Sue Merrilees is an advisor at the Science Philanthropy Alliance

Over the last two years, we at the Science Philanthropy Alliance have led a series of gatherings for philanthropists interested in ocean science. The first took place in May 2017 and was co-hosted with the Gordon and Betty Moore Foundation with the goal of introducing Alliance members and other interested philanthropists to one another and to ocean science priorities in general. The next two workshops (January and April 2018) were held on behalf of Dalio Philanthropies and focused on its OceanX One Big Wave initiative.

As more partners interested in ocean science funding emerged, we convened a workshop in September 2018 at the Global Climate Action Summit in San Francisco on transformative technology for ocean science that featured scientists from several top oceanographic institutions. This event drew philanthropists and senior staff from more than a dozen foundations and concluded with a funders-only discussion. During this discussion, funders asked to convene again to dive deeper into ocean data science. Schmidt Marine Technology Partners agreed to host, and a January 2019 date was set for the workshop. It was agreed that one focus would be to generate a “seascape” of funding opportunities and to encourage collaborations and partnerships.

While there were many animated discussions during the scientific presentations on topics like data platforms, artificial intelligence/machine learning, and modeling and visualization, a panel on Recruiting, Training and Retaining Ocean Data Scientists attracted particularly enthusiastic engagement from the audience. This panel focused on addressing challenges and finding solutions to the shortage and retention of talent in academic research. The panel featured three early career scientists who came to work in ocean data science via different career paths.

Chris Mentzel, from the Gordon and Betty Moore Foundation, served as moderator, and began by noting that ‘ocean data science’ is not yet an established concept and then asked the audience for a definition of data science more broadly. He joked that even though he has been working on an initiative related to data science for over a decade, he still hasn’t figured it out, and concluded that data science is still emerging as a discipline, one largely at the intersection of computer science and statistics.

Mentzel then asked the panelists to introduce themselves, focusing on their career trajectories.

Julia Stewart Lowndes, of the National Center for Ecological Analysis and Synthesis at University of California, Santa Barbara, used a Star Wars analogy, comparing herself to Luke Skywalker standing next to his crashed spaceship on the shores of a swamp when she first confronted data challenges as a PhD student at Stanford University. But, with guidance from Yoda (mentoring and open data science), she learned how to problem solve and improve her science through coding and open software communities. She said, “I was trained as a marine ecologist, but I was not taught data science in graduate school, or how to collaborate efficiently with data.” Frustrated at the limitations of exchanging data with other scientists via email and Excel files, she taught herself how to code since no formal courses were available. In 2018, she received a Mozilla Fellowship, which enables her to apply both data knowledge and collaborative techniques to promote open science through her own mentorship program, Openscapes.

Donglai Gong, an assistant professor at Virginia Institute of Marine Science, College of William and Mary, credited his wide interests (aeronautics, astrophysics, and oceanography) with providing him with diverse and deep mathematical training, and he also noted his willingness to seek out data science tools (e.g. MOOC courses) outside of his formal classes.

Christian Müller, currently at the Simons Foundation’s Flatiron Institute, admitted that while growing up in Germany, his first dream was to play professional soccer. He ultimately turned to his extensive other interests, which included bioinformatics, physics, computational biology, applied math and statistics, and ultimately, computer science. When he began to study the microbiome as a statistical problem, it led him to the ocean microbiome, and he concluded, “Ocean data science is exactly the area where I can use all my different skill sets.”

left to right: Christian Müller, Flatiron Institute, Simons Foundation; Donglai Gong, Virginia Institute of Marine Science, College of William and Mary; Julia Stewart Lowndes, National Center for Ecological Analysis and Synthesis, University of Santa Barbara; and Chris Mentzel, Gordon and Betty Moore Foundation

Mentzel described how, in 2016, the Gordon and Betty Moore and Alfred P. Sloan Foundations partnered to fund the Moore-Sloan Data Science Environments at New York University, University of California, Berkeley, and University of Washington. These environments were intended to build academic homes for data scientists at universities and research institutes. Noting the diversity of the panelists’ background, “someone from a university, a university-adjacent center, and an independent research organization,” Mentzel asked each to comment on the challenges posed by academic resources and culture to attracting talent in data science.

Lowndes noted she was not on a tenure track, and while this allowed her to contribute to team science, it meant much of her funding was on a grant basis with year to year salaries. “We shouldn’t only give job security and status to tenure-track scientists – there are many important ways to contribute to academic research and we need to value those roles and provide stability for teams if we want to retain more talent in science.”

Gong observed that for himself and his students, working on questions they cared about was “worth a lot,” and that when graduate students “choose this path, they are pre-selecting themselves to say that they care more about science than about making money.”

Mentzel noted that in computer science, it was normal for professors to leave to work on a start-up, and then return, but this was uncommon in the natural sciences. Gong had spent six months of his own time during evenings and weekends on a start-up prior to the tenure process and recommended it as it allows scientists to test ideas outside of their disciplines, noting that this is work that NSF and NOAA would not support. “To be able to work with someone else who has training and expertise in another discipline, to flesh out an idea and take that risk, accepting it may fail, is important for science to move forward in the future,” Gong said.

Müller, at the Flatiron Institute, commented that the institute was “in between academia and industry, because we can pay a bit more money than academia, and while we don’t have the typical academic track, we have mostly research scientists and post-doctoral students.” He felt it was “an interesting model, based on Bell Labs, but on a different scale,” and that it was “a good place for outcasts who do not fit the academic formula.”

Mentzel asked the panelists to describe the elements of successful data science collaborations.

Lowndes commented, “I think about this in terms of who is included and excluded from data science. The barriers to engage are exposure to what is possible and confidence that you can engage successfully. It is about creating a culture of support.” She added, “Data and repositories are cross-cutting, uniting scientists across domain-specific research questions. But without training in data science and open practices, many marine scientists will be excluded from the full power of data science.”

Müller said, “A key point is curiosity and openness, because in oceanography there are so many different disciplines, and it requires a lot of effort for all parties to communicate.”

Virginia “Ginger” Armbrust, a professor in the School of Oceanography at the University of Washington, director of SCOPE-Gradients at the Simons Foundation, and the moderator for the ocean data science workshop, commented on the inadequate funding for such efforts, “An important note is that collaborations that span so many disciplines take time. So, a normal three-year grant from NSF is not enough time.”

Mentzel then asked the panelists for final thoughts on recruitment, training, and retention.

Lowndes recommended providing training in coding and data analysis to undergraduate, graduate, and early career science faculty – and supporting the scientists who teach the training sessions.

Gong agreed that training for graduate students in data science is essential and commented, “I like the idea of summer schools that bring students together with professional data scientists to learn they how they can apply these advanced tools to their own research and then take this back to their labs.” He added, “It is my dream to make enough money through a start-up to fund my own science.”

Müller recommended funding programs beyond the standard three years for PhD and post-doctoral fellows, so that teams could be sufficiently trained and fully integrated.

Commenting on the interest from the funders and the enthusiastic discussion generated by the presentation and the other scientists in the room, Mentzel wondered if perhaps ‘ocean data science’ could become a more established concept in future. It wouldn’t be the first time philanthropy has helped nurture an emerging field, and I found it exciting to observe an increasing number of funders recognizing the potential of investment in ocean science.

A rising tide or One Big Wave, it’s a powerful and precious phenomenon to witness.