Clouds Online
DE

Sky now › Method

How we turn weather data into clouds

This page is for anyone who wants the details: which data we use, where it comes from, how often we fetch it, how we choose stations, which rules we use to derive clouds – and where the limits are. Neither weather stations nor weather models deliver cloud species ready-made; nearly every genus, species and probability you see on our pages is our own, traceable derivation.

Description as of 22.11.2026. Whenever we change rules or data sources, we update this page too.

Overview: three building blocks

All weather features of Clouds Online rest on the same three building blocks:

  1. Observations at weather stations – either by a person who identified the cloud using the WMO code (✓ “observed”), or by an instrument that only measures the cover and height of cloud layers.
  2. Weather models – numerical models of the atmosphere that give cover, temperature, humidity and wind at various heights for any place, for the future (forecast) or in hindsight (archive, reanalysis).
  3. Rules from cloud physics that turn these numbers into a genus, species or feature – e.g. “closed layer below 500 m without thermals = stratus”. The rules are our own rules of thumb, based on the WMO definitions and textbooks, and are listed in full below.

Principle: observation before model. What a person at a nearby station has seen is the most reliable information; an instrument only knows height and cover, so we estimate the genus from that; the weather model fills gaps and shows the next hours. That is why the observation comes first on “Sky now”, and the stations come before the model on the photo page.

  • Sky now: observation from the nearest stations (“now”) and forecast for the next hours.
  • Photo page, “The weather on this day”: archive data for the day of the photo – large-scale weather pattern, nearby stations, measurements and weather model or reanalysis.

Data sources

SourceWhatUsed forFrequency and resolutionLicence
DWD, current station reports (SYNOP)
opendata.dwd.de
Staffed stations: genera by WMO code CL/CM/CH. Automatic stations: cover (oktas) and base per layer (ceilometer), total cover, weather, temperature, dew point. Sky now (Germany) Observers report genera on the hour, automatic stations every 10 minutes. We fetch all new files every half hour (minute 12 and 42) and discard reports older than 3 hours. CC BY 4.0
NOAA, airport reports (METAR)
aviationweather.gov
Cover as FEW/SCT/BKN/OVC (we count 2/4/6/8 oktas) with base; the only genera are CB and TCU (towering cumulus); weather, temperature, dew point. Sky now outside Germany (in Germany only if no DWD station is nearby) About 5,000 airports worldwide, mostly every half hour or hour; a collective file we fetch every half hour. US agency; NOAA states no separate licence for other countries’ reports
DWD ICON weather model
via Open-Meteo
Hourly cover, temperature, humidity, wind and height on 12 pressure levels (1000–100 hPa, ground to about 16 km), plus low/middle/high cover, precipitation, weather code, CAPE, freezing level, temperature and dew point at 2 m, wind and gusts. Forecast on “Sky now” The finest model covering the place: ICON-D2 (2 km, central Europe), ICON-EU (7 km, Europe) or global ICON (13 km). 3 days ahead; grid cell 0.1°; fetched again as soon as a new model run is available, at the latest after 3 hours. CC BY 4.0
DWD climate archive, hourly values
opendata.dwd.de
Per station: cloud layers with genus, height and cover; temperature, dew point, humidity, pressure; wind; precipitation. Weather on the day (Germany) Hourly, over decades; genera only as long as an observer reported at the station (mostly from about 1990 until automation). CC BY 4.0
NOAA, Integrated Surface Database
ncei.noaa.gov
Hourly reports worldwide: SYNOP stations with genera (the dominant one per étage), airports with cover and height, measurements. Weather on the day (outside Germany) Hourly to three-hourly; the archive ends in August 2025. US data free; other countries under WMO Resolution 40 non-commercial only
ICON, archived forecasts
via Open-Meteo
The same fields as the forecast (without visibility). Weather on the day, photos from 15 Sep 2023 Hourly; pressure levels in the archive only from autumn 2023. CC BY 4.0
ERA5 reanalysis (Copernicus/ECMWF)
via Open-Meteo
Low/middle/high cover, precipitation, snow, temperature and dew point at 2 m, boundary-layer height, wind and gusts – no vertical levels. Weather on the day, older photos (from 1940) Hourly, grid of about 25 km; a day becomes available after about 6 days. CC BY 4.0
DWD, catalogue of weather types GWL-REA
opendata.dwd.de
Hess/Brezowsky weather type for central Europe, computed from ERA5. Weather on the day (central Europe) Daily from 1940. CC BY 4.0
GeoNames
geonames.org
Places with coordinates (all places with at least 1,000 inhabitants worldwide; in Germany, Austria and Switzerland all inhabited places). Place search; locating where a photo was taken – CC BY 4.0

Why DWD in Germany and other data abroad?

The German Weather Service publishes its station reports openly and free of charge – both current and in the climate archive, with genera identified by observers and ceilometer measurements every 10 minutes. Not every country offers something comparable. Outside Germany we therefore use the worldwide uniform reports collected by NOAA: currently the airport reports (METAR), which contain little more than cover and height, and in the archive the Integrated Surface Database, which also contains SYNOP stations with genera. In Germany we only fall back on airport reports if no DWD station nearby reports clouds – the DWD network is denser and usually more detailed.

Sky now: observation

Observer and instrument

At staffed stations – weather offices and air bases, in Germany only a handful remain – observers identify the clouds on the hour using the WMO code for low (CL), middle (CM) and high clouds (CH). Each code names a genus, often also a species or feature (CL 3 = cumulonimbus calvus, CM 7 = altocumulus duplicatus/opacus …). We take these as “observed” ✓. Where the code says “or” or “sometimes”, the feature only counts as possible. A code names only the dominant cloud per étage; CL 8 (cumulus and stratocumulus at different levels) yields both, CH 5/6 cirrostratus plus cirrus as a possibility. Where a figure names two genera or species, the rest of the report decides: CM 2 (altostratus opacus or nimbostratus) with rain or snow gives nimbostratus, otherwise altostratus; CL 2 (cumulus mediocris or congestus) with showers gives congestus, otherwise the species stays open; CM 6 (altocumulus cumulogenitus or cumulonimbogenitus) with cumulonimbus or thunder reported gives cumulonimbogenitus, otherwise more likely cumulogenitus.

Automatic stations use a ceilometer (a laser pointing straight up) to measure only how much cloud lies at which height above the station – usually only up to about 7 km, and it often misses high ice clouds. We estimate the genus from that (étage by base height: below 2 km low, 2–6 km middle, above high):

  • low, at least 5/8 and below 500 m: Stratus ●●● likely
  • low, scattered shreds below 500 m with precipitation: Stratus fractus (pannus)
  • low, closed with rain: underside of a Nimbostratus
  • showers or thunder reported: Cumulus congestus or Cumulonimbus
  • low above 500 m, otherwise: Cumulus or Stratocumulus – what decides is the sun’s elevation (the driver of thermals) and the condensation level (dew-point spread × 125 m, only with readings from the same station). Base at most 200 m above the condensation level with the sun above 10° → Cumulus (from 7/8 Stratocumulus before Cumulus); more than 500 m above → Stratocumulus. Without readings: up to 6/8 and sun from 30° → Cumulus before Stratocumulus; up to 4/8 and sun from 15° → both nearly equal; from 7/8, with a low sun or at night → Stratocumulus before Cumulus.
  • middle, at least 5/8: Altostratus or Altocumulus; broken: Altocumulus
  • high: Cirrus; at 7/8 Cirrus before Cirrostratus; only at 8/8 Cirrostratus – observers report cirrus in most cases even with almost closed high cloud; the instrument cannot tell fibres from a veil
  • sky obscured (9/8) or fog reported: fog; drizzle: Stratus

The same applies to airport reports; there the report names only CB and TCU as genera. Without “AUTO” in the report a person observed, so we show “observer”.

If a low layer covers 85 % or more of the sky, we show the étage above as “hidden” – nothing can be seen there from below, even if clouds may be present.

Which station? Distance, age, several stations

Search radius

In Germany we look for DWD stations up to 60 km, staffed stations with genera up to 80 km. If none is found, we take airports up to 90 km. At most the four best stations are listed (“Change station”).

Order: distance plus age

The first station is the nearest one with cloud information – we do not prefer staffed stations, because a nearby station usually describes your sky better than a distant one with an observer. Older reports count as further away: for each hour behind the newest report nearby we add 8 km to the distance. Among stations at almost the same distance the more recent one wins, but a very close station with a slightly older report stays ahead. Stations without any cloud information go last (a total cover of 0 % under a clear sky counts as information).

Several stations combined

Cover, weather and estimates come from the first station. On top we add the observed genera of neighbouring stations: stations up to 25 km away, or at most 10 km further than the first, whose report is at most 1.5 hours older than the newest. Why? Staffed stations are rare; the nearest station often has only an instrument. If an observer nearby reports a genus, that is worth more than our estimate.

  • If the first station estimates the same genus, it becomes “observed” ✓ (with the neighbouring station’s name on the chip).
  • If an observer reports a different genus in an étage where the first station only estimates, the remaining estimates there are multiplied by 0.7 – “possible” usually drops out.
  • If the first station has no information at all for an étage (the instrument does not reach that high), we take the observer’s “no clouds” or “hidden”.
  • If you choose a station by hand, only that one is used.

Distance as a traffic light

up to 20 km: the clouds over the station are usually the same as where you are.
up to 50 km: same weather pattern, individual clouds and showers may differ. Estimates from this station × 0.85.
further: only a rough guide; the forecast for your location is then often closer. Estimates × 0.7.

We do not weaken observed genera – what was seen stays ✓ – but from yellow on the chip gets a yellow or red frame with the distance. The limits are empirical: cloud fields of one weather pattern usually span dozens of kilometres, individual heap clouds and showers do not.

Time

Depending on the station, reports are 10 minutes to over an hour old; time and “… minutes ago” are shown for each station, with a warning from 2 hours. Without a running fetch job the page catches up itself if the last fetch is more than 65 minutes ago.

Sky now: forecast from the weather model

A weather model predicts how much cloud there is at which height – not which genus. So we read the vertical profile: adjacent pressure levels with at least 30 % cover are combined into one cloud layer (base and top halfway to the next level). For each layer we then know height, thickness, temperature at base and top, and wind and humidity above.

LayerConditionResult
Deep
base below 2.5 km, top above 6 km or summit ≤ −20 °C
thunder, or summit ≤ −20 °C (glaciated) with showers or CAPE ≥ 300 J/kgCumulonimbus; summit ≤ −40 °C capillatus, otherwise calvus, ≤ −45 °C incus – only as “possible”, since temperature makes glaciation and an anvil more likely but the data cannot show the shape itself; gusts ≥ 60 km/h arcus
showers or unstable (CAPE ≥ 100 J/kg), summit not yet glaciatedCumulus congestus
stable, continuous precipitation (≥ 0.3 mm/h) or without precipitationNimbostratus + pannus or Altostratus
High
base ≥ 5 km and ≤ −25 °C
cover ≥ 90 %Cirrostratus nebulosus
lessCirrus: layer ≥ 3 km thick spissatus; wind shear ≥ 3 m/s per km uncinus; otherwise fibratus; jet stream ≥ 150 km/h radiatus. From 70 % also Cirrostratus fibratus as a second option (model data cannot tell it from dense cirrus)
Middle
base ≥ 2 km
cover ≥ 60 % and ≥ 1.8 km thickAltostratus; no precipitation at the ground and dry air below (a pressure level between cloud base and ground with ≤ 60 % relative humidity) virga
thinnerAltocumulus stratiformis; lapse rate 3–5.5 km ≥ 7 °C per km castellanus (+ floccus); wind ≥ 50 km/h in stable layering over terrain from 400 m lenticularis; two separate layers duplicatus
Low daytime thermals: condensation level below the top of the mixed layerCumulus (less than 1 km tall humilis, otherwise mediocris, unstable and over 2.5 km congestus); from 60 % cover Stratocumulus cumulogenitus
no thermals, ≥ 60 % and base below 500 mStratus (wind ≥ 25 km/h fractus, otherwise nebulosus)
otherwiseStratocumulus (unstable above the layer castellanus, otherwise stratiformis)

Further rules

  • Transparency: layer less than 700 m thick translucidus, otherwise opacus; cover 40–85 % perlucidus.
  • Waves: wind shear at the top ≥ 4 m/s per km in stable layering undulatus; Richardson number below 0.25 (shear beats stability) fluctus.
  • Fog: weather code 45/48 or visibility below 1 km; without visibility (ICON no longer provides it from December 2026) instead dew-point spread ≤ 0.3 K, low cloud ≥ 95 % and no precipitation. Drizzle → Stratus nebulosus.
  • Contrails (simplified Schmidt–Appleman criterion): at 250 or 300 hPa (cruising level) ≤ −40 °C and humidity over ice ≥ 100 % → Cirrus homogenitus; between 70 and 100 % trails form but dissolve quickly.
  • Thin clouds: if the model reports at least 20 % cover in an étage but no pressure level with 30 %, the cloud lies between the levels and is thin – typically Altocumulus, low Stratocumulus, high Cirrus.
  • Thunder, shower and precipitation codes without a matching layer in the coarse profile still yield Cumulonimbus, Cumulus congestus or Nimbostratus.
  • Hidden: middle étage with low cover ≥ 85 %, high étage with low or middle cover ≥ 85 %.

Condensation level: rule of thumb 125 m per degree of difference between temperature and dew point at the ground. Saturation over ice by the Magnus formula. Below every forecast hour, “What we derive this from” shows the values used, and every card shows the steps under “Why?”.

Probability levels

Each rule assigns a value between 0 and 1 for how clearly the data point to this cloud. These become levels:

  • ●●● likely from 0.70
  • ●●○ quite possible from 0.45
  • ●○○ possible from 0.25
  • below that “unlikely” – we never show it.

Observed clouds ✓ always come first and carry no level. If the same genus comes from several rules, the highest value counts. The level is an expert estimate, not a measured hit rate – but we check how well it works against our photos (see Validation).

The weather on this day (photo page)

For photos dated to the day with a known location we fetch that day’s weather data from archives. Everything is fetched once in advance and stored – nothing is loaded when the page is viewed, and no visitor data goes anywhere.

Location

The location is free text with the photo (“Gießen, Hessen”, “Ayaş, Ankara”). We look it up in the place list (GeoNames); countries and regions given must match. Only an unambiguous hit on the first place named counts as certain; uncertain and unfound places are checked by hand. Landscapes and regions without a place get no weather data.

Large-scale weather pattern

The Hess/Brezowsky weather type (30 types, e.g. “westerly, cyclonic”) describes the flow over central Europe on a given day. We use the DWD catalogue GWL-REA, which derives it uniformly from ERA5, and show it only for photos between about 46.5° and 56° north, 3° and 19.5° east – south of the main Alpine ridge the weather is often quite different under the same pattern. The texts and typical genera per type are ours.

Nearby stations

In Germany we read the 5 nearest stations with cloud observations within 80 km and the nearest station for measurements (temperature, dew point, pressure, wind, precipitation) within 40 km. Outside Germany we query the 6 nearest NOAA stations within 80 km; there observers report only the dominant cloud per étage under the SYNOP code, and airports only cover and height.

The sky per hour (6, 9, 12, 15, 18, 21 h and the hour of the photo, local time) is built exactly as on “Sky now”: reported genera = ✓, height and cover only → the same estimation rules, then the stations are combined. As there is rarely anything closer, we include all stations within 80 km; the traffic light shows the distance. We compute the condensation level for heap clouds only from temperature and dew point of the same station.

Weather model: ICON from autumn 2023, ERA5 before

For photos from 15 Sep 2023 we take the archived ICON forecast for the day and place of the photo and read it with exactly the same rules as the forecast. This archive has vertical levels only from autumn 2023.

For older photos (from 1940) we use the ERA5 reanalysis: a uniform after-the-fact reconstruction of the weather from all observations of the time, freely available and globally complete. It is the only dataset reaching back that far. At Open-Meteo, however, it contains only surface values and cover per étage (grid of about 25 km) – no vertical profile, no CAPE, no thunder or shower reports. So separate, more cautious rules apply:

  • Thermals = condensation level below the boundary-layer height in daytime (ERA5 provides boundary-layer height, ICON does not).
  • Thermals + precipitation → Cumulus congestus; from 1 mm/h at 12 °C or more also Cumulonimbus “quite possible”.
  • Precipitation without thermals from a middle and low deck → Nimbostratus.
  • Low: thermals → Cumulus (below 60 %) or Stratocumulus; no thermals and very moist → Stratus, otherwise Stratocumulus; thermals without significant low cloud → Cumulus humilis “possible” (single heap clouds are smaller than a grid cell).
  • Middle: Altocumulus or Altostratus (hard to tell apart without a profile); high: Cirrus, from 70 % also Cirrostratus, from 90 % Cirrostratus first.
  • “Likely” ●●● never occurs, species only roughly, Cirrocumulus never.

Time window and what is public

We compare ±2 hours around the time of the photo, or the day from 5 to 22 h if no time is known. Publicly we only show matches (“also reported at weather stations”, “fits the measurements at weather stations”, “fits the weather model”). Mismatches are only visible to us in the admin area, as a prompt to check classification, date or location – a station 40 km away does not see the same cloud as the camera, so a missing match is not an error in the photo. We make no statement if no station reported genera, if high clouds would have been hidden under a deck of 7/8 or more, for Cumulonimbus when heap clouds are reported (storm towers are often visible from far away), and for Cirrocumulus when high clouds are reported (rarely reported separately).

“Does the weather fit the cloud?”

For each genus on the photo we combine three sources from the time window – observation before model: (1) a station report of that genus by an observer (✓), weighted by the station’s distance 1 (up to 20 km), 0.85 (up to 50 km), otherwise 0.7; (2) the estimate from height and cover at a station with an instrument (level as on “Sky now”, already weakened by distance there); (3) the strongest suggestion from ICON or ERA5 (level). Each source loses 0.05 per hour away from the time of the photo. The source with the highest weight comes first, the others are listed under “Why?”. So a nearby observation beats any model; a report more than 50 km and several hours away can rank behind a “likely” from ICON. Until November 2026 only the model counted here, with the observations listed separately below.

Measurements: notes only for marked changes (pressure ±1.5 hPa in the 3 hours before the photo, temperature drop ≥ 4 °C in 2 hours); the rule of thumb for the base of heap clouds (125 m per degree) only for cumulus and cumulonimbus.

How well do the rules work?

We test the rules against our own photos: the classification of a photo is taken as correct, and we check whether the rules would have expected the same genus from the weather data of that day. Secondly, we take hourly reports from staffed stations, drop the reported genus and estimate it from height and cover alone – showing how good the estimation rules for automatic stations are. Results on the photo archive (as of 22.11.2026):

  • Estimation rules versus observers (about 6,800 station hours): Altocumulus first choice in 98 % of cases, Stratocumulus in 66 %, Cumulus in 63 % (before the sun-elevation rule only 12 %, mostly second choice behind Stratocumulus); Cirrus in 99 % (before 49 %, the other half estimated as Cirrostratus); Cumulonimbus and Nimbostratus never recognised without a weather report.
  • The other way round – when the estimate names a genus, it is right this often (only cases where the observer could see that étage): Cirrus 87 %, Altocumulus 78 %, Stratus 65 %, Cumulus 62 %, Stratocumulus 59 %, Cirrostratus 33 % (before 20 %: a high, dense ice layer seen by the instrument is usually cirrus after all – so it now comes first only at 8/8). When the estimate names Cumulus or Stratocumulus first, that is right in 82 % and 78 % of cases respectively (before: 76 % and 58 %).
  • ERA5 reanalysis versus older photos: hits (at least “possible”) and, in brackets, the share of all photos for which the rules name the genus at all. Only the gap shows whether a rule picks out the right days: altostratus 62 % (26 %), cumulonimbus 43 % (24 %, often distant storms), stratus 36 % (20 %), nimbostratus 29 % (12 %), stratocumulus 56 % (45 %), cumulus 75 % (67 %). Cirrus 82 % (79 %), cirrostratus 73 % (72 %) and altocumulus 70 % (76 %) are named almost always – their high rate says little. Cirrus and cirrostratus cannot be separated from cover alone; the more cautious rule (cirrostratus first only from 90 %) does not change that. Cirrocumulus 0 %.

Limits – what we cannot do

  • A station sees its own sky; a model computes for a grid cell of 2 to 25 km. None of these sources can pin down individual clouds, showers, lenticular clouds over a mountain or a hole in the cloud deck for your exact location.
  • At Open-Meteo the ICON model provides no boundary-layer height. The thermals rule for fair-weather cumulus therefore does not apply in the forecast; flat heap clouds usually appear there as stratocumulus. For cumulus, the observations are far more reliable.
  • Cirrocumulus, features such as mamma or tuba and most varieties can only be hinted at from model data; whatever stays “unlikely” under our rules is not shown.
  • Ceilometers often miss high ice clouds; “no data” in the high étage therefore does not mean “no clouds”.
  • Observers report only the dominant cloud per étage. In the archive, genera mostly exist only from about 1990 and only until a station was automated; the NOAA archive ends in August 2025.
  • All thresholds are rules of thumb. They are based on textbook knowledge and our own testing, but are not statistically calibrated.

By term: formation and derivation

For every genus, species, variety and feature: how it forms, how we recognise it in the data and how well that works.

The derivation is given per data source, exactly as the rules apply it: weather model = ICON (“Sky now” and photos from September 2023), reanalysis = ERA5 (older photos, surface values only), station = automatic station with height and cover per layer, observer = the figure in the WMO code (CL/CM/CH) with which a staffed station reports the term itself – it then counts as observed. If a source is missing, it does not derive the term.

Genera

TermFormationHow we recognise itDerivable
CirrusIce cloud at great height (usually above 6 km, below about −35 °C). Forms when moist air is lifted high up, or as left-over ice from convective clouds and contrails; falling crystals make the fibres.

Weather model: High layer (base from 5 km, below −25 °C), under 90 % cover; from 70 % cirrostratus alongside.

Reanalysis: High cloud cover from 20 % and under 90 %; from 70 % cirrostratus alongside.

Station: Layer above 6 km up to 6/8; at 7/8 narrowly ahead of cirrostratus.

Observer: CH 1–4; with CH 5/6 as a possibility.

fair
CirrocumulusThin high layer with tiny convective cells, often supercooled droplets that freeze quickly. Forms through weak instability or shear waves high up – rare and short-lived.

Weather model: Thin high layer (under 1 km) warmer than −40 °C – but it stays below the “possible” level and is therefore never shown.

Observer: High-cloud code CH 9.

cannot be derived
CirrostratusIce veil, usually formed by slow large-scale lifting ahead of a warm front; as it thickens and lowers it turns into altostratus.

Weather model: High ice layer from 90 % first; at 70–90 % as a second option next to cirrus, at 50–70 % only “possible” – fibres and veil cannot be told apart in the data.

Reanalysis: From 90 % high cover first; at 70–90 % next to cirrus, at 45–70 % only “possible”.

Station: 8/8 above 6 km; at 7/8 as a second option next to cirrus.

Observer: CH 5–8.

fair
AltocumulusLayer at middle height (2–6 km), usually a few hundred metres thick. Radiative cooling at the top and mixing create small overturning cells – hence clumps, rolls and waves.

Weather model: Layer with base from 2 km (above 5 km only if warmer than −25 °C), thinner than 1.8 km or under 60 % cover; also middle cloud that fits between the model levels. Next to thick layers only “possible”.

Reanalysis: Middle cloud cover from 20 % and under 80 %; from 80 % without precipitation level with altostratus.

Station: Layer at 2–6 km up to 4/8; from 5/8 next to altostratus.

Observer: CM 3–9.

fair
AltostratusThick, extensive layer at middle height formed by large-scale lifting (warm front, low pressure system); often followed by nimbostratus with continuous precipitation.

Weather model: Layer with base at 2–5 km, thicker than 1.8 km and from 60 % cover – or a layer from below 2.5 km to above 6 km without precipitation or instability.

Reanalysis: Middle cover from 80 % with precipitation or at least 50 % low cloud beneath.

Station: Layer at 2–6 km from 5/8 (at 8/8 ahead of altocumulus).

Observer: CM 1 (translucidus), CM 2 without precipitation (opacus).

fair
NimbostratusDeep cloud mass from the low into the middle or high étage, formed by sustained large-scale lifting; brings continuous rain or snow.

Weather model: Continuous layer from below 2.5 km to above 6 km (or below −20 °C) with continuous precipitation or at least 0.3 mm/h, without showers or instability; otherwise when continuous rain or snow is forecast.

Reanalysis: At least 0.2 mm/h without thermals from at least 50 % middle cloud, plus 40 % low cloud or 0.5 mm/h.

Station: Rain from a closed low layer (from 5/8) or rain/snow with at least 80 % total cover.

Observer: CM 2 with rain, snow or drizzle reported.

good
StratocumulusLow layer of clumps or rolls beneath a stable layer (inversion). Forms through radiative cooling and mixing at the cloud top, or when heap clouds spread out under the inversion.

Weather model: Low layer (base below 2 km) that does not count as stratus; also low cloud from 20 % without a distinct layer. The rule “heap clouds spread out” does not apply in the model (no boundary-layer height).

Reanalysis: Low cloud from 20 % without thermals; with thermals from 60 % (spread-out heap clouds), from 40 % alongside.

Station: Low layer above 500 m: from 7/8, at night or with a low sun (below 15°), or with its base more than 500 m above the condensation level.

Observer: CL 4, 5, 8.

good
StratusVery low, uniform layer formed by cooling of moist air near the ground (radiation, moist air advection) or when fog lifts.

Weather model: From 60 % low cloud with base below 500 m or condensation level below 300 m; drizzle; lifting fog; shreds beneath nimbostratus.

Reanalysis: From 60 % low cloud without thermals and condensation level below 300 m; lifting fog; shreds beneath nimbostratus.

Station: From 5/8 below 500 m; single shreds below 500 m with precipitation; drizzle.

Observer: CL 6, 7.

good
CumulusHeap cloud: warm air bubbles (thermals) rise, cool and condense at the condensation level – hence the flat base. How tall they grow depends on the instability above.

Weather model: Only towering heap clouds: showers or unstable air (CAPE) in a layer from below 2.5 km to above 6 km whose summit has not yet glaciated. Flat fair-weather cumulus are missing because ICON at Open-Meteo provides no boundary-layer height.

Reanalysis: In daytime the condensation level (dew-point spread × 125 m) below the boundary-layer height: up to 60 % low cloud cumulus, without much low cloud only “possible” (single heap clouds are smaller than the grid).

Station: Low layer above 500 m up to 6/8 with the sun at least 30° high (from 15° up to 4/8) or the base close to the condensation level; showers reported.

Observer: CL 1, 2, 8.

fair
CumulonimbusThundercloud: a heap cloud grows to great heights in very unstable air; its summit glaciates (fibrous) and often spreads out as an anvil. Lightning, thunder, showers, hail.

Weather model: Layer from below 2.5 km to below −20 °C with showers or CAPE of 300 J/kg or more; or thunderstorms forecast.

Reanalysis: Thermals with at least 1 mm precipitation in the hour at 12 °C or more – only “quite possible”, the reanalysis knows no thunderstorms.

Station: Thunderstorm reported or CB in the airport report.

Observer: CL 3, 9.

good

Species

TermFormationHow we recognise itDerivable
fibratusStraight or slightly curved fibres without hooks: ice crystals falling in a uniform wind.

Weather model: Default for cirrus without wind shear and for cirrostratus under 90 %.

Observer: CH 1.

rough
uncinusHooked fibres: falling ice crystals enter air moving at a different speed (wind shear) and are bent.

Weather model: Cirrus when the wind increases by at least 3 m/s per km in the high layer.

Observer: CH 1, 4.

rough
spissatusDense cirrus, grey against the sun – often the remains of a thunderstorm anvil.

Weather model: High layer at least 3 km thick; dense cirrus next to thunderstorms.

Reanalysis: Dense cirrus next to possible cumulonimbus.

Observer: CH 2, 3.

rough
castellanusTurrets on a common base: the air at that height is unstable, so saturated parts rise on their own.

Weather model: Altocumulus: lapse rate between 3 and 5.5 km of 7 °C per km or more (steeper than moist adiabatic, ~6 °C/km); stratocumulus: the same above the layer.

Observer: CM 8.

rough
floccusSeparate tufts with ragged bases: turrets whose base has dissolved, often with fallstreaks.

Weather model: Only as a possibility next to altocumulus castellanus – we do not check the humidity beneath the layer.

Observer: CM 8.

rough
stratiformisExtensive sheet or layer: the normal case for Ac, Sc and Cc.

Weather model: Default for altocumulus and stratocumulus without instability.

Reanalysis: Stratocumulus without thermals.

Observer: CM 3.

fair
nebulosusFeatureless veil or layer – with stratus and cirrostratus.

Weather model: Stratus with wind below 25 km/h, after fog or with drizzle; cirrostratus from 90 %.

Reanalysis: Stratus with wind below 25 km/h and after fog.

Observer: CL 6 (as a possibility next to fractus).

fair
lenticularisLenses or almonds that stay in place: air flowing over mountains oscillates in stable layering (lee waves) and condenses in the wave crests.

Weather model: Altocumulus with wind of 50 km/h or more at its base, stable layering (below 5.5 °C per km), terrain from 400 m – only as a rare hint, without mountain data.

Observer: CM 4.

cannot be derived
fractusRagged shreds: form in gusty wind or beneath rain clouds in moist air (then pannus).

Weather model: Stratus with wind of 25 km/h or more; shreds beneath nimbostratus.

Reanalysis: Stratus with wind of 25 km/h or more; shreds beneath nimbostratus.

Station: Single shreds below 500 m with precipitation, or beneath rain from a closed deck.

Observer: CL 7; CL 6 as a possibility.

rough
humilisFlat fair-weather cumulus: a stable layer just above the condensation level stops growth.

Reanalysis: Cumulus whose depth (boundary-layer height minus condensation level) stays below 600 m; single heap clouds without much low cloud.

Observer: CL 1.

rough
mediocrisModerately tall cumulus with small protuberances: moderate instability.

Reanalysis: Cumulus at least 600 m deep without precipitation.

Observer: CL 2 without showers – as one of two possibilities next to congestus.

rough
congestusTowering cumulus (cauliflower): strong instability, but the summit has not yet glaciated.

Weather model: Towering heap cloud (from below 2.5 km to above 6 km) with showers or instability whose summit has not yet glaciated; showers forecast.

Reanalysis: Thermals with at least 0.1 mm precipitation in the hour.

Station: Showers reported.

Observer: CL 2 with showers; without showers as a possibility next to mediocris.

fair
calvusCumulonimbus with a still smooth, just glaciating summit.

Weather model: Only as a possibility: cumulonimbus with summit between −20 and −40 °C – the data cannot show the smooth shape.

Observer: CL 3.

rough
capillatusCumulonimbus with a fibrous, fully glaciated summit, often an anvil.

Weather model: Only as a possibility: summit colder than −40 °C – it usually glaciates there; the data cannot show whether it looks fibrous.

Observer: CL 9.

rough
volutusA single horizontal roll, e.g. at gust fronts.Cannot be derived from any of our data sources.cannot be derived

Varieties

TermFormationHow we recognise itDerivable
translucidusThin enough for the sun or moon to show through.

Weather model: Layer thinner than 0.7 km (Ac, Sc) or 2.5 km without precipitation (As); middle cloud between the model levels.

Reanalysis: Altocumulus under 50 % cover.

Observer: CM 1, 3, 4; CM 5 as a possibility.

fair
opacusThick enough to hide the sun or moon completely.

Weather model: Thicker layer or precipitation; drizzle.

Reanalysis: Stratocumulus from 70 % cover.

Observer: CM 2 without precipitation; CM 5 and 7 as a possibility.

fair
perlucidusGaps between the elements through which sky or higher cloud can be seen.

Weather model: Ac or Sc layer with 40–85 % cover.

Reanalysis: Stratocumulus under 70 % cover.

rough
undulatusWaves: the wind changes across the layer (shear) in stable layering – the air oscillates.

Weather model: Wind change of 4 m/s per km or more at the layer with stable layering (Richardson number 0.25 or more).

rough
radiatusParallel bands that seem to converge towards the horizon: cloud streets in strong, uniform wind.

Weather model: Cirrus with a jet stream of 150 km/h or more. Cumulus cloud streets are missing (no boundary-layer height).

Reanalysis: Cumulus with wind of 30 km/h or more.

rough
duplicatusTwo or more layers one above the other.

Weather model: Two separate layers in the middle étage.

Observer: CM 7 (as a possibility).

fair
intortusIrregularly entangled cirrus fibres.Cannot be derived from any of our data sources.cannot be derived
vertebratusCirrus like fish bones or ribs.Cannot be derived from any of our data sources.cannot be derived
lacunosusA net of round holes: sinking air between small convective cells.Cannot be derived from any of our data sources.cannot be derived

Supplementary features

TermFormationHow we recognise itDerivable
virgaFallstreaks: precipitation evaporates in dry air beneath the cloud before reaching the ground.

Weather model: Ac or As without precipitation at the ground and a pressure level beneath with at most 60 % relative humidity (without a level beneath: surface dew-point spread of 8 °C or more).

Reanalysis: Altocumulus under 80 % without precipitation, surface dew-point spread of 8 °C or more (no profile).

rough
praecipitatioPrecipitation reaching the ground.

Weather model: Precipitation at the ground forecast.

Reanalysis: Precipitation of 0.1 mm or more in the hour.

Station: Precipitation reported.

good
incusAnvil: the storm tower hits the tropopause and spreads sideways.

Weather model: Only as a possibility: cumulonimbus reaching air colder than −45 °C – the data cannot show whether an anvil really spreads out.

rough
mammaPouches on the underside: cold, ice-laden air sinks out of the anvil into drier air.

Weather model: Only as a rare hint: a possible anvil (below −45 °C) and dry air at the ground (dew-point spread of 6 °C or more).

rough
arcusShelf cloud: the storm’s cold outflow pushes warm air ahead and lifts it.

Weather model: Cumulonimbus with gusts of 60 km/h or more.

rough
tubaFunnel: a rotating updraft, precursor or part of a tornado.

Weather model: Only as a rare hint: cumulonimbus in very unstable air (CAPE of 1,500 J/kg or more) with strong wind shear (20 m/s or more up to 6 km).

cannot be derived
fluctusBreaking waves (Kelvin-Helmholtz): at the boundary of two layers with very different winds, waves curl over.

Weather model: Richardson number at the layer top below 0.25 (shear beats stability).

rough
asperitasWavy, chaotic underside – cause not yet fully understood.Cannot be derived from any of our data sources.cannot be derived
cavumA hole in a layer of supercooled droplets: droplets freeze (often triggered by aircraft) and fall out.

Weather model: Only as a rare hint: a thin altocumulus layer (under 0.7 km) between −10 and −25 °C.

cannot be derived
murusWall cloud beneath a rotating thunderstorm (supercell).Cannot be derived from any of our data sources.cannot be derived
caudaTail cloud attached to a supercell’s wall cloud.Cannot be derived from any of our data sources.cannot be derived

Accessory clouds

TermFormationHow we recognise itDerivable
pileusA cap above a fast-growing tower: it lifts a moist, stable layer above it.Cannot be derived from any of our data sources.cannot be derived
velumA veil through which towers grow: a lifted moist layer.Cannot be derived from any of our data sources.cannot be derived
pannusDark shreds beneath rain clouds: the air is saturated by evaporating precipitation.

Weather model: Shreds beneath nimbostratus.

Reanalysis: Shreds beneath nimbostratus.

Station: Shreds below 500 m with precipitation, or beneath rain from a closed deck.

Observer: CL 7.

fair
flumenInflow bands of a supercell.Cannot be derived from any of our data sources.cannot be derived

Mother clouds

TermFormationHow we recognise itDerivable
cumulogenitusA layer that formed from spreading heap clouds.

Reanalysis: Thermals and at least 60 % low cloud: the deck forms from spread-out heap clouds.

Observer: CL 4; CM 6 without cumulonimbus reported.

rough
cumulonimbogenitusA cloud that formed from a cumulonimbus, such as dense cirrus from the anvil.

Weather model: Dense cirrus next to a cumulonimbus.

Reanalysis: Dense cirrus next to a possible cumulonimbus.

Observer: CH 3; CM 6 with cumulonimbus or thunder reported.

rough
homogenitusMan-made, mainly contrails: engine water vapour freezes in very cold air; if the air is supersaturated with respect to ice, they spread into cirrus.

Weather model: At 250 or 300 hPa below −40 °C and saturated over ice (100 % or more): persistent contrails (cirrus homogenitus); from 70 % only a note on short-lived ones.

fair

Phenomena

TermFormationHow we recognise itDerivable
FogCloud at ground level: cooling (radiation on clear nights) or moist air saturates the air near the ground.

Weather model: Visibility below 1 km or weather code fog; without visibility a dew-point spread of at most 0.3 K, at least 95 % low cloud and no precipitation.

Reanalysis: Dew-point spread of at most 0.3 K, at least 95 % low cloud, under 0.1 mm precipitation.

Station: Fog reported or sky obscured (9/8).

fair

References

Definitions from the WMO International Cloud Atlas (2017) and the WMO cloud code (CL/CM/CH); formation from meteorological textbooks (e.g. Houze: Cloud Dynamics; Wallace & Hobbs: Atmospheric Science; Kraus: Die Atmosphäre der Erde); weather types after Hess/Brezowsky (DWD catalogue); contrails after Schmidt/Appleman. Data: German Weather Service (CC BY 4.0), Open-Meteo (CC BY 4.0), Copernicus/ECMWF ERA5 (CC BY 4.0), NOAA, GeoNames (CC BY 4.0).