NSF Tried to Pull 900 Ocean Sensors. Congress Stopped It. Now the Agency Is Asking Which Ones Are Worth Keeping — and the Answer Window Closes September 30.

In June, a ship was scheduled to start pulling a buoy out of the water off the Oregon coast. It was the first step in removing most of the roughly 900 sensors that make up the Ocean Observatories Initiative — a $386 million system of cabled and moored arrays off Oregon, Washington, Alaska, North Carolina, and Greenland, built for a 30-year design life and about a decade into it.

The National Science Foundation called the plan a "descoping" rather than a cancellation. Senators Jeff Merkley of Oregon and Lisa Murkowski of Alaska called it something else and passed a two-page bill through the Senate barring the agency from spending federal funds to decommission the network before a real review. On June 18, 2026, NSF reversed course: it would "not proceed with further removal or descoping of equipment from the remaining arrays," would continue operations including planned maintenance, and would redeploy the already-recovered Endurance Array after servicing.

NSF also promised two things in that statement — a Dear Colleague Letter soliciting stakeholder input, and an expert panel to weigh it.

The letter arrived on August 27, 2026. It is NSF 26-024, signed by Chief Science Officer Simon Malcomber, and responses are due September 30, 2026.

This is not a routine RFI

Dear Colleague Letters that ask the community what it wants are common at NSF, and most of them are genuinely low-stakes: an agency taking the temperature of a field before writing a solicitation. This one is different in a specific way. It is the procedural record that a review — the one Congress demanded — actually happened, and it is the document the expert panel will read.

The letter's framing is not "what should we build." Question 5 asks, in NSF's words, which elements of NSF's existing portfolio require maintenance or updating. That is a question about what stays in the water.

The full set of nine:

1. How should NSF's ocean observation investments complement or differ from NOAA and other national and international efforts?
2. What observation gaps exist for high-impact science questions, and which oceanic regions need NSF systems?
3. What observation types advance the national interest?
4. Which observation types and technologies are ready for implementation?
5. Which elements of NSF's existing portfolio require maintenance or updating?
6. What operational uses and stakeholders should be considered?
7. What metrics assess scientific and societal impact over time?
8. How should NSF balance developing new technologies against collecting data with established ones?
9. What additional factors warrant consideration?

Read questions 1, 3, and 8 together and the argument NSF is inviting becomes visible. Question 1 offers a path to shed anything NOAA arguably duplicates. Question 3 reframes ocean observing around national interest rather than climate science — the framing under which the original removal was proposed. Question 8 sets up a trade in which maintaining legacy infrastructure competes directly against funding new technology, which is the trade any budget-constrained agency wants to be able to point to a community endorsing.

None of that means the outcome is predetermined. It means the questions are load-bearing, and answering only the ones that feel comfortable is how a community loses an argument it could have won.

The mechanics are unusually restrictive

Responses go through a Microsoft Forms portal linked from the DCL, not to a program officer's inbox. Each question is capped at 3,250 characters including spaces — NSF describes this as roughly a single page of 12-point text.

That cap is the most important operational detail in the letter. Nine questions at 3,250 characters is about 4,000 words total, and there is no mechanism to attach a 40-page white paper. A lab that has spent a decade building a case for its array cannot upload the case. It has to compress it.

The second detail: "All information provided to NSF may be made public without further notice." Responses are not confidential submissions to a program officer. Write accordingly.

Third, and easily overlooked: responses "may be submitted by individuals or as aggregated responses from groups or organizations." Both channels are open, and they do different work. A consortium response carries institutional weight and can present a coherent portfolio argument. Individual responses demonstrate breadth — the number of distinct scientists, students, educators, technology developers, fishery managers, and tribal and state agencies who depend on a given data stream. An expert panel reading two hundred individual responses that all independently cite the Endurance Array's role in West Coast hypoxia forecasting is receiving a different signal than one reading a single society letter that says the same thing.

Nothing in the DCL prohibits doing both. Groups that plan to file an aggregated response should still encourage members to file individually, and should share their draft early enough that individual filers can reinforce rather than duplicate it.

Who this actually affects, beyond oceanographers

The obvious constituency is the ocean sciences community — the researchers who write proposals against OOI data, the students trained on it, the technology developers who build to its interfaces.

The less obvious constituency is everyone whose NSF proposal cites a data source they did not personally collect. Question 7 asks what metrics assess scientific and societal impact over time. Whatever answer NSF adopts becomes the standard by which observing infrastructure justifies its budget line, and observing infrastructure is not confined to the ocean. NSF operates long-baseline facilities across the geosciences, and the precedent set here — how usage is counted, whether downstream publications and operational forecasts count, whether stakeholder dependence counts at all — will not stay in the ocean.

Anyone who has framed a proposal around a long-term dataset has a stake in question 7 specifically. It is also the question most likely to go unanswered by a community focused on defending hardware.

There is a funding-strategy dimension too. NSF's Geosciences directorate spent 2026 consolidating: the GEO Core solicitation, NSF 26-516, merged atmospheric, earth, and ocean core programs into a single anytime-submission opportunity, part of the same portfolio-wide restructuring that replaced twelve solicitations and abolished submission deadlines. Facility decisions and core-program decisions are being made by the same agency in the same year under the same pressure. A community that treats NSF 26-024 as a fight about buoys and skips the question about how NSF should balance new technology against established measurement is answering half the letter.

What a useful response looks like

The DCL rewards specificity that survives compression. Some concrete guidance for the 3,250-character format:

Name the array and the science question in the same sentence. "The Coastal Endurance Array" plus "hypoxia and ocean acidification forecasting for the Oregon and Washington shelf fishery" is a claim a panel can evaluate. "Long-term observations are foundational to ocean science" is not.

Answer question 1 head-on rather than avoiding it. If your data stream is genuinely not duplicated by NOAA, say exactly why — measurement depth, temporal resolution, subsurface coverage, real-time cabled power and bandwidth. If parts of it are duplicated, say that too. A response that claims nothing overlaps reads as advocacy and gets discounted; a response that concedes overlap and then draws a clean line around what does not reads as expertise.

Bring numbers to question 7. Publications, theses, downstream operational products, distinct data users, classroom deployments. If you can count it, count it in the response rather than assuming NSF already has the count.

Say something about question 8. The new-versus-established trade is real, and a community that refuses to engage it hands the framing to whoever does.

Practically: draft in a text editor with a character counter, not in the form. The portal is a submission mechanism, not a workspace, and losing a partial draft to a session timeout is an avoidable way to miss a deadline that is now under four weeks out.

The expert panel NSF promised will convene after September 30 to assess observational needs, evaluate available data sources, weigh these responses, and identify what the agency calls a sustainable path for its ocean observing systems. The letter is the only formal channel into that process that is open to anyone who wants it. For researchers tracking which federal comment windows are worth the hours and which are theater, Granted surfaces the ones with real decisions attached — and this is one of them.