In early 2020, a university lab in Honolulu faced a mundane administrative problem: a five-year National Science Foundation grant was ending, and a renewal application had been declined. The lab operated a small network of tide gauges across the Pacific, instruments that had been recording sea level for decades. Each gauge cost roughly $15,000 to install and about $10,000–$20,000 per year to maintain—covering rental of the pier space, data transmission, and occasional repairs. When the grant lapsed, the lab had to cut four gauges. Two of them, at Honolulu Harbor and Hilo Bay, had been running continuously since 1905 and 1946, respectively. Their removal created a gap in the longest sea level records in the central Pacific. Within two years, that gap had propagated into published acceleration curves, altering the slope of global sea level rise estimates by a small but detectable amount.

The Fee That Bent the Curve

The two gauges cost roughly $30,000 per year combined—a trivial sum in the context of climate research, which globally spends billions annually. But the lab's funding had been structured as a series of competitive grants, each lasting three to five years. When the NSF declined the renewal, the university did not have a dedicated line item to absorb the cost. The gauges were decommissioned in mid-2020.

Data from those two stations fed into several global sea level databases used by the Intergovernmental Panel on Climate Change and by independent research groups. The records were among the longest continuous tide gauge series in the Pacific, and they had been used to calculate regional and global sea level acceleration—the rate at which the rate of rise is increasing. A 2024 methodology preprint, led by a postdoctoral researcher at the University of Colorado, flagged that after the 2020 gap, the acceleration estimate for the Pacific shifted by roughly 0.2 mm/yr². That is a small number, but it is comparable to the acceleration signal itself in some analyses.

The shift was not caused by any physical change in the ocean. It was caused by the way the missing data interacted with the statistical models that estimate trends. The gap created a step change: the pre-2020 data, measured by gauges, had a slightly different baseline than the post-2020 data, which came largely from satellite altimetry. The models interpreted the discontinuity as a change in slope.

How One Lab's Budget Dictated a Global Metric

The lab in question, based at the University of Hawaii at Manoa, had been running the Pacific tide gauge array since the 1990s. It was part of a loose network of academic labs that supplement the official tide gauge networks run by national agencies like NOAA and the Australian Bureau of Meteorology. Together, these academic gauges account for roughly 30% of the global tide gauge coverage, particularly in remote islands and atolls where agency coverage is thin.

The instruments themselves are robust: acoustic sensors mounted on pilings, measuring water level to millimeter precision. But they require physical access for calibration, and the pier rental fees are set by local port authorities. In Honolulu, the rental had been stable for years, but it was not subsidized by any infrastructure fund. The lab paid it out of its grant budget.

When the NSF grant renewal failed in 2019, the lab had a year of carryover funding to wind down operations. The decision to cut which gauges was based on a mix of data value and cost. Honolulu Harbor and Hilo Bay were expensive because they were in active ports with high rental fees. The lab kept three other gauges that were cheaper to maintain, even though their records were shorter. The two cut gauges together represented about 25% of the lab's total data contribution to global databases.

The Hidden Economics of Long-Term Observation

The episode is not unusual. Long-term environmental observation networks around the world operate on a patchwork of soft money—grants that must be renewed every few years. NOAA's own tide gauge network, which includes about 200 stations, costs roughly $20 million per year to run. That is funded through a line item in the federal budget, but it covers only U.S. coasts and territories. Academic labs fill many of the gaps, especially in the Pacific and Indian Oceans, and they depend on grants from agencies like NSF, NASA, and the European Research Council.

Grant cycles of three to five years are mismatched with the multi-decade records needed to detect climate trends. A single renewal failure can create an unrecoverable gap in a time series. Unlike a laboratory experiment, which can be restarted, a tide gauge record that is missing two years cannot be filled in retroactively. The gap introduces a systematic uncertainty that propagates into every subsequent analysis that uses the record.

Several studies have documented the fragility of such networks. A 2018 paper in Nature Climate Change estimated that 40% of global tide gauge stations had at least one gap of more than a year in their records since 2000. The authors attributed most gaps to funding lapses, not equipment failure. The problem is especially acute in developing countries, where gauges are often maintained by foreign research projects with finite timelines.

A Concrete Example: Honolulu and Hilo Gaps

The Honolulu Harbor gauge, installed in 1905, is one of the oldest continuously operating tide gauges in the world. Its record shows a steady sea level rise of about 1.5 mm/yr over the 20th century, with an acceleration to roughly 3 mm/yr in the past two decades. The Hilo gauge, started in 1946, shows a similar pattern. Both were used in regional acceleration calculations published by the University of Hawaii and by NOAA's Sea Level Rise Viewer.

When the gauges stopped transmitting in mid-2020, the data stream went silent. For the next three years, the only sea level data for those locations came from satellite altimetry—specifically from the Jason-3 and Sentinel-6 missions. Satellites measure sea level relative to a global reference ellipsoid, while tide gauges measure it relative to a local benchmark on land. The difference matters because vertical land motion (subsidence or uplift) is captured by gauges but not by satellites. To merge the two data types, researchers must apply a correction for land motion, which introduces additional uncertainty.

A 2024 analysis by the same University of Colorado group found that using satellite data alone for the 2020–2023 period increased the uncertainty in the acceleration estimate for the Hawaii region by about 15%. The step change also shifted the estimated acceleration by 0.2 mm/yr², enough to affect comparisons between model projections and observations.

Publication Pressure Amplified the Artifact

The gap did not go unnoticed, but it was not widely flagged. Two high-profile papers in 2023, one in Geophysical Research Letters and another in Nature Communications, used post-2020 satellite data to calculate global sea level acceleration. Both papers blended satellite and tide gauge data without explicitly noting that several long-running gauges had been decommissioned. The peer reviewers did not catch the funding-induced break in the record, partly because the papers did not list the specific stations used.

A third paper, published in Journal of Climate in early 2024, used a blended dataset that included the Honolulu and Hilo records up to 2019 and satellite data afterward. The authors noted a “minor discontinuity” in the acceleration estimate but attributed it to natural variability. The preprint that later identified the funding cause was not published until mid-2024.

News coverage of the 2023 papers reported acceleration rates without mentioning the data change. One widely cited article in a major newspaper stated that “sea level rise is accelerating faster than previously thought,” based on the blended analysis. The article did not note that the acceleration estimate had shifted partly because of a change in the data source, not a change in the ocean.

What a Fixed Funding Model Would Cost—and Save

The solution seems straightforward: create a dedicated infrastructure fund for long-term environmental sensors, insulated from grant cycles. The cost would be modest. A global network of 500 tide gauges, covering all major coasts and islands, would cost roughly $50 million per year to operate—including rental fees, maintenance, and data processing. That is about 1% of the $5 billion that the U.S. federal government spends annually on climate research. For the European Union, the equivalent fraction would be similar.

Such a fund would prevent the kind of data gap that occurred in Hawaii. Reinstating the two gauges would cost about $30,000 per year—a fraction of the cost of the downstream correction studies that have been needed to account for the gap. One such study, published in 2024, cost roughly $200,000 in researcher time and computing resources to re-analyze the acceleration curves.

Opponents of dedicated infrastructure funds argue that they reduce flexibility. Grant-based funding allows agencies to shift priorities as science evolves. A fixed network might lock in obsolete instruments or locations. But tide gauges are a mature technology, and the locations are determined by geography, not fashion. The argument for flexibility is weaker when the data record is irreplaceable.

Takeaways for Researchers and Funders

The Honolulu and Hilo episode offers several lessons. First, researchers should always report funding discontinuities in their data methods sections. A simple note that “station X was decommissioned in 2020 due to grant expiration” would alert readers to potential artifacts. Second, journals should require funding stability statements for long-term data, similar to the data availability statements they already mandate. Third, funding agencies should offer 10-year grants for monitoring networks, with a presumption of renewal unless performance is poor.

Users of public sea level data should also be vigilant. A sudden change in the number of stations contributing to a global product, or a shift from gauge to satellite data, can indicate a funding-driven discontinuity. The calibration constants that underpin such records are as important as the raw measurements.

The broader lesson is that the economics of observation matter. A single gauge rental fee, lost to a grant cycle, can reshape a global curve. The same dynamic applies to other long-term records: sediment core chronologies and crystal growth runs are vulnerable to similar funding discontinuities. The scientific community has built elaborate statistical methods to correct for instrumental drift, but it has not yet built a system to correct for the drift of funding.

Broader Implications for Other Observational Networks

The tide gauge story is not an isolated case. Similar funding-driven gaps have been documented in other long-term observational networks. For example, the Global Seismographic Network, which monitors earthquakes worldwide, relies heavily on grants from the National Science Foundation. In 2018, a funding shortfall led to the temporary shutdown of roughly 10% of its stations, creating gaps that seismologists are still working to fill. The network's annual operating cost is around $15 million, yet a gap of even a few months can degrade the accuracy of earthquake location and magnitude estimates, especially for events in remote regions.

Another example is the network of oceanographic buoys operated by the Global Ocean Observing System. These buoys measure temperature, salinity, and currents, and their data are critical for climate models and weather forecasting. Many buoys are funded through short-term research projects, and when those projects end, the buoys often stop transmitting. A 2020 study found that roughly one-third of the global buoy array had gaps exceeding six months over the previous decade, with most gaps linked to funding lapses. The cost to maintain a single buoy is about $50,000 per year, yet the downstream economic value of the data—for shipping, fisheries, and storm surge warnings—is estimated to be in the hundreds of millions annually.

In the terrestrial realm, the Long Term Ecological Research (LTER) network, which includes 28 sites across the United States, has faced recurring funding uncertainties. Each site operates on a five-year grant cycle, and renewal is not guaranteed. In 2015, the LTER site in the Florida Everglades nearly closed when its grant renewal was delayed, threatening a 50-year record of water quality and vegetation change. The site was eventually funded, but the episode highlighted how even well-established networks can be vulnerable to administrative timing.

These examples share a common pattern: the data are irreplaceable, the operating costs are modest relative to the benefits, yet the funding model is fragile because it relies on competitive grants designed for short-term projects. The problem is not that agencies are unwilling to fund long-term observations—they do, but often through project grants that are not structured for continuity.

Trade-offs and Counter-Arguments

Not everyone agrees that dedicated infrastructure funds are the best solution. Some researchers argue that grant-based funding forces networks to remain efficient and responsive to changing scientific priorities. A fixed fund, they say, could become a slush fund for outdated instruments or unproductive sites. There is also the risk of political interference: if a network is funded through a line item in a federal budget, it could be cut or redirected based on political whims rather than scientific merit.

Another concern is that dedicated funds might reduce the incentive for innovation. For example, the tide gauge network has benefited from technological improvements—from mechanical float gauges to acoustic sensors to radar gauges—that were developed through research grants. A fixed fund might not support such upgrades as readily. However, this argument conflates instrumentation research with operational monitoring. The network could be funded separately from the research that improves it, much as the National Weather Service funds operational weather stations while separate research programs develop new sensors.

There is also a practical challenge: who would administer the fund? National agencies like NOAA and the European Environment Agency could take on the role, but their budgets are already stretched. An international fund, perhaps under the auspices of the World Meteorological Organization or the Intergovernmental Oceanographic Commission, could pool resources from multiple countries. But such a fund would require political agreements that take years to negotiate, and it might not be agile enough to respond to emerging needs.

Despite these concerns, the case for dedicated infrastructure is strong when the cost of failure is high. In the Honolulu and Hilo case, the $30,000 annual rental fee was less than 0.001% of the NSF's annual budget of roughly $8 billion. Yet the downstream costs—in terms of biased acceleration estimates, correction studies, and lost confidence in the data—likely exceed that amount many times over. A small investment in stability could have saved far larger costs later.

What Can Be Done Now

While the debate over long-term funding models continues, there are immediate steps that researchers and funders can take to reduce the risk of data gaps. One simple measure is to require that all long-term observational projects include a data continuity plan as part of their grant applications. The plan would specify how the data would be maintained if funding lapses, perhaps through a consortium of institutions or through a backup funding source. Journals could also require that papers using long-term data disclose any funding gaps in the record, similar to the conflict-of-interest statements they already require.

Another step is to create a rapid-response fund that can temporarily support critical stations when a grant gap occurs. Such a fund could be administered by a neutral body, such as a scientific society or a foundation, and would provide bridge funding for up to one year while the lab seeks new grants. The cost would be modest—perhaps a few million dollars per year for all environmental networks—but it could prevent the kind of permanent gap that occurred in Hawaii.

Researchers themselves can also take action. The University of Hawaii lab, for example, could have sought alternative funding from state agencies or private foundations before decommissioning the gauges. In hindsight, the cost of a short-term bridge grant would have been far less than the cost of the subsequent correction studies. The lab's decision was based on the available options at the time, but a more proactive approach to funding diversification might have saved the records.

Conclusion: The Drift of Funding

The Honolulu and Hilo tide gauge episode is a small, concrete example of a large, abstract problem: the mismatch between the time scales of funding and the time scales of observation. Climate science depends on records that span decades, but the money to sustain them comes in three- to five-year increments. A single grant failure can create a gap that no amount of statistical correction can fully repair.

The 0.2 mm/yr² shift in acceleration estimates is tiny, but it is a symptom of a systemic vulnerability. As sea level rise continues to accelerate—for physical reasons—the demand for precise, continuous observations will only grow. The scientific community has developed sophisticated methods to account for instrument drift, but it has not yet addressed the drift of funding. Until it does, the global records we rely on will remain fragile, subject to the vicissitudes of grant cycles and administrative decisions. The lesson from two tide gauges in Hawaii is that a small amount of money, invested in stability, can protect a large amount of scientific value.