Scientists aboard a research vessel off the coast of Brazil identified 31 previously unknown species during just two weeks of deep-sea exploration, drawing them from the ocean’s midwater zone, a vast habitat that remains one of the least studied environments on Earth. The Monterey Bay Aquarium Research Institute, known as MBARI, supplied laser imaging systems and artificial intelligence tools that allowed the team to document organisms in place rather than hauling fragile specimens to the surface. The discoveries add fresh urgency to questions about how much life the mesopelagic layer holds and what stands to be lost before science can catalog it.
Why 31 new midwater species change the deep-sea calculus
The mesopelagic zone, roughly 200 to 1,000 meters below the surface, contains what marine biologists often describe as the largest animal migration on the planet. Every night, billions of organisms rise toward the surface to feed, then sink back into darkness. Yet taxonomic knowledge of this layer lags far behind shallower coral reefs or even the abyssal seafloor, where remotely operated vehicles have spent decades collecting samples. Pulling 31 new species from a single two-week cruise suggests the true species count in midwater habitats could be dramatically higher than current estimates.
The speed of these identifications matters as much as the number. Traditional deep-sea taxonomy requires physical specimens, months of lab work, and comparison against museum collections. On this expedition, MBARI’s technology team deployed laser imaging and AI-driven analysis that compressed that timeline from years to days. If the same toolkit can be replicated on future cruises, the proportion of midwater organisms identified to species level during a single voyage could rise sharply. A reasonable projection, based on the jump from near-zero real-time species-level IDs to 31 in one outing, is that combining these data streams with improved machine-learning models could push single-cruise identification rates from under 10 percent to well above 40 percent within a few years. That projection depends on continued training of the algorithms and broader deployment of the hardware, neither of which is guaranteed.
Laser sheets, 3D cameras, and AI models behind the count
Three technologies worked in tandem to produce the 31-species tally. The first, called DeepPIV, is a laser-and-optics system that projects a thin illuminated sheet into the water column. That sheet lets researchers quantify fluid motion and particle movement around an organism, and it is especially effective at revealing gelatinous structures in jellyfish and other soft-bodied animals that would collapse or disintegrate if brought to the surface in a net. By recording how water flows through and around these animals, DeepPIV generates morphological data that physical collection often destroys.
The second instrument, EyeRIS, is a 3D light-field camera system designed to capture quantitative three-dimensional visual data about the structure and movement of deep-sea organisms in situ. Where a standard underwater camera produces flat images, EyeRIS records depth information across its entire field of view. That depth data lets taxonomists measure body proportions, fin shapes, and tentacle arrangements without ever touching the animal. For species that exist only as single observed individuals, digital morphology from EyeRIS may be the only measurement record available.
The third component is FathomNet, an open underwater image training database whose seed data comes from MBARI’s VARS video-annotation system, according to a technical preprint authored by MBARI researchers. FathomNet includes midwater classes and draws on external imagery sources, giving its machine-learning models a growing reference library against which new footage can be compared. During the Brazil expedition, AI tooling built on this database contributed to rapid identification of the 31 new species by flagging organisms whose features did not match any known entry. That real-time screening allowed scientists to focus their limited dive time on the most promising unknowns rather than re-documenting familiar animals.
Stanford’s Prakash Lab also contributed to the expedition. The lab’s work on squid imaging and related optical methods added to the overall species count, though the precise number of species attributable to each team’s instruments has not been broken out in available records. What is clear is that combining multiple, independently developed systems yielded a richer picture of midwater biodiversity than any single platform could have provided on its own.
What the expedition left unanswered about midwater biodiversity
For all the speed of the new identifications, several gaps remain. No public source from the expedition lists the 31 species by name, voucher number, or depth range. Without formal taxonomic descriptions published in peer-reviewed journals, the “31 new species” figure reflects field-level assessments that still need confirmation through established scientific review. History shows that initial field counts sometimes shrink when closer lab analysis reveals that two apparent species are actually life stages of the same organism, or expand when genetic sequencing splits what looked like one species into several.
The expedition also raises questions about how many of the 31 organisms were documented only digitally versus collected as physical specimens. For gelatinous animals in particular, imaging-based descriptions may be the only feasible option, because nets and bottles can shred their tissues beyond recognition. That reality forces taxonomists to grapple with whether high-resolution 3D imagery and behavioral footage can stand in for traditional type specimens, or whether new standards are needed for naming species that cannot survive capture.
Another unknown is how representative the Brazil survey area is of the broader South Atlantic midwater community. The cruise covered a limited set of transects and depth bands, constrained by ship time and weather. If 31 candidate species emerged from that narrow slice, the implication is that many more undescribed organisms inhabit adjacent basins and under-sampled regions. But until similar campaigns deploy comparable instruments in other oceans, it is difficult to extrapolate from one expedition to global biodiversity estimates.
There are also open questions about ecological roles. The imaging systems documented body shapes, swimming behaviors, and in some cases predator-prey interactions, but they could not, on their own, reveal diet, reproductive strategies, or population sizes. Without that context, it is hard to assess how vulnerable these newly observed species might be to climate-driven changes in oxygen levels, temperature, and food supply in the mesopelagic zone.
Implications for conservation and future surveys
Despite those uncertainties, the Brazil expedition underscores how rapidly perceptions of deep-sea diversity can shift once new tools are applied. If imaging and AI systems continue to accelerate the pace of discovery, policymakers may find themselves confronted with evidence that midwater ecosystems are far richer and more specialized than previously assumed. That, in turn, could influence debates over activities such as deep-sea mining, high-seas fishing, and carbon sequestration projects that target the ocean interior.
For researchers, the immediate implication is methodological. The combination of laser-based flow visualization, volumetric cameras, and machine-learning classifiers offers a template for future cruises. Standardizing these approaches could make it easier to compare data across regions and years, turning what are now isolated case studies into a coherent global survey of the midwater biome. Training AI models on a broader range of taxa and environmental conditions would also reduce the risk of misclassification as more expeditions contribute imagery to shared databases.
At the same time, the expedition highlights the need for careful calibration between speed and certainty. Rapid, AI-assisted identifications can guide in-the-moment decisions about where to point cameras and which organisms to prioritize, but formal species descriptions still require painstaking work. Balancing those two tempos-real-time triage in the field and slower, rigorous taxonomy in the lab-will determine how robust the next generation of deep-sea species lists ultimately becomes.
For now, the 31 candidate species recorded off Brazil serve as a reminder that the dark, midwater expanses between surface and seafloor are not empty corridors but densely populated habitats. Each new organism brought into focus by laser sheets and 3D cameras expands the known boundaries of marine life and complicates any assumption that human impacts on the deep ocean will be limited or easily reversible. As more expeditions adopt similar technologies, the picture of what lives in the planet’s largest, least visible ecosystem is likely to grow both richer and more fragile at the same time.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
