Climate.ruthner.at← ruthner.atWeather Amstetten →DEEN

Methods, questions and glossary

How was it calculated, where do the data come from, what do the terms mean? Plus the method register and the overview of sources.

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.

10 — Questions and answers

Terms and methods briefly explained

What do the technical terms mean, what was calculated, and where does the data come from? Click a question and the answer expands.

What is a temperature anomaly?

Instead of the absolute temperature, most charts show by how many degrees a value lies above or below a reference mean. Depending on the chart, the reference mean is 1961–1990 (mostly) or 1979–1983 (Spencer’s original); it is stated alongside each one. Red/yellow means warmer, blue colder than in this reference period.

Why anomalies and not the temperature itself?

Locations and instruments differ in their level (altitude, site, measurement height). Anomalies from each series’ own mean make changes comparable even when series lie at different levels. This is standard practice for trends.

What is a trend “per decade”?

The slope of a straight line fitted through all values in the period (least-squares method), converted to ten years. A trend of +0.13 °C per decade means: on average an increase of 0.13 °C per ten years over the chosen period. It depends on the period and says nothing about individual years.

What is NOAA ERSSTv5?

A sea surface temperature series from the US weather agency NOAA, since 1854, based on ship and buoy measurements, with gaps filled statistically, on a 2° grid (about 200 km). The page uses 1979–2025.

What are Berkeley Earth and HadCRUT5?

Two independent analyses of the Earth’s surface temperature from weather stations (over land) and sea measurements. Berkeley Earth (University of California) also supplies a 1° grid for land, which the page uses for the land colours. HadCRUT5 comes from the British Met Office. Both were used here only as global monthly means.

What are CMIP6 and the “climate models”?

CMIP6 is an international comparison project in which research centres run their climate models (computer simulations of atmosphere and ocean) with the same inputs. Used here: 41 models, one run each, sea surface and air temperature. These are calculation results, not measurements.

What do “historical” and “SSP2-4.5” mean?

“historical” are hindcasts of the past up to 2014 with the actual forcings (greenhouse gases, volcanoes, Sun). SSP2-4.5 is a scenario from 2015 with an assumed, intermediate development of emissions. Everything from 2026 onwards is therefore a projection under this assumption, not a measurement and not a prediction of the actual course.

How do you get from raw data to the finished chart?

From the raw data to the line in the chart there are six steps, the same for the models (A) and the measurements (B).

A · Climate modelsRaw datatos, tas · per cellSelection1 run per modelWeightingcos φ-weightedAnomalyminus 1961–1990Smoothing12-month meanResult41 lines + meanB · MeasurementsRaw dataERSSTv5, etc.SelectionBerkeley EarthWeightingcos φAnomalyminus 1961–1990Smoothing12-month meanResultMeasurement lineC · ComparisonBoth on the same basisTrend 1979–2025 per model and measurementChart and tableModel range, mean, measurement in the same pictureAll steps run in Python (xarray, dask); the raw data stay unchanged, and each step produces a new intermediate file.

Model example: A grid of several tens of thousands of grid cells becomes a single number per month. After anomaly and smoothing, one curve remains (1850–2100), from which the trend is calculated. The formulas for this are in the next question.

Which formulas are behind this?

The five calculation steps of this page and the two basic physical relationships without which the model discussion cannot be understood.

1 · Area-weighted mean. Grid cells at the poles are smaller than at the equator, so each cell counts with the cosine of its latitude φ:

T¯=∑iwiTi∑iwi,wi=cos(φi)

2 · Anomaly. From each value the mean of the same series over the reference period 1961–1990 is subtracted: for monthly series the mean of the 360 months, for annual series that of the 30 years. For monthly values the seasonal cycle is therefore initially retained; it drops out only through the annual means or the 12-month smoothing (the formula shows the case of monthly series):

TAnomaly(t)=T(t)−1360∑m∈1961–1990Tm

3 · Trend. The straight line through the points (time t, temperature T) by the least-squares method; b is the slope:

b=∑(tj−t¯)(Tj−T¯)∑(tj−t¯)2(°C per year, times 10 = °C per decade)

4 · Radiative forcing of CO₂. With C as the CO₂ concentration and C₀ as the baseline value (approximation after Myhre et al. 1998). A doubling gives about 3.7 W/m²:

ΔF=5.35·ln(CC0)W/m²

5 · Climate sensitivity. The warming ΔT is the forcing ΔF times a sensitivity parameter λ (°C per W/m²). It is precisely this λ that distinguishes the models (see above); it is not prescribed but results from clouds, water vapour, ice and ocean in the model:

ΔT=λ·ΔF

On the greenhouse effect: The equilibrium temperature of the Earth without an atmosphere follows from the Stefan-Boltzmann law, with albedo α ≈ 0.3, solar constant S ≈ 1,361 W/m² and σ = 5.67·10⁻⁸ W/(m²K⁴). The measured average is about 288 K; the difference of about 33 K is the greenhouse effect:

Te=(1−α)·S4σ4≈255 K
What is a climate model, and how does it work?

A climate model is a computer program that divides the Earth into a grid of cells (atmosphere, ocean, ice, land) and, for each cell, steps forward the physical equations for flow, heat, water vapour and radiation. What the grid shows too coarsely (for example individual clouds) is replaced by simplified assumptions (“parameterisations”). A model does not predict tomorrow’s weather, but calculates how the climate develops on average over decades. Each research centre builds its own model, which is why they differ.

What are “forcings”?

Forcings are the external influences prescribed to a model: the concentration of greenhouse gases (CO₂, methane), particles from volcanoes and industry, solar radiation and land use. In the “historical” run, the forcings actually measured or reconstructed are used. Beforehand, a control run (“piControl”) with constant, pre-industrial forcings is carried out; the model starts from it. It shows how much the climate fluctuates by itself without any external change.

What is a scenario, and what do SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5 mean?

Nobody knows the forcings for the future, so the models run several assumptions (“scenarios”). SSP stands for “Shared Socioeconomic Pathway”, a development path for population, economy and energy use. The second number is the additional radiative forcing in 2100 in watts per square metre compared with pre-industrial: SSP1-2.6 (sharply falling emissions), SSP2-4.5 (middle road, used here), SSP3-7.0 (high emissions) and SSP5-8.5 (very high emissions). Scenarios are assumptions, not predictions. Which of them occurs depends on decisions, and no probability is assigned to them. On this page only SSP2-4.5 is calculated; until about 2030 the scenarios differ very little.

What is the difference between a forecast, a projection and a hindcast?

A forecast concerns the actual course of events (for example the weather of the next few days). A projection is an “if-then” calculation: if the forcings follow the scenario, this development results. A hindcast re-calculates the past with known forcings. In the charts, everything up to the end of 2014 is hindcast, and from 2015 it is a projection under SSP2-4.5. The models do not know the actual events from 2015 onwards (for example a real volcanic eruption or a real El Niño); the comparison with the measurement for 2015 to 2025 is therefore not a pure test of the models.

What does “r1i1p1f1” mean, and what is an ensemble?

The abbreviation names a single model run: r = realisation (start from a different state), i = initialisation, p = physics version, f = forcing version. Because the climate has chaotic fluctuations (for example El Niño), the same model gives a slightly different course with a slightly different start. Several runs of the same model are called an ensemble. The average over many runs shows the development caused by the forcing, a single run additionally contains random fluctuations. In the two main charts one run per model is used, mostly r1i1p1f1; how much several runs differ is shown in the “For specialists” block.

Why do the models give different results?

The models have different grids, calculate clouds, sea ice and aerosols differently, and react with different strength to more CO₂. One measure of this is the equilibrium climate sensitivity, that is, the warming for a doubling of the CO₂ content: for the CMIP6 models it ranges from about 1.8 to 5.6 °C (Zelinka et al. 2020). This is why the spread between the lines is large, and why the page shows not just the mean but all models. The model mean is a calculated value and not a better truth than the individual models.

What is the greenhouse effect?

The Earth is warmed by the Sun and radiates heat into space. Water vapour, clouds, carbon dioxide and other gases in the atmosphere retain part of this heat radiation, which is why the surface is on average about 33 °C warmer than it would be without these gases. Water vapour contributes the largest part (see the NASA analysis under “Links”). The physical effect of CO₂ is well established. What is studied and discussed scientifically is in particular the exact strength of the warming, the feedbacks in the climate system and the regional impacts. The IPCC assesses the human influence on the current global warming as dominant (AR6: for 2010–2019 about 1.07 °C of human-caused warming at about 1.09 °C of observed warming relative to 1850–1900).

What is the difference between sea surface temperature and air temperature?

Sea surface temperature is the water temperature of the uppermost layer and rises slowly, because water stores a lot of heat. Air temperature is measured 2 m above the ground or water and fluctuates more, especially over land. In the measurement series of surface temperature (HadCRUT5, Berkeley Earth), air temperature is used over land and water temperature over the sea; the model air temperature is at 2 m height everywhere. This is why the page has separate charts for sea and air.

Why is the data before 1900 less certain?

Before 1900 there were only a few measurements at sea, taken by ships along the trade routes, and the instruments (buckets, thermometers) and measurement methods changed. Large areas, especially in the Southern Hemisphere, were never measured. NOAA and others fill these gaps statistically; the course there is therefore partly calculated. Over land, too, there were only a few stations before 1880, and the gridded values from Berkeley Earth are heavily interpolated there. The maps and curves before 1900 are therefore shaded lightly on the page and marked with a note. Statements about individual years are weak there, statements about decades better.

Why does CO₂ fluctuate over the year?

From May to September, plants in the Northern Hemisphere, where most of the forest and farmland lies, take CO₂ from the air through photosynthesis; the level falls and reaches its low in September or October. From October the leaves fall and photosynthesis rests, but soils and leaf litter continue to decompose and release CO₂, and heating adds to it; the level rises until about May. The swing at Mauna Loa is about 7 ppm, considerably more near the ground over Siberia or northern China, and hardly anything over the South Pacific. Plants balance out the seasonal cycle within a year; the rise of about 2.4 ppm per year remains. See the section CO₂, tab “Memory”.

What is a proxy, and what is an ice core?

A proxy is a natural archive from which past temperatures are derived indirectly: tree rings, sediments, stalagmites, ice. An ice core is a drill core from the ice sheet: the ratio of the hydrogen isotopes (deuterium) in the ice indicates the temperature of Antarctica at the time, and the air trapped in the ice bubbles indicates the CO₂ content at the time. The age is accurate only to within a few hundred to a thousand years.

What is the Ljungqvist reconstruction?

A reconstruction of the temperature of the extra-tropical Northern Hemisphere (30–90 °N) over the last 2,000 years from 30 proxy series in decadal values (Ljungqvist 2010). It is smoothed over decades and applies to this region, not to the whole Earth.

What does “Calculate offset” mean for temperature and CO₂?

The program shifts the CO₂ curve in time against the temperature curve and looks for the shift at which the two lie best on top of each other (highest correlation). Because of the dating uncertainty of ice and air, the result is a guide value for the timing and says nothing about the cause.

What are El Niño and La Niña?

Fluctuations in sea temperature in the tropical Pacific that influence the weather worldwide: El Niño = unusually warm water, La Niña = unusually cold water. Here, years are marked as El Niño or La Niña years if the Niño 3.4 region (5°S–5°N, 170°W–120°W) was at least 0.5 °C warmer or colder, respectively, in the annual mean than the mean of all seas (60°S–60°N); this removes the long-term warming trend. The official classification is finer (running three-month means).

Why are volcanic eruptions marked on the timeline?

Large explosive eruptions throw sulphur gases into the stratosphere. There they form fine suspended particles (aerosols) that reflect part of the sunlight, and the Earth cools slightly for one to three years. Three eruptions of the measurement period are marked: El Chichón (Mexico, 1982), Pinatubo (Philippines, 1991; the strongest cooling, up to about 0.5 °C in the global mean) and Hunga Tonga–Hunga Haʻapai (Pacific, 2022; an underwater eruption with a great deal of water vapour, whose temperature effect is small and not yet conclusively settled). They are shown as markers so that fluctuations in the curve can be put in context; in 1982 a strong El Niño overlaps the effect. The years come from generally known information; they are not derived from the temperature data.

How large are the individual calculation points (grid cells)?

The data lie as a grid over the Earth: each cell stands for the mean over its whole area. The sea temperature (NOAA) has cells of 2° × 2° (at the equator about 222 × 222 km, at 60° latitude only 111 × 222 km), the land (Berkeley Earth) 1° × 1° (about 111 km). Depending on the model, the 41 climate models have cells of about 56 to 417 km width (median about 139 km), and the ocean components mostly about 100 km. The exact table is in the “Statistics on the data” in the section “Land and sea”. A place like Amstetten lies in a cell of about 150 × 220 km (2° grid) and is averaged there with its surroundings.

What is cosine weighting?

When averaging over the globe, grid cells near the poles count for less because they are smaller. Weighting is done with the cosine of the latitude, so that every area counts equally.

Where do the colours and images on the globe come from?

The Earth’s surface comes from the NASA “Blue Marble” tiles (public domain), the coasts and national borders from Natural Earth. The temperature colours are calculated in the browser from the grid data (bilinearly from 2° or 1° cells). This is why coastal edges look softer than they are in reality.

What does the page not show?

The page’s own data analyses alone do not provide complete proof of the causes: temperature series alone do not show why it became warmer or colder. For the scientific assessment the results of the IPCC and further research are therefore used in addition. The IPCC statement and the objections to it are in the section “Positions”. The data gaps, the selection (in the main charts one run per model) and the uncertainties are explained in the expandable notes. The weather at my front door in Amstetten is on the weather page.

11 — Glossary

Technical terms briefly explained

The most important terms of this page in plain words, in alphabetical order.

Deviation (anomaly)
The difference between a measured value and a reference mean, here mostly the 1961–1990 mean. "+1 °C" means one degree warmer than this mean, not "1 °C warm".
CBAM
The EU carbon border levy: importers of certain goods (for example steel, aluminium, fertiliser) pay for the CO₂ emitted during production.
CMIP6
The sixth coupled model intercomparison project. Many research groups run the same cases (the past and future scenarios), and the results are freely available. Here, 41 models are plotted against the measurements.
Ice core
A cylinder of glacier or polar ice. Trapped air bubbles and the composition of the ice give information about CO₂ and temperature in earlier times (here Vostok in Antarctica).
Ice sheet
The ice on land (Antarctica, Greenland). Not to be confused with sea ice, which floats on the sea.
El Niño and La Niña
A natural oscillation in the tropical Pacific that can change the global temperature by a few tenths of a degree. El Niño is the warm phase, La Niña the cool phase.
GIA
Uplift and subsidence of the Earth's crust after the last ice age. It is removed from sea level data because it has nothing to do with the water.
Marginal costs and abatement costs
The cost of avoiding one more tonne of CO₂ (marginal cost) and the total cost of all measures (abatement cost). They usually rise with the share that is avoided.
HadCRUT5, ERSSTv5, Berkeley Earth
Temperature measurement series: HadCRUT5 (air temperature, Met Office and University of East Anglia), ERSSTv5 (sea surface, NOAA) and Berkeley Earth (land and air). They are based on thermometer, ship and buoy measurements.
IPCC
The Intergovernmental Panel on Climate Change of the United Nations. It summarises the state of research in assessment reports, most recently the sixth (AR6, 2021 to 2023).
Annual cycle
The regular fluctuation over the course of a year, for example less CO₂ in the air in summer (plants take it up) and more in winter.
Keeling curve
The measurement series of the CO₂ content at Mauna Loa (Hawaii) since 1958, with an annual cycle and a steady rise.
Little Ice Age
A cool phase from about 1300 to 1900, coldest around 1700. The causes have not been conclusively established.
Climate sensitivity
The warming for a doubling of the CO₂ content. In its latest report, the IPCC gives 2.5 to 4 °C as the likely range.
Climate model
A computer program that simulates heat, wind and water in many boxes of the Earth. The result is a calculation under certain assumptions, not a prediction.
Sea ice extent
The area where at least 15% of the sea is covered with ice (satellite measurement). This is not the same as the ice volume.
Model mean
The mean of all model lines in a chart.
ppm (ppmv)
Parts per million. 426 ppm CO₂ means: out of one million air molecules, 426 are CO₂ molecules.
Proxy data
Indirect measurements of temperature in earlier times, for example from ice, tree rings or sediments, because there were no thermometers then.
Reconstruction
A temperature history of the past calculated from proxy data. It is less accurate than a measurement, as the uncertainty bands show.
SSP2-4.5
An IPCC scenario for the future: a middle path for population, economy and emissions ("middle of the road"). It is not a prediction.
Urban heat (UHI)
Cities are warmer than their surroundings (asphalt, buildings). This can distort measurement series from stations in growing cities.
TCRE
The relationship between the total CO₂ emitted and the warming. The IPCC gives 0.45 °C per 1,000 billion tonnes of CO₂ (range 0.27 to 0.63). This is the basis of the calculators on this page.
Trend
The slope of a best-fit line through the values, for example "°C per decade". It depends on the chosen period.
Gt (billion tonnes)
One gigatonne (Gt) is one billion tonnes. "bn t CO₂" on this page means the same.
About this page

Who is behind it, and how is the work done?

Author and editorial responsibility: Ing. Wolfgang Ruthner, a private project on ruthner.at (Imprint, contact form). The page is not a scientific institution and does not replace a specialist publication: it evaluates published data and presents studies.

Aim: A data page, not an opinion page. It shows measurements, reconstructions, model calculations and studies, each with its source, and where uncertainties exist. The statements are not personal opinions of the author; differing positions are named and linked. Results stay on the page even when they do not fit an expectation.

  • Source and data status: Figures name their source and date; the footer gives the date of the evaluation.
  • Type of statement: Labels separate measurement, reconstruction, model, study and own calculation.
  • Quality grades: In the “Cost studies” sections A means directly from the study, B derived by us and C approximation.
  • Uncertainty and gaps: Bands, ranges and data gaps are shown rather than smoothed over; where a figure could not be checked, this is stated.
  • Data basis: For the model calculations about 400 GB were transferred from the public archive (estimate with date and derivation in the register); only the results are kept, which are available as a small data package.
  • Calculation methods open: Own calculations have expandable explanations with steps and formulas. For the model, quality-check and glacier charts there is a register with sources, downloads and scripts: Data and methods.
  • Media coverage: How climate topics are presented in headlines is checked against the original in a separate case collection, with the verbatim headline, source and assessment: Zwischen Klimawissenschaft und Klimapanik (German-language page).
  • Corrections: Please report errors via the feedback; they are checked and corrected.

As of: 20 Nov 2026.

And on the doorstep? The weather in AmstettenTo the weather page →

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”.

Feedback

Values wrong, an idea or a question? The message goes directly to the author of the page. Name and email address are optional and are only used for a reply.