Assessing Accuracy and Precision of Measurement and Sampling Schemes in Aquaculture Research

Measurement and sampling techniques commonly used in aquaculture research studies are varied and range from bulk-tank measurements (i.e. netting fish from a tank into a smaller calming receptacle on a scale to determine the total weight of fish in a tank) to individual weighing of all or a netted subset of fish from a tank. Beyond considerations for animal welfare and research logistics, concerns about differential data quality call for an evidence-based approach to inform selection of measurement and sampling schemes.

To this end, we assess accuracy and precision of measurement and sampling schemes commonly used for the evaluation of weight performance in dietary studies conducted on Atlantic salmon (Salmo salar). We use pilot data consisting of weekly weights, both individual and bulk-tank, for 1080 pit-tagged parr-sized fish (average initial weight 58g, SD 17g) sourced from 45 families, stocked into 24 0.19m3 recirculating aquaculture system tanks and fed the same commercial diet. Using a combination of graphical tools, descriptive statistics and mixed-model-based analyses fitted to tank-level data and to fish-level data, we first evaluate relative accuracy of weight observations across measurements and sampling schemes. We then use a variance components approach to characterize hierarchical sources of non-systematic variability in a typical aquaculture production system and quantify their differential relative magnitudes under each scheme. Finally, we explore inferential precision for detecting putative dietary effects across measurement and sampling schemes using mixed-models based precision analyses tailored to the design of aquaculture studies. Ultimately, we endeavor to elicit recommendations for high-quality data collection schemes in aquaculture research. This is work in progress; we seek input from the group and welcome open discussion.