Repository logo

Browsing by Author "Matos, João Duarte"

Now showing 1 - 1 of 1
  • Development of models for European seabass feed intake prediction
    Publication . Matos, João Duarte; Conceição, Luís Eugénio Castanheira da; Silva, Tomé Pereira de Azevedo Santos
    The aquaculture production of finfish has become one of the fastest-growing animal food-producing sectors worldwide and is today a key provider of seafood. As the magnitude of production expands, so does the probability that new biological, economic, and social issues will emerge, posing a threat to the industry's ability to maintain ethically sound, productive, and ecologically friendly fish production. To combat these challenges, mathematical models of feed intake, in particular, have become essential tools in aquaculture, both in commercial and research circumstances. Feed intake in the scientific literature is usually modelled as a function of BW and temperature, while other important environmental and nutritional factors are often neglected. This thesis aims to improve European seabass feed intake modelling by developing feed, energy, and nutrient intake models that explicitly consider dissolved oxygen's effects and combine them to predict feed intake under different environmental and nutritional contexts. To achieve this, three major tasks were set. In Task 1, data from publicly available sources (aquafeed companies and scientific literature) was used to generate oxygen-independent intake models of feed (FI), energy (EI), protein (PI), and lipid (LI) intake models. In Task 2, data from three European seabass (Dicentrarchus labrax) trials showed that: feed gross energy had a negative marginal effect on FI and PI, and a positive marginal effect on LI, and no clear effect on EI; dissolved oxygen had a positive marginal effect on FI, EI, PI, and LI; and no clear evidence of “oxygen × nutrient” interaction effects was observed. In Task 3, two oxygen-dependent models (sigmoid and segmented) were developed to isolate the effects of dissolved oxygen on the FI, EI, PI, and LI intake responses. These models were then evaluated regarding their capacity to predict observed feed intake rates in Trials 1, 2, and 3. This work showed that the addition of oxygen effects largely led to improved predictions, and that the feed intake prediction errors of the models generally range from 32 to 40% (either in sigmoid or piecewise-linear form). The comparison of the different models also revealed that the feed intake model based on EI seems to display the lowest error levels and also displayed the largest improvement when adding the effect of oxygen. The opposite was observed for LI.
    2021-11-16Master thesis Restricted Show more