Blog / METAR & TAF

Temporal Architecture of Aerodrome Forecasts: Parsing Change Indicators as Bounded Windows

An analysis of TAF change indicators—FM, BECMG, TEMPO, and PROB—as bounded temporal windows with distinct baseline, persistence, and probability rules.

September 20, 2026 · VectorWX · 8 min read

Temporal Architecture of Aerodrome Forecasts: Parsing Change Indicators as Bounded Windows

Introduction

Terminal Aerodrome Forecasts (TAFs) encode projected atmospheric states within a five-statute-mile radius of an airport runway complex. While introductory aviation curricula often treat the alphanumeric string of a TAF as a continuous qualitative narrative, operational meteorological data structures depend on deterministic temporal segmentation [1]. The core structural components of these forecasts are change-indicator groups: FM (From), BECMG (Becoming), TEMPO (Temporary), and PROB (Probability) [1], [2].

A persistent pedagogical failure mode occurs when students interpret these indicators as simple descriptive annotations rather than mathematically bounded time intervals. When a forecast student reads a change group as an open-ended narrative remark, the distinction between permanent baseline state transitions and transient perturbation overlays degrades. Correct syntactic parsing requires modeling each change group as a formal time window defined by discrete entry boundaries, exit boundaries, and parameter inheritance rules [2], [4]. Approaching these indicators through interval logic transforms an ambiguous alphanumeric broadcast into a predictable timeline of baseline conditions and overlaid statistical variations [3].

Syntactic Mechanics: Formal Deconstruction of FM, BECMG, TEMPO, and PROB Windows

The syntax of international and domestic aerodrome forecasts relies on standardized change indicators governed by International Civil Aviation Organization (ICAO) Annex 3 and National Oceanic and Atmospheric Administration (NOAA) / Federal Aviation Administration (FAA) orders [1], [4]. Each indicator enforces unique temporal boundaries and state-persistence rules.

00:00Z               06:00Z               12:00Z               18:00Z               24:00Z
  |--------------------|--------------------|--------------------|--------------------|
  [===== Base Forecast Group =====)
                       [===== FM120600 New Prevailing Base State =====================)
                                            [-- BECMG 12/14 --) ==> [ New Prevailing )
                                            [--- TEMPO 12/16 ---] (Intermittent <50%)
                                                       [-- PROB30 14/18 --] (30% Risk)

The FM (From) Group: Complete Baseline State Replacement

The FM indicator denotes an abrupt, permanent transition from one prevailing atmospheric baseline to another [1], [2]. It is represented syntactically as FMddhhmm (day, hour, and minute in Coordinated Universal Time, UTC), establishing a strict start boundary at minute-level resolution:

tstart = tFM
tend = min(tnext FM, tnext BECMG end, tTAF expiry)

Mechanically, an FM group executes an unconditional overwrite of all atmospheric variables [1], [4]. Every parameter—surface wind vector, horizontal visibility, present weather, sky cover, and non-convective low-level wind shear—must be restated [1]. No meteorological element carries over from the preceding block.

If an initial prevailing block forecasts broken cloud ceilings at 2,000 feet and a subsequent FM group specifies only high overcast layers without referencing low cloud, the low cloud layer ceases to exist in the forecast model precisely at tstart [2], [4]. The FM window remains the active baseline until bounded by the start of a subsequent FM timestamp or the completed termination of a BECMG interval [3], [4].

The BECMG (Becoming) Group: Gradual Ramp Transitions and Selective Evolution

The BECMG group models a gradual, sustained atmospheric transition across a defined start-and-end time interval [1], [4]. It is encoded as BECMG ddhh/ddhh, where the first four digits define the window opening (t0) and the second four digits mark the window closing (t1) [2], [3].

Unlike the instantaneous parameter reset of an FM step-function, BECMG functions as a temporal linear ramp:

  • During the interval [t0, t1]: The atmosphere is actively transitioning. Observed values may fluctuate between the previous prevailing baseline and the newly introduced parameters [1], [4].
  • At and after t1: The newly stated conditions reach completion and establish the new prevailing baseline for the aerodrome, which persists until the next structural change group [1], [3].

A critical parsing rule that separates BECMG from FM is selective persistence [3], [4]. The BECMG group only specifies elements that are actively changing; unmentioned variables automatically carry over from the preceding prevailing state [3]. If a BECMG window between 1400 UTC and 1600 UTC (BECMG 0814/0816) encodes an increase in surface wind speed but omits ceiling and visibility groups, the ceiling and visibility values from the prior prevailing block remain entirely operational through and beyond t1 [2], [3].

The TEMPO (Temporary) Group: Bounded Perturbation Overlays

The TEMPO indicator models transient, short-lived excursions from the active prevailing baseline [1], [2]. Formatted as TEMPO ddhh/ddhh, it establishes a bounded validity interval [t0, t1] subject to rigid mathematical duration limits under WMO and FAA specifications [1], [4]:

  1. Single-Instance Constraint: Any continuous occurrence of the specified weather phenomenon must last strictly less than 60 minutes [1], [4].
  2. Aggregate Duration Constraint: The cumulative duration of all temporary occurrences combined must total less than 50% of the entire window length:
∑k Δtk < ½ (t1 − t0)

Because TEMPO conditions represent intermittent excursions rather than persistent shifts, the prevailing baseline does not collapse [1], [2]. During a TEMPO window, the meteorological state oscillates stochastically between the underlying prevailing baseline and the temporary overlay [2], [3].

Once the terminal boundary t1 is reached, the TEMPO overlay vanishes immediately, and the atmosphere reverts exclusively to the active prevailing baseline without requiring an explicit canceling code [1], [4].

The PROB (Probability) Group: Quantified Risk Envelopes

The PROB group introduces probabilistic risk modeling into the aerodrome timeline, typically restricted to PROB30 or PROB40 (a 30% or 40% probability of occurrence) [1], [4]. The syntax defines a discrete temporal window: PROB30 ddhh/ddhh.

Meteorologically, PROB denotes hazardous phenomena—such as convective squalls or low visibility ceilings—whose occurrence within the time envelope [t0, t1] is uncertain [1], [4]. Under standard forecast protocols:

  • Probabilities below 30% are omitted as statistically negligible noise [4].
  • Probabilities of 50% or greater cannot be encoded as PROB; forecasters are required to integrate them directly into prevailing FM, BECMG, or deterministic TEMPO groups [3], [4].

When combined as a compound group (PROB30 TEMPO ddhh/ddhh), both probabilistic and duration constraints intersect [2], [4]. The parsing model must treat the syntax as a 30% probability that temporary fluctuations—each lasting under one hour and aggregating to less than half the window duration—will manifest inside the [t0, t1] interval [2], [4].

Change Group Syntax Window Definition Baseline Interaction Parameter Inheritance Rule
FMddhhmm tstart to subsequent baseline marker Complete replacement at tstart Zero inheritance; all elements must be explicitly restated [1], [4]
BECMG ddhh/ddhh Gradual ramp over [t0, t1] Transitions during window; prevailing at t1 Selective persistence; omitted elements carry over [3], [4]
TEMPO ddhh/ddhh Bounded interval [t0, t1] Intermittent overlay; baseline remains active Transient; specified elements revert to base at t1 [1], [2]
PROB[30|40] ddhh/ddhh Risk window [t0, t1] Probabilistic envelope; base remains active Non-deterministic; baseline unaffected if unfulfilled [2], [4]

Comparative Methodologies in Meteorological Timeline Modeling

Aviation education has historically relied on narrative visualization, where an individual mentally aggregates text lines to form a general impression of arriving weather. However, cognitive friction surfaces when multi-layered forecasts feature nested intervals—such as a 4-hour TEMPO bracket situated inside a broader 12-hour FM prevailing block alongside an overlapping PROB30 group [1], [2].

Time (UTC)       12:00    13:00    14:00    15:00    16:00    17:00    18:00
Prevailing:      [================== FM121200 Baseline ======================>
Overlays:                 [------- TEMPO 1213/1216 -------]
                                   [------- PROB30 1214/1218 -------]
Effective State: [ Base ] [ Base ~ ][ Base ~ TEMPO ~ PROB30 ][ Base ~ PROB30]

To resolve these ambiguities, modern weather analysis methods contrast unstructured human heuristic decoding against deterministic, state-machine processing:

[Raw Alphanumeric TAF]
         |
         v
[Lexical Parser] ---------> Split into FM, BECMG, TEMPO, PROB tokens
         |
         v
[Timeline Projection] ----> Map tokens to [t_start, t_end) coordinate intervals
         |
         v
[State-Machine Evaluator]
         |--> Layer 0: Continuous Prevailing Baseline (FM step / BECMG ramp)
         |--> Layer 1: Intermittent Stochastic Perturbations (TEMPO overlays)
         +--> Layer 2: Quantified Risk Envelopes (PROB30/PROB40 windows)
  1. Sequential Human Parsing: The student reads line by line. When an overlapping TEMPO or BECMG code appears out of chronological order, comprehension requires backtracking to verify which parameters are active. This frequently introduces baseline corruption, where temporary values are mistakenly retained after the window has closed [3].
  2. Interval State-Machine Modeling: The parsing algorithm decomposes the text into discrete time coordinates [tstart, tend). The forecast becomes an array of continuous states: Layer 0 contains the prevailing baseline (updated via FM steps or BECMG transitions), Layer 1 applies intermittent TEMPO variations within active bounds, and Layer 2 applies PROB risk flags [1], [4].

VectorWX presents TAF change groups as interval-mapped data so students can keep baseline, TEMPO, and PROB layers on a shared timeline. That framing makes it easier to treat BECMG as a distributed gradient interval rather than an instantaneous step, and to keep parameter persistence inside each group's stated bounds.

Digital Schemas and the Evolution of Temporal Weather Data

The global transition from legacy Traditional Alphanumeric Codes (TAC) to the ICAO Meteorological Information Exchange Model (IWXXM) underscores the industry's shift toward explicit temporal windowing. While TAC strings rely on human-interpreted keywords such as FM, BECMG, and TEMPO, IWXXM encodes aerodrome forecasts using Extensible Markup Language (XML) and Geography Markup Language (GML) schemas compliant with ISO 19136.

Under IWXXM schemas, implied time windows are converted into explicit XML elements:

  • An FM group is parsed into an instantaneous gml:TimeInstant, terminating the previous om:OM_Observation and instantiating a completely new target record.
  • BECMG, TEMPO, and PROB groups are mapped into explicit gml:TimePeriod elements defined by distinct <gml:beginPosition> and <gml:endPosition> ISO 8601 timestamps.
  • The conditional nature of TEMPO and PROB is captured via schema attributes (changeIndicator="TEMPORARY_FLUCTUATIONS") rather than inferred from text layout.
This transition highlights the enduring structural premise of TAF change groups: they are mathematical time containers designed to divide the future aerodrome environment into predictable, testable windows.

Whether parsed from a compressed alphanumeric string or ingested as a structured GML schema, decoding a change group requires strictly mapping its start boundary, end boundary, and element persistence rules.

References

  1. https://www.weather.gov/jetstream/taf
  2. https://pilotefb.com/learn/taf-change-groups/
  3. https://flightkneeboard.app/guides/decoding-tafs
  4. https://skybrary.aero/articles/terminal-aerodrome-forecast-taf
TAF TEMPO BECMG data reading