Collection:
Meroplankton larvae and benthopelagic invertebrates were collected using DeepZoo, a novel autonomous sampler developed at Woods Hole Oceanographic Institution's Autonomous Vehicles and Sensor Technologies Innovation Hub (WHOI AVAST). DeepZoo is a cost-effective, lightweight (6.4 kg), all-depth device currently at the working prototype stage. It integrates two systems: (1) sample collection, composed of a lid with an oil-filled 10 RPM gear motor (ServoCity, USA), a 63-µm Nitex mesh net, and a T200 thruster (Blue Robotics, USA) to drive water flow; and (2) system control, housed in titanium and containing electronics, including a 14.8 V, 10 Ah rechargeable lithium-polymer battery (Blue Robotics). Pre-programmed in Arduino IDE 2.3.6 (Arduino 2024), DeepZoo runs autonomously, controlling motor and thruster operations (i.e., timing and speed) according to dive plans. The thruster operates at a specific rate, defined by a pulse-width modulation (PWM) signal, which ranges from 1500 to 1900 microseconds. The two systems are contained within a lightweight PVC frame with SpeedRail connectors and a wire-mesh base (91.4 × 27.9 × 30.5 cm).

DeepZoo was deployed three times in a horizontal configuration, once on the MISO Lander and two on the HOV Alvin science basket. On the MISO lander, DeepZoo was deployed in parallel with a large-volume plankton pump (McLane WTS-LV50, Falmouth, MA, USA). Before sampling, HOV Alvin repositioned the lander to be within meters of YBW-Shimmering Forest. Both systems were sampled over the same 21.5-hour period. This sampling design makes the two systems comparable, where the key difference is that the pump is situated 1.5 m above the seafloor and samples vertically. On HOV Alvin, DeepZoo was mounted on the starboard side of the basket with the mouth facing forward. During these two deployments, DeepZoo sampled while HOV Alvin was moving on the seafloor, including during transit to and from a station. Compared to the lander, DeepZoo sampled for a shorter period at a higher thrust rate.

For the named vent sites in this dataset, we used the most recent benchmarked, georeferenced positions from Table 1 in Wu et al. (2022). Bottom depth for positions near 9°50’ N was determined in QGIS using bathymetry acquired in 2018, 2019, and 2021 (Parnell-Turner et al. 2022).

Shipboard sample processing:
Upon recovery, the net section of DeepZoo was placed into a 5-gallon bucket with chilled, filtered seawater (CFS; 0.5-micron) and brought to the cold room (4°C). The net was rinsed with CFS and then poured over a 63-micron sieve. The sieve was washed with 95% non-denatured ethanol into a 50 mL Falcon tube.

Laboratory sorting and morphological identification:
Samples were poured over nested 250- and 63-micron sieves and sorted in dishes with 96% ethanol under a Leica S9i stereo microscope at magnifications up to 55x. Meroplankton larvae and benthopelagic invertebrates were transferred to a 6-well plate with 96% ethanol.

Morphological identifications followed the identifications set by the “EPR pump time series” dataset (986309; see Related Datasets section below). Individuals were identified to the lowest taxonomic level as morphotypes, with an emphasis on larval gastropods (Mills et al., 2009). Gastropods were identified by Susan Mills and Johanna Weston. All morphotypes of meroplankton larvae and benthopelagic invertebrates (e.g., amphipods) were enumerated and placed into taxon-specific containers. The presence or absence of “possibly benthic forams” was denoted. Holoplanktonic taxa (e.g., planktonic copepods and chaetognaths) were excluded from counts but denoted for their presence or absence.

A subset of 35 individuals spanning morphotypes were selected for the DNA barcoding of the cytochrome c oxidase subunit I (COI) region (target ~650 bp) using the HotShot method (Truett et al., 2000). Each individual was placed into a 0.5 mL PCR tube with 10 μL of HotShot lysis reagent (25 mM NaOH, 0.2 mM EDTA). Lysis reactions were incubated at 95°C for 30 min, then cooled to 4°C for 15 min. Subsequently, 10 μL of the neutralization reagent (40 mM Tris-HCl) was added, the mixture was spun down, and it was incubated at 4°C for 10 min. PCR amplification was performed as described by Meyer-Kaiser et al. (2025) with the COI primers jgLCO1490 [5’-TITCIACIAAYCAYAARGAYATTGG-3’] and jgHCO2198 [3’-TAIACYTCIGGRTGICCRAARAAYCA-5’] (Geller et al., 2013) and Promega Go Taq™ Master Mix. PCR products were visualized using a blueGel™ electrophoresis system (miniPCR). All products with banding were sent to Sequegen Inc. (Worcester, MA) for silica-glass purification and Sanger sequencing.

PCR products and clean sequences were generated from 15 individuals. Electropherograms were manually inspected and trimmed in MEGA 11 (Tamura et al., 2021). Any ambiguous base calls were denoted by 'N'. Sequences were translated to assess for the presence of stop codons. Sequences were compared against existing published sequences in the Barcode of Life Database (BOLD v5, Ratnasingham & Hebert, 2007) and in GenBank using the BLASTn algorithm (Camacho et al., 2009). Sequences with >97% identity were considered species-level matches, while those that matched >93% to published records were considered a genus-level match. Sequences with >80% match to published records were assigned to the family or superfamily level based on morphological identification and tree placement in BOLDv5. Genetic sequences are submitted to GenBank (accession numbers: PX280612-PX280626).

The scientificName determination came from the combination of the morphotype and genetic evidence, when available, to match the lowest-level identification to the World Register of Marine Species (WoRMS) taxonomic database. For this dataset, the verbatimIdentifications include the 64 morphotypes found across the “EPR pump time series” (https://www.bco-dmo.org/dataset/986309), allowing for direct relative abundance and community composition comparison. In addition to the shared 64 morphotypes, eight morphotypes were included. Six of the eight morphotypes represent lower taxonomic levels identified (e.g., amphipods to Ventiella sulfuris) or new morphotypes not seen in the pump dataset (e.g., an unknown gastropod bubble). The other two of eight morphotypes, chaetognaths and copepods, were denoted in this dataset as presence/absence.

Funding Acknowledgement
Support for this work and the generation of these data was provided by the U.S. National Science Foundation Division of Ocean Sciences; the Rinehart Initiative for “Access to the Sea” at Woods Hole Oceanographic Institution; and the Cecil H. and Ida M. Green Technology Innovation Award II Fund and the Alfred M. Zeien Endowed Fund for Innovative Ocean Research at Woods Hole Oceanographic Institution.