A TAF looks worse than a METAR because it keeps changing its mind mid-sentence. A METAR is one set of conditions. A TAF is several, stacked in time, with rules about which parts of the earlier ones survive.

Learning how to decode a TAF is mostly learning those rules. The weather vocabulary is the same one you already read in a METAR.

Read a TAF in three passes. First the header: station, issue time, valid period. Then the prevailing conditions, which run until something changes them. Then the change groups: FM starts a clean new block, BECMG transitions gradually, TEMPO overlays a temporary condition without replacing anything else, and PROB30 attaches a probability.

What's New in TAF Reading for 2026

The code has not moved. What has changed is how many pilots meet a TAF through a decoder rather than raw, and that changes which mistakes are common.

Nobody misreads a wind group any more, because the tool spells it out. What tools handle badly is carry-forward: what a TEMPO block leaves unchanged, and which conditions survive a BECMG. Those are rules about the structure of the forecast, not about any one group, and a list of decoded groups can be entirely correct while leaving that structure invisible.

So the part worth learning by hand in 2026 is the part a decoder is worst at showing you.

The Header: Three Things Before Any Weather

Every TAF opens the same way.

KDFW 281730Z 2818/2924

• KDFW the station, always the ICAO identifier
• 281730Z issued on the 28th at 1730Z
• 2818/2924 valid from 1800Z on the 28th to 0000Z on the 30th

The issue time and the valid period are different things and both matter. A TAF issued at 1730Z that becomes valid at 1800Z has a short gap where it is not yet in force. More usefully, the issue time tells you how stale the forecaster's picture is.

Two flags can appear before the station. AMD means this is an amended forecast, replacing an earlier one because it went wrong. COR means corrected, a fix to a transmission error rather than a change of mind. They are not the same thing and the difference is worth knowing.

The Prevailing Conditions

Straight after the valid period come the conditions expected at the start, in the same order a METAR uses.

17012G20KT P6SM FEW045

Wind from 170 at 12 gusting 20, visibility better than 6 statute miles, few cloud at four thousand five hundred feet.

Two habits from METAR reading carry across intact. Wind is degrees true, not magnetic. Cloud heights are above ground level, not above sea level.

One thing differs. A TAF uses P6SM for good visibility where a METAR would give you an actual measured value like 10SM. A forecaster is not going to predict the difference between 7 and 10 miles, so the format does not ask them to.

Outside the US you will see 9999 instead, meaning 10 kilometres or more, and visibility in metres throughout.

FM: A Clean Break

FM is the easiest change group because it wipes the slate.

FM290200 15008KT P6SM SCT120

From 0200Z on the 29th, everything changes to what follows. Nothing carries across an FM group. The block it starts is self-contained, and it runs until the next FM or the end of the valid period.

The time is a full six digits: day, hour, minute. The change is treated as happening rapidly, within about an hour.

BECMG: A Gradual Transition

BECMG is where carry-forward starts to matter.

BECMG 2906/2908 24010KT

Between 0600Z and 0800Z on the 29th the wind becomes 240 at 10. That is the only thing this group mentions, so that is the only thing that changes. Visibility and cloud carry forward from whatever was in force before it.

This is the single most misread group in the format. A BECMG that names only a wind has not cancelled the cloud, and a period with no cloud group in it is not a period with no cloud in it.

By the end of the transition window, the new condition is in force and stays that way until something else changes it.

TEMPO: An Overlay, Not a Replacement

TEMPO describes conditions expected to occur temporarily within a window.

TEMPO 2820/2824 3SM TSRA BKN025CB

Between 2000Z and 0000Z, temporarily 3 statute miles in thunderstorms with rain and broken cumulonimbus at two thousand five hundred feet.

Two rules make TEMPO different from everything else:

• The conditions are expected to last less than an hour at a time, and to cover less than half the period in total
• What the group does not mention is unchanged. The prevailing conditions are still running underneath

So a TEMPO block is the bad half hour inside an otherwise acceptable period. For planning, it is usually the deciding factor, because you have to be able to handle the worst case in the window you will be there.

PROB30 and PROB40: Attaching a Probability

PROB30 2902/2906 1SM FG

A 30 percent chance of 1 statute mile in fog between 0200Z and 0600Z.

Only PROB30 and PROB40 exist. Below 30 percent the forecaster says nothing; at 50 percent or more they would use a different group, because a coin flip is not a probability worth qualifying.

PROB30 TEMPO appears often and reads exactly as it looks: a 30 percent chance of temporary conditions.

The Groups at the End

After the last change group, a few things can appear.

• TX26/2919Z and TN16/2910Z are forecast maximum and minimum temperatures, with the time each is expected
• WS020/18040KT is non-convective low level wind shear, here at two thousand feet with wind from 180 at 40 knots
• AMD NOT SKED means the field has no forecaster on duty, so this forecast will not be amended if things change
• NSW means previously forecast significant weather is expected to end

AMD NOT SKED is worth pausing on. It looks like it says amended. It says the opposite.

Decoding One All the Way Through

Rules stick once you have run them over a whole forecast. Here is a complete one:

KDFW 281730Z 2818/2924 17012G20KT P6SM FEW045 TEMPO 2820/2824 3SM TSRA BKN025CB FM290200 15008KT P6SM SCT120 BECMG 2906/2908 24010KT PROB30 2910/2914 2SM BR

Reading it in order:

• Issued 1730Z on the 28th, valid 1800Z on the 28th to 0000Z on the 30th
• From 1800Z: wind 170 at 12 gusting 20, better than 6 miles, few at four thousand five hundred
• Between 2000Z and 0000Z, temporarily 3 miles in thunderstorms with rain and broken cumulonimbus. The wind is unchanged, because the group does not mention it
• From 0200Z on the 29th, a clean break: wind 150 at 8, better than 6 miles, scattered at twelve thousand. The thunderstorms are gone from the picture because FM replaced everything
• Between 0600Z and 0800Z the wind becomes 240 at 10. Visibility and cloud carry forward from the FM block, so it is still better than 6 miles with scattered at twelve thousand
• Between 1000Z and 1400Z, a 30 percent chance of 2 miles in mist

Now ask the question that matters. If you arrive at 0700Z, what do you plan for? Wind somewhere in the middle of a transition to 240 at 10, better than 6 miles, scattered at twelve thousand. Not the thunderstorms, which ended at the FM.

The Three Mistakes Worth Naming

Mistake 1: treating a TEMPO as the new normal. The prevailing conditions run underneath it. A TEMPO telling you 3 miles in thunderstorms has not said the wind changed, and it has not said the rest of the period is bad.

Mistake 2: assuming a BECMG cancels what it does not mention. It changes one thing, or two, and everything else stays exactly as it was. This is how pilots end up reading a period as cloud free when the cloud simply was not restated.

Mistake 3: reading the far end of a TAF as confidently as the near end. A forecast 24 hours out is a much weaker claim than one two hours out, and nothing in the format marks that difference for you. The PROB groups are the only explicit uncertainty in the whole product, and their absence does not mean certainty.

FAQ

What is the difference between FM and BECMG in a TAF?

FM is a clean break: everything after it replaces everything before, and nothing carries forward. BECMG is a gradual transition that changes only what it names, so any group it does not mention stays in force from before.

Does TEMPO replace the conditions around it?

No. TEMPO overlays a temporary condition on top of whatever is already running. Anything it does not mention is unchanged. Reading a TEMPO block as a new baseline gives you a worse picture than the forecast actually contains.

How long is a TAF valid?

Most US TAFs cover 24 hours. Busier airports get 30 hours. The valid period is written as two day-and-hour stamps right after the issue time, so 2818/2924 runs from 1800Z on the 28th to 0000Z on the 30th.

What does P6SM mean in a TAF?

Visibility greater than 6 statute miles. Forecasters are not asked to predict the difference between 7 miles and 10, so the format caps good visibility at this one value. Outside the US the same idea is written 9999.

Are TAF cloud heights AGL or MSL?

Above ground level, the same as a METAR. This is the opposite of a PIREP, where cloud heights are above sea level unless the pilot says otherwise, which is why mixing the two products up causes trouble at high elevation fields.

Verdict on How to Decode a TAF

Decoding a TAF is two skills, and only one of them is vocabulary.

The vocabulary is borrowed from the METAR you already read. The part that is genuinely new is the structure: which group replaces, which group overlays, and which conditions survive when a change group does not mention them.

Learn FM replaces, BECMG changes only what it names, and TEMPO overlays, and most of the format falls into place. The rest is worth letting a decoder handle, as long as it shows you the raw group beside the plain English so you can check it.

Last updated August 30, 2026