2.2.1 Physiology

After possible recovery from stress related to transport from farm to our laboratory, the first simulated transport experiment was carried out (Open transport). Just before transfer to the transport tank (see below), six individual salmon were netted individually from the holding tank and rapidly killed by a sharp cranial blow within 5–10 s (Control open). Blood samples were drawn from the caudal vasculature for analysis of cortisol, blood pH, glucose, lactate, sodium, potassium, calcium, chloride, partial pressures of oxygen (P_O2) and carbon dioxide (P_CO2), and haematocrit (Hct). Subsequently, pH in the white muscle was measured. Finally, fish appearance was checked for potential scale loss and haemorrhages (‘red bellies’). The same procedure was carried out before closed transport (Control closed) and after recovery (Recovery-19 h and Recovery 43-h) where 10–12 fish were sampled.

2.2.2 Simulated transport

The duration of each simulated transport was 5 h, corresponding to an intermediate duration of a commercial well-boat transport. For the simulated open transport, a similar adjacent tank (referred to as the transport tank) was used as a common FT system, whereas for the simulated closed transport (same tank), the water supply was stopped and pure oxygen was supplied to the tank through two diffusers (Point Four Systems Inc.). The water level in the tank was adjusted, before fish were added, to a height of 56 cm to mimic a typical commercial fish density. The resulting volume of water was 2.24 m³. Fish density was 54 kg m⁻³ in the open transport and 42 kg m⁻³ in the closed transport. Water quality was measured (DO, CO₂, pH and temperature) and 0.5-L water samples were taken after 0, 0.5, 2.5 and 5.0 h for later analysis of total ammonium-N (including NH₃), total organic carbon (TOC), turbidity and colour. After 5 h transport, ten fish were individually sampled and killed as described above to monitor the stress condition of the fish. The remaining fish in the transport tank were then netted back to the FT holding tank for post-transport recovery.

2.2.3 Experimental timeline

Sixty-five hours after the fish arrived from farm and were transferred to the laboratory holding tank (Day 0), the simulated open transport started. It was anticipated (see Discussion) that this time period was adequate for fish to recover from transport stress. The 5-h open transport was conducted on Day 3. On Day 4, after 14.5 h of recovery, the 5-h closed transport began. In this case, fish were also sampled on Day 5 and Day 6, as well as after 19 and 43 h recovery following the closed transport. An overview of the timeline for the experimental fish groups is shown in Figure 1.

2.3.1 Velocity and turbulence measurements

Measurement techniques and post-processing analysis similar to those reported by Plew et al. (2015) and Klebert et al. (2018) were used. The velocities inside the different tanks were recorded using acoustic Doppler velocimeters (Vector, Nortek AS). These instruments have a cylindrical sample volume that is approximately 15 mm high by 15 mm in diameter, located approximately 0.15 m from the instrument probe. Two velocimeters were used at different locations in the tank. During the measurement, the velocities were continuously recorded with high frequency (32 Hz) over several hours. Velocity data from the Nortek Vectors were filtered using a phase space filter (Goring & Nikora, 2002), which removes the velocity spikes that occurred due to Doppler noise, signal aliasing (fish in the sampling volume) or other instrument noise. The mean velocity (time-averaged) values were calculated from the velocity data with the spikes removed. For each velocity components in three spatial directions (U, V and W), the instantaneous fluctuations from the mean values are denoted u’, v’ and w’ respectively. Turbulent kinetic energy (TKE) is evaluated like the following $k=\frac{1}{2}\left(\overline{{u^{\prime }}^2}+\overline{{v^{\prime }}^2}+\overline{{w^{\prime }}^2}\right)$, and the turbulence intensity (TI) is calculated from the TKE as $I=\sqrt{\frac{2k}{3}}$ and is used in this study to evaluate the effect of fish on the flow.

2.3.2 Water quality

The dissolved oxygen level was measured using a YSI ProODO meter (YSI Inc.), and the concentration of carbon dioxide was measured by using the OxyGuard® Portable CO₂ Analysator (OxyGuard). The instrument measures the level of free dissolved CO₂ in the range of 0–50 mg l⁻¹. Water acidity was measured with a shielded glass electrode (WTW SenTix 41, WTW) connected to a portable pH meter (model WTW 315i). For measuring water temperatures, a Testo 110 thermometer (Testo AG) was used. Total ammonium nitrogen (TAN = NH₄⁺ + NH₃) was determined according to Norwegian Standard NS 4746, (1975) by measuring the absorbance at 630 nm after a reaction with hypochlorite and phenol at pH 10.8–11.4. The detection limit is 5 μg l⁻¹. The toxic ammonia fraction (NH₃) was calculated from data given by Emerson et al. (1975) where the ammonia concentration is computed based on the measured TAN, pH, water temperature and salinity values. The contents of total organic carbon (TOC) in the samples, including precipitate, were analysed by subjecting samples to combustion and IR spectrophotometric analysis of TOC according to NS-EN 1484. The detection limit is 0.5 mg l⁻¹. Turbidity (cloudiness of water) was analysed according to NS-EN ISO 7027. Water samples, transferred to cells, were inserted into a Hach 2100 AN IS Turbidometer (Hach) for measurement of light scattering. The results are presented as NTU (Nephelometric Turbidity Unit). The detection limit is 0.03 NTU. In case of Colour, samples were determined spectrophotometrically according to ISO 7887:2011. The limit of detection is <1 mg Pt l⁻¹. Colour values are reported dimensionless.

2.3.3 Blood chemistry

Cortisol in blood was sampled with heparinized syringes and centrifuged (6000 rpm, 5 min) with a Galaxy Mini Star Silverline C1413-VWR230 centrifuge (Radnor, USA) for extracting blood plasma. The plasma was subsequently stored at −20°C until later analysis. Cortisol was determined by using a radioimmunoassay method as described by Iversen et al. (1998). Whole blood pH was measured immediately after blood withdrawal by a syringe. A similar instrument was used for measurement of water pH. Glucose was measured by dipping the tip of a test strip in whole blood after the blood was sampled with a 5-ml syringe. The strip was then inserted into an Ascensia Contour meter (Bayer HealthCare LLC), and the glucose concentrations were provided in mmol l⁻¹ on the display. Whole blood lactate was assessed with a Lactate Scout+ meter (EKF Diagnostics GmbH) with a measuring range of 0.5–25 mmol l⁻¹. A test strip was inserted into the instrument after it was briefly soaked in blood immediately after the blood was sampled by syringe. While point-of-care devices can be useful as field methods to assess relative differences between fish groups, it should be pointed out that portable lactate and glucose meters generally underestimate the concentrations of glucose and lactate compared with traditional analytical methods (Bakke & Woll, 2014; Stoot et al., 2014; Venn Beecham et al., 2006; Wells & Pankhurst, 1999). Plasma chloride was determined by silver chloride titration using a Radiometer CMT 10 chloride titrator (Radiometer AS). The sodium content of the plasma was analysed by ICP-MS by an Agilent series 7700 instrument according to NS-EN ISO 17294-2 (2007).

2.3.4 Initial pH in white muscle

The initial pH of white epaxial muscle was measured directly in the muscle just after killing. Similar instrument and electrode were used for measuring pH in water.

2.3.5 Respiration and behaviour

Respiration frequency was calculated as a count of the number of opercular beats per minute observed from underwater films (filmed with GoPro 5 cameras). It was determined for five individuals per time point. Recovery after commercial transport was observed 1 and 2 h following commercial transport. Observations during open and closed transport were done at 0, 1, 2, 3, 4 and 5 h after the onset of simulated transports. Potential differences in behaviour among fish in open and closed systems were evaluated by visual observations. Gaping behaviour was defined as a disturbance of water surface by fish with the mouth open. It was measured by counting the number of fish exhibiting this behaviour in a 1-min period and standardized as a proportion by dividing by the number of fish in the tank at that time. The mean of this proportion was calculated for each transport phase, that is during early recovery after commercial transport, as well as during open and closed transports.

3.1.1 Water quality

The water quality in the three closed transport tanks on the vehicle just after arrival, as well as in the FT holding tank during the period when the swimming and ‘bottom’ fish were sampled (see above) is shown in Table 1. The water temperature (9.1°C) was close to that at the farm (9.4°C), similar to the acclimation temperature of the fish. In all tanks, the water was supersaturated with respect to DO (120–128%). In the holding tank, DO varied between 49% and 122% saturation since conditions had not yet stabilized after fish had recently been transferred to the tank. Carbon dioxide levels were only somewhat elevated in all cases (2–5 mg l⁻¹), also verified by the fact that seawater pH was comparatively high in all cases (pH 7.26–7.62). In the closed tanks, TAN had accumulated during transport to 1.12–1.47 mg l⁻¹, whereas in the FT holding tank, the level was lower (0.70 mg l⁻¹). The toxic fraction (NH₃) of TAN was, however, low in all cases (2.0–3.0 μg l⁻¹). The TOC levels were clearly elevated after transport (6.2–8.6 mg l⁻¹). Turbidity in the transport tanks was between 2.7 and 4.8 NTU, and 1.8 NTU in the holding tank, whereas the Colour values varied between 8 and 10 in the transport tanks as opposed to 3 in the holding tank.

3.1.2 Stress

Table 2 shows the stress load of swimming fish and possible moribund fish in the holding tank shortly after arrival from the farm. The cortisol levels (577 and 653 nmol l⁻¹) show that both groups of fish were severely stressed although they were not different (p > 0.05). The mean blood pH of ‘bottom fish’ (pH 7.32) was significantly lower than ‘swimmers’ (pH 7.54) indicating that acidosis was less severe in the latter group. Glucose, sodium and chloride concentrations did not differ between groups (Table 2, p > 0.05). In contrast, the lactate levels of swimming fish (6.0 mmol l⁻¹) were strikingly different from ‘bottom fish’ (17.6 mmol⁻¹). In line with that, the initial muscle pH of the two groups of fish was also significantly different.

3.2.1 Hydraulic analysis

Hydraulic analysis was used to monitor environmental conditions in which the fish were submitted during all measurements and also to possibly gather information about their behaviour. Figure 2 shows that the turbulence intensity recorded in the tanks was increasing in the holding tank (Turbulence 2/holding tank) with the stocking density (number of fish). Similar trend occurred in the flow-through transport case (Turbulence 3). The turbulence intensity values in both cases were quite similar. In contrary, in the closed transport case, the intensity (Turbulence 4) was four times higher than the two other cases. Regarding the mean velocities, they decreased in both holding and flow-through tanks with increasing stocking densities and the magnitudes were quite similar. For the closed transport case though, a very low mean flow velocity was registered, where it was basically due to the movement of the fish. The hydraulic conditions affected the swimming pattern of the fish.

3.2.2 Water quality

Throughout the open system transport, water temperatures varied between 9.3 and 9.7°C, whereas in the closed system, the temperature increased from 9.6 to 10.4°C during the 5-h transports. Changes in water quality during both transports, as well as during closed transport recovery for 19 and 43 h, are shown in Table 3. At DO saturation levels between 92% and 108% in the FT transport and holding tanks (used for recovery), the supply of oxygen was always adequate. In the closed transport tank, DO varied between 90% and 134% saturation. Carbon dioxide was not detected during the open transport (instrument detection limit: 0.0 mg l⁻¹), and during closed transport, CO₂ increased gradually, from 7 mg l⁻¹ after 0.5 h up to 24 mg l⁻¹ after 5 h. Acidity in the open system was relatively stable at about pH 8.0 during the entire transport. Due to the gradual accumulation of carbon dioxide during closed transport, acidity dropped to pH 6.65 after 5 h. For both transports, the initial concentrations of TAN were lower than the detection limit of the analytical method (<0.010 mg l⁻¹). After 0.5 h of open transport, TAN had increased to 0.554 mg l⁻¹ followed by a gradual decrease after 2.5 and 5.0 h. During closed transport, TAN increased to 2.95 mg l⁻¹ after 2.5 h and remained at a similar level until the experiment was terminated after 5 h. After subsequent recovery in the FT holding tank for 19 and 43 h, TAN was still somewhat elevated at 0.113 and 0.068 mg l⁻¹. When it comes to the toxic ammonia (NH₃) fraction, it is evident that the ammonia levels were very low (≤2.7 μg l⁻¹) at all times in both systems. The level of total organic carbon was fairly stable, between 1.4 and 1.9 mg l⁻¹, during open transport. After closed transport for 2.5 and 5 h, TOC was somewhat higher at 6.2 and 4.9 mg l⁻¹. In line with the elevated levels of TOC during closed transport, turbidity was elevated similarly (3.9 and 3.7 NTU) although not dramatically higher than during open transport (0.4–1.7 NTU). The Colour values were very low in all cases, below or near the detection limit of the method (<1), meaning the water did not become discoloured by, for example, possible blood from haemorrhages (see ‘red-bellied’ fish below).

3.2.3 Respiration and behaviour

As shown in Table 4, the mean respiration frequency after commercial transport and during open and closed simulated transports was 161, 137 and 153 beats min⁻¹ respectively. The swimming pattern was very different for fish in the open system, where the water flow was unidirectional with a circular motion, compared with the stagnant water in the closed system (Figure 2). In the open system, the fish distributed themselves in an organized manner against the water current, as commonly observed in fish cultures. In the closed system, however, single fish swam slowly around in the tank in a random fashion. By the end of the simulated closed transport, fish occasionally surfaced and exhibited a coughing behaviour. Gaping behaviour was also observed where fish stayed in a tilted position (tail downwards) at the water surface. Notably, when the fish were transferred to the FT holding tank for recovery, they almost immediately stopped coughing and they no longer stayed close to the surface.

3.2.4 Blood chemistry and white muscle biochemistry

4.1.1 Water quality

Overall, it was clear that water quality had deteriorated during transport but probably not to the extent where water quality alone (Table 1) unambiguously could explain altered behaviour and post-transport mortalities for some of the fish. Otherwise, see below for a more detailed discussion on water quality and recommended safe levels for the various water quality parameters. It is likely that TOC predominantly consisted of mucus (glycoproteins) and faeces (observed at the bottom of the transport tank since the fish were not fasted). Colour values in seawater have been determined as <2 (unpublished data). Hence, the Colour values (8–10) in transport tanks must be considered elevated. Possibly, the yellowish tint of the added AQUI-S™ may have affected the Colour values. For comparison with transport water quality (Table 1), another case can be mentioned where fasted freshwater rainbow trout were transported by vehicle at a fish density of 169 kg m⁻³ for 3.3 h at about 11°C. By the end of the transport, the levels of DO, acidity, TAN, NH₃, TOC and Colour were 310% saturation, pH 6.88, 5.0 mg l⁻¹, 8.0 μg l⁻¹, 0.79 mg l⁻¹ and 6 (−) respectively. In this case, no fish exhibited extreme abnormal behaviour and there were basically no delayed mortalities (Shabani et al., 2016).

4.1.2 Stress

The fish, sampled from a large production cage, were not fasted before transport. It is generally recommended that fish should be fasted for at least 48–72 h before transport to reduce metabolic rates, oxygen consumption and ammonia excretion (Wedemeyer, 1996a). Sufficient fasting time for emptying of gut contents before transport would also be beneficial to avoid water fouling during transport. Sufficient fasting time for emptying of gut contents before transport is also beneficial to avoid water fouling during transport. On the contrary, post-exercise recovery rates are not significantly affected by whether fish are fasted or not (Scarabello et al., 1991). The stress level, shortly after transport (Table 2), shows mean cortisol levels 577 and 653 nmol⁻¹. By comparison, the cortisol level of rested salmon is approaching zero (Gamperl et al., 1994; Iwama et al., 2004) while, for example, severe crowding of salmon in a commercial net pen at 12.6°C for 0–2.6 h resulted in cortisol levels of 665–736 nmol l⁻¹ (Erikson et al., 2016). The mean blood pH values after transport were 7.32 and 7.54. When rested salmon, with blood pH 7.85, were exercised to exhaustion, this resulted in extracellular acidosis and a pH-drop to 7.316 (Tufts et al., 1991), similar to the ‘bottom fish’ in the present study. It is difficult to make conclusions regarding the levels of glucose and sodium since their levels were within normal ranges (see below), and we did not know pre-transport levels at the farm. On the contrary, the mean chloride levels (181 and 191 mmol l⁻¹) of both fish groups were extremely high compared with typical levels (approximately 135–165 mmol l⁻¹) from previous studies of salmon in cage cultures (unpublished results). This indicates ionic and osmotic disturbances, something that can be associated with anaerobic exercise (see Cooke et al., 2008). The mean lactate levels of swimming and ‘bottom’ fish were 6.0 and 17.6 mmol⁻¹, respectively, both high compared with 3.1 mmol l⁻¹ after a 3.3-h closed transport of trout (Shabani et al., 2016) as well as with salmon severely crowded in a net pen, where the highest lactate value was 3.4 mmol⁻¹ (Erikson et al., 2016). The same analytical point-of-care method for lactate was used in all cases. In particular, the lactate level of ‘bottom fish’ must be considered dramatically high at >17.6 mmol⁻¹ (the values of two fish which exceeded the upper range of the instrument, 25 mmol l⁻¹, were excluded). At pH 7.24 and 6.83, swimming and ‘bottom’ fish confirmed that the latter group of fish had struggled to near exhaustion, whereas the swimming fish were moderately stressed since it has been shown for rested and exhausted Atlantic salmon that the initial pH are typically within pH 7.5 ± 0.1 and pH 6.7 ± 0.1 respectively (Erikson & Misimi, 2008). By comparison, after the already mentioned live transport of rainbow trout (Shabani et al., 2016) and the crowding of Atlantic salmon in the net pen, the initial pH values were about 6.9 and 7.0–7.1 respectively (Erikson et al., 2016). Regarding the severely disturbed acid–base status including the extremely high lactate levels of the ‘bottom fish’, it could be that both respiratory (water quality issues) and metabolic (exercise to exhaustion) acidosis contributed to the severe condition where the fish might have been moribund. Another indication of this was observed 20 h later where nine other fish were found dead in the holding tank. The appearance of these fish was normal with no prominent occurrences of red bellies, haemorrhages or loss of mucus and scales. These findings (extreme acidosis, exercise to exhaustion and delayed mortalities) may support the hypothesis put forward by Wood et al. (1983) suggesting that post-exercise mortality of rainbow trout may be related to severe intracellular acidosis. In their case, mortalities were observed 4 h post-exhaustive exercise at lactate levels about 12 mmol l⁻¹ (as measured by an established, well-validated laboratory method) and a blood acidity of approximately pH 7.35. AQUI-S™ was added to the closed tanks before the last leg of the journey to our facilities. Although we cannot rule out that it did not provide a stress mitigating effect, the fish nevertheless arrived in a stressed condition. A likely explanation could be that the fish were already severely stressed before AQUI-S™ was introduced into the vehicle's transport tanks, that is, after crowding in net pen, loading to live-fish carrier, and transfer of fish from vessel to vehicle.

4.2.1 Water quality

As a result of repeated sampling of fish, the fish density gradually decreased from 54 kg m⁻³ in the open system to 17 kg m⁻³ at final sampling of recovered fish. When fish density per se is taken into consideration, it is unlikely that it would be a significant factor in our study since Thorarensen and Farrell (2011) concluded that there is little, or no consistent effect of density up to about 80 kg m⁻³ on growth, survival and welfare of Atlantic salmon. Regarding DO, the question is whether the elevated levels of DO up to 134% in the closed system (Table 3) would be sufficiently high to significantly reduce respiration rates, which in turn would cause a build-up of CO₂ in the blood and cause acidosis (Hobe et al., 1984). Gas supersaturation can also induce emboli in tissues causing gas bubble disease, which would cause even greater problems if associated with nitrogen (Noga, 2000). When Atlantic salmon juveniles are exposed to DO supersaturated water (110–220%), changes in behaviour with regards to swimming activity, number of turns and panic reactions have been observed, demonstrating signs of pain and discomfort (Espmark et al., 2010). Hyperoxic water can also lead to elevated oxygen levels internally in the fish and might cause oxidative damages (Lygren et al., 2000). Furthermore, exposure of coho salmon smolts (Oncorhynchus kisutch) to hyperoxia (around 340% saturation) for 6 h resulted in osmoregulatory problems (Brauner, 1999). On the contrary, when rainbow trout were reared in FT systems at DO levels 94–180% saturation for 125 days, no differences were observed in growth and feed conversion, nor was mortality affected (Edsall & Smith, 1990). Moreover, no detrimental effects were observed when Atlantic salmon smolts were exposed to 123% DO saturation during a 3-h transport (Hosfeld et al., 2008). Notably, moderate oxygen supersaturation (<140%) did not cause harmful effects on blood chemistry and hepatic gluathione status of rainbow trout (Ritola et al., 1999). Taken together though, it is unclear to what extent DO supersaturation played a role regarding behaviour and stress during the present 5-h closed transport. The highest level of carbon dioxide encountered in this study was 24 mg l⁻¹ after 5 h closed transport (Table 3). The recommended maximum carbon dioxide levels in aquaculture to maintain good fish welfare and to support maximum growth of salmonids varies from 6 to 10 mg l⁻¹ (Fivelstad et al., 1998; Wedemeyer, 1996b) to <20 mg l⁻¹ (Portz et al., 2006; Timmons & Ebeling, 2007). According to Wedemeyer (1996a), the CO₂ level in a hauling tank water should be <20–30 mg l⁻¹ to prevent blood CO₂ from rising (hypercapnia), which would impair adequate transport of oxygen to tissues (Bohr effect). Rainbow trout change their normal swimming behaviour when CO₂ exceeds 35–60 mg l⁻¹ and equilibrium is lost at about 150 mg l⁻¹ (Clingerman et al., 2007). Furthermore, Atlantic salmon are lightly sedated at 70–80 mg l⁻¹ (Erikson, 2011). Therefore, it seems that the duration of the closed transport was not long enough to induce detrimental effects due to elevated levels of carbon dioxide per se. However, the elevated CO₂ levels resulted in a drop in acidity where the lowest level was pH 6.65 at the end of the closed transport. This value is still within the recommended levels for fish in aquaculture, namely pH >6 (Randall, 1991) and pH 6.5–8.5 (Timmons & Ebeling, 2007). The increase in TAN after 0.5 h might be explained by an initial stress response (increased metabolic rate) due to the transfer of fish from the holding tank. As the fish subsequently acclimated to the new tank environment, they may have become less stressed (less excretion of ammonia) resulting in diminishing TAN concentrations after 2.5 and 5.0 h. In the closed system, however, TAN gradually accumulated to about 2.9 mg l⁻¹ after 2.5 and 5.0 h. The dominant fraction of TAN, ammonium (NH₄⁺), has generally been considered not harmful (Tabata, 1962). The ammonia fraction was always ≤2.7 μg l⁻¹ in both systems. Although TAN increased significantly in the closed system, ammonia was always kept at bay due to the steadily decreasing pH. With respect to NH₃, suggested safe limits for salmonids varies between 12 and 25 μg l⁻¹ (Fivelstad et al., 1995; Timmons & Ebeling, 2007; Wedemeyer, 1996b, 1997). The TOC levels in the present study varied between 1.1 and 6.2 mg l⁻¹. By comparison, Davidson et al. (2009) reported TOC values of 4.64 and 20.52 mg l⁻¹ in RAS with high and low water exchange rates, respectively, where fish survival was high (>99%) in both RAS regimes. It is well-known that adverse water quality and stress can lead to excessive production of mucus (see review by Shepard, 1994). Mucus on gills arches can cause hypoxia, and in such cases, it has been observed that fish try to clear mucus from gills by a coughing behaviour (Ultsch & Gros, 1979). The turbidity levels of both closed (0.2–3.9 NTU) and open (0.4–1.7 NTU) transport were well below the recommended maximum limit of <20 NTU over ambient levels (Wedemeyer, 1997). Moreover, in seawater cage cultures, we have previously measured values around 0.2 NTU (unpublished results).

All in all, the water quality during open transport can be considered good since there were no plausible issues that could have induced severe stress reactions or compromised fish welfare. In case of the closed transport, water quality clearly deteriorated gradually. After 2.5 h, although several water quality parameters were elevated, they were still within acceptable levels according to recommended single-factor values. However, a joint impact of lowered pH and somewhat elevated levels of DO, CO₂, TAN, TOC and turbidity might have affected the fish in terms of behaviour, stress and welfare although we are unable to draw categorical conclusions. It should also be kept in mind that the recommended water quality safe limits in aquaculture mostly addresses long-term farming issues such as optimal on growth and health. Fish can probably tolerate less favourable conditions for a relative short period of time such as short, closed transports. Under the environmental conditions prevailing in the present study, it seemed that closed transport could be carried out safely when transport duration is less than approximately 2–3 h. For longer transports, at some point in time, water treatment will eventually become necessary to avoid excessive stress reactions, compromised welfare or mortality.

4.2.2 Respiration and behaviour

The respiration frequencies of fish 2 h after arrival to laboratory as well as during open and closed transports varied between 137 and 161 beats min⁻¹ at 9°C (Table 4). By comparison, the respiration frequency of severely stressed Atlantic salmon at 13°C in a commercial net pen increased to about 80 beats min⁻¹ after crowding for up to 80–100 min from 55–64 beats min⁻¹ in the bulk volume before crowding (Erikson et al., 2016). In rested rainbow trout at 15°C, values of about 53 beats min⁻¹ have been reported (Altimiras & Larsen, 2000). Just after about a 3-h commercial closed transport of adult rainbow trout at hyperoxic conditions at 10°C, the mean respiration rate was 130 beats min⁻¹. Concurrent analysis of blood chemistry showed the fish were severely stressed and the respiration rate remained high at 120 beats min⁻¹ after recovery for 48 h in a semi-closed system (Shabani et al., 2016). Consequently, the high respiration rates (Table 4) show that the fish were severely affected by commercial transport and that they remained high for several days post transport, showing the dominant role the commercial transport had on our fish.

In the closed system, the fish exhibited gaping behaviour and coughing during the last part of the transport. By then, the levels of DO, CO₂, TAN, TOC and turbidity were most elevated and the pH was low (Table 3). Related observations include brook trout (Salvelinus fontinales) where fish exposed to low pH often surfaced, a behavioural response indicating that fish are suffering deprivation of oxygen (Packer & Dunson, 1970), whereas gaping and coughing behaviour have been observed during closed transport with heavy oxygen supersaturation (Erikson, 2001). Furthermore, when rainbow trout were exposed to acidic water, an increase in coughing behaviour and ventilation frequency has been observed (Ye & Randall, 1991).

4.2.3 Blood chemistry and white muscle biochemistry

As discussed above, the fish had not yet recovered after the commercial transport before the simulated open transport started. Subsequently, the already elevated values of cortisol basically remained high throughout the duration of the entire experiment (Table 5). Similarly, blood pH did not change significantly and remined about 0.45 pH units below typical values of rested fish (see above), in spite of the drop in water acidity during the closed transport, from pH 8.05 to pH 6.65 after 5 h (Table 3). The levels of glucose varied between 3.5 and 6.1 mmol⁻¹ (Table 5). In rested fish, the concentration of glucose is typically around 3 mmol l⁻¹ and increases slowly after exposure to a stressor. Glucose then remains elevated for a long period of time (Iwama et al., 2004). In our study, the lowest glucose values appeared shortly after the commercial transport (Table 2), whereas the highest levels occurred 19 and 43 h after the simulated closed transport indicating a slow stress response. The lactate levels were low (0.9–3.2 mmol⁻¹, Table 5) throughout the simulated transport experiment. Typical lactate levels in rested salmon are close to, or below, the detection limit of the instrument (0.5 mmol l⁻¹). Lactate recovery from exhaustive exercise in rainbow trout has been shown to take place within 6–12 h depending on blood and white muscle pH (Milligan, 1996; Milligan et al., 2000), or in some cases, within 24–48 h (Franklin et al., 1990). Furthermore, slow recovery of fish can also be explained by elevated levels of cortisol (as in our case) delaying restoration of acid–base status and metabolite levels (Eros & Milligan, 1996; Pagnotta et al., 1994). The low lactate levels (Table 5) confirm our visual observations, that the fish did not respond to open and closed transport with excessive swimming activity. Of the ions (Na⁺, Cl⁻, K⁺ and Ca²⁺), the only significant change that took place was the marked increase in sodium after the simulated closed transport and the subsequent recovery after 19 h indicating the changes in water quality (Table 3) resulted in an ionic imbalance. Regarding the partial pressures of oxygen and carbon dioxide, the only significant change during the study was the increase in p_CO2 from 3.68 to 13.28 mm Hg in case of the closed transport (Table 5). During the transport, water CO₂ and pH changed from 0 to 24 mg l⁻¹ and 8.07 to 6.65 respectively (Table 3). On the contrary, DO supersaturation up to 134% did not reduce respiration frequency (Table 4), which could otherwise cause accumulation of CO₂ in the blood. In rainbow trout, the venous P_CO2 tension ranges from about 3.8 to 11.3 mm Hg (Harter et al., 2014) and according to data from several authors, in rested trout with blood pH of 7.74–8.02, P_CO2 ranges from 1.4 to 4.3 torr (as determined by traditional laboratory methods), see Thomas and Poupin (1985). In rainbow trout exercised to exhaustion, arterial pH was reduced from pH 7.88 to pH 7.37. This was accompanied by an increase in arterial P_CO2 from 3.03 torr (rested fish) to 7.79 torr. In this case, the fish recovered completely after 2 h (Turner et al., 1983). In another study, a 0.5 pH-unit drop was observed as P_CO2 rose from 1.88 mm Hg (rested trout) to 4.50 mm Hg (exhausted trout) as measured by established laboratory methods. The fish fully recovered after 2–4 h (Kieffer et al., 1994). In RAS and FT systems containing Atlantic salmon smolts, mean values of P_CO2 have been reported as 11.7–12.7 mmHg, respectively, as determined by the i-STAT system (Kolarevic et al., 2014). Previously reported haematocrit values in salmonids vary considerably making our values (21–32%) difficult to interpret. For example, as based on the traditional centrifugation method, Turner et al. (1983) reported Hct values in rested rainbow trout as about 21%, whereas Kolarevic et al. (2014) measured values around 35% when using the i-STAT system. Furthermore, Morgan and Iwama (1997) reported that normal values in fish are in the range of 30–50%. The initial pH in white muscle (pH 7.25–7.45, Table 5) were approaching the levels of rested salmon (pH 7.5 ± 0.1, see above). This is in line with visual observations and the low lactate levels discussed above. The gradual increase in the number of red-bellied fish (Table 5) might be a delayed effect of the severe stress bout during the commercial transport. Previously, we have occasionally observed more extreme cases ultimately leading to death several days after transport from farms (unpublished results). Such fish were always kept in FT tanks with good water quality and were not subjected to additional handling or experimentation.

Before the simulated transports were carried out, the fish were clearly stressed (control open, Table 5). It is, however, difficult to assess how stress responses would have been if the fish had been in a rested state before the simulated transport experiments were initiated. It could well be that moderate stress effects may have been masked to some extent (overshadowed by stress due to the commercial transport to the laboratory).