to capture the observations (data from [43,44]) while explicitly accounting for experimental temperature and manually adjusting the value of f to best match the observations (Figure 1 and Figure 2). All other parameters were kept fixed at values listed in Table 2. For S. solea, the best fit to the data for the ad libitum treatment was obtained by setting the f parameter to 0.18 (Figure 1). For larvae at MediumFood and LowFood treatments, the f values were 0.15 and 0.10, respectively. For S. senegalensis, the f values were 0.4, 0.45, and 0.55, for larvae that were fed with low, medium and high frequency (Figure 2). We hypothesize these relatively low values implied for f in the discussion.

3.3. Simulation Results

3.3.1. Temperature Increase Is Speeding Up the Development and Thus Reducing the Stage Duration

The stage duration, i.e., the time elapsed from one life event to the next one, decreased in both species within the simulation scenarios experiencing higher temperature (S4: +TbX and S5: ++TbX), higher food availability (S3:bT+X), and even lower food availability if combined with a temperature increase (S6:+T−X and S7:++T−X) (Figure 3 and Figure 4).

For S. solea, the S5 (++TbX) and S7 (++T−X) scenarios hastened development the most, which is especially interesting for S7, where a larger temperature increase compensated for a 50% decrease in food availability compared with S6 (+T−X). Early spawned individuals (i.e., on 1 January) in scenario S7 exhibited a more pronounced difference for age at metamorphosis (aj, Figure 3) than those spawned mid and later in the year. Compared with the baseline scenario S1, a complete metamorphosis occurred 14.7 days earlier for early-spawned individuals and 13.4 and 7.4 days faster for those spawned mid and later on the year, respectively (Figure A5 in Appendix C). Similarly, the time to complete puberty shortened the most in the S5 and S7 scenarios, with a maximum of 58.6 days difference between early-spawned larvae and baseline scenario ones in S5 individuals.

The differences in the stage duration until complete metamorphosis were less marked for S. senegalensis, which showed a maximum of 9.5 days difference between baseline and the S5 (++TbX) scenario for the early-spawned individuals, while the difference was less than a day and a half in late-spawned individuals (Figure 4, top).

Regarding age at puberty, late-spawned soles were generally a bit faster than mid and early-spawned fish in reaching puberty. The highest difference from baseline was obtained for S. senegalensis individuals issued from the last spawning event of the year (day 210, 1 August) and under the S7 (++T−X) scenario, where it would take up to 694 days less for individuals to reach puberty than individuals from its corresponding baseline scenario.

Changes in food availability and temperature impacted, to a lesser extent, the age at hatching and age at birth. In both species, a slight reduction in the age at birth was observed, notably for the scenarios involving an increase in temperature. In S. senegalensis, changes in the age of birth—the age at which exogenous feeding starts—were more evident for early-spawned individuals but always lower than 2.7 days, which was the maximum difference recorded for S7 (++T−X) scenario (early spawns). In S. solea, the highest time variance for age at birth was obtained in S7 individuals from the mid-spawn, which started exogenous feeding 4.5 days earlier than baseline individuals. The food availability did not alter the age at hatching or birth, which at this period are processes regulated by the amount of energy available in the egg yolk and yolk sac.

3.3.2. Food Availability Reduction Is Delaying Development and Thus Increasing the Stage Duration

For both species, age at metamorphosis and puberty were consistently higher in larvae from the S2 scenario (bT−X, Table 1), where food availability was reduced by 50% compared with the baseline and temperature was unaltered (S1: bTbX) (Figure 3 and Figure 4). The increase in time for completing metamorphosis varied among spawning events, although, for both species, it was always less than three days compared with the baseline scenario (Figure 3 and Figure 4). The age at puberty was notably delayed for both species: S. solea juveniles took, on average, 42 more days (compared with S1 juveniles) to complete maturation (Figure 3, bottom panel), whereas S. senegalensis juveniles needed over 1000 additional days to complete maturation (Figure 4, bottom, see also Table S1).

3.3.3. Larval and Juvenile Growth Rates

As a broad pattern, a rise in temperature increased growth rates in both species, and a decrease in food availability had the opposite effect, which was offset or even fully compensated when the two factors acted together (S6: +T−X, S7:++T−X). The effect depended on the time of spawning and the species (Figure 5 and Figure 6, see also Figure A6 in Appendix C): for example, a temperature increase affected larval phases of early S. solea spawners more than late S. solea spawners (Figure 5b,c), but the decrease in food availability hit late spawners of this species harder (Figure 5a). Late spawners of S. senegalensis also exhibited higher sensitivity to decreased food availability than earlier batches (Figure 6a), but the responses to temperature increase were similar between batches, with relatively little effect within the first few months (Figure 6b,c).

Predicted growth rates (in length) varied according to the developmental stage (Figure A5 and Figure A6), especially for S. senegalensis individuals. The growth rates in length declined with fish age and were highest for larvae stages—i.e., from birth to metamorphosis—ranging from 0.080–0.140 cm d⁻¹ in S. senegalensis and 0.037–0.091 cm d⁻¹ in S. solea larvae. For both species, the lowest growth rates were observed in individuals under the S2 (bT−X) scenario (0.08 and 0.037 cm d⁻¹, for early-spawned S. senegalensis and mid-spawned S. solea, respectively). In contrast, the maximum growth rate corresponded to late-spawned S. senegalensis (0.140 cm d⁻¹) and S. solea (0.091 cm d⁻¹) larvae under the S5 (++T-bX) scenario.

When comparing growth rates between the two species, S. senegalensis larvae showed, for all simulated scenarios, higher growth rates than S. solea larvae, which has a more homogeneous growth rate between larvae and juveniles, and as such, a less drastic difference in developmental stage growth rate than the Senegalese sole (Figure S1). Interestingly, the highest temperature increase (scenarios S5: ++T-bX and S7: ++T−X) affected, overall, the late spawners of both species more severely than earlier spawners, but in the opposite direction: while S. solea individuals grew at times slower than their temp-baseline counterparts, S. senegalensis individuals grew faster (Figure 5b,c and Figure 6b,c).

4. Discussion

Studying the impact of CC-induced changes in food availability and temperature on the ontogenetic development of fish larvae seems necessary in the context of global change and biodiversity loss, coupled with the ongoing changes the planktonic communities have faced for several decades [7]. Here, we used DEB models for forecasting and comparing fish early life-history traits of two commercially relevant species (S. solea and S. senegalensis) under seven distinct future CC scenarios. The direct comparison (simulating identical environmental conditions to forecast the responses of two distinct species) was possible after independently parameterizing and validating models with species-specific parameter sets. We will discuss (i) the assumptions and choices made while parameterizing models and designing the study, (ii) the main results of model parameterization, validation, and simulations, (iii) the implications of our results in the wider context of climate change, and (iv) further research.

4.1. Study Design—Assumptions and Choices

We have made several explicit assumptions and choices while designing the study. The process was motivated by literature and experience, but as the choices determine the scope and applicability of the results, we feel it is useful to discuss each of the major assumptions and choices.

All individuals are physiologically equal. To simplify the study design and results interpretation, all individuals of the same species are assumed to be physiologically equal, i.e., they are described by the same set of (species-specific) parameters [20]. Clearly, this is not the case in nature, where much of the inter-individual variability stems precisely from physiological differences. When sufficient information is available and inter-individual variability is in focus, variability can be integrated into DEB models via one or more physiological parameters (e.g., [59,60]). However, our choice was motivated by the overall aim to focus on the role of the two environmental factors (temperature and food availability) in shaping observable responses in growth and maturation rates while excluding differences resulting from other factors (e.g., as was done in [38,54,55]).

Characterizing the hypothetical location. For the baseline conditions (temperature and food availability), we used a temperature year function representing a hypothetical location from the west coast of France and Portugal and have coupled this temperature function with constant food availability, represented by a scaled functional response of f = 0.85 for both species, which decreases during migration. The choice was, again, motivated by study focus and supporting literature. Even though the location temperature characteristics favor the growth and maturation of S. solea over S. senegalensis—a more subtropical species—the represented location does support both species [61,62,63] and might become more accommodating for S. senegalensis in the context of CC-induced temperature increase. Constant food availability—as applied in similar studies [38,54,57,64]—was preferred over some function of food availability because proxies generally used to construct such functions (e.g., chlorophyll a) vary irregularly between different years and micro-locations, as does water turbidity, hindering the design of a representative food function. However, the methods and models applied here could be used for a particular area, for which specific temperature and food density functions could be implemented, as was done in, e.g., [55,56].

Mother’s nutritional status determines the egg size. Egg size and the availability of energy from the yolk are conditioned to the mother’s status (i.e., see the maternal effect in [20,65]. Without information on the mother’s status and in the context of constant food availability in simulated environments, eggs of a species were assumed to be of the same size throughout the simulation. Devauchelle et al. [66], however, reported smaller S. solea eggs with increasing temperature, implying that these eggs have a lower energy content. Because higher temperatures are usually associated with being later in the spawning season, this could imply energy limitations experienced by mothers later in the spawning season, which could be included in the model. Under the current setup, the models’ predictions fail to illustrate the natural variability in individuals hatching with different energetic contents. However, should a need arise, it is possible to include variability in egg size—via the maternal effect and/or stochastic variability [59]—where the initial energy in an egg is adjusted to the temperature and/or spawning time and corresponds to a lower f (energy limitation) experienced by the mother, followed by a higher f after settlement into estuaries, i.e., nutrient-rich environments.

Migration as a passive journey. We modeled the larval migration as passive transport lasting 30 days. During this period, we assumed larvae have a 25% lower f compared with that experienced by fish outside the migration period. In addition, larvae were gradually exposed to the temperature of the estuary as opposed to that of the sea. The duration of the migration period was assumed to be dependent on abiotic factors only (mainly sea currents) and therefore was the same for both studied species. The assumptions for the ‘migration’ were motivated by the literature suggesting mostly passive drift [3,5] and larvae reaching estuaries almost or completely metamorphosed [3,5,46,47]. In reality, the migration duration—as well as recruitment success in the estuaries—depends on multiple biotic and abiotic factors; and includes, to an extent, active swimming by larvae, which not only differs between individuals of different species but also between individuals of the same species, showing temporal and spatial variation (e.g., [3,5]). Duration of larval development—happening in parallel with the migration—is, in turn, also highly dependent on biotic and abiotic factors and can have a range of values up to 40 days (Table S1 in [3,53]; see also our results). Setting the migration duration as fixed 30 days in our simulations is a reasonable and approximative choice but can be modified if more precision is needed.

Food and temperature as the main drivers of CC-induced effects. In this work, we simulated the effects of food and temperature—two important environmental drivers—in sole metamorphosis success separately, but in nature, they are unlikely to operate independently. Furthermore, each (or both) of these factors might be further combined with the effects of ocean acidification and/or hypoxia [67]. Even though some preliminary work has been carried out on the effects of acidification on sole development [64], we have decided to focus on the better-documented and much better-studied effects of temperature and food availability. To elucidate better the effects of food availability, we simulated a theoretical increase in food availability (scenario S3), but research indicates that a lower scaled functional response f is much more likely than a higher one, as shown for anchovy with f values estimated mostly below 0.8 [55,56], so more focus is given to interpreting the food reduction scenarios. In addition to exploring the effects of food and temperature (scenarios S1–S5) separately, the two simulated scenarios combining the effects of temperature and food (S6: +T−X and S7:++T−X) capture situations where CC-induced temperature increase results in energy-depleted oceans, which seems quite likely [7].

Parametrization of the model using aquaculture data. Finally, a minor but valid point is that a lot of the data used for model parameterization (see Appendix B) is obtained from aquaculture, whereas the simulations focus on wild populations. Generally, individuals reared within aquaculture facilities are well fed and raised at higher temperatures, which enhances growth, and as a result, aquaculture soles grow and mature faster than wild individuals. Aquaculture of the Senegalese sole has managed to make fish reach puberty within 2–2.5 years when the individuals reach an average weight of 700–1000 g, while in the wild, the transition from juvenile to adult can take between 4 to 7 years, depending on the location [30]. Regardless of using aquaculture data for the model parameterization, our models were able to reproduce these characteristics with an initial spawning between 5 and 7 years of age (Figure 6b). In fact, one might argue that it is precisely because aquaculture data were available, with monitored and recorded temperature and feeding conditions—which could then explicitly be taken into account—that we were able to parameterize the model for both species with such success (see Appendix B for model fit to data from both aquaculture and the wild).

4.2. Implications of the Main Results

4.2.1. Model Parameterization and Validation

Generally, parameter sets for both species produce good fits to data used for model parameterization (Appendix B) as well as good fits to validation data (Figure 1 and Figure 2). As expected for closely related species belonging to the same genus, parameter values are similar between the two species, with some exceptions (see Results, Table 2), which might have been influenced by the availability of data used for model parameterization. Namely, the relative abundance of larval data used for S. solea parameterization might have influenced a somewhat higher value for the parameter T_A, implying a higher temperature sensitivity. We discuss this in more detail later in the discussion.

Model validation was carried out on experimental laboratory data. Finding relevant data for validation is often a challenge, particularly when evaluating the effects of multiple variables. Here, the experimental data imply a substantial difference between the expected and estimated food availability (f): while the experimental protocols aimed for ad libitum food (analogous to f = 1), estimates for high food treatments were much lower for both species (f ranging from 0.10 to 0.45). The relatively low estimates for f might result from insufficiently nutritious food items provided under an ad libitum feeding regime —either due to quantity or quality (variety) of food— as suggested for S. solea by Mounier et al. [38] and partially addressed for S. senegalensis by Sardi et al. [44], or due to inappropriate (too small) rearing containers. Feed not matching the realistic needs of commercial fish has already been demonstrated for another well-studied species (bluefin tuna [68]).

4.2.2. Larval and Juvenile Growth Are Affected by Temperature and Food Availability

As a general pattern, growth rates were positively affected by an increase in both temperature and food availability (Figure A6), and ultimate size (implied by the growth curve) was affected by food availability (Figure 5 and Figure 6). These results are in line with simulations of similar type [38,54,57] and have much to do with the model setup, but also do reflect what we can observe in nature.

Growth retardation over colder months (winter) was much more evident for S. senegalensis, the species for which simulations (lasting until puberty was attained) also lasted substantially longer: up to 10 years, compared with those for S. solea, lasting up to 2.5 years. S. senegalensis also had a more pronounced positive response to the more drastic temperature increase (+3.7 °C), whereas juveniles of S. solea that were spawned late even exhibited a negative response compared with the baseline (see Figure 5b). The patterns can most likely be explained by the different ecology of these species—S. senegalensis generally prefers warmer waters (but see [69])—which was accounted for in the model in the form of critical high and low temperatures (Table 2). Research shows that raising larvae at higher temperatures accelerates their growth rate (e.g., [70]), but the ultimate size decreases with increasing temperature—also known as the temperature-size rule (TSR [71]). Consequently, at higher temperatures, individuals do not necessarily attain their max size, resulting in smaller fish body sizes [72]. The reason may be directly related to the feeding rate, which also increases with temperature (i.e., food supplies will likely become limited at higher temperatures, reducing the ultimate size [20]), or to oxygen limitation, i.e., oxygen availability as the critical factor limiting growth becomes especially relevant at higher temperatures [73,74,75]. Ultimate size (of adults) could be relevant for recruitment: larger body size is (with high variability) positively correlated to fecundity [76,77], which, in turn, has been linked to higher recruitment variability [78]. However, the mechanisms behind the TSR in ectotherms are still an issue of debate in the scientific community [79] and are out of the scope of this study.

Size has been hypothesized to alter survivorship in several ways. Predation and “the bigger and faster relationship” depend on individuals’ size; as such, faster growth and larger body size are assumed to increase the survival probability of fish (but see [80,81]). In flatfish, the size at settlement—which has also been identified as a proxy for juveniles’ survival—is also influenced by the pelagic larvae duration (PLD) [53], which is, in turn, affected by the same abiotic factors (see next section). Larger sizes at settlement may offer some survival advantages, but small individuals settling in warmer conditions may grow faster after settlement and quickly reach equivalent sizes to fish settling in cooler conditions [82]. When studying the size composition of fish stock at any given time, it is important to keep in mind that the observed size can be a result of many complex interactions, as shown by our simulations. Even while assuming all individuals of the same species are physiologically identical and thus excluding inter-individual variability, our simulations predict that, for example, in January of the second year, the hypothetical nursing grounds will contain S. solea juveniles ranging from 15–20 cm total length, and S. senegalensis juveniles ranging from 5–10 cm total length. The size range, in our case, is a result of (i) age: S. solea individuals are between 8 and 12 months old, and S. senegalensis juveniles are between 5 and 8 months old; (ii) temperature: all of the S. solea individuals experienced peak high temperatures of the summer (June–August), whereas some S. senegalensis were spawned at the end of the warm season and consequently experienced only several months of winter temperatures, and (iii) food: some simulations included a higher, and some a lower food availability. In nature, the inter-individual variability will be affected by even more factors.

4.2.3. Developmental Rates Are Affected by Temperature and Food Availability

The simulation results demonstrate that both temperature and food availability affect development rates. A temperature rise reduced the time required for the occurrence of exogenous feeding and metamorphosis in pre-metamorphic larvae. The decrease in food availability, however, had a more pronounced effect on the period required for juveniles to reach puberty, especially for S. senegalensis. When the simulations combined temperature rise with food reduction (S6 and S7), the stage duration was slightly prolonged, indicating that the reduction of energy availability partially counterbalanced the acceleration effect of temperature rise. These results directly affect the population renewal, and they also imply—especially for later batches—more prolonged use of the nursery, where fish are more exposed to anthropogenic pressures, including exposure to pollutants, fishing, and heatwaves.

Metamorphosis is an energy-demanding process; as such, a food reduction simulated in scenario S2 delays the onset of metamorphosis, and larvae reach metamorphosis at an older age. However, the reduction in food availability did not delay the onset of metamorphosis in both species when combined with higher temperatures (S6 and S7). This counterintuitive result might indicate the need to include survival predictions and, more explicitly, model the physiological stress potentially experienced at higher temperatures [69]. Specifically, the survival probability of larvae might decrease if hastened growth (and development) is not coupled with abundant food [80,81], which was not the case in scenarios where a temperature increase was coupled with food reduction (S6, S7). In our results, the lesser effect of changes in food availability—particularly between scenarios S2, S6, and S7—on the early stages of larvae development (hatching and birth = mouth opening), as well as on the age to complete metamorphosis, reflects the influence of running simulations using the same initial reserve energy. Namely, the availability of energy from the yolk was assumed to be the same within the models because all eggs of the same species were assumed to be of the same size. In nature, larvae issued from large eggs with a large yolk sac have a higher energy supply (yolk) when they hatch [83] and thus have a greater chance of surviving than those with a small yolk sac, especially if food is scarce. As a result, individuals’ variability in the energy reserve would potentially result in differences in the stage duration between birth and the end of metamorphosis. From a simulation point of view, this could be solved by comparing larvae growth using different initial reserve energy content values, but detailed information is needed to compile a realistic range of values.

4.3. Context: Phenology, Phenological Adaptations, and Potential Mismatches

Soles are batch spawners, with the timing of spawning induced by temperature: for S. solea, the optimal temperature for spawning ranges from 8 °C to 12 °C, whereas for S. senegalensis, the water temperature should be between 15 °C and 20 °C [45]. We took this into account by simulating three spawning events (early-middle-late) for each species, corresponding to early January, early March, and early May for S. solea and for S. senegalensis early May, mid-June, and early August [45]. We did not evaluate/simulate the effect of temperature on the timing of spawning, but the observed variations in the intensity of the effect did depend on the spawning batch, with the fish spawning the latest being the most affected (Figure 3, Figure 4, Figure 5, Figure 6). For S. solea, the effects were negative for the highest simulated temperature, which reached the thermal tolerance for this species when larvae hatched in May. For S. senegalensis, the simulated temperature rise had a positive effect, but the food limitation drastically prolonged the maturation process. This was especially evident in late spawners who, possibly due to water temperatures decreasing soon after spawning (occurring in August), underwent growth and development retardation as low temperatures reached their thermal tolerance. Indeed, the times until yolk absorption, metamorphosis, and PLD of fish larvae are all negatively correlated with temperature [82]. All temperature rise simulations (S4–S7) advanced the onset and reduced the stage duration of metamorphosis in both species, and because mortality is generally very high during the larval phase, faster growth and shorter PLD at higher temperatures could positively affect larval survivorship [82]. However, should larvae metamorphosize entirely before reaching the estuarine nursery areas, they will be vulnerable to predation in an environment low in prey and will need to spend their energy trying to reach the estuaries, which will probably negatively impact their survival probability, as also suggested in Pepin et al. [84]. In that context, it might be advantageous for both species to shift spawning earlier in the season, to avoid high temperatures (S. solea) or lower temperatures with low food availability (S. senegalensis).

Organisms that are dependent on temperature to stimulate physiological development and larval release—such as S. solea and S. senegalensis—have significantly moved forward in their seasonal cycle in response to temperature [10]. Further, another study demonstrated that globally, spring phenologies of all marine species have advanced by 4.4 ± 1.1 days per decade since the mid–20th century [11]. Phenological adaptations also constitute, for fish, an important mechanism to tackle the changing climate [12]. Indeed, a study that estimated the peak spawning time in seven different subpopulations of sole from 1970–2010 demonstrated that at least four of the seven studied stocks of S. solea showed a significant long-term trend toward earlier spawning at a rate of 1.5 weeks per decade [13].

The impacts of temperature-induced changes in sole phenology are hard to predict and demonstrate. Even though spawning earlier might mitigate a mismatch between metamorphosis and arrival into nurseries, temperature-driven changes in phenology could cause a mismatch between different trophic levels and negatively affect recruitment by directly increasing mortality and extending the metamorphic period. Climate change is causing variations in the peak and seasonality of primary production in the oceans [85], pointing to a reduction of food availability for pelagic larval stages. Although many pelagic organisms are responding to water warming, it is, however, the intensity of the response that varies. The onset of metamorphosis requires larvae to acquire a competent size [53], and if the conditions are not favorable for growth, metamorphosis takes longer and renders larvae vulnerable to predation [82]. This was observed in S2 simulations, which accounted for the highest median duration for metamorphosis in S. solea (30 days) and S. senegalensis (~27 days). The scenario S2 (bT−X) can well represent the phenological adaptation of fish spawning earlier in the season (to experience lower temperatures in the future, akin to current temperatures at spawning) but, due to a phenological mismatch with the (rest of the) planktonic community, experience a lower food availability.

4.4. Further Research

4.4.1. Temperature Tolerance Shifts during Ontogenesis—Parametrization of T_A

Changes in thermal tolerance along the life cycle imply that different developmental stages might react differently to temperature fluctuations. A narrow tolerance range suggests that the individual is exposed to small temperature changes, and this, in turn, is usually linked to a relatively high T_A value. In DEB models, the Arrhenius temperature parameter (T_A) describes how metabolic rates are affected by temperature. The value of the T_A is the resulting slope after plotting the log of any metabolic rate against the inverse absolute temperature [20]. The T_A is typically low (~6 kK) for species that experience extreme temperature changes, like in the intertidal zone, and higher (~12 kK) for species that experience temperature changes in a narrower range as would be, for instance, pelagic species [20].

One of the challenges we encountered while parametrizing the Solea spp. models was related to the T_A parameterization, which seemed to be different for larvae vs. adult individuals. More precisely, data for developmental rates (hatching and birth) and growth in larvae suggest a higher slope of ~10–12 kK, while data describing juvenile and/or adult growth rates result in a slope of ~7000 K. As we could not source sufficiently detailed data on various metabolic rates for larvae, juveniles, and adults, we opted for a constant (average) T_A value estimated by DEBtool parameter estimation routines (Table 2), which results in acceptable predictions for both larvae and later life stages (Appendix B). Having said that, we did put a stronger emphasis on larval data (as early life stages are the focus of this study), either by including more data for early life stages during parameter estimation (S. solea) and/or by giving higher ‘weights’ to early life-stage data during parameter estimation (both species). The estimated value of T_A is around 8000 K for S. solea and 6500 K for S. senegalensis (Table 2), compared with T_A of around 4500 K for S. solea in work by Mounier et al. [38]. As a result, our model predictions for S. solea match better data for early life stages, which is the opposite situation than in work by Mounier et al. [38], who focused on later life stages, while sacrificing the fits on early life stages. For S. senegalensis, the T_A value might have been higher if more larval development data had been included, but the only other available dataset [86] did not match other data for the species and was therefore excluded.

In general, it is assumed that T_A, like most other parameters, remains constant throughout the ontogeny of an individual and is species-specific [20]. The same holds for other parameters (T_AL, T_AH, T_L, T_H) that together define the optimal thermal niche, critical high and low temperature, and physiological responses to temperature shifts. However, evidence suggests that thermal tolerance shifts during the life cycle of fish species [87] and that the optimum temperature for growth decreases with the increasing body size of juveniles and adults [14]—a difference in thermal responses of life stages suggested also by our analysis. Thermal tolerance range might differ between fish larvae and adults due to respiration capacity varying between developmental stages [88]. In that sense, the changes in thermal tolerance during the life cycle correspond to the development of aerobic capacities and cardiorespiratory organs. Specifically, tolerance ranges are hypothesized to widen from embryo to larval and adult stages. During reproduction (spawning stage), tolerance ranges may narrow again due to a net decrease in the aerobic capacity associated with additional energy and, thus, O2 demand for gamete production and biomass [87]. This hypothesis supports the common observation that shows larvae have the highest sensitivity to environmental stressors [82]. It would also explain the differences observed while calibrating the T_A parameter using larval or juvenile/adult data. This could be included as a simple model extension, akin to the modification enabling several parameters to change their value throughout ontogeny in the abj-type model, [36,37] also applied here. The thermal response might also be population specific, where (sub)tropical populations tolerate higher temperatures which cause stress in colder-water populations of the same species [69,89]. However, for a meaningful model extension, detailed experiments and data are needed, particularly data on metabolic rates (e.g., development, growth, ingestion, respiration) for all life stages of both species under controlled feeding regimes and several different temperatures. Such data were not available at the time of the study (so constant T_A was assumed as a good approximation) but developing life stage- or population-specific temperature responses could be a valuable study direction, especially in the context of climate perturbations, including heat waves.

4.4.2. Survival Calculations Made Explicit

We have focused on energetic pathways and consequences of environmental perturbations (temperature increase, food availability increase or decrease) on the growth and development of sole individuals. This is an appropriate tool for studying fish recruitment, as developmental changes require energy and allocation, and the critical size for switching between life stages can be quantitatively linked to DEB parameters and modeled processes [90]. However, larvae survival modeling simulations should also be considered for forecasting recruitment. DEB (individual) models can easily be coupled to a population model that includes survival, as was done, e.g., for the early life stages of bluefin tuna [91]. As another example based on soles, Van De Wolfshaar and collaborators [92] estimated the number of viable S. solea larvae surviving the phase from the egg until settlement in a nursery by calculating the survival of post-settlement juveniles for a particular set of estuaries, years, and climate scenarios using the length-based mortality rate function. The authors based their calculations on temperature-dependent mortality losses [92]. Because size determines mortality rate and early arrival is associated with low growth rates, early arrivals stay small for a longer period than late arrivals, and post-settlement survival of late arrivals is, therefore, higher than for early arrivals. We believe that by coupling the proposed modeling framework to a larval dispersion model, as in [92,93], we could provide detailed information on recruitment success for specific locations/nursery grounds.

5. Conclusions

Given the fast pace of change in the ocean environment due to climate change and pollution, it is likely that the status of the future soles will not be improved. The rapid development of fish larvae in warmer waters requires highly nutritive and available food, but warmer oceans—due to stratification and a change in predominant plankton communities—might exhibit long and inefficient food webs with poor nutritional quality [7]. This will impact the food availability of mothers, larvae, and juveniles. As a result, we could anticipate two scenarios, one with smaller egg size and another with larger but fewer eggs, larvae from which will again face energy-impoverished environments. These scenarios logically denote a more extended period to reach metamorphosis and, thus, longer PLD and a reduction in survival and recruitment, respectively.

Modeling tools are necessary and practical for predicting the effects of climate change on fish recruitment. The DEB models calibrated and validated here showed potential, particularly by fitting the data very well. When confronting the model predictions to the experimental data used for validation, where feeding frequency is used as a food availability proxy, we noticed that fish were not fed until satiation, suggesting that the feeding regime was potentially insufficient in nutrition and quantity. This result points to the need to adjust and improve experimental larvae-rearing protocols, particularly regarding providing enough energy without compromising the larvae’s health and rearing conditions.

Regarding timing for reaching a developmental stage, the effects of temperature rise seemed to counterbalance the effect of reduced food availability. Our results indicate the critical effect of temperature rise on the phenological cycles of both species, mainly by advancing the onset and the completion of metamorphosis. The growth simulations in both species indicate shorter pelagic stages responding to increased temperatures. However, changes in the PLD, which are a priori beneficial for larvae survival, when accompanied by a potential mismatch between hatching and the zooplankton cycle timing, result in more extended periods for maturation and smaller ultimate size. In that regard, the combined temperature-rise and food-reduction conditions simulated in scenarios 6 and 7 strongly affected the maturation age of S. senegalensis, which was significantly prolonged, notably for the latest batches, and would result in prolonged use of the nursery areas. This result might have important consequences in population renewal as it will take longer for the cohort to reach maturity and spawn new eggs. Additionally, we believe that coupling the proposed modeling framework to a larval dispersion model, like in [93], would provide detailed information on recruitment success for specific locations/nursery grounds.

To get a more holistic (and realistic) conclusion on the subject, it is necessary to include potential temperature-induced damage and mortality in the model, which might change between different life stages and/or populations [69,89]. Without including these potential negative effects, the overall conclusion is that temperature rise reduces the pelagic larvae’ duration, potentially increasing the survival rates of larvae and, thus, recruitment. Additionally, survival should be explicitly included in the models, accounting for physiological temperature tolerance ranges, as well as size and habitat-linked mortality.

Finally, the survival and ontogeny of fish larvae may directly influence the future abundance of adult fish stocks. The latter has been, and still is, the most important reason for studying ichthyoplankton, as most processes determining the recruitment strength and the spatial distribution of fish populations occur during the planktonic stage of fishes, resulting in critical interannual fluctuations in fish stock biomasses. While questions on recruitment success and larvae survival were not our focus and thus remain unanswered, understanding the physiological mechanisms behind temperature increase and food limitation via DEB models provides a powerful tool for studying the role of early-life history traits on population renewal under climate change.