Inside Alphabet X’s new effort to combat climate change with seagrass
In late September, Bianca Bahman snorkeled above a seagrass meadow off the western coast of Flores, a scorpion-shaped volcanic island in eastern Indonesia. Bahman was flutter-kicked across the seabed and steered an underwater camera suspended from a pair small pontoons. The stereoscopic camera captures high resolution footage from two angles. This creates a three-dimensional map showing the ribbon-shaped seafloor leaves.
Bahman is a project manager for Tidal, whose team wants to use these cameras, along with computer vision and machine learning, to get a better understanding of life beneath the oceans. Tidal has used this camera system for years to monitor fish in aquafarms off Norway’s coast.
Now, MIT Technology Review can confirm that Tidal hopes its system will help preserve and restore seagrass beds around the world and accelerate efforts to harness the oceans’ ability to store and absorb more carbon dioxide.
Tidal is a project within Alphabet’s X division, the so-called moonshot factory. Its mission is to increase our knowledge of underwater ecosystems to inform and encourage efforts to protect the oceans from rising threats from overfishing, pollution, ocean acidification and global warming.
Its tools “can unlock areas which are desperately needed in ocean world,” Bahman states.
Studies indicate that the oceans could absorb a significant portion of the billions of tons of carbon dioxide that are needed to be removed from the atmosphere each and every year to maintain a stable temperature by midcentury. To make that happen, it will be necessary to restore coastal ecosystems, grow more seaweed, add nutrients to stimulate plankton growth and other similar interventions.
Tidal chose to initially focus on seagrass as it is a fast-growing plant and is particularly effective at absorbing carbon dioxide in shallow waters. These meadows could absorb more carbon dioxide from the ocean if they are expanded by communities, businesses, and nonprofits.
Scientists have a limited understanding of the role seagrass plays in climate regulation and the amount of carbon it sequesters. It will be difficult to track climate progress and create credible carbon credit markets that would pay for such practices without this knowledge and affordable methods to verify that restoration efforts actually store more carbon.
Tidal hopes to solve the problem by creating models and algorithms that convert the three-dimensional seagrass maps it captures into reliable estimates for the carbon below. Automated versions of Tidal’s data-harvesting technology may be able to provide the missing verification tool. This could help to kick-start and lend credibility marine-based carbon credit markets and projects, helping to restore ocean ecosystems.
A workbench within X’s building, where Tidal develops and tests its underwater camera systems.
The team envisions creating autonomous versions of its tools, possibly in the form of swimming robots equipped with its cameras, that can remotely monitor coastlines and estimate the growth or loss of biomass.
“If we can quantify these systems, then we can drive investment to preserve and conserve them,” says Neil Dave (general manager of Tidal). Some scientists are skeptical that Tidal’s technology can accurately estimate shifting carbon levels in faraway corners of the globe. Indeed, nature-based carbon credits have faced growing criticism: studies and reporting find that such efforts can overestimate climate benefits, create environmental risks, or present environmental justice concerns. Dave acknowledges that they aren’t sure how it will work yet. He says that the Tidal team, along with a group from Australia, went to Indonesia to find out.
Google launched what was then called Google X in early 2010, with a mandate to go after big, hard, even zany ideas that could produce the next Google. This research division took over Waymo’s self-driving car project. It created the Google Brain machine-learning tool, which powers YouTube recommendations, Google Translate, as well as many other core products of its parent company. And it gave the world the Google Glass augmented-reality headset (whether the world wanted it or not). Even teleportation and space elevators were briefly considered.
X has pursued climate-related initiatives from the beginning, but has had a mixed track record in this field to date.
It acquired Makani, an effort to capture wind energy from large, looping kites, but the company shut down in 2020. It also pursued a project to produce carbon-neutral fuels from seawater, dubbed Foghorn, but abandoned the effort after finding it’d be too hard to match the cost of gasoline.
The two official climate “graduates” still operating are Malta, a spinout that relies on molten salt to store energy for the grid in the form of heat, and Dandelion Energy, which taps into geothermal energy to heat and cool homes. Both are still relatively small and trying to gain traction in their respective markets.
After 12 years, X has yet to deliver a breakout success in climate or clean tech. It is unclear whether shifting strategies at X and current climate-related efforts such as Tidal will improve this track record.
Astro Teller was the head of X and said that the division initially “pushed hard for radical innovation”. It has been focusing more on the feasibility and not the “rigor dials” since then, he said.
The X climate projects were high-risk and heavy-tech projects that directly addressed energy technology and climate emissions. They also produced fuels and stored them in novel ways.
There are clear differences between the climate projects that X is currently pursuing. The two aside from Tidal are Mineral, which is using solar-panel-equipped robots and machine learning to improve agricultural practices, and Tapestry, which is developing ways to simulate, predict, and optimize the management of electricity grids.
X has created tools to help industries address environmental threats and ensure that ecosystems can thrive in a warmer, harsher environment. It’s also leaning heavily in to its parent company’s areas of strength, drawing on Alphabet’s robotics expertise as well as its ability to derive insights from massive amounts of data using artificial intelligence.
Such efforts might seem less transformative than, say, flying wind turbines–less moonshot, more enabling technology.
Teller admits that their new thinking may change the character of things at X today, but he argues against the suggestion that the problems it is pursuing aren’t as difficult, big or important as they were in the past.
“I don’t know that Tidal has to apologize for some sort of scope problem,” he says.
Humanity is destroying the oceans,” he says. “We must find a way to extract more value from the oceans for humanity while simultaneously regenerating them instead of continuing to deplete them.” We must find a way to automate the oceans .
A better protein source
Tidal, founded in 2018, grew out of informal conversations at X about the mounting threats to the oceans and the lack of knowledge required to address them, Dave says.
“The goal was too simple: save the oceans, save our world,” he said. “But it was based upon the understanding that the oceans were critical to humanity, and probably the most neglected and misused resource we possess .”
They decided to start by focusing on one application: aquaculture. This uses land-based tanks, protected bays, open ocean pens, and sheltered bays to raise fish, shellfish and seaweed. Today, these practices produce just over half the fish consumed by humans. They can be used more often to reduce overfishing, emissions from fishing fleets, as well as the environmental impact of trawling.
Tidal believed that it could offer tools that would enable aquafarmers to monitor and spot problems sooner, optimize their processes, and ensure faster growth and better health at a lower cost.
The researchers developed and tested a variety of prototypes for underwater camera systems. They also started training computer vision software that can recognize objects and attributes in footage. They started by using goldfish in a child’s pool to get the process off the ground.
For the past five years, they have been stress-testing their tools under the harsh North Sea conditions, in partnership with Mowi, a Norwegian seafood company.
During a Zoom call, Dave pulled up a black-and-white video of the chaos that ensues at feeding time, when salmon compete to gobble up the food dropped into the pen. The scene is difficult to see with the naked eye. The computer vision software tags each fish with tiny colored labels. It identifies individual fish swimming through the frame or captures them feeding.
Dave says fish farms can use this data in real-time, even in an automated manner. For example, they might stop dropping food in the pen when the fish stop eating. The software and cameras can also detect other important information, such as how heavy the fish are, their sexual maturity and any health issues. They can detect spinal deformities and bacterial infections. They can also detect the presence of parasites called sea lice. These parasites are often too small for the human eye. “We knew from the beginning that aquaculture would be us getting out of our comfort zone, so to speak,” Grace Young, Tidal’s scientific lead. “We knew it would be an stepping stone to working on other difficult problems .”
Tidal is confident that it has created one commercially viable application.
“Now is a big moment for us,” she adds, “because we’re able to see how the tools that we built can apply and make a difference in other ocean industries.”
Restoring our coasts
Seagrasses form thick meadows that can run thousands of miles along shallow coastlines, covering up to about 0.2% of the world’s ocean floors. They provide nutrients and habitat to marine populations, filter pollution, and protect coastlines.
The plants are photosynthetic and produce the food they require from sunlight, water and carbon dioxide dissolved into ocean waters. They store carbon in biomass and transport it to the seabed sediments. They also capture and bury carbon in organic matter that floats above.
Globally, seagrass beds may sequester as much as 8.5 billion tons of organic carbon in seafloor sediments and, to a much, much smaller degree, in their biomass. On the high end, these meadows draw down and store away about 110 million additional tons each year.
But estimates of the total range and carbon uptake rates of seagrass vary widely. The reason is that it is not possible to map all of the planet’s coastlines. Only about 60% of seagrass meadows have been surveyed in US waters, with “varying degrees of accuracy because of difficulties in remote sensing of underwater habitat,” according to a National Academies study.
Whatever their full expanse, though, we know they are shrinking. Overfishing, pollution, and development are all factors that destroy coastal ecosystems. These include carbon-sucking habitats such as mangrove forests or salt marshes. These shallow biological communities are being drained and excavated each year, releasing hundreds of millions of tons of CO2. Climate change is also making oceans more acidic and deeper, putting greater strains on many species.
Nations could help halt or reverse these trends by converting developed shorelines back into natural ones, actively managing and restoring wetlands and seagrass meadows, or planting them in new areas where they may do better as ocean levels rise.
Such work would however be extremely expensive. It is not clear who would pay for it, especially if it is at the cost of lucrative coastal development. The main possibility is that governments or companies could create market incentives to support conservation and restoration by granting credits for the carbon that mangroves, seagrasses, and salt marshes absorb and store away. Tens of billions of dollars’ worth of carbon credits are likely to be traded in voluntary markets in the coming decades, by some estimates.
The carbon market registry Verra has already developed a methodology for calculating the carbon credits earned through such work. At least one seagrass project has applied to earn credits: a long-running effort by the Nature Conservancy’s Virginia chapter to plant eelgrass around the Virginia Barrier Islands.
But some marine scientists and experts in carbon markets argue that it is necessary to have more stringent ways to ensure that these efforts are removing carbon as they claim.
We risk allowing people and businesses to purchase and sell carbon credits without actually helping the climate.
Tidal began exploring whether its tools could be used for seagrass late last year, as a growing body of studies underscored the need for carbon removal and highlighted the potential role of ocean-based approaches.
“We started to double-click and read a lot of studies,” Dave says. “And we found out, ‘Wow. We do have some technology that could be applied here ‘,” Dave says.
Seagrass is difficult to map at large scales due to the difficulty of distinguishing from other dark spots in shallow water. Andy Steven, a marine scientist, oversees coastal research at CSIRO.
Researchers with the Commonwealth Scientific and Industrial Research Organisation pull up and examine seagrass and sediments from research plots.
“The world needs to move to being able to map and then measure change on a far more frequent basis,” Steven says. “I see Tidal technology as part of an arsenal that helps us quickly survey, process and deliver information to decision-makers in the timeframes that we need.” It is addressing a really fundamental issue.”
CSIRO agreed to help Tidal test how well its system works. They worked together on an earlier field trial off Fiji’s coast this summer, and the subsequent experiment in September in Indonesia. The country’s thousands upon thousands of islands is home to one of the largest and most diverse seagrass meadows in the world.
Tidal used an off-the shelf autonomous underwater vehicle with a basic camera to pair its software. Their general approach would be easier to share if they could scan meadows with standard hardware.
It didn’t work. The seagrass was higher than expected and the tides were lower. Bahman states that the thruster and rudder quickly became clogged with seaweed, causing the team to stop every few seconds.
After a brainstorming session, the Tidal group decided to take its own camera system and turn it upside down on a floating platform that could be pulled by a boat. The Hammersled has fins to keep it straight, and a set ropes and cleats to allow researchers to dip the camera further into the water.
Tidal’s researchers test out the “Hammersled” at a pool in the middle of Alphabet’s campus in Sunnyvale, California, by pulling it over patches of plastic seagrass.
The system worked well enough during a few tests in a large pool in the middle of Alphabet’s campus in Sunnyvale, California, where team members pulled it by hand over patches of plastic seagrass on the bottom. The bigger test is whether Tidal can translate the maps into an accurate estimate for the carbon seagrass holds or buries in the seafloor.
‘We’ve got it’
After Steven and his colleagues arrived in Labuan Bajo, on the western tip of Flores, they rented a 14-cabin liveaboard, the Sea Safari VII, and began sailing around the islands. They launched surveillance drones from the deck in search of promising seagrass beds. This was to train Tidal’s algorithms and models for the wide variation that occurs in nature.
Once the CSIRO researchers selected, measured, tagged, filmed, and photographed their 100-meter transects, the Tidal team passed through.
They pulled the Hammersled along with a small Indonesian fishing boat. Bahman, Hector Yee, a software engineer, and other staffers took turns jumping in the water with flippers and goggles to hold a pontoon. They then kept the camera straight as they crisscrossed around the test area.
After the process was completed, the CSIRO researchers used spades and peat borers to remove the seagrass and deep sediments.
Back on the main island, the Australian scientists used makeshift ovens, including some created from hair dryers, to dry out the plant materials and sediments. They ground them up and placed them in hundreds of plastic bags. Each bag was marked with different depths and locations.
In the months ahead, they will analyze each batch of carbon at their labs in Adelaide and determine the total amount in each plot.
“If our algorithm takes a look at the data we gathered before they took the core samples and comes up with the same answer, then we’ve got it,” says Terry Smith, a solutions engineer with Tidal.
Not everyone, however, is convinced that seagrass is a particularly promising path for carbon removal, or one whose climate benefits we’ll be able to accurately assess.
Among the suite of approaches to carbon removal that the National Academies has explored in its studies, those focusing on coastal ecosystems rank near the bottom in terms of the potential to scale them up. These ecosystems are limited in their ability to exist within narrow bands along shorelines and face intense competition from human activity.
We need to do everything possible to preserve seagrass,” states Isaac Santos, a professor of maritime biogeochemistry at University of Gothenburg, Sweden. This is because of the important roles these plants play in protecting coastlines and marine biodiversity. But the big question is: Are they going to save us all from climate change?” He says. “They don’t have enough area to sequester enough carbon to make a big impact.”
Accurately determining the net carbon and climate impact from seagrass restoration is also problematic, as studies have highlighted.
Carbon sequestration varies dramatically in these coastal meadows, depending on the location, the season, the mix of species, and how much gets gobbled up by fish and other marine creatures. The carbon in seafloor sediments may also leak into the surrounding waterways, where it is dissolved and remains in the ocean for many millennia. Some may escape into the atmosphere. Additionally, coastal ecosystems also produce methane, nitrous oxide, and other potent greenhouse gases. These should be considered when estimating the overall climate impact.
Finally, the vast majority of carbon in seagrass beds is buried beneath the seafloor and not in the plant material Tidal intends.
“And we also know that the correlation between biomass and sediment carbon is not straight forward,” Santos said in an email. “So, any approach that is based solely on biomass will need all sorts of validations,” Santos stated in an email. This is to ensure that it provides accurate estimates of stored carbon.
An essay in The Conversation late last month highlighted another concern: environmental justice. Sonja Klinsky of Arizona State University, and Terre Satterfield of University of British Columbia were the authors. They stressed that local communities should have a lot of say in such projects. Some coastal communities may not wish to see their active harbor transformed into a salt marsh.
“Much of the global population lives near the ocean,” they wrote, and some interventions “might impinge on places that support jobs and communities” and provide significant amounts of food.
Unlocking the secrets
Addressing the scientific questions will require better understanding of coastline ecosystems. Steven from CSIRO says he hopes Tidal’s technology can make it easier to conduct the necessary research. He says, “It’s definitely a challenge.” “But you have to start somewhere
” Tidal emphasizes that nature-based carbon removal strategies can provide multiple benefits for local communities and natural ecosystems. They could, for example, help sustain fishery population. Tidal also works with CSIRO in Fiji and Indonesia to train local communities, including students from universities, to participate in carbon markets.
” “Ultimately, our vision for these communities is to give them tools to be in a position to manage, protect and repopulate local systems locally,” Dave stated in an email.
So what’s next for Tidal It will still take months for the Australian team’s analysis of the seagrasses and sediments. The teams will continue field experiments to refine their models and algorithms, and ensure accurate carbon estimates for a variety seagrass types and different conditions.
For instance, Tidal may look to partner with other research groups focused on the Bahamas, another major seagrass region. Tidal believes that if it works, other ocean-based methods of carbon removal could be supported by its suite of tools, such as the restoration and growth of mangrove forests and seaweed.
Dave says he can envision a variety of potential business models, including providing carbon measurement, reporting, and verification as a service to offsets registries or organizations carrying out restoration work. They may also develop autonomous robotic systems that can plant seagrass without human intervention.
Even though the systems aren’t reliable enough to estimate carbon, Tidal believes that their efforts will still support scientific efforts to understand important ocean ecosystems and international efforts to protect them. Dave suggests that this could include monitoring coral reef health, which is gravely endangered by rising sea levels.
It may not sound like a moonshot, but it is the original concept of X. It’s not a space elevator.
Tidal may have found a new way of solving difficult problems by creating tools that can be used by many organizations.
I’m a journalist who specializes in investigative reporting and writing. I have written for the New York Times and other publications.