Impact of Temperature Reduction from 14 °C to 12 °C in RASs on Atlantic Salmon: Increased Mineral Accumulation in RASs and Enhanced Growth Post-Transfer to Seawater

Abstract

Robust, healthy, and fast-growing smolt is of high importance for fish farmers as a way of reducing the mortality and production time of Atlantic salmon (Salmo salar) in open sea pens. Lowering the water temperature in flow-through systems (FTSs) compared to recirculating aquaculture systems (RASs) has shown promising results for the growth and health of fish post-transfer to sea; unfortunately, limited information is available on the same parameters in replicated RAS setups. Hence, the current study aimed to compare the performance of Atlantic salmon reared at 12 and 14 °C over a 9-week RAS period and a subsequent 10-week post-transfer period, while also investigating the accumulation pattern of minerals in RASs. The results showed a 100% survival and comparable condition factors and cardiosomatic index (CSI) across both temperatures. During the RAS period, the thermal growth coefficient (TGC) was higher at 12 °C, but body weight gain and feed consumption were lower. No differences in mineral retention or fecal stability were observed. However, the production water accumulated more dissolved phosphorus (DP) and total iron (Fe) at 12 °C. Post-transfer, the TGC remained higher for fish with a rearing history of 12 °C. This fish also had higher body weight gain and feed consumption while exhibiting a lower hepatosomatic index (HSI) and viscerosomatic index (VSI), indicating improved muscle growth. Overall, lower water temperature reduced growth and increased DP and Fe in RASs. However, it improved post-transfer weight gain of muscle tissue, highlighting its benefits for seawater performance.

1. Introduction

Effluents generated from aquaculture production are well known to have potential detrimental effects on water quality and the biodiversity of local fauna [1,2]. This is because aquaculture effluent is rich in nutrient substances like phosphorus (P) and nitrogen (N) and can therefore pose a risk of eutrophication in natural waterbodies when released in excessive amounts [1,2,3]. Due to the potential negative environmental impact, N and P are regulated in the global aquaculture industry, especially for countries in the European Union (EU), where farmers are subject to strict guidelines in terms of effluent regulations on what and how much can be emitted [4,5,6,7]. While N originates from protein digestion in fish [8,9], P, among other things, is a key bone-forming mineral, along with calcium (Ca) and magnesium (Mg), and is thus an essential part of the diet for several aquaculture species [8,10]. The strict regulations on N and P also motivate salmon farmers to invest in technology-intensive farming solutions as a way of reducing and controlling their emissions. One such solution is the recirculating aquaculture system (RAS), which has improved the operational sustainability of fish production by reducing the environmental impact compared to traditional flow-through systems (FTSs) and has generally increased biomass productivity in fish farming [11,12,13]. The multiple benefits of RASs can be credited to the system’s capability of temperature control, reduced water consumption, and its potential for effluent discharge treatment [14,15,16].

The high reuse capability of RASs can cause the production water within the system itself to contain higher levels of total suspended solids (TSS), minerals, metals, nitrogenous compounds, and total gas pressure (TGP) compared to a FTS [17,18,19,20]. This can, in a worst-case scenario, pose a risk to the operation of the RAS and the safety of the fish. In most aquaculture systems, nitrate-N (NO₃⁻) concentrations are below 50 mg L⁻¹, but in RASs with low water exchange rates, the concentrations can exceed 400 mg L⁻¹ [21,22]. For the production of salmonids, elevated levels of NO₃⁻ in the production water can reduce growth, health, and general performance or increase the mortality of the fish group [23,24,25]. Conversely, the concentration of P in the production water exceeding the recommended levels of 3+ mg L⁻¹ [18] does not necessarily affect fish performance in a negative manner [26]. However, in terms of waste production, elevated P levels increase the risk of eutrophication when combined with elevated N concentrations [1,2,3]. It is important to remember that even in a RAS with a high degree of water reuse, the nutrients that are not removed by the drum filter or foam fractionator, will accumulate in the production water and ultimately be released from the facility in a highly concentrated manner.

While RASs offer operational benefits, the impact of temperature manipulation on fish performance remains understudied. For the production of Atlantic salmon (Salmo salar), the technical system itself will not necessarily affect the performance of the fish post-transfer as long as environmental parameters are kept equal [27]. However, manipulating the temperature can significantly affect the performance of the fish post-transfer to seawater due to its effect on the general biology of the fish, such as heart weight and intestinal fat accumulation [28,29]. Growth rate, feed intake, FCR, and stomach evacuation rate are all affected by temperature [30,31,32]. These parameters are also directly linked to the waste production of fish and, thus, the nutrient release to the production water [33,34,35]. While optimal temperature for freshwater smolts up to 300 g ranges around 14 °C, the optimal temperature for the feed conversion ratio (FCR) is 2 °C lower [31,36]. Since post-transfer seawater performance studies have primarily been conducted by comparing FTSs to RASs, it could be useful for the industry to understand how temperature manipulation within these ranges relates to a pure RAS comparison.

The effect of temperature in RASs has also been documented to strongly correlate with total ammonia nitrogen (TAN) and nitrite (NO₂⁻) removal rates in the biofilters, where nitrification rates decrease with decreasing temperature [37,38,39]. Moreover, the biophysical characteristics of feces, such as settling velocities, from channel catfish (Ictalurus punctatus) and bighead carp (Aristichthys nobilis) vary with temperature, affecting how waste disperses and accumulates [40]. Increased particle size and improved fecal stability can be crucial to better encapsulate the particle fraction of N and P compounds until they can be removed by drum filtration [41,42], thus reducing the total emission.

The primary objective of this experiment was, therefore, to compare fish performance of Atlantic salmon reared at a water temperature of 14 °C and a cooler temperature of 12 °C over a 9-week RAS period and a subsequent 10-week post-transfer seawater period in FTSs. A secondary objective was to determine the accumulation pattern of minerals in the production water of the RAS operated at the two water temperatures.

2. Materials and Methods

2.1. Atlantic Salmon and Trial Design

Atlantic salmon post-smolt used in this study were derived from fertilized eggs provided by Aquagen (Atlantic QTL-innOva SHIELD, Aquagen, Kyrksæterøra, Norway) and were hatched and raised on site (Cargill Innovation Center, Dirdal, Norway). The salmon were vaccinated (40–50 g) (Alpha Ject Micro 6, PharmaQ, Norway), pit-tagged, and smoltified (at 70 g using a diet containing salt [EWOS ADAPT FLEX 40, Cargill, Florø, Norway]). On 19 October 2022, a total of 840 post-smolts (240 ± 35 g) were randomly stocked into six RAS units and reared at water temperatures of 12 °C and 14 °C (n = 3). The fish were housed under 24 h light in brackish water (approximately 14‰). The initial biomass density was 33 kg m⁻³ per RAS tank, and the fish were reared for 9 weeks under the experimental temperature conditions. The fish were transferred into a FTS for a post-transfer period after the 9-week RAS period. During this post-transfer period, 90 fish from each individual RAS unit was distributed into three FTS units, resulting in 30 fish per FTS unit originating from the same RAS tank, or 270 fish in total per temperature treatment (n = 9). This was carried out to ensure a reasonable biomass density and enough fish to sample for the post-transfer period as the fish grew. The post-transfer fish were all reared under identical environmental parameters with temperatures of 9.4 ± 0.6 °C, salinity of 28.8 ± 0.8‰, oxygen saturation in inlet water of 140 ± 13.1%sat, and a light regime with 9 h light and 15 h darkness. The initial biomass density was 16 ± 0.3 kg m⁻³ per tank, and the fish were reared for 10 weeks under the experimental conditions before the trial was completed. The experimental design is illustrated in Figure 1.

2.2. Production System

This study was conducted in six replicate RASs (Figure 2), each consisting of a total volume of 1500 L, a 1000 L cylindrical rearing tank, a mechanical drum filter (40 μm mesh) with a backwash loop, a three-step biofilter filled with approximately 0.23 m³ of saddle-chip biomedia (Dania plast AS, Mariager, Denmark) in each step, and a pump sump combined with a gas balancing filter (GB filter). Pellet and feces were settled and separated through a swirl separator and collected two to three times a day. An external blower supplied each biofilter chamber with air to maintain stable conditions. CO₂ was removed from the GB filter and ventilated out of the systems. Water temperature was regulated with a ceramic heating element and a cooling coil placed in the pump sump, and setpoints were adjusted to achieve target temperatures of 12 °C and 14 °C for the two groups of fish.

The degree of recirculation was roughly 90%, equal to a hydraulic retention time of approximately 10.13 days. All tanks were equipped with flow meters on the seawater and freshwater inlets for constant measurement of water consumption. Make-up water was automatically added to compensate for the wastewater discharged from the drum filter backwash and swirl separator. Alkalinity was maintained at >150 mg L⁻¹ by manually adding sodium bicarbonate (NaHCO₃) after wastewater discharge from the drum filter. Dissolved oxygen concentration in the culture tank was operated at approximately 100% saturation throughout the trial. A more detailed description of the RAS can be seen in [26].

Percent daily recirculation and feed per liter of water were calculated as:

$$\text{Daily recirculation}\left(\mathrm{\%}\right)=1-\left(\frac{\text{Seawater used}+\text{Freshwater used}}{\text{Total water volume in the system}}\right)\ast 100$$

$$\text{Feed pr.Litre of new water}\left({\mathrm{g}\mathrm{L}}^{-1}\right)=\left(\frac{\text{cumulative feed delivered to tank}}{\text{total new water consumption of tank}}\right)$$

The FTS tanks used for the post-transfer period were 1.5 m in diameter and 0.95 m³ in volume.

2.3. Diet and Feeding

A commercial extruded smolt diet (EWOS, Cargill, Bergneset, Norway) suited for Atlantic salmon within the size range and life stage of the current study [43] was fed during the RAS period (

Table 1. Formulation of ingredients and chemical analysis of the content for the commercial diet.

Parameter	Diet
Ingredient (%)
Fish meal (LT94)	27.9
Plant ingredients	43.5
Marine oils	10.8
Plant oils	16.9
Micro ingredients *	5.0
Water balance	−4.1
Composition (%, “as-is basis”)		Method of analysis
Protein	40.6	Elementar Rapid Max N system (Dumas principle) (Elementar Analysesysteme GmbH, Langenselbold, Germany)
Fat	34.7	NMR Analyzer Bruker minispec mq10 system (Bruker, Billerica, MA, USA) LfNMR scan
Moisture	3.9	Leco TGA 701 analyzer (Leco, Geleen, The Netherlands)
Ash	6.7
Gross energy (MJ kg−1)	25.4	Leco AC 600 gross energy bomb calorimetry system (Leco, Geleen, The Netherlands)
Carbohydrate a	18.0	na
Minerals (g kg−1, “as-is basis”)
Phosphorus	9.3	PUMA s2 EDXRF (Bruker, MA, USA)
Calcium	10.1
Zinc	0.18
Iron	0.22

Notes: * Vitamins, minerals, and amino acids. ^a Carbohydrate = 100 − Ash − Protein − Fat. NMR: nuclear magnetic resonance. LfNMR: low-field nuclear magnetic resonance. EDXRF: energy-dispersive X-ray fluorescent spectrometer.

). The water content of the raw material was balanced against the target for the finished diet to ensure that the nutrient contents of the recipe remained accurate and unaffected by variations in the water content; this is considered water balance (Table 1). The feed requirements were determined by internal Cargill feeding tables, and the feed was weighed daily and delivered without interruption by a pre-programmed belt feeder (Alpha Aqua, Esbjerg, Denmark). All salmon were fed to apparent satiation, which was determined to have been reached when excess feed was visually detected at the tank’s bottom. The excess feed, along with feces, was removed gently by a feed collector, designed based on the Guelph system principle [44]. Excess feed and feces were collected twice daily (in the morning at 07:30 and again at 14:30) and manually separated before the uneaten feed was weighed (wet and dry weights).

For the post-transfer period, both groups of salmon were fed a commercial salmon diet containing 42% protein, 32% fat, and 25 MJ kg⁻¹ gross energy (EWOS RAPID HP 500, Cargill, Florø, Norway). The feed was weighed according to the internal Cargill feeding tables and distributed to the tank via a belt feeder (Hølland Teknologi AS, Sandnes, Norway) at an interval of 1.5 h, with 1 h breaks, for a total of four meals during the 9 h light period. Excess feed was removed and weighed from each FTS and recorded every week.

2.4. Water Quality Sampling and Analysis

The measurement and sampling of water quality parameters were performed as described in [26] due to the similarity in the trial designs. The sensor well (Figure 2) provided access to parameters that required monitoring five times a week (Monday to Friday). Water samples were manually collected from each tank and tested for a variety of parameters at different intervals three times a week (Monday, Wednesday, and Friday) and once a week (Monday) (

Table 2. Description and frequency of testing for each water quality parameter monitored and sampled in RASs stoked with Atlantic salmon (Salmo salar).

Parameter	Method of Analysis	Frequency
Temperature	Multi-parameter portable meter WTW multi 3620 IDS (Xylem Analytics, Mainz, Germany)	5 times a week
pH		5 times a week
Salinity		5 times a week
Oxygen		5 times a week
CO2	Franatech dissolved CO2 sensor HR (Franatech AS, Oslo, Norway)	5 times a week
TAN	Spectroquant cell test kits (Merck, Darmstadt, Germany) Spectroquant Prove 300 (Merck, Darmstadt, Germany)	3 times a week
NO2−		3 times a week
NO3−		3 times a week
Alkalinity		3 times a week
Turbidity	WTW Turb750 IR (Xylem Analytics, Mainz, Germany)	3 times a week
TSS	Standard methods NS-EN872:2005 [45]	1 time a week
TGP	Oxyguard Handy Polaris TGP (Oxyguard, Farum, Denmark)	1 time a week

Notes: TAN: total ammonia nitrogen; NO₂⁻: nitrite; NO₃⁻: nitrate; TSS: total suspended solids; and TGP: total gas pressure.

).

Unionized ammonia (NH₃-N) was calculated according to the Henderson–Hasselbach relationship when the water temperature and pH were measured [46]:

$${\mathrm{N}\mathrm{H}}_3-\mathrm{N}=\frac{\mathrm{T}\mathrm{A}\mathrm{N}\ast 1}{(1+{10}^{\mathrm{p}\mathrm{K}\mathrm{a}-\mathrm{p}\mathrm{H}})}$$

where the acidic dissociation constant (pKa) was estimated as:

$$\mathrm{p}\mathrm{K}\mathrm{a}=0.09018\frac{2729.92}{\mathrm{T}\mathrm{e}\mathrm{m}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e},\mathrm{K}}$$

2.5. Mineral Sampling and Analysis

To measure mineral accumulation in the RAS, 500 mL of water sample was collected at the start (19 October), middle (16 November), and end (13 December) of the experiment. To understand the distributional pattern of these minerals at different stages of water treatment in the RAS, the water samples were collected directly from the water stream: A. before the fish tank, B. in the swirl separator, C. after the drum filter, and D. after the microparticle filter (Figure 2). The inlet water was collected by removing the flow sensor on the inlet pipe. When the flow sensor was removed, the production water was allowed to flow freely for five seconds before the sample was retrieved. This was performed so that the sample was not contaminated by the sedimented dirt and grime from the opening. The production water was collected in a bucket and returned to the RAS via the swirl separator. Water after the fish tank, after the drum filter, and after the biofilter was collected directly from the water stream by submerging the 500 mL container. After 13 December, turbidity and TSS were also mapped for each water treatment step.

Continuous flow analysis (CFA) was used to determine total phosphorus (TP), dissolved phosphorus (DP), represented as P particles smaller than 0.45 µm, and orthophosphate (PO₄-P) in samples by following the NS-EN ISO 15681-2 standard [47]. Bone-forming minerals (Ca and Mg) and sludge classification metals (Fe and Zn) were analyzed by inductively coupled plasma mass spectrometry (ICP-MS) after nitric acid digestion, ensuring that all forms were converted to a measurable solution (SS-EN SS-EN ISO 17294-2:2016 [48]/ISO 15587-2:2002 [49]).

2.6. Fish Sampling, Chemical Composition, and Fecal Stability

All sampled fish, for both the RAS and post-transfer period, were euthanized with a lethal overdose (over 250 mg L⁻¹) of tricaine mesylate (Pharmaq AS, Oslo, Norway) before obtaining the weight and length for each individual, allowing for calculations of the condition factor (K) [50,51], specific growth rate (SGR) [29,52], and thermal growth coefficient (TGC) [53] as per the equations below:

$$\mathrm{K}=100\ast \frac{\mathrm{weight},\mathrm{g}}{\mathrm{total} {\mathrm{length}}^3,\mathrm{cm}}$$

$$\mathrm{SGR}(\% {\mathrm{day}}^{-1})=100\ast \frac{\ln \left(\mathrm{final} \mathrm{biomass},\mathrm{g}\right)-\ln (\mathrm{initial} \mathrm{biomass},\mathrm{g})}{\mathrm{time},\mathrm{days}}$$

$$\mathrm{TGC}=\frac{\left(\mathrm{end} \mathrm{weight},{\mathrm{g}}^{1/3}\right)-(\mathrm{start} \mathrm{weight},{\mathrm{g}}^{1/3})}{\mathrm{days} \mathrm{in} \mathrm{sampling} \mathrm{period}\ast \mathrm{mean} \mathrm{temperature},°\mathrm{C}}\ast 1000$$

At start of the RAS period, 20 fish in total were sampled for the heart, liver, and viscera. At the end of the RAS period, the heart, liver, and viscera were collected from 10 fish per RAS unit, for a total of 60 fish. For the post-transfer period, the heart, liver, and viscera were collected from 10 fish per FTS tank, for a total of 180. All of the organs were weighed using a precision weight (Ohaus Pioneer PA2102, T.A. Elektronikk AS, Sandnes, Norway) so that the hepatosomatic index (HSI) [54], cardiosomatic index (CSI), and viscerosomatic index (VSI) could be calculated as per the equations below:

$$\mathrm{HSI}=\frac{\text{liver weight (g)}}{\text{weight of whole fish (g)}}\ast 100$$

$$\mathrm{CSI}=\frac{\text{heart weight (g)}}{\text{weight of whole fish (g)}}\ast 100$$

$$\mathrm{VSI}=\frac{\text{visceral weight (g)}}{\text{weight of whole fish (g)}}\ast 100$$

For chemical analysis of the whole body of the fish, 5 fish per RAS unit were collected in the beginning and again at the end of the RAS period. To ensure that the sampled fish were analyzed purely on the content of the body, the fish were dissected to remove stomach and intestinal content (undigested feed). Both the stomach and intestine were scraped clean with a scalpel and then thoroughly cleaned with fresh water before they were placed back into each individual fish. Subsequently, the whole-body samples were frozen at −20 °C until further processing. The frozen fish were then ground down to a fine paste with a Robot-Coupe R7VV industrial grinder (Robot-Coupe, Vincennes, France) before being weighed and dried. The dried paste was then ground again with a coffee grinder to secure a homogenous sample for the chemical analysis. The average retention of the selected minerals in the whole body was estimated as per equation below:

$$\mathrm{Retention}(\mathrm{\%})=\frac{\frac{\mathrm{w}2\left(\mathrm{g}\right) \ast \text{mineral in fish}(\mathrm{g} {\mathrm{kg}}^{-1})}{100}-\frac{\mathrm{w}1\left(\mathrm{g}\right) \ast \text{mineral in fish}(\mathrm{g} {\mathrm{kg}}^{-1})}{100}}{\frac{\left(\mathrm{w}2\left(\mathrm{g}\right)-\mathrm{w}1\left(\mathrm{g}\right)\right) \ast \mathrm{FCR} \ast \text{mineral in feed}(\mathrm{g} {\mathrm{kg}}^{-1})}{100}}\ast 100$$

where w1 represents the initial fish weight and w2 is the final fish weight.

Feces from each RAS were retrieved directly from the swirl separator using a 12.5 cm ø mesh stainless steel strainer (Grunwerg, Chesterfield, UK). The retrieved fecal particles were subjected to 5 min of stirring in a Mastersizer 3000 instrument (Malvern Panalytical Ltd., Malvern, UK), where each passage in the instrument reduced the particle size of the fecal sample. A higher fraction of particles larger than 50 µm indicated a higher fecal stability.

The homogenized samples of the whole body and feces were analyzed for minerals with inductively coupled plasma optical emission spectroscopy (ICP-OES) (PerkinElmer, Shelton, CT, USA). The mineral analyses were validated using certified reference materials (Bovine Muscle Certified Reference Material [BOVM-1] for trace metals and European Reference Material [ERM-BB422] for other constituents).

2.7. Statistical Analysis

For the mineral accumulation in RASs, to prevent pseudo-replication issues, generalized mixed linear models (with “log” link) were used to account for the presence of non-independent samples, focusing on the effect of water temperature in RASs on the different minerals tested at different points in time. Every model considered fixed effects for time and temperature as categorical variables (as well as their interaction), along with random effects for the “tank” variable and for the “measurement location”, which was considered to be nested within “tank”. Generalized mixed linear models were also used for the samples collected on a fish basis (TGC, HSI, VSI, and CSI) in RASs and post-transfer. The analysis was performed with a “log” link (the effects were assumed to be multiplicative) for all parameters except the TGC. For the TGC, we utilized a linear approach to better capture its specific relationship with temperature. This was performed by looking at the linear relationship between the cumulative temperature and the difference between the cube roots of the final body weight and the initial body weight.

Water quality parameters, average whole-body retention, and fish performance on a tank basis were analyzed by a one-way ANOVA. The body composition of salmon was tested with a two-way ANOVA, with time and temperature as variables. Due to the lack of fecal samples from the start of the experiment, the chemical analysis of selected minerals in the fecal sample was tested with a one-way ANOVA only for the samples at the end of the experiment to determine differences in the composition. For the analysis of fecal stability, the confidence interval (CI) was determined through simulation and is represented as the 95% CI for the mean of each sample.

Equal variances for the data were investigated with a Levene test (significance level α = 0.05), and post hoc tests were performed either with a Tukey test for pairwise comparisons or with a Dunnett’s test for comparisons against a single reference group (significance level α = 0.05 and highly significant at α = 0.001). The statistical software R (version 4.3.2) and Minitab (20.4, 2021 Minitab, LLC, State College, PA, USA) were used to manage the data.

2.8. Ethical Statement

The experiment adhered to the guidelines and protocols sanctioned by the European Union (EU Council 86/609; D.L. 27 January 1992, no. 116) and the National Guidelines for Animal Care and Welfare issued by the Norwegian Ministry of Education and Research.

3. Results

3.1. Fish Performance

There was no mortality during the entire experimental period. There was significantly higher body weight gain (p < 0.05), length (p < 0.05), and SGR (p < 0.05) for the salmon reared at 14 compared to 12 °C (

Table 3. Fish performance parameters in Atlantic salmon (Salmo salar) post-smolts reared at 12 °C and 14 °C in RASs and after the post-transfer stage. Data are expressed as averages on a tank basis ± SD for each temperature regime in RASs and for post-transfer. Data were analyzed by a one-way ANOVA (p < 0.05), with weight, length, and K factor per tank and with thermal growth coefficient, specific growth rate, feed consumption, feed conversion ratio, and mortality per temperature regime.

	RAS	Post-Transfer
Initial weight (g)
12 °C	239 ± 3	563 ± 12
14 °C	241 ± 2	570 ± 11
Final weight (g)
12 °C	554 ± 9 *	1197 ± 30
14 °C	575 ± 6 *	1156 ± 53
Body weight gain (g)
12 °C	315 ± 7 *	634 ± 29 *
14 °C	334 ± 6 *	586 ± 50 *
Length (cm)
12 °C	34 ± 0.2 *	43 ± 0.44
14 °C	35 ± 0.03 *	42 ± 0.77
K factor
12 °C	1.39 ± 0.01	1.51 ± 0.02
14 °C	1.39 ± 0.02	1.51 ± 0.03
Feed consumption (kg)
12 °C	35.4 ± 0.3 *	16.4 ± 0.7 *
14 °C	36.1 ± 0.1 *	15.0 ± 1.1 *
Feed conversion ratio
12 °C	0.80 ± 0.01	0.84 ± 0.02
14 °C	0.77 ± 0.01	0.86 ± 0.03
Thermal growth coefficient
12 °C	2.99 ± 0.05 **	3.85 ± 0.2 *
14 °C	2.67 ± 0.04 **	3.53 ± 0.3 *
Specific growth rate
12 °C	1.50 ± 0.03 *	1.09 ± 0.04 *
14 °C	1.55 ± 0.02 *	1.02 ± 0.06 *

Notes: Significance levels: * p < 0.05 and ** p < 0.001.

) in the RAS period. While the TGC was significantly lower (p < 0.001) in the salmon reared at 14 °C, compared to 12 °C. No differences were observed in the K factor (p = 0.86). At the end of the post-transfer period, salmon with a rearing history of 12 °C had a significantly higher body weight gain (p < 0.05), TGC (p < 0.05) and SGR (p < 0.05) compared to 14 °C. Also, in this period, no significant differences were observed for the K factor (p = 0.94).

Feed consumption was significantly higher for the salmon reared at 14 °C in the RAS period (p < 0.05). Fish reared at 12 °C in the RAS period generally had higher FCR (p = 0.08) compared to fish reared at 14 °C. For the post-transfer period, the tendency shifted to a lower FCR (p = 0.07) for the salmon with a rearing history of 12 °C. However, there was no evidence of difference in FCR either in the RAS or post-transfer period.

In the RAS period, no significant differences were detected for HSI, CSI, or VSI. Post-transfer, there were significant differences in HSI (p < 0.001) and VSI (p < 0.001) between the fish with rearing history of 12 and 14 °C (Figure 3). From the end of the RAS period to the end of the post-transfer period, an increase in HSI was observed for the fish with a rearing history of 14 °C, while VSI increased for both groups.

3.2. Water Quality in RASs

From all the water quality parameters analyzed, significant differences were only observed for the concentration of nitrate (p < 0.05) between the two temperature regimes for the 9-week experimental period in the RAS (

Table 4. Water quality measures in brackish recirculating aquaculture systems operated at water temperatures of 12 °C and 14 °C for 9 weeks. Data are expressed as averages for the trial period ± SD (n = 3). Data were analyzed by a one-way ANOVA (p < 0.05).

	12 °C	14 °C
pH	7.6 ± 0.05	7.7 ± 0.02
Salinity (ppt)	14.6 ± 0.13	14.5 ± 0.19
Dissolved oxygen (ppm)	8.70 ± 0.04	8.88 ± 0.16
CO2 (mg L−1)	8.7 ± 0.19	8.6 ± 0.20
Alkalinity (mg L−1)	159 ± 4.24	151 ± 3.25
TAN (mg L−1)	0.6 ± 0.11	0.5 ± 0.04
Nitrite—Nitrogen (mg L−1)	2.4 ± 0.71	2.6 ± 0.13
Nitrate—Nitrogen (mg L−1)	55 ± 0.61 *	62 ± 2.98 *
Unionized ammonia—Nitrogen (μg L−1)	5.5 ± 1.6	6.0 ± 0.1
TSS (mg L−1)	8.31 ± 1.06	8.70 ± 0.60
Turbidity (NTU)	1.1 ± 0.40	0.9 ± 0.18
TGP (<102%)	96.0 ± 0.67	95.5 ± 1.78
Freshwater consumption (L day−1)	94.1 ± 8.87	92.8 ± 8.06
Seawater consumption (L day−1)	53.9 ± 5.60	52.2 ± 4.13
New water consumption (L day−1)	148 ± 13.3	145 ± 11.9
Total feed per liter of new water (g L−1)	5.3 ± 0.16	5.3 ± 0.50
Recirculation degree (% day−1)	90 ± 0.91	90 ± 0.81

Notes: TAN: total ammonia nitrogen; TSS: total suspended solids; and TGP: total gas pressure. * Significant at p < 0.05.

).

3.3. Accumulation of Minerals in RASs

No significant differences were observed in the concentration of any mineral or metal following the flow of the production water through the treatment loop. At the end of the RAS period, significant differences between the temperature regimes were only observed for TP (p < 0.001), DP (p < 0.001), and total Fe (p < 0.001), where the RAS operating with water temperatures of 12 °C accumulated higher concentrations compared to systems with a water temperature of 14 °C (Figure 4). No interactions were found between the time of sampling and location of sampling in the RAS.

3.4. Body Composition of Fish in RASs

The concentrations of selected minerals in the whole body of salmon post-smolt did not differ between temperature regimes (

Table 5. Chemical analysis of dry matter and selected minerals in the whole body of Atlantic salmon reared in RASs at 12 °C and 14 °C measured at the start and end of the RAS period (19 October to 13 December). The average initial weight was 240 ± 35 g, and the final weight was 554 ± 9 g at 12 °C and 575 ± 5 g at 14 °C. Data were analyzed by a two-way ANOVA (time, temperature, and time × temperature [interaction]) and are given as mean ± SD (n = 3) in wet weight (as-is).

				p-Value
	Initial	Final	Time	Temperature	Interaction
Dry matter (%)
12 °C	31.2 ± 0.6	36.3 ± 0.2
14 °C	31.6 ± 0.2	36.3 ± 0.7	<0.001	0.515	0.510
Calcium (g kg−1)
12 °C	4.4 ± 0.1	4.0 ± 0.3
14 °C	4.6 ± 0.3	4.4 ± 0.3	0.123	0.108	0.659
Iron (mg kg−1)
12 °C	8.2 ± 0.2	11.2 ± 0.2
14 °C	8.9 ± 0.5	12.3 ± 3.8	0.020	0.454	0.858
Magnesium (mg kg−1)
12 °C	294 ± 13	320 ± 22
14 °C	302 ± 5	321 ± 7	0.019	0.593	0.714
Phosphorus (g kg−1)
12 °C	4.4 ± 0.2	4.4 ± 0.3
14 °C	4.5 ± 0.1	4.4 ± 0.1	0.395	0.484	0.768
Zinc (mg kg−1)
12 °C	44 ± 1	41 ± 3
14 °C	47 ± 1	41 ± 1	0.004	0.284	0.080

) from the start to the end of the RAS period. Dry matter, Fe, and Mg concentrations increased between initial and final sampling, while Zn concentrations decreased.

The estimated average retention of selected minerals in the whole body of the salmon did not show evidence of differences between the two fish groups reared at 12 °C and 14 °C (

Table 6. Estimated average retention of selected minerals in the whole body of Atlantic salmon reared in RASs at 12 °C and 14 °C + SD (n = 3). The average initial weight was 240 ± 35 g, and the final weight was 554 ± 9 g at 12 °C and 575 ± 5 g at 14 °C. Data were analyzed by a one-way ANOVA.

Average Whole-Body Retention (%)		p-Value
Calcium
12 °C	47 ± 6
14 °C	55 ± 5	0.179
Iron
12 °C	8 ± 0.3
14 °C	9 ± 4	0.730
Phosphorus
12 °C	58 ± 9
14 °C	60 ± 2	0.695
Zinc
12 °C	28 ± 4
14 °C	26 ± 1	0.438

).

3.5. Chemical Composition of Feces and Fecal Stability in RASs

The chemical composition of feces gathered from the fish reared in RASs showed that only the Mg concentrations were significantly lower for the fish reared at 12 °C; all other minerals and metals were similar between the two temperature regimes (

Table 7. Chemical analysis of selected minerals (mg kg⁻¹) in feces of Atlantic salmon reared in RASs at 12 °C and 14 °C from 19 October to 13 December. The average initial weight was 240 ± 35 g, and the final weight was 554 ± 9 g at 12 °C and 575 ± 5 g at 14 °C. Data were analyzed by a one-way ANOVA and are given as mean ± SD (n = 3).

	G kg−1	p-Value
Calcium
12 °C	44.7 ± 2.08
14 °C	46.7 ± 2.08	0.305
Iron
12 °C	0.83 ± 0.02
14 °C	0.84 ± 0.02	0.588
Magnesium
12 °C	5.27 ± 0.32
14 °C	6.83 ± 0.06	0.001 *
Phosphorus
12 °C	24.3 ± 1.53
14 °C	26.7 ± 1.15	0.102
Zinc
12 °C	0.55 ± 0.01
14 °C	0.55 ± 0.02	0.768

Notes: * Significant difference between temperatures (p < 0.05).

).

In terms of fecal stability, there was no evidence of differences between the fish groups. The fish reared at 12 °C had a probability of 85.1 ± 2.3% to generate fecal particles smaller than 50 μm vs. 85.5 ± 4.1% for the fish reared at 14 °C.

4. Discussion

Fish reared at 14 °C showed higher growth during the RAS period compared to those at 12 °C, aligning with previous studies on Atlantic salmon [31,36,55,56,57] and findings from the Freshwater Institute in West Virginia [58]. Although the TGC was higher for fish reared at 12 °C, it remained within the target proposed for Atlantic salmon in RASs [59]. The K factor was similar between the two temperature regimes, indicating a comparable slaughter yield [50,51]. Although the FCR was not significantly different between 12 and 14 °C, there was a tendency for higher FCR at lower temperatures (p = 0.08), which was opposite to what was expected [31,36]. Interestingly, the tendency shifted post-transfer to seawater, and the fish with a rearing history of 12 °C ended up with a tendency of lower FCR (p = 0.07) compared to those reared at 14 °C. For the post-transfer period, fish with a rearing history of 12 °C maintained a higher TGC and SGR. This is in line with results showing that lower freshwater temperatures (4–6 °C) before transfer can enhance post-transfer growth in different Atlantic salmon strains, and the pattern in the current study seems to be consistent with studies on compensatory growth [60,61,62]. Additionally, reducing temperatures from 14 to 12 °C during the freshwater period can reduce sexual maturation [58] and improve physiological adaptation to seawater [28,31,61]. Acclimation of fish to post-transfer temperatures is also highlighted as beneficial for growth parameters [31,63,64]. In the current trial, the post-transfer water temperature of 9.4 ± 0.6 °C, being closer to 12 °C, likely gave an advantage to the fish with a freshwater rearing history of 12 over those reared at 14 °C.

Post-transfer, both the HSI and VSI were significantly lower in fish reared at 12 compared to 14 °C. These observations in the RAS, along with the significantly lower HSI post-transfer for the fish reared at a lower temperature, are consistent with a previous study examining fish raised at lower temperatures in a FTS compared to a RAS [28]. Although the current study did not show significant differences in the CSI, it is believed that the small difference in water temperature of 2 °C was not sufficient to inflict a different CSI compared to the previous study, which had a larger difference in water temperature (approximately 5 °C) [28]. The similar post-transfer K factor, combined with a lower VSI for fish with a rearing history of 12 °C, suggests that their weight gain was primarily due to muscle growth. Fish grow faster early in life, with new muscle fibers forming until they reach 40 to 50% of their maximum body length [65]. The growth period where they reach up to approximately 1 kg is thus of high importance due to the recruitment of new muscle fibers [66,67], regardless of smolt type [68]. However, further research is needed to determine if temperature reduction in RASs affects muscle fiber development and fish robustness.

Although the energy requirements for cooling water differ between different RAS setups [69], the RAS units in the current trial generated water temperatures of 20 °C, which needed to be cooled to 12 and 14 °C. Based on the heat capacity of the water, the energy required to cool water an additional 2 °C, from 14 to 12 °C, can be calculated. In the current trial, the total consumption of new water did not differ between the groups (148–145 L day⁻¹). The following equation can be used to determine the energy required to cool 146.5 L day⁻¹, representing the average new water consumption of the tanks in the current study [70]:

$$\mathrm{Q}=\mathrm{m}\mathrm{c}∆\mathrm{t}$$

where Q is the heat energy (in joules), m is the mass of the water (in kg), c is the specific heat capacity of water (approximately 4.18 J g°C⁻¹ or 4180 J kg°C⁻¹ [70]), and Δt is the change in temperature (in °C).

$$\mathrm{Q}=146.5 \mathrm{k}\mathrm{g}\times 4180 \mathrm{J}/\mathrm{k}\mathrm{g}°\mathrm{C}\times 2 °\mathrm{C}$$

To cool the water by an additional 2 °C, 1 224 740 J is required. Using the average price of electricity in Norway for Q3 24 at NOK 1.123 kWh⁻¹, the cost can be calculated as follows [71]:

$$\mathrm{E}\mathrm{n}\mathrm{e}\mathrm{r}\mathrm{g}\mathrm{y}\left(\mathrm{k}\mathrm{W}\mathrm{h}\right)=\left(\frac{1224740 \mathrm{J}}{3600000}\right)\times \mathrm{N}\mathrm{O}\mathrm{K} 1.123 {\mathrm{k}\mathrm{W}\mathrm{h}}^{-1}$$

An extra NOK 0.382 day⁻¹ per RAS unit is needed to cool the water by an additional 2 °C. In relation to the accumulated biomass production of 44.3 kg, this translates to an additional cost of NOK 0.00058 per kg fish produced per day (NOK 0.058 ton fish produced day⁻¹). These calculations are based purely on the effect of the cooling of water, and to relate this to specific RAS productions, energy efficiency, cooling equipment, and additional production time must be considered to adapt the estimate. For the current trial, which only included differences in the water temperature, the extra cost of NOK 0.058 ton fish produced day⁻¹ achieved an approximately 8% increase in body weight gain after 66 days in seawater.

Water samples taken in the RAS showed no differences in compound concentrations between different water treatment steps. Crouse et al. (2022) experienced slightly elevated TSS and nitrate for the salmon reared at 14 compared to 12 °C due to higher feed consumption driven by the elevated temperature, as also shown by previous studies [58,72,73,74]. Similar results were found in the current study, with significantly elevated nitrate at 14 °C. However, the upper recommended limits for nitrate-N (443 mg L⁻¹) and TSS (15 mg L⁻¹) were not exceeded [75,76,77]. Both at 12 and 14 °C, the concentrations of nitrite-N (2.4 and 2.6 mg L⁻¹, respectively) were within the recommended levels reported for salmonids reared in freshwater and seawater (0.1 and 0.5 mg L⁻¹, respectively) [22,78,79]. The concentration of unionized ammonia nitrogen, which has sublethal effect on Atlantic salmon at water concentrations of 10 μg L⁻¹ [80], was also below the lowest recommended level of 12.5 μg L⁻¹ [79]. Since intermediate salinities reduce ammonia toxicity, it is concluded that the concentrations in the current study were well within acceptable levels for salmon [81]. Phosphates play a crucial role in bacterial development, and PO₄ can act as a supporting or limiting factor for bacterial growth and activity [3,82,83,84]. The biofilter bacteria, including Nitrosomonas and Nitrobacter, consume PO₄ to grow [85]. In the current experiment, a spike in PO₄ was observed mid-trial for both temperatures before the levels were reduced again, unlike TP, DP, and dissolved Fe. This could indicate periodic drops in bacterial culture due to the moving bed itself, where biochips regularly collide and detach bacterial pieces that ultimately exit through the recirculation loop [86], potentially also causing variations in TP and DP concentrations. Since the spikes were observed for both temperature regimes and only for PO₄, they are more likely to represent the natural development of bacterial cultures [87]. When the biofilter bacteria die, it will take some time to build up the new culture, and in this period, PO₄ could increase, as observed in the current trial. Thus, PO₄ had the highest fluctuations between the sampling dates of the P components measured in this experiment.

A solution with settling columns was used for fecal removal in the RAS, which is not normal in commercial systems, but the wastewater discharge rates were similar to those of other studies [17,18]. Both the Guelph system and total urine and feces columns are confirmed as suitable methods for fecal collection, ensuring minimal nutrient leakage and damage to the fecal particle [88]. The current experimental system successfully removed larger particles, leaving mainly microparticles to accumulate within the system. This was indicated by analyses that showed no significant difference between TP and DP (NS-EN ISO 15681-2) [47]. Higher TP and DP accumulation at 12 °C were also observed, suggesting that the dissolved fractions will control the accumulation rate in the RAS. This study therefore highlights the challenge of capturing DP in a mechanical filter and supports the suggestions that precipitation or enhanced biological phosphorus removal must be applied to remove DP from the production water [89,90]. Although not proven in the current study, the removal of settleable particles (> 100 μm) [91] may have generated a lack of cake filtration in the drum filter. Accumulated material on the filter (cake) has the potential to provide greater filtration efficiencies than the filter screen alone [92]; subsequently, this lack of cake could have caused a lower removal efficiency of P and particles in general. Drum filters can easily remove particles larger than 30 μm, but smaller microparticles tend to accumulate [77,93,94,95,96]. The material of the drum filter may also affect the removal rates or general accumulation [97]; for example, in the current experiment, polyester was used, but filter media consisting of glass or metal components could generate a different effect within the system. However, this remains unclear and needs to be examined in further studies.

By using an equation to determine concentrations in the outlet of a reuse system [89], it is possible to predict the accumulated concentration measured in RASs compared to what would be expected in a FTS:

$$\mathrm{C}=\frac{1}{1-\mathrm{R}+\left(\mathrm{R}\ast \mathrm{R}\mathrm{E}\right)}$$

where C is the concentration measured in the outlet or, in our case, the tank, R equals the degree of reuse, and RE equals the removal efficiency.

$$\mathrm{C}=\frac{1}{1-0.90+\left(0.90\ast 0\right)}$$

$$\mathrm{C}=10$$

The removal efficiency is set to zero since the current study suggests that DP will accumulate. With 90% reuse, the concentration of P in the RAS is estimated to be ten times higher than in a FTS. If the current experiment had been conducted in a FTS, the concentrations in the fish tank would have been approximately (4.7 and 3.8 mg L⁻¹)/10 = 0.47 and 0.38 mg L⁻¹ for the 12 and 14 °C regimes, respectively. Cooling the water by an additional 2 °C resulted in a 23.7% increase in DP, which, from an environmental perspective, means that temperature affects how concentrated or dilute the emitted P effluent will be. To illustrate this effect in a practical example, where it is assumed that an average of 18% dietary P is emitted as DP and 52% is emitted as particles in feces [35,98,99,100], reducing the water temperature from 14 to 12 °C could therefore change the emission characteristic to 22.3% DP and 47.7% particle-emitted P, making it harder to collect in sludge treatment. These levels are used as an example, and the availability of P in different raw materials is still the most important factor affecting the composition of the excreted P waste [26]. The elevated DP at 12 °C, compared to 14 °C, may most likely be attributable to a combination of reduced feed consumption and altered P excretion rates, both driven by temperature changes.

In rainbow trout (Oncorhynchus mykiss), the Mg requirement of 330–600 mg kg⁻¹ [43] can be met through the diet or water if water-borne Mg concentrations are at least 46 mg L⁻¹ [101], as seen in low water exchange in brackish RASs [26], similar to that in the current trial. While the Mg requirements can be sufficiently met, the observed 20.5% increase in DP in the current trial is unlikely to cover the total requirement of P for the fish. Estimating water concentrations of P against drinking rates for seawater-adapted salmon (4 mL kg h⁻¹–6.4 mL kg h⁻¹ [102,103]) and assuming, using a simplified example, that a salmon needs approximately 70 days to reach 1 kg, it drinks approximately 6.72–10.75 L of the production water. If the average production water in a RAS contains 5 mg P L⁻¹ (reflecting the 12 °C regime in the current trial), the salmon can only have 33.6–53.7 mg kg⁻¹ growth, which is insignificant compared to the dietary requirement of 6000–10,000 mg kg⁻¹ [43]. To cover the P requirement, a water concentration of 893–930 mg P L⁻¹ would be needed, nearly 200 times higher than that measured in the current trial.

The Ca/P ratio in the salmon was around 1, which was expected and has been confirmed in several studies [104,105,106]. The whole-body composition thus provided an estimated average P retention of 59%, implying a P excretion of 41%. Fecal stability was similar for both temperature treatments, meaning that the amount of feces removed from the system was equal. The feces composition only showed significantly lower Mg for fish reared at 12 °C. The unaffected Zn accumulation in water and in the feces was a positive observation as it indicated that reducing the temperature from 14 to 12 °C did not alter sludge composition. This would be beneficial since the sludge can be graded similarly when used as a fertilizer [107]. However, the 23.7% higher DP concentration in the production water at 12 °C suggests a difference in the nutritional retention of P. Although not significant, P in the feces showed a tendency to be lower (p = 0.1) for fish reared at 12 °C, along with Ca, another bone-forming mineral. It has been proven that a faster growth rate initiated by temperature will promote a more rapid bone development [43,74], meaning the fish will have a higher dietary P requirement at higher temperatures. Exceeding this dietary requirement leads to higher urinary P excretion [108], measurable as DP. While the current trial cannot determine this, future studies with a nutritional trial design and longer RAS period could investigate it. The stable conditions of a RAS can prevent drops in appetite initiated by sudden fluctuations in temperature, thus securing a stable growth rate throughout a production period [109,110,111]. Lowering dietary P to counteract the increased excretion of P when reducing the temperature from 14 to 12 °C while still maintaining growth and health would be beneficial for the industry and environment.

5. Conclusions

This study concludes that rearing fish at 12 °C in a RAS offers several benefits post-transfer to seawater, including a higher TGC and lower VSI and HSI. A reduction from 14 to 12 °C will not negatively affect the Ca/P ratio in Atlantic salmon. Although cooling water from 14 to 12 °C resulted in an 8% higher body weight at the end of post-transfer to seawater, it caused the RAS to accumulate higher levels of DP and total Fe in water at 12 °C, which can increase the effluent emitted if left untreated.