You check the app over coffee and there it is: 91 percent, logged at 3am. Or you land in Denver and your readings sit several points below the range you have watched for months.

These are the two most common oxygen surprises, they have completely different causes, and in both cases the number is behaving exactly as it should.

Key takeaways

• On a real commercial flight, healthy adults' saturation fell from 97 percent to 92 percent at cruising altitude (Kelly et al. 2006, Aviation, Space and Environmental Medicine).

• Cabins are pressurised to a maximum equivalent of about 2,440 m, and long-haul passengers showed a 3 to 4 percent decline (Geertsema et al. 2008, BJSM).

• Low oxygen genuinely begins to appear between 2,000 and 3,000 m of elevation (StatPearls).

• Overnight readings are the least reliable your device produces, and its error grows exactly where the numbers look most alarming.

Why do overnight readings run lower?

Two reasons, and they are not equally important.

The real one: breathing naturally becomes slower and shallower during sleep, particularly in REM, when the muscles that support your airway relax. Saturation running slightly below your daytime values overnight is expected in everyone.

The bigger one, for most alarming numbers: the measurement is at its worst overnight. An overnight value comes from a sensor pressed against a moving, sleeping body. An arm tucked under a pillow, a bent wrist, pressure from the mattress, a band that slipped down in the night, cold skin at 4am. Each of those produces low outliers that have nothing to do with your oxygen.

That would matter less if devices were accurate at the low end. They are not. In controlled testing against arterial blood, a smartwatch fell within 2 percent of the true value only 32 percent of the time, and its error grew from there to 5.82 percent at saturations below 88 percent (Jiang et al. 2026, JMIR Formative Research). The lower the reading looks, the less you should trust it. We go into this further in what is a normal SpO2 reading.

Is a single overnight dip worth worrying about?

Almost never. The useful question is not "did I dip" but "is there a pattern."

Nearly everyone shows occasional overnight dips. One low blip on one night, in a metric measured this awkwardly by a device this imprecise, is close to meaningless.

What carries information:

• Your own typical overnight range, built from many nights, rather than any single value.

• A night's average, which tells you far more than its single lowest minute. That lowest minute is often just the moment you rolled onto the sensor.

• Repetition. The same kind of dip appearing night after night is a genuine pattern.

This is the same logic that makes walking heart rate readable despite a noisy sensor: consistent error mostly cancels across a trend and is actively misleading in a snapshot.

What happens to my oxygen on a plane?

It drops by about five points, and this is designed behaviour rather than a fault.

Aircraft cabins are not pressurised to sea level. They are held at a maximum equivalent of roughly 2,440 m, around 8,000 feet.

Two studies measured what that does to passengers:

A simulator study running 36 eight-hour flights at ground, 4,000, 6,000 and 8,000 feet found saturation varied as expected, lowest at 8,000 feet and lowest of all in passengers with existing heart or lung conditions (Aerospace Medicine and Human Performance 2017).

So if you spot-check mid-flight, expect low 90s and read it as the aircraft, not yourself. It resolves shortly after landing.

Why do readings drop at altitude?

Because there is genuinely less oxygen available per breath.

Air pressure falls as you climb, so each breath delivers less oxygen and your blood picks up less of it. In most people, hypoxemia begins to appear between 2,000 and 3,000 m (StatPearls). Denver sits at about 1,600 m, which is enough to shift a baseline by a few points without being anywhere near a problem.

The higher you go, the larger the effect, and it is not linear. Somewhere above 3,000 m, desaturation becomes essentially unavoidable regardless of fitness.

Then your body adapts. Over the first days, breathing quietly deepens and quickens. Over weeks, you produce more red blood cells to carry oxygen. Readings drift back toward your usual range, which is why the interesting comparison at altitude is mountain day three against mountain day one, not against your sea-level baseline.

The standard acclimatisation advice is unglamorous and holds up: keep the first day or two easy, drink more than usual, add exertion gradually, and sleep generously. Sleep itself often feels lighter for the first nights up high, which is part of the same adjustment.

If you live at elevation, none of this applies as a deviation. Your everyday numbers simply are your baseline, and the only fair comparison is against your own history.

When is a pattern worth raising?

Some overnight patterns deserve more than a shrug, and this is where the metric earns its place.

Consistently low overnight readings across many nights, especially alongside loud snoring, gasping or choking awakenings, morning headaches or heavy daytime sleepiness, form a concrete story worth bringing to a clinician. None of that is a diagnosis, and no consumer device diagnoses anything. But a trend screen makes it possible to show rather than describe from memory, which is genuinely more useful in an appointment than trying to recall how last month felt.

Given how much devices overestimate, the reverse caution matters too: a normal-looking overnight average does not rule out a problem, and symptoms outrank a reassuring percentage.

Frequently asked questions

Is it normal for oxygen levels to drop while sleeping?

Yes. Breathing slows and becomes shallower during sleep, particularly in REM, so overnight saturation running slightly below daytime values is expected. Most dramatic single dips are measurement artifacts from arm position, a loose band or pressure on the sensor rather than real changes in your oxygen.

Why is my SpO2 low at night on my watch?

Usually the measurement. Overnight is the hardest condition your device faces, and its accuracy is worst at exactly the low values that look alarming: error reached 5.82 percent below 88 percent saturation in controlled testing (Jiang et al. 2026). Wear the band snug, a finger-width above the wrist bone, and judge the week rather than the minute.

How much does oxygen saturation drop on a plane?

About five percentage points. Healthy adults went from 97 percent before flight to 92 percent at cruising altitude (Kelly et al. 2006), and long-haul passengers showed a 3 to 4 percent decline (Geertsema et al. 2008). Cabins are pressurised to a maximum equivalent of roughly 2,440 m, so this is normal and temporary.

What is a normal SpO2 at altitude?

Lower than your sea-level baseline, by an amount that grows with elevation. Low oxygen genuinely begins between 2,000 and 3,000 m (StatPearls). There is no single correct number, which is why the useful comparison at altitude is against your own readings on previous days there, watching for the upward drift that means you are acclimatising.

When should I be concerned about overnight oxygen dips?

When they are consistent rather than occasional, and especially when they come with loud snoring, gasping awakenings, morning headaches or daytime sleepiness. That combination is worth showing a clinician. A single low night, or a scattering of outliers, is almost always the measurement.

Related reading

The other overnight signal worth knowing is breathing rate, which is steadier than oxygen and measured rather differently. See your breathing rate while you sleep.

Knit shows patterns, not diagnoses. Talk to a clinician about readings that concern you.