Predicting farmed rainbow trout weight distribution to improve feeding practice: an individual-based model approach.

Abstract

In the near future, the role of aquaculture in human diet is likely to increase due to the rising demand for fish proteins. With a long tradition of rainbow trout (Oncorhynchus mykiss) farming, Italy is one of the main producers of this species in the European Union (EU). The EU is allocating economic resources to foster the sustainable development of the aquaculture sector, aiming to produce more while using less resources. Precision fish farming (PFF) is a promising approach to achieve this goal and its implementation is being facilitated thanks to the reduction of costs of sensors. PFF will likely lead to a new generation of mathematical dynamic models based on sets of control variables and external forcing functions, coping with Big Data and machine learning techniques. In this work, we developed an individual-based dynamic model for the simulation of the fish size distribution and total biomass of a population of rainbow trout within a raceway. At its core, there is a bioenergetic individual model that can simulate weight changes taking into account water temperature and feeding regime. This model was tested against weight observations collected by a non–invasive monitoring system, that was deployed for the first time in a trout farm. The model allows one to estimate the optimal feeding ration based on fish weight and water temperature.  The results indicate that current methodologies, based on the estimation of the average weight, lead to slightly overestimate the feed ration: therefore, the model proposed here would allow one to save feed, thus reducing operational costs.