DATA

Explore opens a plain reference globe. Coastal names become more detailed as you zoom; named seabed features use representative label markers. Coastal names, seabed names and reef outlines can be switched independently. This is a geographic atlas, not a navigation chart.

Seabed names: IHO-IOC GEBCO Gazetteer of Undersea Feature Names, www.gebco.net, distributed by NOAA NCEI / IHO DCDB. Saved 18 October 2026 from all three official point, line and polygon layers: 5,310 distinct named features after combining repeated feature IDs and omitting one unknown-type record. Generic types include banks, shoals, canyons, ridges, trenches and channels. AO uses the spherical mean of each feature’s source geometry vertices as a representative label position; markers are not surveyed hazard positions or feature boundaries. Feature coordinates shown are rounded label anchors, not a statement of positional accuracy. The red pointer marks a location selected on the reference globe; its coordinates are not a surveyed depth or hazard position. Find location and shared Explore links can retain this pointer.

Coastal names and major reefs: Natural Earth reference geography. Existing ocean, sea, gulf, bay, strait, sound, channel, fjord, island and cape labels are retained. Major reef outlines are from Natural Earth’s 1:10m reefs distribution (version 4.0.0), derived from WDB2. These are generalised, incomplete reference outlines; Natural Earth documents poor definition in some areas, including the Great Barrier Reef. They do not show current reef condition, water depth or safe passage. Natural Earth data are public domain.

Names are selected by zoom, available space and proximity to the visible hemisphere. A missing label or outline does not imply that a feature is absent. Banks and shoals here are named geographic features, not a complete inventory of underwater hazards. Buoys, lights, wrecks, obstructions, surveyed depths, tides and safe shipping routes are not supplied. Use appropriate current official charts for navigation.

The saved date records retrieval, not an observation, survey or legal effective date. Source versions, scope, adaptation methods and checksums · Reef source and known limitations ↗ · Open Explore ↗.

NOAA Coral Reef Watch v3.1 Degree Heating Week (DHW), provided through NOAA CoastWatch ERDDAP. Native 0.05° grid (about 5 km); daily satellite-derived product. The globe shows the latest verified field, 20 October 2026, with a source reference time of 12:00 UTC. This is a daily date marker, not a claim that every satellite observation occurred at noon. Checked 22 October 2026: the primary NOAA STAR archive and CoastWatch mirror both reach 20 October. A source publication date and retrieval/check time are separate. This layer is a saved snapshot; an automatic production refresh is not yet configured.

DHW combines the intensity and duration of thermal exposure over the preceding 84 days, in °C-weeks. NOAA accumulates daily Coral Bleaching HotSpot values of at least 1 °C above its Maximum Monthly Mean reference, divided by seven. The threshold climatology represents 1985–1990 plus 1993, derived from a longer record and recentered by NOAA. It is different from our OISST temperature-anomaly baseline; the two fields must not be subtracted or interpreted as interchangeable. AO uses the supplied DHW field rather than calculating a new index from OISST.

The numerical snapshot retains every valid native grid value in hundredths, split into four lossless tiles. Land, ice, missing values and values masked by the provider’s declared 0–100 valid range remain unavailable; they are never replaced with a nearby ocean value or clipped into validity. The manifest records excluded water values separately. Selecting the globe reads the nearest native grid cell. Zoom does not add source resolution.

The globe colour overview uses the maximum available native value within each 0.1° display cell (2 × 2 native cells). Devices with smaller texture limits use 0.2° or 0.4° blocks with the same maximum rule. These maxima preserve higher local exposure in the broad overview; they do not change numerical point readings. The selected point can therefore differ from its surrounding overview colour. Entirely missing blocks remain missing. The colour scale saturates at 20 °C-weeks; point readings retain valid higher values.

Heat stress can indicate potential coral bleaching risk. This is not observed bleaching, a mortality estimate, a forecast or an overall ocean-health score. The global numerical field includes ocean cells without coral reefs; an offshore colour does not establish a local coral impact. DHW alone does not determine NOAA’s Bleaching Alert Area category, which also depends on HotSpot and a defined time window. We do not infer those alert categories from DHW. NOAA’s current alert scheme includes Levels 1–5; older mirror category metadata must not truncate it.

Major reef outlines reuse the Natural Earth public-domain reference geography already used by Explore. They are generalised and incomplete, become visible on closer views, and carry no current reef-condition information. Source and coastline masks can differ around small islands. This view does not resolve reef-depth temperatures, local shading, species sensitivity, acclimatisation or other causes of ecological variation.

NOAA product methods ↗ · Numeric data service ↗ · Dataset citation ↗ · Saved date, coverage and checksums. Credit: NOAA Coral Reef Watch. Current CRW products are freely reusable with attribution; NOAA does not endorse this adaptation. Reuse and citation guidance ↗.

NOAA/NCEI Optimum Interpolation Sea Surface Temperature (OISST) v2.1. Daily analysed surface fields on the native global 0.25° grid. This layer uses the preliminary 13 October 2026 analysis, retrieved 15 October 2026. The globe displays the source analysis date, not the retrieval date; noon UTC in the source identifies the daily field, not the time every observation was made. This is a saved snapshot; automatic updates have not yet been configured.

Temperature is sea-surface temperature in °C, separate from air temperature and the heat stored through the water column. Anomaly uses NOAA’s supplied difference from its 1971–2000 daily seasonal climatology, without rebasing or calculating a new anomaly. Blue is cooler, red warmer and zero is neutral. An older baseline is not today’s climate normal. A warm anomaly alone does not establish a marine heatwave, ecological harm or its cause.

The numerical fields share one mask: source land/missing SST or anomaly, and cells with a known NOAA ice-proxy fraction of at least 15%, are excluded from both views. The NOAA ice field helps its SST analysis estimate temperatures near ice; it is not AO’s independent sea-ice concentration layer. Missing ice-proxy values are retained as missing, not changed to zero; otherwise valid SST and anomaly remain available. Grey ocean has no displayed reading and should not be interpreted as cold water or verified absence of ice. Coastline geography can differ from the analysis grid.

AO preserves all native grid-cell values and rotates the longitude ordering into −180° to 180° for display. Tapping a location reads the containing native grid cell; it does not move to a nearby valid cell or interpolate a new value. Display readings round to 0.1 °C. Temperature colours saturate at −2 and 35 °C; anomaly colours at −5 and +5 °C. Readings retain source values beyond these colour limits. Land and ice-masked waters remain excluded while changing views.

NOAA combines satellite, ship and buoy observations into a spatially complete analysis over its valid ocean domain. This is a grid estimate, not a thermometer measurement at the selected point. Preliminary fields may be revised. The source includes an SST analysis-error field; this first global view does not display uncertainty or infer anomaly confidence. It is not intended for navigation or operational decisions.

NOAA product and methods ↗ · Dataset citation ↗ · Source daily NetCDF ↗ · Saved field dates and checksums · Snapshot provenance. NOAA data are publicly available; NOAA does not endorse this adaptation.

FNMOC ESPC-D V02 / HYCOM, distributed by HYCOM.org. The dated saved field uses forecast hour zero, valid at the displayed model run time. The map and saved manifest show the run and valid times; retrieval and the latest refresh check are separate records. These are instantaneous model velocities at 0 m, not daily means, direct measurements or a map of the deep overturning circulation. The model metadata describe its operational status as pre-operational. A daily scheduled source check publishes only validated paired fields. It verifies units, coordinates, depth, time, mask pairing and coverage; rejects older, future or over-72-hour-old source runs; and keeps the previous manifest and field if retrieval or validation fails. Fixed source metadata are checked for revisions before reusing an unchanged run. The refresh-status record exposes success or retained data without changing the scientific timestamp. Fields older than 72 hours are labelled as older saved data on the map.

AO takes every tenth longitude and twentieth latitude from the provider’s regular 0.08° × 0.04° delivery grid, producing a 0.8° display sample. The saved vectors are not averaged or interpolated. Display streamlines use the separate interpolation described below. The delivered grid covers 80°S to 90°N; retained latitude centres end at 89.6°N. The source model’s native grid is finer than this delivery and preview. Land and missing vectors stay missing. Display geography and model coastlines can differ, especially near islands and narrow channels.

Curved streamlines illustrate the instantaneous surface-flow pattern, with arrowheads towards the flow. AO traces short paths through the frozen field in both directions using a spherical midpoint method at 0.2° spatial steps, with bilinear interpolation only where all four surrounding retained grid points are valid. Tracing stops at missing data, weak flow below 0.01 m/s, the coverage limits or 85° latitude; completed small loops are not repeatedly traced. Paths are display samples, not measured tracks or predictions. Traced lines are projected onto the globe and drawn as local quadratic Bezier curves; these short control segments avoid the overshoot of long decorative handles. Lines thin with zoom and screen spacing. Line colour represents local interpolated speed from 0 to 2 m/s, saturating at 2 m/s. Arrowheads show direction; line length and spacing do not represent speed, volume or elapsed time. Gaps can reflect display thinning, weak flow or missing coverage.

Tapping still reads the nearest retained model grid point without interpolating its numerical value or substituting another ocean point. Speed is √(u²+v²); bearing is clockwise from true north, towards the flow. Readings retain speeds above the colour limit. These are instantaneous streamlines, not time-evolving particle trajectories.

The model assimilates observations and is driven by atmospheric forcing. This view does not display uncertainty or separately identify tidal and wave-driven drift components, and is not for navigation. It cannot establish the present movement of plastic: the separately labelled accumulation zones represent a historical modelling study with different dates and methods.

Official product, grid and model documentation ↗ · Public source catalogue ↗ · Saved run, valid time and checksum · Latest refresh check. Source metadata permit public release with unlimited distribution; the catalogue labels the data freely available. Credit: FNMOC / HYCOM.org. No endorsement is implied.

World Bank / IMF World Seaborne Trade Monitoring System (Cerdeiro, Komaromi, Liu and Saeed, 2020). Commercial-vessel AIS position reports accumulated from January 2015 to February 2021, including moving and stationary ships. Retrieved 13 October 2026. This is historical reporting activity, not current routes, vessel counts or measured environmental impact.

AO summed the source grid into a smaller display raster and applied a logarithmic colour and opacity scale. One worldwide scale runs from lighter teal (#7bc8c5) through pale green (#9bcdb4), cream (#efd792) and peach (#e28b60) to red (#b0394e) as recorded activity increases. Cream is a colour transition, not an average or normal baseline. The stored intensity follows a squared logarithmic scale, capped at the 99th percentile of positive cells. A smooth, increasing display curve softens the middle tones so busier corridors stand out; the faintest readings retain their opacity. This styling does not change the source counts or their ordering. There is no lower-percentile cutoff; very small readings remain faint, and the top 1% shares maximum colour and opacity. Angular cells have unequal physical areas, so this is relative activity, not density per square kilometre. Coverage stops near 85° north and south. Blank areas can reflect no reports, weak displayed activity or unavailable coverage.

Shipping shading is clipped to water in the Natural Earth reference map; the underlying activity raster is unchanged. Coastal water and estuaries remain visible where the map resolves them. Narrow rivers and most inland waters are not represented by this land mask, so this view is not a complete inland-waterway map. Clipping removes inland marks from view; it does not validate the remaining AIS reports.

Port names and reference points come from Natural Earth’s 1:10m ports, version 5.0.0 (1,081 records; retrieved 21 October 2026). More names appear as you zoom, using the source’s cartographic scale ranks and spacing to avoid overlapping labels. These ranks do not measure traffic. Locations may be approximate and names historical; this is not a complete current port inventory or a navigation chart. Port markers do not connect individual AIS reports to ports or show live arrivals.

Natural Earth port source and location limitations ↗ · Public-domain terms ↗ · Port layer provenance and checksum.

World Bank source and methods ↗ · CC BY 4.0 licence ↗. AO’s aggregation and display styling are adaptations of the source raster.

Global Wind Power Tracker, Global Energy Monitor, February 2026 release. Retrieved from GEM’s public map data on 13 October 2026. The globe includes 2,085 offshore project-phase locations across 57 countries and areas. A point represents a project phase, not an individual turbine; a farm can have several phases.

GEM principally tracks phases above 10 MW, with some smaller projects included. Coverage is incomplete and most locations in this subset are approximate. Operating, construction, announced, pre-construction, shelved, cancelled and retired records remain distinct. The release date is not a common observation date or a current operational check; project facts have different dates. A start year may be actual or planned, depending on status. Capacity, where given, is phase nameplate capacity rather than measured generation or environmental impact.

The source’s exact/approximate location flag is retained. “Exact” is the provider’s classification, not an AO survey or a legal boundary. Country and area labels follow the publisher’s economic contexts rather than strict political boundaries. This subset comes from GEM’s separately published public map data; its totals can differ from tracker headline figures. Shared locations and legitimate phases are retained.

AO selected offshore installation types, retained source status and location-accuracy fields, normalized the records and rounded coordinates to six decimals. © Global Energy Monitor. Tracker, methods and citation ↗ · GEM reuse terms ↗ · CC BY 4.0 licence ↗.

Nyadjro, Ebenezer; Webster, Jennifer; Boyer, Tim; Cebrian, Just; Collazo, Leonard; Kaltenberger, Gunnar; Larsen, Kirsten; Lau, Yee; Mickle, Paul; Toft, Tiffany; Wang, Zhankun (2025). The NOAA NCEI Global Marine Microplastics Harmonized Database during 1972-04-20 to present, NCEI Accession 0306475. NOAA National Centers for Environmental Information. AO’s subset contains 16,988 ocean-water measurements in particles per cubic metre, collected from 20 April 1972 to 23 October 2024. Accessed 13 October 2026. The default selection is 2020–2024. NOAA updates its archive quarterly; this view is a saved snapshot.

AO selected valid dated ocean-water records in the same units, retaining zero readings, sampling method, mesh size, depth and original-study links. Beach and sediment samples are excluded. No interpolation, averaging, unit conversion or concentration classification is added. Globe readings use up to three significant digits for readability, with ≈ marking rounded values; selected records on this page retain the supplied precision. Equal pink dots mark samples; outlined groups collect nearby samples for readability. Size and colour do not encode concentration.

Sampling is uneven. A gap does not mean clean water, and a zero result does not show that surrounding water is plastic-free. Methods, depths and mesh sizes differ; matching units alone do not make studies directly comparable. Select a sample to open its original study and read its methods.

Dataset citation ↗ · NOAA metadata and reuse terms ↗ · Methods and limitations ↗ · CC0 1.0 public-domain dedication ↗. Source links and credits are retained; NOAA does not endorse this adaptation.

NOAA/NCEI OISST v2.1, supplied via NOAA CoastWatch ERDDAP. ANALYSIS · °C · native 0.25° grid · daily surface fields. Atlantic map coverage is 65°S–75°N, 85°W–25°E. Preliminary values can be revised. Saved daily dates appear in the map time controls.

SST anomaly is the difference from NOAA’s 1971–2000 daily climatology. Positive values are warmer; negative values are cooler. Colours saturate at the legend endpoints; point readings retain the actual value. Temperature and anomaly use different scales.

NOAA product and methods ↗

The air toggle uses Copernicus Climate Change Service / ECMWF ERA5 reanalysis at 2 metres above land, ocean and sea ice. It is a grid-cell estimate combining observations and modelling, not a local thermometer reading. The map displays its own valid UTC day. A separate daily updater retrieves saved public fields directly from the Copernicus Climate Data Store; page loading does not start a download.

Temperature is the average of all 24 hourly records in a UTC day. The anomaly subtracts the average for the same calendar day over 1991–2020, using the same ERA5 variable and grid. The Copernicus daily statistics service averages hourly records and removes partial periods. The historical normal requires all 30 requested yearly daily fields; the updater checks their dates, grid and coverage before publication. February 29 uses the mean of the February 28 and March 1 normals.

The regular 0.25° source grid is sampled every 0.5° for display. Colour shading blends adjacent valid grid-node colours. Point readings use the nearest sampled source node; missing data is not filled. Screen shading and zoom do not add scientific resolution. Missing data remains unavailable. Blue denotes cooler and red warmer than the historical mean. Anomaly colours stop at ±10 °C; numeric readings retain the actual value. No separate anomaly uncertainty is supplied.

Reanalysis arrives after a delay; recent ERA5 values may be preliminary and revised. The air map keeps its own date, which can differ from the ocean and ice layers. A single daily anomaly does not establish a long-term trend. Forecasting is not included in this air comparison.

Contains modified Copernicus Climate Change Service information (2026). Official ERA5 source, DOI and licence ↗ · Copernicus daily statistics, DOI and CC-BY 4.0 licence ↗. Contains modified Copernicus Climate Change Service information. The European Commission and ECMWF are not responsible for use of this information. Daily source processing retains the previous snapshot if retrieval or validation fails. The map flags daily air records older than fourteen days. Visitors receive saved public data and never receive the download credential.

FNMOC ESPC-D V02 / HYCOM · MODELLED, with INITIAL STATE or FORECAST as appropriate. Velocity at 0 m, in m/s; speed and direction are DERIVED from eastward and northward components. Source output spacing is 0.08° longitude × 0.04° latitude, sampled at 0.32° over the Atlantic map domain.

This map shows modelled surface currents from HYCOM. All available times come from the same model run. Missing times are left out, and areas without data remain blank.

Streamlines show the direction the water is moving. This map is for exploration and learning, not navigation. Surface speed does not measure the full-depth Atlantic overturning circulation (AMOC).

The blue ocean with static grain is reference geography, matched to the globe; dark land retains its geographic mask. Flow lines and wave marks are clipped to the reference water area. This display clipping does not change the underlying numerical vectors; reference and model coastlines can differ. Line colour represents current speed from 0 to 2 m/s, saturating at 2 m/s; point readings retain higher values. AO traces short instantaneous streamlines through the displayed frame, using midpoint steps in the map projection and bilinear interpolation of eastward/northward components only where all four surrounding saved nodes are valid. Local quadratic Bezier segments smooth the display without joining gaps. Arrowheads point toward flow. Line length, spacing and density do not show speed, volume or travel time, and are not particle trajectories. Lines stop at weak flow, missing coverage or the map boundary; the saved point readings and their numerical sampling are unchanged. Currents below 0.02 m/s have no displayed streamline. Blue water without a line can reflect thinning, calm or missing coverage; it is not a zero-current reading. No interpolation is made between forecast times.

HYCOM data service ↗

Surface wind is an ECMWF IFS forecast at 10 metres above the surface. It differs from the Jet stream layer, which shows winds at 250 hPa high in the atmosphere. Both are modelled forecasts.

Matched eastward (10u) and northward (10v) components come from the same model run and forecast time. The 0.25° source is sampled at every other original node for a 0.5° display grid, covering 65°S–75°N and 85°W–25°E. Components are stored to 0.01 m/s; this is encoding precision, not a measure of forecast accuracy.

Nine saved forecast times span 0–48 hours in six-hour steps from the 00 UTC run. The selected valid time, model issue time and retrieval time are separate. “Latest available” chooses the saved forecast closest to the present. The scheduled updater retains the last validated snapshot if a source fails or a candidate is incomplete; older runs are labelled stale.

Speed is the magnitude of the two wind components, in m/s. Single-colour curved lines and arrowheads show travel toward; the selected direction reports wind from. Colour encodes speed on the existing 0–30 m/s scale, capped at 30; numerical readings retain higher values. AO traces short instantaneous streamlines through the displayed frame, using midpoint steps in the map projection and bilinear interpolation of eastward/northward components only where all four surrounding saved nodes are valid. Local quadratic Bezier segments smooth the display without joining gaps. Arrowheads point toward flow. Line length, spacing and density do not show speed, volume or travel time, and are not particle trajectories. Lines stop at weak flow, missing coverage or the map boundary; the saved point readings and their numerical sampling are unchanged. Winds below 0.5 m/s have no streamline.

These are model grid values, not station observations or gust forecasts. Coastal and local terrain effects may be unresolved. Zoom enlarges the saved grid rather than adding source detail. Use official forecasts for decisions at sea.

Contains modified ECMWF forecast data (2026), CC BY 4.0. ECMWF Open Data source and documentation ↗.

Expanded Surface wind loads one 0.4° forecast time on demand and caches it while the page is open. Eastward and northward wind components are interpolated from the full 0.25° source grid before deriving speed and direction. The normal map uses 0.5° sampling. Both views come from the same model run and matched source fields. If finer detail cannot be validated, the overview remains visible and is labelled 0.5°. Small coastal and terrain effects may still be finer than the source grid.

ECMWF IFS · MODELLED. The jet map uses 250 hPa wind, with speed and direction DERIVED from U/V components, in m/s. Source grid: 0.25°; displayed vectors: 0.5°. Domain maximum uses all source nodes, including land. A fixed pressure level does not capture every jet at every altitude.

The wind map uses a custom lavender-to-blue-purple scale with a wider light-to-dark range, informed by a visual reference, blending pale lavender for lower speeds into deep blue-purple for higher speeds. The colours progress from light to dark; they show speed, not temperature, heat transport or good/bad conditions. The jet map uses a fixed 0–100 m/s scale, capped at 100; point readings retain their values.

Single-colour curved lines and arrowheads show wind travel toward at 250 hPa. The existing raster colour encodes speed from 0 to 100 m/s; line length is not a speed scale. AO traces short instantaneous streamlines through the displayed frame, using midpoint steps in the map projection and bilinear interpolation of eastward/northward components only where all four surrounding saved nodes are valid. Local quadratic Bezier segments smooth the display without joining gaps. Arrowheads point toward flow. Line length, spacing and density do not show speed, volume or travel time, and are not particle trajectories. Lines stop at weak flow, missing coverage or the map boundary; the saved point readings and their numerical sampling are unchanged. Wind below 5 m/s has no jet streamline.

The main air view uses matched ERA5 UTC daily temperature and anomaly, as described in daily air methods. If ERA5 is unavailable, the separately labelled fallback temperature snapshot uses ECMWF IFS modelled air at 2 m above land and ocean, converted from kelvin to °C. It supplies no air anomaly. Source grid: 0.25°; display sampling: 0.5°. One daily initial state is saved; no air-temperature forecast sequence is shown. This is not a station observation or sea-surface temperature. Jet forecasts retain separate run and valid times; available saved frames appear in the controls.

Contains modified ECMWF forecast data. ECMWF source and licence ↗

NOAA GFS-Wave · MODELLED initial state and forecasts. Source resolution 0.25°; display sampling 0.5°. Height is in metres and mean swell period in seconds. Single-colour curved crest marks are perpendicular to swell propagation; arrowheads show travel toward, converted from the source FROM convention. These local marks describe swell partition 1 only, including when combined height or mean period is the selected colour field. Fixed mark size, length and spacing do not encode height, period, speed or energy. No propagation ray or water-particle trajectory is inferred. For visual placement, unit direction components are interpolated only where all four surrounding nodes have valid direction and partition 1 height at least 0.05 m; opposing, ambiguous directions are omitted. Point readings keep the original nearest-grid values and FROM bearing.

The wave view uses NOAA GFS-Wave. Combined significant height includes wind waves and swell; swell height, mean period and direction refer to partition 1 only. No total-swell, wave-energy or coastal breaking-wave value is inferred. Dark ocean cells indicate unavailable or source-masked data. Forecasts from separate models retain separate runs and valid times.

NOAA NOMADS source service ↗

Copernicus Marine / ACRI-ST GlobColour · satellite-derived estimates. Product OCEANCOLOUR_GLO_BGC_L3_NRT_009_101; dataset cmems_obs-oc_glo_bgc-plankton_nrt_l3-multi-4km_P1D. Chlorophyll-a is shown in mg/m³ over the Atlantic map domain. This is a seven-day observation window, not a forecast.

Copernicus Marine / ACRI-ST GlobColour merges satellite ocean-colour observations to estimate chlorophyll-a, a pigment used by phytoplankton. Clouds, ice and low light can leave gaps. Coastal sediments and coloured dissolved material can affect estimates.

Our view averages available daily estimates over seven consecutive source dates. Different locations may represent different days within that window; select a point to see how many daily values contributed. No gap filling is applied here. The source grid is approximately 4 km; the displayed grid samples it every 0.125°. Zooming does not add detail.

This layer does not measure species diversity, fish stocks, harmful algae or the health of the ocean. It is separate from the weather forecasts and historical records above.

Contains modified Copernicus Marine Service information. Source product and documentation ↗

Copernicus Marine / Mercator Ocean GLO12 · product GLOBAL_ANALYSISFORECAST_PHY_001_024 · MODELLED daily means. Source grid 1/12°; displayed sampling 0.25° across 65°S–75°N, 85°W–25°E. Salinity is shown in PSU and current speed in m/s. Saved depths: 0.49402499198913574, 47.37369155883789, 92.3260726928711 m; select the available date and model depth in the explorer.

Datasets: cmems_mod_glo_phy-so_anfc_0.083deg_P1D-m and cmems_mod_glo_phy-cur_anfc_0.083deg_P1D-m. Analysis/forecast collection; no individual model issue time is supplied here. Retrieval is not the model issue time. These daily means are separate from HYCOM surface currents. Single-colour curved lines and arrowheads show current direction only; the existing raster colour encodes daily mean speed at the selected depth. AO traces short instantaneous streamlines through the displayed frame, using midpoint steps in the map projection and bilinear interpolation of eastward/northward components only where all four surrounding saved nodes are valid. Local quadratic Bezier segments smooth the display without joining gaps. Arrowheads point toward flow. Line length, spacing and density do not show speed, volume or travel time, and are not particle trajectories. Lines stop at weak flow, missing coverage or the map boundary; the saved point readings and their numerical sampling are unchanged. Currents below 0.02 m/s are omitted. Salinity has no streamlines. Missing values remain unavailable; zoom does not increase source resolution.

Contains modified Copernicus Marine Service information. Product, methods and limitations ↗

The optional Protected areas overlay in Atlantic → Geography and Satellite uses salmon outlines and transparent salmon shading. It is off by default. The selected sources cover parts of the North and South Atlantic, including the Gulf and US Caribbean. This is not a complete Atlantic inventory. Unshaded water does not mean unprotected water.

Coverage. Canadian Oceans Act sites include the Gully, St. Anns Bank and Laurentian Channel. The US inventory adds Atlantic, Gulf and US Caribbean sites. Brazil contributes federal coastal/marine conservation units, including marine national parks and archipelagos. South Georgia and the South Sandwich Islands use their official MPA outline. Ascension and St Helena use clearly labelled reference outlines derived from whole-EEZ designations. The OSPAR network remains a historical North-East Atlantic source.

Remaining gaps. Canadian sites under other Acts, non-US Caribbean and Mexican sites, other South American jurisdictions and African Atlantic jurisdictions are not comprehensively mapped. Tristan da Cunha has a large protection zone, but its exact reusable boundary has not been verified for this overlay. Its whole EEZ must not be shaded as no-take: inshore and seamount fishing zones are excluded. Tristan government management plan ↗. The global Protected Planet database is not redistributed here because its reuse conditions differ from these sources.

Dates. Retrieved 9 October 2026. Retrieval is separate from a source release, designation or legal effective date. These are dated reference records, not daily observations or fortnightly changes. A revised catalogue date is not proof of every boundary’s effective date. OSPAR retains its 2021 revision 003 release; NOAA’s hosted item calls itself the 2024 inventory while its service URL retains 2023.

What protection means. A designation does not establish no-take status, current permitted activities, enforcement or ecological outcomes. Some sites include land or overlap. Site details reproduce fishing restrictions only where supplied and identify them as dated source attributes. Check the authority’s current regulations before interpreting restrictions. The South Georgia outer MPA is not entirely no-take.

Display method. Added polygons are simplified with topology preservation at a 0.005-degree tolerance and coordinates rounded to five decimals. Holes are retained; Canadian management zones are dissolved into source-named sites. NOAA geometry was requested with 0.005-degree generalisation. The original simplified OSPAR trial is retained: 558 designated sites are drawn and 41 proposed records are excluded. Source coordinates for Brazil and South Georgia are transformed to WGS84. Approximate display boundaries are not navigation charts or authoritative legal geometry. Zoom adds no boundary accuracy.

Area readings. Reported marine area is shown only where supplied. Brazil’s total site area is labelled separately and can include land. Canadian zone figures are not summed into a misleading site total. No marine area is inferred from an outline or a satellite image. The backgrounds retain their own dates: GEBCO 2026 relief and a 2024 satellite mosaic.

Credits. Attributes have been reduced and geometry simplified for display. Each source retains its own reuse terms; no provider endorsement is implied. The combined file does not replace those terms.

OSPAR Marine Protected Areas Network. OSPAR / ODIMS Historical revised release, not proof of current protection. Includes some land; proposed sites are not drawn.

DFO Oceans Act marine protected areas. Fisheries and Oceans Canada. Contains information licensed under the Open Government Licence – Canada. Does not include every Canadian protected area, other Acts or areas of interest. Zones dissolved; zone areas are not summed. Regulations determine legal boundaries.

NOAA Marine Protected Areas Inventory. NOAA National Marine Protected Areas Center Dated inventory, not proof of current rules. Source polygons can include terrestrial portions. NOAA generalisation was requested at 0.005 degrees; site areas are source attributes, not recalculated.

ICMBio federal conservation units. ICMBio, Brazil · federal conservation-unit boundaries, August 2026 Not all state or municipal sites. Some polygons include land and coastal habitats; total site area is not marine area. SIRGAS 2000 transformed to WGS84. Biome tags do not establish fishing restrictions.

South Georgia and South Sandwich Islands MPA. Hogg, O., Gerrish, L., & Belchier, M. (2026). South Georgia and the South Sandwich Islands Marine Protected Area boundaries, 2025 update (v1.0). NERC EDS UK Polar Data Centre. Legal measures effective 22 April 2025. The outer MPA is not entirely no-take; separate fishing zones are not drawn. Custom Albers equal-area coordinates transformed to WGS84.

Marine Regions maritime boundaries. Flanders Marine Institute (2023). Maritime Boundaries Geodatabase: Maritime Boundaries and Exclusive Economic Zones (200NM), version 12. DOI: 10.14284/632. Derived display outline, not an official MPA legal boundary. EEZ source does not itself confer protection. Tristan is excluded because its protection zone is not the whole EEZ.

Boundary snapshot & source provenance (JSON, 7.2 MB) · Return to the Atlantic map

The Geography tabs use saved shaded-relief imagery reproduced from the GEBCO_2026 Grid, retrieved on 30 September 2026. The Atlantic image covers 85°W–25°E and 65°S–75°N. The Arctic image is reprojected to the existing polar stereographic view. Both use the land and ice-surface elevation version: Greenland shows its ice surface, not the bed beneath it.

GEBCO is a global terrain model at 15 arc-second grid intervals. It combines surveyed depths with estimated and interpolated areas; grid spacing is not measurement accuracy or evidence that every location has been surveyed. Geographic labels are orientation aids. The saved display images are lower-resolution representations of the source grid; zooming does not add source resolution.

This reference imagery does not refresh with weather or sea-ice snapshots, does not show current ice cover, and supplies no numerical point-depth reading. Updated GEBCO editions require a reviewed reference-image replacement. The separate Atlantic seabed section retains its EMODnet depth queries and regional detail. Neither view is for navigation or safety at sea.

GEBCO Bathymetric Compilation Group 2026 (2026). The GEBCO_2026 Grid — a continuous terrain model for oceans and land at 15 arc-second intervals. NERC EDS British Oceanographic Data Centre NOC. Dataset citation ↗ · Source, methods and terms of use ↗

The GEBCO grid is public domain, with source acknowledgement required. The imagery has been adapted for these maps; no endorsement is implied. Atlantic reference provenance ↗ · Arctic reference provenance ↗

EMODnet DTM 2024 provides detailed regional bathymetry at approximately 115 m grid spacing. Some areas use GEBCO estimates; grid spacing is not survey accuracy. The global relief background uses an older EMODnet / GEBCO compilation. Zooming loads additional imagery, up to the source detail available.

Depth readings come from the EMODnet mean-depth service, not our former 0.1° sampled grid. They are rounded to a metre for display; this does not imply metre accuracy. EMODnet depths are referenced to Lowest Astronomical Tide where available; source references vary, including in GEBCO-filled areas. These are compiled reference maps, not live soundings or navigation charts.

BGS sediment mapping covers the UK Continental Shelf at 1:250,000 scale. Blank areas are unclassified in this layer. Map services and point readings require an internet connection; an unavailable service is reported separately from missing coverage.

EMODnet tile edition: 2024. Service configuration checked 16 September 2026. Survey dates vary. Contains British Geological Survey materials © UKRI 2026.

EMODnet depth sources ↗ · BGS offshore viewer ↗ · BGS data & licence ↗

The global background is EMODnet World Base Layer (2020 service; documented inputs include EMODnet 2018 and GEBCO 2019). Genuine EMODnet contours and IHO-IOC GEBCO Gazetteer labels load on demand. Outside mean-depth query coverage, no numeric depth is inferred from the picture. Place labels are orientation aids, not sampling locations.

Modelled wind 10 metres above the surface, including land, ocean and sea ice. ECMWF IFS medium-range control forecast, matched eastward (10u) and northward (10v) components from a single 00 UTC run. Saved times run from lead 0 to 48 hours in six-hour steps. Every frame retains its own valid time and model run; past valid times remain predictions from that run. Lead zero is a model initial state, not a station observation. These are wind-speed fields, not gusts or 250 hPa jet-stream winds.

Source grid: 0.25°; saved display sampling: 0.5°. Point readings keep the nearest saved node and its coordinates. Meteorological bearing reports wind from. Single-colour curved lines and arrowheads show travel toward, rotating physical east/north components into the Arctic polar stereographic view centred on 90°N, 45°W. AO traces short instantaneous streamlines through the displayed frame, using midpoint steps in the map projection and bilinear interpolation of eastward/northward components only where all four surrounding saved nodes are valid. Local quadratic Bezier segments smooth the display without joining gaps. Arrowheads point toward flow. Line length, spacing and density do not show speed, volume or travel time, and are not particle trajectories. Lines stop at weak flow, missing coverage or the map boundary; the saved point readings and their numerical sampling are unchanged. The colour field retains its existing 0–30 m/s wind-speed scale. Wind below 0.5 m/s and within 0.5° of the pole has no streamline; bearing remains unavailable at the pole or below 0.1 m/s.

Map colour uses a custom lavender-to-blue-purple scale with a wider light-to-dark range, informed by a visual reference, with pale lavender for lower speeds and deep blue-purple for higher speeds. Colours show speed, not temperature or good/bad conditions. The fixed scale runs from 0 to 30 m/s, capped at 30 m/s; point values retain the saved component-derived speed. The colour map blends the saved components between adjacent nodes for readability; this visual interpolation adds no observations or source resolution. Point values retain nearest-node sampling. All four nodes must be valid; missing data are not filled. No uncertainty estimate is supplied. This wind layer does not show sea-ice drift, currents or wave height. The ice-class boundary is hidden on this layer so that different product dates are not mixed.

The map serves a validated saved forecast. A separate source check runs alongside the Atlantic data updates; failed validation retains the previous forecast. Opening or reloading the map does not fetch a new model run. Runs older than 36 hours carry an older-forecast notice; missing or invalid data remain unavailable while the other layers continue working. The builder validates matched metadata, Earth-relative vectors, grid coordinates and quantisation before replacing a saved snapshot.

Contains modified ECMWF forecast data (2026), sampled and reprojected by Atlantic Observatory. ECMWF Open Data, source and terms · CC BY 4.0. The saved file includes retrieval time, source byte ranges and SHA-256 checksums: wind snapshot and provenance. For exploration and learning, not navigation.

Modelled wind 10 metres above the surface, including land, ocean and sea ice. ECMWF IFS medium-range control forecast, matched eastward (10u) and northward (10v) components from a single 00 UTC run. Saved times run from lead 0 to 48 hours in six-hour steps. Every frame retains its own valid time and model run; past valid times remain predictions from that run. Lead zero is a model initial state, not a station observation. These are wind-speed fields, not gusts or 250 hPa jet-stream winds.

Source grid: 0.25°; saved display sampling: 0.5°. Point readings keep the nearest saved node and its coordinates. Meteorological bearing reports wind from. Single-colour curved lines and arrowheads show travel toward, rotating physical east/north components into the Antarctic EPSG:3031 stereographic view centred on 90°S, 0°E. AO traces short instantaneous streamlines through the displayed frame, using midpoint steps in the map projection and bilinear interpolation of eastward/northward components only where all four surrounding saved nodes are valid. Local quadratic Bezier segments smooth the display without joining gaps. Arrowheads point toward flow. Line length, spacing and density do not show speed, volume or travel time, and are not particle trajectories. Lines stop at weak flow, missing coverage or the map boundary; the saved point readings and their numerical sampling are unchanged. The colour field retains its existing 0–30 m/s wind-speed scale. Wind below 0.5 m/s and within 0.5° of the pole has no streamline; bearing remains unavailable at the pole or below 0.1 m/s.

Map colour uses a custom lavender-to-blue-purple scale with a wider light-to-dark range, informed by a visual reference, with pale lavender for lower speeds and deep blue-purple for higher speeds. Colours show speed, not temperature or good/bad conditions. The fixed scale runs from 0 to 30 m/s, capped at 30 m/s; point values retain the saved component-derived speed. The colour map blends the saved components between adjacent nodes for readability; this visual interpolation adds no observations or source resolution. Point values retain nearest-node sampling. All four nodes must be valid; missing data are not filled. No uncertainty estimate is supplied. This wind layer does not show sea-ice drift, currents or wave height. The ice-class boundary is hidden on this layer so that different product dates are not mixed.

The map serves a validated saved forecast. A separate source check runs alongside the Atlantic data updates; failed validation retains the previous forecast. Opening or reloading the map does not fetch a new model run. Runs older than 36 hours carry an older-forecast notice; missing or invalid data remain unavailable while the other layers continue working. The builder validates matched metadata, Earth-relative vectors, grid coordinates and quantisation before replacing a saved snapshot.

Contains modified ECMWF forecast data (2026), sampled and reprojected by Atlantic Observatory. ECMWF Open Data, source and terms · CC BY 4.0. The saved file includes retrieval time, source byte ranges and SHA-256 checksums: wind snapshot and provenance. For exploration and learning, not navigation.

BAS is the default reference map. The Geography toggle switches to saved shaded-relief imagery reproduced from the GEBCO_2026 Grid, retrieved on 1 October 2026. Both views retain the same EPSG:3031 projection, selected location and curated place/station markers.

The BAS Antarctica and the Southern Ocean map includes ADD v7.11 coastline, REMA v2 land hillshade and IBCSO v2 bathymetry. Reference tiles load from its ArcGIS service; unavailable tiles are reported. BAS map and full credits ↗ · Open Government Licence 3.0 ↗

The GEBCO alternative uses standard land and ice-surface elevation, not under-ice topography. Its saved 1,000 × 1,000-pixel image is sampled by nearest neighbour from WMS imagery into the existing polar view. Shading shows terrain context and supplies no numeric point depths. Zoom enlarges the saved image; it does not add survey resolution. Source quality and coverage vary. This reference image does not refresh with weather or sea-ice snapshots and is not for navigation.

Imagery reproduced from the GEBCO_2026 Grid. GEBCO Bathymetric Compilation Group 2026 (2026), NERC EDS British Oceanographic Data Centre NOC. Dataset and attribution ↗ · GEBCO imagery service ↗. Place and research-station references retain their individual original source links; they do not indicate current station occupancy.

Daily extent and retrieval dates are shown beside each hemisphere’s published record. Monthly records begin in November 1978 and end at the latest published complete month in the saved record. The selected month and year appear beside each chart. Coverage and missing years vary by calendar month.

NOAA/NSIDC Sea Ice Index v4 (G02135) · satellite-derived estimates · both hemispheres. Daily extent is in million km², counting ocean grid cells with at least 15% sea-ice concentration. It does not measure ice thickness, volume, land-ice loss or sea-level rise. Recent daily values can be revised.

Seasonal charts compare completed monthly means with each calendar month’s 1981–2010 average. Reference means use available published years; their counts appear in the monthly table. Missing months are not extrapolated. Historical charts compare the same calendar month across years, not exact annual minima or maxima.

Daily and monthly images use the 25 km source grid. The orange edge is a spatial median for 1981–2010, a different statistic from average total extent. Historical comparison images load directly from NOAA/NSIDC and follow the chosen month, year, reference year and extent/concentration setting. Missing images are not estimated.

Dataset and citation ↗ · Data and image archive ↗

Antarctic daily chart

The Antarctic seasonal chart has Daily and Monthly views. Daily extent uses accepted NOAA/NSIDC Sea Ice Index v4 observations with no reported missing area. The daily average and shaded historical range use available observations for the same calendar day in 1981–2010. The range is the mean plus or minus two sample standard deviations; it describes historical variation, not measurement error or a confidence interval.

The Anomaly view divides the difference between daily extent and that day’s reference mean by the sample standard deviation. Zero is the reference mean; negative values show less extent. At least 20 accepted reference years and a positive standard deviation are required. The daily table shows the available count. February 29 has insufficient reference years, so its historical comparison and anomaly remain unavailable. Other missing dates remain gaps; recent daily values can be revised. Daily observations are not substituted for incomplete monthly means.

Download the Antarctic daily chart record and reference statistics · NSIDC daily source CSV

The sea-ice reading beside the map

The large extent and percentage describe the latest accepted daily record for the whole hemisphere, separate from the selected map location and the dates of individual map layers. The small chart is monthly: it uses completed, published monthly means for the year of that daily record. It is not a daily curve. Selecting its link opens the matching Monthly view. No value is drawn for an incomplete or missing month.

Chart dates describe the observations; retrieval dates in the downloadable records describe when AO saved the source. The Antarctic daily chart record includes its retrieval timestamp, source checksum, accepted observations by year and same-calendar-day reference statistics. Its 1981–2010 baseline and NOAA/NSIDC source differ from charts using JAXA data or a 1991–2020 baseline, so their percentages and standard-deviation anomalies should not be treated as identical.

Daily historical comparison

The daily range is the same calendar day’s 1981–2010 mean plus or minus two sample standard deviations, using available daily records; at least seven valid reference years are required. Below, above and within refer to this range for the whole hemisphere, not ice at a selected map location. The percentage compares the latest extent with the same-day mean. A single day does not establish ice health or a long-term trend.

Download the saved reference statistics · Download the saved daily and monthly records

Satellite imagery: NASA Global Imagery Browse Services (GIBS), part of NASA’s Earth Science Data and Information System (ESDIS). NOAA-20 VIIRS corrected reflectance, true colour; daily mosaics, not sea-ice measurements. NASA imagery documentation.

Further-reading sources: NERC Arctic Office reuse policy (Open Government Licence v3, except where stated); SCAR Antarctic Digital Database (CC BY 4.0; dataset-specific citations); Bedmap3 data, Pritchard et al. (2024), version 1.0 and Pritchard et al. (2025), Bedmap3 publication. These links open external reference resources; the separate Antarctic Geography layer is documented above.

Sources for the landscape illustration and hidden-lakes story on our Antarctic page. These are published research references, separate from the map’s dated weather, ocean and sea-ice layers.

Bedmap3: the hidden landscape

Bed height, ice-surface height and ice thickness, with uncertainty and survey-coverage grids. Antarctica south of 60°S · 500 m grid spacing · heights and thickness in metres · GeoTIFF and NetCDF.

Compiled from surveys collected over many years, with estimates between measurements. Grid spacing is not the same as measurement accuracy; uncertainty varies across the continent.

Dataset and downloads ↗ · Research paper ↗

Data: Pritchard, H., Fretwell, P., Fremand, A. et al. (2024), version 1.0, NERC EDS UK Polar Data Centre. Paper: Pritchard, H. D., Fretwell, P. T., Fremand, A. C. et al. (2025), Scientific Data 12, 414. Both use CC BY 4.0; cite the dataset and paper when reusing the data.

Thwaites Glacier: water beneath the ice

A study of a lake-drainage event in 2013 combines CryoSat-2 measurements of ice-height change with ice-velocity and ocean observations, and glacier–ocean modelling. The observations and simulations have different coverage and units; see the paper and its source-data files.

Study and source data ↗ · ESA’s explanation ↗

Gourmelen, N., Jakob, L., Holland, P. R. et al. (2025), Nature Communications 16, 2272. Published 6 March 2025. The paper is CC BY 4.0. Its Data availability section links the original datasets, each with its own reuse terms. We link to these files at their source.

NOAA/NSIDC Sea Ice Index, Version 4 (G02135). Satellite-derived estimates; recent daily values can be revised. Maps and daily values have their own dates; monthly charts end at the latest published complete month.

We calculate each monthly reference average from available published values for 1981–2010; the monthly table shows how many years contribute. Missing values remain gaps. The map’s orange edge is a spatial median; it is a different statistic from the charts’ average total extent.

A scheduled check runs daily. Daily maps and extent are refreshed when available; monthly records include published complete months only. If a download fails, the last validated snapshot remains visible. Reloading this page loads the published snapshot; it does not trigger a new source download.

Dataset and citation ↗ · Data archive ↗ · Arctic daily source CSV ↗ · Antarctic daily source CSV ↗

Concentration maps and historical monthly comparisons: NOAA/NSIDC Sea Ice Index v4, 25 km grid. Historical map images are requested directly from the source archive.

Data citation: Fetterer, F., Knowles, K., Meier, W. N., Savoie, M., Windnagel, A. K. & Stafford, T. (2025). Sea Ice Index, Version 4. NSIDC. doi:10.7265/a98x-0f50. Subset: both hemispheres, daily extent and monthly extent records.

Further-reading sources: NERC Arctic Office reuse policy (Open Government Licence v3, except where stated); SCAR Antarctic Digital Database (CC BY 4.0; dataset-specific citations); Bedmap3 data, Pritchard et al. (2024), version 1.0 and Pritchard et al. (2025), Bedmap3 publication. Arctic resources link to external references. The Antarctic explorer’s Geography layer uses the BAS reference map and a curated place and station list; its map disclosure provides credits and individual reference sources.

Antarctic landscape and hidden-lakes sources: Research, data and reuse terms; illustration and video credits.

Photographs: Matthew Stephenson, Colin Ruthven and Bob Brewer / Unsplash. Illustrative photography; not satellite observations or measurements of sea-ice extent.

Educational context: Why sea ice matters ↗. Sea-ice maps are not navigation or ice-safety guidance.

The Greenland feature now focuses on the ice sheet and its glaciers. Ice motion uses the source’s polar stereographic projection, EPSG:3413. See ice-velocity sources and methods below for observation dates, sampling, uncertainty and units. Ocean, floating sea ice and air conditions remain on the Arctic map.

The velocity preview is a reviewed, dated mosaic. No daily velocity update is claimed. The annual mass-balance record retains its existing source and update process; its period and date remain beside the figures.

The ice-sheet overview uses the GEUS/PROMICE mass-balance product, Mankoff et al. (2021), source dataset under CC BY 4.0. It shows complete hydrological years (1 September–31 August), with the source’s uncertainties. Annual net loss is compared with 1990/91–2019/20; the historical range is the historical mean plus or minus two sample standard deviations. The percentage compares annual net loss with mean annual net loss; it is not the percentage of the ice sheet lost. These whole-sheet estimates do not measure local melt, thickness change or today’s glacier front.

Land, coastline and generalized glaciated-area outlines use Natural Earth, public domain. The ice outline is a geographic reference, not a current glacier front. Approximate glacier markers link to NASA Earth Observatory or ESA background information. The satellite view retains the EOX reference mosaic and its attribution.

The Greenland map uses the PROMICE Edition 5 velocity mosaic for 16 August–9 September 2026, derived from Copernicus Sentinel-1 radar measurements. This is a dated 24-day composite, not live imagery or a forecast.

Source resolution is 200 m in EPSG:3413. The whole-island speed preview samples source cells at 2 km spacing. Sermeq Kujalleq, Petermann and Helheim previews retain native 200 m cells. Point readings use the original sampled values and source speed error estimate. Missing values are not filled.

Speeds supplied in metres per day are multiplied by 365.25 to display metres per year equivalent. This expresses a rate; it does not claim a full-year average. Flow lines follow measured horizontal velocity components, spatially smoothed for legibility over approximately 5 km in the overview and 1 km in the glacier views. Directions below twice the source speed error estimate are omitted. Very slow motion is omitted from the illustrative line layer. The paths illustrate a spatial vector field, not observed trajectories of individual ice particles.

Natural Earth’s generalized glaciated-area outlines and land geography provide orientation; they are not contemporary glacier fronts. Surface-speed shading is restricted to this generalized ice outline to separate it from exposed rock and ocean; some measured outlet flow may extend beyond it. The satellite option remains the separately credited 2024 EOX reference mosaic. Ice speed, annual mass balance and melt extent are distinct measurements.

Ice velocity maps were produced as part of the Programme for Monitoring of the Greenland Ice Sheet (PROMICE) using Copernicus Sentinel-1 SAR images distributed by ESA, and were provided by the Geological Survey of Denmark and Greenland (GEUS).

Solgaard & Kusk · Greenland Ice Velocity from Sentinel-1 Edition 5 · GEUS · CC BY 4.0 ↗ · Solgaard et al. 2021 · scientific methods ↗

For Sermeq Kujalleq, Petermann and Helheim, the Observatory compares fixed inland 1 km squares using PROMICE Edition 5. One mosaic is selected per year with its start date nearest 16 August. The 2017–2025 reference windows start between 10 and 21 August; the latest period is 16 August–9 September 2026. Exact windows and native source filenames are retained with each observation. The additional 2016 point is shown where all common cells were observed; incomplete coverage is a chart gap and is excluded from the reference average.

Only identical 200 m source cells with valid speed and source error in every reference year and the latest period contribute to the area mean. At least 20 of 25 cells must pass. Coordinates and cell counts are retained in the downloadable comparison record; a square marks the reference area when zoomed in. These are Observatory reference samples, not provider monitoring stations, contemporary calving fronts or averages of the entire glacier.

The historical average is the arithmetic mean of nine season-matched area averages. The recent range is that mean plus or minus two sample standard deviations, truncated at zero. Faster than usual or Slower than usual appears only when the latest area mean plus or minus twice its mean source speed error lies entirely beyond this range. Otherwise it is Within the recent range. A comparison is withheld for insufficient coverage or speeds small relative to source error.

Percentages measure speed difference from the 2017–2025 reference average. They do not measure ice loss. The mean source error is not divided by the square root of cell count because nearby cell errors may be correlated. Chart bars show one mean source error estimate. These descriptive labels are not a significance test, provider-issued alert or a 30-year climate normal. Radar processing, seasonal window differences and source uncertainty limit interpretation; speeds are annual equivalents of short-period observations, not annual means.

The source files are read using verified HTTP byte ranges from the public GEUS archive. The derived comparison retains the file ID, provider checksum, sampled coordinates, common-cell offsets, dates and raw speed/error estimates outside the website, with reproducible processing scripts.

GEUS / PROMICE · Greenland Ice Velocity from Sentinel-1 Edition 5 · CC BY 4.0 ↗

The dashed line shows the 2017–2025 average; shading is the recent range and bars show source speed error. Rates in metres per day are multiplied by 365.25 to show metres per year equivalent, rather than a full-year average. Download the saved comparison values, exact dates and reference locations.

PROMICE mass balance product, Mankoff et al. (2021) · whole-ice-sheet model and measurement estimates · gigatonnes (Gt) per hydrological year. One billion tonnes is one Gt. Each year runs from 1 September to 31 August; the selected period remains beside the figures.

Surface mass balance accounts for snowfall and other surface gains minus runoff and surface losses. Subtracting solid-ice discharge and basal loss gives the overall balance. These are whole-ice-sheet estimates, not local melt measurements; annual estimates may be revised. Uncertainties on the cards are supplied by the source.

The historical comparison uses 30 complete annual estimates for 1990/91–2019/20. Its range is mean annual net loss plus or minus two sample standard deviations. Categories use central estimates; source uncertainties remain separate. The percentage compares annual net loss with the historical mean, not the percentage of Greenland’s ice that disappeared. Within the historical range can include substantial ice loss and does not mean the ice sheet is stable.

Saved snapshot retrieved 4 October 2026. GEUS/PROMICE dataset and citation · Source annual balances · Download the saved annual record and uncertainties · CC BY 4.0.

Place markers are approximate geographic references, not current glacier fronts. Nuuk is Greenland’s capital. Gunnbjørn Fjeld, in the Watkins Range, is its highest mountain, 3,694 m above sea level; this is distinct from the highest surface of the ice sheet. Smaller town labels appear on closer views.

Visit Greenland · Nuuk ↗ · Visit Greenland · Watkins Range ↗ · Visit Greenland · highest-point elevation ↗ · GeoNames · mountain coordinates ↗

ESA · Kangerlussuaq Glacier ↗ · ESA · 79°N Glacier ↗ · ESA · Zachariæ Isstrøm ↗

Reconstructed ANALYSIS · °C. Saved annual coverage: 1854–2025. Retrieved 2026-09-17. NOAA basin definitions ↗.

Compare the whole Atlantic, North Atlantic and South Atlantic. Each line shows annual temperature differences from that region’s own 1981–2010 average, on the same chart scale.

NOAA ERSSTv5 reconstructed monthly temperatures on a 2° grid. NOAA WOA98 surface basin definitions supply fractional ocean-area weights from their 1° cells, between 65°S and 75°N. Includes the Caribbean and Gulf waters and the Atlantic sector of the Southern Ocean (63°W–20°E); excludes separately classified Arctic and Mediterranean waters. The equator divides North and South; equatorial cells contribute half to each. Twelve equally weighted monthly means form a complete calendar year. Each region is rebased to its own 1981–2010 mean. Whole Atlantic uses area weights, not a simple average of the two hemispheres. Months with more than 1% missing area are excluded. The incomplete current year is omitted. Earlier reconstructed estimates have greater uncertainty; no uncertainty band is supplied here. This replaces the previous 20–60°N regional series and differs from the daily map product and baseline.

Source and documentation ↗

NOAA ERSSTv5, provided by NOAA PSL. Source temperature data · Source basin mask. The saved snapshot was downloaded on 17 September 2026. These historical records update separately from the maps.

NOAA NCEI · observation-based ANALYSIS. Saved annual coverage: 1955–2025. Retrieved 2026-09-11.

Annual heat-content difference from the 1981–2010 average, across the Atlantic including the Arctic, in the upper 700 metres.

NOAA’s whole Atlantic basin series, including the Arctic, rebased to its 1981–2010 mean. Units are zettajoules (ZJ): 1 ZJ = 10²¹ joules. The shaded band is ±1 source standard error around each annual estimate; it does not include uncertainty in our reference-period mean. This measures stored heat, not surface temperature or heat flowing through the basin.

Source and documentation ↗

NOAA NCEI ocean heat content. Source heat-content data. The saved snapshot was downloaded on 11 September 2026; historical records update separately from the maps.

PSMSL · tide-gauge observations. Saved annual coverage: Newlyn · Cornwall: 1916–2024; Brest · France: 1846–2023; Reykjavík · Iceland: 1957–2024; New York · The Battery: 1856–2025. Retrieved 2026-09-17. Holgate et al. (2013) ↗.

Annual mean sea level relative to the land, shown against each station’s own 1981–2010 average. Choose a station to explore its local history. The available records currently cover North Atlantic coasts; they do not represent the whole Atlantic. These are historical records, not current sea levels.

Revised Local Reference (RLR) annual records, converted from millimetres to centimetres and rebased using available accepted years in 1981–2010. Missing years, attention or mean-tidal-level flags, and years marked incomplete are omitted; gaps remain visible. Stations have different record lengths and reference-period coverage. At least 20 accepted annual values are required in 1981–2010. The number of available reference years is shown for each station. Local land movement and ocean conditions both affect this record. It is not an Atlantic-wide sea-level average or a tide forecast. No uncertainty band is available in these annual files.

Source and documentation ↗

Station sources

• Newlyn · Cornwall · Source annual values
• Brest · France · Source annual values
• Reykjavík · Iceland · Source annual values
• New York · The Battery · Source annual values

PSMSL, accessed 17 September 2026; Holgate et al. (2013). These historical records update separately from the maps.

The NAO describes a broad North Atlantic atmospheric circulation pattern. Positive and negative values identify opposite phases; values close to zero indicate a weak expression of this index pattern. Neither sign is inherently good or bad.

Positive phases generally correspond to a stronger subtropical-high / subpolar-low contrast; negative phases to a weaker or reversed contrast. Weather effects vary by region and season. This index alone does not predict your local weather or measure AMOC strength.

This NOAA CPC index is calculated from 500 hPa height anomalies, not simply two surface pressure readings. Its standardisation uses 1950–2000. Bars above zero are positive; bars below zero are negative. Bar height shows the daily index, not wind speed or temperature. No forecast values are plotted.

The reading shows the latest date in the downloaded file, which can lag today. The page uses the published saved record; reloading does not start a source download. Map selections and RAPID dates remain separate.

NOAA CPC NAO ↗ · Method ↗ · Daily data ↗

MEASURED / DERIVED · Sv. Historical published record — not a live reading. 1 Sv = one million cubic metres of water per second. The selected date, release/version and full record coverage come from the loaded record beside the chart. Surface-current maps are not AMOC measurements.

Explore the published record ↗

The approved illustration explains global circulation pathways; it is not live data, and not every pathway is part of the AMOC. Research findings are separate from the calculations used to display the RAPID series.

Adapted from ESA’s global ocean conveyor belt ↗. Redrawn, recoloured and relabelled by Atlantic Observatory; coastlines: Natural Earth, public domain. This adapted diagram is available under CC BY-SA 3.0 IGO ↗.

Xing, Q., Elipot, S., Johns, W. E., Smeed, D. A., Moat, B. I. & Loder, J. W. (2026). Meridionally consistent decline in the observed western boundary contribution to the Atlantic Meridional Overturning Circulation. Science Advances 12(15), eadz7738. Published online 8 April 2026. Paper DOI ↗

Baker, J. A. et al. (2025). Continued Atlantic overturning circulation even under climate extremes. Nature 638, 987–994. Published 26 February 2025; author correction 17 April 2025. Paper DOI ↗ · Correction ↗. The experiments sustain a weakened circulation; they do not establish a collapse date.

Mechanism and unit: Met Office · Atlantic Meridional Overturning Circulation ↗.

Editorial review: 22 September 2026. This review date is separate from the observation release and selected data time.

ANALYSIS combines observations using estimation methods. MEASURED identifies instrument-based records. MODELLED identifies numerical model output. DERIVED identifies calculations from other fields. FORECAST identifies a future prediction.

Valid time describes the conditions shown. Model run identifies when a forecast began. Retrieval date records downloading, not the time measured. Record coverage is the period included in a published series. A source check may succeed without finding newer data.

Read each layer’s own units, dates and classification. Daily air anomalies use 1991–2020; daily SST anomalies use 1971–2000; historical temperature, ocean heat and sea-level differences use 1981–2010. The NAO is standardised against 1950–2000. These reference periods are not interchangeable.

Each reading represents a grid cell from the selected dataset, not a basin average or a reading from a sensor at that location. The main Atlantic map uses an equirectangular projection with a 30°N standard parallel and Natural Earth 1:50m land. Ocean colours are smoothly blended between displayed cells for readability; this is visual interpolation, not additional observations or model resolution. Point readings retain original cell values. Missing-data boundaries may appear softened. High-density display rendering sharpens coastline outlines, text and arrows without adding scientific grid detail. Coastlines are generalised; narrow coastal cells can be obscured. The RAPID line is a geographic reference, not the measured instrument layout.

Missing or masked data are not zero. No values are invented for gaps. Where supplied, uncertainty or analysis error accompanies the reading; absence of an uncertainty band does not mean certainty. Read the product-specific limitations below.

The side panel reports the selected source grid cell, with its units, classification and valid date or time. It is not a basin average or an instrument reading at that exact location. Wind direction is meteorological FROM; current direction points TOWARD. Swell direction is FROM for partition 1 only. Map arrows point toward travel.

A short Forecast comparison uses the next available frame from the same model run, at the same selected cell. Both values and the actual elapsed hours are displayed; missing values are not compared. Saved analyses comparisons use consecutive available daily analyses instead. These are derived comparisons, not an independent forecast or evidence of a long-term trend. There is no later comparison at the last available time.

The jet domain maximum is computed from all source nodes, including land; it can occur away from the selected location. It is not the local reading. Near-zero swell can have no meaningful direction. Missing or masked source values remain unavailable. SST influenced by sea ice can use proxy estimates; the selected-location panel keeps this warning when applicable. Preliminary analyses and recent reanalysis can be revised. Product-specific resolution, processing, source components and licences are listed below; source check and retrieval dates are in published source status.

Each map layer keeps its own dates and source. Valid time describes the conditions shown; model run identifies when a forecast started. Jet, current and wave frames are predictions from one run, even when the selected time is in the past. Their initial state is modelled too. SST shows saved daily analyses, without an SST forecast. ERA5 air uses a full UTC daily mean reanalysis; its publication delay and revision limits are described in daily air methods.

The slider steps through available saved frames, not equal intervals of elapsed time. Missing times are omitted; no intermediate forecast is invented. Previous, Next and Play follow this same sequence. Latest available selects the saved forecast closest to now, or the latest saved analysis. The fallback ECMWF air layer holds one model initial state rather than a forecast sequence.

Reload published data reloads the website’s latest published snapshot. It does not request a new forecast or start scientific source retrieval. Validation and publication run separately; failed checks retain the last validated data. Older data and failed refresh warnings remain beside the map. Source checks and publication.

The Atlantic map includes North and South Atlantic forecast regions and UK Shipping Forecast areas. Selecting an area moves the map to a reference point. The reading uses the selected dataset’s nearest grid cell; it is not an area average or the official forecast.

Links open the Met Office or WMO marine forecast pages for the selected area. The Observatory’s NOAA, ECMWF and HYCOM map readings remain separate model and analysis products. Dashed lines are reference guides, not complete official boundaries. Southern regions extend beyond this map’s 65°S limit. The map is for exploration and learning, not navigation.

Met Office Shipping Forecast · Met Office high seas forecasts · WMO marine forecasts

The static grain on dark land is decorative. It does not represent terrain, ice or an environmental measurement. Country boundaries use Natural Earth’s generalized 1:110 million Admin-0 land boundary lines, version 5.1.0, in the public domain. The map includes lines classified as international boundaries and omits disputed or claim lines. Boundaries are clipped to the land shapes and provide geographic orientation.

Globe, Atlantic and Mercator currents, Atlantic surface wind and jet stream, and both polar wind layers share rounded quadratic Bezier lines, directional arrowheads and a screen-sized stroke of 2.2 CSS pixels on desktop and 2 on phones. Globe and Atlantic surface-current line colour encodes speed; on Mercator and wind maps single-colour lines indicate direction over the existing quantitative colour fields. Waves use short curved crest marks and arrowheads for first-partition swell travel, rather than connecting them into current streamlines. Line length and spacing do not encode magnitude. Source dates, units, depth/height, point readings and existing regional update schedules stay separate.

Polar land edges are drawn from the original Natural Earth 1:50 million land polygons used to create the reference land masks, rather than enlarged mask pixels. This sharpens geographic outlines without adding scientific resolution or changing source-cell readings. These generalized coastlines are orientation references, not current glacier fronts, ice-shelf boundaries or navigation charts.

Ocean and continent names are geographic labels, not data contours. The Atlantic overview uses straight, spaced lettering; place and ocean names on the polar maps follow arcs centred on the projected pole. Labels remain upright and use straight fallback text where a tight arc would reduce readability. The faint dashed Equator, Tropic of Cancer, Tropic of Capricorn and Arctic Circle are orientation references. Tropical and polar circles are shown at approximate conventional latitudes of 23.44° and 66.56°; they are separate from the red ice-class boundary. They appear at the overview scale and give way to local labels on closer views.

Natural Earth · Admin-0 boundary lines ↗

Map dates show the selected conditions; model run dates show when a forecast began. Historical charts have separate records, reference periods and update schedules. A successful check does not necessarily mean newer data is available. Map times are shown in UTC.

Atlantic Observatory · Data updates
Update check information is not available yet. Published coverage and retrieval dates are shown below.

Seabed depth, relief & sediments

Detailed imagery is requested when the seabed section opens. EMODnet DTM 2024 tiles and its mean-depth query service provide European and northeast Atlantic detail, with GEBCO-filled areas. The wider global background is EMODnet World Base Layer (2020 service; documented inputs include EMODnet 2018 and GEBCO 2019). Outside mean-depth query coverage, no numeric depth is inferred from the picture.

BGS marine sediments: UK Continental Shelf, 1:250,000 mapping. Point results include the record’s release date where provided. Place names provide orientation; label positions are approximate, not sampling points or boundaries. Detailed seabed feature names appear as you zoom in, from the IHO-IOC GEBCO Gazetteer served by EMODnet. Toggle “Show place names” for an unobstructed map. Services checked 16 September 2026; this is a service check date, not a survey date. No daily data updater is needed. EMODnet publishes periodic terrain-model releases; the pinned tile edition needs review when a new release appears. BGS map content is served directly by BGS.

Contains British Geological Survey materials © UKRI 2026. BGS source and Open Government Licence · EMODnet sources and methods

The Observatory serves published validated assets. Loading or reloading a page does not start a scientific source download. Source retrieval runs separately from the website. External maps and comparison images can request their providers directly when opened.

Atlantic SST, jet, waves, currents and NAO have a published check report. A completed check need not produce newer data. Air and polar products have separate source dates and check reports. Arctic and Antarctic daily update checks are linked above; the Greenland velocity preview is a separate dated mosaic. Missing check evidence is unknown, not success or failure.

The repository defines separate source-update workflows, but their existence does not prove they are enabled or have run successfully. Source processing validates candidates before publication; failed checks retain previously validated snapshots. Historical records and RAPID remain reviewed saved releases. Their end dates are not failed daily updates.

Read valid time, model run or release, retrieval and recorded check outcome separately. Selected dates remain beside the visualisations; published coverage and recorded outcomes are listed in Atlantic source status ↗.

Some metadata remains product-specific: Arctic per-layer retrieval times are not recorded; Mercator individual model issue time is unavailable. Where no product-specific reuse terms are recorded here, consult the linked provider before reuse.

EOxCloudless 2024 by EOX IT Services GmbH (contains modified Copernicus Sentinel data 2024). A photographic mosaic assembled from different acquisitions. Ocean and polar infill can be older and coarser. It is not a current observation of ice, melt or weather.

Imagery and coverage · CC BY-NC-SA 4.0 licence · EOX::Maps. Displayed on this noncommercial site in geographic and equal-area projections; reference imagery remains separate from the dated environmental data.

Satellite place names use Natural Earth physical labels (regions 5.0.0; marine areas 5.1.0), in the public domain. Continents and oceans appear at wide views; seas, islands, archipelagos and coastal features appear as you zoom. Names are selected for available space, so this is not a complete gazetteer. The Labels control hides them.

Label anchors are approximate orientation aids, not navigation points, protection boundaries or evidence of current conditions. Tristan da Cunha is anchored at the main island using Smithsonian Global Volcanism Program coordinates. Label sources and snapshot details.

These viewers have their own controls, dates, scales and attribution. They do not change the main map’s selected reading. The starting date is a viewer request, not evidence of data availability.

Sea-surface temperature · Analysis · Met Office OSTIA · daily · 0.05° grid. Sea-surface temperature in °C, combining satellite and in-water observations. This is not the temperature of the ice surface. Source ↗ · Product DOI ↗

Surface currents · Model · Mercator Ocean · daily mean · 1/12° grid. Surface water velocity in m/s at the shallowest model level (about 0.5 m). These are different model estimates from our North Atlantic HYCOM maps. Source ↗ · Product DOI ↗

Salinity · Model · Mercator Ocean · daily mean · 1/12° grid. Salinity at the shallowest model level; use the depth control inside the viewer to explore below the surface. Read the provider’s scale and units. Source ↗ · Product DOI ↗

Significant wave height · Model · MFWAM · 3-hourly fields · 1/12° output grid. Combined significant wave height in metres. Individual waves can be higher. This differs from our North Atlantic NOAA GFS-Wave forecast. Source ↗ · Product DOI ↗

Primary swell height · Model · MFWAM · 3-hourly fields · 1/12° output grid. Significant height of the primary swell partition, in metres; not total swell or coastal breaking waves. Source ↗ · Product DOI ↗

Chlorophyll-a · Satellite-derived · Copernicus GlobColour / ACRI-ST · daily · 4 km grid. Chlorophyll-a in mg/m³. Clouds, darkness and ice leave gaps; blank areas are not zero. This daily layer differs from our seven-day composite. Only recent dates are available in this near-real-time product. Source ↗ · Product DOI ↗

Seabed depth — model grid · Static model bathymetry · Mercator Ocean · 1/12° grid. Depth below the geoid in metres, simplified for the ocean model. This is not a detailed seabed survey or sediment map. Source ↗ · Product DOI ↗

Antarctic hidden-lakes video

Antarctic subglacial lakes threaten ice sheet stability, 3 April 2025. Credit: © ESA / University of Edinburgh / Planetary Visions. ESA’s supplied YouTube embed loads only when selected. Used for educational context under the ESA Standard Licence.

Bedmap3 relief map

Bedmap3 bed-topography illustration by Peter Fretwell, from Pritchard, H. D., Fretwell, P. T., Fremand, A. C. et al. (2025), Scientific Data 12, 414, via British Antarctic Survey. CC BY 4.0. Image unchanged; resized for display, with the original height scale preserved. Dataset, methods and limitations.

Site photography

Photographs from Unsplash:

• Open ocean — Yiran Ding
• Rocky coastline — Alexei Scutari
• Underwater — Aleksandr Barsukov
• Waves — Wolfgang Hasselmann
• Polar coastline — Colin Ruthven / Unsplash

Photographs: Matthew Stephenson, Colin Ruthven and Bob Brewer / Unsplash. Illustrative photography; not satellite observations or measurements of sea-ice extent.

Explore the Observatory

Photographs: Clemens Jezler · Francesco Ungaro · Matt Palmer / Unsplash.

Photography page

Original photographs and credits for the Photography gallery. Dates below are publication dates on Unsplash, not verified dates of capture.

• Mountain hero — Robert Haverly. Antarctica. Published 28 November 2020 (UTC). Camera: Sony ILCE-6500. Original photograph on Unsplash ↗.
• Icebergs — Robert Haverly. Published 28 November 2020 (UTC). Camera: Sony ILCE-6500. Original photograph on Unsplash ↗.
• Coastline from above — Matt Palmer. Antarctica. Published 22 June 2021 (UTC). Camera: Canon EOS 5D Mark III. Original photograph on Unsplash ↗.

These three source listings identify the photographs as available under the Unsplash License ↗. Photographer names remain beside the photographs.

Arctic photograph strip — mahlersilvan / Unsplash. An ice-filled fjord beneath low cloud, shown above the Greenland feature. Original photograph on Unsplash ↗. Cropped and resized for the page; illustrative photography, not a dated observation of ice conditions.

Explore Greenland header — Annie Spratt. Coastal village photograph. Original photograph on Unsplash ↗. Displayed as a cropped, resized header on the Arctic page.

Founder photograph: James, Iceland. The founder’s original caption and statement are on About ↗. Photography is illustrative, not a dated scientific observation.

ESA adaptation and CC BY-SA 3.0 IGO terms are listed under Circulation illustration ↗. Natural Earth geography is public domain. Other providers retain their individual attribution and reuse requirements; no single licence is asserted for all resources.

• World Bank / IMF · Shipping activity
• Global Energy Monitor · Offshore wind
• NOAA NCEI · Microplastics
• NOAA NCEI OISST · Global Ocean Climate
• NOAA OISST
• FNMOC / HYCOM · Atlantic currents
• FNMOC / HYCOM · Global currents
• Copernicus / ECMWF ERA5 daily air
• ECMWF IFS
• NOAA GFS-Wave
• Copernicus / ACRI-ST GlobColour
• Copernicus / Mercator GLO12
• OSPAR / DFO / NOAA / ICMBio / BAS / VLIZ · Protected areas
• GEBCO 2026 · Geography reference maps
• EMODnet / GEBCO / BGS
• OSI SAF / MET Norway / NOAA / WMM2025
• NOAA/NSIDC Sea Ice Index v4
• NOAA ERSSTv5
• NOAA NCEI ocean heat
• PSMSL tide gauges
• NOAA CPC NAO
• RAPID release
• ESA illustration and AMOC research
• NASA GIBS and polar references