RAIM Prediction Before an RNP Approach: What the Announced Outage Really Changes

Poste de préparation de vol avec cartes de navigation et carte de couverture satellitaire

Night flight to an airfield whose only instrument procedure is an RNP APCH, and the flight file announces an unfavorable integrity prediction over the arrival window, with a scheduled satellite outage behind it. Give up, delay the arrival, switch to an alternate — or notice that the prediction does not say what it is being credited with saying? Everything hinges on the mechanism behind the rule: it is what determines what the announced outage actually takes away, and what it leaves intact.

What RAIM monitors, and what “available” means

Receiver autonomous integrity monitoring is not an accuracy indicator. It does not answer “what is my position error”, but a much narrower question: “would the receiver notice, quickly enough, that a measurement is wrong?”

The mechanics are arithmetic. A three-dimensional position with clock bias counts four unknowns: four satellites determine it exactly, with no remainder, hence no possible check. A fifth creates a residual, the system becomes overdetermined and the inconsistency of a measurement becomes measurable — fault detection. A sixth makes it possible to identify the offending measurement and discard it — detection and exclusion. The satellite count, however, never concludes on its own: the sensitivity of the test depends on geometric conditioning, and six satellites clustered in the same sector of the sky are worth less than a well-spread fifth.

From four to six measurements: position, detection, exclusionThree celestial domes above a receiver. Four measurements determine the position without allowing any check. The fifth creates a residual and opens detection. The sixth makes it possible to discard the false measurement. A note recalls that a clustered geometry cancels this gain.4 measurements5 measurements6 measurementsexact positionno checkmeasurable residualdetectionfalse measurement discardedexclusiona clustered geometry cancels this gain
The satellite count does not give an accuracy, it gives a right to check. Four measurements exhaust the unknowns, the fifth opens detection, the sixth opens exclusion. The cross marks the discarded measurement; for an equal count, a clustered distribution is worth less.

What the receiver produces is therefore not the actual error, which remains unknown, but a statistical upper bound — the horizontal protection level — computed from the geometry of the moment and an a priori model of pseudorange error. Integrity is available when this upper bound stays below the alert limit for the flight phase. Three terms define the contract and must be read together: the alert limit, the maximum time to announce an exceedance, the probability of failing to detect it. Tightening the alert limit, which is what the approach does, mechanically makes availability more fragile.

RAIM doesn’t tell you your exact position, but whether a satellite is lying. It needs at least five satellites to detect a problem and six to exclude it. The tighter the approach rules, the harder it is to meet these checks.

At an airport like KJFK, an A320 with a GPS-only receiver might lose RAIM availability for 15 minutes during a specific approach window due to a satellite outage. However, if the same aircraft uses a multi-constellation receiver (GPS + Galileo), it might maintain availability throughout, showing how equipment assumptions matter.

Two additions often change the verdict. Barometric aiding injects pressure altitude as an extra measurement and frees up a satellite for checking: it improves availability, but assumes a correct setting, which shifts part of the problem onto altimetry. And the gap between detection alone and exclusion is not academic: the former results in an alert, hence a loss of function, the latter allows the flight to continue. Over oceanic or remote airspace, the requirement frequently bears on exclusion.

Why this availability can be predicted, and why it shifts

Satellite geometry is deterministic. The orbits are known and broadcast, and the ground track repeats: for the GPS constellation, the repeat period is close to the sidereal day, so that a configuration observed at a given time recurs about four minutes earlier the following day. The other constellations have their own, longer repeat periods, so this daily shift does not carry over mechanically to a multi-constellation receiver. This determinism is what allows prediction, and explains why an unfavorable window is local and brief, not a state of the system.

Removing a satellite does not bring the coverage down: it removes margin. The upper bound degrades where the residual geometry is least favorable, at certain latitudes and certain times. At a given point, unavailability is typically counted in minutes or tens of minutes. Hence the very shape of the “prediction” object: a location, a window bracketing the estimated time of arrival, a flight phase, an equipment assumption. None of the four is optional — changing one changes the answer.

Horizontal protection level over the hour, compared with two alert limitsA solid curve represents the computed protection level. Two dashed lines mark the en-route alert limit, wide, and the approach alert limit, tightened. The curve stays below the en-route limit throughout the period but crosses the approach limit over a narrow band bracketing the estimated arrival: the unavailability window.protection leveltimeen-route limitapproach limitunavailable windowestimated arrival
An announced outage does not bring the coverage down: it brings the upper bound closer to a limit, and only the tightest one is crossed. The solid line is the computed protection level, the dashes are the alert limits. Reading the width of the band gives the order of magnitude of the arrival shift that would suffice.
Orders of magnitude of alert limits and time-to-alert by flight phase
Flight phase Horizontal alert limit Time to alert Effect on availability
Oceanic and remote airspace The widest, a few nautical miles The most permissive, on the order of several minutes Rarely at issue; the requirement bears on exclusion
Continental en route On the order of two nautical miles Same order as oceanic, several minutes Met with a nominal constellation
Arrival and terminal area On the order of one nautical mile On the order of ten seconds Sensitive to multiple failures
RNP approach, lateral minima The tightest, tenths of a mile Same order as the terminal area: it is the alert limit, not the time to alert, that discriminates The only discriminating level

Two predictions that do not tell the same story

There are two ways to obtain a prediction, and their blind spots are opposite. The onboard function computes with the equipment’s actual algorithm, its error model, its mask angle, its constellation — exactly the system that will fly the approach. But it reasons from the received almanac, which can be old if the equipment has been powered off for a long time, and above all it ignores what is not in the navigation message: a satellite declared unavailable by a notice to airmen but still broadcast as healthy remains, for it, usable.

The ground service does the opposite. It incorporates announced outages, that is its purpose, but reasons on a generic receiver, with conservative assumptions and often a single constellation. It does, however, produce a dated document, attached to the flight file and defensible at audit — which the receiver’s screen is not.

The two answers therefore diverge in both directions. A single-constellation service will declare unavailable a window that a multi-constellation receiver would cross without a hitch; a receiver that ignores the announced outage will display a reassuring and false availability. The question is never “which one to believe” but “which one matches the approval under which the flight is operating”.

Onboard prediction and ground service: what each one knows and does not know
Criterion Onboard function Ground service
Source of geometry Almanac received on board, possibly outdated Orbital data kept up to date
Announced outages Ignored, except a satellite broadcast as unhealthy Taken into account: that is the purpose of the service
Receiver model Exactly that of the equipment Generic receiver, conservative assumptions
Constellation The one actually used on board Often a single one
Mask angle That of the installation Conventional value
Scope queried What the crew enters, often the destination Destination, alternates, requested window
Record for the file Display, not archived Dated document, attachable

What the prediction does not cover

An integrity prediction is a geometry calculation under assumptions: everything that is not geometry escapes it. Jamming and spoofing first. A GNSS interference notice and an unfavorable prediction do not stem from the same mechanism and do not call for the same response. The former announces a loss of signal tracking, hence a loss of function: no position, or — more insidiously — a position whose falseness the receiver cannot measure if the spoofing signal is consistent across all channels. Integrity checking compares measurements against each other: it detects a satellite that is lying, not a ground transmitter that makes them lie in concert.

Local masking next: nearby terrain, ground structures, multipath, the prolonged bank of an interception turn that removes a portion of the low sky from the antenna. Ionospheric scintillation, at equatorial and auroral latitudes, is only partially represented in the a priori models.

The actual state of the installation, finally, which the prediction assumes to be nominal: a degraded receiving chain, an aging antenna, a second receiver deferred to the minimum equipment list. And everything that has nothing to do with the signal: database validity, amended procedure, operational approval. A favorable prediction does not make usable a procedure that is unusable for some other reason.

The trap. A prediction is bounded in place, time and flight phase, and these bounds are quickly forgotten. Querying only the destination means discovering in flight that the unfavorable window covers precisely the alternate. Sticking to the estimated time at filing means discovering during descent that a holding pattern has slid the arrival into the gap. And a prediction made for lateral minima says nothing about vertical minima.

Integrity produced on board or by the ground: the question changes nature

On a receiver whose integrity is derived on board, the approach down to lateral minima does indeed depend on autonomous monitoring, and the prediction is the right question. As soon as the approach descends to vertical minima based on a satellite augmentation, the mechanism is different: integrity is produced by a ground segment, broadcast as corrections and confidence bounds, and the protection levels no longer come from a consistency test between satellites.

Two consequences follow, and they are not symmetrical. An unfavorable autonomous integrity prediction does not entail the loss of vertical minima if the equipment and the procedure are augmented. But the reverse also exists: near the edge of the service area, or when ionospheric corrections degrade, vertical minima drop while lateral minima remain available. An augmentation service is subject to its own predictions and its own notices, which are not interchangeable with satellite outage notices.

Remaining minima depending on autonomous integrity and augmentation service statusTwo-by-two matrix. Columns: autonomous integrity available or unavailable; rows: augmentation service nominal or degraded. The four cells give respectively lateral and vertical minima, vertical minima only, lateral minima only, and no satellite-based minimum. The two asymmetric cells are highlighted.autonomous availableautonomous unavailablenominalaugmentationdegradedaugmentationlateral + verticalvertical onlylateral onlyno minimumremaining minima in each case
The two integrities fail independently, and their losses do not remove the same minima. The two highlighted cells are the least expected: an unfavorable autonomous prediction leaves vertical guidance alive, and a degraded augmentation service removes vertical guidance while leaving lateral guidance.

The translation into flight preparation is concrete: decision height, margins and diversion fuel differ with and without vertical guidance. Falling back on lateral minima during descent means discovering late that the chosen alternate no longer works.

Unfavorable prediction: what remains playable

A negative answer is almost never a blanket lock. It comes apart when you read its four bounds.

  • The time bound. How long does the window last? If it is counted in minutes, shifting the arrival time is often enough.
  • The phase bound. An unavailability during approach generally leaves en-route and arrival integrity intact, because their alert limits are much wider. It is not the flight that is called into question, it is one approach option.
  • The equipment bound. Under what constellation and receiver assumption was the answer produced, and does it match what is approved?
  • The geographic bound. Does the window cover the alternates, including the one that was not queried because it seemed obvious?

The rest is a matter of the usual trade-off: fall back on an approach independent of GNSS, choose an alternate equipped with a conventional aid, or accept the fuel penalty of a more distant diversion — a penalty decided when the flight plan is filed. An airfield whose every procedure is GNSS-based is not a relevant alternate during a regional unavailability, whatever its weather.

The right instinct. Query the prediction over the full scope — destination, alternates, window widened to foreseeable holding — and record the equipment assumption together with the result. An answer kept without its assumption is unusable, both in flight and afterward.

Decision matrix: what an unfavorable notice removes, depending on what it announces
What is announced What is removed What remains usable What to prepare
Insufficient integrity on approach over a window Lateral minima, on that window only En route, arrival, non-GNSS approaches, augmented minima Arrival shift, backup approach, conventional alternate
Augmentation service unavailability Vertical minima based on that service Lateral minima if onboard integrity holds Minima, margins and fuel without vertical guidance
Long-duration satellite outage Margin across the whole constellation The rest of the constellation, more fragile geometry Destination and all alternates queried
Regional GNSS interference notice The navigation function, not just its check Ground aids, inertial if the aircraft is so equipped Loss-of-position procedure, cross-checking, dedicated briefing
Equipment deferred to the minimum list Operations requiring redundancy or exclusion What the list explicitly allows Review of the associated conditions before departure

What to go and check in your own documentation

The mechanism is the same everywhere, its translation is not. Four points deserve to be found in black and white, in four documents.

Two men in an aircraft cockpit review a flight plan on a tablet, with a navigation display in the background.
RAIM outages require proactive flight plan adjustments for RNP approaches.

In the operations manual: which cases trigger a check, which service is designated, who keeps the answer and in what form, what flight preparation does when it is unfavorable. In the flight manual and the approval documents: equipment class, constellations covered, any requirement for barometric aiding, and whether the airspace flown requires exclusion and not detection alone. In the manufacturer’s documentation: the system’s behavior when integrity is lost on final — what annunciation, and what response is expected depending on whether the aircraft is before or after the final approach fix; this is the line that is reread the least and that decides the fastest. In the aeronautical information publications, finally: the way the State concerned disseminates these notices, which varies more than one might think.

A structural gray area remains. The approval of multi-constellation receivers is progressing faster than the assumptions built into prediction tools: a crew may have a more capable system than the model that judges its availability, with no convenient way to assert that gap. The question belongs to the authority and the operator, it is not settled in the briefing room. For anyone preparing a type rating, this is exactly where the examination bears: not on the definition of integrity monitoring, but on the reasoning held when the answer is negative, the alternate is far away, and the fuel is loaded. The PBN modules at formations.fl330.com cover this type of case.

The applicable regulatory text, the aeronautical information publications of the States concerned, and the operator’s approved documentation take precedence over this article, whose sole purpose is to explain the mechanism.

Questions fréquentes

Does an unfavorable integrity prediction prohibit the flight?

No, it removes a capability over a bounded scope. An outage announced at approach level generally leaves en-route and arrival integrity available, because the alert limits there are much wider. What is lost is one approach option over a window, and the appropriate response is to shift the arrival time, choose an approach independent of GNSS, or select an alternate equipped with a conventional aid.

Should you trust the receiver's prediction or the ground service's?

Both are useful because their blind spots are opposite. The onboard function uses the equipment’s actual algorithm and constellation, but ignores outages announced by notice to airmen as long as the satellite is still broadcast as healthy. The ground service incorporates these notices, but reasons on a generic receiver and often on a single constellation. The selection criterion is the equipment assumption that matches the operator’s approval, and it is the service designated by the operations manual that is authoritative for the file.

Does the RAIM prediction apply to an approach flown down to vertical minima with satellite augmentation?

No, the integrity mechanism there is different. With satellite augmentation, integrity is produced by a ground segment and broadcast as corrections and confidence bounds, instead of resulting from a consistency test between satellites seen by the aircraft. The availability of these minima must therefore be queried from the prediction service specific to the augmentation, and satellite outage notices do not substitute for these predictions.

What is an integrity prediction worth in an area covered by a GNSS interference notice?

It does not answer the question being asked. An integrity prediction models satellite geometry, whereas an interference notice announces a loss of signal tracking, hence a loss of navigation function. A spoofing signal that is consistent across all channels can moreover remain undetectable by a consistency test between measurements. The remedy lies elsewhere: loss-of-position procedure, cross-checking of conventional ground aids, and a briefing specific to the area.

Over what scope should the prediction be queried before departure?

Over the four parameters that define the answer: location, window, flight phase and equipment assumption. In practice, this means querying the destination and all the chosen alternates, and widening the window to foreseeable holding delay rather than sticking to the estimated time at filing. The equipment assumption used must be recorded together with the result, otherwise the answer kept is neither usable in flight nor defensible afterward.

Does a multi-constellation receiver change the reading of the prediction?

It changes the actual availability, not necessarily the tool’s answer. A prediction service that reasons on a single constellation can declare unavailable a window that a multi-constellation receiver would cross without any alert, which creates a gap between the aircraft’s capability and the model judging it. This gap is addressed with the authority and the operator, by checking in the approval documents which constellations are covered, not by a judgment call in the briefing room.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

FL330, c’est toute l’aviation professionnelle — explorez nos autres espaces.