LEVERAGING MACHINE LEARNING TO INFORM SPATIAL AND TEMPORAL PATTERNS IN DEEP-SEA CORAL BIODIVERSITY, GULF OF MEXICO AND US WEST ATLANTIC

Characterizing deep-sea coral biodiversity is essential to evaluate the current state of deep-sea ecosystems and to assess vulnerability to anthropogenic threats such as offshore drilling and ocean acidification. Thousands of deep-sea coral records from the past six decades are compiled in the publicly accessible NOAA database, but few large-scale analyses of deep-sea coral biodiversity through time have been conducted. This study provides the longest temporal analysis of deep-sea coral generic biodiversity (58 years), and the first such study for US waters. Patterns of deep-sea coral generic biodiversity were calculated for observed and simulated data from 1960-2018 in the US marine ecoregions of the Gulf Stream Slope, Carolinian Atlantic Shelf and Slope, South Florida/Bahamian Atlantic Shelf and Slope, and Northern Gulf of Mexico (NGOM) Shelf and Slope. There are statically significant decreases in deep-sea coral generic biodiversity for all ecoregion depth pairs in simulated data and most ecoregion depth pairs in observed data. There is relative stability in biodiversity from 1960 to the mid-2000’s followed by generally rapid decreases between 2007-2011. Biodiversity stabilized after 2011, though it remains at the lowest average level in the past 58 years. Most ecoregions and depths show the lowest generic richness during the most recent time interval (2010-2018) compared to the preceding five decades. The most persistent deep-sea coral genera from 1960-2018 vary based on ecoregion and depth. The rapid decreases of biodiversity support previous work showing deep-sea corals can undergo biodiversity loss in less than four years, which is alarming given their slow growth and decadal recovery times. This temporal, machine learning based analyses provides a robust temporal framework for future researchers to evaluate casual mechanisms affecting deep-sea coral biodiversity and ultimately may inform conservation strategies.