Weather

loading…

Domain

Chart

Model

Shading

Overlay

Animation

Speed

Export

Export animated GIF

Width Frames Frame delay

The colour bar and the valid time are drawn into the image, so the file still says what it is once it leaves this page.

— local

—

F000

2 m temperature (°C)

Hover the map Temp— Wind— Dir— Rain 3 h—

Peak— DWD ICON-EU · © Ryan Vella

Source

Animation

Speed

—

NDVI

Cloud Sea

Point forecast

—

Place

Range

Data sources, projection and keyboard shortcuts

Data. DWD ICON-EU 0.0625°, native over Malta and subsampled to 0.25° over the Central Mediterranean; ItaliaMeteo-Arpae ICON-2I 0.025 × 0.02° (CC BY 4.0), hourly to +72 h, Malta only — its domain stops at 33.7 °N; ECMWF IFS 0.25° open data (CC BY 4.0) on every frame — the four Europe charts, 6-hourly to +144 h and 12-hourly to +360 h beyond it, and the surface fields over the other two, held to ICON-EU's +120 h so that switching model on a frame lands on the same day rather than dragging the timeline back a day. The full fifteen days are only published on the 00 and 12 UTC runs, so the Europe charts prefer those; on a morning when ECMWF is late they fall back to the newest run and stop at +144 h, and the next cycle repairs it. IFS is offered there as a second opinion rather than a better one: at 0.25° it is the coarsest of the three, and over Malta it is a dozen grid points, but it is run independently of ICON, so where the two disagree is a reading of how settled the forecast is. Coastlines are GSHHG full resolution, the same source the rendered Malta charts use. Precipitation is the total over each step interval, differenced from the model's accumulation — six hours on IFS against three on ICON-EU and one on ICON-2I, which is why the same shower shades deeper on the coarser model.
Meteogram. Three models spliced end to end, each carrying the chart as far as it goes: ICON-2I hourly to +72 h, ICON-EU three-hourly to +120 h, then IFS to +360 h. A rule and a label mark every handover and the shading deepens behind each one, because a change of model is not a change in the weather and must not read as one. Forecast shows the first two, which is the chart as it stood before the long range existed; Extended adds the third. Two views rather than one because the axis is linear in time — drawn together, the hourly detail anyone opens this page for would be squeezed into a fifth of the width to make room for nine days that cannot resolve an hour. Both come out of the same download, so the button is instant. The IFS section steps three-hourly to +144 h and six-hourly after it — not a preference but the gust field's own window, which is the maximum over the preceding three hours up to +144 and over the preceding six beyond it; sampling more coarsely than that would report part of an interval as though it were all of it. Past about day ten the temperature is a weak signal and the rain bars are close to noise, so the section is drawn to be read as a pattern rather than as a forecast for an hour, and the summary line above the chart deliberately speaks only for the first five days.
24 h rainfall. The IFS total-precipitation field differenced across a fixed 24-hour window, so each chart is the total for the day ending at the time shown — rain and melted snow, convective and large-scale together. Consecutive frames overlap: six hours apart, two charts share eighteen hours of the same rain. That is what a running daily total is, and it is also what makes an accumulation affordable on a timeline whose cadence changes — a fixed window means the same thing at six-hourly, at twelve-hourly and in the archive's twelve-hourly past, where a since-the-last-frame bucket would not. The loop starts at +24 h, the first hour with a full day behind it. Read the totals as a pattern rather than a measurement: at 0.25° a convective cell is smaller than a grid box, so the model spreads an hour's downpour over a wider and gentler area, and orography smoothed to 25 km under-rains the windward slope it was smoothed off.
Vapour transport. Derived here rather than downloaded: specific humidity and wind on the 1000, 925, 850, 700, 600 and 500 hPa surfaces, integrated over pressure and divided by g. The column is cut off at the model's own surface pressure, so the levels that lie inside the Alps or under the Greenland ice — which the model fills by extrapolation — contribute nothing. Stopping at 500 hPa rather than the conventional 300 leaves out under 5% of a typical transport and saves roughly a third of what is already the heaviest download in this pipeline. Contours are 700 hPa geopotential height, the level the plume is steered at. Mercator, unlike the Lambert used by the rendered PNGs — a conformal projection is what lets the streaklines be advected in screen space without distorting direction.
Observed cloud. The Satellite overlay on the map panels is one scan per past step of whichever timeline is showing — 3-hourly over the last day on the surface frames, 12-hourly over the last week on the synoptic ones. Nothing is drawn over forecast hours. Which imagery depends on how finely the frame is drawn. The close-ups are limited by the instrument, so they take MTG FCI's high-resolution fast imagery: 0.5 km in the visible, grey rather than colour because it is a single channel, falling back to the 10.5 µm window after dark. The wide frames are limited by the size of image they can afford to carry rather than by the instrument, so they take Geo Colour instead — true colour by day blended into infrared and city lights across the terminator, which a window spanning whole days needs. The synoptic frame is two satellites. From 0° the Earth's limb crosses it near 50° W, which would cut off the western half of the Atlantic the chart exists to watch, so GOES-East at 75.2° W is composited in to fill it — each pixel taken from whichever satellite sees it less obliquely, feathered across the join near 38° W. Both are the same Geo Colour product, which is what lets them meet without a seam. MSG is not used: EUMETView publishes no HRV, the SEVIRI channels it does publish are 3 km and visibly blocky at these zooms, and SEVIRI has no green channel, so there is no true-colour product for it at all.
Satellite. EUMETSAT MTG FCI (CC BY 4.0) for the 24-hour loop, 20-minutely at about 1 km. The two stills are Copernicus Sentinel data (ESA): Sentinel-3 OLCI at 300 m, roughly daily, on the same frame as the maps, and Sentinel-2 MSI at 10 m every two to three days, zoomed to the islands because at map width its detail would not survive. All three are observations, not forecasts, and all are resampled into this page's Mercator box server-side.
Satellite. EUMETSAT Meteosat Third Generation FCI Level 1c (CC BY 4.0), the Malta subset of the full disk at 20-minute cadence over the last 24 hours. True colour from the visible channels while the sun is up; after dark the 10.5 µm infrared window, inverted so cold cloud tops are white in both. It is observation, not forecast: the loop ends at the newest scan.
Keyboard: ← → step, space play/pause, Home/End jump — on whichever view is showing. On the synoptic charts, ↑ ↓ move between the four levels — 250 hPa jet, 500 hPa, vapour transport, surface pressure — holding the forecast hour, so the same moment is read at four heights.