Poster Presentation Australian Society for Fish Biology Conference 2017

Comparing indices of abundance from standardised catch rates calculated using alternative imputation methods (#154)

Ross J Marriott 1 , Berwin Turlach 1 , Kevin Murrach 1 , David Fairclough 2 , Gary Jackson 2 , Jeffrey Norriss 2 , Stephen Newman 2
  1. University of Western Australia, Nedlands, W.A.
  2. Department of Fisheries, Government of Western Australia, Hillarys, W.A.

An index of fish abundance is typically calculated from the estimated marginal means predicted from a generalised linear model fitted to fishery catch rate data and suitable explanatory variables. However, fishing grounds can change, because fleets often shift their activity to target different areas of a fish population over time, which can create gaps in the dataset and lead to biases in the index. It has been shown that such biases can be reduced by using imputation methods to fill the gaps. Some published methods impute a constant relative abundance over the period with missing data. Our simulations demonstrated that imputations accounting for a possible increase in relative abundance over time generally resulted in less biased estimates. Here, we follow on from that work by applying these methods to real fisheries datasets to explore if, and how, the use of alternative types of imputations can affect results. Catch rate data for indicator species of several Western Australian demersal scalefish fisheries were obtained from commercial fishing logbook returns. We compared the standardised catch rates that were calculated for each of these species using different imputation methods and statistical models. Results show that, at least for some datasets, the choice of imputation method can have an important effect on the imputed index trend, and on the estimated relative fish abundance in the most recent year.  This poster provides a brief summary of those findings.