Rapid prediction of nitrogen compounds concentrations in water of Recirculating Aquaculture Systems (RAS) using portable near-infrared spectroscopy combined with a principal component analysis- neural network-based model

Recirculating aquaculture systems (RAS) are land-based, closed-loop systems that reuse water by passing it through a filtration system, reducing the amount of fresh, clean water used and the space required for fish farming. They are therefore sustainable systems that provide a controlled and biologically safe environment in which to grow fish. The success of a fish farm is significantly depending on whether the fish can live in an environment with optimal water quality. A key process in RAS for purifying water is nitrification, a process in which bacteria convert ammonia excreted by fish into nitrite and then nitrate. Ammonia and nitrite are extremely toxic for fish, so they need to be prompt detected and monitored. The aim of the study was to assess the ammonia, nitrite and nitrate concentration in water of RAS using an ultra-compact Near Infrared (NIR) spectrometer. In the context of the Fish-PhotoCAT project, 32 samples of water were collected from six experimental RAS in which adult rainbow trout were reared at a density of 15 kg/m3 for 30 days. NIR calibrations were developed by means of principal component analysis (PCA)-neural network obtaining models with a fairly good coefficient of determination (R2C = 0.83 and R2CV = 0.85 for NH3-N, R2C = 0.79 and R2CV = 0.80 for NO2-N, R2C = 0.89 and R2CV = 0.88 for NO3-N) and a reasonable prediction error (RMSECV = 0.05, 0.12 and 12.33 for NH3-N, NO2-N and NO3-N respectively). Based on these results, the portable spectrometer would be useful for providing a fast screening of NH3-N, NO2-N and NO3-N in water samples at farm level, enabling proper management of recirculating aquaculture systems and rapid turnaround in plants advisory systems.