Orechromis niloticus (Nile tilapia) is the most cultivated fish species in Brazil, representing 47.1% (239.09 × 10 ³ metric tons) of total Brazilian fish production. This is linked to its resistance and adaptability to different growing environments, and the species' rapid growth rate ( Costa & Fróes, 2012 ; NG & Romano, 2013 ).

Despite these interesting zootechnical attributes found in cultivation systems using net cages and ponds with water exchange, more study of the zootechnical performance of Nile tilapia raised in intensive systems with minimal or no water exchange is necessary, and of the use of resources in systems with low flow and high levels of dissolved salts not suitable for human consumption, mainly due to the scarcity of freshwater in various parts of Brazil and the world.

Nile tilapia culture in biofloc systems is a very promising alternative, since it uses minimal or zero water exchange, high stocking densities and can reach high yields, due to the C:N ratio in the water, strong aeration and alkalinity control using microorganisms to remove and recycle nutrients ( De Schryver et al. , 2008 ; Crab et al. , 2012 ; Emerenciano et al. , 2013 ; Avnimelech, 2012 ).

Biofloc systems have been used with great success for shrimp ( Brito et al. , 2014 , 2016 ; Marinho et al. , 2017 ; Ray & Lotz, 2017 ) and tilapia cultivated in freshwater ( Ekasari et al. , 2015 ; Long et al. , 2015 ; Pérez-Fuentes et al. , 2016 ; Alves et al. , 2017 ; Miranda-Baeza et al. , 2017 ; Zapata-Lovera et al. , 2017 ). Tilapia can adapt to a biofloc system because of their resistance to high solids levels and filter-feeding ability, thus allowing the absorption of the suspended flocs ( Avnimelech, 2011 ), although more information on tilapia performance in brackish water is needed.

Tilapia are excellent for cultivation not only in freshwater, but also in brackish water, due to their tolerance for salinity between 0 and 12 g/L ( Likongwe et al. , 1996 ; El-Sayed, 2006 ). Other studies have shown salinity tolerance of 22.5 g/L ( Kamal & Mair, 2005 ) and up to 35 g/L ( Mapenzi & Mmochi, 2016 ). In addition, freshwater fish cultivated in brackish water may save energy, which could be made available for growth ( Takata & Luz, 2015 ), and increased salinity may also reduce the toxicity of nitrogen compounds (ammonia-NH ₃ and nitrite-NO ₂ ) ( Colt, 2006 ), which increases at higher stocking densities.

Nevertheless, higher stocking densities of tilapia cultivated in traditional freshwater systems have been found to cause the following negative effects: decreased growth ( Abou et al. , 2007 ), decreased survival ( Ferdous et al. , 2014 ), increased FCR (feed conversion ratio) ( Ridha, 2006 ), decreased crude protein content of the fish carcass ( Osofero et al. , 2009 ); decreased cost-benefit ratio ( Moniruzzaman et al. , 2015 ), decreased levels of hematological parameters: red blood cells, white blood cells, hemoglobin, hematocrit and platelets ( Kpundeh et al. , 2013 ), and decreased plasma cortisol concentrations, which may indicate chronic stress ( Barcellos et al. , 1999 ), despite the positive effect on increased yield ( Suresh & Lin, 1992 ) .

The effects of stocking densities (25, 50, 100 fish/m ³ ) of Oreochromis sp. in culture systems with and without bioflocs have been reported by Widanarni et al. (2012) , with higher zootechnical performance obtained in the biofloc system. In this sense, this study evaluated the effects of a low-salinity biofloc system with different stocking densities on water quality and zootechnical performance of Nile tilapia fingerlings.

The water quality variables of temperature (28.8 to 29.5°C), dissolved oxygen (4.3 to 6.5 mg/L), pH (6.2 to 8.8) and salinity (10.4 to 11.5 g/L), were not significantly affected (a > 0.05) by stocking density (Table 1).

TSS increased during culture time with increasing stocking densities, reaching average values close to 600 mg/L (573.49 mg/L in BFT-1000; 677.19 mg/L in BFT-1250) (Table 1). SS also increased during culture time and were significantly affected (a <0.05) by stocking density (Table 1).

Significant differences in alkalinity were found in BFT-1250 and BFT-1000 as compared to BFT-750 and BFT-500 (Table 1). A reduction in alkalinity occurred in the second week, although the addition of inorganic carbon (hydrated lime-Ca(OH) ₂ ) twice a week at 20% (by weight) of the total weekly feed was sufficient to maintain alkalinity levels above 100 mg/L.

The principal dissolved inorganic nitrogen com­­po­unds were NO ₃ , which ranged from 0.85 to 0.90 mg/L, followed by NO ₂ with concentrations ranging from 0.48 to 0.49 mg/L, and TAN ranging from 0.30 to 0.44 mg/L. No significant differences (a <0.05) were observed in the nitrogen compounds with stocking density increases (Table 1). PO ₄ ^-3 concentrations were significantly higher in BFT-1250 (2.82 mg/L) and BFT-1000 (2.85 mg/L) as compared to BFT-500 (2.60 mg/L) (Table 1).

In relation to water consumption, BFT-1250 and BFT-1000 treatments had the highest consumption (101.54-94.77 L/kg) and BFT-500 the lowest (52.48 L/kg). The use of settling chambers was necessary in all experimental units, but its sedimentation time in BFT-1250 was significantly higher (305 h/kg) than in the other treatments (238-271 h/kg) (Fig. 1).

Fish survival rates were all above 96% during the 42-day experimental period and significant differences (a ≤ 0.05) were found in BFT-500, BFT-750 and BFT-1000 compared to BFT-1250. Fish growth (final weight and final length) were more accelerated at the lowest stocking density (BFT-500) as compared to the highest density (BFT-1250), although resulted in a lower yield. No significant differences (a > 0.05) were observed for SGR and FCR, but significant differences (a ≤ 0.05) were observed for PER between BFT-1250 and the other stocking densities (Table 2).

Weight gain and length had positive correlations with alkalinity and TSS, and negative correlations with density, NO ₃ and PO ₄ ^-3 . For FCR, the inverse was found, negative correlations with alkalinity and TSS, and positive correlations with PO ₄ ^-3 . As for yield, a positive correlation with density and TSS was recorded, and a negative correlation with survival and density. Water consumption had a negative correlation with density, TSS and PO ₄ ^-3 (Table 3).

The water quality variables (dissolved oxygen, temperature, salinity and pH) in the culture water were within the range recommended for Nile tilapia culture ( El-Sayed, 2006 ).

TSS and SS increased in higher stocking densities, probably due to increasing amounts of feed, molasses and feces. Peaks of TSS (1,005 mg/L) were found in BFT-1250 on the fourth week of culture, but its mean value was close to 500-600 mg/L for the stocking densities evaluated. These amounts are similar to those observed by Long et al. (2015) who investigated the effects of biofloc technology on growth, digestive activity, hematology, and immune response at a C:N ratio of 15, and Zapata-Lovera et al. (2017) who evaluated C:N ratios of 10, 15 and 20. According to Avnimelech (2012) TSS can reach 1,000 mg/L in tilapia culture with bioflocs.

SS also presented the same rise as stocking density increased, however, the values were within the range (5 to 50 mL/L) recommended by Avnimelech (2011) , possibly due to the settler used to maintain SS values under this limit. High concentrations of solids are not favorable for fish growth, since they contribute to a greater consumption of oxygen, and may cause accumulation of organic matter and obstruction in fish gills ( Hargreaves, 2006 ; Avnimelech, 2011 ).

Although the treatments with higher stocking densities presented lower alkalinity values, these were higher than 100 mg/L, as recommended by Hargreaves (2013) , so as not to interfere with nitrification processes by nitrifying bacteria and absorption of ammonia by heterotrophic bacteria. The alkalinity values were maintained at ideal values by corrections made with calcium hydroxide.

The concentrations of TAN and NO ₂ were low and related to the use of water from a biofloc matrix tank (50% of the tank volume), so that only 30% molasses was used in relation to the quantity of feed supplied. Low TAN and NO ₂ concentrations were also observed in other studies that used previously prepared biofloc water ( Pérez-Fuentes et al. , 2016 ; Zapata-Lovera et al. , 2017 ). The use of such water prior to fish stocking led to an increase in heterotrophic and nitrifying bacteria, which transform ammonia into microbial biomass and NO ₃ , respectively. A higher concentration of NO ₃ than other nitrogen compounds indicates that the controlled addition of molasses does not inhibit the development of nitrifying bacteria. The stocking densities tested did not seem to influence these nitrogen compounds, contrasting with results obtained in a system without addition of an organic carbon source ( Luz et al. , 2012 ).

In biofloc systems there is an increase of PO ₄ ^-3 concentrations in the water ( Luo et al. , 2014 ; Day et al. , 2016 ). According to Thakur & Lin (2003) , a large portion (38.8-66.7%) of the phosphorus that enters pond systems is deposited in the sediment, which does not exist in biofloc systems. This accumulation was influenced by the increased stocking density, with increasing input of phosphorus by feed addition.

Effects of increased stocking density on WC and ST were observed, where the treatments with higher densities of 1,000 and 1,250 fish/m ³ (respectively 94.77 and 101.54 L/kg) presented significantly higher values (Fig. 1). These higher values are related to the higher concentration of TSS and SS, leading to the need for a longer total time of settling chambers, and consequently greater water replenishment. Despite the higher use of water at the higher densities tested, the amount is still much lower than in traditional systems with water renewal. Zapata-Lovera et al. (2017) evaluated water consumption in tilapia culture in a water exchange (1,800 L/kg) and a biofloc system (90-190 L/kg) and found that a biofloc system can be used to produce tilapia, especially in places with limited water availability.

The lowest survival (96.82%), final weight (12.4g) and final length (8.84 cm) was found in BFT-1250, indicating the negative effect of high density. Water-flow systems, such as raceways ( Silva et al. , 2002 ) and cages ( Maeda et al. , 2010 ), also have reductions in final weight at increased stocking densities. These results suggest greater competition for feed and space due to cannibalism and stress, influencing the performance of the fingerlings.

SGR and FCR were not influenced by the stocking density, and our results were similar to those observed in other studies with tilapia in freshwater biofloc systems ( Zapata-Lovera et al. , 2017 ), thus indicating that the effect of a salinity of 10 g/L in a biofloc system does not seem to influence the productive performance of tilapia at the densities studied. However, PER was influenced by the increased stocking density, since the lowest values were found in the BFT-1250 treatment, due to the lower performance of the organisms.

In relation to salinity, some studies have found a decrease in the hematological parameters and histopathological alterations at increased water salinity levels ( Azevedo et al. , 2015 ), however, in biofloc systems salinity does not seem to be the main stress factor for fish. According to Lima (2017) , tilapia grown in a biofloc system with salinity of 10 g/L and high levels of settleable solids (55 mL/L) show signs of stress (reddish body extremities) and hemorrhaging, leading to reduced growth and decreased survival. However, the higher solids levels did not appear to have negatively influenced yield, since the treatments with higher stocking density presented a yield of 15 kg m ^-3 . Biomass produced in biofloc systems can result in a yield of 10-40 kg/m ³ ( Avnimelech, 2007 ), similar to the results observed in this study (9.4-15 kg/m ³ ), therefore indicating good results even at a salinity of 10 g/L.

To summarize, the results of this study confirm the zootechnical potential of Nile tilapia fingerling cultivated in a low-salinity biofloc system, with a positive effect of stocking density increase on yield, but a negative effect on zootechnical performance (survival, final weight and final length), water quality (TSS, SS and PO ₄ ^-3 ), WC and ST. It is possible to increase the stocking density of Nile tilapia fingerlings up to 1000 fish/m³ in a low-salinity biofloc system, since this has no effect on the zootechnical potential. However, studies are needed to determine the economic viability and compensatory growth for Nile tilapia fingerling culture in the growth stage.