Evaluating predictability of catch in a highly mixed trawl fishery using stacked and joint species models (110152)
James Smith
1
,
Daniel Johnson
1
- NSW DPI Fisheries, TAYLORS BEACH, NSW, Australia
| Evaluating drivers and the predictability of catch is valuable for the management of mixed fisheries. Drivers can represent or help to identify levers for management, and predictable catch compositions are a key component of simulation tools and dynamic management strategies. But modelling mixed fisheries can be challenging due to the large number of taxa, and analysis typically focuses on a few key species or highly aggregated taxa. Here we present results from stacked and joint species models to explore the drivers and predictability of trawl-level catches in an ocean prawn trawl fishery, in New South Wales, Australia. We also evaluate the value of ‘conditional joint prediction’ as a tool for predicting unobserved discards from observed catches. We found that predictive skill was reasonably high, with most of the 130 taxa examined showing a useful level of predictability. We also found that a random forest hurdle model in a stacked framework was the most accurate modelling approach. Conditional joint prediction was not a useful addition, showing that it’s more important to include key environmental variables than to focus on incorporating co-occurring taxa. These results will aid the development of predictive models to enhance and supplement on-board observer catch records. |