Lessons from two high CO 2 worlds – future oceans and intensive aquaculture

Abstract Exponentially rising CO 2 (currently ~400 μatm) is driving climate change and causing acidification of both marine and freshwater environments. Physiologists have long known that CO 2 directly affects acid–base and ion regulation, respiratory function and aerobic performance in aquatic animals. More recently, many studies have demonstrated that elevated CO 2 projected for end of this century (e.g. 800–1000 μatm) can also impact physiology, and have substantial effects on behaviours linked to sensory stimuli (smell, hearing and vision) both having negative implications for fitness and survival. In contrast, the aquaculture industry was farming aquatic animals at CO 2 levels that far exceed end‐of‐century climate change projections (sometimes >10 000 μatm) long before the term ‘ocean acidification’ was coined, with limited detrimental effects reported. It is therefore vital to understand the reasons behind this apparent discrepancy. Potential explanations include 1) the use of ‘control’ CO 2 levels in aquaculture studies that go beyond 2100 projections in an ocean acidification context; 2) the relatively benign environment in aquaculture (abundant food, disease protection, absence of predators) compared to the wild; 3) aquaculture species having been chosen due to their natural tolerance to the intensive conditions, including CO 2 levels; or 4) the breeding of species within intensive aquaculture having further selected traits that confer tolerance to elevated CO 2. We highlight this issue and outline the insights that climate change and aquaculture science can offer for both marine and freshwater settings. Integrating these two fields will stimulate discussion on the direction of future cross‐disciplinary research. In doing so, this article aimed to optimize future research efforts and elucidate effective mitigation strategies for managing the negative impacts of elevated CO 2 on future aquatic ecosystems and the sustainability of fish and shellfish aquaculture.

Introduction -Climate change, high CO 2 and global food security In 2015, atmospheric CO 2 concentrations had risen to an annual average higher than 400 latm the first time in over 800 000 years (L€ uthi et al., 2008;Dlugokencky & Pieter, 2016), as a result of anthropogenic CO 2 emissions. The potential implications of this postindustrial rise in CO 2 were predicted over 110 years ago (Krogh, 1904); yet, it was only recently that governments agreed to take action on this issue. Despite 196 nations taking an unprecedented stance on climate change last year by signing the COP21 agreement to curtail emissions, CO 2 concentrations are still projected to approach 1000 latm by 2100 (P€ ortner et al., 2014). Around a quarter of anthropogenic CO 2 emissions have been absorbed by the oceans (P€ ortner et al., 2014). Whilst this results in a phenomenon commonly referred to as ocean acidification, elevated atmospheric CO 2 is also driving a large elevation in the average aquatic CO 2 in fresh and brackish water systems, regardless of diurnal and seasonal variation. What is more, seasonal oscillations of aquatic CO 2 in the future are predicted to amplify over time which will likely result in CO 2 levels that exceed 1000 latm for several months each year well before 2100 (McNeil & Sasse, 2016). Occurring simultaneously with warming, pollution, habitat degradation, disease outbreaks and overfishing, this aquatic acidification is therefore threatening not only aquatic ecosystems but also global food security (FAO, 2014, Porter et al., 2014. Anthropogenic CO 2 emissions accelerate alongside growth of the global human population, which is projected to exceed 9.6 billion by 2100 (Gerland et al., 2014). This same growth has also resulted in at least 80% of world fish stocks being overexploited (FAO, 2014, Pauly & Zeller, 2016. Aquaculture is therefore crucial to ensure the continued provision of fish and shellfish protein for human consumption, particularly for developing countries and small island nations (Bennett et al., 2016). Indeed, aquaculture is one of the fastest growing food-producing industries globally (8.8% annual growth for the last 30 years) (FAO, 2014), and it is the only foreseeable way of increasing seafood 1 production in the face of this human population expansion. However, to ensure aquaculture is able to maximize its potential for addressing global food security, a number of challenges need to be resolved concerning water availability and quality, environmental impacts and vulnerability to changing climatic conditions. Recirculating aquaculture systems (RAS) address many of these issues (Martins et al., 2010) and enable the sustainable intensification of aquaculture. These systems significantly reduce water requirements, relocate production of aquatic organisms away from a natural environmental setting and minimize environmental impacts. They also enable a tighter control of pathogens and other environmental parameters, potentially improving animal welfare and biosecurity, but they create some additional problems, particularly associated with accumulation of CO 2 .

A common problem, two perspectives
Physiologists have known for decades that raising the CO 2 partial pressure in water to well above atmospheric levels (e.g. 10 000 latm) has a direct effect on aquatic organisms in terms of acid-base and ion regulation, respiratory function and aerobic performance (Cameron & Randall, 1972). More recently, climate change studies have shown that CO 2 levels projected for end of this century (e.g. 800-1000 latm) can negatively affect development, physiology and fitnessrelated behaviours in aquatic animals (see below). Due to the very high stocking densities achieved in most aquaculture settings, as well as the methods employed to control pH and O 2 , CO 2 often accumulates, particularly in RAS. However, despite recent evidence on the potential detrimental effects of CO 2 exposure at a level projected for 2100 (1000 latm), the aquaculture industry was intensively farming fish and shellfish successfully at much higher CO 2 levels long before the term 'ocean acidification' was coined. The levels at which the effects of CO 2 are perceived as problematic, therefore, appear to differ greatly between the connected yet traditionally disparate fields of climate change and aquaculture (Fig. 1).
Current guidelines for intensive RAS propose safe CO 2 levels ranging from 15 to 40 mg L À1 (Fivelstad et al., 1999(Fivelstad et al., , 2015Blancheton, 2000;Petochi et al., 2011). These equate to an upper limit of CO 2 ranging from >5000 to >30 000 latm which are 12.5 to 75 times higher than current atmospheric levels, respectively. Furthermore, far from being an issue exclusively associated with RAS and finfish production, elevated CO 2 levels appear synonymous with intensive aquaculture more generally. For example, over 40% of Norwegian salmon smolt hatcheries (flow-through and RAS) report CO 2 levels >5400 latm (Noble et al., 2012), whereas Bangladeshi shrimp ponds are shown to experience CO 2 levels averaging >17 000 latm (Saksena et al., 2006;Sahu et al., 2013).
In stark contrast, recent studies emerging from aquatic acidification research have demonstrated that just 2.0-to 2.5-fold increases in CO 2 levels projected for the end of this century (e.g. 800-1000 latm) can have dramatic and long-lasting effects on the development, physiology and behaviour of both fish and invertebrates (Briffa et al., 2012;Schalkhausser et al., 2012;Heuer & Grosell, 2014;Watson et al., 2014;Welch et al., 2014). For example, exposure to 1000 latm during early life cycle stages has been shown to result in reduced survival as well as a number of sublethal effects including tissue damage (e.g. Frommel et al., 2012Frommel et al., , 2014Chambers et al., 2014), altered calcification (e.g. Arnold et al., 2009;Maneja et al., 2013), reduced size (e.g. Talmage & Gobler, 2009;Maneja et al., 2014), reduced metabolic rate (e.g. Small et al., 2016), delayed development and altered gene expression (e.g. Tseng et al., 2013;Goncalves et al., 2016) in a range of different marine organisms. What is more, similar effects are also demonstrated in freshwater, with Ou et al. (2015) showing a significant effect of elevated CO 2 (1000-2000 latm) on the larval development of pink salmon Oncorhynchus gorbuscha. The authors reported a reduction in larval length, total wet and dry mass and reduced production efficiency (conversion of yolk into tissue growth).
In impacting a diverse array of aquatic organisms during early life stages, increased partial pressure of CO 2 in aquatic environments above present-day atmospheric levels is likely a bottleneck for organism production. This in turn would significantly impact aquaculture practices that depend upon a reliable source of larvae or juveniles. In 2007, these impacts were realized with the upwelling of elevated CO 2 , aragonite undersaturated sea water off the US west coast, significantly impacting oyster hatchery production as a 1 Seafood in this context refers to all fish and shellfish species produced under fresh, brackish or marine conditions and intended for human consumption direct result of changing climatic conditions (Barton et al., 2012). In addition to providing a case study in which to investigate the impact of ocean acidification on shellfish production globally, this event highlighted the significant advances achieved when climate change scientists and aquaculture practitioners work closely together. Unifying their research efforts to overcome this phenomenon, the climate change community and shellfish growers were able to successfully identify the root cause of this issue and put in place a number of mitigation strategies and monitoring protocols to minimize impacts in the future (Barton et al., 2015).
Far from being restricted to early life stages, a growing number of studies have also shown sublethal physiological impacts of elevated CO 2 (range 1000-2000 latm) in a number of species which include impacted respiratory gas transport, acid-base balance and gut carbonate excretion (e.g. Lannig et al., 2010;Esbaugh et al., 2012Esbaugh et al., , 2016Heuer et al., 2012;Wei et al., 2015). Rapid and efficient acid-base compensation has been demonstrated in a number of species at elevated CO 2 concentrations (e.g. Melzner et al., 2009;Lewis et al., 2016). However, such physiological responses incur energetic costs and could therefore have negative implications for production efficiency and body condition both in aquaculture and natural settings. Likewise, a wide range of behaviours are shown to be disrupted under elevated CO 2 , such as Fig. 1 Diagrammatic representation of the levels at which elevated carbon dioxide is considered problematic within recirculating aquaculture systems (RAS) (caused by accumulation of excreted CO 2 due to high stocking densities) and under global aquatic acidification (marine and freshwater, caused by rising atmospheric CO 2 ). Numbered arrows, and corresponding key indicate the levels at which CO 2 is demonstrated to have significant impacts on fish development, physiology and behaviour. The expanded view on the right side highlights CO 2 levels in relation to climate change scenarios in greater detail (0-3000 latm or 0-4 mg L À1 ). Conversion of CO 2 levels between latm and mg L À1 in this diagram is based on 35 psu sea water at 15°C. Fish images Kovalevska and Kazakov maksim/shutterstock.com. References corresponding to numbered arrows indicate levels of CO 2 shown to have a significant impact of fish development, physiology or behaviour; 1) Hamilton et al. those linked to sensory stimuli (including smell, hearing and vision; e.g. Simpson et al., 2011;Nilsson et al., 2012;Roggatz et al., 2016) and cognitive-related functions (such as lateralization, learning, bold-shy phenotypes and escape behaviour; e.g. Schalkhausser et al., 2012;Hamilton et al., 2014;Watson et al., 2014), which will have clear detrimental implications at the population level (Munday et al., 2009(Munday et al., , 2010Chivers et al., 2014). However, animals reared in many aquaculture settings are living in a relatively benign environment, being provided with abundant food, relatively constant environmental conditions, protection against disease and absence of a predation threat. Therefore, it is perhaps not surprising that the ecologically relevant physiological and behavioural disruptions caused by end-of-century CO 2 levels in OA studies have not emerged from aquaculture studies. Equally it may be possible these behavioural effects have not been noted as they are not typically measured in aquaculture studies. Nevertheless, this does not mean that animals reared in an aquaculture setting are not facing problems associated with elevated CO 2 that potentially influence their health and/or production efficiency.

Cross-discipline interaction to improve understanding of CO 2 consequences
Given these contrasting views, combining the knowledge that has arisen from climate change and aquaculture research is crucial to allow a more in-depth understanding of the physiological and ecological responses of aquatic animals to elevated CO 2 . The opportunity to compare these two fields directly is appealing, and should enable a more accurate prediction of the consequences of changing climatic conditions for wild populations and intensive aquaculture practices alike. However, at present, such comparison is not straightforward. This is partly due to the different experimental measures and reporting protocols typically adopted by each of these scientific fields. To facilitate this process, it would be fruitful to develop a collective research agenda and implement standard operating procedures with respect to hypothesis development, experimental outcomes and data reporting.
The comparison is also complicated by rather different species often being used in aquaculture compared to OA research, with the former inevitably relying on species that are amenable to domestication, which may go hand in hand with greater environmental tolerance. Indeed, when considering contrasting results from aquatic acidification and aquaculture fields, it is worth noting that responses from even closely related species can often vary significantly. For example, Ferrari et al.
(2011) demonstrated a striking and unexpected difference for the impact of CO 2 on the antipredator response of closely related damselfish species. Similarly, Lefevre (2016) and Heuer & Grosell (2014) highlight heterogeneity in physiological responses to elevated CO 2 that argues against a unifying physiological theory for defining CO 2 tolerance, and which needs to be accounted for when modelling and predicting the impacts of climate change. Indeed explaining such interspecies variability with respect to CO 2 tolerance may provide a mechanistic understanding of why species used in aquaculture may be relatively tolerant to the CO 2 levels prevalent within intensive production. However, it is important to note that even cod reared under end-of-century CO 2 levels (1000 latm) exhibit avoidance behaviour towards these conditions when presented with a choice, indicating negligible habituation and suggesting these conditions are unfavourable (Jutfelt & Hedg€ arde, 2013). Furthermore, a growing body of evidence shows that levels of CO 2 experienced in aquaculture may be more detrimental than traditionally perceived (Heuer & Grosell, 2014). For example, Tirsgaard et al. (2015) and Ou et al. (2015) demonstrated detrimental effects of elevated CO 2 in cod and salmon, respectively, species traditionally grown successfully under aquaculture settings. Exposure to 9200 latm resulted in longer meal processing time and less efficient digestion in cod (Tirsgaard et al., 2015), whilst exposure to 2000 latm reduced growth and production efficiency in salmon larvae (Ou et al., 2015), end-point measures that are of specific importance to aquaculture production. Thus, differences between these two fields in the perceived impact of elevated CO 2 cannot be explained solely by variability in interspecific responses. Measuring the impact of elevated CO 2 on a diverse array of physiological and behavioural endpoints, not just those traditionally perceived as important for aquaculture production, is thus vital. It is also crucial to measure these responses in as many species as possible, both finfish and shellfish, as well as those traditionally perceived as CO 2 tolerant and CO 2 sensitive. By doing so, it will be possible to optimize water quality parameters within aquaculture, based on a species-specific suite of physiological and behavioural CO 2 tolerance endpoints. Targeting these conditions has the potential to maximize growth efficiency and health of aquaculture species, enhancing the sustainability of seafood production. With that aim, it is critical to understand the practical considerations of reducing and maintaining environmental conditions, particularly CO 2 , in an aquaculture context. Targets should thus be set that optimize productivity and welfare of the aquaculture species, but which are equally achievable in a practical and economical context (Noble et al., 2012).
To optimize research efforts and ensure data are both scientifically robust and comparable, a unified protocol for selecting, manipulating, measuring and finally reporting carbonate chemistry parameters is also needed. This is of particular importance given the methods of carbonate chemistry manipulation employed within intensive aquaculture, for example the addition of a strong alkali to buffer changes in pH, such as sodium hydroxide (NaOH), sodium bicarbonate (NaHCO 3 ), calcium hydroxide (Ca(OH) 2 ) or calcium oxide (CaO). This is the most commonly used of all water chemistry quality management practices in aquaculture, being typically employed in a diverse array of aquaculture settings (Boyd et al., 2016). However, this method of pH compensation additionally elevates alkalinity, often significantly beyond any natural analogue (Ellis et al., in preparation), and depending on the alkali used can have dramatic indirect effects on additional water chemistry parameters, some of which are shown themselves to influence a number physiological processes in aquatic organisms (Boyd et al., 2016;Middlemiss et al., 2016). A further crucial issue is the selection of experimental controls representing present-day CO 2 levels (400 latm), and we propose this should be a common reference point for both climate change and aquaculture researchers. Control levels employed within aquaculture research typically exceed 1000 latm (range 1000-3000 latm) (Fivelstad et al., 1999;Petochi et al., 2011) and thus surpass most of the 'high CO 2 ' treatments used as end-of-century projections in climate change studies. To complicate matters further, reporting CO 2 levels as mg L À1 in aquaculture studies overlooks the impact of temperature and salinity on the solubility of CO 2 and the resulting impact these have on the partial pressure of this gas (Weiss, 1974;Dickson, 2011). For example, for the same mg L À1 concentration, the actual partial pressure of CO 2 varies by more than threefold between cold freshwater and warm sea water (Fig. 2). This is critical because it is partial pressure (not the mg L À1 concentration) that determines the internal (blood) levels of CO 2 and its impact on physiology, behaviour, growth, etc. At present, the scarcity of sufficient water chemistry parameters being presented, the lack of environmentally relevant controls and the prevalence of reporting CO 2 levels as mg L À1 in aquaculture literature preclude an unambiguous comparison between data from these two fields.
Finally, understanding and reporting the provenance of the study species/population will be important to enable a more in-depth assessment of CO 2 tolerance, that is whether animals are wild-caught, laboratorybred or reared within an aquaculture setting (potentially already at very high CO 2 when considered in a climate change experimental context). It is fair to say that many (though not all) laboratory-based climate change studies benefit from easy access to study species available from aquaculture. The systematic selection of traits of interest by the aquaculture industry, such as fast growth and resistance to pathogens, has inherently selected for good performance under intensive farming conditions. In that context, it is possible, and even likely, that additional nontarget traits have also been selected, potentially including those involved in CO 2 tolerance. Indeed, enhanced CO 2 tolerance has been demonstrated in selectively bred populations of the Sydney rock oyster, compared to its wild-type congeners (Parker et al., 2011Thompson et al., 2015). Furthermore, as demonstrated by Malvezzi et al. (2015), early life survival at elevated CO 2 concentrations can have a significant additive genetic element (i.e. highly heritable), which under sufficient selection pressure could elicit a strong and rapid evolutionary response. It is highly likely therefore that aquaculture practices operating at elevated CO 2 concentrations would elicit sufficient selection pressure to directly select for CO 2 tolerance during early life stages, leading to the rapid Fig. 2 Schematic representation of the conversion of 1 mg L À1 dissolved CO 2 concentration into partial pressure (latm) at a range of different temperatures and salinities. This shows the very large influence of temperature in particular (up to 3.2-fold higher partial pressure at the warmest temperature compared to the coolest) but also salinity (up to 26% higher partial pressure at the highest salinity compared to freshwater) on the CO 2 partial pressure due to the impact these abiotic factors have on the solubility of CO 2 in water (Dickson, 2011;Weiss, 1974). Conversion of dissolved CO 2 in mg L À1 to partial pressure in latm was undertaken using the CO2SYS programme (Pierrot et al., 2006), using dissociation constants from Mehrbach et al. (1973), refit by Dickson & Millero (1987), and KSO 4 using Dickson (1990), with values for CO 2 solubility at different temperatures and salinities checked against Weiss (1974). evolution of the population in just a few generations. Thus, exploring the traits selected for in broodstock within intensive aquaculture offers a fascinating opportunity to investigate multigenerational adaptation to CO 2 levels experienced under intensive production conditions in aquaculture species. In addition, it will be vital to undertake multigenerational studies in order to discern the transgenerational acclimation to elevated CO 2 of different fish species with respect to different behavioural (e.g. Welch et al., 2014) and physiological (e.g. Miller et al., 2012) endpoints. Combining the understanding from these two fields will therefore help determine the physiological basis for CO 2 tolerance, determine its true ecological consequence and determine its ecological impacts over relevant timescales.

Conclusions
The yield from wild capture fisheries has plateaud since the late 1980s and human consumption from aquaculture exceeded that from wild sources for the first time in 2014 (FAO, 2014). Furthermore, as stated previously, aquaculture is likely to be the only pathway for increasing seafood production in the future. Moving from a capture to a culture mentality requires a shift in attitude that will require time, a luxury that is illafforded in the rapidly changing environment of the Anthropocene. Creating opportunities for the aquatic acidification community and the aquaculture industry to work together should help to speed up this process and enable the aquaculture industry to rapidly adapt by using better-informed decisions to a) optimize the water chemistry conditions within intensive aquaculture to suit the species and/or b) select traits within the species to suit intensive aquaculture conditions. This will help address the environmental, economic and social impacts of this developing sector towards a sustainable intensification of production, enhancing food security and its resilience to climate change. Equally, this cross-discipline interaction should also improve our capability to predict and mitigate the consequences of the changing chemistry for natural ecosystems in a future 'high' CO 2 world. FAO (2014) (2015) Growth and physiological models for Atlantic salmon (Salmo salar L.) parr exposed to elevated carbon dioxide concentrations at high temperature. Aquaculture, 436, 90-94.