How accurate are TAFs?

Tafrob editorial team · Updated · 6 min read

Short answer

A TAF gives the most probable value of each element, not a guarantee. EASA guidance sets operationally desirable targets: wind direction within 20 degrees and wind speed within 5 knots in 80% of cases, visibility within set margins in 80%, and cloud amount, cloud height and temperature in 70%. Forecast offices verify their TAFs against observations and amend them when needed.

Key points

  • Every value in a TAF is the most probable value, not a certainty.
  • EASA guidance targets 80% of cases for wind, visibility and precipitation, and 70% for cloud and temperature.
  • The targets are tolerances: 20 degrees, 5 knots, 200 m or 30% for visibility, 100 ft or 30% for cloud height.
  • The NWS verifies TAFs with its Stats on Demand program and stresses accuracy at flight category limits.
  • Amendments, TEMPO and PROB groups are how a TAF handles its own uncertainty.
  • The latest METAR is the quickest check of whether a TAF is on track.
A US Army aviation weather forecaster checking satellite imagery in a staff weather office at Baghdad International Airport
Photo: Sgt. 1st Class Thomas Mills, Public domain, via Wikimedia Commons

A forecast of the most probable weather

A TAF reads like a precise statement: a wind of 240 degrees at 15 knots, visibility 6000 metres, a broken layer at 1,200 feet. None of these figures is a promise. WMO-No. 782 states that, owing to the variability of meteorological elements in space and time, to the limitations of forecasting techniques and to the definitions of some elements, the value of any element given in a forecast should be understood as the most probable value it is likely to take during the period.

The Transport Canada AIM uses the same wording: a TAF describes the most probable weather conditions expected at an aerodrome, together with their most probable time of occurrence. The pillar page on reading a TAF covers the code; this page covers how close that code is expected to come to reality, and how forecasters check it.

The accuracy targets

The clearest published targets come from European guidance. GM3 MET.TR.220 in the EASA Acceptable Means of Compliance and Guidance Material to Part-MET says the accuracy of TAFs should be monitored using verification methods agreed with the competent authority. It lists the elements to verify (wind direction, wind speed, visibility, precipitation, cloud amount and cloud height, plus maximum and minimum temperature where forecast) and gives the operationally desirable accuracy:

Element Operationally desirable accuracy Minimum share of cases within range
Wind direction Within 20 degrees 80%
Wind speed Within 2.5 m/s (5 knots) 80%
Visibility Within 200 m up to 800 m; within 30% between 800 m and 10 km 80%
Precipitation Occurrence or non occurrence 80%
Cloud amount One category below 450 m (1,500 ft); occurrence or non occurrence of BKN or OVC between 450 m and 3,000 m (10,000 ft) 70%
Cloud height Within 30 m (100 ft) up to 300 m (1,000 ft); within 30% above 300 m 70%
Air temperature Within 1 degree Celsius 70%

Two points stand out. First, the targets are tolerances, not exact hits: a forecast of 4000 metres visibility counts as accurate if the observed value is between about 2800 and 5200 metres. Second, the targets are not 100%. Even a TAF that meets every European goal can be outside the tolerance for wind in one case out of five, and for cloud height in nearly one case out of three.

Reading the table

  • Wind. A forecast of 24015KT is on target if the observed wind is between 220 and 260 degrees and between 10 and 20 knots, in at least 80% of cases.
  • Visibility. At low values the margin is absolute (200 metres up to 800 metres), because a few hundred metres change the approach minima. Above 800 metres it becomes a percentage.
  • Cloud. Below 1,500 feet the target is to get the amount within one category, for example SCT instead of BKN. Between 1,500 and 10,000 feet the test is only whether a ceiling (BKN or OVC) occurs.
  • Precipitation. The test is occurrence or non occurrence, not intensity or type.

How TAFs are verified

Verification compares each forecast with the observations that follow it, usually METARs and SPECIs from the same aerodrome. WMO-No. 782 notes that the preparation, amendment and cancellation of TAFs depend, among other things, on receipt of the aerodrome's METAR or SPECI.

The NWS approach

NWS Instruction 10-813 sets out quality assurance in section 7. It states that the goal of aviation forecasting is to improve forecast services continually by identifying weaknesses and developing methods to strengthen or eliminate them. It names Stats on Demand as the primary program for TAF verification, encourages offices to send performance reports to staff and to stakeholders such as airport managers and airlines, and states that results tracked in the program will never be used against forecasters.

The instruction also says where accuracy matters most. Appendix C explains that changes between the flight categories VFR, MVFR, IFR and LIFR have significant operational impact, for example on fuel and alternates, and that the TAF should be especially accurate for those values. A forecast that misses the ceiling by 300 feet in clear weather matters little; the same miss across an alternate minimum matters a great deal.

Probability and reliability

The Aviation Weather Handbook explains why probability is part of the picture. A deterministic forecast gives a single solution; a probabilistic forecast expresses uncertainty as the probability of different outcomes. The handbook states that weather forecasts contain uncertainty, especially further out in time, and that a probabilistic forecast must be reliable: if thunderstorms are forecast at 70%, they should occur seven times out of ten.

For a TAF, this applies to the PROB30 and PROB40 groups. Over many forecasts, conditions given as PROB30 should occur roughly three times out of ten. A PROB30 that never verifies is as unreliable as one that always does.

How a TAF handles its own uncertainty

The code has several tools for showing that the weather is not certain.

  • Change groups. TEMPO marks temporary fluctuations, BECMG a gradual change, FM a rapid change at a set time. NWS Instruction 10-813 limits TEMPO groups to four hours and PROB groups to six hours or less.
  • Probability groups. TC AIM states that a probability below 30% of actual values deviating from the forecast does not justify a PROB group, and that when the alternative is 50% or more, BECMG, TEMPO or FM should be used instead.
  • Amendments. WMO-No. 782 states that amendments are issued as and when necessary, and that a later TAF automatically replaces the earlier one. The thresholds that trigger an amendment are listed on the amended TAF page.
  • Regular reissue. WMO-No. 782 states that TAFs valid for 12 to 30 hours are issued every 6 hours and shorter ones every 3 hours, so the forecast for any given hour is refreshed several times before that hour arrives.

When observation and forecast differ

A difference between the METAR and the TAF is not always a forecast error. TC AIM gives an example with automated stations: an AWOS cloud height sensor tends to under read during precipitation, and the forecaster may have inferred conditions from satellite, radar, lightning data, pilot reports and nearby stations. It states that on the few occasions when a METAR AUTO and a TAF differ, this may not imply that the TAF is inaccurate, or that an amendment is required.

The opposite case is also covered. When observations cross an amendment threshold and the TAF does not reflect it, the forecast is out of date until an amendment is issued. The TAF versus METAR page shows how to read the two together.

What this means for a pilot

  • Read values as central estimates. A visibility of 5000 metres in a TAF means visibility around that value, not exactly it.
  • Watch the margins near minima. A ceiling forecast just above an alternate or approach limit leaves little room inside the 30% cloud height tolerance.
  • Take TEMPO and PROB seriously. They are where the forecaster puts the conditions most likely to cause a surprise.
  • Use the latest issue. Check for TAF AMD and compare with the latest METAR and any SPECI.
  • Apply the planning rules. Regulatory rules on alternates and change groups, described on the flight planning page, build margins on top of the forecast.

Method and limits

The accuracy table is taken from GM3 MET.TR.220 in the EASA AMC and GM to Part-MET, Issue 1 Amendment 2. The columns of the source PDF are misaligned in text extraction; the rows above are reconstructed from the listed elements, tolerances and percentages, which appear in the same order as in the extract. The EASA guidance mirrors ICAO Annex 3, which was not consulted directly. The table states targets, not measured results. No published statistics on the actual verification scores of TAFs, in Europe or in the United States, were found in the sources cited here, so this page does not give any; NWS Stats on Demand results are internal. This page has no "In real reports" box because accuracy cannot be measured from the code of a TAF alone.

Training and reference only. For flight planning, always use official briefing sources (your national weather service, AIS or approved provider).

Frequently asked questions

How accurate is a TAF supposed to be?

EASA guidance on Part-MET gives operationally desirable accuracy for TAFs: wind direction within 20 degrees, wind speed within 5 knots, visibility within 200 m up to 800 m and within 30% from 800 m to 10 km, all in 80% of cases. Cloud amount, cloud height and temperature have a 70% target.

Is a TAF wrong if the METAR shows different weather?

Not necessarily. WMO-No. 782 states that each value in a TAF is the most probable value during the period. A small difference is expected. When the difference crosses an amendment threshold, such as a flight category limit, the forecaster should issue an amended TAF.

Who checks whether TAFs are accurate?

The forecast provider. EASA guidance asks for verification methods agreed with the competent authority. NWS Instruction 10-813 names Stats on Demand as the primary NWS verification program and says its results will never be used against forecasters.

Why does a TAF use TEMPO and PROB groups?

They express uncertainty in time and in occurrence. TC AIM states that a probability below 30% does not justify a PROB group, and that 50% or more calls for BECMG, TEMPO or FM. The groups show what else may happen besides the most probable conditions.

Are TAFs less accurate later in the validity period?

The Aviation Weather Handbook notes that forecasts contain uncertainty, especially further out in time. TAFs are reissued every 6 hours for long validities, and amended when needed, so the most recent issue is the one to use.

Sources

  1. European Union Aviation Safety Agency (2022). AMC and GM to Part-MET, Issue 1 Amendment 2 (Annex IV to ED Decision 2022/004/R), GM3 MET.TR.220
  2. National Weather Service (2024). NWS Instruction 10-813, Terminal Aerodrome Forecasts (October 30, 2024), section 7 and Appendix C
  3. World Meteorological Organization (2025). Aerodrome Reports and Forecasts: A Users' Handbook to the Codes (WMO-No. 782), 2025 edition
  4. Federal Aviation Administration (2024). Aviation Weather Handbook (FAA-H-8083-28A), section 3.2
  5. Transport Canada (2025). TC AIM, MET section 7.0 Aerodrome forecasts (TAFs)