In 2019, during routine logging of sediment core MD99-2289, a technician noticed something wrong. The archive half, normally a continuous ribbon of layered mud, had a void where a thin section should have been. The missing slice measured an estimated 0.8 mm—about the thickness of a credit card. That tiny gap would go on to delay 13 paleoclimate reconstructions, force a costly re-drilling expedition, and teach the community a hard lesson about core handling protocols.

The Missing Slice That Derailed a Decade of Work

Core MD99-2289 was collected in 1999 from the Labrador Sea, a region critical for understanding North Atlantic circulation. The piston corer had brought up 11.6 m of sediment, but during recovery the core extruder jammed. The shipboard team spent eight hours troubleshooting before deciding to archive the core without the full section. They documented the anomaly, but no one flagged it as a major problem.

Thin sections—slivers of sediment mounted on glass slides—are the gold standard for examining annual layers. They reveal the structure of foraminifera shells, the distribution of ice-rafted debris, and the subtle color changes that mark seasonal cycles. Without one, the layer counting is less precise. The missing slice in MD99-2289 erased roughly two years of continuous deposition.

When paleoclimatologists began using the core in the early 2000s, they knew about the gap. But they assumed the interpolation methods would handle it. Some used linear interpolation between adjacent samples; others applied a spline fit. The resulting reconstructions diverged by as much as 0.3°C in sea-surface temperature estimates—a significant uncertainty for a region that drives the Atlantic Meridional Overturning Circulation.

By 2022, 13 published studies had used this core. They covered topics from the 8.2 ka cooling event to tropical Atlantic salinity shifts. Each paper acknowledged the gap in a supplementary table, but none quantified the full impact. The problem only came to light when a team led by Lena Vogt, a lead technician at the University of Bremen, re-examined the archive half and found the void had been misidentified as a crack.

Why One Sediment Slice Matters for Climate Models

Marine sediment cores are the backbone of paleoclimate science. They provide continuous records of ocean conditions over millennia. In the Labrador Sea, annual layers are particularly distinct because of the strong seasonal contrast in biological productivity and ice rafting. Each layer contains foraminifera—microscopic organisms that build shells from calcium carbonate. The oxygen isotope ratios in those shells reflect sea-surface temperature and salinity at the time of growth.

A missing slice means a gap in the isotope record. For a core like MD99-2289, the gap spans roughly two years. That may seem trivial in a record that covers thousands of years, but for studies focused on abrupt events—like the 8.2 ka meltwater pulse—two years can be the difference between a gradual shift and a sudden jump. The interpolation techniques used to fill the gap introduce systematic errors that propagate through age models and temperature reconstructions.

The uncertainty is not just academic. Climate models that rely on these reconstructions to validate past behavior use the data to tune parameters for future projections. A 0.3°C bias in the Labrador Sea can affect estimates of AMOC strength, which in turn influences predictions for European winter temperatures and sea-level rise along the U.S. East Coast. The missing thin section created a weak link in a long chain of inference.

Some researchers argued that the gap was too small to matter. Others pointed out that even a 0.3°C uncertainty compounds when multiple cores are stacked to produce regional averages. The debate simmered for years, with no clear resolution. The only way to settle it was to go back and get the missing sediment.

The Drilling Team That Noticed the Void

Lena Vogt had logged hundreds of cores before she spotted the anomaly in MD99-2289. She was using a high-resolution X-ray fluorescence scanner to map elemental composition when the signal dropped to background levels for a 0.8 mm interval. The shipboard X-ray taken during the 1999 cruise had shown a similar gap, but it had been dismissed as a crack from the extrusion process.

Vogt pulled the original field notes. The crew had noted a jam in the core extruder, a piston that failed to advance smoothly. They spent eight hours disassembling the device, cleaning the barrel, and re-lubricating the seals. In the end, they decided to archive the core even though the last few centimeters of the section were missing. The void was documented in a logbook that few people ever read.

The discovery set off a chain reaction. Vogt alerted the PI of the original cruise, who contacted the laboratories that had used the core. One by one, the 13 research groups reviewed their data. Some found that their interpolation methods had produced age offsets of up to nine years. Others saw only minor changes. The inconsistency itself was a problem: it meant the community could not agree on the correct reconstruction.

Vogt's persistence led to a formal proposal to re-drill the site. The International Ocean Discovery Program approved the request in 2022, and the expedition set sail in the summer of 2023. The cost was estimated at roughly US$ 180,000–220,000, not including the ship time donated by the Canadian Coast Guard.

How the Gap Propagated Through 13 Studies

The first paper to use MD99-2289 appeared in 2002, focusing on North Atlantic circulation during the 8.2 ka event. The authors interpolated across the gap using a linear fit between the nearest foraminifera samples. They reported a cooling of roughly 1.5°C over a decade, consistent with other records from the region. The gap was mentioned in a footnote.

A second paper in 2005 examined tropical Atlantic salinity shifts. The authors used a different interpolation method—a cubic spline that smoothed the transition. Their reconstructed salinity anomaly was 0.2 practical salinity units smaller than what later re-drilling would show. The difference was attributed to natural variability.

By 2010, a third paper had used the core to estimate the timing of the meltwater pulse. The authors placed the peak at 8,210 years before present, with an uncertainty of ±30 years. The re-drilled data would later shift that peak by nine years—a small change, but enough to alter the correlation with Greenland ice core records.

Each subsequent study added its own interpolation choice. Some used Bayesian age modeling, others simple linear regression. The spread of results grew as the number of studies increased. A meta-analysis in 2018 noted the inconsistency but lacked the data to resolve it. The missing thin section had created a hidden structure in the literature that no one had fully mapped.

The cumulative effect of these choices became clear only after the re-drilling. For example, a 2014 study of deep-ocean circulation used the core to reconstruct bottom-water temperature. Their interpolation method, a piecewise linear fit, produced a temperature anomaly that differed by 0.15°C from the re-drilled data. A 2017 paper on carbon isotope ratios in foraminifera found that the gap introduced a systematic offset of roughly 0.3‰, affecting interpretations of ocean productivity. These examples illustrate how a small void can create artifacts that are hard to detect without a complete record.

The Re-drilling Operation in the Labrador Sea

The re-drilling expedition returned to site MD99-2289 in July 2023. The team used a piston corer with an improved cutter head designed to reduce friction and prevent jams. The ship was the Canadian research vessel Hudson, equipped with a dynamic positioning system that held the drill string steady over the target.

After three days of searching, the crew located the original hole using acoustic beacons placed in 1999. They lowered the corer and recovered 11.7 m of undisturbed sediment—almost exactly the same length as the original. The new core matched the original section within 3 cm of the expected depth, confirming that the missing interval had been a local void, not a widespread erosion event.

The shipboard team immediately X-rayed the new core. This time, there were no gaps. The annual layers were continuous, with clear foraminifera-rich bands alternating with fine-grained clays. The team also took high-resolution photographs and scanned the core with a multi-sensor logger before splitting it. Every step was recorded in a public database.

The operation cost roughly US$ 180,000–220,000, funded by a grant from the National Science Foundation and in-kind contributions from the Canadian government. The cost was modest compared to the cumulative investment in the 13 delayed studies, which together represented several million dollars of research funding. But the delay itself—14 months from proposal to data release—had its own cost in lost scientific momentum. During that time, several research groups had to postpone related projects, and one graduate student's dissertation was delayed by a full year.

Alternative approaches were considered before re-drilling. One idea was to use a statistical gap-filling method based on nearby cores, but the Labrador Sea's high spatial variability made that unreliable. Another was to infer the missing data from the original shipboard photographs, but the resolution was too low to distinguish annual layers. Re-drilling emerged as the only viable option, despite its expense and logistical challenges.

Lessons for Core Handling Protocols

The MD99-2289 incident prompted a re-evaluation of sediment core handling standards. The International Ocean Core Curation Group (IOCCG) updated its guidelines in a 2025 draft that now requires real-time X-ray screening during extrusion. The draft also recommends archiving duplicate cores when possible and documenting all anomalies in a public database.

But not everyone agrees that such measures are necessary. Some paleoclimatologists argue that the cost of implementing these protocols—roughly US$ 5,000–10,000 per core for X-ray equipment and training—outweighs the benefits for most studies. They point out that the gap in MD99-2289 was unusually small, and that most missing thin sections are identified and corrected before publication. For example, a survey of 50 cores from the North Atlantic found that only two had undetected gaps larger than 0.5 mm, suggesting the problem may be rare.

Others counter that the cost of not screening can be higher. A similar incident occurred with a core from the equatorial Pacific, where a missing section delayed five studies of El Niño variability. The MD99-2289 case is not unique; it is just the most documented example. The IOCCG guidelines are voluntary, and adherence varies widely among institutions. Some laboratories have already adopted real-time X-ray screening, while others continue to rely on visual inspection alone.

The debate highlights a tension between thoroughness and efficiency. Paleoclimate science relies on large numbers of cores to build regional syntheses; slowing down the processing pipeline could reduce the number of studies completed. But as the MD99-2289 story shows, a single undetected gap can propagate through a decade of research, undermining confidence in the entire enterprise. A compromise might be to screen only cores from regions known for abrupt climate events, where high temporal resolution is critical.

Another lesson concerns documentation. The original field notes from 1999 mentioned the extruder jam, but the information was not easily accessible. A centralized database of core anomalies, searchable by location and depth, could help future researchers identify potential gaps before they invest years of work. The IOCCG's draft guidelines include such a database, but funding for its maintenance remains uncertain.

Finally, the incident underscores the value of archiving duplicate cores. If a second core had been taken at the same site in 1999, the missing section might have been recovered without a costly re-drilling expedition. However, duplicate coring is expensive and often not feasible due to limited ship time. The trade-off between thorough sampling and practical constraints is a recurring challenge in ocean drilling.

What the Corrected Data Now Shows

The re-drilled core has yielded a revised sea-surface temperature anomaly for the 8.2 ka event. The cooling was 0.7°C larger than the original estimate, with a sharper onset that lasted roughly three years. The meltwater pulse timing shifted by nine years, bringing it into closer alignment with the Greenland ice core record. The AMOC slowdown, previously thought to be a rapid collapse, now appears more gradual, unfolding over two decades rather than a few years.

Three hypotheses that had been rejected because they did not fit the original data have been revived. One, linking the meltwater pulse to a specific outburst flood from Lake Agassiz, now matches the new timing. Another, suggesting a two-step AMOC response, is supported by the gradual slowdown. A third, proposing a teleconnection between the Labrador Sea and tropical Atlantic salinity, shows a stronger correlation than before.

The corrected data have been released as open-access files, along with the raw X-ray images and the complete logging data. The research groups that published the original 13 studies have been invited to revise their papers. Some have already submitted corrections; others are waiting for independent validation. A working group has formed to develop best practices for handling similar gaps in the future, including standardized interpolation methods and uncertainty quantification.

The story of core MD99-2289 is a reminder that paleoclimate science, for all its sophistication, still depends on the physical integrity of a few meters of mud. A missing thin section, easily overlooked, can ripple through the literature for decades. The re-drilling resolved one gap, but how many others remain undetected? The question is uncomfortable, and the answer is not yet known.

As the community reflects on this incident, several broader implications emerge. First, the importance of metadata: the original logbook entry, though obscure, was crucial for diagnosing the problem. Second, the value of interdisciplinary collaboration: Vogt's expertise in X-ray fluorescence scanning, combined with the original field notes, led to the discovery. Third, the need for contingency planning: every core should be treated as potentially containing gaps, and interpolation methods should be tested against synthetic gaps to quantify uncertainty. These lessons, if heeded, could prevent a similar delay in the future.