Methodology
Three kinds of numbers, never mixed
Every temperature on this site carries one of three labels.
- Satellite analysis. The daily value shown for each location and on the map. It comes from NOAA Coral Reef Watch’s CoralTemp product: a gap-free, global, 5 km3.1 mi grid of night-time sea surface temperature built from several satellites and blended with in-situ data. It is an analysis, which means a best estimate for every 5 km3.1 mi cell, not a thermometer reading at your beach.
- Measurement. Where a buoy or pier sensor exists within about 15 km9.3 mi of a location (NOAA CO-OPS and NDBC stations in the United States, Fisheries and Oceans Canada buoys in Canada), we show its latest reading separately, with the station name, its distance from the location and the time of the reading.
- Typical temperature. The long-term average for the current month at the same 5 km3.1 mi cell, calculated from the monthly means of the satellite record from 2003 to the most recent complete year. The “warmer or colder than typical” line compares today’s analysis with that average.
We do not show forecasts, and we do not fill gaps with model guesses. If today’s analysis is not available we say so and show the latest one we have, with its date.
Why the beach can feel different
Satellite sea surface temperature describes the top layer of open water in a 5 km3.1 mi cell. Water at a beach can differ from it by several degrees:
- Shallow water warms and cools faster than the open sea, especially on sunny or windy afternoons.
- Tides move water in and out; an incoming tide over sun-warmed sand can be noticeably warmer.
- Upwelling, common on the Pacific coast of North America, pulls cold deep water to the surface close to shore.
- River mouths, harbors and lagoons hold their own water, which the satellite cell may not represent. Locations in that situation are marked on their page and left out of the rankings.
Where a nearby station exists, its measurement is usually the better guide for swimming. Where none exists, treat the satellite value as the open-water temperature just offshore.
How a location is placed
Each location is a named beach, waterfront or coastal town with verified coordinates. We select the nearest ocean cell within 15 km9.3 mi of that point and show the distance on the page. Cells over sheltered water that is known to behave differently from the open coast are flagged rather than silently used.
Freshness rules
- The daily analysis is normally published the following day. A value up to three days old is shown as current, with its date.
- Between three and ten days old, the page says “latest available analysis” and explains that no newer data has arrived.
- Older than ten days, the current value is hidden and only the long-term averages remain, with a notice.
- Station measurements are shown only if they are less than 24 hours old.
Averages and the record we use
The satellite record begins in 1985, but the 1985 to 2002 portion is a different reanalysis with a research-only license. We therefore use the record from January 2003 onward, which is also the period produced by one consistent method. Monthly averages, the typical range for each month and the warmest and coldest months are all computed from that period. This differs from the 1991 to 2020 climate normals used by meteorological agencies; we say so here rather than quietly mixing two sources with different resolutions.
A month’s average normally rests on at least 15 years. The monthly satellite product leaves a cell empty for any month in which sea ice was reported nearby, so cells in the Gulf of St. Lawrence and around Nova Scotia and Newfoundland lose many winter months. Where a month has 8 to 14 years we still show the average and mark it as based on fewer years; where it has fewer than 8 we show no average at all, only the range of the few values that exist, and we label the month ice-affected. A location is published only if at least five months, the open-water season, have the full 15 years. When winter months are ice-affected, the “coldest month” is not named, because the true coldest month is hidden inside that gap.
Sea temperature by year
The “Sea temperature by year” section of a place page gives, for each complete calendar year, the average of the twelve monthly means at the place’s 5 km3 mi cell, and compares the average of the last five complete years with the first five. Years with a month missing are left out, so the record starts in 2004 (the monthly record begins in February 2003).
A single coastal cell can drift from its surroundings over twenty years, through changes in the satellites behind the record or real changes close to shore. So we check each place’s change against a second, independent NOAA record, OISST, which also uses ship and buoy measurements, at its nearest open-water cell of 0.25°, about 25 km16 mi across. Where the two changes differ by more than 0.5°C0.9°F, or OISST has no cell nearby, the section is left out: 12 places on 13 November 2026, most of them on the coasts of the Carolinas and Virginia, where both records show cooling but disagree on how much. The OISST change is printed under the chart. Places whose winters are often hidden by sea ice have too few complete years and get no section.
Historical ocean temperatures
The historical explorer uses NOAA’s OISST version 2.1, which blends satellite, ship and buoy measurements into a map in cells of 0.25°, about 25 km16 mi, with monthly means from September 1981. Each month is compared with the 1991–2020 average for the same calendar month from the same record, and each year’s picture is the mean of its twelve monthly differences. Water at the freezing point (a monthly mean of −1.5°C29.3°F or colder) is treated as sea ice and left out.
“The world ocean” means the sea between 60°S and 60°N, the usual way of summing up the world’s sea surface temperature, because sea ice nearer the poles hides the water under it. All averages, for the world and for each ocean and sea, are weighted by the area of each cell. The change picture compares the average of the last ten complete years with that of 1982–1991. The daily maps elsewhere on the site use NOAA CoralTemp instead, a finer product whose early years (before 2003) cannot be used here, so the same month can differ slightly between the two.
Rankings
The warmest and coldest water lists compare locations using the same product on the same date. Locations with a stale value, or flagged as sheltered water, are excluded from the ranking for that day. Rankings cover the locations in our inventory, not every stretch of coast on Earth.
Each list exists for the world, the United States, Canada, the United Kingdom, Europe (the European countries in our inventory, including Turkey’s Mediterranean and Aegean coasts) and each hemisphere (split at the equator). Lists show at most 100 rows; ties are broken alphabetically so the order is stable from one day to the next. A list is published only when at least five locations qualify on that date.
Surf forecasts and the conditions score
The surf pages are built on a second data layer: a wave and wind forecast from NOAA’s GFS-Wave model (WAVEWATCH III), sampled at open-water cells near each break, with measured waves from NOAA National Data Buoy Center buoys where one sits nearby. The rules below are fixed and public; the score is a filter, not a judgement.
- Model and cells. We read the 16 km10 mi regional grids for the coasts of the United States and Canada and the 25 km16 mi global grid elsewhere, from the 00 and 12 UTC runs, hourly for two days and every three hours out to seven. Each spot is tied to the nearest sea cell within 35 km22 mi on the regional grids or 60 km37 mi on the global grid, preferring a cell that lies in front of the break rather than behind a headland. The cell’s distance is printed on every page.
- Not a surf height. The big figure is the model’s significant wave height in open water at that cell, with the main swell’s height, period and direction beside it. Breaking waves at the beach can be smaller or larger; we do not estimate a face height, because that would need per-break calibration we do not have.
- Coarse cells. When the cell lies more than 20 km12 mi from a break on a regional grid or 35 km22 mi on the global grid, sits behind the facing direction, is hemmed in by land on most sides, or the spot is marked sheltered, the page says so, gives no rating and leaves the spot out of the best-conditions lists.
- Wind classes. Relative to each break’s facing direction: offshore within 45° of blowing off the land, onshore within 45° of blowing onto it, cross-shore in between, light below 8 km/h5 mph.
- Conditions score (0 to 100, with the words Very poor, Poor, Fair, Good and Excellent). Three parts are each scored from 1 to 10 and multiplied together, then divided by ten, so a perfect day is 10 × 10 × 10 ÷ 10 = 100 and one bad part drags the whole score down the way it does in the water. Swell energy follows the main swell’s height squared times its period on a logarithmic scale (an energy of 1.5 scores 2, of 10 scores 6, of 100 scores 10), capped at 3 when the period is under 7 s and at 5 under 9 s. Direction is 10 inside the break’s swell window, 7 within 15° outside it and 1 further outside. Wind is 10 when light, under 8 km/h5 mph; offshore 9, or 6 above 40 km/h25 mph; cross-shore 9 below 12 km/h7 mph, 6 up to 25 km/h16 mph and 4 above; onshore 8 below 12 km/h7 mph, 6 up to 20 km/h12 mph, 4 up to 30 km/h19 mph and 1 above that, which is blown out. The word follows the score: very poor below 25, poor from 25, fair from 45, good from 64 and excellent from 85. Every rating on the site can be opened to show the three parts and their reasons. No score is shown when any input is missing or the cell is coarse.
- Daily summary. Each local day is summarised over its daylight steps (06:00 to 18:59): the range of wave heights, the energy-weighted swell direction, the majority wind class, the best score and the hour that reaches it.
- Freshness. A run up to 18 hours old is shown as the forecast; up to 36 hours it is labelled “latest available forecast”; older than that the forecast sections are replaced by a notice and the page is withdrawn from search engines until a newer run is processed. Buoy readings are shown only when less than three hours old.
- Summaries. The sentences under “Forecast summary” are generated from the numbers by fixed templates; no language model is involved.
Wave heights
The wave height map shows significant wave height, the average height of the highest third of the waves from trough to crest, from NOAA’s GFS-Wave model at the start of its 00 and 12 UTC runs: a model analysis on a 25 km16 mi grid, not a measurement. Single waves can be nearly twice as high. The figures per ocean and sea are area-weighted over every open-water cell inside the same outlines as the ocean pages; “rough end” is the 90th percentile. Near the coast the sea bed, headlands and reefs change the waves, so a beach can differ a lot from the nearest cell. Water the model leaves out near the poles is drawn as sea ice.
Ocean currents
The ocean currents map shows the surface current from NOAA’s RTOFS ocean model: a model analysis of the top layer of the ocean at 00 UTC, averaged from its 9 km6 mi grid into cells of about 25 km16 mi. Colours give the speed, and white streaks (moving trails on the interactive map) the direction the water flows toward, the opposite of winds and waves, which are named for where they come from. “The big currents” are the fastest cells inside fixed boxes drawn around six well-known currents; the figures per ocean and sea are area-weighted inside the outlines of the ocean pages, and “fast water” means faster than 1.8 km/h1.1 mph. The model describes the open ocean: tidal streams, rip currents and the flow at harbour mouths are not in it.
Temperature anomalies
The anomaly map shows the difference between each day’s sea surface temperature and the 1991–2020 average for the same date, as computed by NOAA Coral Reef Watch from its CoralTemp analysis in 5 km3 mi cells. The figures per ocean and sea, per named sea, gulf or bay (outlined as in Natural Earth’s marine areas) and for the world are area-weighted averages over the water that has a value; ice-covered water has none. “Furthest from normal” lists 1° cells, about 110 km70 mi across, at least half measured, within 70° of the equator and at least 1,500 km930 mi apart, named by their sea area and the nearest country. This is a different comparison from the location pages, which compare each place with its own 2003–2025 average for the month.
Marine heatwaves
The marine heatwave tracker follows Hobday and colleagues (2016, 2018) against the 1991–2020 climate period. Each day we compare NOAA’s CoralTemp sea surface temperature, in cells of about 10 km6 mi, with the average and the 90th percentile of 1991–2020 for that place and day of the year, computed by the Danish Meteorological Institute from the European Space Agency’s climate record (11-day window, 31-day smoothing). A cell is in a marine heatwave once it has been above its 90th percentile for at least five days in a row, and its category counts how far above, in steps the size of the gap between the average and the 90th percentile: Moderate (up to twice the gap above the average), Strong (two to three times), Severe (three to four) and Extreme (four times or more). Water with sea ice today is not assessed. The two records agree closely: over 1991–2020 their averages differ by less than 0.1°C0.2°F on average. Until 13 November 2026 the tracker used NOAA’s Marine Heatwave Watch, which measures against 1985–2012 and leaves out all water that had sea ice in those years.
Because the ocean has kept warming since 1991–2020, a large share of it counts as in a heatwave on most days. For each ocean and named sea we give the share of its water in each category, the strongest category covering at least a tenth of it, and how many days in a row at least half of its water has been in a heatwave, from a daily record we keep since 13 November 2026 (longer runs show as “at least”). The home page names a sea only after five such days, and only when at least a tenth of it is Strong or worse.
Coral bleaching heat stress
The coral bleaching tracker shows NOAA Coral Reef Watch’s Bleaching Alert Area, the highest alert of the last 7 days in each 5 km3 mi cell on its 2023 scale, and the figures of NOAA’s virtual stations for 214 reef areas. The alert follows two measures: the HotSpot, how much warmer the water is than the usual temperature of the warmest month, and Degree Heating Weeks, the HotSpots of at least 1°C1.8°F added up over 12 weeks. Bleaching Warning means a HotSpot of at least 1°C1.8°F; Alert Levels 1 to 5 begin at 4, 8, 12, 16 and 20 degree heating weeks, counted in °C-weeks. These are satellite measures of heat stress and the risk of bleaching, not observations of bleached corals.
Tide times
Tide pages show predictions: the times and heights of high and low tide computed years ahead from the tide’s astronomical rhythm, by NOAA in the United States and by the Marine Institute in Ireland. They are not measurements. Strong onshore wind and low air pressure push the water above the prediction, offshore wind and high pressure hold it below, and in a storm the difference can reach 1 m3 ft or more. They are for planning a beach day and are not for navigation.
Which station. Each place uses one prediction station within 25 km16 mi, named on its page with its distance. Many stations sit behind the beach, in creeks, rivers, the Intracoastal Waterway or back bays, where the tide turns later and rises less than on the open coast, so a station named as open coast (a pier, “ocean”, “outer coast”) is preferred when it is not much farther away, and a station on inside water is used only when nothing else is within reach. Every choice was reviewed, partly by comparing the station’s average tidal range with that of the nearest open-coast station: a much smaller range gives a back-bay station away. Places with no suitable station, such as the ocean beaches of Fire Island, have no tide page.
Times and heights. Times are in the place’s own local time, with the time zone shown. Heights are measured from chart datum, the zero of local charts and tide tables: Mean Lower Low Water in the United States and Lowest Astronomical Tide in Ireland. Where the provider publishes an hourly curve, the chart draws it through the exact highs and lows. Where it publishes only the highs and lows, the chart joins them with a smooth half-cosine curve, the usual way to estimate the water level between high and low tide, and its caption says the line is an estimate. Times follow the °C/°F switch: 24-hour with °C, 12-hour with °F.
The description of the tide. The paragraph about each place’s tide is computed from the next 30 days of its predictions: whether there are two tides a day of about equal height (semidiurnal), two unequal ones (mixed) or mostly one (diurnal); the average rise or fall between a high and the next low; and the size of the bigger tides of the month (the 90th percentile of those ranges). A range of the bigger tides under 2 m6.6 ft is called small, 2 m to 4 m6.6 ft to 13 ft moderate and over 4 m13 ft large.
Units
Temperatures are shown in Celsius or Fahrenheit, wave heights in metres or feet and wind speeds in km/h or mph. Until you choose, each page uses the convention of the location’s country (Fahrenheit for the United States, Celsius elsewhere); the °C/°F switch in the header changes every figure on the site, including the text of the articles, and remembers your choice on this device. The same switch chooses metres and km/h (with °C) or feet and mph (with °F).