Every pilot learns to read a METAR the same way: badly at first, then fast, then — after a few hundred of them — with a kind of grudging fluency. You stop spelling out BKN and start seeing a ceiling. You stop converting Q1011 and start feeling the altimeter.
But fluency is not the same as understanding, and the gap between them is where most weather-related go-arounds live.
Here is an actual observation from London Heathrow, taken at 07:20Z this morning:
METAR EGLL 100720Z AUTO 05007KT 020V080 5000 HZ OVC005 06/06 Q1011
TEMPO 4000 BR BKN004
A competent pilot reads that in four seconds: north-easterly at seven, five kilometres in haze, overcast at five hundred, six over six, 1011. IFR. Fine.
But four seconds of reading does not answer the questions that actually matter:
- Six over six. The temperature and dew point are identical. Is this going to become fog, or burn off?
- Overcast at 500 ft. Is that ceiling rising or falling? The METAR is a snapshot; snapshots don't have direction.
- 05007KT with a variable range of 020–080. Which runway, and how much crosswind on the worst swing?
- And the one nobody can answer from a METAR at all: what does this look like in six hours, when I actually plan to leave?
The raw code contains none of that. It contains the ingredients. The cooking is left to you.
That gap is the reason I built metarandtaf.com — a live aviation weather site that takes the same official data and does the arithmetic, the trending and the drawing that pilots otherwise do in their heads, on a kneeboard, or not at all.
What follows is the same observation, taken apart layer by layer.

Every reporting station on one map, each coloured by flight category, with animated precipitation and SIGMETs over the top.
1. The decode, with the parts that actually matter pulled forward
Here is that Heathrow observation on the site. The raw METAR and TAF stay visible at the top — you should always be able to see the source — but underneath, every field is unpacked.

The raw report stays on top; the decode, the runway diagram and the wind components sit underneath it.
Three things worth noticing, because they are the things a plain text decoder usually skips.
The dew point spread is given its own line, not buried. Six over six is a 0.0 °C spread and 100 % relative humidity. The site says so in words: a narrow spread that favours mist or fog forming. A spread under about 2 °C with light wind and a clear night is the classic radiation-fog setup — there is a whole guide on dew point spread and fog forecasting if you want the physics rather than the rule of thumb.
Visibility is shown in both systems at once. 5 km and 3.1 SM, on the same bar, with the IFR/MVFR/VFR boundaries marked. If you fly in a country that reports in metres and trained in one that reports in statute miles — or the reverse — that conversion is where the mistakes happen.
Ceiling is separated from cloud. OVC005 is a cloud layer. Ceiling 500 ft AGL is an operational fact. Scattered and few layers are not ceilings; broken and overcast are. The site computes the ceiling and shows the layers, which also matters for vertical visibility and obscuration codes.

Pressure in hPa and inHg, visibility in km and SM, ceiling in feet AGL, and the temperature–dew point spread with relative humidity.
2. Which runway — and what happens on the worst swing of a variable wind
05007KT 020V080 is a seven-knot wind. Nobody loses an aeroplane to seven knots. But the method of answering the runway question is the same whether the wind is seven knots or twenty-seven, and the method is where it is worth being precise.
The site computes head, tail and crosswind components for every runway at the field, sorts them, and marks the best one:
| RWY | HDG | Head/Tail | Crosswind | Length | Fit |
|---|---|---|---|---|---|
| 09L | 090° | 5 kt headwind | 4 kt left | 12,799 ft | 100 |
| 09R | 090° | 5 kt headwind | 4 kt left | 12,001 ft | 100 |
| 27R | 270° | 5 kt tailwind | 4 kt right | 12,799 ft | 0 |
| 27L | 270° | 5 kt tailwind | 4 kt right | 12,001 ft | 0 |
The diagram animates the wind arrow against the actual runway geometry, including parallel pairs, and the variable range 020V080 is drawn as a sweep rather than a single line — so you can see what the crosswind becomes at the unhelpful end of the swing, not just at the reported mean.
If you want to do this on paper instead, the crosswind component calculation walks through the trigonometry and the 30°/60° mental shortcuts.
3. The TAF, broken into periods you can actually plan against
A TAF is a paragraph of nested change groups, and that nesting is precisely what makes it hard to read under pressure. Here is Heathrow's, as a table:

Each FM, TEMPO, BECMG and PROB30 group as its own row, with a flight category computed per period.
| Valid | Change | Category | Conditions |
|---|---|---|---|
| 06:00Z – 09:00Z | FROM | IFR | 060° 6 kt, 6 km, broken 500 ft |
| 06:00Z – 09:00Z | TEMPO | LIFR | 4 km, BR, broken 300 ft |
| 06:00Z – 07:00Z | PROB30 TEMPO | LIFR | 0.5 km, FG, broken 100 ft |
| 09:00Z – 18:00Z | BECMG | VFR | 060° 6 kt, 10 km+, few 800 ft |
That PROB30 TEMPO 0500 FG BKN001 is the line that decides the morning. Thirty percent chance of half a kilometre in fog with a hundred-foot ceiling, until 07:00Z — and then a BECMG to 10 km and a few at 800 by 09:00Z. The field is not marginal all day. It is marginal for about ninety more minutes.
You cannot see that shape in the paragraph. You can see it immediately in the table. The full method is in the guide on how to read a TAF forecast.
4. Trend: a METAR has no direction, but five of them do
This is the part that most weather sites skip entirely, and it is the part that changes decisions.

The archive of recent observations, with the direction of travel computed from them.
07:20Z 05007KT 5000 HZ OVC005 06/06 Q1011
06:50Z 06006KT 7000 OVC004 06/06 Q1011
06:20Z 06005KT 7000 OVC004 06/05 Q1011
05:50Z 07006KT 7000 OVC003 06/05 Q1011
05:20Z 06006KT 6000 OVC003 06/05 Q1012
Read down the column rather than across the line, and the morning has a plot: the ceiling has lifted from 300 to 500 ft over two hours, visibility has dropped from 7 km to 5 km as haze thickened, and the pressure is falling — 1013 to 1011 over five hours.
The site states that explicitly — pressure falling 1013 → 1011 hPa — and projects the next observation from the trend: next METAR (estimated): ~6 kt, ~1011 hPa, ~6 °C. Not a forecast model. Just the honest extrapolation of what the last five reports are doing, which is exactly the calculation an experienced pilot does by eye and a tired one skips.
Every observation is archived permanently, so the trend is computed from real history rather than from whatever happens to be cached.
5. Past the TAF: where the next five days are
A TAF covers 24 or 30 hours. Most flying decisions — especially the general-aviation ones, where the question is which day this week rather than which hour today — live well outside that window.

Hour-by-hour to 14 days, with every source model selectable and a computed best flight window.
Three things are happening in that screenshot.
A flyable window is identified, not just plotted. Best flight window: Thu 22:00 – Sat 01:00Z (27 h). The site scores every hour against VFR-ish thresholds and finds the longest run.
The official TAF is overlaid on the model forecast. The TAF band runs along the top, so you can see where the forecaster and the models agree, and — more interestingly — where they don't.
Every model is one click away. ECMWF IFS, ECMWF AIFS, NOAA GFS, DWD ICON, CMC GEM, Météo-France, UK Met Office — and the blended line. When the models disagree, you can see that they disagree, which is far more useful than a single confident number.
And then the rows that matter operationally, which no raw forecast gives you:

Fog risk, ceiling in feet AGL, precipitation type and probability, gusts, turbulence, low-level shear, freezing level, icing and density altitude — hour by hour.
Fog risk flagged as MOD for the next three hours. Ceiling trending 280 → 400 → 640 → 1,100 ft. Gusts to 19 kt mid-afternoon. Low-level shear. Freezing level at 9,300 ft and dropping to 8,000. Icing flagged MOD overnight. Aircraft icing from METAR and TAF covers how to read those together.
6. The vertical slice: what the air looks like on the way up
The meteogram is time across the bottom. The profile below it is time across the bottom and altitude up the side — a cross-section of the atmosphere you are about to climb through, to 40,000 ft.

Cloud layers, precipitation, the 0 °C isotherm and winds aloft in 1,000 ft steps, hour by hour.
Cloud layers drawn where they actually are. The freezing level as a line, not a number. Winds aloft at every level with temperature — 123°/15 at 4,000 and −5 °C, 236°/11 at 10,000 — which is the shear profile, the icing layer and the climb plan in one picture. More on that in winds aloft and multi-model forecasts.
7. And then: look at it
The parts above are instruments. This part is a window.

The field rendered in 3D from the live report: runway layout, cloud base at the reported height, visibility as actual visibility.
The 3D view builds the airport from its real runway geometry and the current observation: overcast at 500 ft is drawn at 500 ft, 5 km in haze looks like 5 km in haze, the wind blows from the reported direction at the reported speed. You can orbit it freely, or take one of the fixed cameras — tower, approach, or top-down.

Tower camera: the windsock, the low overcast, and the murk — rendered from the same two lines of code we started with.
It is a strange thing to say about weather software, but the 3D view is the part that makes briefing stick. "Overcast zero-zero-five" is an abstraction. A grey lid a hundred and fifty metres over the threshold is a memory.
8. Where the forecast actually comes from
Every forecast site owes you this section, and most don't provide it.
The forecast line is not a new numerical model. It is a blend of seven global ones — ECMWF IFS and AIFS, NOAA GFS, DWD ICON, CMC GEM, Météo-France ARPEGE, UK Met Office — combined in two steps:
- Robust weighting. For each hour, the median of the models' wind vectors is found, and each model is weighted by its distance from that median using a Tukey biweight. Outliers are reduced to a token 5 %, not deleted — a lone model can still be right, but no single run can take the blend hostage.
- Live calibration against METAR. Each model's current-hour forecast is scored against the nearest reporting station. Models matching the observation gain weight; models missing it lose weight, down to a floor of 35 %. This is classical Model Output Statistics applied to a single live observation instead of a long archive — which is what makes it usable the moment you open the page.
With explicit guard rails: calibration only runs when the observing station is within about 55 km and the wind is above 4 kt. Beyond that, terrain dominates. Below that, direction is unstable by nature and scoring against it adds noise rather than skill.
And with honest limits, stated on the page itself: Kalynda inherits everything its inputs get wrong. If all the models miss a sea breeze or a mountain wave, so does the blend. No radar assimilation, no nowcasting, no local terrain model. The full method, including which model is trusted for which variable and why, is on the model page.
That last point is why the site stops at 14 days rather than the 16 the sources publish: past day 14 only one model remains, and a single model's tail is not a consensus.
9. What costs money, and what doesn't
The live map, the decoder, the plain-language reading, the runway and crosswind diagram, a 3-day meteogram and the vertical cloud profile to 40,000 ft are free — with a limit of three METAR/TAF lookups a day.
Beyond that: Pilot at €20/year extends the forecast to 10 days and adds briefings, route briefing with TAF at ETA for destination and alternate, push and e-mail alerts, the real-time 3D simulation and the per-model comparison. Pro at €40/year goes to 14 days and adds the observation archive, trends, world rankings and data export. Fleet/API at €79/year is for schools and operators.
Twenty euros a year is roughly four minutes of flight time in a rented C172. The plans page has the full comparison; there's a 7-day trial.
Try it with your own field
The fastest way to judge any weather tool is to open it on an airport you know well, on a day whose weather you already understand, and see whether it tells you something true.
- Your airport: metarandtaf.com — type an ICAO or IATA code, or a city
- Browse by country: every reporting station, grouped
- The guides: how to read METAR and TAF, properly
- Examples: EGLL London Heathrow · LTBS Dalaman
Eight languages, no app to install, and the raw report always visible at the top of the page — because the decode is a convenience, and the source is the authority.
METAR & TAF 3D is built on official data: observations from the NOAA Aviation Weather Center, forecasts from seven global models via Open-Meteo, airport and runway geometry from OurAirports. It is intended for education and situational awareness and is **not a substitute for an official pre-flight briefing.

Top comments (0)