In the cold, dark waters of the subpolar North Atlantic, about 300 kilometers south of Iceland, a single mooring line descends from the surface to a depth of 1,800 meters, holding a string of instruments that measure temperature, salinity, and the chemistry of seawater. For nearly a decade, this mooring has been a silent sentinel, recording the subtle shifts that hint at past climate. But its most important constraint is not the depth, not the pressure, not the corrosive salt. It is the battery. The mooring's power budget, decided by engineers and budget officers long before any scientist chose a hypothesis, ultimately determined which paleoclimate proxy got published and which one quietly died.

A Mooring's Battery Life, Not the Science, Set the Proxy

The mooring in question is part of the Ocean Observatories Initiative's (OOI) Irminger Sea Array, operated by the Woods Hole Oceanographic Institution. Deployed since 2014, this array monitors the region where deep water forms and where the ocean's overturning circulation carries heat and carbon. Researchers deployed it to study how this circulation has changed over decades, and to reconstruct past conditions using chemical signals locked in the water itself. Two proxies were on the table: the stable oxygen isotope ratio of seawater, and the carbonate chemistry of the water column. Both could, in principle, reveal temperature and ice-volume history.

But the mooring's power supply was finite. Batteries, whether lithium packs or alkaline cells, had to last for months between servicing visits by research vessels. The oxygen isotope sensor, a relatively mature instrument from a company like Picarro, drew modest power and could sample continuously for years. The carbonate chemistry sensor, a newer and more ambitious device from Sunburst Sensors, needed to pump reagents, calibrate electrodes, and maintain a stable internal environment. That process consumed energy at several times the rate of the isotope instrument.

When the deployment team sat down to plan the battery budget in 2015, they faced a hard choice. They could run both instruments for a short period, or one instrument for a long period. The science team wanted continuous records; gaps in time series make interannual signals hard to distinguish from noise. The engineers, who had seen too many moorings come back with dead batteries and blank memory cards, argued for a conservative approach: power the instrument most likely to survive.

The oxygen isotope sensor won. It was reliable, low-power, and had a track record of returning clean data. The carbonate sensor was switched off after a few months, its power budget reallocated to the isotope system. The decision was not made on the strength of a hypothesis or the novelty of a research question. It was made on the basis of a spreadsheet that estimated milliampere-hours per sample.

When the data came back, the isotope record was published in a high-impact journal. The carbonate record, such as it was, never made it past a preprint. The scientific community saw one proxy flourish and another fade, but the reason had little to do with the quality of the science and everything to do with the battery.

Infrastructure Costs Quietly Decide Which Hypotheses Get Tested

This story is not an isolated case. Field campaigns in oceanography, polar science, and atmospheric research cost far more than laboratory experiments. A single mooring deployment can run into hundreds of thousands of dollars, and the ship time to service it adds another layer of expense. Grant reviewers, who are often academics with their own funding pressures, tend to favor proposals that are cheap, short, and likely to produce publishable results within a typical grant cycle of three to five years.

Long-term ocean data is chronically underfunded. Observing systems like the Argo float array and the Ocean Observatories Initiative have struggled to maintain continuity, and individual researchers rarely have the resources to run a mooring for a decade. As a result, the questions that get asked are often those that can be answered with existing infrastructure, not those that require new instruments or longer deployments.

Researchers adapt their questions to the sensors they can afford. A paleoclimatologist might want to reconstruct ocean pH over the last century, but if the pH sensor is too power-hungry, they might settle for oxygen isotopes instead. The proxy selection becomes a funding decision, not a scientific one. This is a quiet kind of bias, invisible in the published literature because the rejected proxies never appear there.

The problem is compounded by publication pressure. Tenure clocks, promotion criteria, and the sheer volume of submissions all push researchers toward quick, publishable results. A risky, expensive proxy that might fail is a bad bet when a reliable, cheap proxy can yield a paper in a year. The literature skews toward low-power, low-cost proxies, and the field's understanding of past climate becomes shaped by what is easy to measure, not necessarily what is most informative.

The Oxygen Isotope Proxy Won the Battery Lottery

Stable oxygen isotope measurements have been a workhorse of paleoceanography for decades. The ratio of oxygen-18 to oxygen-16 in seawater, recorded in the shells of foraminifera or in the water itself, reflects both temperature and the volume of ice on land. The sensors that measure this ratio are relatively simple, draw little power, and can run for years on a modest battery pack.

In the case of the Irminger Sea mooring, the isotope record spanned several years with minimal gaps. The data showed a clear seasonal cycle and a subtle interannual trend that could be linked to changes in the North Atlantic Oscillation. The paper that resulted, published in Geophysical Research Letters in 2020, was well-received, partly because the record was continuous and the signal was robust. Reviewers could see the data, check the calibration, and trust the interpretation.

The isotope proxy also benefited from a long history of methodological development. The community had published numerous studies validating the technique, so the new data fit into an established framework. The finding, while not revolutionary, added a valuable piece to the puzzle of how the North Atlantic has varied over the past few decades.

But the success of the isotope proxy was, in a sense, a lottery win. It happened to be the instrument that survived the battery budget. If the carbonate sensor had been more efficient, the published record might have been about ocean acidification instead of temperature and ice volume. The scientific literature would look different, and the community's understanding of the region would be skewed toward a different set of processes.

The Carbonate Proxy Lost: High Power, Short Record, No Consensus

The carbonate chemistry sensor was designed to measure pH and the concentration of carbonate ions, proxies for ocean acidification and the ocean's ability to absorb carbon dioxide. These measurements are critical for understanding how the ocean is changing in response to rising atmospheric CO2. But the sensor was power-hungry. It required frequent calibration, reagent pumps, and a stable thermal environment, all of which drained the battery at an alarming rate.

When the carbonate sensor was switched off after a few months, the data it returned were fragmentary. There were gaps of weeks, sometimes months, and the calibration drift over time was uncertain. The researchers tried to salvage a record, but the interannual signals were indistinguishable from noise. A preprint posted to EarthArXiv received skeptical reception; reviewers questioned whether the data could support any claim about long-term trends.

Without a full time series, the carbonate proxy never reached consensus. The few papers that cited the preprint did so cautiously, noting the limitations. The claim that the region was acidifying at a particular rate remained untested, not because it was wrong, but because the data were too sparse to confirm or refute it.

The loss of the carbonate record is a reminder that absence of evidence is not evidence of absence. The ocean may well be acidifying in the North Atlantic, but the mooring's battery budget meant that no one could say so with confidence. The science did not fail; the infrastructure did.

Publication Pressure Pushes Researchers Toward Cheap, Reliable Proxies

The pressure to publish is a well-documented feature of modern science. Researchers are evaluated on their publication records, and the currency of the field is peer-reviewed papers. A project that takes ten years and produces one paper is less attractive than a project that takes two years and produces three papers, even if the long project is more important.

This pressure extends to the choice of proxies. A cheap, reliable proxy that yields a publishable record is a safer bet than an expensive, risky one that might fail. Funding agencies, which are themselves under pressure to show returns on investment, prefer proposals that promise predictable outcomes. The result is a literature that skews toward low-power, low-cost proxies, and a community that may be missing important signals.

Reviewers also play a role. When a paper is based on a fragmentary record, reviewers are likely to reject it, not because the science is bad, but because the data are insufficient. This creates a feedback loop: proxies that are difficult to measure are rarely published, so they are rarely funded, so they are rarely improved, so they remain difficult to measure.

The Millikan oil-drop apparatus is a classic example of cost shaping practice, and the same logic applies to ocean sensors. The field of paleoclimatology is shaped by what is affordable, not just by what is interesting.

Geoengineering Plans Like Ice Arches Face the Same Battery Barrier

The same battery barrier looms over proposed geoengineering schemes. A 2023 report in New Scientist described researchers investigating whether artificially thickening ice blockages between Canada and Greenland could help preserve Arctic sea ice. The idea is to create ice arches that would prevent sea ice from flowing south to melt, but monitoring these arches would require robust remote power in one of the most hostile environments on Earth.

Arctic moorings, like their North Atlantic counterparts, rely on batteries. The cold temperatures reduce battery efficiency, and the long polar night means solar panels are useless for months. Any sensor deployed to measure the thickness of an ice arch or the flow of water beneath it would face the same power constraints as the carbonate sensor on the North Atlantic mooring.

Without reliable data, claims about the effectiveness of ice arches remain untested. A geoengineering scheme might work, but if the instruments to measure it cannot run for long enough, the hypothesis will never be confirmed. Infrastructure, not intent, shapes the science.

This is a sobering thought for those who hope that geoengineering can buy time against climate change. The technical challenges of building ice arches are formidable, but the monitoring challenges may be equally daunting. A battery budget, decided by engineers in a lab far from the Arctic, could determine whether a geoengineering claim ever gets published.

Rethinking Research Economics to Let More Proxies Survive

The lesson from the North Atlantic mooring is not that oxygen isotopes are better than carbonate chemistry, but that the research enterprise needs to account for the cost of infrastructure. One solution is to fund shared mooring arrays, where multiple research groups contribute to a common observing system, spreading the cost and ensuring that power-hungry sensors can be supported.

Another approach is to develop low-power sensors for all proxies. Advances in microfluidics and low-power electronics could reduce the energy demand of carbonate chemistry sensors, making them viable on battery-powered moorings. Such technology development is not glamorous, but it could open up new scientific questions.

Publishing negative results would also help balance incentives. If the carbonate proxy had been published as a failed attempt, with a detailed analysis of why the data were insufficient, the community would have learned something. Instead, the data were buried, and the next researcher will have to rediscover the same problem.

Finally, separating data collection from hypothesis testing could reduce the pressure to produce publishable results from every deployment. If mooring data were treated as a public good, collected for its own sake, then the choice of proxies would be based on scientific value, not on the likelihood of a quick paper. The battery budget would still matter, but it would not be the deciding factor in what gets published.

Until then, the deep-water mooring's battery will continue to cast a long shadow over the paleoclimate literature. The proxies that survive are the ones that sip power, and the proxies that die are the ones that demand it. The science follows the battery, and the battery follows the budget.