Addressing measurement error in lobster growth modelling

Abstract

Mark and recapture tagging data is key for determining growth in species that are hard or impossible to age. Obtaining this data can be an expensive process, consequently there is an increasing reliance on measurements from non-scientists including commercial and recreational fishers and industry volunteers. A challenge with relying on data collection from these groups is that they are likely to have higher measurement errors. We demonstrate that this measurement error can introduce substantial bias. To account for this we developed a Bayesian model that is robust to a broad range of measurement uncertainty, which enables stochasticity in model outputs to be attributed to more appropriate causes, such as environmental drivers for further investigation. The model was tested through application to Southern Rock Lobster (SRL), Jasus edwardsii data. This was applied in two distinct areas in Tasmania, with different biological characteristics and data collection regimes. Application to this case study demonstrates that high measurement error can be problematic for commonly used methods, however our developed approach allows unbiased growth estimation from such datasets.