Your pace lies in the heat and at altitude — here's the correction

Pace is an output, not an input#

The number on the screen is the result of an effort, not a description of one. Between two runs at 5:00/km, the difference in what your body actually did can be enormous — and every time you compare those two runs as equals, you draw a wrong conclusion about your fitness.

This matters twice. Once in the moment, when you push to hold a pace that today costs far more than it usually does. And once afterwards, when “I was slower than last month” becomes a belief about your training rather than a fact about the weather.

Three things break the pace-to-effort relationship more than anything else: heat, altitude, and terrain.

Heat#

The mechanism is competition for blood. Working muscle needs blood for oxygen; skin needs blood for cooling. In the heat both demands rise and the heart resolves the conflict the only way it can — by beating faster for the same output. Add fluid loss through sweat, which reduces plasma volume, and stroke volume falls too, so heart rate climbs further through the session.

That is cardiac drift: the same pace, a heart rate that keeps rising. On a long run in heat, a drift of 10% or more between the first and second half is unremarkable.

What it means practically:

Pace at a given effort falls. Reported penalties vary widely with humidity, acclimatisation and event duration, but the direction is unambiguous and the size grows with both temperature and duration.
Humidity matters more than temperature. Sweat that does not evaporate does not cool you. 28 °C at 80% humidity is a harder day than 33 °C in dry air.
Acclimatisation is real and fast. Most of the adaptation — plasma volume expansion, earlier and more dilute sweating — arrives within one to two weeks of regular heat exposure, and is lost about as quickly.

The practical rule in heat is to run to effort or heart rate, not to pace, and to judge the session afterwards on adjusted numbers rather than raw ones.

Altitude#

Different mechanism, same consequence. Air pressure falls with elevation, so each breath delivers less oxygen, arterial oxygen saturation drops, and the amount of work you can sustain aerobically drops with it.

The effect is small at 1,000 m, clearly noticeable by 1,500–2,000 m, and severe above that — and it is duration-dependent: a short sprint barely cares, while anything above about two minutes cares a great deal, because the longer the effort the more of it is aerobic.

Two things routinely surprise people arriving from sea level:

Easy pace slows more than you expect, and that is correct. Easy is defined by effort, not by a number you brought with you from home.
The first two or three days are often the worst, and partial acclimatisation takes one to two weeks — with the deeper haematological adaptations taking considerably longer.

Terrain#

The most common of the three and the easiest to correct. Running uphill at 5:30/km is harder than running on the flat at 5:30/km; running downhill is cheaper metabolically, but not proportionally — the saving going down is much smaller than the cost going up, and steep descents add mechanical damage that a metabolic model does not see at all.

Grade-adjusted pace is the correction: an estimate of the flat-ground pace that would have cost the same metabolically. On a rolling route it turns an uninterpretable set of splits into one comparable number. It is a model, not a measurement, and it is at its weakest on steep descents, on technical trails and on soft ground — where the mechanical cost dominates and the gradient no longer predicts it.

Why comparing raw pace across days is worse than useless#

Here is the failure mode in full. You run your usual loop in August at 5:20/km, having run it in April at 5:00/km. You conclude the block is not working. You add intensity. You are now training harder in conditions that already cost more, at exactly the point in the season when accumulated fatigue is highest.

Nothing about that chain of reasoning is unusual, and every step of it follows from treating pace as if it were effort.

The corrections exist so that the comparison you actually want — am I fitter than I was — is made between comparable numbers. Adjust for gradient, normalise for heat and altitude, and the August run may turn out to be the better one.

The honest limits of any correction#

Adjusted pace is an estimate produced by a model, and models have edges:

They are population averages applied to one person. Heat tolerance in particular varies hugely between individuals and within one individual across a season.
They cannot see what you did not record. Wind is rarely in the data at all, and it can dwarf every other factor on an exposed course.
They handle steep and technical ground poorly, for the reasons above.
They do not know your acclimatisation state. The same 30 °C day is a different event for someone in week three of a hot summer and someone who flew in yesterday.

So treat corrected pace as a better comparison than raw pace, not as truth. If the adjusted number disagrees violently with how the session felt, the session is usually right.

The second read: decoupling#

There is one measure that partly sidesteps the whole problem. Decoupling compares the ratio of output to heart rate in the first half of a steady session against the second half. If pace held while heart rate climbed 8%, the session cost you more than it looked like it did.

Decoupling is useful in exactly the conditions where pace is least trustworthy, because it is internal: it asks what the effort did to you rather than what the watch recorded. Rising decoupling across a heat block is a familiar pattern, and falling decoupling on similar days is one of the cleaner signs that acclimatisation is happening.

What to do with all this#

Run hard days to effort, not to pace, whenever conditions are not ordinary.
Do not chase a paced target in the heat. The pace will come back in September.
Compare adjusted numbers, not raw ones, when you want to know whether the block is working.
Expect two weeks of feeling bad after arriving somewhere hot or high, and plan the block around it rather than fighting it.
Watch decoupling as the second opinion, especially on long steady sessions.

Shindo computes grade-adjusted pace and normalises for heat and altitude, and reports decoupling on steady sessions, so comparisons across different days are made between comparable numbers. It does this to sessions that have already arrived through Apple Health — it does not record workouts, there is no GPS tracking and nothing runs during your session. Shindo runs on an iPhone 11 or newer. The methodology documents the thresholds and adjustments.