The Argo float program is the world’s primary source of systematic, real-time ocean temperature and salinity data below the sea surface. Since 2000, a fleet of autonomous, battery-powered instruments has drifted with ocean currents, periodically diving and resurfacing to profile the water column and beam its readings to shore by satellite. For research administrators and data stewards outside oceanography, Argo is worth understanding on its own terms and as a working example of a specific design pattern: a large, internationally distributed sensor network whose raw output is placed in the open, quality-controlled by a formal two-stage process, and made citable through persistent identifiers rather than routed through any single national agency.
What Argo Is and What It Measures
Argo is an international program, not a single instrument or dataset. Its core hardware is the profiling float: a roughly 1.3-2 metre cylindrical instrument that spends most of its life below the surface, adjusting its own buoyancy to sink to a “parking depth,” drift with the local current for several days, then dive further before rising back to the surface while recording temperature and salinity (via conductivity) at intervals through the water column. On reaching the surface, the float transmits its profile by satellite, typically via the Iridium network, then submerges again to repeat the cycle, usually every ten days. A single float has a working life of roughly four to five years before its battery is exhausted, which is why sustaining the array requires continuous replacement deployment, not a one-time launch.
The original “Core Argo” mission measures temperature and salinity from the surface to about 2,000 metres depth. That data feeds directly into operational ocean and weather forecasting, seasonal climate prediction, and the long-term climate record used to track ocean heat content and sea-level rise — the ocean absorbs the large majority of excess heat trapped by the climate system, and Argo is the dataset that lets researchers actually measure where that heat is going.
How the Float Network Is Governed and Funded
Argo has no single owner. It is coordinated by the Argo Steering Team under the auspices of the intergovernmental Global Ocean Observing System (GOOS) and the Joint WMO-IOC Collaborative Board (the successor to JCOMM), with metadata and deployment tracking supported by the international project office OceanOPS (formerly JCOMMOPS). Roughly 30 national programs contribute floats and funding — each country funds, builds, and deploys its own instruments and then contributes the resulting data to the shared pool under a common data policy, rather than any single funder underwriting the whole array. Major contributors include national oceanographic and meteorological agencies in the United States, the European Union (via the Euro-Argo European Research Infrastructure Consortium), Japan, Australia, and others; the array’s overall design and target size are set collectively through the Argo Steering Team rather than by any one national program.
From 3,000 Floats to OneArgo: Core, Deep, and BGC Missions
Argo’s original design target, set at program launch in 2000, was a global array of roughly 3,000 floats providing even spatial coverage of the ice-free open ocean — a target the program reached in November 2007. Since then, the array has both grown and diversified in mission. Two newer float types now sit alongside Core Argo:
- Deep Argo floats profile the full water column to a depth of about 6,000 metres, extending measurement below Core Argo’s 2,000-metre limit into the deep ocean, which conventional Argo cannot observe.
- BGC-Argo (Biogeochemical-Argo) floats carry additional sensors — typically for dissolved oxygen, pH, nitrate, chlorophyll fluorescence, optical backscatter, and downwelling irradiance — extending the array from a purely physical-oceanography instrument into one that also tracks ocean biogeochemistry and the carbon cycle.
The current planning framework, known as OneArgo, targets an integrated global network of about 4,700 floats by 2030: roughly 2,500 Core floats, 1,200 Deep floats, and 1,000 BGC floats, along with expanded coverage into seasonal sea-ice zones and marginal seas that the original design excluded. Because floats are continuously being deployed and retired, and coverage is never perfectly static, published figures for the number of floats currently reporting vary somewhat by source and month; Argo’s own program pages and data centers (linked below) are the authoritative live count rather than any fixed number repeated here.
Data Management Architecture: DACs, GDACs, and Two-Stage Quality Control
Argo’s data pipeline is a useful case study in distributed-but-standardized data stewardship. Each national program operates its own Data Assembly Centre (DAC), which receives raw transmissions from its own floats and applies a first pass of automated real-time quality control — range checks, spike detection, and similar tests designed to catch obviously erroneous readings without requiring human review. Real-time profiles are typically made available within 24 hours of a float surfacing, which is what makes Argo usable for operational forecasting, not just retrospective research.
Every DAC then forwards its full data stream, real-time and delayed-mode alike, to one of two mirrored Global Data Assembly Centres (GDACs): one operated by Ifremer’s Coriolis data centre in France, the other in the United States. The two GDACs hold identical copies of the complete global Argo dataset and are the canonical distribution point — this redundant, mirrored design is itself a deliberate resilience choice, not an accident of history.
Separately from the automated real-time pass, trained scientific experts apply delayed-mode quality control to each profile, typically within six to twelve months of collection. This second pass adjusts for sensor drift and calibration issues that automated checks cannot catch, and produces the scientifically citable, research-grade version of the dataset. The distinction between real-time and delayed-mode data matters for anyone reusing Argo data: real-time profiles are fast but provisional, delayed-mode profiles are slower but authoritative, and Argo’s documentation is explicit about which flag applies to a given profile.
Accessing, Citing, and Reusing Argo Data
Argo data is released under an open data policy with no restrictions on use, redistribution, or the requirement to seek permission — a policy choice that predates, but aligns closely with, what the research-data field now calls FAIR and open-data norms. Data can be retrieved directly from either GDAC via FTP or HTTPS, and the GDAC snapshot dataset itself is assigned a persistent DOI (via SEANOE, Ifremer’s data repository), giving researchers a citable, versioned reference point rather than an informal pointer to a live directory that changes daily. A range of secondary tools and portals — including Argovis, the Euro-Argo data selection tools, and national meteorological agency ERDDAP servers — provide search, subsetting, and visualization on top of the same underlying GDAC data, rather than maintaining separate copies with their own provenance.
For a research data steward, Argo’s citation model is a practical illustration of a recurring problem: how to make a continuously updated, operational dataset citable without freezing science in place. Argo’s answer — a periodically snapshotted, DOI-assigned reference dataset sitting alongside the live, continuously updated operational feed — is a pattern worth recognizing when evaluating repository or data-infrastructure design elsewhere, not just in oceanography.
Why Argo Is a Useful Reference Point for Research Data Management
Outside its direct scientific value, Argo is instructive as a model of large-scale, internationally governed open research infrastructure:
- Federated funding, unified policy. No single funder or country owns Argo, yet all contributors operate under one shared data policy and one technical data format — a governance pattern relevant to any multi-institutional or multi-national data-sharing consortium.
- Two-tier quality control as a documented lifecycle stage. The explicit real-time versus delayed-mode distinction gives data users a documented, machine-readable way to know how much scrutiny a given record has received — a more granular model than a simple published/unpublished binary.
- Redundant, mirrored distribution. Maintaining two full GDAC copies rather than one central repository is a preservation and continuity choice with direct relevance to how institutions evaluate repository trustworthiness and long-term access risk.
- Persistent identifiers for a moving target. Snapshotting a continuously updated live dataset under a DOI, rather than treating “live” and “citable” as incompatible, is a solvable-problem pattern other operational or sensor-network datasets can follow.
Frequently Asked Questions
What does an Argo float actually measure?
Core Argo floats measure temperature and salinity (via electrical conductivity) from the surface to about 2,000 metres depth on each dive cycle. Newer BGC-Argo floats add sensors for dissolved oxygen, pH, nitrate, chlorophyll, light backscatter, and irradiance; Deep Argo floats extend the same physical measurements down to about 6,000 metres.
How often does an Argo float report data?
The standard cycle is about ten days: a float parks at depth, drifts with the current, dives deeper, then rises to the surface measuring as it goes and transmits its profile by satellite before submerging again.
Is Argo data free to use?
Yes. Argo operates an open data policy with no cost, registration requirement, or use restriction for either real-time or delayed-mode data, accessible directly from the GDACs or through secondary tools built on top of them.
What is the difference between real-time and delayed-mode Argo data?
Real-time data has passed only automated quality-control checks and is typically available within about 24 hours of a float surfacing, prioritizing speed for operational forecasting. Delayed-mode data has additionally been reviewed by trained scientists, usually within six to twelve months, to correct for sensor drift and calibration issues, producing the research-grade version of the record.
How many Argo floats are currently active?
The array has fluctuated around 3,800-4,000 actively reporting floats in recent years, with the OneArgo design targeting an expanded network of about 4,700 floats (Core, Deep, and BGC combined) by 2030. Because floats are continuously deployed and retired, the current count is best checked directly against Argo’s own program pages or GDAC status reports rather than a fixed figure.
Can Argo data be cited in a publication?
Yes. The GDAC dataset is assigned a persistent DOI through SEANOE (Ifremer’s data repository), which researchers can cite as a versioned snapshot rather than pointing to a live, continuously changing directory.
Who funds and governs the Argo program?
Argo is coordinated by the Argo Steering Team under the Global Ocean Observing System and the Joint WMO-IOC Collaborative Board, with deployment and metadata tracking supported by OceanOPS. Funding and float deployment are distributed across roughly 30 participating national programs rather than centralized in a single agency.
Related CASRAI Resources
Argo sits alongside other domain-specific ocean and earth-science data infrastructures covered elsewhere on this site: see OBIS, the Ocean Biodiversity Information System, for how marine species-occurrence data is aggregated and shared under a comparable open-data model; PANGAEA, for how earth and environmental science datasets more broadly are published and made citable; and NOAA NCEI, for the US federal archive that holds long-term climate and ocean records including Argo-derived products. For the broader concepts Argo’s design illustrates, see this site’s guides on choosing a metadata schema for a dataset and the CASRAI dictionary entry on the data management plan (DMP).
