University of Rhode Island

Peter Cornillon

Emeritus Professor of Oceanography / Graduate School of Oceanography

pcornillon@uri.edu +1 401 742 2911 Watkins Laboratory, Rm 112 Narragansett, RI 02882

I study the ocean surface as satellites see it — the fronts, eddies and temperature structure that thermal infrared and microwave sensors have been recording for more than forty years — and I build the data systems that make those records usable by other people.

I joined the University of Rhode Island in 1975, on the Ocean Engineering faculty, and moved to the Oceanography faculty at the Graduate School of Oceanography in 1980. I became emeritus in 2023. Emeritus at URI means I have stopped teaching; it does not mean I have stopped doing science. I work on research full time, and I remain based at GSO.

My training is not in oceanography. I took my B.S. in engineering physics and my Ph.D. in experimental high energy physics, both at Cornell, and came to the ocean by way of the instruments. That has shaped how I work ever since: most of my career has been spent on the measurement itself — what the sensor actually recorded, how much of it is noise, and whether the number you pull out of an archive means what you think it means.


Current work

What I am working on now

Four threads, all of them active. The first two are the reason the archive described below exists.

Dataset reconstruction

Rebuilding two decades of MODIS Level-2 sea surface temperature

A four-year effort to reprocess the R2019 MODIS SST record from Terra and Aqua, correcting problems in the delivered product that matter for anyone using the data to look at gradients, fronts or long-term change. Processing ran in the cloud and touched more than four million granules. The result — roughly 40 TB per satellite, about 80 TB in all — is served publicly and openly at sst.uri.edu, and the paper documenting it is in final preparation.

Measurement uncertainty

The structure of noise in satellite SST fields

Satellite SST error is not white, and treating it as though it were quietly corrupts every derived quantity that involves a derivative — gradients, fronts, and the small-scale variability that mesoscale and submesoscale work depends on. Recent work characterises correlated noise in the difference between coincident AMSR-E and MODIS fields; earlier work established the pixel-to-pixel uncertainty in the AVHRR record back to 1982.

Ocean structure

Fronts, rings and zonal structure in the surface ocean

The long thread. The Cayula–Cornillon edge-detection algorithm came out of this work in the early 1990s and is still the basis of much satellite frontal analysis; the same line runs through Gulf Stream ring and meander statistics, wind response over SST fronts, and quasi-zonal banding in the microwave SST field.

Research methods

Generative AI as a scientific instrument — and as a problem for universities

I now do a substantial share of my analysis and software development in collaboration with large language model agents, and the change in what one researcher can accomplish has been large enough to be worth writing about. Separately, with J. X. Prochaska, I have argued that higher education needs to take seriously how much of the teaching role an agentic AI system could assume, and how soon — “The Agentic Professor: Exploring GenAI-Supported Futures in Higher Education,” in press in EDUCAUSE Review.

In preparation

Three papers are currently in preparation and are expected to be submitted before the end of 2026. Work now underway is expected to produce two more shortly after.


1990–2005

Distributed data systems and OPeNDAP

For roughly fifteen years, from the early 1990s through the mid-2000s, the larger part of my working life went into a problem that was not oceanography at all: how a scientist at one institution could read a binary dataset held at another without first downloading it, converting it, and hoping the two formats agreed.

I was PI on the development of what became the OPeNDAP data access protocol, which lets a client open a remote dataset over the network as though it were a local file, request only the subset it needs, and receive it in a form its own software already understands. I served as president of the OPeNDAP project from 2000 to 2006. The related work of those years included the National Virtual Ocean Data System (NVODS), serving GODAE data and products to the ocean community, the OPeNDAP Data Connector, and the integration of remote data access into scientific workflow systems.

The protocol outlived the project that produced it. It is now a routine part of how the earth science community distributes data — and it is what the SST archive described below is served over, which is the sense in which this thread and the current one are the same thread.


Resources

Data and software

A recurring theme of my career: the science is only as good as the community's ability to get at the data. Everything below is open and in active use.

Improved MODIS L2 SST archive

More than 80 TB of reprocessed Level-2 sea surface temperature from MODIS Terra and Aqua, spanning the full mission record, served over OPeNDAP so that users can subset remotely rather than downloading whole granules.

OPeNDAP

The data access protocol described above: remote binary datasets read and subset over the network as though they were local. Open source, widely deployed across the earth science community, and the transport behind the SST archive.

Cayula–Cornillon front detection

The histogram-based edge detection algorithm developed with J.-F. Cayula for SST imagery, together with the companion sequential cloud-detection method. Both are described in the publications listed here and have been independently reimplemented many times over.

Claude Switchboard

A macOS status panel for running several AI coding sessions at once. It reports, at a glance, every desktop on the machine, the project each one is working on, and which sessions are running, finished, or stopped to ask a question. Written for my own use in day-to-day research computing and released publicly.

Claude configuration

The working standard my AI-assisted research sessions run under: one rules file that loads in every repository, a project template that gives each project the same spine of status, decision, task and log files, and hooks that keep a record of what was done and why. Assembled from a configuration in daily use on the projects described here.


Background

Education and appointments

1973Ph.D., Experimental High Energy Physics — Cornell University. Advisor: J. Orear.
1969B.S., Engineering Physics — Cornell University.
2023–Emeritus Professor of Oceanography, University of Rhode Island
1990–2023Professor of Oceanography, University of Rhode Island
2000–2006President, Open Source Project for a Network Data Access Protocol (OPeNDAP)
1987–1990Associate Professor of Oceanography and Ocean Engineering, URI
1983–1987Associate Research Professor of Ocean Engineering and Oceanography, URI
1981–1983Assistant Research Professor of Ocean Engineering and Oceanography, URI
1980–1981Visiting Assistant Professor, Department of Meteorology and Physical Oceanography, MIT
1979–1986Partner and Senior Scientist, Applied Science Associates
1975–1980Assistant Research Professor of Ocean Engineering, URI

Service and teaching

Selected activities

ongoingMember, GHRSST Science Team (Group for High Resolution Sea Surface Temperature)
2009–2012Team lead, NASA Sea Surface Temperature Science Team
2019–2022Created and taught the URI Grand Challenge course “Envisioning the Future”, averaging 110 students per offering
2018–2019Sabbatical at NASA Jet Propulsion Laboratory with Dimitris Menemenlis, working on SST in LLC4320 MITgcm output; machine learning coursework at Caltech
2011Co-chaired the URI Honors Colloquium “Are You Ready for the Future” and co-taught the companion honors course
30+ yrsTaught or co-taught at least one course every semester for over thirty years, across ten or more distinct preparations in physics, oceanography, and technology and the future
careerBrought more than $20M in federal funding to the university for oceanographic research and distributed data systems
careerAdvised 11 Ph.D. and 8 M.S. students and 8 postdoctoral scholars; served on numerous data-related committees and science panels