Workshop 2 on Fish Distribution (WKFISHDISH2; outputs from 2022 meeting)
The objective of WKFISHDISH2 was to develop a standardized and open-source way of routinely using trawl survey data to produce distribution maps which can be easily updated. To do so, workshop participants (i) reviewed models that can produce distribution maps from survey data in DATRAS and MEDITS formats, (ii) considered best practice guidance for data, and models, (iii) implemented best practice to produce distribution maps in a transparent manner, and (iv) populated an ICES repository with distribution maps and associated scripts. Nine models were reviewed, seven of which were considered appropriate to produce distribution maps using survey data. The two models excluded failed to estimate gear standardization factors. Best practice for data should consider the following: correct/remove erroneous observations, account for changes in haul duration/timing affecting catchability, use species-specific modelled areas, remove gears/surveys with no observations to reduce model converging time, and include appropriate explanatory variables. A step-by-step list to preprocess survey data was provided. Best practice for models should consider the following: account for skewed distribution of survey data by exploring different statistical error distributions, use open-source models able to reproduce distributions from simulated data, consider the complexity of the model required vs. the run-time (e.g. number of knots), avoid extrapolating the model to areas/depths where the species is not observed, and perform relevant model diagnostics/model selection. To compare distribution estimated by different models, the SPAtial EFficiency metric (SPAEF) was used together with centres of gravity, biomass hot spots (90th percentile), and effective occupied areas. All seven models were able to produce satisfactory distribution maps for at least one species within the workshop time constraints, both with DATRAS data in the Atlantic and the MEDITS data in the Mediterranean. Comparisons between model estimates showed reasonable consistency overall, although discrepancies were noted owing to different model structures. Distribution maps produced during the workshop and the associated script are available at the ICES SharePoint. Participants identified priorities for future research to be tackled during a subsequent workshop: combine outputs of different models in a single distribution map, include fish length in models, investigate environmental covariates and their use for predictions, investigate models’ abilities to reproduce known distributions and further consider the impact of vessel/gear effects.
Published under the auspices of the following ICES Steering Group or Committee