concentration during the experiments for each membrane. The TMP was chosen to produce the maximal flux before the %AR begins to diminish as a result of the pressure applied or the force produced through the membrane (Cancino Madariaga et al., 2011). For the NF the TMP was 16 bar and for the RO it was 24.5 bar. The highest ammonia concentration in the feed solution studied was 11.45 mg/L, levels that are greater than the values permitted in RAS; high ammonia concentrations in the feed solution were tested to observe the AR efficiency of the membrane. With the idea to project this work later, to operate in a separate equipment of the fish culture itself, recycling the water in this special membrane system to concentrate the ammonia. Then, the membrane system can operate separately from the RAS itself, as a unit of operation for ammonia treatment. The water can betreated in a different tank to that of the culture, without damage to the biomass. The water can be recirculated through the mem- brane within this unit until the concentration is 11.45 mg/L. Higher concentrations could also be studied in future research.
Fig. 2 shows the different times taken to achieve maximum ammonia concentration depending on the membrane used. This is reflected in the permeate flux, where higher flux gives shorter times. NF270 was the fastest membrane with the highest flux (see Table 3), with a maximum ammonia concentration of 11.45 mg/L. This is in accordance with the membrane resistance (Rm) value, were higher Rm means lower flux, and longer times. NF270 was the membrane with the lowest Rm, 2.58×1013 m−1, while NF90 presented the highest Rm, at 6.51 × 1013 m−1 , with the lowest flux (Cancino Madariaga et al., 2011). NF270 was also the membrane with lowest AR.
Fig. 3 shows the AR for the membranes during the concen- tration process. No significant difference in AR was observed for membranes NF90 and DSS-HR98PP during ammonia concentration. However, NF200 changed the AR by a significant degree (p = 0.0234) with a decrease in AR after 400 min. This effect could be explained by the increased NH4+ concentration leading to higher ionic force, producing a decrease in the charge on the membrane. This idea is in line with the Donnan exclusion theory. This behaviour has also been observed by other authors (Peeters et al., 1998) and they agree that it is a characteristic of negatively charged membranes. Simi- lar behaviour was observed in NF270 with a significant decrease in AR (p=0.0090). In the experiences carried out in the previous works, with a constant NH4+ concentration (Cancino Madariaga et al., 2011), we reported a similar discussion because the results obtained. The interaction with ions, counter ions, and the mem- brane surface are explained with a model (Bruni and Bandini, 2009; Cancino Madariaga et al., 2011). One of the mechanisms that could be produced, is the adsorption of the NH4+ ions on the active layer of the membrane forms a positively charged superficial layer, those charged functional groups attract ions of the opposite charge, which managed the repulsion of other ammonium ions.
Fig. 4 shows the permeate flux during the concentration process without significant differences throughout the process (p ≥ 0.1648). With all membranes, the layer of NH4+ ions over the membrane surface does not represent a significant additional resistance to Rm. This is seen as the permeate flux shows no change throughout the process, and is also evident in the concentration process. This sta- bility in the flux, as the permeate passes through the membrane, enables the calculation of some figures for the application of RAS
systems.
Table 4 shows the AR and Flux average at maximum TMP, for the
RO and NF membranes under steady-state operation for 180 min, were the concentration of ammonia in the feed is constant at 5.44mg/L ammonia at 15±1◦CTable 5 shows a projection for RAS with the use of an NF and RO membrane. It is assumed that the flow to be treated with the membrane system is 2000 L/h with an ammonia concentration of 5.44 mg/L, with the AR and Flux values presented in Table 4.
All the AR for NF membranes (shown in Table 3) were statisti- cally different from that of the RO membrane. This result indicates that the RO membrane requires 2.5 times the area of NF270 mem- brane to obtain the same permeate flow, though using higher pressure, which means more energy, thus increasing operating costs. In terms of application in RAS systems, as seen in Table 5, it is possible to measure the specific mass flow of ammonia removal for each membrane, evaluated as the mass of ammonia removed per day and per unit of membrane area. NF270 presents the highest spe- cific mass of removed ammonia; more than double that of the RO membrane. If we compare the energy used for separation in accor- dance with Table 4, labelled specific ammonia removal per energy, NF270 is undoubtedly the best membrane. Similar behaviour is seen with NF200 and the RO membrane, and NF90 is the worst in terms of energy consumption. In addition to this energy con- sumption, the refrigeration system must also be considered for the fluid to be maintained at the correct temperature, especially when the pump has been running for some time. In this case, energy con- sumption increases depending on the pressure used, which means that for the RO membrane, though its results are close to NF200, consumption is considerably higher because of the energy used by the cooling and pump system.
The conclusion of this analysis is therefore that the best mem- brane is the NF270.
Application of membrane processes to recover and reduce ammonia in water is not only relevant for RAS, but also in the trans- port of live fish, such as for use in Wellboats or Welltrucks. In these transport systems recovery of ammonia is performed by biofilters. However, as was stated in the introduction, biofilters are difficult to control due to the use of bacteria and a continuous flow of ammo- nia is needed to keep these microorganisms alive; however, this is not possible in fish transportation because the quantity of fish changes constantly, journey time is variable, the time for the charge and discharge operation is also variable. All of these factors mean that the bacteria have no feed or diminished feed. Nevertheless, as was demonstrated in this research, membrane processes can be operated in continuous or batch systems, adjusting the number of membranes according to necessity.
In RAS systems, the potential of NF membranes is relevant to three cases:
Case (1): As an emergency system in closed-flow fish-farming. In this case the NF membranes could be used in parallel to the traditional biofilters. Thus, when the efficiency of the biofilters diminishes the nitrification capacity, the NF system will be acti- vated automatically.
Case (2): As an emergency system in incubation chambers. In incu- bation chambers, hatching of the fish eggs produces a peak in ammonia concentration, and fish farms traditionally increase the flow of water to depress this peak, sometimes incorporating con- taminants into the system, thus losing innocuousness. In this case, the system could be activated to restore water quality in the closed system without the risk of infection.
Case (3): As an alternative to biofilters in RAS. This case depends of the flow and the cost of investment, on the water quality and the price of the fish produced. In comparison with biofilters, mem- brane processes remove ammonia, and biofilters transform it into nitrites and nitrates, implying the additional cost of denitrification.
Additional research is needed to evaluate economics and oper- ational characteristics this process for aquaculture applications.