Ensemble habitat suitability modeling of vulnerable marine ecosystem indicator taxa to inform deep-sea fisheries management in the South Pacific Ocean

Abstract

Abstract Resolutions of the United Nations General Assembly (UNGA) require states and competent authorities to protect vulnerable marine ecosystems (VMEs), ecologically important habitats in the deep sea that are considered to be especially at risk from anthropogenic disturbances such as fishing. The lack of data concerning the location and extent of VMEs poses a significant obstacle to their protection. Habitat suitability modeling is increasingly used in spatial management planning due to its ability to predict the distribution and niche of marine organisms based on limited input data. We generated broad-scale, medium-resolution (1\u2009km2) ensemble models for ten VME indicator taxa within the New Zealand Exclusive Economic Zone and a portion of the South Pacific Regional Fishery Management Organisation (SPRFMO) convention area. Ensemble models were constructed using a weighted average of three habitat suitability model types: Boosted Regression Trees, Maximum Entropy, and Random Forest. All models performed well, with area under the curve scores above 0.9, and ensemble models marginally outperformed any of the individual modeling approaches. Highly suitable habitat for each VME indicator taxa was predicted to occur in relatively small areas throughout the region, typically associated with elevated seafloor features with steep slopes. Determining the spatial distribution of VME indicator taxa is critical for assessing the current and historical extent of bottom trawling impacts on benthic communities, and for supporting the improved spatial management of fisheries in the South Pacific Ocean. Given the additional threats of climate change and ocean acidification to VME indicator taxa throughout the deep sea, habitat suitability modeling is likely to play an increasing role in designing effective, long-term protection measures for cumulative impacts on VMEs.