Data and methods← Climate of the Earthruthner.at
Transparency · Climate of the Earth

Data and methods

For every chart on the climate page this register states where the figures come from, how they were calculated, what is excluded and where the limits lie. The derived figures and the scripts can be downloaded and recalculated. For the data sources themselves (who collects them, what they provide, where their limits lie) see Sources and assessment.

Rules of this register

  • All statements on licences come from the metadata or the source pages, not from memory. What has not been read is marked “not read”.
  • Measurement series are not passed on, only derived single figures. Please download the original data at the source.
  • Model data keep the licence of their model (20 models CC BY-SA 4.0, 3 models CC BY-NC-SA 4.0), stated in the header of each file.
  • No selection by match with the measurement; runs and models are not independent.
Data package

Data package

Data basis

Data basis

For the model calculations the data were read from the public archive Pangeo-CMIP6 (Google Cloud), one run after the other, and only the global time series were kept. The amount of data transferred is large: about 400 GB (estimate); the model calculations are complete. Only the results are kept (the time series in the data package, a few MB).

SeriesRunsread, unpackedtransferred, estimated
Air temperature (tas), global286147 GB87 GB
Precipitation (pr), global283145 GB125 GB
Sea surface (tos), 60°S–60°N233240 GB122 GB
Europe and Alps (tas and pr)97–66 GB
Totalabout 399 GB

Status: 20.11.2026 17:05. The runs are complete; the smaller measurement datasets (E-OBS, Berkeley Earth and others) are not included in the total.

How this was estimated: per model the size of the stored field (months × grid cells × 4 bytes) times the actual number of months of each run; the compression ratio (1.2 to 2.0, depending on quantity) was measured on ten stores in the archive. Not included are the model series of the charts with the 41 single runs and the measurement data (about 2 GB). The result is an estimate, not a counter reading; after the Europe runs are complete (97 runs) about 30 GB will be added.

Overview

Overview

ChartDateStatusDownloads
Climate models against measurement: air temperature20.11.2026documented1
Climate models against measurement: sea surface20.11.2026documented1
Several runs per model20.11.2026documented3
Time windows, internal variability and model difference20.11.2026documented3
Quality check of the models20.11.2026documented2
Glacier map20.11.2026documented0
Glaciers: length since 185020.11.2026documented1
Glaciers: mass balance20.11.2026documented1
Glaciers: summer temperature and mass balance20.11.2026documented1
Glaciers: finds and forecast signs (reports)20.11.2026documented0
Europe, the Alps and Austria: models against measurements20.11.2026documented2
Time travel: stations, thermometers and reconstructions20.11.2026documented0
Time travel back 66 and 540 million years20.11.2026documented0
Vostok ice core: temperature and CO₂ over 420,000 years20.11.2026documented0
Reconstruction of the last 2,000 years (Ljungqvist)20.11.2026documented0
Rates of rise: how fast was warming in the past?20.11.2026documented0
Map and globe: sea, land, El Niño, CO₂ in the air20.11.2026documented0
CO₂ in the air: amount, origin and fate20.11.2026documented0
CO₂ seasonal cycle, Keeling spiral and residence time20.11.2026documented0
Countries and scenarios: calculator for CO₂ savings, warming and costs20.11.2026documented0
Alpine trend map and height profile (E-OBS)20.11.2026documented0
Model map Alpine region: temperature trend, models against measurement20.11.2026documented0
Scenarios to 2100: SSP1-2.6, SSP2-4.5, SSP5-8.520.11.2026documented0
Sea ice and sea level20.11.2026documented0
Cost studies Austria (climate neutrality 2040)20.11.2026documented0
Cost studies Germany (KfW/Prognos, climate neutrality 2045)20.11.2026documented0
Cost studies European Union (impact assessment 2040)20.11.2026documented0

Climate models against measurement: air temperature

documented · To the chart →

Original data source
Models: CMIP6 archive “Pangeo” (Google Cloud), variable tas (air temperature at 2 m, monthly means), 41 models. Measurement: HadCRUT5 (Met Office Hadley Centre, version 5.2.0.0 according to the page) and Berkeley Earth.
Models and runs used
One run per model (mostly r1i1p1f1). In the “blended” comparison 30 of the 41 models (only models whose land mask is available). historical to 2014, ssp245 from 2015.
Period and reference period
Display 1850–2046 (default from 1979). Reference period 1961–1990, the own mean of the 360 months subtracted from every series (measurement and each model). Trend 1979–2020.
Calculation method
For each month the area mean over all grid cells, weighted by the cosine of latitude; departure from the 1961–1990 mean; 12-month moving average (for legibility only); trend as the linear slope of the monthly values (least squares) per decade. “Blended”: air temperature over land, sea surface over water, with the land fraction of the respective model.
Exclusions and limitations
No selection by how well models match the measurement. The filters “Without hot models” and “ECS 2.5 to 4 °C” are simple selections by climate sensitivity, not a statistical weighting. Not corrected: gaps in the coverage of the measurement, scenario SSP2-4.5 against the actual forcing since 2015. The models are not independent.
What can be derived from it
Whether the climate models warm more or less than the measurement on average since 1979, and how widely they scatter.
What cannot be derived from it
Which model is “right”, what causes the warming, or a forecast (one run per model, one scenario).
Download of the data used
  • tas_global_monatlich.zip (2.2 MB)
    Monthly series of the 23 models with at least three runs (the 18 models with fewer runs are not in the package; script 01 fetches them from the archive) Licence: CC BY-SA 4.0 bzw. CC BY-NC-SA 4.0 je Modell (siehe Dateikopf)
Calculation script
01_modellreihen_holen.py, 02_auswertung.py
Date
20.11.2026

Climate models against measurement: sea surface

documented · To the chart →

Original data source
Models: CMIP6 (Pangeo), variable tos, 41 models. Measurement: NOAA ERSSTv5.
Models and runs used
One run per model, historical to 2014, ssp245 from 2015.
Period and reference period
1850–2046 (measurement from 1854), reference period 1961–1990, trend 1979–2020.
Calculation method
Mean sea surface temperature between 60° South and 60° North, area-weighted (cosine of latitude); departure from the 1961–1990 mean; trend as linear slope per decade. Model and measurement describe the same quantity.
Exclusions and limitations
No selection by match. The sea poleward of 60° is left out. Scenario uncertainty is missing (one scenario).
What can be derived from it
How well the models match the warming of the sea surface between 60° S and 60° N; here model and measurement describe the same quantity.
What cannot be derived from it
Statements about the polar seas or about the causes.
Download of the data used
  • tos_global_monatlich.zip (1.7 MB)
    Monthly series of sea surface (60°S–60°N) of the 23 models with at least three runs Licence: CC BY-SA 4.0 bzw. CC BY-NC-SA 4.0 je Modell (siehe Dateikopf)
Calculation script
01_modellreihen_holen.py, 02_auswertung.py
Date
20.11.2026

Several runs per model

documented · To the chart →

Original data source
CMIP6 (Pangeo), variable tas, historical + ssp245.
Models and runs used
All runs with the same physics and forcing variant (label r<run number>i1p1f1) as the run shown first that exist for both periods: 23 models with at least three such runs, 286 runs; 236 in the charts. 50 of 72 runs of EC-Earth3 only start in 1970, have no 1961–1990 mean and are missing. 24 runs end before 2100 (7 until 2020, 17 until 2039) and end earlier in the charts.
Period and reference period
1850–2100 (display from 1950 or 1850), reference period 1961–1990 per run. Trend 1979–2020.
Calculation method
Global area mean per month (cosine of latitude), annual mean from 12 months, departure from the 1961–1990 mean of the respective run. Mean and range per year across the runs of a model, only where at least three runs have values. Trend per run as the linear slope of the annual means.
Exclusions and limitations
Plain air temperature, no “blended” comparison (multiple runs not available for it). Runs and models are not independent. Forcing uncertainty is not included. Measurement uncertainty is shown as the 95 % band of the 200 HadCRUT5 realisations (version 5.2.0.0), plus Berkeley Earth as a second measurement series.
What can be derived from it
How much of the scatter between runs is chance (initial conditions) and whether a single run is typical of its model.
What cannot be derived from it
A “blended” comparison with the measurement (the multi-run data show pure air temperature) or independent models: runs and models are related.
Download of the data used
  • jahresanomalien_tas_MIROC6.csv (76 KB)
    Example: annual anomalies of all 50 MIROC6 runs (the other 22 models are in the same folder) Licence: CC BY-SA 4.0
  • laeufe_liste.csv (18 KB)
    All 286 runs with period, licence and use Licence: je Modell
  • tas_global_monatlich.zip (2.2 MB)
    Monthly series of all runs Licence: CC BY-SA 4.0 bzw. CC BY-NC-SA 4.0 je Modell (siehe Dateikopf)
Calculation script
01_modellreihen_holen.py, 02_auswertung.py
Date
20.11.2026

Time windows, internal variability and model difference

documented · To the chart →

Original data source
As “Several runs per model”; measured trends HadCRUT5 and Berkeley Earth from annual means.
Models and runs used
The same 236 runs in 23 models.
Period and reference period
Trend windows 1950–2020, 1979–2020, 1990–2020 and 2000–2020; scatter per year 1950–2100.
Calculation method
Within: mean variance of the runs of a model (shown as standard deviation). Between: variance of the model means minus the noise of the runs (variance within divided by the number of runs). Only models with at least three runs with a value; at least five models per year. The measurement counts as “within range” if its trend lies between the smallest and largest trend of the runs of a model.
Exclusions and limitations
The 1961–1990 reference period makes the model differences look small at the start. One scenario (SSP2-4.5), scenario uncertainty is missing. Measurement uncertainty only indicated as the difference between the two measurement series.
What can be derived from it
How the result depends on the chosen time window and from when the difference between models exceeds the random scatter.
What cannot be derived from it
Whether a model is “too warm” from a short window: the shorter the period, the less it says about a model.
Download of the data used
Calculation script
02_auswertung.py
Date
20.11.2026

Quality check of the models

documented · To the chart →

Original data source
Models: CMIP6 (Pangeo), tas, pr, tos. Measurement: HadCRUT5 (trend, variability), Berkeley Earth (absolute mean and seasonal cycle 1951–1980), GPCP v2.3 and CMAP (precipitation), NOAA ERSSTv5 (sea).
Models and runs used
23 models with at least three runs (tas 286, pr 283 runs); sea 22 models (232 runs, MRI-ESM2-0 without tos); plus six Europe quantities (trend of temperature and precipitation in Central Europe, the Alpine region and Austria, measurement E-OBS, the first up to five runs per model).
Period and reference period
Trend and variability 1979–2020, mean and seasonal cycle 1951–1980. Precipitation and sea 1979–2020.
Calculation method
Eight figures per run: trend, absolute mean, seasonal cycle (largest minus smallest month of the climatology), variability of the annual means about the trend (air temperature); global precipitation mean and trend in percent per decade; sea trend and variability. Per model mean, smallest and largest value across the runs; comparison with the measurement: above / within range / below. With two precipitation products the measurement is the range of both.
Exclusions and limitations
No overall score. For the absolute mean the measurement depends on the definition (air above sea ice 14.10 °C, water below sea ice 14.70 °C). The two precipitation products contradict each other for the trend (GPCP +0.18, CMAP −1.49 % per decade, correlation of the annual series −0.13); the cause has not been checked. Models deliver air temperature everywhere, measurement series mix air and water and have gaps. Few runs give narrow ranges.
What can be derived from it
For which quantities the measurement lies within the range of a model’s runs, and whether the model lies above or below.
What cannot be derived from it
An overall grade or ranking of the models; a reliable verdict on the precipitation trend (the two measurement products contradict each other).
Download of the data used
Calculation script
02_auswertung.py
Date
20.11.2026

Glacier map

documented · To the chart →

Original data source
Outlines: Natural Earth (public domain). Glacier locations: WGMS (2026), doi:10.5904/wgms-fog-2026-02-10. Weather stations: GeoSphere Austria.
Models and runs used
Six glaciers: Pasterze, Hintereisferner, Vernagtferner, Kesselwandferner, Stubacher Sonnblickkees, Goldbergkees; three stations: Sonnblick, Obergurgl, Rudolfshütte.
Period and reference period
–
Calculation method
Simplified outlines (Douglas-Peucker), simple map projection (longitude and latitude linear). Info card from the series of the following entries.
Exclusions and limitations
The map is an orientation aid, not a survey document; glacier positions are the coordinates according to WGMS.
What can be derived from it
Where the six glaciers shown and the three weather stations are located.
What cannot be derived from it
Statements about all glaciers of the Alps: these are six selected ones.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Glaciers: length since 1850

documented · To the chart →

Original data source
WGMS (2026): Fluctuations of Glaciers (FoG) Database, table front_variation. Open access with correct citation.
Models and runs used
Six Austrian glaciers (147 measurements Pasterze 1851–2023, Hintereisferner 107, Vernagtferner 104, Kesselwandferner 61, Stubacher Sonnblickkees 63, Goldbergkees 174).
Period and reference period
From the first measurement period ending after 1850 to the last measurement (mostly 2023).
Calculation method
Sum of the measured changes in length; the sum starts at zero at the end of the first measurement period. Retreat per decade: difference of the sums at the start and end of a period, linear between the measurement years, converted to ten years.
Exclusions and limitations
Gaps in the series (up to three per glacier) are shown joined. Length reacts with a delay and depends on shape, slope and debris cover; it is not a direct climate signal.
What can be derived from it
How far the measured glacier ends have retreated since 1850 and in which periods the retreat was faster.
What cannot be derived from it
A direct climate signal: the length depends on shape, slope and debris cover and reacts with a delay.
Download of the data used
Calculation script
03_gletscher.py
Date
20.11.2026

Glaciers: mass balance

documented · To the chart →

Original data source
WGMS (2026), table mass_balance (annual_balance). Open access with correct citation.
Models and runs used
Six glaciers; series 1946–2025 (Stubacher Sonnblickkees from 1949, Hintereisferner and Kesselwandferner from 1953, Vernagtferner from 1965, Pasterze from 1980, Goldbergkees from 1989).
Period and reference period
Periods of the table: 1946–1979, 1980–1999, 2000–2025 (mean only with at least five years).
Calculation method
Annual total balance of the glacier (hydrological year, metres of water equivalent, averaged over the glacier area); the sum is the sum of the annual values.
Exclusions and limitations
Years without measurement are missing. The area to which the balance refers changes over time.
What can be derived from it
How much ice a glacier gains or loses per year (mass balance in water equivalent) and how it adds up.
What cannot be derived from it
The cause from these data alone; the balance also depends on winter snowfall.
Download of the data used
Calculation script
03_gletscher.py
Date
20.11.2026

Glaciers: summer temperature and mass balance

documented · To the chart →

Original data source
Mass balance: WGMS (2026). Temperature: GeoSphere Austria, station data v2 (1 month), doi:10.60669/923n-p390 (CC BY 4.0).
Models and runs used
Station per glacier: Sonnblick (Pasterze, Goldbergkees), Obergurgl (Hintereisferner, Vernagtferner, Kesselwandferner), Rudolfshütte (Stubacher Sonnblickkees); 37 to 73 years per glacier.
Period and reference period
Years with balance and all four months June to September.
Calculation method
Summer temperature as the mean June to September; relationship as correlation and slope. Additionally without the shared trend: the linear trend is subtracted from both series (r between −0.51 and −0.65).
Exclusions and limitations
The stations are not at the glacier but at a different height and distance. Winter snow, radiation and dust are not included. A relationship is an indication, not a complete proof of the cause.
What can be derived from it
That warm summers coincide with greater ice loss (correlation and slope, also without a common trend).
What cannot be derived from it
A complete proof of the cause: snowfall, radiation and dust are not in the data, and the stations lie lower than the glaciers.
Download of the data used
Calculation script
03_gletscher.py
Date
20.11.2026

Glaciers: finds and forecast signs (reports)

documented · To the chart →

Original data source
Reports from the University of Graz, the University of Innsbruck, the Historical Lexicon of Switzerland, E&G Quaternary Science Journal (CC BY 4.0), CNN/KSL, Yale Climate Connections and USGS; each with a link.
Models and runs used
–
Period and reference period
Reports from 2015 to 2026.
Calculation method
The reports were read at the original (20 November 2026); the summaries reproduce their statements. The screenshots show the header of the page on the reading date.
Exclusions and limitations
The USGS page blocks automated access, so there is no screenshot there. The section makes no statements of its own about the cause of earlier glacier extents; it reproduces the classification of the reports.
What can be derived from it
What documented reports of finds under the ice (wood, paths, objects) show in connection with the retreat.
What cannot be derived from it
A frequency or a climate proof from individual finds.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Europe, the Alps and Austria: models against measurements

documented · To the chart →

Original data source
Models: CMIP6 (Pangeo), tas and pr. Measurement: E-OBS (grid, temperature and precipitation); station Amstetten: GeoSphere Austria (monthly values).
Models and runs used
23 models, the first five runs of each model by run number (not by result), 97 runs; EC-Earth3 runs starting in 1970 are missing.
Period and reference period
Display 1950–2100, reference period 1961–1990, trend 1979–2020. Regions: Central Europe 45–55° N, 5–20° E; Alpine region 44–48.5° N, 5.5–16.5° E; Austria 46.3–49.1° N, 9.5–17.2° E.
Calculation method
Mean per region (cosine of latitude; measurement land cells only), departure from the 1961–1990 mean (temperature in °C, precipitation in percent), annual and seasonal values, trend as the linear slope per run; per model mean and range of runs. Station: months with at least 25 days, reference mean from 28 of 30 years, trend with at least 35 years.
Exclusions and limitations
Model boxes (about 100 km) average valleys and peaks and contain water; no height correction; only departures and trends are comparable. The uncertainty of E-OBS is not included. From E-OBS the package contains only the derived trends, not the series. The calculation is not yet a script in the package (status: numbers and method published, script to follow).
What can be derived from it
Whether models and measurement show similar trends in Central Europe, the Alpine region and Austria, for the year and seasons.
What cannot be derived from it
Compare absolute values (model boxes average valleys and peaks and contain water), or reliable small-scale statements; for precipitation the scatter is large.
Download of the data used
  • jahresreihen_gebiete.csv (155 KB)
    Annual series per region, quantity and period: model mean and range of model means, station Amstetten (without the E-OBS series, whose terms do not allow redistribution) Licence: abgeleitet; Modelle CC BY-SA 4.0 bzw. CC BY-NC-SA 4.0
  • trends_1979_2020_je_modell.csv (35 KB)
    Trend 1979–2020 per model (mean and range of runs) with the E-OBS trends as derived numbers and the station Licence: abgeleitet; Modelle CC BY-SA 4.0 bzw. CC BY-NC-SA 4.0
Calculation script
–
Date
20.11.2026

Time travel: stations, thermometers and reconstructions

documented · To the chart →

Original data source
Official station Amstetten (GeoSphere Austria), own station (Vantage), Ljungqvist (2010), Vostok ice core (Petit et al. 1999), CO₂: Law Dome (MacFarling Meure et al. 2006, NOAA NCEI) and Mauna Loa (NOAA GML).
Models and runs used
Official station: 84 annual means (1936–2025, gap 1989–1992) from 31,299 daily values, only years with at least 11 months. Own station: 12 annual means (2003–2015) from 4,484 daily values in the official time window. Ljungqvist: 200 decadal values. Vostok: 3,311 temperature and 283 CO₂ values. CO₂ of the last 2,000 years: Law Dome (164 values to 1958) and Mauna Loa (67 annual means 1959–2025).
Period and reference period
Shown are 90 years, 2,000 years and 12,000 and 420,000 years. Between 2,012 and 2,342 years ago there are no values in the CO₂ sources.
Calculation method
Annual means from daily values; the own station only in the time window in which the official one also has data. The reconstruction and ice-core series are displayed, not recalculated.
Exclusions and limitations
The series measure different things (station, Northern Hemisphere outside the tropics, Antarctica) and cannot simply be equated. The own station’s values are not adjusted. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
See orders of magnitude and developments over very different periods side by side, from the station Amstetten to millions of years.
What cannot be derived from it
Equate the series: station, Northern Hemisphere and Antarctica measure different things, and older values are reconstructed from traces.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Time travel back 66 and 540 million years

documented · To the chart →

Original data source
To 66 million years: CenCO2PIP 2023 (CO₂) and Westerhold et al. 2020 (temperature). Older: Scotese et al. 2021 (temperature) and Foster et al. 2017 (CO₂).
Models and runs used
CO₂ from 66 million years from indirect measurements (marine sediments, plant remains, soils) with 95 % uncertainty; the temperature from 66 million years is a model calculation fitted to such traces (Scotese et al. 2021, about every 5 to 10 million years); CO₂ before 66 million years from about 1,500 indirect single estimates (Foster et al. 2017, up to 420 million years).
Period and reference period
Last 66 and 540 million years respectively.
Calculation method
The series are displayed, not recalculated; at dataset changes they are not adjusted to each other (except ice core and reconstruction in their common period of the last 2,000 years, by mean offset). Changes are marked in the chart.
Exclusions and limitations
Only the last roughly 170 years (thermometers) and CO₂ since 1958 are measured; everything older is reconstructed from traces, the uncertainty grows with age, and other reconstructions (for example Judd et al. 2024) arrive at different ranges. The lines are not measurement series and their swings are not to be read as short-term fluctuations. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
Rough developments of CO₂ and temperature over 66 and 540 million years respectively.
What cannot be derived from it
Short-term fluctuations or exact values: the lines are reconstructions and model calculations with growing uncertainty.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Vostok ice core: temperature and CO₂ over 420,000 years

documented · To the chart →

Original data source
Petit et al. (1999), Vostok ice core, Antarctica.
Models and runs used
Temperature (deuterium in the ice): 3,311 measurements, age 0 to 422,766 years, mean spacing about 130 years. CO₂ (trapped air): 283 measurements, gas age up to 414,085 years, mean spacing about 1,500 years.
Period and reference period
Last 420,000 years.
Calculation method
Temperature in 500-year means, CO₂ as single values, connected linearly in between. The correlation temperature–CO₂ is calculated without and with a time offset (−3,000 to +6,000 years in 50-year steps), with the best offset per section.
Exclusions and limitations
The age of the ice is only known to a few hundred to a thousand years; the temperature applies to Antarctica and not globally. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
The common course of Antarctic temperature and CO₂ over four ice-age cycles and, on average, the time sequence.
What cannot be derived from it
A cause-effect relation (the search for the best offset is exploratory), the global temperature, or an exact dating (ice and air are known to a few hundred to a thousand years).
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Reconstruction of the last 2,000 years (Ljungqvist)

documented · To the chart →

Original data source
Ljungqvist (2010), temperature reconstruction for the Northern Hemisphere outside the tropics; measured temperature in the same dataset.
Models and runs used
200 decadal values (year 0 to 1999) with an uncertainty band (±2 standard deviations), from 30 proxy series; measured temperature: 15 decades (1850–1999).
Period and reference period
Year 0 to 1999.
Calculation method
No recalculation of our own: the file is displayed, and the numbers in the tooltip come directly from the table.
Exclusions and limitations
Decadal values smooth out short rises; the area is the Northern Hemisphere outside the tropics. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
The course of temperature in the Northern Hemisphere outside the tropics over 2,000 years in decadal values.
What cannot be derived from it
Global means or short-term fluctuations: decadal values smooth.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Rates of rise: how fast was warming in the past?

documented · To the chart →

Original data source
Vostok (3,311 values), Ljungqvist reconstruction (200 decades) and thermometers (15 decades).
Models and runs used
Six linear slopes over fixed time windows; the strongest 30- and 50-year rises before 1850 from all possible windows.
Period and reference period
Windows depend on the series (see chart).
Calculation method
Linear slope (least squares), scaled to °C per century.
Exclusions and limitations
Older data are smoothed and show short rises only weakly; the ice core is too coarse for short periods. The rates are therefore only comparable to a limited extent. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
Compare slopes over fixed time windows in °C per century.
What cannot be derived from it
Say whether there were similarly fast short rises in the past: the older data are too coarsely resolved for that.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Map and globe: sea, land, El Niño, CO₂ in the air

documented · To the chart →

Original data source
Sea: NOAA ERSSTv5 (Huang et al. 2017), land: Berkeley Earth, near-surface CO₂: NOAA CarbonTracker (a computer model, adjusted with measurements). Display: MapLibre GL (BSD), Natural Earth (public domain), NASA GIBS Blue Marble (public domain).
Models and runs used
Sea: 2,064 monthly maps (Jan 1854 – Dec 2025), each 89 × 180 cells (2°), giving 172 annual maps. Land: 2,100 monthly maps (Jan 1850 – Dec 2024), each 180 × 360 cells (1°), giving 175 annual maps. Delivered to the page: two files (sea 2.8 MB, land 11.3 MB, compressed about 0.9 and 1.5 MB), value step 0.05 °C.
Period and reference period
Sea 1854–2025, land 1850–2024; reference period 1961–1990; trend sea 1979–2025, land 1979–2024.
Calculation method
Annual mean per cell (sea: cells with fewer than nine valid months stay empty); departure from the 1961–1990 mean of the same cell; trend as the linear slope of the annual values per cell (cells with fewer than 80 % valid years stay empty); the global line for the sea is the mean 60° S to 60° N, weighted by the cosine of latitude. The El Niño value is a relative Niño-3.4 series.
Exclusions and limitations
The sea poleward of 60° (sea ice) is left out; coastlines look softer on the globe (bilinear interpolation). The cells are large (sea about 222 km, land about 111 km at the equator). The CO₂ layer is a computer model and not a measurement. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
Where and how strongly sea and land have warmed since 1979 (trend per cell) and how El Niño changes the sea.
What cannot be derived from it
Causes or reliable values before 1900 (sparsely covered, strongly interpolated); the cells are large, coastlines look softer.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

CO₂ in the air: amount, origin and fate

documented · To the chart →

Original data source
NOAA Global Monitoring Laboratory (global annual mean), pre-industrial 278 ppm after IPCC AR6; Global Carbon Budget 2025 (Friedlingstein et al.).
Models and runs used
Human emissions 2024: 11.6 bn t carbon; total emissions 1850–2024: 751 bn t C (503 from combustion and cement, 248 from land use); of which 39 % air, 27 % sea, 23 % land, 11 % unallocated.
Period and reference period
Annual mean 2025, budget 1850–2024, mean 2015–2024.
Calculation method
“Share of the exchange” (about 5 %) is the ratio of annual amounts (11.6 against roughly 200 bn t C per year after IPCC AR6); “share of the stock” (about 35 %) is the increase in the air since 1750 relative to today’s amount. Conversion: 1 ppm ≙ 2.124 bn t C ≙ 7.78 bn t CO₂.
Exclusions and limitations
The budget has a gap of 80 bn t C (11 %) that is unallocated; the natural exchange of about 200 bn t C is a rough figure. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
The order of magnitude of CO₂ in the air and how emissions, sea, land and air balance.
What cannot be derived from it
The origin of single molecules or an allocation of the budget gap of 11 %.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

CO₂ seasonal cycle, Keeling spiral and residence time

documented · To the chart →

Original data source
NOAA Global Monitoring Laboratory, Mauna Loa (monthly means, columns “average” and “deseasonalized”); Joos et al. (2013) for the decay curve.
Models and runs used
Monthly means since 1958; seasonal cycle as the mean 2015–2025.
Period and reference period
Spiral: from 1958; seasonal cycle: mean 2015 to 2025.
Calculation method
The spiral and seasonal cycle come directly from the NOAA monthly means (the difference is the seasonal cycle). The decay curve is the impulse response from Joos et al. (2013), a sum of exponential functions: f(t) = a₀ + Σ aᵢ e^(−t/τᵢ) with a₀ = 0.2173, a₁ = 0.2240, a₂ = 0.2824, a₃ = 0.2763 and τ₁ = 394.4, τ₂ = 36.54, τ₃ = 4.304 years.
Exclusions and limitations
The decay curve is a model of carbon uptake by sea and land and not a measurement; Mauna Loa is a single station. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
The seasonal cycle and rise of CO₂ at Mauna Loa and, as a model, how long a CO₂ pulse stays in the air.
What cannot be derived from it
Read the decay curve as a measured value (it is a model), or take Mauna Loa for the whole Earth (it is a single station).
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Countries and scenarios: calculator for CO₂ savings, warming and costs

documented · To the chart →

Original data source
CO₂ emissions per country: Our World in Data (Global Carbon Project, production-based, fossil sources and industry); relation between amount and warming (TCRE) after IPCC AR6: 0.45 °C per 1,000 bn t CO₂, likely range 0.27 to 0.63.
Models and runs used
Freely selectable countries and targets (for example EU targets, Austria climate-neutral 2040).
Period and reference period
Reference path: emissions stay at the 2025 level; scenarios to 2050.
Calculation method
Saved amount = sum of avoided annual emissions relative to the reference path; temperature effect linear via the TCRE; EU targets (relative to 1990) are converted to the 2025 level of the selection. Costs: rising marginal-cost curve with our own support points (lower estimate 0–10, 10–50, 50–100, 100–200 €/t; upper 0–20, 20–100, 100–200, 200–400 €/t), without discounting.
Exclusions and limitations
Only CO₂ from fossil sources and industry; without land use, methane, nitrous oxide, sinks, feedbacks, production shifts and the behaviour of other countries. The cost support points are our own assumptions, an extrapolation above 50 % reduction, and not a forecast. Climate damage and co-benefits are not included. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
The order of magnitude of what a country scenario roughly means for warming and costs, under the stated assumptions.
What cannot be derived from it
A prediction, real costs (own assumptions, no climate damage, no co-benefits), or effects of methane, land use and other countries.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Alpine trend map and height profile (E-OBS)

documented · To the chart →

Original data source
E-OBS v31.0e (ECA&D/KNMI), daily mean temperature (tg) and elevation grid, 0.25° grid; borders: Natural Earth (public domain); glacier locations: WGMS (2026); stations: GeoSphere Austria.
Models and runs used
1,450 cells with a value in the section 43.1–49.9° N, 4.1–18.4° E (land, without sea).
Period and reference period
Months 1978/79 to 2020; trend 1979–2020.
Calculation method
Monthly mean per cell only with at least 25 days; season only with all 3 months (year: 12), winter with the December of the previous year; trend as the linear slope of the annual values per cell, only with at least 35 years, in °C per decade. Height profile: cells by mean height in five bands (to 500, 500–1,000, 1,000–1,500, 1,500–2,000, from 2,000 m), mean and 10th–90th percentile per band; fitted line of trend against height across all cells.
Exclusions and limitations
Measurement only, no models. Cells are averages over about 18 × 28 km; the height is the mean cell height. In mountains with few stations E-OBS is less certain; neighbouring cells are not independent, and the high cells lie in the centre of the Alps, so the relation with height cannot be cleanly separated from location. The raw E-OBS data are not in the package (reference to the source only). No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
Where in the Alpine region the warming 1979–2020 was stronger or weaker, and whether the trend is related to cell height (as an indication).
What cannot be derived from it
Cleanly separate the height relation from location (the high cells lie in the centre of the Alps), or assess individual valleys and peaks.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Model map Alpine region: temperature trend, models against measurement

documented · To the chart →

Original data source
Models: CMIP6 (Pangeo), tas (air temperature at 2 m), historical to 2014 and ssp245 from 2015. Measurement: E-OBS v31.0e (daily mean temperature, 0.25°), averaged onto the 0.5° grid.
Models and runs used
23 models, the first up to five runs of each model by run number (not by result), 97 runs; the 1979–2020 trend needs no 1961–1990 reference period.
Period and reference period
Trend 1979–2020 (winter with December 1978); section 43.25–49.75° N, 4.25–18.25° E.
Calculation method
Per run annual and seasonal means per cell of the model grid, linear slope (°C per decade), then bilinearly converted to the common 0.5° grid; per model the mean of the runs, “mean of all models” = mean of the model means. Difference = model minus measurement only in land cells with an E-OBS value (367 cells). Figures: mean difference, root mean square of the cell differences, correlation across the cells (pattern similarity).
Exclusions and limitations
The model map is interpolated (model cells about 100–300 km, only a few in the section) and cannot reproduce valleys and elevations; no height correction. With at most five runs a lot of internal variability remains in a cell’s trend. The raw E-OBS data are not in the package (reference to the source only). No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
Whether the models match the area mean and the coarse spatial pattern of the measurement.
What cannot be derived from it
Fine patterns of the Alps: the model map is interpolated from coarse cells; single cells cannot be interpreted.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Scenarios to 2100: SSP1-2.6, SSP2-4.5, SSP5-8.5

documented · To the chart →

Original data source
Models: CMIP6 (Pangeo), tas, experiments historical, ssp126, ssp245, ssp585. Measurement: HadCRUT5 (annual means).
Models and runs used
22 models, 76 runs (for each model the first up to five by run number, not by result, that are available in all three scenarios and have a full 1961–1990 mean); GISS-E2-1-G has no such runs.
Period and reference period
1950–2100; reference period 1961–1990 per run; warming 2041–2060 and 2081–2100 relative to 1850–1900 and 1995–2014.
Calculation method
Global mean per month (cosine of latitude), annual mean only from twelve months, departure from the run’s 1961–1990 mean; mean of the runs per model; model mean and range (smallest and largest model value) per year. Warming per run as the difference of two period means, averaged per model; table: mean, median, range across the models.
Exclusions and limitations
Scenarios are if-then calculations and not forecasts. Nothing was filtered by agreement with the measurement (including no “hot” models). The figures are not the IPCC’s assessed range. SSP3-7.0 and SSP1-1.9 are missing. The measurement mixes air and water, the models show pure air temperature. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
How strongly the warming to 2100 calculated by the models depends on the emission assumptions, and how widely the models scatter within a scenario.
What cannot be derived from it
A forecast (scenarios are if-then calculations) or the IPCC’s assessed range.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Sea ice and sea level

documented · To the chart →

Original data source
Sea ice: NSIDC Sea Ice Index, version 4 (doi:10.7265/a98x-0f50); Antarctic ice sheet: IMBIE (Shepherd et al. 2018) and update to 2020; sea level: University of Colorado Sea Level Research Group, version 2026_rel2 (Nerem et al. 2018, PNAS, doi:10.1073/pnas.1717312115).
Models and runs used
Three measurement series: sea-ice extent Arctic (March and September), Antarctic (September and February), global sea level.
Period and reference period
Sea ice 1979–2026, sea level since 1993.
Calculation method
No model calculation of our own, only the measured values of the sources. Sea ice: monthly means from daily values (every second day until 1987), area with at least 15 % ice cover; trends as the linear slope per decade. Sea level: change relative to the mean of the first year; seasonal cycle and land uplift and subsidence (GIA) are removed by the authors.
Exclusions and limitations
The area is measured, not the ice volume; the ice on land (Antarctic ice sheet) is measured differently. Sea level is a global mean; regionally it rises and falls differently. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
Read satellite measurement series of sea-ice extent (Arctic, Antarctic) and global sea level since 1993.
What cannot be derived from it
The ice volume, the ice on land or regional sea level: the area or a global mean is measured.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Cost studies Austria (climate neutrality 2040)

documented · To the chart →

Original data source
Studies on the “climate neutrality 2040” scenario (KN40) against the WEM scenario (with existing measures, as of 1 January 2022), among others IHS/WIFO; the exact citation is given at the chart.
Models and runs used
Additional investment 6.4 to 11.2 bn € per year (1.1 to 1.9 % of GDP), total investment 46.8 to 64.5 bn € per year, broken down by sector.
Period and reference period
Price basis 2024, period 2024 to 2040 unless stated otherwise.
Calculation method
Reproduction of the study values; additional investment is total investment minus replacement investment and the investment already in the WEM scenario.
Exclusions and limitations
Nothing here is calculated by ourselves: numbers are reproduced from the studies. Quality levels: A directly from the study, B derived by us from study values, C approximation. What the studies do not state is shown as a gap. The studies answer different questions with different reference points and are therefore not combined into one number. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
Read side by side what the respective study states (investment, scenarios, quality levels A/B/C).
What cannot be derived from it
Form one common number, derive net costs without benefits and avoided damage, or calculate own costs: the page calculates nothing itself here.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Cost studies Germany (KfW/Prognos, climate neutrality 2045)

documented · To the chart →

Original data source
KfW Research, Fokus Volkswirtschaft No. 561 (12 October 2026), a study by Prognos commissioned by KfW; the paper reflects the authors’ views.
Models and runs used
Four transformation scenarios; climate-protection investment 270 to 300 bn € per year, net investment 130 to 150, additional to the status quo 60 to 70 bn € per year.
Period and reference period
Period 2025 to 2045, prices of 2024.
Calculation method
Reproduction of the study values; the three investment concepts (climate-protection investment, net investment, additional to the status quo) are shown separately because they must not be confused.
Exclusions and limitations
Nothing here is calculated by ourselves: numbers are reproduced from the studies. Quality levels: A directly from the study, B derived by us from study values, C approximation. What the studies do not state is shown as a gap. The studies answer different questions with different reference points and are therefore not combined into one number. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
Read side by side what the respective study states (investment, scenarios, quality levels A/B/C).
What cannot be derived from it
Form one common number, derive net costs without benefits and avoided damage, or calculate own costs: the page calculates nothing itself here.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026

Cost studies European Union (impact assessment 2040)

documented · To the chart →

Original data source
European Commission, impact assessment SWD(2024) 63 on the 2040 target, based on the official chart data (Annex 8, version 23 April 2024; models PRIMES and E-QUEST) and the summary SWD(2024) 64.
Models and runs used
Target variants S1 (up to 80 %), S2 (at least 85 %, also the reference) and S3 (at least 90 %) and LIFE (demand-side changes); for example investment need in the energy system excluding transport 3.3 % (S2) and 3.7 % (S3) of GDP, annual average 2031–2040.
Period and reference period
Decades 2011–2050, depending on the quantity.
Calculation method
Reproduction of the Commission’s chart data; units and price basis are not stated in the data file and were derived from the value range (level B). The main report was not retrievable automatically, so the definition of “investment needs” could not be checked.
Exclusions and limitations
Nothing here is calculated by ourselves: numbers are reproduced from the studies. Quality levels: A directly from the study, B derived by us from study values, C approximation. What the studies do not state is shown as a gap. The studies answer different questions with different reference points and are therefore not combined into one number. No download and no script in the package; the calculation methods are in the explanation boxes at the chart.
What can be derived from it
Read side by side what the respective study states (investment, scenarios, quality levels A/B/C).
What cannot be derived from it
Form one common number, derive net costs without benefits and avoided damage, or calculate own costs: the page calculates nothing itself here.
Download of the data used
No separate downloads for this chart.
Calculation script
–
Date
20.11.2026
Recalculating in three steps

Recalculating in three steps

  1. Fetch a model series: python skripte/01_modellreihen_holen.py tas MIROC6 r1i1p1f1 series.csv (loads anonymously from the public archive).
  2. Evaluate: python skripte/02_auswertung.py monatlich/tas_global_monatlich.zip MIROC6 gives the 1979–2020 trend and variability per run; for MIROC6 (50 runs) the mean trend is 0.173 °C per decade (range 0.140 to 0.234), as in the figures file.
  3. Glaciers: python skripte/03_gletscher.py <WGMS folder> Pasterze 566; the sum of the change in length of the Pasterze since the first measurement period after 1850 is −2,636.4 m.

The scripts were checked against the published files: a run fetched afresh from the archive agrees month by month (largest difference 0.0), and so do the MIROC6 trends and the Pasterze sum.

Citing

Citing

ruthner.at/klima, Data and methods, data package of 20 November 2026. For model data also the authors of the models (in the file header) and CMIP6 (Eyring et al. 2016, doi:10.5194/gmd-9-1937-2016); for glacier data WGMS (2026), doi:10.5904/wgms-fog-2026-02-10; for temperature data GeoSphere Austria, doi:10.60669/923n-p390.