Europe, the Alpine region and glaciers
What shows up close to home? Europe, the Alpine region and Austria in a comparison of measurement and models, maps of the temperature trend and nine glaciers in Austria.
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.
For specialists: Europe, the Alps and Austria: models against measurements
Europe, the Alps and Austria: how well do the models match the measurements?
The same 23 climate models as above, this time not global but for three regions: Central Europe (45–55° N, 5–20° E), the Alpine region (44–48.5° N, 5.5–16.5° E) and Austria (46.3–49.1° N, 9.5–17.2° E). The measurement is the gridded dataset E-OBS (station data interpolated to a grid). For Austria the official station Amstetten (GeoSphere) is added. The first five runs of each model are used, 97 in total. Temperature and precipitation are shown, for the year and for the seasons.
💬 In plain language
- This is about Europe, the Alps and Austria. The question is: do the computer models match what was measured there?
- The blue line is the measurement, the orange line is the mean of the models, the orange band is the range of the models. For Austria the weather station Amstetten is added.
- In a small region the weather varies a lot from year to year. So the statement is less certain than for the whole Earth.
- For precipitation the models are, on average, somewhat higher than the measurement, and the models scatter widely.
- The charts show how well or badly it fits. They do not say which model is best.
Trend 1979–2020 per model (unit see axis); line = range of the runs, dot = mean of the runs; measurement E-OBS as the vertical line, station Amstetten dashed
All seasons at a glance
Map: temperature trend in the Alpine region and height profile
The same measurement (E-OBS), but not as a regional mean, rather per grid cell (0.25°, about 18 × 28 km): the map shows where and how strongly the Alpine region warmed between 1979 and 2020. Below it shows whether the trend is related to the height of the cell. Clicking a season switches both displays; with the mouse (or by tapping) the location, mean height of the cell and trend appear.
Trend by height of the cell, °C per decade; line = mean of all cells of the height band, band = middle 80 % of the cells (10th to 90th percentile), grey = other seasons, number = cells in the band
Reading notes: (1) Each cell is an average over about 18 × 28 km; the height is the mean height of the cell (E-OBS elevation grid), not that of a summit or valley. (2) In mountains with few stations the E-OBS grid is less certain; values at great height rest on fewer measuring sites. (3) Neighbouring cells are not independent, and almost all high cells lie in the centre of the Alps: the relation with height cannot therefore be cleanly separated from location. (4) The trend per cell needs at least 35 years with values (months with at least 25 days, a season only with all its months). (5) The map shows the measurement only, no models.
Model map: how well do the climate models reproduce the pattern?
The same map from the climate models: for each model the temperature trend 1979–2020 was calculated in every grid cell from its first up to five runs and averaged (97 runs in total, 23 models). On the left the measurement (E-OBS), in the middle the chosen model or the mean of all models, on the right the difference (model minus measurement). All three maps lie on the same coarse grid (0.5°); the models natively have cells of only about 100 to 300 km and are interpolated in between.
Reading notes: (1) The model map is interpolated: a model with 200 km cells has only a few cells in the section, and the gaps in between are smoothed. Fine patterns of the Alps (valleys, elevations) cannot be reproduced by the models. (2) E-OBS is averaged onto the same 0.5° grid (mean of 2 × 2 cells, at least 2 valid) and applies to land only; the difference is formed only there. (3) With at most five runs per model a lot of chance (internal variability) remains in the trend of a single cell; the mean of all models is smoother than any single one. (4) “Pattern” means spatial similarity (correlation across the land cells), not the strength of the trend. (5) No height correction; the measurement is less certain in mountains.
The own weather station compared with the official one
Two series of the own stations can be compared with the official station Amstetten (temperature only, monthly means, months with at least 25 days): the earlier Vantage station (2003 to October 2015, 145 months) is on average +0.08 °C above the official one (scatter of the difference 0.33 °C); the new station (2020 to 2025, 64 months, daily means only) is −0.36 °C below (scatter 0.37 °C). Part of the new station’s difference is due to the averaging method (a mean of 15-minute values lies about 0.3 °C below the mean of daily maximum and minimum). The own measurements are not adjusted. They are not compared with the models here, because the series are short and precipitation and Tmax/Tmin are partly missing.
- The charts show how widely models and measurements scatter in small regions, not which model is “right”.
- The measurement itself is a single series with its own uncertainty, which is not drawn here.
- The station Amstetten is a point, not a region; it sits at a particular altitude and location.
Reading notes: (1) Five runs per model: the range of runs is therefore narrow, and “within the range” is a strict criterion. (2) EC-Earth3: runs that start only in 1970 have no 1961–1990 mean and are missing. (3) Measurement E-OBS is a grid built from stations; in mountains with few stations it is less certain. The uncertainty of the E-OBS series is not included here. (4) Station Amstetten: official monthly values (only months with at least 25 days); reference mean 1961–1990 from 28 of 30 years (1989 and 1990 are missing); trend only with at least 35 years. The station does not lie on the grid and is not adjusted. (5) The series of the own weather station are unchanged (see the paragraph above).
How this was calculated – regions, runs, limits
Data
- Models: CMIP6 via “Pangeo”,
tas(air temperature at 2 m) andpr(precipitation),historicalto 2014,ssp245from 2015; the first five runs of each model (computed from both ends on two computers), 23 models, 97 runs. - Measurement: E-OBS (temperature: daily mean, precipitation: daily total), per region over the land cells in the box, weighted by the cosine of latitude; monthly values, from which annual and seasonal values (winter = December to February).
- Station: GeoSphere Austria, Amstetten, monthly mean (temperature) and monthly total (precipitation), only months with at least 25 days.
Calculation steps
- Mean per region and month, departure from the 1961–1990 mean (temperature in °C, precipitation in percent).
- Seasonal values from the monthly values, annual means from the twelve months.
- Trend 1979–2020 as the linear slope per run, in °C or percent per decade; per model the mean and range of the runs.
- Classification: if the measurement lies within the range of the runs it reads “within”; if the model lies above it, “above”, otherwise “below”. This is a classification, not a grade.
Glaciers in the Alps
Wie haben sich Österreichs Gletscher seit 1850 verändert, und was hat die Sommertemperatur damit zu tun? Neun Gletscher mit langen Messreihen: Pasterze, Hintereisferner, Vernagtferner, Kesselwandferner, Stubacher Sonnblickkees und Goldbergkees (Länge, Massenbilanz und Sommertemperatur) sowie Hallstätter Gletscher, Großer Gosaugletscher und Schladminger Gletscher am Dachstein (nur Länge, weil dort keine passende Station und keine ausreichende Bilanzreihe vorliegt). Weitere Gletscher mit langen Reihen sind nicht enthalten, die Auswahl ist also keine Gesamtbilanz aller Alpengletscher. Es sind Messwerte der Gletscherdienste und der Wetterstationen, keine Modellrechnung für einzelne Gletscher.
Where are the glaciers?
Die Dreiecke sind die neun Gletscher, die grauen Rauten die Wetterstationen, deren Temperaturen unten verwendet werden. Mit dem Mausrad (oder den Knöpfen) lässt sich zoomen, durch Ziehen verschieben. Ein Klick auf einen Gletscher zeigt seine wichtigsten Zahlen und führt zu den Grafiken.
How far has the glacier end retreated?
The lines show by how many metres the end (the tongue) of a glacier has shortened (minus) or lengthened (plus) since the first measurement from 1850 onwards. Each line starts at zero at the end of the first measurement period. The dots are the individual measurements, joined in between.
How fast was the retreat in different periods?
The more direct climate signal: gain and loss of ice
The mass balance states how much ice a glacier gains or loses in a year in net terms, in metres of water equivalent averaged over the whole glacier area. Unlike length, it does not depend on the shape and slope of the glacier and reacts to snow and temperature without delay.
Warm summers, greater ice loss
Each dot is a year: on the left the summer temperature (June to September) of the nearest high-mountain station, above it the annual mass balance of the selected glacier. Dark dots are older years, light ones more recent.
Pasterze, Hintereisferner and Vernagtferner in pictures
Images with a free licence (Wikimedia Commons). Under each image you will find the author, the licence and the link to the original page. Paintings and drawings are not true-to-scale photographs; the photographs do not always show the same viewpoint.
Glaciers are not a given: finds under the ice
Reading notes: (1) The weather stations are not at the glacier but in the same mountains at a different height and distance (Obergurgl 1,941 m, Rudolfshütte 2,317 m, Sonnblick 3,109 m); the glaciers lie higher. (2) A relationship between summer temperature and mass balance is an indication, not a complete proof of the cause: winter snowfall, radiation and dust also play a role and are not in these data. (3) Length depends on shape, slope and debris cover and reacts with a delay; it is not a direct climate signal. (4) Years are missing in the series; where there is no measurement every year, the dots are joined. (5) Climate models (CMIP6) provide temperature and precipitation, not the behaviour of an individual glacier; that would require additional glacier models. There is therefore no forecast for individual glaciers here.
How this was calculated: sources and steps
Sources
- Length and mass balance: WGMS (2026): Fluctuations of Glaciers (FoG) Database, World Glacier Monitoring Service, Zurich, doi:10.5904/wgms-fog-2026-02-10 (version of 10 February 2026, free to use with correct citation). The measurements come from the observers of the individual glaciers.
- Temperature: GeoSphere Austria, monthly means of air temperature (stations Sonnblick, Obergurgl, Rudolfshütte).
Calculation steps
- Length: sum of the measured changes in length from the first measurement period ending after 1850; the sum starts at zero there.
- Retreat per decade: difference of the sums at the start and end of the period, linear between the measurement years, converted to ten years.
- Mass balance: annual total balance of the glacier (hydrological year, m water equivalent); the sum is the sum of these annual values.
- Summer temperature: mean June to September, only years with all four months; relationship as correlation and slope. Additionally without the shared trend: the linear trend is subtracted from both series so that not only the shared course over the years counts.
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”.
