Analysis of Respiratory Behaviour of Thicklipped Grey Mullet (Chelon labrosus) Juveniles Under Different Rearing Conditions

Abstract

Thicklipped grey mullet (Chelon labrosus) shows potential as an appealing species for aquaculture in the EU. Knowledge of its metabolic requirements is essential for species management and control of environmental conditions. We examined routine and postprandial oxygen consumption (OC) in juveniles as a function of body weight (Bw: 2–85 mg) and temperature (T: 14–26 °C), as well as OC and ventilatory frequency (VF) under gradual hypoxia as a function of T (14–22 °C). Multiple regression analyses determined the effects of Bw and T on mean daily (OC_mean), postprandial (OC_SDA), routine (OC_routine), and maximum (OC_max) levels, as well as on OC_max/OC_routine ratio (MSF), postprandial OC duration (D_SDA) and time to reach maximum activity (D_peak). The effects of dissolved oxygen (DO) and T on OC and of T on initial VF (VF_ini), maximum VF (VF_max), critical DO threshold (%DO_crit), and VF change threshold (%VF_ch) were also analysed. All OC levels increased with T and Bw, except MSF, D_SDA, and D_peak, uninfluenced by Bw. Under gradual hypoxia, OC decreased with falling DO, more sharply at higher T, consistent with oxyconformer behaviour. VF remained stable until 50% DO, then rose progressively, reaching higher VF_max at higher T. Simulations using derived equations estimate C. labrosus respiratory response and water flow requirements under aquaculture conditions.

1. Introduction

Mugilids, in particular the thicklipped grey mullet (Chelon labrosus; Risso, 1827), have attracted growing interest as candidate species for aquaculture diversification. This interest stems from their particular biological and ecological characteristics, given that C. labrosus is a low trophic level organism with omnivorous feeding habits [1,2], and its euryhaline and eurythermic nature renders it highly adaptable to different farming conditions [3], as well as its recognition in relevant markets, such as the Mediterranean and Asian [4]. These characteristics align C. labrosus perfectly with the sustainability objectives for aquaculture encouraged by the EU [5].

In order to optimise the growth performance of an emerging fish species, its metabolic requirements must be comprehended. In this case, it is of the utmost importance to do so given that said requirements vary depending on multiple environmental factors and the developmental stage of the fish [6]. In aquatic environments, and especially under aquaculture conditions, the most critical factor controlling the physiological responses and behaviour in fish is the amount of dissolved oxygen available, ultimately influencing aspects such as respiration, growth, and activity patterns [7]. In general, oxygen consumption in fish is mainly influenced by temperature and body weight [8]. In scientific literature, there are numerous studies that attest to the effect of both variables on fish oxygen consumption [9,10,11,12,13]. As in most marine fish, the effect of temperature on the metabolism is strongly reflected in the regulation of metabolic rates and energy behaviour [14]. From a thermodynamic point of view, as temperature increases, catabolic reactions accelerate, increasing the energy demand required for maintenance, growth, and reproduction [15] and therefore the oxygen need. Whereas increases in temperature raise the metabolic rate, specific oxygen consumption shows an inverse relationship with body weight, such that smaller fish present higher values per unit of mass [16,17,18,19,20]. This pattern is associated with an accelerated metabolism in smaller fish that promotes rapid growth, reducing vulnerability to predators and increasing the chances of survival [20].

Oxygen consumption is commonly used as a proxy for energy metabolism in fish [21]. In aquaculture, understanding this aspect is crucial, as it directly influences factors such as survival rate, growth, welfare, and production efficiency of the species being farmed [22,23]. The measurement of oxygen consumption allows the estimation of energy used in metabolic processes, as demonstrated in previous studies with other marine species of interest in aquaculture [17,23,24]. Likewise, being aware of the oxygen consumption is also very useful to design recirculating aquaculture systems (RASs), and so is knowing the response of cultured fish to hypoxic stress conditions.

Hypoxia in fish can lead to slower appetite and growth, reproductive disorders, histological alterations, haematological responses, cardiac dysfunctions, changes in enzyme and hormone activities, energy metabolic responses, oxidative stress adaptations and immunological responses, and/or particular gene expression [25] (and references therein), which may ultimately influence production performance under aquaculture conditions. In response to environmental hypoxia, fish usually exhibit different behavioural responses, such as escaping, lowering movement, or increasing ventilatory frequency [25,26]. At rest, most fish species maintain their ventilatory frequency and oxygen consumption constant even when dissolved oxygen levels drop, i.e., they behave as regulators [27]. Conversely, some fish species show an oxygen consumption directly related to the available dissolved oxygen and are typically referred to as conformers [28]. In any case, if oxygen availability falls below a critical level, oxygen consumption slows down, ventilatory frequency increases [17,23], and metabolism could become anaerobic [29,30], resulting in a less efficient metabolism that negatively impacts the fish’s development. Determining these thresholds is of great importance in fish aquaculture to establish the optimal water flow and the requirements for water renewal and oxygenation in recirculating aquaculture systems and to know the margin for manoeuvre in situations in which water flow or oxygen might be cut off. In this context, it is also very important to understand variations in oxygen consumption throughout the daily cycle under cultivation conditions, such as feeding or resting.

Considering that the knowledge of specific environmental conditions required for the successful grow-out of C. labrosus is currently limited, the present study assesses this species respiratory behaviour, focusing specifically on oxygen consumption and ventilatory frequency in juvenile specimens during a normal daily growing cycle, including feeding and rest periods, under different growing conditions. The selected integrated approach to different fractions of oxygen consumption enables a breakdown of energy requirements linked to basal metabolism, spontaneous activity, and response to feeding [18]. These variables represent relevant physiological indicators for the characterisation of energy demands under different environmental conditions.

2. Materials and Methods

2.1. Fish Used in the Trials

The specimens of Chelon labrosus used in this study were obtained from catches in the wild (W) on the Cantabrian coast, North Spain, and also from the breeding at our facilities (F1) in the “El Bocal” Marine Aquaculture Plant belonging to the Oceanographic Centre of Santander (Spanish Institute of Oceanography, COST-IEO/CSIC). Wild fish underwent a 40–50-day quarantine period, which also served as an adaptation period to captivity and artificial feeding. Once acclimated, both W and F1 fish were kept stocked separately in 1.5-m³ tanks in a flow-through seawater system at natural temperature (12–24 °C), salinity (35‰), and photoperiod (43°29′15.6″ N, 3°49′37.1″ W) conditions.

The fish were always handled (routine management and experimentation) according to the Guidelines of the European Union [31] and the Spanish legislation [32] for the use of laboratory animals. Moreover, every person involved in the experiments was in possession of the required FELASA accreditations for each procedure [33]. The project in which this experiment is included was submitted for official ethics committee review, obtaining a favourable report with the number NTS-ES-285239.

2.2. Experimental Setups and Procedures

2.2.1. Oxygen Consumption Trials

These trials aim to ascertain the influence of body weight (Bw) and temperature (T) on oxygen consumption (OC) in C. labrosus juveniles at different relevant stages of the daily growing cycle, namely feeding and rest (including spontaneous activity), which is of interest to establish the optimal rearing conditions for this species. OC trials were performed between August 2023 and March 2025. For all experimental configurations, fish from the stock (wild or F1) were selected to form homogeneous experimental groups (CV < 20%) for each trial. Fish weighing between 2 and 85 g fresh Bw were used in the trials. Fish were transferred from the stocking tanks to their respective experimental tanks according to their Bw. The trials did not begin until the fish were fully adapted and showing a consistent daily feeding pattern [17]. Photoperiod was set as 14L:10D. The fish remained in the experimental tanks for at least 2 weeks to acclimatise prior the experiments and were starved for 24 h before the trials started. A total of 63 OC measurements were conducted according to the methodology described in Cerezo-Valverde et al. [23], albeit with some modifications. In all trials, OC measurements were taken hourly over 2 consecutive days, starting at 08:00 on the first day and finishing at 14:00 on the second day, for a total of 30 h. Feed was given at 11:00 and 16:00 on the first day only (1.5% of the fish biomass, divided equally between each feeding). According to Cerezo-Valverde and García-García [17], floating plastic sheets were used to cover the surface of the tanks in all trials to prevent oxygen diffusing from the atmosphere into the water (O₂ diffusion is negligible), and aeration was dispensed with.

Three experimental setups were used depending on the size of the fish, as explained in Supplementary Materials (Method S1). In any case, OC is expressed as mg O₂ h⁻¹ based on the number of fish in each tank. After each trial, all fish were weighed again, and the mean weight was recorded. OC was plotted over time for each trial (63 in total). Figure 1 shows how OC measurements were partitioned over time to define data analysis and the variables to be considered:

• Mean daily OC (OC_mean): mean OC over 24 h, including feeding periods (from 11:00 on the first day until 11:00 on the second day), in mg O₂ h⁻¹.
• Postprandial OC (OC_SDA): frequently referred to as the specific dynamic action (SDA; [34] and references therein), it is the OC related to feeding calculated as the mean value of the OC values from the time the first feed is given until the routine level is reached, in mg O₂ h⁻¹.
• Routine OC (OC_routine): mean OC excluding OC_SDA. It is equivalent to standard metabolism and spontaneous activity, in mg O₂ h⁻¹.
• Maximum OC (OC_max): maximum OC peak, usually during feeding time, in mg O₂ h⁻¹.
• Metabolic scope due to feeding (MSF): relationship between OC_max and OC_routine.
• Duration of OC_SDA in hours (D_SDA).
• Time from the first feeding until OC_max is reached (D_peak), in h.

2.2.2. Ventilatory Frequency Trials

The aim of these trials is to understand how OC and ventilatory frequency (VF) are influenced by a progressive decrease in available DO, as well as the effect of T. The results will enable us to determine the critical oxygen saturation values that cause changes in CO and VF according to temperature. Trials were conducted according to experimental configuration 2 described in Method S1. Fish with a mean Bw of 100 g remained in the experimental tanks for at least 2 weeks to acclimatise prior the experiments and were starved for 24 h before the trials started. The trial procedure was adapted from that in Cerezo-Valverde et al. [23] and is described in Method S2. Trials were conducted at three different temperatures: 14 °C (5 fish per tank), 18 °C (3 fish per tank), and 22 °C (3 fish per tank).

The following variables (Figure 2) were analysed according to Cerezo-Valverde and García-García [17,35] and Cerezo-Valverde et al. [23]:

• OC: as mentioned in Method S1 for experimental configuration 2.
• Initial VF (VF_ini): initial ventilatory frequency at 100% oxygen saturation (normoxia).
• Critical DO threshold (%DO_crit): percentage of oxygen saturation at which a change in OC is identified. The determination of this value in each trial is based on the intersection of two regression lines, one in which the VF values were roughly constant and another in which they increased.
• Maximum ventilatory frequency (VF_max) reached by a specimen.
• DO (% sat.) threshold that triggers a change in VF (%VF_ch): intersection of the regression lines in which the VF values that remain constant meet those showing an increase.

2.3. Statistical Analysis

2.3.1. Oxygen Consumption Trials

To assess the influence of the independent variables (Bw and T) on D_SDA and D_peak, partial correlation coefficients were calculated. Partial correlation coefficients for OC_mean, OC_SDA, OC_routine, OC_max, and MSF and independent variables were also launched as a first step to establish the significant independent variables to be included in a regression analysis. Next, a multiple regression analysis was performed to understand the relationship between the dependent variables and the significant independent ones to estimate the former as a function of the latter. The OC data were fitted by multiple regression to Liao’s reference model [36]:

LnOC = Ln a + b ∙ Ln T + c ∙ Ln Bw,

where Ln means natural logarithm, Bw and T are mean body weight and temperature, respectively, and a, b, and c are the parameters in the equation. The natural log was added to or removed from the variables to find the best fit. The significance of the models was determined using analysis of variance (ANOVA); the simple (R²) and adjusted coefficient of determination (R² adj.), as well as the standard error of the estimates (SEE) were calculated. Only models with the highest explained variability are shown.

2.3.2. Ventilatory Frequency Trials

Similar to the statistical approach used for OC trials (Section 2.3.1), partial correlation coefficients were calculated between OC and DO saturation and T as independent variables and between T and VF_ini, %DO_crit, VF_max, and %VF_ch. Next, a simple or multiple regression analysis was performed on the variables that showed significant correlation. The natural log was added to or removed from the variables to find the best fit. Significance of the models was determined using ANOVA; R² or R² adj. and SEE were calculated. Only models with the highest explained variability are shown.

All statistical analysis were performed with SPSS v26.

2.3.3. Simulations

Simulations were performed using equations obtained through simple or multiple regression analysis to better understand the effect of independent variables on dependent ones. Similarly, to demonstrate the effectiveness of OC models in C. labrosus farming, simulations were performed to determine the necessary water flow rate (renewals per hour) within the studied Bw and T ranges for various rearing densities using the OC_mean model. Simulations assumed the DO in the incoming water to be 100% and the DO at the outlet to be 60%.

3. Results

3.1. Oxygen Consumption Trials

The OC_mean, OC_SDA, OC_routine, OC_max, MSF, D_SDA, and D_peak data obtained in all trials are summarised in Table S1 (Supplementary Material) together with the mean Bw and T values corresponding to each trial. Figure 1 is representative of the OC results of all the experimental configurations, as well as Bw and T ranges, and helps us to explain the development of the trials in general. It shows an increase in OC just after the first feeding event maintained or even increased (OC_max) after the second feeding event and then gradually decreased (OC_SDA) until reaching a value maintained over time until the end of the trial, which corresponds to OC_routine. This pattern is consistent across all 63 trials conducted. In all trials, as expected, the OC_SDA was between 1.5 and 2.1 times higher than the OC_routine. Likewise, OC_max was 1.9–2.7 times higher than OC_routine.

A significant correlation was obtained for Bw with OC_mean, OC_SDA, OC_routine, and OC_max and for T with all dependent variables (

Table 1. Partial correlation coefficients obtained between independent (Bw and T) and dependent variables (OC_mean, OC_SDA, OC_routine and OC_max; MSF, D_SDA, and D_peak). (*: p < 0.05; **: p < 0.01; n.s.: non-significant).

	OCmean	OCSDA	OCroutine	OCmax	MSF	DSDA	Dpeak
Bw	0.58 **	0.58 **	0.65 **	0.55 **	0.06 n.s.	−0.10 n.s.	−0.01 n.s.
T	0.59 **	0.57 **	0.52 **	0.58 **	0.27 *	0.71 **	0.31 *

). MSF values relating to OC_max/OC_routine ranged from 1.18 to 4.03, increasing as Bw and/or T increased. D_SDA (8–21 h) and D_peak (1–11 h) values were also greater the higher the Bw and/or T became. In other words, in smaller fish and/or at lower temperatures, the duration of D_SDA is shorter and D_peak is reached earlier, returning to routine mode sooner.

The equations obtained through multiple regression analysis (Table 2) show that OC_mean, OC_SDA, OC_routine, and OC_max tend to increase as T and Bw increase. Simulations performed using the resulting equations show this relation between dependent and independent variables (Figure 3). Overall, as expected, OC_mean was always between OC_SDA and OC_routine, with OC_max being higher than the other OC levels. Simulations revealed that OC_mean, OC_SDA, and OC_max increase progressively as the temperature rises, reaching values around five times higher at 26 °C than at 14 °C and four times higher in the case of OC_routine. Similarly, as Bw increases, OC_mean, OC_SDA, OC_routine, and OC_max increase, reaching values in fish weighing 85 g around 8.5 times higher than those in fish weighing 5 g.

Ln OC_mean = −8.496 + 2.611 ∙ Ln T + 0.760 ∙ Ln Bw

(1)

Ln OC_SDA = −8.187 + 2.561 ∙ Ln T + 0.765 ∙ Ln Bw

(2)

Ln OC_routine = −7.599 + 2.239 ∙ Ln T + 0.762 ∙ Ln Bw

(3)

Ln OC_max = −8.106 + 2.628 ∙ Ln T + 0.756 ∙ Ln Bw

(4)

Table 2. Results of multiple regression analysis for OC_mean, OC_SDA, OC_routine, and OC_max with Bw and T as independent variables. Only the results of the equations that provided the best fit for each of the dependent variables are shown. (a–c: coefficients; SEE: standard error of estimation; ***: p < 0.001).

Dep. Var.	a	b	c	R	ANOVA
	SEE	SEE	SEE	R2 adj.	F
	p-Value	p-Value	p-Value	SEE	p-Value
Equation (1): OCmean	−8.496 0.379 ***	2.611 0.119 ***	0.760 0.024 ***	0.977 0.953 0.183	624.481 ***
Equation (2): OCSDA	−8.187 0.490 ***	2.561 0.154 ***	0.765 0.032 ***	0.962 0.923 0.237	371.657 ***
Equation (3): OCroutine	−7.599 0.320 ***	2.239 0.101 ***	0.762 0.021 ***	0.982 0.963 0.155	810.238 ***
Equation (4): OCmax	−8.106 0.584 ***	2.628 0.184 ***	0.756 0.038 ***	0.947 0.893 0.283	260.997 ***

Table 2. Results of multiple regression analysis for OC_mean, OC_SDA, OC_routine, and OC_max with Bw and T as independent variables. Only the results of the equations that provided the best fit for each of the dependent variables are shown. (a–c: coefficients; SEE: standard error of estimation; ***: p < 0.001).

Dep. Var.	a	b	c	R	ANOVA
	SEE	SEE	SEE	R2 adj.	F
	p-Value	p-Value	p-Value	SEE	p-Value
Equation (1): OCmean	−8.496 0.379 ***	2.611 0.119 ***	0.760 0.024 ***	0.977 0.953 0.183	624.481 ***
Equation (2): OCSDA	−8.187 0.490 ***	2.561 0.154 ***	0.765 0.032 ***	0.962 0.923 0.237	371.657 ***
Equation (3): OCroutine	−7.599 0.320 ***	2.239 0.101 ***	0.762 0.021 ***	0.982 0.963 0.155	810.238 ***
Equation (4): OCmax	−8.106 0.584 ***	2.628 0.184 ***	0.756 0.038 ***	0.947 0.893 0.283	260.997 ***

Figure 3. Simulations of OC_mean, OC_SDA, OC_routine, and OC_max performed using the corresponding equations (Equations (1)–(4); Table 3) for the T range studied and for three Bw within the range studied.

Figure 3. Simulations of OC_mean, OC_SDA, OC_routine, and OC_max performed using the corresponding equations (Equations (1)–(4); Table 3) for the T range studied and for three Bw within the range studied.

Table 3. Partial correlation coefficients obtained between DO saturation and T as independent variables and OC, and between T and % DO_crit, VF_max and %VF_ch. (**: p < 0.01; n.s.: non-significant).

	OC	%DOcrit	VFini	VFmax	%VFch
T	0.754 **	―	0.57 n.s.	0.92 **	0.23 n.s.
DO (% sat.)	0.393 **

Table 3. Partial correlation coefficients obtained between DO saturation and T as independent variables and OC, and between T and % DO_crit, VF_max and %VF_ch. (**: p < 0.01; n.s.: non-significant).

	OC	%DOcrit	VFini	VFmax	%VFch
T	0.754 **	―	0.57 n.s.	0.92 **	0.23 n.s.
DO (% sat.)	0.393 **

The simulation performed with the OC_mean model (Equation (1)) to estimate the renewal flow rate required for the cultivation of C. labrosus juveniles at different temperatures and rearing densities (Figure 4) reveals flow rates higher the lower the Bw gets (the number of specimens for the same density is higher the smaller the specimens) and the higher the T gets. Thus, if a flow rate < 2 renewals h⁻¹ is to be maintained and T maximised, for specimens with Bw = 5 g, a density of 10 kg m⁻³ should not be exceeded, while for specimens with Bw = 45 g, a density of up to 15 kg m⁻³ could be reached. For specimens with Bw = 85 g, a density of even almost 20 kg m⁻³ could be reached without compromising the oxygenation of the rearing environment.

3.2. Ventilatory Frequency Trials

Results of OC and VF are summarised in Table S2 (Supplementary Material). A consistent pattern was observed in all these trials, consisting in a reduction in OC as DO decreases (Figure 5). Consequently, no % OS_crit values were observed, suggesting that C. labrosus behaved as a conformer organism. The OC results in this trial and the partial correlation test (Table 3) confirmed the results of the previous trial, i.e., that OC significantly increases as T increases, and that OC decreases as DO (% sat.) decreases. Multiple regression analysis provided Equation (5) (

Table 4. Results of simple and multiple regression analysis for OC with T and DO (% sat.) as independent variables and for VF_max with T as independent variable. Only the results of the equations that provided the best fit for each of the dependent variables are shown (a–c: coefficients; SEE: standard error of estimation; **: p < 0.01; ***: p < 0.001).

Dep. Var.	a	b	c	R	ANOVA
	SEE	SEE	SEE	R2:R2 adj.	F
	p-Value	p-Value	p-Value	SEE	p-Value
Equation (5): OC	−2.745 0.177 ***	1.603 0.061 ***	0.006 0.001 ***	0.867 0.751 0.179	419.467 ***
Equation (6): VFmax	3.211 0.191 ***	0.055 0.011 ***	― ― ―	0.919 0.845 0.088	27.329 **

), which allows the OC to be estimated based on DO (% sat.) and T for specimens with Bw = 100 g. Simulation performed with Equation (5) confirmed the pattern mentioned above (Figure 6). Furthermore, under normoxic conditions (90–100% saturation), the OC values at the temperatures tested in this trial (7.75, 12.35, and 16.98 mg O₂ h⁻¹ at 14, 18, and 22 °C, respectively) were very similar to the OC_routine values estimated with the model developed in the previous trial for fish weighing 100 g (6.16, 10.82, and 16.96 mg O₂ h⁻¹ at 14, 18, and 22 °C, respectively).

The partial correlation test (Table 3) revealed that only VF_max was significantly and positively correlated to T. This means that the higher the T, the higher the VF_max, which, under our experimental conditions, was reached at the lowest DO achieved, around 25–30%. At the beginning of the trials under normoxic conditions, VF_ini was around 50 beats min⁻¹ in the bulk of trials, regardless T. VF remained roughly constant over the course of the trials until DO dropped to around 50% (Figure 5), with no significant effect of T on this variable (Table 4). From this critical level (VF_ch ≈ 50%), VF increased until it reached VF_max at the end of the test (Figure 5). VF_max can be estimated for individuals with Bw = 99.5 g as a function of T with Equation (6) (Table 4). The simulation performed with Equation (6) showed that the abovementioned relationship between VF_max and T (Figure 7) for individuals with Bw = 99.5 g: VF_max was 1.25 and 1.55 times greater at 18 °C and 22 °C than at 14 °C, respectively

Ln OC = −2.745 + 1.603 ∙ Ln T + 0.006 ∙ DO (% sat.)

(5)

Ln VF_max = 3.211 + 2.611 ∙ T

(6)

4. Discussion

In many freshwater, brackish, and marine fish, Bw and T represent key factors in determining OC [15,17,37,38,39,40,41], and this is also true for Chelon labrosus. The OC models obtained in the present study provide insight into the impact of these variables on the development of C. labrosus juveniles, thus enabling their different OC levels to be predicted. This provides valuable resources for the production management of the species under aquaculture conditions. The analysis of the ventilatory behaviour suggests a conformer performance of C. labrosus juveniles in terms of OC under conditions of progressive hypoxia. This controversial aspect will be discussed further on.

In the present work, the Bw coefficient for the different OC levels considered ranges between 0.75 and 0.77, which lies in the middle of the range observed in many fish species [8,11,42] (0.40–1.30). The values obtained for this coefficient by García-García [43] for gilthead seabream (Sparus aurata) and García-García et al. [44] for sharpsnout seabream (Diplodus puntazzo) were slightly lower than those in the present study. The latter authors suggested that a higher Bw coefficient is related to a higher metabolic rate, meaning that C. labrosus would show higher energy expenditure than that for those species.

While the effect of Bw on OC is related to the developmental stage and overall metabolism, the effect of T is related to the modulation of physiological processes and behaviour [6,18], covering processes from a cellular to an ecological scale from single-celled organisms, plants, and ectothermic and endothermic animals without distinction [45,46]. It has been widely documented in fish that an increase in temperature within tolerance ranges is associated with a boost in metabolic rate, feed intake, growth, and consequently OC [19,34,47,48,49,50,51,52]. The present study shows how an increase in T significantly delays the appearance of D_peak after feeding and extends the duration of the effect of feeding on metabolism (D_SDA). However, Guinea and Fernández [49] did not find a significant relationship between these variables in Chelon saliens (syn. Mugil saliens), a species related to C. labrosus, despite a tendency to increase with T. Nevertheless, these authors did observe a tendency for the maximum peak of OC to take longer to appear with the increase in T. Both studies, as well as the one by Niklitschek and Secor [53], showed feeding as the main cause of variation in OC throughout the day, or at least it proved to be important enough to mask any other daily rhythm.

The models developed here demonstrate the usefulness of these tools for managing aquaculture facilities by allowing water flow rates to be adjusted to environmental and husbandry conditions. The simulations performed in the present work revealed that, for C. labrosus juveniles, the lower their Bw, the higher the water flow renewal rates were needed to rise T if the stocking density increased. In flow-through aquaculture, water oxygenation mostly occurs with the inflow of new water, while in RASs, this occurs through water surface aeration, renewal with external water or liquid oxygen injection. Therefore, OC modelling is particularly important under RAS conditions, especially considering that these production systems conceptually tend to minimise the renewal rate. This means that a thorough monitoring of DO levels in RASs at all times is essential to ensure the welfare of the organisms being cultivated and to maximise production efficiency [54]. The current trend is to incorporate modelling into automated systems to control various variables relevant to cultivation in RAS, such as T, pH, and DO [55]. In this regard, the equations obtained in the present study are particularly useful for the cultivation of C. labrosus in RASs.

The most noteworthy, albeit controversial, finding of this study regarding the OC and VF trials is that C. labrosus juveniles showed a progressive decrease in OC as DO decreased, which is consistent with conformer behaviour in response to progressive hypoxia. This is quite unusual, since the most common behaviour in fish subjected to hypoxia is to maintain their metabolic rate up to a critical point, beyond which their metabolism slows down, i.e., they exhibit regulatory behaviour [27,30,56]. Steffensen [27] indicates water stratification and the resulting accumulation of metabolites (CO₂, NH₄⁺), spontaneous activity, and stress to be the key factors that can influence the results of respirometry experiments in closed systems such as the one used in our study. To minimise potential sources of error under our experimental conditions, we took several steps, such as to provide an acclimatisation period long enough for full adaptation (until normal feeding) to be achieved, to install complete visual isolation in the respirometer walls to avoid external sources of stress that could cause sudden activity, and to incorporate a pump/filter into each respirometer to prevent stratification and remove ions. However, this filter is not capable of removing the CO₂ released into the environment, which could interfere with the determination of OC. Nevertheless, CO₂ can begin to interfere with oxygen uptake only when hypoxia becomes severe [57], something that did not happen in our trials (only could have occurred at the end of some trial in the worst-case scenario).

In an oxyconformer fish such as sturgeon (Acipenser transmontanus), Burggren and Randall [58] observed that VF also decreased as DO and OC decreased, which is consistent with their adaptive mechanism, namely simply reducing the metabolic rate instead of entering anaerobic metabolism (as occurs in oxyregulator fish). Conversely, in our study, C. labrosus exhibited a ventilatory behaviour common to other oxyregulator fish, consisting in maintaining a constant VF until DO fell below ≈50% saturation (%VF_ch), at which point it increased until reaching a maximum at the end of the trials. In Diplodus puntazzo and Dentex dentex, both species coexisting with C. labrosus, %VF_ch is noticeably higher (66.5 and 64.7%, respectively) [17,23], which would indicate that C. labrosus is better adapted to inhabit hypoxic environments. In any case, oxyconformer behaviour is uncommon and is often determined by the masking of species-specific behaviours. This is the case of the Adriatic sturgeon (Acipenser naccarii), which behaves as an oxyconformer in static conditions but also as an oxyregulator when allowed to swim at a low sustained speed [59]. Mullets in general, and C. labrosus in particular, inhabit estuarine and lagoon environments prone to hypoxia events, and, throughout their evolution, they have developed adaptations such as aquatic surface respiration (ASR) [60,61]. Under our experimental conditions, C. labrosus juveniles were unable to display this behaviour in response to gradual hypoxia, as the floating plastic sheet placed on the surface of the respirometers to prevent gas exchange with the atmosphere prevented them from doing so. This impediment may have led to atypical behaviour, with C. labrosus acting as an oxyregulator in circumstances under which it can deploy its full range of mechanisms against hypoxia, similar to the case of the Adriatic sturgeon mentioned above.

Oxyconformer fish adjust their consumption and metabolism to the oxygen available in water rather than keeping it constant like oxyregulators do. This behaviour offers advantages, such as energy savings in cases of hypoxia, but also disadvantages, as it prevents the fish from maintaining a normal growth rate under these conditions [62]. Although the present study is far from being the first case in which rigorous experimentation has demonstrated oxyconformer behaviour, previous research into this aspect has revealed findings that highlight the ability to regulate OC under particular circumstances [27]. This opens the door to further research in this area with species such as C. labrosus.

5. Conclusions

As with so many fish species, the different OC levels (daily average, maximum, postprandial, and routine) in juvenile Chelon labrosus can be predicted based on their Bw and T. The models developed also allow the water renewal flow requirements to be determined based on fish density, their Bw, and T. These tools are very useful for managing the farming conditions of this species, particularly in RASs. Under conditions of gradual hypoxia, C. labrosus showed consistent oxyconformer behaviour with respect to DO, although the ventilatory behaviour displayed corresponds to an oxyregulator response. As has been demonstrated in other species such as the Adriatic sturgeon, C. labrosus may have behaved as an oxyconformer because it was unable to display normal behaviour; in the case of C. labrosus this may have occurred due to its ASR capacity, but this is more of a hypothesis than a conclusion.