Ontogenetic habitat associations of a demersal fish species,Pagrus auratus, identified using boosted regression trees

• Published in: Marine ecology. Progress series (Halstenbek) Vol. 462; pp. 219 - 230
• Main Authors: Compton, Tanya J., Morrison, Mark A., Leathwick, John R., Carbines, Glen D.
• Format: Journal Article
• Language: English
• Published: Inter-Research 21-08-2012
• Subjects: Aquatic habitats; Forest habitats; Habitat preferences; Marine; Marine fishes; Modeling; Orbital velocity; Pagrus auratus; Sediments; Snapper; Spatial models; Young animals; PSE, New Zealand, North I., Hauraki Gulf; PSE, New Zealand
• ISSN: 0171-8630; 1616-1599
• DOI: 10.3354/meps09790

In coastal areas, the identification of habitat types critical to fish life history strategies can provide useful information for ecosystem-based management. Recent studies show that species distribution modelling can be a cost effective tool for describing fish habitat. However, few modelling studies have examined ontogenetic habitat associations. This is critical, as fish species often have different habitat preferences depending on their life stage. In this study, we used boosted regression trees (BRT) to describe ontogenetic habitat associations in snapperPagrus auratusacross the inner Hauraki Gulf of New Zealand. The BRT models identified that juvenile snapper were most frequently associated with slow orbital velocities and slow tidal current speeds, as well as biogenic sedimentary structures (area under the receiver operating curve or AUC, a measure of model performance, 0.79). In contrast, larger snapper were associated with faster tidal currents and faster orbital velocities (AUC 0.78). Juvenile and adult snapper were spatially separated; juvenile snapper occurred in waters close to shore, whereas large snapper occurred mainly in the channels between the islands and the waters around the islands. The successful discrimination of adult and juvenile habitat associations suggests that a modelling approach such as this could be useful for ecosystem-based management.