Temperatures on land and sea, sea ice and sea level
Where is it getting warmer, and by how much? The globe and map show land and sea since 1850, below them the sea ice of the poles and sea level from satellite measurements.
How to read: Measurement measured values · Reconstruction from indirect data such as ice or tree rings · Model computer model · Own calculation calculated by me from raw data. Technical terms are explained in the glossary.
Where is it getting warmer? – On land and sea.
Annual mean of sea surface temperature (NOAA ERSSTv5, 1854–2025) and, optionally, of land temperature (Berkeley Earth, 1850–2024) as an anomaly from the 1961–1990 mean. The slider goes back to 1854; “Play” starts at 1979. Play, choose a year or switch to “Trend”; the globe can be rotated and zoomed. Where has it become warmer since 1979 – on land, at sea or both?
💬 In plain language
- The globe shows how warm it was compared with earlier times (1961 to 1990 mean). Red means warmer, blue means colder.
- “Warmest year” shows, for each place on Earth, the year in which it was warmest there. Dark red means: only since 2015.
- “El Niño”: every few years the water in the Pacific off South America becomes unusually warm. This changes the weather in many countries. The map shows the difference from the cool “La Niña” years.
- “CO₂ in the air” shows where there is more or less CO₂ near the ground. It is a computer model that is checked against measurements.

Before about 1900 the data are sparse (light area on the left): few ship measurements, large ocean areas filled in statistically; the identification of El Niño and La Niña years is uncertain there. El Niño year · La Niña year · Volcano: major eruptions. Sulphur-containing gases rise into the stratosphere and form fine particles there that reflect sunlight; the Earth then cools by a few tenths of a degree for one to three years. Krakatau (Indonesia, August 1883), Santa María (Guatemala, October 1902), Agung (Bali, March 1963), El Chichón (Mexico, March/April 1982), Pinatubo (Philippines, June 1991, the strongest eruption of recent times, with up to about 0.5 °C of global cooling) and Hunga Tonga–Hunga Haʻapai (Pacific, January 2022, an underwater eruption that released mainly water vapour; the effect on temperature is small and still disputed). After Pinatubo the curve stays low for several years (1991/92 is also overlaid by an El Niño); in 1982 the eruption coincided with a strong El Niño, so the effect there cannot be separated out.
What the map shows – expand for explanation
What it shows: For each 2° × 2° grid cell, the annual mean sea surface temperature, subtracted from the mean of the years 1961–1990. Red is warmer, blue colder than in this reference period. Land and sea ice are left out in grey. With “Trend” the map instead shows how many degrees per decade each cell changed from 1979 to 2025. The slider goes back to 1854; before about 1900 there are only a few ship measurements at sea, and large areas are then filled in statistically (the map looks smoother there than the data actually are).
What you see: The tropical Pacific fluctuates the most. Since 1950, the warmest years in the Niño 3.4 region (5°S–5°N, 170°W–120°W), measured against the mean of all oceans, were 1987 (+1.1 °C), 2015 (+1.1 °C), 1997 (+1.1 °C), and the coldest 2022 (−1.2 °C), 1999 (−1.2 °C), 2021 (−0.9 °C): these are El Niño and La Niña. In the trend, not all seas warm equally: the temperature rises most in marginal and inland seas, in the Caspian Sea (up to about +0.9 °C per decade), the Black Sea, the Baltic Sea and the Mediterranean (mean +0.4), and also in the Barents Sea. Across all cells between 60°S and 60°N the trend is about +0.13 °C per decade; in the eastern equatorial Pacific it is barely present at +0.04, and in the south-eastern Pacific off southern Chile it cools slightly (−0.14).
On scope: ERSSTv5 is a series statistically filled in from ship and buoy measurements, especially in the early years and in remote areas. The map shows the sea surface, not the air temperature over land.
How the map was calculated
- Data: NOAA ERSSTv5 (Huang et al. 2017, J. Climate), monthly sea surface temperature on a 2° grid, one file of 162 MB from psl.noaa.gov. The years 1854–2025 were used.
- Annual mean: for each cell the average of the twelve months; cells with fewer than nine valid months are left empty.
- Anomaly: from each annual mean, the 1961–1990 mean of the same cell is subtracted.
- Trend: linear slope of the annual values 1979–2025 for each cell (least-squares method), in °C per decade; cells with fewer than 80% valid years are left empty.
- Display: The values were rounded to steps of 0.05 °C and loaded as a file (sea 2.8 MB, land 11.3 MB, about 0.9 and 1.5 MB respectively when transferred compressed) on opening; the map is drawn in the browser. The line in the timeline below is the mean temperature anomaly of the sea between 60° south and 60° north (excluding the polar regions with sea ice). Each cell counts according to the area it actually has: cells near the poles are smaller than at the equator and are therefore weighted less, by the cosine of latitude (1.0 at the equator, only 0.5 at 60°).
- Globe: The sphere uses the MapLibre GL library (BSD licence), coastlines and country borders from Natural Earth (1:110 million, public domain) and, as the coloured Earth surface, the “Blue Marble – Shaded Relief and Bathymetry” tiles from NASA (GIBS, public domain), which are loaded directly from NASA when viewed. The colours are calculated in the browser from the same 2° grid (bilinear interpolation) and laid onto the sphere as tiles; this is why the coastal edges look softer than in the flat map.
- Events in the timeline: El Niño and La Niña years are determined from the same data: a year counts as El Niño-dominated if the sea surface in the Niño 3.4 region (5°S–5°N, 170°W–120°W) was at least 0.5 °C warmer in its annual mean than the mean of all oceans between 60°S and 60°N, and as La Niña-dominated if it was at least 0.5 °C colder (relative index; it removes the long-term warming and measurement differences of the early years). This is simplified; the official classification uses running three-month means and changing reference periods. The dashed marks show major volcanic eruptions (Krakatau, August 1883; Santa María, October 1902; Agung, March 1963; El Chichón, March/April 1982; Pinatubo, June 1991; Hunga Tonga–Hunga Haʻapai, January 2022).
- Land temperature: The “Land temperature” layer comes from the gridded land data of Berkeley Earth (Rohde et al. 2013, 1° grid, CC BY 4.0 licence): annual mean per cell, anomaly from the 1961–1990 mean, as with the sea data; available from 1850 to 2024 (very sparsely covered before 1900, so the grid values are heavily interpolated there). When switched on, the colour scale changes (±2 °C or ±0.6 °C per decade), because the values over land fluctuate more.
- Land series in the timeline: With land temperature switched on, the timeline additionally shows the area-weighted mean (cosine of latitude) of all land cells of the Earth including the polar regions, 1850–2024, as an anomaly from the 1961–1990 mean. The sea line applies only to 60°S–60°N; the two lines therefore do not describe the same area.
- Limits: Sea ice and coastal cells are uncertain; the reconstruction is less accurate before the satellite years. An annual mean shows no seasons. On the globe the coastlines come from a coarse source and do not match the 2° cells exactly everywhere.
Statistics on the data – oceans and land temperature
| Sea temperature (NOAA ERSSTv5): monthly maps read | 2,064 (Jan 1854 – Dec 2025), each 89 × 180 grid cells, about 33.1 million individual values |
| From this: annual maps | 172 years, 1,889,936 valid cell values (68.6%); the rest is land, sea ice or too patchy |
| Sea cells with data (2° grid) | about 11,000 |
| Land temperature (Berkeley Earth): monthly maps read | 2,100 (Jan 1850 – Dec 2024), each 180 × 360 cells (1° grid) |
| From this: annual maps | 175 years, 3,454,307 valid cell values (30.5%), 24,311 land cells per map |
| Calculated | Annual mean per cell, anomaly from the 1961–1990 mean, trend 1979–2025 (sea) and 1979–2024 (land), relative Niño 3.4 series, global time series 60°S–60°N |
| Delivered to the page | 2 files: 2.8 MB (sea) and 11.3 MB (land), about 0.9 and 1.5 MB when transferred compressed, value step 0.05 °C |
How large are the calculation points (grid cells)?
| Dataset | Cell size | Cell width at the equator / at 48° N (Amstetten) / at 60° N | Area at the equator / at 48° N |
|---|---|---|---|
| Sea temperature (NOAA ERSSTv5) | 2° × 2° (89 × 180 cells) | 222 km / 149 km / 111 km | about 49,600 / 33,000 km² |
| Land temperature (Berkeley Earth) | 1° × 1° (180 × 360 cells) | 111 km / 74 km / 56 km | about 12,400 / 8,300 km² |
| Climate models, air temperature (CMIP6, 41 models) | depending on the model from 0.5° to 3.75° longitude, 64 to 360 latitude rows; median 1.25° × 1° | 56 to 417 km, median 139 km (cell height 56 to 313 km, median 125 km) | Median about 17,400 km² (smallest cells about 3,100, largest about 100,000 km²) |
| Climate models, sea temperature (CMIP6) | mostly about 1° (about 100,000 cells); two models 0.25° (about 1.5 million cells), the coarsest 2° (about 13,000 cells) | mostly about 100 km, at 0.25° about 28 km, at 2° about 220 km | mostly about 10,000 km² |
| Computer model ERA5 (only in the comparison with Amstetten) | 0.25° (about 31 km, given as “about 28 km”) | 31 km / 21 km / 16 km | about 960 / 640 km² |
Important: A cell is a mean over the whole area, not a measurement at a single point. A single place (for example Amstetten, 48° N) can therefore deviate from the cell value, especially with a coarse grid. The width of the cells decreases towards the poles (by the cosine of latitude), while the height stays the same. For the global means each cell counts with this weight, so that every area counts equally. For the models: each model calculates on its own grid; I have not converted anything to a common grid.
Sea ice and sea level
Three measurement series from satellite data: how much sea ice the Arctic and the Antarctic have over the course of the year, and how sea level has changed since 1993. Shown here without any model calculation of my own, only the measured values of the sources.
💬 In plain language
- In winter the sea around the poles freezes over; in summer part of it melts away again.
- The first chart shows this for the Arctic (North Pole). Since 1979 the ice area in September has been getting smaller. The values vary strongly from year to year.
- The second chart shows the Antarctic (South Pole). There the sea ice grew slightly until 2014; since then there has been markedly less.
- The third chart shows sea level. It has risen by about 10 centimetres since 1993.
What you see: In September 2026 the ice area was 4.8 million km², which is −25 % compared with the 1981 to 2010 mean. The trend since 1979 is −11.6 % per decade. In March 2026 it was 14.1 million km² (trend −2.5 % per decade). The lowest September value was in 2012 with 3.6 million km².
Scope: What is measured is the area with at least 15% ice cover, not the ice volume. The values are monthly means of daily values (every second day until 1987). Source: NSIDC, Sea Ice Index, Version 4, doi:10.7265/a98x-0f50.
What you see: Until 2014 the sea ice area around Antarctica grew slightly (September: +1.2 % per decade). 2014 was the year with the largest September area, at 19.8 million km². Since then it has shrunk markedly (−5.4 % per decade since 2015); the lowest September value was in 2023 at 16.9 million km². Over the whole period 1979 to 2026 there is practically no trend (−0.4 % per decade). In September 2026 it was 17.3 million km².
Scope: This is the sea ice around the continent, not the ice on land. The Antarctic ice sheet is measured differently (satellites for height and gravity field). According to the IMBIE assessment, between 1992 and 2017 it lost on average about 109 billion tonnes of ice per year, mostly in West Antarctica and on the Peninsula; East Antarctica gained slightly (about 5 billion tonnes per year, mainly through more snowfall). Sources: NSIDC (as above), summary of the IMBIE study (Shepherd et al. 2018) and IMBIE update to 2020.
What you see: Global mean sea level has risen by about 115 millimetres since 1993, on average by 3.4 millimetres per year. The series runs until June 2026.
Scope: This is a mean over the whole ocean; regionally, sea level rises and falls differently. The annual cycle and land uplift and subsidence (GIA) have been removed by the authors. Shown is the change relative to the mean of the first year. Source: University of Colorado Sea Level Research Group, version 2026_rel2, using radar altimeter data (RADS, provided by NOAA and TU Delft); Nerem et al. 2018, PNAS, doi:10.1073/pnas.1717312115.
Data as of November 2026. The data series end in different years depending on the source (mostly 2024 or 2025); the end point is given in the source note under each chart, an overview is on “Data and methods”.