Fabio Cabrera De Leo

We present a comprehensive study of deep‐sea bioluminescent organisms in the Cascadia Basin, Northeast Pacific Ocean, based on four years (2020–2023) of in situ measurements using STRings for Absorption length in Water instruments. Analysis of over 1.3 million camera images and continuous photomultiplier tube recordings revealed that bioluminescent activity at depths of 2.0–2.6 km is driven by mechanical stimulation from tidal currents, with emission rates showing strong periodicity across quaddiurnal, semidiurnal, and diurnal scales. Machine‐learning‐assisted chromatic characterization of emission peaks identified a taxonomic composition primarily consisting of blue‐emitting organisms (470–500 nm), likely including glass squid (Leachia lemur) and various planktonic species, alongside a non‐negligible population of yellow‐green emitters (5̃65 nm) tentatively associated with tomopteris polychaete worms. We quantified an effective organism abundance of , a value consistent with earlier indirect assessments but here confirmed through long‐term, direct observation. These findings establish a baseline for the biological optical background in the Cascadia Basin, providing critical parameters for the design and background‐modeling of large‐scale neutrino telescopes such as Pacific Ocean Neutrino Experiment (P‐ONE).

Organic-rich whale bones and wood falls occur widely in the deep sea and support diverse faunal communities, contributing to seafloor habitat diversity. Changes in community structure through succession on deep-sea bone/wood substrates are modulated by ecosystem engineers, i.e., bone-eating Osedax annelids, and wood-boring Xylophaga bivalves. Here, we use a comparative experimental approach and Ocean Networks Canada’s (ONC) cabled observatory in hypoxic Barkley Canyon to study, at high temporal resolution, colonization and succession on whale-bone, Douglas fir wood, and control carbonate rock over 8.3 mo. Experimental substrates were similar in size and mounted on PVC plates near the seafloor at 890 m depth and monitored by high-definition video camera for 5-min intervals every 6-12 h over a period of 8.3 mo. A broad range of seafloor and sea-surface environmental variables were also monitored at this site over the 8.3 mo to account for environmental variability and food input. Following loss of the high-definition camera, substrates were surveyed approximately annually with lower resolution ROV video for an additional 8.5 y. We find that megafaunal abundances, species diversity, and community structure varied substantially over 8.3 mo on each substrate, with markedly different patterns on whale bones due to the development of extensive white bacterial mats. A combination of seafloor and sea surface variables explained < 35% of bone/wood community variation. Bone-eating Osedax annelids failed to colonize whale bones even after 9.2 years, and boring Xylophaga bivalves colonized the wood at much lower rates than in better oxygenated deep-sea locations. Species diversity on whale-bone and wood substrates appeared to be substantially reduced due to the absence of ecosystem engineers and the low oxygen concentrations. We hypothesize that Osedax/Xylophaga colonization, bone/wood degradation, and bone/wood community development may be limited by oxygen concentrations of 0.22 - 0.33 ml.l on the NE Pacific margin, and that OMZ expansion due to climate change will reduce whale-bone and wood degradation, and the contribution of whale falls and wood falls to beta diversity, on the NE Pacific margin.

Biological data sets are increasingly becoming information-dense, making it effective to use a computer science-based analysis. We used convolution neural networks (CNN) and the specific CNN architecture Unet to study sponge behavior over time. We analyzed a large time series of hourly high-resolution still images of a marine sponge, Suberites concinnus (Demospongiae, Suberitidae) captured between 2012 and 2015 using the NEPTUNE seafloor cabled observatory, off the west coast of Vancouver Island, Canada. We applied semantic segmentation with the Unet architecture with some modifications, including adapting parts of the architecture to be more applicable to three-channel images (RGB). Some alterations that made this model successful were the use of a dice-loss coefficient, Adam optimizer and a dropout function after each convolutional layer which provided losses, accuracies and dice scores of up to 0.03, 0.98 and 0.97, respectively. The model was tested with five-fold cross-validation. This study is a first step towards analyzing trends in the behavior of a demosponge in an environment that experiences severe seasonal and inter-annual changes in climate. The end objective is to correlate changes in sponge size (activity) over seasons and years with environmental variables collected from the same observatory platform. Our work provides a roadmap for others who seek to cross the interdisciplinary boundaries between biology and computer science.