Seismic disturbance, productivity and depth shape hadal benthic habitats and biodiversity in the Japan, Ryukyu and Izu-Ogasawara trenches (northwest Pacific Ocean)

This dataset contains biological annotations from six hadal submersible dives conducted in Japan, Ryukyu, and Izu-Ogasawara (Izu-Bonin) subduction trenches in the Northwest Pacific Ocean during August–September 2022. Video footage spanned depths of 6939 m to 9775 m and allowed for observations of benthic biodiversity and habitat characteristics across trenches differing in seismic activity and productivity regimes. The dataset contains nearly 30,000 organism observations, representing 70 morphotaxa across 11 phyla and eight habitat categories. The dataset supports the associated publication, investigating the effects of multiscale environmental drivers on benthic community structure and habitat characteristics in hadal environments. Accompanying R code is provided for generating dive-level visual summaries, as well as basic community analysis and visualisation (ordination, indicator species, and assessments of community composition).

Methods
Video data were collected during six crewed hadal submersible dives in Japan, Ryukyu, and Izu-Ogasawara (Izu-Bonin) trenches in August–September 2022. Dives followed transect-style surveys at depths of 6939–9775 m, with high-definition forward- and downward-facing video recorded throughout.

Megafauna (benthic, demersal, and benthopelagic organisms visible in the video) were annotated using EventMeasure software. Only footage with good visibility, even lighting (~1 m above seafloor), and steady submersible movement was included. Organisms were identified to the lowest possible taxonomic level; unidentified forms were grouped into morphologically distinct morphotaxa and assigned feeding and mobility traits.

Annotation sare given for each dive ("Dive_xx"). To account for differences in dive duration and usable video footage, relative abundance was also calculated as the number of individuals per minute of analysed video. 

In addition, each dive was segmented into three minute non-overlapping sampling units ("Subperiod_xx"). Each sampling unit was assigned a dominant habitat category using a majority-rule approach. Eight habitat categories were defined across the dataset based on substrate type and slope gradient.