What this converts
A runner living in Flagstaff at 2,106 m whose marathon pace is 5:30/km would run about 5:13/km at sea level for the same effort: 17 seconds per kilometre faster, or 5.3%, with a likely range of 5:08 to 5:18.
The trip is not symmetric: sea level to Flagstaff on arrival day costs a marathon effort 11.2%, while the return as a resident gives back 5.3%, because a resident has already absorbed half of what a visitor is about to feel.
Getting the direction and the settings right
Enter the pace you actually run today, at the place you actually run it: if you train in Boulder and race in Chicago, Boulder is the From side. That side assumes you are adapted to where you train, which is right for almost everyone. Change it in the advanced panel only if the pace you typed is one you ran while visiting somewhere higher than home.
- Race start lines are their own entries. A point-to-point course is converted at its start, not at the town it finishes in: the REVEL Mt Charleston start sits at 2,316 m, Las Vegas below it at 610 m.
- The range is the answer. Two runners at the same elevation can differ by more than the penalty itself, so open at its slow end.
What altitude costs per kilometre, by distance
Each column uses its own reference pace: 4:00/km for the 5K, 4:30 for the 10K, 5:00 for the half, 5:30 for the marathon. The figures are for a sea-level runner on arrival day.
| Elevation | 5K, from 4:00/km | 10K, from 4:30/km | Half, from 5:00/km | Marathon, from 5:30/km |
|---|---|---|---|---|
| 500 m | +2 s/km (1.0%) | +3 s/km (1.1%) | +4 s/km (1.2%) | +4 s/km (1.2%) |
| 1,000 m | +9 s/km (3.6%) | +11 s/km (4.1%) | +13 s/km (4.3%) | +14 s/km (4.3%) |
| 1,500 m | +15 s/km (6.2%) | +19 s/km (7.0%) | +22 s/km (7.4%) | +25 s/km (7.5%) |
| 2,000 m | +21 s/km (8.9%) | +27 s/km (10.0%) | +32 s/km (10.6%) | +35 s/km (10.6%) |
| 2,500 m | +28 s/km (11.5%) | +35 s/km (12.9%) | +41 s/km (13.7%) | +45 s/km (13.7%) |
| 3,000 m | +34 s/km (14.1%) | +43 s/km (15.9%) | +50 s/km (16.8%) | +56 s/km (16.8%) |
The penalty grows with distance because the longer the race, the more of it your aerobic system carries alone. A review of four decades of altitude research found the impairment of submaximal exercise proportional to both the elevation and the duration of the effort (Fulco 1998), and across 132,104 elite track performances, events over 800 m lost 2 to 4% above 1,000 m while sprints above 1,500 m, the hurdles aside, ran 0.3 to 0.7% faster (Hamlin and colleagues, 2015).
A few weeks there cuts these to roughly 61% of the table, our model's estimate rather than a measured curve: at 2,000 m the marathon penalty drops from 35 seconds per kilometre to about 21.
Altitude training bases and what they cost
The bases in the search box, with the band the calculator assigns each one: low is 500 to 2,000 m, moderate is 2,000 to 3,000 m.
| Base | Elevation | Band | Why runners go |
|---|---|---|---|
| Mammoth Lakes, USA | 2,402 m | Moderate | Mammoth Track Club base, with lower ground near Bishop for faster work |
| Iten, Kenya | 2,400 m | Moderate | The Kamariny dirt track and the High Altitude Training Centre |
| Addis Ababa, Ethiopia | 2,355 m | Moderate | Ethiopia's centre of gravity; the Sululta camp belt above it sits at 2,613 m |
| Kaptagat, Kenya | 2,341 m | Moderate | Eliud Kipchoge's NN Running Team camp, on forest trails around 2,400 to 2,500 m |
| Sierra Nevada, Spain | 2,320 m | Moderate | Spain's high-altitude centre, with lower ground around Granada for faster sessions |
| Park City, USA | 2,134 m | Moderate | Utah Olympic Park campus, 40 minutes above Salt Lake City |
| Flagstaff, USA | 2,106 m | Moderate | HOKA NAZ Elite and NAU base; Lake Mary Road and Buffalo Park |
| Eldoret, Kenya | 2,095 m | Moderate | Rift Valley hub and airport, with camps along the Eldoret to Iten road |
| Kunming Haigeng, China | 1,892 m | Low | one of China's longest-running altitude camps, on the shore of Dianchi lake |
| Font-Romeu, France | 1,850 m | Low | French national altitude centre, built for the 1968 Mexico City Games |
| Colorado Springs, USA | 1,839 m | Low | United States Olympic and Paralympic Training Center, the original US campus |
| St. Moritz, Switzerland | 1,822 m | Low | Europe's classic camp: a 400 m track plus the Stazerwald forest loops |
| Livigno, Italy | 1,816 m | Low | Alpine valley with a synthetic track and heavy summer camp traffic |
| Boulder, USA | 1,655 m | Low | Dense professional-runner community, with Magnolia Road at about 2,600 m above town |
| Albuquerque, USA | 1,619 m | Low | Year-round dry altitude; the Sandia foothills add 300 to 600 m |
| Potchefstroom, South Africa | 1,350 m | Low | North-West University High Performance Institute, a southern-hemisphere winter camp |
The same box holds marathon host cities and race start lines too, so you can convert straight from the camp to the race you entered.
Converting workout paces, not just race times
Three rules cover almost every session:
- Easy and long runs. Keep the effort and accept the slower number.
- Intervals of five minutes or longer. Use the tempo or 5K row of the effort table in the result.
- Short reps. No adjustment at all, on Jack Daniels' own guidance.
A 4:30/km tempo at sea level becomes about 4:48/km at Flagstaff after three weeks there, 18 seconds per kilometre slower, range 4:42 to 4:53. The effort table splits the entered pace six ways, but the coefficients separating those rows are assumptions of this calculator, not measured values.
Heart rate will not tell you whether the altitude is biting. Maximum heart rate falls only about 1.7 bpm per 1,000 m in a pooled analysis of 86 studies, roughly 4 to 5 bpm at 2,500 m, less than the 5 to 8 bpm your submaximal heart rate drifts from day to day. Your pace moves several percent while your heart rate zones barely move, which is why chasing the usual pace at the usual heart rate comes apart. Convert the pace, then build the week around it in the training pace calculator.
Why altitude costs you pace
At a given effort your pace follows the oxygen you can actually use, and in trained runners that ceiling falls roughly 6.3% per 1,000 m. For a 6:00/km marathon effort the first 1,000 m of elevation costs about 16 seconds per kilometre: 6:16 at 1,000 m, 6:38 at 2,000 m, on the day you arrive.
Running economy does not change at altitude, so a given pace costs the same oxygen it costs at home, and the thinner air gives almost nothing back: air resistance is only about 2% of the energy cost at marathon speed, so at 2,100 m the density credit for a 6:00/km runner works out at 0.12% against an aerobic-ceiling drop of about 11%.
What the model cannot give you is your own number: two trained athletes at the same elevation can land on opposite sides of the average. That is why the result is a range.
The evidence behind each coefficient
- 6.3% per 1,000 m. Eight endurance-trained athletes ran to exhaustion at six simulated elevations from 300 to 2,800 m in a hypobaric chamber. VO2max fell linearly at 6.3% per 1,000 m, individual slopes 4.6 to 7.5%, and the drop was already measurable between 300 and 800 m (Wehrlin and Hallen 2006, doi 10.1007/s00421-005-0081-9). The exposure was acute.
- Economy unchanged. A pooled analysis of 153 subjects across four research centres concluded that "exercise economy remains unchanged after acclimatization to high altitude" (Lundby and colleagues, 2007).
- Air resistance. Wind-tunnel work put the energy cost of overcoming air resistance at 2% at marathon speed (5 m/s), 4% at middle-distance speed and 7.8% sprinting (Davies 1980). Pugh's earlier figures are about double these; we use the lower set, which leaves the conclusion that thin air does not rescue distance runners true either way.
- Individual spread. At 580 m, trained cyclists lost 6.8% of VO2max on average, but individuals ranged from 1.2% better to 12.3% worse (Gore 1996). In the only real running race in this evidence base, 27 elite runners were 48.5 seconds slower over 3,000 m at 2,100 m, 48 hours after arriving; those who desaturated most lost 54.0 s and those who desaturated least 38.9 s (Chapman 2011, doi 10.1249/mss.0b013e318211bf45). The sea-level baseline time is not in the published abstract and the full paper is paywalled, so that gap cannot honestly be turned into a percentage here.
- Acclimatisation. Staying longer improves what you can do, but it does not restore VO2max (Fulco, Rock and Cymerman 1998, PMID 9715971; Bärtsch and Saltin 2008). Our steps, 100%, 78%, 61% and 50% of the penalty remaining, are estimates taken from the performance column of a three-week study of elite cyclists at 2,340 m (Schuler 2007), with the resident value a judgement call.
- Recreational runners lose less. Twelve recreational distance runners showed no significant VO2max loss at 914 m and a clear one at 1,219 m (Squires and Buskirk 1982), while trained cyclists lost measurably at 580 m (Gore 1996). If you are not racing at an elite level, read these figures as an upper bound.
- Where the model stops. It is validated to 3,000 m and understates the loss above that. It models no heat, no humidity, no wind and no hills. The marathon output is the least trustworthy figure it produces, and the likely direction of the error is that it is too kind.
When to arrive
As early as you practically can. If that is only the evening before, that is genuinely fine.
The one study that tested this directly took 15 male adolescents to 1,700 m: a shuttle test came out about 37% below sea level six hours after arrival, 18% at 18 hours and 14% at 47 hours. It improved steadily, with no window worse than arrival day, and the authors concluded that "travel to moderate altitude should occur as early as is practical before competition" (Weston 2001).
Shaving it finer does not help: cyclists given only two hours of simulated 2,500 m before a time trial did no better than those who had spent the whole night in it, and "would not gain an advantage by delaying their arrival until a few hours before the competition" (Foss and Chapman 2017).
If you are going early, sleep at the altitude of the race rather than above it. Among 48 collegiate runners racing 3,000 m at 1,780 m, those living 300 to 1,000 m higher may run significantly slower on arrival, as the 2,454 m and 2,800 m groups did in that trial, and that arrangement may need up to 19 days to settle (Chapman 2016).
Not the same thing as elevation gain
This converter deals with barometric altitude: how much oxygen pressure the air holds where you are standing. It says nothing about whether the course goes up.
A flat 10K at 2,000 m and a sea-level 10K with 300 m of climbing are different problems, and only the first belongs here. For climbing, use the elevation gain and loss calculator, or the grade adjusted pace calculator.
Courses that do both are converted at their start line, where the air is thinnest and where the pacing decision gets made.