Health assessment and size measurement of Atlantic salmon from pictures and videos in a commercial sea cage at 4 depths
Description
This repository contains the raw data and R scripts used in the statistical analyses for our publication: "Exploring physical health and size stratifications of Atlantic salmon (Salmo salar) with depth in a commercial sea cage." We encourage readers to consult the article for detailed information on data collection, variable definitions, analytical methods, and conclusions. We present the raw health scoring data for our two datasets, the stereovision dataset and the CView Eye (automatic sensor; CView Eye 2023-2024, Createview AS, Molde, Norway) dataset, along with size measurements: fork length (measured using EventMeasure; V6.43–64 bit, SeaGIS Pty Ltd 2006–2024, Bacchus Marsh VIC, Australia) for the stereovision dataset, and weight (automatically measured by the stereovision cameras of the sensor) for the CView Eye dataset. Data were collected at four depths (1 m, 5 m, 9 m, and 14 m) in a commercial sea cage (Gudmundset fish farm in Møre og Romsdal county, Norway) between June and August 2024. For the health assessments, fish were scored from 0 = no damage to 3 = serious damage, following the Laksvel protocol ("A standardised, operational welfare monitoring protocol for Atlantic salmon held in sea cages", Nilsson et al., 2025) for the selected indicators: spinal deformity (SD), jaw deformity (JD), emaciation (E), operculum damage (O), scale loss (SL), Body wound (BW). In the Stereovision dataset, scoring was performed manually from video footage. In the CView Eye dataset, scoring was performed manually on a random selection of images selected by the sensor. Dataset Summary Stereovision dataset: 4,053 fish scored and measured for fork length → File: HealthScoring-LengthMsrts-Stereovision.xlsx CView Eye dataset: 8,641 fish scored (note: not all indicators were possible to score due to bad fish orientation, visibility, or lighting on the pictures) → File: HealthScoring_fromsensorpictures.xlsx 57,616 fish retained for weight estimation → File: SizeMsrts_automaticsensor.xlsx Statistical Analyses Principal Component Analysis (PCA) was used to explore correlations among health indicators → Script: PCAscript.R Generalized Linear Mixed Models were applied to each health indicator as an outcome variable for both datasets → Scripts: HealthScores_analysis_Stereovision.R, HealthScores_analysis_CViewdata.R Linear Mixed-Effects Models were used to analyze size data (length or weight) → Scripts: Length_analysis_Stereovision.R, Weight_analysis_CViewdata.R