Using multivariate analysis of water quality in RAS with Nile tilapia (Oreochromis niloticus) to model the evolution of macronutrients

Since fish are very sensitive to water quality, their welfare is greatly influenced by the environment. Little is known about the most suitable levels of ions for optimum growth in fish, although their concentration tends to increase with the accumulation of waste and uneaten feed. Maintaining good water quality is important since it will also affect biofilter function and provide optimal growth and better fish health. Multivariate analysis was used to study the evolution of water quality in thirteen feeding trials with Nile tilapia (Oreochromis niloticus). Specifically, thirteen different variables were measured: pH, electrical conductivity, ammonium, nitrite, nitrate, bicarbonate, sodium, potassium, calcium, magnesium, chloride, phosphate and sulfate. A random regression model was assessed to determine the evolution of these variables in time, resulting in an increase in its level except for ammonium and nitrite, which tended to decrease with time, and calcium, that was more variable among trials. A posteriori, trials were sorted into three groups, depending on the fish stocking density (less than 1, between 1 and 2 and more than 2 kg m(-3)). Random regression analysis allowed to find equations that describe the behaviour of each ion, showing different patterns depending on the variable. Principal component (PC) analysis suggests that most of the variance is described by two PCs, the first explained the total content of dissolved ions and the second most important PC was related to fish and feed. That implies that changes in water chemistry are separate from changes in fish density or feeding and they explain more of the variation in ion concentrations. The application of random regression models and PCA provides a meaningful characterization of RAS water samples based on water quality criteria. (C) 2012 Elsevier B.V. All rights reserved.