Developing prediction equation for digestibility of nutrients in faeces from individual atlantic salmon

The purpose of this study was to develop a robust and reliable approach to predict apparent digestibility coefficients of fat and protein of individual Atlantic salmon (Salmo salar), based on multivariate regression models from measured NIR (near infrared reflectance) spectra of their faeces. The pe...

Full description

Bibliographic Details
Main Author: Appiah, Augustine Kwarteng
Format: Master Thesis
Language:English
Published: Norwegian University of Life Sciences, Ås 2015
Subjects:
fat
Online Access:http://hdl.handle.net/11250/2364816
Description
Summary:The purpose of this study was to develop a robust and reliable approach to predict apparent digestibility coefficients of fat and protein of individual Atlantic salmon (Salmo salar), based on multivariate regression models from measured NIR (near infrared reflectance) spectra of their faeces. The period of the experiment from rearing to faecal analysis was from July 2014 to March 2015. A total of 180 faecal samples of protein collected from 10 different experiments and 115 faecal samples of fat also all collected from 9 different experiments were used to calibrate the NIR instrument. Full cross-validation was used for the fat calibration, with 5 samples randomly selected and chemically analyzed for fat to complement the fat validation. 23 faecal samples from a different experiment were also used to validate the protein calibrations. Faeces from 60 Atlantic salmon (Salmo salar) of different families were used for the prediction in the study. The developed equation had better prediction precision (R2; 0.97, RESD; 0.19, and Bias; -0.002) for protein than for fat (R2; 0.92, RESD; 0.29, and Bias; 0.02). In comparing with results of other studies performed in similar ways, the protein model had good prediction results (15.6, of data set A) and the fat (5.26, of data set A) as compared with other studies. However, detailed description explanation of the prediction (data set A) of individual is treated in parallel thesis. M-AA