Cold Weather Altimetry: Why Your Altimeter Over-Reads in the Cold

A pressure altimeter does not measure height. It measures static pressure and converts it into feet using the International Standard Atmosphere — and on the winter days when terrain clearance matters most, that is not the atmosphere you are flying in. In air colder than ISA the instrument reads higher than you actually are, and the error grows with the height you have climbed above the station whose setting is on your subscale. What follows is where that error comes from, the datum it is measured from — the point most notes get wrong by hundreds of feet — the rule of thumb and exactly where it stops being adequate, the ICAO correction table, which altitudes you must correct and who you must tell, and how to find the lowest usable flight level when QNH is below 1013.

Why a Pressure Altimeter Over-Reads in Cold Air

A pressure altimeter contains nothing that can sense height. It contains an aneroid capsule that senses static pressure, and a mechanism that converts that pressure into feet using one fixed, built-in relationship: the International Standard Atmosphere. The subscale moves the datum pressure. It does not change the atmosphere the instrument believes it is flying in.

That atmosphere is 1013.25 hPa and +15°C at mean sea level, cooling at 1.98°C per 1000 ft (6.5°C/km) to the tropopause at 36 090 ft, then constant at −56.5°C — the same structure drilled in the ATPL meteorology questions. So the instrument never answers the question you asked. You asked how high am I; it answers in a standard atmosphere, what height would this pressure correspond to? Those are the same number only on a standard day.

What links them is the thickness of the air: the vertical distance between two pressure surfaces is proportional to the mean absolute temperature of the layer between them. Cold air is dense, so pressure falls away faster with height and the pressure surfaces lie closer together. Climb until the capsule sees the pressure the mechanism has labelled "3000 ft" and, in cold air, you have physically climbed less than 3000 ft. The indication is right for the pressure; the height it represents is not there.

Hence the direction of the error, which is the one thing you must never get backwards: colder than ISA, the altimeter over-reads — true altitude is lower than indicated. Warmer than ISA it under-reads and you are higher than indicated, which matters to a baro-VNAV path but not to terrain.

Indicated, pressure, true and density altitude

Four terms are worth separating, because examiners mix them deliberately. Indicated altitude is what the altimeter reads on whatever is set; pressure altitude is what it reads on 1013.25; true altitude is your actual geometric height above mean sea level, the one terrain cares about; and density altitude is pressure altitude corrected for temperature — a performance quantity with nothing to say about terrain clearance.

QNH, QFE, 1013 — and 1 hPa ≈ 27 ft

Setting the subscale moves the datum the instrument counts up from. It leaves the ISA assumption inside the mechanism untouched, which is why no altimeter setting will ever remove the temperature error. Altimeter setting procedures are examined in their own right across the EASA Instrument Rating theory papers, and this is the distinction they turn on.

• QNH — station pressure reduced to mean sea level using the ISA relationship. The altimeter reads altitude above mean sea level, and on the ground at the issuing station it reads that station’s elevation.
• QFE — pressure at a specified datum, normally aerodrome or threshold elevation. The altimeter reads height above that datum and zero on the ground there.
• 1013.25 hPa — the standard setting. The altimeter reads pressure altitude, expressed as a flight level, used above the transition altitude climbing and above the transition level descending.
• Regional or area QNH — the lowest forecast QNH for a region over a period. Deliberately pessimistic, so it buys a little terrain margin, but it is not the pressure at any particular place.

Feet per hectopascal

The hectopascal-to-feet conversion is set by air density, so it is not a constant. In ISA at mean sea level 1 hPa spans 8.3 m, which is 27.3 ft. The familiar 30 ft per hPa is a rounded working figure: close to correct a couple of thousand feet up, and the safer of the two when correcting for a QNH below 1013 because it gives the larger number. Both grow steadily with height:

Two things follow. At an aerodrome the gap between the two surface settings is roughly the elevation divided by 27, so at a 540 ft aerodrome the QFE is about 20 hPa below the QNH. And at cruising levels a 1 hPa error in a setting is worth far more than 27 ft, which is part of why everyone above the transition level shares the single 1013.25 datum.

Carry the 27 ft figure forward as exactly what it is: the ISA value at sea level, and the number the lowest usable flight level calculation later in this article has to convert with — alongside its deliberately conservative 30 ft alternative. Because feet per hectopascal grows with height, 27 ft slightly understates the offset of a low QNH, which is why the 30 ft figure is the one to reach for when the question does not say. EASA questions normally state which conversion to use; if not, say which you used.

Where the Error Is Measured From

This is the part most notes get wrong, and getting it wrong moves the answer by hundreds of feet — it sits alongside magnetic compass errors as an instrument topic students routinely learn backwards. The temperature error is zero at the elevation of the altimeter setting source and grows with height above it. It is not measured from sea level, and not from the terrain beneath you.

The reason is that QNH is built from the pressure actually measured at that station and reduced to sea level using exactly the same ISA relationship the altimeter mechanism uses. The two fictions cancel. Stand on the apron there, set that QNH, and the altimeter reads the aerodrome elevation precisely — at −30°C just as at +30°C. Error accumulates only over the layer of real, non-standard air you then climb through.

So the correction is a percentage of the height above the elevation of the altimeter setting source:

Height to correct = published altitude − elevation of the altimeter setting source

Worked example: a 4600 ft MSA from a 1200 ft aerodrome

Aerodrome elevation 1200 ft, minimum sector altitude 4600 ft, reported temperature −20°C. The height above the source is 4600 − 1200 = 3400 ft, not 4600 ft. The ICAO table reproduced two sections below gives 480 ft for that height and temperature, so the corrected figure is 5080 ft. Round the figure you actually fly up to the next 100 ft, and you fly 5100 ft indicated.

Had you taken the whole 4600 ft as the height to correct, the same table would have given 660 ft, a corrected 5260 ft and 5300 ft to fly on the same rounding convention. That is safe — it is more climb, not less — but it is 200 ft of needless climb that can put you into airspace you were not cleared for, and in an exam it is simply the wrong answer.

• On QFE, the reading is already a height above the datum, so correct it directly — nothing to subtract.
• On a regional or area QNH, or any setting whose source elevation you do not know, treat the source as mean sea level. The height to correct becomes the whole published altitude, which yields the larger correction and is the conservative choice.
• Never use height above the terrain. The terrain is what you are trying to miss, not the datum at which your altimeter was made truthful.

The 4% per 10°C Rule — and Where It Runs Out

The rule of thumb is: add 4% of the height above the setting source for every 10°C the temperature is below ISA, often written as 4 ft per 1000 ft per °C of ISA deviation. The temperature is the reported temperature at the station whose setting you have, and the deviation is worked out against ISA for that station’s elevation — not against +15°C, unless the station happens to be at sea level.

Take a sea-level aerodrome reporting −20°C. ISA there is +15°C, so the deviation is 35°C and 3.5 × 4% = 14%. Over 3400 ft that is 476 ft, against 480 ft from the ICAO table. The agreement is real because it is like for like: the table is itself computed for a station at sea level.

Now close the loop on the 1200 ft aerodrome of the previous section. ISA at 1200 ft is about +12.6°C, so a reported −20°C there is ISA−32.6, not ISA−35. That gives 3.26 × 4% = 13.05%, and 0.1305 × 3400 = about 444 ft — less than the table’s 480 ft, because the table is built for a sea-level station and is therefore deliberately generous at an elevated one. Fly the table value. The point of the arithmetic is that the deviation belongs to the source elevation, and quoting ISA−35 at a 1200 ft aerodrome is exactly the error the examiner is looking for.

Where the rule runs out

Knowing why 4% is the number tells you when it stops working. The exact ratio is the ISA deviation divided by the mean absolute temperature of the layer — the station value less half the layer’s own lapse. For −20°C over 3400 ft that mean is about 250 K, giving 35/250 = 14.0%. So the rule’s 0.4% per °C is really 1/250 per °C: it has a mean layer temperature of roughly 250 K, or −23°C, baked into it.

ICAO Doc 8168 places the limit explicitly: the 4% approximation is stated to be safe for all altimeter setting source elevations at temperatures above −15°C. Against the published table that produces a clear pattern:

• Down to about −15°C at the station the rule is generous or exact, over-correcting slightly at the warmer end — the safe direction.
• Below −15°C it turns non-conservative, and the shortfall grows: around 5% short at −30°C, 8% at −40°C, 13% at −50°C. At −50°C and 5000 ft above the source the rule gives 1300 ft where the table demands 1500 ft — 200 ft you do not have.

A second, quieter assumption is that the real atmosphere cools at the ISA lapse rate throughout, so the deviation measured at the station persists upwards. Usually that is kind to you: the conditions producing extreme surface cold — clear skies, light wind, snow cover, long nights — also produce a strong low-level inversion, the mechanism the PPL meteorology questions drill, in which the station is the coldest point in the column and the correction over-estimates. Watch for the reverse, a station in a mild coastal or föhn-warmed pocket beneath a much colder airmass aloft, where the station temperature understates the deviation over the high ground you need to clear.

The ICAO Cold Temperature Correction Table

The published table is in ICAO Doc 8168, PANS-OPS. Rows are the reported aerodrome temperature; columns are the height above the elevation of the altimeter setting source, in feet; the values are in feet and are always added to the published altitude or height. They are already rounded up to the next 10 ft.

Reading the table

Take the reported temperature, dropping to the colder row if you fall between two; work out the height above the setting source; interpolate linearly across the columns; add the result to the published figure and round up. Then look at the numbers. At −30°C a published 3000 ft above the aerodrome puts you 570 ft lower than you believe; at −40°C and 5000 ft the deficit is over 1200 ft, more than the terrain separation most minimum altitudes are built on.

The formula behind the table

Two properties are worth carrying:

• It is computed for a station at sea level. At a higher-elevation aerodrome the same reported temperature is a smaller ISA deviation — ISA at 4000 ft is about +7°C, not +15°C — so the tabulated value exceeds what is strictly needed. That is conservative, and you use it as published.
• Every value comes from one relationship, usable for heights or temperatures off the end of the table:
  Correction = H × (15 − t) ÷ (273 + t − 0.5 × L × H), where H is height above the source in feet, t the station temperature in °C and L = 0.00198°C per foot. Rounded up to the next 10 ft it reproduces the published table exactly.

When a correction must be applied

There is no single universal trigger temperature. PANS-OPS requires correction where the temperature is significantly below standard, and the actual threshold is set by the State: look for it in the AIP, in the GEN and AD sections, and on the approach chart itself. The benchmark most commonly published is a reported aerodrome temperature at or below 0°C for approach minima, with some States setting a separate and considerably colder threshold for en-route minima. Treat that as the shape of the answer rather than the answer: in the exam the correct response is that the trigger is State-published, not universal, and in the aeroplane it is whatever the AIP and your operator’s procedures say. The commercial-operations side of that sits in the CPL theory exams.

What is published and absolute is the temperature range for baro-VNAV on an RNP APCH. The approach chart carries both a minimum and a maximum aerodrome temperature for the barometric vertical path, and the same limits appear in the State AIP. Below the minimum the barometric vertical guidance is not authorised, because the path is computed from barometric altitude and in cold air the aeroplane flies a shallower, lower profile than the nominal one. Above the maximum the same mechanism runs the other way — the profile flown is steeper and higher than nominal — which is where warmer-than-ISA air, harmless to terrain, finally does matter.

Which Altitudes You Correct, and Who Tells ATC

The correction only helps if it goes on the right numbers. The working principle: if it is a barometric figure printed on a chart and you are relying on it for terrain or obstacle clearance, you correct it. Note too that the approach where this matters most — night, snow cover, a dark aerodrome in a dark valley — is also the one where visual illusions are working hardest against you.

The pilot corrects

Add the correction to:

• Minimum sector altitudes and minimum safe altitudes on the approach chart.
• En-route minima — area minimum altitude, MEA, MOCA, MORA and grid MORA.
• Minimum holding altitude.
• Procedure altitudes and heights: initial and intermediate approach altitudes, the final approach fix or final approach point altitude, and every step-down fix.
• DA/H and MDA/H, and missed approach altitudes including the level-off altitude.

Not corrected by you, or already accounted for

• Flight levels, for separation from other traffic. Every aircraft above the transition level shares the single 1013.25 datum and is wrong in the same direction, so the vertical order is preserved, the pressure separation ATC is working to still stands, and no aircraft closes on another. The true geometric gap does shrink, because the error grows with height above the datum and cold pressure surfaces lie closer together: at ISA−35, 1000 ft of pressure-altitude separation is only about 870 ft of real distance, some 13% less. That loss is accepted — it applies to everyone equally and the separation minima are built to absorb it. What is not absorbed, and what you must correct, is the distance between the lowest aircraft and the ground.
• ATC-assigned altitudes under radar vectoring. The service provider corrects its own minimum vectoring altitudes.
• The DH of a Cat II or Cat III ILS, which is a radio-altimeter height. Radio altitude is geometric and immune to all of this, as is GNSS geometric height.
• The ILS glide path itself — a radio beam, fixed in space, so the descent profile does not change with temperature. The barometric numbers on the same chart, the platform altitude, the altitude check at the fix and the DA, are still wrong in cold air, and the most exposed part is the level segment before glide path intercept.
• Anything already compensated. Where the FMS applies temperature compensation, or a State publishes pre-corrected charted altitudes, do not correct twice.

Two errors of direction to finish. Never subtract in warm air: a minimum is a minimum, and while warmer-than-ISA air leaves you higher than indicated, that gives no authority to descend below a published figure. And density altitude is not a terrain correction — it tells you what your take-off run and climb gradient will do, and is never applied to a minimum altitude.

Low QNH and the Lowest Usable Flight Level

Above the transition altitude everyone sets 1013.25. When QNH is below 1013 you have set a datum pressure higher than the one that actually exists at sea level, so the altimeter counts up from a fictitious level beneath the real one and reads high. You are lower than your flight level says, by (1013 − QNH) hectopascals converted into feet. Cold air does the same again from a different cause, and both must be paid for before you choose a level.

1. Start with the minimum altitude AMSL you must clear — MEA, MORA or area minimum altitude.
2. Add the cold temperature correction, giving the indication you would need on QNH.
3. Add (1013 − QNH) × 30 ft to convert that indication into a pressure altitude. Thirty feet is the conservative figure and the one to use when the question does not specify: 27 ft is the ISA sea-level value and slightly understates the offset, because feet per hectopascal grows with height. At QNH 983 the 27 ft figure gives 810 ft, where the 983 hPa surface in ISA actually sits 836 ft below the 1013.25 surface.
4. Take the lowest flight level from the applicable table of cruising levels at or above that pressure altitude.
5. Apply the direction-of-flight rule, which will often push you higher again.

Worked example: 5000 ft minimum, QNH 983, −20°C

Minimum altitude 5000 ft AMSL, regional QNH 983 hPa, reported temperature −20°C at a station near sea level.

• Temperature. 5000 ft above the source at −20°C gives 710 ft, so you need 5710 ft indicated on QNH.
• Pressure. 1013 − 983 = 30 hPa. At 30 ft per hPa that is 900 ft, so 5710 + 900 = 6610 ft of pressure altitude. At 27 ft per hPa it is 810 ft and 6520 ft; the exact ISA value, 836 ft, lands between them at 6546 ft.
• Level. FL60 is only 6000 ft of pressure altitude — between 520 and 610 ft short, whichever conversion you use. The lowest usable level is FL70.

Note what was never a candidate: FL65. In the ICAO table of cruising levels the levels ending in 5 — FL35, FL45, FL55, FL65, FL75 — are VFR levels. An IFR flight is choosing between FL50, FL60 and FL70, so even had the arithmetic landed at 6010 ft the answer would still have been FL70. Step 4 is a lookup in the right half of the table, not a round-up to the nearest 500 ft.

A minimum altitude of 5000 ft requiring FL70 is worth sitting with. Omit either correction and FL60 looks acceptable — but at FL60 in these conditions your true altitude is about 4540 ft on the 27 ft conversion, some 460 ft below the minimum. Omit both and FL50 looks fine, at roughly 3660 ft true: about 1340 ft low, over terrain that was surveyed to need 5000. Check the sign every time: QNH below 1013 means the flight level is higher than your true altitude, so the lowest usable level moves up.

Transition level

The same arithmetic sets the transition level. With a transition altitude of 5000 ft and QNH 983, that altitude sits at 5810 ft of pressure altitude on the 27 ft conversion and 5900 ft on the 30 ft figure, so the lowest flight level clear of it is FL60 either way — and where the State prescribes a minimum 1000 ft transition layer, FL70. A falling QNH pushes the transition level up, which is why it is issued to you rather than assumed.

What the IR papers ask

Expect the direction of the error set as a trap; the height the correction is applied to; a table lookup with interpolation; which altitudes the pilot corrects and which ATC does; a lowest-usable-flight-level calculation; and the 27 or 30 ft per hectopascal conversion. Every one of them is arithmetic on a datum, and the datum is the half of the question candidates get wrong.