Wilson A. Lennard

University of South Australia, UniSA STEM, South Australia, 5000, Australia

Keywords:

ABSTRACT

1.Introduction

Recirculating aquaculture systems (RAS) require additions of a buffer or basic salt to counteract the acidification (hydrogen ion production) that occurs as a consequence of the nitrification of the ammonia waste released by the fish within the biofilter by nitrifying bacteria(Espinal & Matulic, 2019; Masser et al., 1999; Timmons et al., 2002).The nitrification process may be represented in chemical equation form thus:

Overall, the combined equations become:

The conversion of 1 mol of ammonia to 1 mol of nitrate requires 2 mol of oxygen and causes the net release of two hydrogen ions and 1 mol of water. Because of this, the bacterial mediated conversion of ammonia to nitrate within a biofilter causes an influx of hydrogen ions into the recirculating water, which leads to a fall in pH and the acidification of culture waters (Ebeling, 2000). Bacterial mediated nitrification occurs most efficiently in a pH range of 6.5–8.0, so buffer additions of basic salts are generally applied to maintain pH levels within this range (Masser et al., 1999; Timmons et al., 2002).

The most common basic salts used in RAS fish culture are sodium bicarbonate (NaHCO) and calcium carbonate (CaCO) (Masser et al.,1999). However, calcium carbonate tends to dissolve too slowly and may not buffer pH efficiently if rapid accumulation of nitrification-driven hydrogen ion production occurs and therefore, sodium bicarbonate is generally recommended (Masser et al., 1999).Hydroxide-ion based buffers may also be used, but these tend to be highly caustic in nature and fish may be harmed if zones of very high pH are created (Masser et al., 1999).

Recirculating aquaponic systems (also known as “coupled” aquaponic systems), which grow both fish and plants, contain either a dedicated biofilter (as seen in fish-only RAS) and a hydroponic component for the plants, or a hydroponic component only that has the ability to act as a biofilter (e.g. gravel beds; the gravel acts as a medium for the colonisation of nitrifying bacteria) (Lennard, 2017; Palm et al., 2019;Rakocy & Hargreaves, 1993). If balances between the biomass of fish grown (and hence, the amount of metabolic waste produced) and the nutrient removal ability of the plants are matched, the amounts of buffer required in aquaponic systems should be lower than that of fish-only systems (Lennard, 2017). This is because plants are known to release negatively charged, alkalising ions (OH, HCO) when they are actively taking up nutrient ions such as nitrate (NO) (Imsande &Touraine, 1994; Maucieri et al., 2019, p. pp77). This is generally a one-to-one exchange mechanism (when one nitrate ion is assimilated,one alkalising ion is released) since its primary function is to balance the homeostatic charge status within the plant root (Imsande & Touraine,1994; Lennard, 2017; Maucieri et al., 2019, p. pp77). However, the bacterial mediated nitrification of one ion of ammonia to one ion of nitrate releases two hydrogen ions (see equations above); therefore, the release of alkalising ions during active plant assimilation of nitrate cannot completely counteract the acidification caused by the nitrification of soluble fish waste ammonia. Therefore, as in fish-only RAS,aquaponic systems often require the addition of a basic buffer, however due to the activity of the plants, this usually require less buffer additions than equivalent Fish-only RAS (Lennard, 2017).

When buffer is added to a system (fish-only or recirculating aquaponic), it is only the negative ion component that is used for pH buffering (i.e. when sodium bicarbonate is added, it is the bicarbonate portion which does the buffering). Buffers are universally added as salts,as these are easier to acquire and handle (Lennard, 2017). Hence, when buffers are added to aquatic systems, the positive ion component is unused and begins to accumulate within the system water (Lennard,2017; Rakocy & Hargreaves, 1993). In fish-only systems, when buffered with sodium bicarbonate, sodium begins to accumulate. Sodium ion accumulation within fish-only systems is dealt with in the same manner as the build-up of other deleterious ions (e.g. nitrate, phosphate); ionic concentrations are diluted by water exchange (Espinal & Matulic, 2019;Timmons et al., 2002).

One of the key advantages of recirculating aquaponic systems is that the addition of plants counteracts nutrient and other ionic accumulations, so that the reliance on water exchange to reduce ion accumulation is either lowered or negated completely (Lennard & Goddek, 2019;Rakocy & Hargreaves, 1993). Therefore, aquaponic researchers have advocated the use of buffer salts to maintain pH that are based upon positive ions that are essential to plant growth and health, and therefore will not accumulate because the plants are actively using them in metabolic processes (Lennard, 2017; Lennard & Goddek, 2019; Rakocy et al., 1992). Because fish feeds are generally lacking in potassium and calcium required for good plant growth, buffers based upon potassium and calcium salts are the most appropriate for aquaponic systems(Lennard, 2017; Lennard & Goddek, 2019; Rakocy & Hargreaves, 1993),as they provide the additional inputs of potassium and calcium that are essential for normal hydroponic plant metabolism (Maucieri et al., 2019,p. pp77; Morgan, 1999; Resh, 2013).

In previous experiments (Lennard, 2006), sodium bicarbonate was used as the buffer species since it was the most common buffer used in fish-only, RAS culture (Masser et al., 1999). Whilst good plant growth was achieved using sodium bicarbonate as the buffer in this previous experiment, the current experiment was devised to test the suitability of both potassium-based and calcium-based buffers as alternatives, and to determine whether either of these buffers (or a mixture of the two)conferred advantages.

The fish species used was the Australian native, freshwater fish Murray Cod,

Maccullochella peelii

, shown to be well adapted to aquaponic systems in previous experiments (Lennard & Leonard, 2004;Lennard and Leonard, 2006) and the hydroponic vegetable was lettuce,

Lactuca sativa

(Green Oak variety). The system treatments were compared for fish growth, plant growth, buffer usage rate, nutrient accumulations (nitrate and phosphate), dissolved oxygen, pH, conductivity and water replacement rate.

The current experiment was done to answer several questions relating to buffer use in the research-scale, recirculating aquaponic system. These questions included:

1. Are Murray Cod (fish) growth rates and feed conversion ratios (FCRs)using either a potassium-based buffer, a calcium-based buffer, or a mixture of the two, comparable to (better than, equal to or worse than) each other and to those of sodium-based buffered controls?

2. Are lettuce growth rates and yields using either a potassium-based buffer, a calcium-based buffer, or a mixture of the two, comparable to (better than, equal to or worse than) each other and to those of sodium-based buffered controls?

3. Is nutrient (nitrate and phosphate) removal from the research-scale aquaponic system water, using either a potassium-based buffer, a calcium-based buffer, or a mixture of the two, comparable to (better than, equal to or worse than) each other and to those of sodiumbased buffered controls?

4. Does the use of either a potassium-based buffer, a calcium-based buffer, or a mixture of the two, confer any advantage, with respect to other measured water quality parameters (i.e. dissolved oxygen,pH, conductivity and water replacement rate), when compared against each other and to those of sodium-based buffered controls?

2.Materials and methods

2.1.Fish origin and holding

Murray Cod were obtained from Australian Aquaculture Products Pty. Ltd., Melbourne, Victoria, Australia. Fish were held indoors at the Royal Melbourne Institute of Technology University Aquaculture Annex and ranged in size from 120 g to 220 g. All fish were kept in 1000L,cylindrical tanks receiving flow-through water at a flow rate of 3000 L day, until required for experimentation. Size distributions of the Murray Cod were similar among replicates and treatments. Water was of domestic origin (Melbourne tap water – catchment rainwater sourced;specific conductance =100 μS/cm; Total alkalinity as CaCO=10 mg/L), carbon-filtered and heated to approximately 22C.

2.2.Experimental recirculating aquaponic system

The experimental set-up was located, and all research was performed, within the Aquatic Culturing Laboratory Annex, Building 223,RMIT Bundoora Campus.

The experimental, recirculating aquaponic system array consisted of 12 individual, identical aquaponic units, allowing replication of experimental treatments. Each aquaponic unit consisted of a fish holding tank,an associated biofilter and a hydroponic component (Figs. 1 and 2).

The fish tank of each aquaponic unit was a round 100 L, opaque,white plastic tank (570 mm diameter x 460 mm deep). As well as fish,the tank contained an airlift pipe (to the biofilter), a submersible water pump (to the hydroponic component) and a 100 W thermostatically controlled electrical resistance aquarium heater. A plastic “core flute”(3 mm) lid covered the fish tank to lower evaporation and to stop fish from jumping from the tank.

Each fish tank had an associated, 20L biofilter (360 mm L x 330 mm W x 290 mm D), made from a plastic storage box. This biofilter sat above the fish tank and was of a wet/dry trickling design. Water entered the biofilter by way of a 20 mm airlift pipe (at an average of 250 L h), running from the base of the fish tank and into the top of the biofilter. A 6 mm plastic hose delivered air to the airlift pipe via an air stone. Water from the airlift entered the top of the biofilter via a “spray bar”, trickled across the biological filter medium (polystyrene “bean bag” beads @ approx. 300 mm: area/volume) and out through a series of 6 ×10 mm holes (drilled in the bottom, front area of the biofilter) and back into the fish tank. The biofilter had a plastic “core flute”(3 mm) lid to lower evaporation. This lid contained a breathing hole in one corner made from a short length of 65 mm PVC pipe to allow gas exchange within the biofilter.

Fig.1.Schematic representation of a single, recirculating aquaponic test unit.

Fig.2.Images of the research-scale, recirculating aquaponic system; top-left: view of the grow beds, lights and fan; top-right: view of the fish tanks and biofilters;bottom-left: view of a fish tank and biofilter of one unit; bottom-right: view of 12 individual aquaponic units with fish tanks below and hydroponic beds above.

Each tank and biofilter unit had an associated hydroponic plant growth component which contained standard, washed aquarium gravel(6 mm average particle size) to a depth of approximately 200 mm. This component was rectangular in shape (780 mm L x 670 mm W x 220 mm D) and was placed above the fish tank/biofilter unit on a separate shelving system. A submersible water pump (Rio 1700, 1200 L h@1.2 m head) in the fish tank continuously delivered water to the hydroponic component via a 19 mm pipe. Water from the hydroponic component was returned to the fish tank via a 20 mm drainpipe, situated at the opposite end of the hydroponic bed from the water inlet.

A continuous flow of water through the hydroponic gravel bed was used because previous experiments (Lennard, 2005) demonstrated that this was the most efficient method for the research-scale aquaponic system for the experimental duration applied.

Water for all experimental tanks was supplied from the aquaculture laboratory water supply system. Air for all biofilters and associated airlifts was supplied by a centralised, pressurised air supply that delivered air to the entire aquaculture lab.

Lighting (for plant growth in the hydroponic sub-systems) consisted of six x 400 W metal halide lamps. Lights were situated above the hydroponic beds at a height of 700 mm above the gravel surface, with one lighting unit located at the interface between two hydroponic subsystems (i.e. one lighting unit supplied the required light for two hydroponic beds). Lights (i.e. day-length) were controlled by a digitally timed electrical switch.

2.3.Experimental methodology

The experiment was designed to compare the effects of four different buffer supplied to the aquaponic unit(s). The hydroponic component of each aquaponic unit for all treatments and the control used a pumping rate of 540 L/h (9 L/min) through the gravel bed. One control and three separate treatments, each with three replicates, were tested:

1.

Sodium Control

; fish in tank, plants in the hydroponic bed, sodium bicarbonate (NaHCO) buffer - this was a standard culture practice control regime to compare nitrate and phosphate accumulation, fish growth and plant growth to those of treatments.2.

Potassium

; fish in tank, plants in the hydroponic bed, potassium bicarbonate (KHCO) as the buffer - this was a test treatment to compare nitrate and phosphate accumulation, fish growth and plant growth to the sodium bicarbonate control and the other two buffer treatments.3.

Calcium

; fish in tank, plants in the hydroponic bed, calcium hydroxide (Ca (OH)) as the buffer - this was a test regime to compare nitrate and phosphate accumulation, fish growth and plant growth to the sodium bicarbonate control and the other two pH buffer treatments.4.

Mixed

; fish in tank, plants in the hydroponic bed, a 50:50 ratio mixture of potassium bicarbonate and calcium hydroxide as the buffer – this was a test regime to compare nitrate and phosphate accumulation, fish growth and plant growth to the sodium bicarbonate control and the other two pH buffer treatments.

The lighting regime for plant growth was 10 h on: 14 h off, with lights coming on at approximately 08:30 a.m. (AEST) and going off at 18:30 p.m. (AEST).

All aquaponic units had operated with fish and plants present for several months prior to experimentation, to establish stable biofiltration treatment capacity (confirmed by daily ammonia determinations – data not presented here). At the initiation of the experiment, all aquaponic units were flushed and refilled with fresh water to 100 L and initial nutrient levels (nitrate and phosphate) were recorded. Fish weights were determined for each aquaponic unit and adjusted to a treatment biomass of approximately 1000 g (individual tank fish biomass was recorded).

Fish were fed at a percentage of the total initial fish biomass per day(for six of seven days per week) with a 9 mm, sinking pellet (43% protein) (Skretting Classic SS, Skretting Pty. Ltd. Hobart, Tasmania,Australia) (Table 1). Feeding rates were set at 1.0% of fish biomass (per day) for the first 5 days, then adjusted to 1.5% for the remaining 15 days of the experiment. Weights of fish feed fed to each aquaponic unit daily were recorded.

Twenty lettuce seedlings (

Lactuca sativa

Green Oak variety) were planted using an evenly distributed planting scheme into each of the replicate hydroponic beds. The individual initial weight (biomass) of these 20 seedlings was recorded (weight with attached soil plug).Because seedlings had attached plugs of soil, initial leaf weight was estimated by recording the weights of 15 additional seedlings from the same batch as those used for the experiment with and without attached soil plugs. These weights were used to establish a mean ratio of leaf-only to leaf +plug weight. This ratio was then used to estimate the initial leaf-only weight of the tested seedlings.

Six of seven days a week (at the same time every day – 9:30 a.m.,AEST; Monday to Saturday, inclusive), the amount of fish feed fed (g),air temperature (C), water replaced per tank (L) (to adjust for evapotranspiration – see below), the amount of specific buffer added (to adjust pH to between 6.80 and 7.00; added directly to the fish rearing compartment) (g), pH, temperature of the tank water (C), conductivity(μs cm) and dissolved oxygen (mg L) were recorded. Twice a week,tanks were sampled for ammonia (mg L) and nitrite (mg L), whilst once a week, tanks were sampled for nitrate (mg L) and phosphate(mg L). All samples for analysis were taken from the fish-rearing compartment of the aquaponic unit(s).

The amount of feed fed, and the buffer added per tank (to maintain pH) were measured using a top loading balance (A&D HL-200). The amount of water replaced per tank was determined by re-filling the tank to a pre-measured 100 L mark with a measuring cylinder. Temperature(tank water), pH, conductivity and dissolved oxygen were determined using a TPS 90-FL multi-parameter meter and associated probes.Ammonia, nitrite and nitrate were determined using a Hanna, C203 Multiparameter ion-specific meter (H025463) and Hanna Ammonia LR reagent (HI 93700-01), Hanna Nitrite LR reagent (HI 93707-01) and Hanna Nitrate HR reagent (HI 93728-01), respectively. Phosphate was determined using a Merck Spectroquant, colour reagent test (code:1.1482.0001)

,

read against a standard curve using a spectrophotometer at 400 nm (Varian Cary 50 Bio UV–Vis).

Table 1 Mean wet-weight gain, specific growth rate (SGR), feed conversion ratio (FCR)and feed consumption for Murray Cod for sodium control, potassium, calcium and mixed treatments over the 21-day trial.

The entire experiment ran for 21 days from water flushing and planting to harvest. At the end of the experiment, fish biomass was determined by wet weight (A&D. HL-200) and plant (leaf only) biomass was determined by wet weight (A&D HL-200). Gains in both fish biomass and plant biomass per unit (or hydroponic component) were determined by the difference between initial and final wet biomasses recorded. Fish biomass was determined on a per-tank basis, whilst plant biomass (leaf only) was determined on an individual, per plant basis.

Comparisons between treatments and controls at the end of the experimental period for fish biomass, fish feed conversion Ratio(FCR =feed fed {g}/fish weight gain {g}), fish specific growth rate(SGR =[(ln final wt. – ln initial wt.)/(time (days))]x100), nitrate and phosphate concentrations were analysed using Mann-Whitney, two independent population, non-parametric analysis. Comparisons between all other parameters were analysed using ANOVA and Least Significant Difference (LSD) post-hoc analysis where appropriate. For lettuce growth and yield parameters, the replicates within a treatment (

n

=

3

)were initially analysed for homogeneity (using individual plant weight gains; (

n

=

20

) and when confirmed, replicate individual plant weight gains were pooled (

n

=

60

) for overall treatment comparisons. All statistical analyses were performed using SPSS (Version 10.0) software.

3.Results

3.1.Fish

Survival of Murray Cod in all replicates (all treatments) was 100%for the 21-day trial. Fish biomass gain (wet weight gain/treatment replicate) averaged 263.3 g, 280.0 g, 316.7 g and 293.3 g for control,potassium, calcium and mixed treatments respectively, whilst specific growth rate (SGR) averaged 1.09%, 1.15%, 1.32% and 1.20% for the same treatments, respectively (Table 1). Feed conversion ratios (FCR)were 0.93, 0.88, 0.78 and 0.84 for control, potassium, calcium and mixed treatments respectively (Table 1). No significant differences(Mann-Whitney:

P >

0.05,

n

=3) were detected between any treatments in terms of any fish growth parameters.

3.2.Lettuce

Lettuce production values for control, potassium, calcium and mixed treatments are presented in Table 2. There was a significant difference(ANOVA:

F

=28.143,

P

=0.000) between the control and all other test treatments in terms of both biomass gain and yield, with both the potassium and mixed treatments exhibiting higher values for these parameters (Table 2), and the calcium treatment exhibiting lower values for these parameters (Table 2). A significant difference (ANOVA:

F

=26.667,

P

=0.000) was detected between the potassium and calciumtest treatments (potassium higher) and between the calcium and mixed test treatments (ANOVA:

F

=33.462,

P

=0.000) (mixed higher),whilst no significant difference (ANOVA:

F

=6.795,

P

=0.094) was detected between the potassium and mixed treatments. Yields averaged 4.15 kg/m, 4.75 kg/m, 3.72 kg/mand 5.00 kg/mfor the control,potassium, calcium and mixed treatments, respectively (Table 2). Statistical differences in yields followed the same associations as for lettuce production values. No qualitative indications of sub-optimal plant growth (tip burn, bolting or wilting) were noted in any control or experimental treatments.

Table 2 Mean biomass gain and mean yield (g/plant & kg/m2) for lettuce plants for sodium control, potassium, calcium and mixed treatments over the 21-day trial.

3.3.Metabolites, nitrates and phosphates

Ammonia and nitrite concentrations were recorded daily to ascertain biological filter conversion efficiency. All replicates in all treatments showed undetectable ammonia concentrations (

<

0.25 mg/L) over the duration of the experiment. Nitrite concentrations remained at zero(

<

0.05 mg/L) for all replicates in all treatments for the duration of the experiment.Final net phosphate accumulations averaged 3.92 mg/L, 3.50 mg/L,2.60 mg/L and 2.81 mg/L for control, potassium, calcium and mixed treatments, respectively (Table 3). A significant difference (Mann-Whitney

P <

0.05,

n

=3) was detected between the control and both the calcium treatment (higher phosphate removal) and the mixed treatment(higher phosphate removal), whilst no significant difference was detected between the control and the potassium treatment. No significant differences (Mann-Whitney:

P >

0.05,

n

=3) were detected between any of the three test treatments for final phosphate accumulation(Table 3).Final net nitrate accumulations averaged 10.40 mg/L, 7.80 mg/L,13.77 mg/L and 8.77 mg/L for control, potassium, calcium and mixed treatments respectively (Table 3). No significant differences (Mann-Whitney:

P >

0.05,

n

=3) were detected between any of the control or test treatments with respect to final nitrate accumulation (Table 3).

3.4.Physical/chemical parameters

Air temperature was measured daily and remained steady at 24C(data not shown). Fish tank water temperatures averaged 22.1C across all control and treatment replicates (Data not shown).

Mean daily dissolved oxygen (D.O.) concentrations did not differ significantly between any treatments (ANOVA:

F

=0.047,

P

=0.987).D.O. concentrations fell over the length of the experiment in all replicates but remained above 6.73 mg/L for all treatments (Fig.3).Buffer additions (Fig.4) and pH (Fig.5) were integrally linked in all treatments. Control buffer additions (sodium bicarbonate) were statistically similar (ANOVA:

F

=0.265,

P

=0.122) to potassium buffer additions and significantly higher than both calcium buffer additions(ANOVA:

F

=0.451,

P

=0.009) and mixed buffer additions (ANOVA:

F

=0.382,

P

=0.026).Potassium buffer additions, when compared with the two other buffer regimes, were significantly higher (ANOVA:

F

=7.767,

P

=0.000). There was no significant difference (ANOVA:

F

=0.069,

P

=0.688) between calcium treatment buffer additions and mixed treatment buffer additions (Fig.4).

Table 3 Mean net (final-initial) phosphate and nitrate accumulations, for sodium control, potassium, calcium and mixed treatments over the 21-day trial.

Fig.3.Mean daily dissolved oxygen concentrations for sodium control, potassium, calcium and mixed treatments.Error bars represent standard errors.

Fig.4.Mean daily buffer additions (per treatment replicate) for sodium control, potassium, calcium and mixed treatments.a & b, x & y: Treatments showing the same letter are not significantly different (P >0.05, n =51).Error bars represent standard errors.

Mean daily pH readings are presented in Fig.5 pH levels within potassium buffer treatments were significantly lower (ANOVA:

F

=7.662, p =0.000) than within all other control and test treatments(Fig.4), even though potassium replicates had the highest amount of buffer additions.Calcium treatment pH levels were significantly higher (ANOVA:

F

=0.102,

P

=0.021) than the control treatment and statistically similar(ANOVA:

F

=0.029,

P

=0.514) to the mixed treatment. Mixed treatment pH levels were statistically similar (ANOVA:

F

=0.096,

P

=0.688) to both sodium control treatment pH levels and calcium treatment pH levels (Fig.5). A distinct fall in pH was seen in the potassium treatment on Day 9 of the trial. This was associated with a lack of buffer being added to those treatment replicates on Day 8 of the trial(raw data not presented here).Mean daily conductivity readings are presented in Fig.6. Conductivity in all control and test treatments exhibited a very similar slope(Fig.6), and no significant differences (ANOVA:

F

=1.644,

P

=0.180)were detected between any control or test treatments.Water was replaced daily to compensate for evapotranspiration(Fig.7). Overall, daily water replacement averaged 1.86 L/replicate,1.59 L/replicate, 1.83 L/replicate and 1.60 L/replicate for the control,potassium, calcium and mixed treatments respectively. Significant differences (ANOVA:

F

=0.096,

P

=0.688) were detected between control and potassium treatments (control required more water) and control and mixed treatments (control required more water). No significant differences (ANOVA:

F

=2.511,

P

=0.060) were detected between any other combinations of control or test treatments (Fig.7).

Fig.5.Mean daily pH readings for sodium control, potassium, calcium and mixed treatments.Error bars represent standard errors.

Fig.6.Mean daily conductivity readings for sodium control, potassium, calcium and mixed treatments.Error bars represent standard errors.

4.Discussion

Previous researchers have noted that the nutrient make-up of the water in standard RAS fish systems is sub-optimal for plant growth in recirculating aquaponic systems (Karimanzira et al., 2016; Lennard,2017; Rakocy & Hargreaves, 1993; Suhl et al., 2016). This research-scale experiment was designed to determine whether the ionic make-up of the buffer could improve the performance of a recirculating aquaponic system, allowing a controlled pH without negative impacts on fish growth and feed conversion, lettuce growth, yield and health and other test parameters.

Fish mortality in all treatments was zero. Ingram (2002) obtained less than 5% mortality for Murray Cod exceeding 50 g in weight in culture trials in tanks. Therefore, the mortality in the present study is what should be expected for Murray Cod of this size (250–400 g) in standard recirculating aquaculture conditions. In terms of feed conversion efficiency, Ingram (2002), using feed containing 43% protein, obtained a mean FCR for Murray Cod over 150 g in weight of 1.2 therefore,the FCR values obtained in the present study (Table 1) are comparable with research results using industry-standard, recirculating culture methods. Whether the buffer addition regime affects fish growth is also an important question. In the present research-scale study, no significant differences in any fish growth parameter (biomass gain, SGR or FCR)were detected between any treatments or controls (Table 1), therefore suggesting that none of the pH buffer treatments tested in this study had a deleterious effect on fish growth or survival.

Fig.7.Mean daily water replacement (per treatment replicate) for sodium control, potassium, calcium and mixed treatments. a & b, x & y: Treatments showing the same letter are not significantly different (P >0.05, n =54)Error bars represent standard errors.

Plant growth, yield and health are the parameters most likely to be significantly affected by the constitution of the buffer added to the recirculating aquaponic system (Rakocy & Hargreaves, 1993). Whilst the products of fish metabolism supply both nitrogen and phosphorous-based nutrients to the system (as well as unknown quantities of essential micronutrients), the buffer that is added to control the pH drop caused by fish metabolism and biofilter nitrification is an additional source of the macronutrients needed for optimal plant growth(Eck et al., 2019; Lennard, 2017; Rakocy & Hargreaves, 1993). Lettuce production as wet, leaf weight gain or yield (weight gain per unit area)within the four treatments in the present study followed the relationship mixed =potassium

>

sodium control

>

calcium (Table 2), with a significant difference (Mann-Whitney:

P <

0.05,

n

=60) detected between all treatments, except between the potassium and mixed treatments which exhibited statistically similar results (Mann-Whitney:

P >

0.05,

n

=60). Therefore, the ionic make-up of the buffer added to the system did affect the efficiency of the aquaponic system in terms of plant production. Yields were equal to or better than those in the studies of Burgoon and Baum (1984), Seawright et al. (1998), Lennard and Leonard (2004), Lennard and Leonard (2006), Geisenhoff et al. (2016),Johnson et al. (2017), Jordan et al. (2018) and Maucieri et al. (2018).

It is well established in both terrestrial plant production (Hopkins &Huner, 2004) and standard hydroponic plant production (Jensen &Collins, 1985; Morgan, 1999) that beyond nitrogen and phosphorous,potassium and calcium are the next most-important macronutrients for plant growth. Potassium plays an essential role in several functions crucial to plants, including the formation of sugars and proteins, carbohydrate synthesis, cell division, water balance and structural rigidity(Morgan, 1999). Calcium also plays an important role, being a major component of cell walls contributing to the support of plant tissues, as well as contributing to enzyme activation and regulation of water movement into and out of cells (Morgan, 1999). All standard hydroponic nutrient solutions contain macro amounts of both potassium and calcium (Jensen & Collins, 1985), and it therefore is expected that aquaponic systems need to contain appropriate concentrations or proportions of these two macronutrients for efficient plant production (Lennard,2017; Maucieri et al., 2019, p. pp77; Rakocy & Hargreaves, 1993; Resh,2013).

In the present study, whilst mixed (potassium and calcium) and potassium treatments produced statistically similar lettuce growth and yields, these were statistically significantly higher than those of both the control and calcium treatment systems (Table 2), with the calcium treatment producing the lowest plant yield. The results of this study therefore, do not fully agree with the findings of other aquaponic studies(Lennard, 2017; Rakocy et al., 1992) that contend that potassium and calcium-based buffers are the most appropriate for recirculating aquaponic systems in terms of plant growth and production, since this study indicates an advantage with a potassium-containing buffer (mixed or potassium treatments).

Net nutrient accumulation within the recirculating aquaponic system is an indicator of balance between fish waste production and plant nutrient use. In the present study, all treatment systems contained statistically similar amounts of nitrate at the end of the experiment(Table 3); therefore, no buffer addition was better than any other in terms of nitrate removal by the plants. However, results showed that both potassium and mixed treatments achieved higher plant growth and yield (Table 3). Therefore, it may be inferred that, whilst the lettuce plants within the potassium and mixed treatments did not remove any more nitrate than did the plants within the calcium and control treatments, they may have used that nitrate more efficiently to achieve more plant tissue growth. Resh (2013), Morgan (1999) and Jensen and Collins(1985) noted that potassium is essential to carbohydrate synthesis in plants. If potassium levels are too low, carbohydrate and sugar synthesis are blocked, leading to lowered overall plant growth. It seems reasonable therefore, that treatments containing no additional potassium supplementation (control and calcium treatments) may have exhibited lower plant growth and yield than those treatments containing potassium (potassium and mixed treatments), due to this potential blocking of sugar synthesis.

Net accumulation of nitrate within all treatments (Table 3) and ranged from 7.80 ±2.20 mg/L (potassium treatment) to 13.77 ±2.23 mg/L (calcium treatment). This accumulation is comparable to previous research results using the same aquaponic system with a similar constant flow regime through the gravel plant-growing bed of 11.80 ±1.78 mg/L over the same time period (Lennard & Leonard,2004). Delaide et al. (2017) achieved nitrate accumulations of 58 mg/L in their small-scale, deep-water culture aquaponic system growing lettuce and basil. Hasan et al. (2017) achieved average nitrate accumulations of approximately 40 mg/L after three weeks growing Sangkuriang Catfish (

Clarias gariepinus

) and Nile Tilapia (

Oreochromis niloticus

) with Water Spinach (

Ipomoea aquatica

) and Lettuce (

Lactuca sativa

). Dediu et al. (2012) observed nitrate accumulations of 34.52 ±6.26 mg/L and 32.25 ±7.06 mg/L in an aquaponic system applying high and low hydraulic retention times growing Bester Sturgeon (

Huso huso x Acipenser ruthensis

) and Lettuce (

Lactuca sativa

). Therefore, nitrate accumulations in the current study compared well with other studies.In terms of final treatment phosphate concentrations, control and potassium treatments were statistically similar (Mann-Whitney:

P >

0.05,

n

=3), whilst calcium and mixed treatments removed more phosphate than did controls (Mann-Whitney:

P <

0.05,

n

=3). However,potassium treatments were statistically similar (Mann-Whitney:

P >

0.05,

n

=3) to both calcium and mixed treatments, and calcium and mixed treatments were statistically similar to each other (Mann-Whitney:

P >

0.05,

n

=3) (Table 3). From these results it can be inferred that phosphate removal is a complex process. Adler, Harper, Takeda, et al.(2000) argued that when other macronutrients are in limiting supplies in hydroponic systems, plants will remove phosphate only to certain levels.The only way to get plants to remove further phosphate from the system is to supply those nutrients that are known to be limited (Adler, Harper,Wade, et al., 2000). From the results in the present study, it may be interpreted that some other phosphate removal mechanism may have been involved. Because the control treatment removed statistically no more phosphate than did the potassium treatment, the addition of potassium to the system had little effect on system phosphate removal by plants. However, it is evident from the results that the addition of calcium to the buffering system had a positive statistical effect on system phosphate removal; systems containing calcium removed more phosphate than did those not contain calcium (Table 3). This does not necessarily mean that the addition of calcium to the system, via a calcium-based buffer, allowed plants to remove more phosphate. It is known that when excess calcium is added to aquatic systems, it can form a complex with the available phosphate, which then has the ability to precipitate out of the system water, thus lowering available system phosphate levels (Lennard, 2017; Resh, 2013; Seawright et al., 1998).This may be the reason why, in the present study, more phosphate was removed from those treatments with calcium supplementation, even though no precipitate was noticed.Net accumulation of phosphate within all treatments (Table 3) and ranged from 2.60 ±0.11 mg/L (calcium treatment) to 3.92 ±0.33 mg/L(control sodium treatment). This is comparable to previous research results using the same aquaponic research system with a similar constant flow regime through the gravel plant-growing bed of 3.87 ±0.71 mg/L over the same time period (Lennard & Leonard, 2004). Makhdom et al.(2017) observed phosphate accumulations in an aquaponic system growing Pearl Gourami (

Trichopodus leerii

) and Cherry Tomato (

Solanum lycopersicum var. cerasiforme

) at the highest planting density of approximately 15 mg/L after 30 days. Liang and Chien (2013) achieved phosphate accumulation of 38.1 mg/L (lowest concentration achieved) when growing Red Tilapia (

Oreochromis sp

.) and Water Spinach (

Ipomoea aquatica

) in an aquaponic system testing feeding frequencies and photoperiods. Therefore, phosphate accumulations in the current study compared well with those of other studies.

Dissolved oxygen concentrations (Fig.3) show that the buffer added(sodium-control, potassium, calcium or potassium and calcium-mixed)had no effect on the ability of the system water to maintain dissolved oxygen concentrations. D.O. was maintained at levels above the minimum requirement for lettuce (2.1 mg/L – Goto et al., 1996; Resh, 2013),warm water, native Australian fishes (5.0 mg/L – Masser et al., 1999;Ingram, 2002; Lennard, 2017) and nitrifying bacteria (2.0 mg/L –Alleman & Preston, 2002).

Previous experiments (Lennard, 2005) demonstrated that the inclusion of plants in the research-scale, recirculating aquaponic system led to an outcome whereby buffer additions to control pH may be lowered(when compared to fish-only controls) due to the ionic exchange mechanisms that are prevalent when plants are actively assimilating nitrate and phosphate ions (Imsande & Touraine, 1994; Rakocy &Hargreaves, 1993; Salsac et al., 1987; Lennard, 2017; Maucieri et al.,2019, p. pp77). When nitrate and phosphate ions are assimilated by plants, negative ions (OH, HCO) are released in order to maintain homeostatic, cellular pH levels within the roots (Imsande & Touraine,1994; Lennard, 2017; Maucieri et al., 2019, p. pp77). It is the negative ion portion of the buffer salt added that directly impacts the buffering capacity of the recirculating water. However, different negative ions have differing capacities to counteract acidification and to buffer pH to desired levels. Because the buffers used in the present study possessed different negative ion constituents (and the addition of these buffers was purely based upon maintenance of pH), it is difficult to directly compare the amounts of the relative buffers used. However, results (Fig.4) suggest that the amount of buffer required was variable and dependent on the negative ion content of the buffer. There was no significant difference in the amount of buffer required between the two buffers that used bicarbonate as the pH-buffering component (control and potassium treatments) (Fig.4). However, significantly less buffer was required in both calcium and mixed treatments (Fig.4). This is because these treatments used the hydroxyl ion as the ion to counteract acidification,and less hydroxyl ion is required to maintain a similar pH in a similar system than bicarbonate ion (Lennard, 2017). This explains why pH levels were more difficult to maintain with the bicarbonate-containing buffers, and why these treatments required significantly higher additions than those treatments using hydroxyl-containing buffers (Fig.5).Therefore, whilst the positive ion component of the buffer used had effects upon plant growth, the relative amounts of the negative ion component of the buffer was the determining factor in pH buffering and maintenance (Lennard, 2017; Timmons et al., 2002). In addition, this opens the possibility of customising the buffer species and mixtures added, so that the amounts or proportions of the positive (plant-assimilated) ion portion of the buffer delivered to the system may be adjusted or controlled by choosing buffer species or mixtures based on the buffering capacity of the negative ion.

Conductivity continually increases in fish-only systems because of the build-up of ions which are produced as the by-products of fish metabolism (Rakocy & Hargreaves, 1993; Timmons et al., 2002). The inclusion of plants into recirculating aquaponic systems leads to active uptake of waste nutrients and ions by the plants, which counteracts the ionic build-up seen in fish-only systems (Rakocy & Hargreaves, 1993;Seawright et al., 1998; Lennard, 2017; Palm et al., 2019). The efficiency of plants to take up nutrients and ions in hydroponic systems, and thus maintain zero conductivity accumulation within system waters, is dependent upon whether the correct mix and concentrations of those nutrients are provided (Morgan, 1999; Hopkins & Huner, 2004; Resh,2013). The conductivity curves (Fig.6) from the present study indicate that, in the final third of the experiment, plants within the mixed and potassium treatments maintained slightly lower water conductivities than observed in control or calcium treatments, although overall conductivities did not differ significantly between any treatments across the 21-day course of the research-scale experiment. This apparent lowering of conductivity levels in the last third of the experiment for the mixed and potassium treatments is probably since plants within these treatments had access to increased levels of potassium during their fastest-growing phase. Potassium is an essential macronutrient to plants(Hopkins & Huner, 2004; Resh, 2013) and is known to play a key role in a plant’s ability to synthesise proteins and carbohydrates (Morgan,1999; Resh, 2013), and thus grow. Adler, Harper, Takeda, et al. (2000)also argued that when other nutrients limit plant growth, nutrient removal can be increased by adding those nutrients that are most limiting, and therefore, other essential nutrients (usually limited in integrated systems) must be added to the aquaponic system. These nutrients typically include iron, manganese and potassium (Adler, Harper,Takeda, et al., 2000). The lowered conductivities in the two treatments containing higher additions of potassium (mixed and potassium treatments) may suggest that the plants in these treatments had a slightly increased ability to achieve elevated carbohydrate and protein synthesis, thus allowing them to remove and assimilate slightly greater amounts of ions from the surrounding water medium and therefore,lowering the conductivity of the system water. This hypothesis also may be supported by the observation that lettuce plants within these two potassium-containing treatments exhibited the highest growth rates and yields (Table 2) and exhibited the greatest removal rates of nitrate from recirculating waters (Table 3).

In terms of water use, results suggest that those treatments containing potassium were also more efficient (Fig.7), with significantly less water used by the potassium and mixed test treatments, when compared to the control treatment. Again, this result is probably correlated with the plant’s requirement for potassium to achieve carbohydrate and protein synthesis and subsequent growth. The plant’s requirement to transpire more water from those treatments not containing additional potassium may be explained by the fact that the plants may have been transpiring greater amounts of water to try and gain greater access to limiting ions, such as potassium.

In conclusion, results from the present study suggest that, when using buffers for maintenance and control of pH in recirculating aquaponic systems, it is advantageous to use buffers with positive ion constituents that are essential to optimal plant growth (i.e., potassium and calcium).Parameters such as plant growth and yield, nitrate removal, conductivity and water use demonstrate that either a potassium-based buffer or a mixture of potassium and calcium-based buffers are the most appropriate buffers to use to achieve optimal plant growth efficiencies in the research-scale aquaponic system, whilst parameters such as fish growth and FCR, dissolved oxygen and pH maintenance are unaffected by the positive ion make-up of the buffer. In addition, it is apparent from buffer use and pH maintenance parameters that the negative ion constituent of the buffer may affect the amount of buffer required. It is therefore recommended that this work be replicated at a commercial scale and through an entire fish production cycle and that future aquaponic researchers and industry individuals use buffers containing potassium as the positive ion constituent and consider periodically using calciumbased buffers to maintain pH in recirculating aquaponic systems.

Ethics approval

Peer reviewed ethics approval granted by University where research was performed.

Consent to participate

All consents to participate have been granted

Consent for publication

All authors consent for publication.

CRediT authorship contribution statement

Wilson A. Lennard:

Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Project administration,Resources, Writing - original draft, Writing - review & editing.

Declaration of competing interest

The authors declare no conflict of interest.

Acknowledgments

This research was partially funded by the Australian Federal Governments, Rural Industry Research and Development Corporation(RIRDC).