Dissolved oxygen criteria for freshwater fish in New Zealand: a revised approach

Abstract

Maintenance of suitable conditions in lowland rivers for both fish passage and resident species is crucial to ensuring the long-term sustainability of fish populations. The dissolved oxygen concentration of water is a key factor controlling habitat quality for fish and a critical measure of stream health. Continued land use intensification and greater exploitation of water resources has contributed towards increasing the frequency and duration of low dissolved oxygen events in lowland rivers and the associated risk of adverse effects on fish communities. Revised guidelines are therefore proposed to support setting of biologically relevant dissolved oxygen limits for the protection of New Zealand freshwater fish communities. These guidelines account for both event magnitude and duration, identify different protection levels based on the risk of negative impacts and are based on current scientific knowledge on the tolerances of New Zealand fish species.

Keywords:

• freshwater fish
• New Zealand
• dissolved oxygen
• guidelines
• limits
• lowland rivers
• fish communities

Introduction

Reports describing the quality of New Zealand's fresh waters have indicated a decline in water quality, and strong relationships between the degree of agricultural development within a catchment, lowered water quality, and reduced biodiversity of stream fauna (Quinn & Hickey 1990; Quinn et al. 1997; Hamill & McBride 2003; Larned et al. 2004; Wilcock et al. 2009). Typical pressures include elevated nutrient and sediment loads, loss of riparian shading, increased water temperatures, low dissolved oxygen, excessive growth of aquatic macrophytes and loss of physical habitat diversity (Wilcock et al. 1999; Larned et al. 2004). Lowland waterways are particularly vulnerable, being subject to multiple stressors and the cumulative effects of upstream pressures.

There is increasing evidence to suggest that in some areas, dissolved oxygen concentrations in lowland streams and rivers are falling below the recognised lethal thresholds for some fish species (Wilcock & Nagels 2001; Wilding et al. 2012). Oxygen is essential for the process of respiration and is therefore a limiting substance to most aquatic organisms. Consequently, the dissolved oxygen concentration of water is a key control of habitat quality and a critical measure of stream health.

Lowland rivers, including estuaries and tidal reaches, are important habitats for fish. They also act as critical migratory pathways to and from upstream habitats for diadromous fish species that undertake obligatory, regular and seasonally timed migrations between the sea and fresh water to complete their life cycles (McDowall 1990). Fish species such as inanga (Galaxias maculatus) and smelt (Retropinna retropinna) also utilise lowland river habitats for spawning. Maintenance of suitable conditions in lowland rivers for both fish passage and resident species is therefore crucial to ensuring the long-term sustainability of fish populations. Achieving this will be a significant challenge for managers of freshwater systems against a background of continued land use intensification and increased exploitation of water resources.

Low dissolved oxygen is most commonly encountered in un-shaded, slow-flowing, lowland rivers where aquatic plants are abundant. As the frequency and duration of low dissolved oxygen events becomes greater, the potential for impacts on fish community structure and population dynamics increases. In order to help control these impacts, managers require biologically relevant water quality limits. These limits must reflect both lethal and sub-lethal effects if they are to prevent losses of fish diversity and abundance, and provide an appropriate level of protection under the influence of multiple stressors.

This paper reviews existing knowledge of the dissolved oxygen tolerances of New Zealand fish species and considers the potential impacts of low dissolved oxygen on fish community structure. It then evaluates existing approaches to setting dissolved oxygen limits for protection of fish communities in New Zealand. Finally, new guidelines are proposed and the implications for management of lowland river systems are discussed. Directions for future research are also discussed.

Factors controlling dissolved oxygen in water

The main processes controlling dissolved oxygen concentrations in rivers are well established and widely described in the scientific literature (e.g. Chapra 1997; Wilcock et al. 1998; Cox 2003). They can be described by:

(1)

where C is dissolved oxygen concentration, C_s is the temperature dependent saturation value of dissolved oxygen, k₂ is the re-aeration coefficient, P is the rate of photosynthesis by plants at time t, R is the rate of respiration by plants at time t, and k₃ is the biochemical oxygen demand (BOD) decay coefficient.

Re-aeration is one of the most important factors controlling the concentration of oxygen in water and is affected by factors such as temperature, wind mixing, water depth and velocity, and the presence of morphological features such as waterfalls, dams and rapids (Cox 2003). Aquatic vegetation, including algae, can also exert a considerable impact on dissolved oxygen dynamics in lowland rivers (Wilcock et al. 1995; Kaenel et al. 2000; Desmet et al. 2011). Dissolved oxygen is produced by photosynthesis (P) during the day and consumed by respiration (R) continuously. The combination of these two processes can therefore produce strong diel and seasonal effects on dissolved oxygen concentrations (Wilcock et al. 1998; Goodwin et al. 2008). BOD is the amount of oxygen required by microorganisms as they consume organic matter in the water and reduces oxygen concentrations in water. As organic matter is processed and therefore decreases, the BOD is also reduced. Water temperature and salinity also control potential dissolved oxygen concentrations through their influence on the saturation capacity of water. As temperature and/or salinity increase, saturation capacity is reduced.

The importance of dissolved oxygen for fish

Dissolved oxygen is one of the most important environmental variables affecting the biology of fish (Alabaster & Lloyd 1982). During respiration, fish, like other animals, take in oxygen and release carbon dioxide. In most fish this is done using the gills, although some can also use the skin (Urbina et al. 2012) or have lung-like structures that can be used in addition to gills (Kind et al. 2002). When a fish respires, water is passed across the gills and oxygen diffuses into the blood through the gill filaments, and is subsequently transported to the tissues in the bloodstream. Simultaneously, carbon dioxide in the bloodstream diffuses into the water and is carried away from the body. A reduction in external dissolved oxygen levels can result in a shortage of oxygen in the tissues and elicit physiological and behavioural responses to compensate (Davis 1975; USEPA 1986; Kramer 1987; Pollock et al. 2007). Typical responses may include a reduction in activity to reduce energy expenditure, increased ventilation of the gills, increased use of aquatic surface respiration (ASR), increased use of air breathing or movement to habitats with a higher oxygen concentration (Kramer 1987; Kind et al. 2002; McNeil & Closs 2007; Pollock et al. 2007; Urbina et al. 2011; Urbina & Glover 2012).

Reduced oxygen availability inevitably results in changes in fish activity due to the coupling between oxygen and energy budgets within organisms (Kramer 1987). If oxygen availability is reduced, then the energy allocated to breathing must be increased in order to maintain oxygen supply to the tissues. Alternatively, if the energy allocated to breathing is to remain constant, then the oxygen allocated to other energetic requirements must be reduced. Consequently, changes in both breathing and activity are likely under conditions of reduced oxygen availability.

The most frequently observed alteration in behaviour following exposure to reduced dissolved oxygen levels is an increase in the ventilation rate of the gills (Doudoroff & Shumway 1970; Kramer 1987; McNeil & Closs 2007). This increases the flow rate of water across the gills in a bid to compensate for the reduced concentration of dissolved oxygen within the water. As the oxygen deficit increases, non-essential activity is often reduced in order to conserve energy. Feeding is often strongly affected because search, digestion and food assimilation are significant components of many fishes energy budget and thus are limited by oxygen availability (Doudoroff & Shumway 1970; Remen et al. 2012). Predator avoidance may also be altered as a result of differential tolerances to low dissolved oxygen between the predator and prey species, reduced swimming capability or enforced changes in habitat selection (Kramer 1987; Robb & Abrahams 2002; Landman et al. 2005; Pollock et al. 2007; Urbina & Glover 2012). Another compensatory response displayed by fish is ASR. Diffusion of oxygen from the atmosphere into the water occurs at the air-water interface meaning that oxygen levels are elevated in the surface film. Under low dissolved oxygen levels, ASR utilises this thin layer of higher dissolved oxygen to help meet the oxygen demand of fish (Kramer 1987; Dean & Richardson 1999; McNeil & Closs 2007; Urbina et al. 2011). However, the use of ASR comes at a cost of increasing predation risk by being close to the surface.

As dissolved oxygen progressively decreases, the energetic costs of breathing will increase and eventually, where possible, fish are likely to move to habitats with a higher dissolved oxygen concentration (Miranda et al. 2000; Pollock et al. 2007). Fish have frequently been shown to display a preference for locations with higher levels of dissolved oxygen (Doudoroff & Shumway 1970; Burleson et al. 2001), and have shown avoidance of normally preferred locations in the presence of hypoxic water (Richardson et al. 2001). However, habitat shifts may have costs in terms of food availability, predation risk and less desirable physico-chemical conditions (Pollock et al. 2007). If movement to a higher dissolved oxygen environment is not possible and low dissolved oxygen conditions persist, oxygen supply may be insufficient to meet the minimal energy demands of essential functions and fish will ultimately suffocate.

Dissolved oxygen tolerances of New Zealand fish species

Acute tolerances

Information regarding the dissolved oxygen tolerances of New Zealand's native fish species is relatively limited. Dean & Richardson (1999) assessed the acute tolerances of seven native New Zealand freshwater fish species and rainbow trout (Oncorhynchus mykiss) to low levels of dissolved oxygen by holding them in the laboratory at constant dissolved oxygen levels of 1, 3 and 5 mg/l for 48 h at 15°C. Common smelt at both the juvenile and adult life stages, juvenile common bullies (Gobiomorphus cotidianus) and juvenile rainbow trout were found to be the most sensitive to low dissolved oxygen, with 50% mortality at dissolved oxygen levels of 1 mg/l occurring after 0.6–0.7, 0.6 and 1 h, respectively, and 100% mortality for all species within 4 h. Juvenile banded kokopu (Galaxias fasciatus) were also relatively sensitive with 50% mortality at dissolved oxygen levels of 1 mg/l occurring in less than 8 h and 100% mortality by 12 h. Juvenile torrentfish (Cheimarrichthys fosteri) showed no mortality for the first 24 h of exposure, but 100% mortality by 48 h. Juvenile inanga (Galaxias maculatus) were shown to be more sensitive than the adult life stage, with 61% mortality after 48 h at 1 mg/l relative to 38% for the adult life stage. At a dissolved oxygen level of 3 mg/l, only juvenile trout responded, with fish moving to the surface to breathe indicating stress, but mortality was only 5% after 48 h. Shortfin (Anguilla australis) and longfin (Anguilla dieffenbachii) eels showed no response under the conditions tested.

Landman et al. (2005) tested the effects of constant low dissolved oxygen under laboratory conditions on a number of fish and invertebrate species by evaluating the dissolved oxygen concentration at which half of individuals died over 48 h exposure at 15 °C (48-h lc ₅₀). Their experimental set-up also prevented ASR by blocking access to the water surface. They found that juvenile inanga were the most sensitive fish, with a 48-h lc ₅₀ at a concentration of 2.6 mg/l. Common smelt and juvenile trout displayed similar thresholds with lethal concentrations of 1.8 and 1.6 mg/l respectively. Shortfin eel and common bully were the most tolerant species at this temperature with lethal thresholds of less than 1 mg/l. Shrimp (48-h lc ₅₀ 0.82) and freshwater crayfish (48-h lc ₅₀ 0.77) were also shown to be tolerant to low dissolved oxygen under the experimental conditions.

The results of these two studies provide a good illustration of the effects of differences in experimental methodology and hence the need for caution when interpreting such results for management purposes. Landman et al. (2005) observed 50% mortality of adult inanga after 48 h at 2.6 mg/l. However, Dean & Richardson (1999) observed only 38% mortality after 48 h at a lower dissolved oxygen concentration of 1 mg/l. Urbina et al. (2011) demonstrated the importance of ASR in inanga exposed to low dissolved oxygen and observed the use of emersion as an avoidance strategy. The disparity in results between the Landman et al. (2005) and Dean & Richardson (1999) studies therefore most likely reflects the importance of these strategies, which were prevented in the Landman et al. (2005) study, as a behavioural response for inanga to overcome low dissolved oxygen concentrations. The difference in results for smelt (50% mortality in 0.7 h at 1 mg/l (Dean & Richardson) compared with 50% mortality in 48 h at 1.8 mg/l for the Landman et al. study) suggests that ASR is less important as a coping strategy for this species. It also demonstrates that there is a very narrow threshold range over which the lethal effect of low dissolved oxygen is rapidly increased, an effect that has been observed for other species (Seager et al. 2000). Another factor that is also important to recognise about these studies is that in both cases fish were acclimated and trials carried out at a temperature of 15 °C. This is significantly lower than summer water temperatures in some lowland streams of the North Island (e.g. Wilcock et al. 1999; Wilcock & Nagels 2001). The metabolism and hence oxygen demand of fish varies with temperature. Consequently, oxygen tolerance thresholds have been shown to get higher with increasing temperature. Downing and Merkens (1957), for example, observed that lethal dissolved oxygen concentrations for a number of fish species increased by an average factor of 2.6 over a temperature range of 10–20 °C. Criteria based on the data presented by Landman et al. (2005) and Dean & Richardson (1999) are therefore likely to be under-protective at higher water temperatures.

Compared with New Zealand's native fish species, much more is known about the effects of low dissolved oxygen on salmonid species, including those present in New Zealand. This literature has been reviewed extensively by other authors (e.g. Doudoroff & Shumway 1970; Davis 1975; Alabaster & Lloyd 1982; USEPA 1986) and therefore will not be reviewed in detail here. However, it has been demonstrated that the acute tolerances of the more sensitive of New Zealand's native species are similar to those of rainbow trout (Dean & Richardson 1999; Landman et al. 2005). It is therefore useful to provide a brief summary of the critical findings in the salmonid literature to provide context for the New Zealand specific studies.

Studies on lethal tolerances of salmonid species vary considerably in terms of the test procedures used, duration, exposure regime, temperature and reported endpoints (Doudoroff & Shumway 1970). Consequently, reported acute thresholds can vary even for the same species. However, Alabaster and Lloyd (1982) and USEPA (1986) concluded that where conditions are otherwise favourable, acute lethal effects are likely to be avoided for most salmonid species by maintaining dissolved oxygen levels above 3 mg/l. It was noted nonetheless that sensitivity varied both between species and life stages and tended to increase with increasing temperature and duration of exposure (USEPA 1986). Silver et al. (1963), for example, observed 100% egg mortality for steelhead trout (Salmo gairdneri) at 1.6 mg/l, but greater than 80% survival at dissolved oxygen concentrations of 2.5 mg/l. Côte et al. (2012), however, observed reduced egg survival in Atlantic salmon (Salmo salar) under hypoxia (4.5 mg/l) relative to normoxic conditions (10.0 mg/l). Seager et al. (2000) showed for juvenile rainbow trout that mortality was significant at 1.6 mg/l for 6-h exposure, but not at 1.5 mg/l for 1-h exposure. The work by Seager et al. (2000) also showed that for a given duration of exposure, the threshold concentration range above which mortality does not occur and below which mortality rapidly becomes high was less than 1 mg/l. The difference between dissolved oxygen concentrations causing total mortality and those allowing complete survival (incipient thresholds) was also highlighted in the USEPA (1986) review and was suggested to frequently be about 0.5 mg/l when exposure duration was less than a week.

Data on other exotic species present in New Zealand, for example carp (Cyprinus carpio), goldfish (Carassius auratus) and tench (Tinca tinca), indicate that these species are generally able to survive for longer and at lower dissolved oxygen than salmonids and many of New Zealand's native fish species (Doudoroff & Shumway 1970; USEPA 1986). For example, Downing and Merkens (1957) reported 24-h constant dissolved oxygen lc ₅₀ values for carp ranging from 0.4 mg/l at 10 °C to 0.8 mg/l at 20 °C. Equivalent thresholds for tench at 10 °C were even lower at 0.2 mg/l. McNeil & Closs (2007) investigated the behavioural response of a number of fish species to progressive hypoxia and found that goldfish and carp were highly tolerant of hypoxia under laboratory conditions and that they may be able to survive in hypoxic (<1 mg/l) habitats for sustained periods of time through the use of ASR.

Chronic tolerances

Acute tolerance thresholds provide information suitable for derivation of limits to avoid direct mortality. However, to offer greater protection to the integrity of aquatic ecosystems, it is necessary to consider potential sub-lethal or chronic effects caused by low dissolved oxygen. This may include, for example, impacts on recruitment, growth, productivity, behaviour and/or habitat use. Information on the chronic tolerances of New Zealand species is again relatively limited, but considerable data are available for salmonid species (Davis 1975; USEPA 1986).

Richardson et al. (2001) investigated the avoidance behaviour of smelt, inanga and common bully to low dissolved oxygen. In this study, the fish were acclimated and trials were carried out at 20 °C. Fish were placed in a fluvarium, half of which was held at a dissolved oxygen of approximately 2 mg/l and the other at 8.5 mg/l, with free access between the two sides of the fluvarium. The behaviour of fish in response to the differences in dissolved oxygen was then observed over a 15-min trial period. Only smelt displayed avoidance behaviour to the low dissolved oxygen water, with inanga showing no significant negative response and adult bullies displaying a significant preference for low dissolved oxygen. No explanation was suggested for the preference for low dissolved oxygen displayed by bullies; however, they have been shown to have quite a high tolerance to low dissolved oxygen levels (Dean & Richardson 1999).

Bannon & Ling (2003) explored the effects of low dissolved oxygen and temperature on sustained swimming capability of inanga and rainbow trout parr. Trials were carried out at 10, 15 and 20 °C under both normoxic (>96% saturation) and mildly hypoxic (75% saturation) conditions, with fish acclimated to the respective trial temperatures prior to testing. Temperature and percentage saturation of oxygen are co-dependent. This means that in these experiments, whilst the percentage saturation was held constant between trails at different temperatures, the concentration of dissolved oxygen (which is not affected by temperature unless percentage saturation is greater than 100%) was different in every trial. That is, in the normoxic trials (96% saturation), dissolved oxygen concentrations were 11.3, 10.1 and 9.1 mg/l at 10, 15 and 20 °C respectively, whilst in the hypoxic trials (75% saturation) dissolved oxygen concentrations were 8.5, 7.6 and 6.8 mg/l at the three different temperatures respectively. This has potential implications for interpretation of the results as discussed below, depending on whether fish respond to thresholds in concentration or percentage saturation.

Both species in the Bannon & Ling (2003) study displayed temperature dependency in sustained swimming capacity. Maximum sustained swimming speed for juvenile inanga under normoxic conditions was 5.1 body lengths per second (BL/s) at 17 °C, but was reduced at both lower and higher temperatures. Under mild hypoxia no difference in swimming capability was observed at 10 °C (sustained swimming speeds c. 4.0 BL/s), but at 15 and 20 °C sustained swimming speeds were significantly reduced (c. 5.0 BL/s at both 15 and 20 °C under normoxia; 4.1 and 3.6 BL/s at 15 and 20 °C respectively under mild hypoxia).

Maximum sustained swimming speed for rainbow trout parr (5.8 BL/s) occurred at 15 °C under normoxic conditions, but decreased at lower (4.4 BL/s at 15 °C) and higher temperatures (4.5 BL/s at 20 °C). Under conditions of mild hypoxia, no effect on sustained swimming speed was observed at temperatures of 10 °C (4.4 BL/s) and 15 °C (5.9 BL/s), but at 20°C a significant reduction in swimming capability was observed with sustained swimming speed reduced from 4.5 to 2.5 BL/s.

The results of the Bannon & Ling (2003) study indicate that the influence of dissolved oxygen saturation on sustained swimming speeds in these species appears to vary with water temperature. However, due to the co-dependence of temperature and percentage saturation, it is unclear whether this reflects increased metabolic demands for oxygen at higher temperatures and hence reduced performance at constant oxygen saturation levels, or whether it is an indication that the fish are responding to the concentration of dissolved oxygen, which varies between each of the trials. The threshold oxygen concentrations (≥7.6 mg/l for inanga; ≥6.8 mg/l for trout) that correspond with altered behaviour in these experiments are similar to those observed in other studies (e.g. Urbina et al. 2011; Remen et al. 2012), potentially suggesting that oxygen concentration is more important.

Urbina et al. (2011) investigated behavioural and physiological responses of inanga to exposure to progressive hypoxia. They observed significant changes in swimming activity as dissolved oxygen declined below 7.3 mg/l. The time that inanga spent performing ASR also increased progressively as dissolved oxygen concentrations declined from normoxia (9.7 mg/l), but only significantly so when a concentration of 1.9 mg/l was reached. At this level, fish spent an average of 16.4% of the time performing ASR and at 1.5 mg/l, this increased to 29.0% of the time (Urbina et al. 2011). Avoidance behaviour, defined as when inanga tried to jump out of the water, was only observed at the two lowest oxygen concentrations that were tested. On average, 70% and 94% of fish exhibited this behaviour at 1.9 and 1.5 mg/l, respectively.

No further data have been published regarding the chronic effects of low dissolved oxygen on New Zealand native fish species. However, more wide ranging information is available for salmonids as reviewed by Doudoroff & Shumway (1970), Davis (1975) and USEPA (1986). For salmonid species, research has been carried out on the chronic impacts of low dissolved oxygen across all life stages (eggs, larval, juvenile and adult). It has been shown that lower dissolved oxygen concentrations can retard egg development (Coble 1961; Silver et al. 1963; Shumway et al. 1964; Ingendahl 2001; Malcolm et al. 2011; Côte et al. 2012), reduce growth and alter behaviour of larvae and juvenile life stages (Jones 1952; Whitworth 1968; Roussel 2007; Remen et al. 2012) and impact on the growth, behaviour and habitat use of adults (Bushnell et al. 1984; Plumb & Blanchfield 2011; Poulsen et al. 2011).

Silver et al. (1963) observed an increase in the median time to hatching for steelhead trout eggs subject to lower dissolved oxygen. At 2.6 mg/l, median time to hatching was 42.5 days. This decreased to 36 days at 7.9 mg/l and remained relatively constant at higher concentrations (Silver et al. 1963). Size at hatching was also affected, with mean size increasing from 15.75 mm at 2.6 mg/l to 19.80 mm at 11.2 mg/l, with a threshold occurring at around 6–7 mg/l. Shumway et al. (1964) observed similar effects for Chinook salmon (Oncorhynchus tshawytscha) at similar dissolved oxygen thresholds. Côte et al. (2012) also observed reduced survival and delayed hatching of Atlantic salmon (S. salar) eggs under hypoxic conditions (4.5 mg/l) compared with normoxic conditions (10.0 mg/l), whilst Ingendahl (2001) only observed larval emergence of brown trout (Salmo trutta) from redds in the Rhine at dissolved oxygen concentrations of 7 mg/l or greater.

Roussel (2007) investigated the effects of embryonic hypoxia (3 mg/l) on emergence and survival of the alevin life stage in brown trout. He observed that compared with those that developed under normoxic conditions (10.3 mg/l), alevins that had been subjected to hypoxia during development had delayed emergence from the gravel, had reduced swimming activity (20%) and greater predation mortality (14%). This suggests that there may be carry-over effects of early exposure to hypoxia into later stages of the life cycle.

Remen et al. (2012) researched the impact of cyclical exposure to hypoxia on feeding and growth in Atlantic salmon. They observed a reduction in feed intake in fish exposed to dissolved oxygen saturations down to 40–70% saturation (3.9–6.9 mg/l) at 16 °C. They suggested that 70% (6.9 mg/l) may be a threshold for impacts of reduced growth, and that dissolved oxygen may need to be maintained at greater than 60% (5.9 mg/l) to maintain a minimum level of protection against impairment.

Behavioural changes were also observed by Poulsen et al. (2011), who undertook a choice chamber test to evaluate the response of rainbow trout to progressive hypoxia. Changes in behaviour were first observed at a dissolved oxygen level of 60% saturation (c. 5.7 mg/l), with a reduction in swimming speed compared with normoxic conditions (c. 9.5 mg/l). However, avoidance behaviour in the form of reduced average residence time per trip was not observed until 50% saturation (c. 4.7 mg/l) was reached. The number of trips to hypoxia did not decline until dissolved oxygen was reduced to 30% saturation (c. 2.8 mg/l). This suggests that behavioural responses to reduced dissolved oxygen may be hierarchical, with an increasing range of responses utilised as the severity of hypoxia increases.

These examples demonstrate the wide range of mechanisms through which the effects of low dissolved oxygen can act on fish. They also highlight the important role that chronic effects could have in determining population and community scale responses to increasing hypoxia. Alabaster and Lloyd (1982) suggested that a minimum dissolved oxygen concentration of 5 mg/l should be sufficient to satisfactorily support activities at most life-stages. However, in several of the studies described above, initial responses were being detected at higher concentrations of 6–7 mg/l. The severity of impact then increases as dissolved oxygen concentrations progressively fall below this threshold. Recognition of this increasing level of impairment with falling dissolved oxygen concentrations led the USEPA (1986) to specify a sliding scale of dissolved oxygen thresholds associated with differing degrees of impact on fish production (Table 1). These figures are the basis of the dissolved oxygen criteria for freshwater ecosystems in the USA (USEPA 1986) and Canada (CCME 1999). Dean & Richardson (1999) also suggested that the USEPA figures may be an appropriate basis for defining dissolved oxygen criteria in New Zealand based on their results showing that the most sensitive of New Zealand's native fish species display similar acute tolerances to trout.

Table 1 Summary of USEPA (1986) dissolved oxygen thresholds (independent of temperature) associated with differing levels of production impairment in salmonids (mg/l).

Degree of impairment	Early life stages	Other life stages
No production impairment	11.0^a (8.0)	8.0
Slight production impairment	9.0^a (6.0)	6.0
Moderate production impairment	8.0^a (5.0)	5.0
Severe production impairment	7.0^a (4.0)	4.0
Limit to avoid acute mortality	6.0^a (3.0)	3.0

^a Recommended water column concentrations to achieve the required inter-gravel oxygen concentrations shown in parentheses.

Effects of dissolved oxygen on fish community structure

Different fish species and life stages vary in their tolerance and behavioural responses to low dissolved oxygen (Doudoroff & Shumway 1970; Davis 1975; Dean & Richardson 1999; Garvey et al. 2007). Consequently, dissolved oxygen concentrations will contribute towards determining species assemblages and community structure and functioning in aquatic systems, particularly when dissolved oxygen levels fall below tolerable limits (Kramer 1987; Smale & Rabeni 1995; Killgove & Hoover 2001).

The ability of fish to survive in low oxygen environments depends on the duration and timing of exposure, the level and constancy of dissolved oxygen, the species, life stage and health status of the fish, as well as on other environmental conditions, e.g. water temperature. Most fish species are capable of adapting their behaviour to compensate for short-term exposure to depressed dissolved oxygen levels (Kramer 1987). However, as the severity and duration of low dissolved oxygen conditions increase, the costs in terms of energy expenditure and vulnerability to predation also increase. Subsequently, changes in fish community structure caused by low dissolved oxygen are increasingly likely (e.g. Smale & Rabeni 1995; Justus et al. 2012).

Fish community structure and composition can be affected by both lethal and sub-lethal dissolved oxygen concentrations. It is expected that as dissolved oxygen concentrations progressively decline below acute thresholds (i.e. 3 mg/l), species richness at an impacted site will reduce as the acute lethal thresholds for individual species are surpassed.

Sub-lethal effects on fish communities may manifest in a range of ways, across a range of dissolved oxygen concentrations (USEPA 1986). Impacts on recruitment success, such as retarded embryonic or larval growth, may result in reduced abundance and eventual loss of more sensitive species and a shift in community composition towards more tolerant species. Limitations on migration, for example through avoidance behaviour or reduced swimming performance, could result in reduced inland penetration of species migrating upstream and/or reduced abundance either upstream or downstream of the impacted reach (e.g. Maes et al. 2007). In the long-term, persistent sub-lethal dissolved oxygen concentrations will likely translate through changes in population dynamics to reduced abundance and/or loss of more sensitive species, and a shift towards simpler communities dominated by the most tolerant taxa.

Justus et al. (2012), for example, observed reductions in maximum fish taxa richness and Ephemeroptera and Tricoptera richness with decreasing dissolved oxygen minima in lowland streams of southwestern Louisiana. The average threshold concentration observed for fish metrics was 2.3 mg/l. Smale & Rabeni (1995) also identified strong correlations between hypoxia tolerances of fish assemblages and observed dissolved oxygen minima in headwater streams in Missouri. They concluded that dissolved oxygen levels had a substantial effect on the composition of fish assemblages.

Understanding the impacts of low dissolved oxygen on New Zealand's fish communities is limited by the lack of knowledge about sub-lethal effects and the confounding influence of multiple stressors. However, inferences can be made regarding likely community responses based on the lethal tolerance thresholds that have been identified and comparisons with better studied salmonid fish species.

New Zealand's lowland fish communities, where land use intensification and exploitation of water resources is greatest, are at most risk of being impacted by low dissolved oxygen. However, given the high prevalence of diadromy in New Zealand's native fish communities (McDowall 1990), poor dissolved oxygen in lowland rivers and streams may also impact on recruitment to upstream habitats.

Where mean dissolved oxygen concentrations are persistently below 3 mg/l (i.e. below lethal thresholds), it is likely that the presence of sensitive fish species, e.g. inanga, smelt and trout, will be limited and that fish communities will be dominated by species tolerant of low dissolved oxygen such as the native shortfin eel and exotic species such as goldfish. At intermediate dissolved oxygen concentrations (<6 mg/l) the impacts are more likely to be a consequence of chronic effects, for example poorer growth rates, reduced fecundity or lowered recruitment success. The long-term implications for fish communities are, however, likely to be similar to those described above for more acute dissolved oxygen problems.

Observations from a number of small lowland Waikato streams demonstrate this shift towards fish communities dominated by more tolerant species. Communities where summer dissolved oxygen concentrations were low were characterised by low species richness, dominance by shortfin eel and common bully, and an absence of less tolerant species such as inanga (P Franklin, National Institute of Water and Atmospheric Research, Hamilton, unpublished data). There was also evidence to suggest a greater prevalence of exotic fish species including goldfish, rudd (Scardinius erythrophthalmus), catfish (Ameiurus nebulosus), gambusia (Gambusia affinis) and tench (T. tinca) (e.g. Franklin & Hodges 2012). However, in streams where dissolved oxygen was higher, species such as inanga, smelt and torrentfish (Cheimarrichthys fosteri) were present (P Franklin, National Institute of Water and Atmospheric Research, Hamilton, unpubl. data).

Dissolved oxygen guidelines for the protection of fish

Limits to water resource use are typically applied for two reasons. Firstly, they are imposed to constrain human-induced alteration of water quality and quantity to levels that are considered sufficient to sustain environmental values. Secondly, they are necessary to ensure that the total availability of water resources and their reliability is quantified and understood (Snelder et al. 2013). In order to set effective and transparent limits that meet environmental objectives, managers require biologically relevant guidelines that reflect both lethal and sub-lethal effects, and indicate an appropriate level of protection under the influence of an environmental stressor.

Establishing limits for dissolved oxygen is crucial for maintaining the life supporting capacity of freshwater ecosystems due to the importance of oxygen for respiration. However, defining appropriate guidelines for dissolved oxygen is challenging due to the natural spatial and temporal variability in the environment, limited knowledge of biological responses to changes in dissolved oxygen for New Zealand species, and the potentially confounding effects of additional stressors. To be of greatest benefit to managers, the guidelines should account for both the magnitude and duration of exposure, define how natural diel variability is taken into consideration, have a risk-based framework, be informed by the current state of biological knowledge and consider the potential impacts of multiple stressors.

In New Zealand, the Resource Management Act 1991 (RMA) proposes a default dissolved oxygen standard for the protection of aquatic ecosystems of at least 80% saturation. This guideline has been widely adopted by regional councils across the country. The aim of this threshold was to provide a default level of protection and to be an indicator of potential degradation if exceeded. However, it is unclear what level of protection this threshold was meant to provide and no account was made for how compliance should be measured against the background of natural diel and seasonal variability. In addition, by defining the standard as a percentage of maximum saturation, the threshold dissolved oxygen concentration decreases as water temperatures increase (i.e. 80% saturation at 10 °C is 9.02 mg/l and at 25 °C is 6.59 mg/l). This seems counterintuitive for ecosystem protection purposes given that the oxygen demand of aquatic fauna is generally considered to increase with increasing temperature (e.g. Downing & Merkens 1957). Davis (1975) also presented criteria based on percentage saturation values, but the saturation thresholds varied with water temperature. More commonly, criteria have been defined as a concentration (e.g. EIFAC 1973; USEPA 1986; CCME 1999). Support for this approach is primarily based on the apparently good fit between dissolved oxygen concentrations and ecological response thresholds in the existing literature (Doudoroff & Shumway 1970; USEPA 1986; Justus et al. 2012). Criteria expressed as a concentration also provide a greater protection level at higher water temperatures, without the need for a sliding scale of thresholds.

As pressures on water resources increase, the frequency and duration of low dissolved oxygen events in impacted rivers and streams is likely to become greater, and thus the potential for impacts on fish community structure and population dynamics also increases. Consequently, there is a need for improved guidelines to support limit setting in freshwater environments. Accordingly, the limits set out in Table 2 are proposed for New Zealand. These new guidelines have been designed to address not only the magnitude of low dissolved oxygen concentrations, but importantly also the duration of those concentrations. This is achieved by integrating both daily minima for prevention of acute mortality, and longer term averages for better protection against chronic impacts caused by episodes of continuous or regularly occurring low dissolved oxygen events (Table 2).

Table 2 Proposed dissolved oxygen levels for protection of New Zealand freshwater fish communities (independent of water temperature).

	Dissolved oxygen concentration (mg/l)
7-day mean
Guideline	8.0
Imperative	7.0
7-day mean daily minimum
Guideline	6.0
Imperative	5.0
Instantaneous minimum
Guideline	5.0
Imperative	3.5

Imperative protection level is the minimum recommended protection level for adult fish. Guideline protection level should be the target protection level or minimum for salmonids and the early life stages of all species. For salmonid spawning redds or embryonic stages of fish, this should be the interstitial concentration.

Two levels of criteria are proposed to provide both minimum and optimum levels of protection and to account for different fish communities and life stages (Table 2). This supports a risk-based approach for determining dissolved oxygen limits. The ‘imperative’ targets are the minimum acceptable criteria and are designed to minimise the likelihood of significant detrimental effects for the majority of fish species. Exceedance of these thresholds indicates that the risk of impacts on fish communities is likely to be high, but it does not provide certainty that an impact has occurred or will occur. The ‘guideline’ thresholds target optimising the protection of fish communities, but should also be considered the minimum requirement for communities dominated by salmonids or other sensitive species, e.g. smelt and inanga, and for early life stages. For spawning redds and embryonic life stages this should be the interstitial dissolved oxygen concentrations. The USEPA (1986) has suggested that water column concentrations of 3 mg/l higher are required to achieve the appropriate interstitial concentrations. Exceedance of these thresholds indicates increasing potential for negative impacts on fish communities.

The imperative instantaneous threshold (Table 2) is based on the results of Dean & Richardson (1999) who showed limited acute effects for New Zealand fish species at dissolved oxygen concentrations of less than 3 mg/l, and the consensus of other reviews (Alabaster & Lloyd 1982; USEPA 1986; Table 1), which suggest 3 mg/l as providing a default level of protection from significant acute impacts on fish. An additional 0.5 mg/l has been added to provide a buffer against multiple stressor effects, especially higher water temperatures (i.e. >15 °C as used by Dean & Richardson 1999). The guideline instantaneous level has been set at a level considered to avoid any greater than moderate chronic effects on most fish communities (Alabaster & Lloyd 1982; USEPA 1986).

In recognition of the influence of exposure duration, 7-day mean thresholds are also proposed (Table 2). Evidence suggests that under conditions of cyclical exposure to reduced dissolved oxygen, both the minimum and mean concentrations are correlated to outcomes for fish (Doudoroff & Shumway 1970; USEPA 1986). The 7-day mean daily minimum (i.e. the average daily minima over 7 days) is set to avoid regular exposure to the instantaneous limits. The imperative level is based on the threshold for avoiding any greater than moderate chronic effects on most fish communities. The guideline level is based on the USEPA (1986) value in Table 1 for no more than slight production impairment. The 6 mg/l threshold has also been identified by some authors (Poulsen et al. 2011; Remen et al. 2012) as the point at which chronic responses to low dissolved oxygen concentrations are first observed or as providing protection from significant impairment.

The 7-day mean thresholds (i.e. the average daily mean over 7 days) are set to provide long-term protection from impairment. The imperative level is set at 7 mg/l, which has been observed as a threshold for behavioural responses in inanga (Bannon & Ling 2003; Urbina et al. 2011) and salmonids (Ingendahl 2001; Remen et al. 2012). The guideline threshold is set at the no impairment level identified in the USEPA (1986) review (Table 1). This represents 100% saturation at water temperatures of 26 °C.

The proposed criteria should be achieved most of the time, but it is recognised that natural variation may result in some deviation beyond these thresholds. The most critical time is likely to be under summer low flows when water temperatures are high. Consequently, evaluation of compliance should focus on these conditions. It is also important that compliance be assessed through the use of continuous monitoring of dissolved oxygen. The significant diel variations that can occur in dissolved oxygen concentrations mean that one-off spot samples can be misleading and are highly dependent on the time of sampling.

The significance of conditions that fail to meet the recommended levels will depend on the magnitude, duration, frequency of occurrence and spatial extent of the breach. This can only be evaluated through the use of continuous monitoring. Consequences must also be dependent on the ecological value of the site. Evaluation of an event's significance will typically be on a case-by-case and site-by-site basis. Where natural conditions alone result in dissolved oxygen concentrations less than the proposed criteria, an imperative criterion of 90% of the natural concentration is proposed, with a guideline criterion of no reduction in the natural concentration.

It is important to note that these guidelines are designed to offer protection to fish communities, but do not take account of the effects on other aquatic fauna. Adherence to the criteria is likely to provide reasonable protection to other fauna, but there may be some sensitive species of macroinvertebrates which require a higher level of protection (Gaufin et al. 1974; Davis 1975).

Conclusion

As pressure on water resources continues to increase, the need to set water resource use limits on both the quantity and quality of water becomes ever greater. Limits should ideally be quantitative, measureable, transparent, founded in scientific knowledge and linked to objectives. Oxygen is a critical resource for the maintenance of aquatic ecosystem structure and functioning. To ensure adequate protection of the life supporting capacity of aquatic systems, there is a need for dissolved oxygen limits. Historical guidelines set out in the RMA and widely used as the foundation for setting dissolved oxygen limits in New Zealand do not effectively reflect current knowledge of the tolerances of New Zealand fish species. They also fail to provide guidance on their interpretation within the context of natural temporal and spatial variability in dissolved oxygen concentrations. The revised guidelines set out here attempt to address the main limitations of the old guidelines to help support improved limit setting for freshwater ecosystems.

While these revised guidelines reflect current published understanding of the dissolved oxygen tolerances of New Zealand fish species, it is recognised that there are still significant knowledge gaps regarding sub-lethal effects and the impact of diel variability on species. There is also little knowledge of how tolerances vary with temperature and under the influence of other environmental stressors. Consequently, particularly where instream values are high it is recommended that a precautionary approach to setting limits for dissolved oxygen be applied until these knowledge gaps are filled.

Future research to support the derivation of more robust, data driven criteria for New Zealand should focus on clarifying both acute and chronic incipient thresholds (i.e. no effect thresholds) for New Zealand species and how these vary under differing exposure regimes. For example, understanding how thresholds vary with increasing duration of constant exposure and under the influence of cyclical dissolved oxygen regimes representative of natural diel variations would be beneficial. It would also be valuable to investigate the influence of temperature and other potential confounding stressors on responses to different dissolved oxygen concentrations. An area currently lacking consensus is the effect of high dissolved oxygen concentrations (>100% saturation) on fish. Reported instances of mortality caused by gas bubble trauma in fish under conditions of gas super-saturation below dams (Lutz 1995; Backman & Evans 2002) are not consistent with results of investigations into the effects of oxygen super-saturation in aquaculture, with a review by Dong et al. (2011) suggesting that no adverse effects or abnormal behaviour is observed in fish when exposed to dissolved oxygen saturations up to 200%. It may also be beneficial to resolve whether percent saturation or concentration is the main control on fish responses.

Acknowledgements

Thanks to Dave Rowe, the editor and two anonymous reviewers for constructive comments on the manuscript. This work was funded by NIWA capability fund project CF102230 and supported by Waikato Regional Council.