NASA science moves toward greater sharing of data and software

NASA’s Science Mission Directorate has reconciled its open-access policies across divisions and proposed steps to more quickly and thoroughly share data and software. The move is in line with broader tendencies across the US government toward open data.

The proposed changes for NASA-funded science are spelled out in document SPD-41a, which was released in draft form in November. Public comments will be accepted through 4 March. It is anticipated that the final policy will be adopted later this year and will be applicable to grant solicitations starting in early 2023, according to Steven Crawford, science data officer for NASA’s Science Mission Directorate.

Astronomers broadly applaud the space agency’s moves toward increased openness. But they worry about unintended consequences—chiefly that the policy could result in proposers getting scooped if proprietary periods for data disappear. Another worry is that sharing software would impose heavy burdens of time and cost. “The devil is in the details,” says Lisa Storrie-Lombardi, director of Las Cumbres Observatory, who previously worked on NASA’s Spitzer mission.

SPD-41a stipulates that data must be released as soon as they have been calibrated and validated. “There shall be no period of exclusive access to Mission data,” the draft policy states, and the period for preparing data for release “shall be as short as possible and shall not exceed six months.”

The new rules would apply to all future missions, unless superseded by international agreements. That’s the case, for example, for the James Webb Space Telescope, for which the default proprietary period is 12 months, although about one-third of the first-year observations now getting underway have zero proprietary time. Rules for existing missions are grandfathered in—Hubble Space Telescope users, for example, typically get a six-month proprietary period.

Opposition is minimal when it comes to zero proprietary time for data from surveys, such as those from TESS (Transiting Exoplanet Survey Satellite) or the Roman Space Telescope. Because survey data are broad rather than based on competitively awarded observations, releasing them is the norm and is generally recognized as maximizing their science value. But scientists worry about principal investigator–driven observations, for which 6 to 12 months of exclusive data access is customary. “For missions where proposers pick particular targets, I would be inclined to keep proprietary periods,” says Chris Lintott, an astronomer at Oxford University and cofounder of the crowdsourced project Galaxy Zoo. If data are open from the start, “others may be quicker to analyze them than the original proposer. That happens already in highly competitive fields,” he says. “It’s a freeloading problem.”

The person or group who put creative thought into getting the data should receive credit for that, agrees Jonathan McDowell of the Center for Astrophysics | Harvard & Smithsonian. One solution, he says, would be to publish winning proposals that could then be cited in later publications.

Others point out that PIs typically have a built-in head start: They have been thinking about the work and likely were awarded funding for it. “There is tension between openness and proprietary periods, but the benefits of sharing data outweigh the dangers,” says Lintott.

Ditching proprietary periods would most punish researchers at small institutions and graduate students. “Those of us at small institutions don’t have big teams of postdocs. My graduate student is a teaching assistant and does research on the side,” says Gerard Williger, who studies quasars and protoplanetary disks at the University of Louisville in Kentucky. Given his heavy teaching load, he adds, “if I only had six months’ exclusive access to data, and the data came in August or September, my hands would be tied. I wouldn’t have the capability to turn around and analyze data that fast.”

Similarly, graduate students “are put at a disadvantage if another, more experienced group can move faster,” says Storrie-Lombardi. “Early-career people and people who have a higher teaching load should get to do research too.” If data are released a bit later, she says, “the science will still get done. It’s not cheating the taxpayer in any way, shape, or form. It’s an equity issue. NASA should take those things into account.”

Joshua Peek is PI for the Mikulski Archive for Space Telescopes (MAST), the repository for data from NASA’s optical, UV, and near- and mid-IR missions at the Space Telescope Science Institute in Baltimore. He says that “a broad group of people use data from the MAST archives, especially early-career scientists and researchers from smaller institutions and poorer countries. I and my institute are big supporters of open data.” Still, he agrees that “it’s reasonable for graduate students to have proprietary access to data.”

FAIR use

Sharing software presents other challenges. The SPD-41a draft says that software developed with funding from the Science Mission Directorate “shall be made available in a publicly accessible repository.” The big change for software release, says NASA’s Crawford, is the shift from “should” to “shall.” The aim, he adds, “is for missions and individual researchers to release software and make it readily accessible. We are encouraging FAIR—findable, accessible, interoperable, reusable—data and software.”

Rich Townsend of the University of Wisconsin–Madison has for years been publishing and maintaining the open-source stellar-oscillation code GYRE. “It’s been a wonderful adventure seeing the flourishing and transformative science and building a vibrant community of users,” he says. Still, he notes that “writing a comprehensive manual is not recognized as a scientific activity or funded by NASA.”

Townsend points to a long history of proprietary code and says that some of his colleagues worry that their graduate students will lose a competitive edge if software is open source. Another concern he has heard, he adds, is that “some bits of code may not work as they should. The coders may be slightly embarrassed. They may not want to publicly display all the details of how the sausage is made.”

The SPD-41a document makes exceptions for commercial software and “single-use” software. But “single-use” is not clearly defined, say many researchers. “Writing hacky stuff that works may take a day,” says McDowell, “but making it usable for others could take six months. And even if you put in that work, the software may not be useful to others.” He worries that the software release requirement will come at the expense of doing science: “To the extent that you layer on requirements to produce the code, you are putting grit in the wheels of making science.” In his comment to the Science Mission Directorate, McDowell requests a “blanket exception” for single-use software.

NASA needs to provide specific guidelines about the open software requirements, say concerned scientists. Townsend likens “just releasing software” to “illegally dumping—throw the software out on the roadside and drive off.” That will be useless for science, he says.

“We want to balance expectations of openness with the burden on individual researchers,” says Crawford. “Research papers don’t always capture all the nuances that went into a project. Reading software could significantly increase the reproducibility of results,” he adds. “We are hoping to develop more guidance on this.”

The intent of the software-sharing policy is good, says Peek, but there should be variances. As written, he says, SPD-41a would require “a ton of unnecessary work from both centers and investigators,” and if the requirements are overly onerous, scientists may be deterred from using open data. “Interpreted strictly, the policy could have a chilling effect.”