Building use-inspired species distribution models: Using multiple data types to examine and improve model...

Area of Science:

• Marine ecology
• Conservation biology
• Fisheries science

Background:

• Species distribution models (SDMs) are crucial for marine conservation and management.
• Increasing marine biodiversity data necessitates guidance on leveraging diverse data types for robust SDMs.

Purpose of the Study:

• To explore the effect of different data types on SDM fit, performance, and predictive ability.
• To compare models trained with fishery-dependent and fishery-independent data for the blue shark (Prionace glauca).

Main Methods:

• Compared SDMs trained with conventional mark-recapture tags, fisheries observer records, satellite-linked electronic tags, and pop-up archival tags.
• Evaluated model fit, performance, and spatial predictions for the blue shark in the Northwest Atlantic.
• Assessed model ensembles and pooled data models for integrating inferences.

Main Results:

• All four data types produced robust SDMs, but spatial predictions varied due to sampling biases.
• Differences in data sampling and absence representation influenced model outcomes.
• Model ensembles and pooled data models yielded more ecologically realistic predictions than individual models.

Conclusions:

• Guidance is provided for practitioners developing SDMs with diverse data sources.
• Future work should focus on integrative modeling approaches that leverage individual data type strengths and account for limitations.
• Ecological realism in model selection and interpretation is vital, regardless of data type used.