How to Train for a Mountain Ultra When You Live Somewhere Flat

Flat-terrain runners usually fail mountain ultras in the same place: somewhere after halfway, on a descent, with a heart rate that looks fine and quads that have stopped answering. The cardiovascular system was never the problem. The training block bought aerobic fitness it already had and skipped the thing the race actually charges for, which is thousands of metres of vertical loading in both directions.

You cannot manufacture a mountain in Rotterdam or Chicago. You can, however, work out exactly how much vertical your race demands, and then pay off as much of it as your terrain allows while knowing precisely which part of the bill you are leaving unpaid. That is a much better position than hoping volume covers it.

Start with the number your race actually charges

Distance alone tells you almost nothing about a mountain race. The international shorthand for this is km-effort, and ITRA’s public formula is simple:

km-effort = distance (km) + elevation gain (m) / 100

That is the formula behind our ITRA index calculator, and it reframes races immediately:

Race	Distance	Gain	km-effort
OCC	60 km	3,500 m	95
CCC	101 km	6,100 m	162
Western States	161 km	5,600 m	217
UTMB	171 km	10,000 m	271

Read that table twice if you are a road runner. OCC is a 60 km race that costs about as much as a flat 95 km. Western States, famous as a “runnable” hundred, still bills at 217. And UTMB at 271 is not “a bit more than Western States” - it is a different category of event, which is exactly why UTMB’s own qualifying tiers are keyed to km-effort rather than distance.

One honest caveat: ITRA’s public formula counts gain only. Descent is not free, and TrailMath’s own internal kmEffort adds a loss term at a gentler divisor of 150. We keep the two separate on purpose, because the ITRA number is the one race organisers and UTMB categories are defined in, and quietly redefining a public standard would make our numbers incomparable with everyone else’s.

From race cost to training target

Knowing your race is 271 km-e does not tell you what to do on Tuesday. The bridge is a coverage target: how much climbing the whole block should contain, not any single week.

TrailMath’s coverage model uses 3x the race’s elevation gain across the training block as the climbing target. For OCC that is 10,500 m. For CCC, 18,300 m. For UTMB, 30,000 m.

That multiplier is a planning heuristic, not a finding from a controlled trial - we are labelling it as such rather than dressing it up. What it is good for is converting a vague anxiety (“I should do more hills”) into a number you can divide by the weeks you have and then actually check against. Climbing is one of six weighted components in that model, worth 20% alongside volume (20%), long run (20%), intensity (15%), consistency (15%) and descent preparation (10%).

The long run target in the same model comes from firmer ground: 70% of estimated race duration, following Jason Koop’s argument in Training Essentials for Ultrarunning that the longest session should rehearse the race’s time-on-feet rather than its distance. For a 30-hour race that is a very different session than “run 32 km.”

Why flat volume does not quietly substitute

The intuitive fix is more mileage. It does not work, for two separate reasons.

The metabolic cost curve is steep and non-linear

Minetti and colleagues (2002) measured the energy cost of locomotion across gradients and produced the polynomial our incline calculator implements directly. Run through it, the numbers are stark:

Grade	Energy cost vs flat
5%	1.30x
10%	1.66x
15%	2.06x
20%	2.50x
25%	2.98x
30%	3.49x

A 20% pitch costs two and a half times what flat ground costs at the same speed. There is no volume of flat running that rehearses that, because the limiter changes: on a sustained steep climb you are no longer bounded by aerobic capacity but by the local muscular endurance of hip extensors and calves, which is the central argument House, Johnston and Jornet make in Training for the Uphill Athlete (2019).

Note the range on that table. Minetti’s data is valid from -25% to +30%, and plenty of alpine trails exceed +30%. Beyond that boundary the model is extrapolating, and so is anyone quoting it - including us.

Descent damages tissue that flat running never loads

Uphill is metabolically expensive. Downhill is mechanically destructive, and the two are not interchangeable. Eccentric contractions - muscle lengthening under load, which is what quadriceps do on every descending stride - produce far more muscle damage and delayed soreness than concentric work at the same intensity. Guillaume Millet’s research on neuromuscular fatigue in ultra-endurance events consistently finds that force-generating capacity after long mountain races is degraded well beyond what metabolic fatigue explains.

This is why our load model scores the two directions separately rather than summing “elevation.” Uphill contributes a bonus of gain / 300 * 0.10 to a session’s load score; downhill contributes loss / 300 * 0.08. Descent scores lower per metre but it is not zero, because the cost lands on your tissue rather than your lungs.

Flat runners do not blow up on the climbs. They blow up on the descent after the climbs, which is a different training problem entirely.

The three substitutes, honestly ranked

When an athlete has no access to sustained climbs, TrailMath substitutes hill sessions automatically, in a fixed priority order. Here is what each one is worth and what it actually trains.

1. Treadmill incline - the best available proxy

A hill session on a treadmill stays a hill session: TrailMath keeps the prescribed intensity and adds 10% to the duration. Credited at 8 metres per minute toward the climbing component, it is by a wide margin the most efficient substitute available.

Its limits are real, though. Most commercial treadmills stop at 12-15%, which puts you at roughly 1.8x to 2.1x flat cost - well short of the 3.0x-plus of a genuine alpine pitch. And a belt does not require the balance, foot placement or stride-length variation of a rocky climb.

2. Stair repeats - high specificity, different rhythm

Stairwells, stadium steps and bleachers deliver genuine vertical. TrailMath converts a hill session to stair climbing at Tempo intensity, same duration, credited at 6 metres per minute.

Stairs are excellent for the muscular endurance side and, unlike a treadmill, the descent is real - though repeated concrete descents concentrate the eccentric load in a way that a trail does not, so this is a session to build into rather than jump into.

3. Weighted flat tempo - the fallback that keeps the block honest

With no treadmill and no stairs, the substitution becomes a weighted flat tempo at 15% extra duration, credited at 2 metres per minute. That credit rate is deliberately low. It is a real session and it is much better than deleting the workout, but the model does not pretend it is climbing.

The gap you cannot close, and what to do about it

Every substitute above trains ascent. None of them trains sustained descent, and that is the honest limit of flat-terrain preparation for a mountain race.

Descent preparation carries 10% weight in the coverage model precisely because it is both important and frequently at zero. If you live somewhere flat, assume it is at zero for you and plan around that fact rather than discovering it at 60 km:

Spend your travel budget on descent, not ascent. If you can reach real terrain a handful of times before the race, use those sessions for long descending, not for climbing. Climbing is the part you can approximate at home.
Use overpasses, levees, parking structures and dune paths. A 40 m descent repeated is not a 1,200 m alpine plunge, but repeated eccentric loading beats none.
Add eccentric strength work. Slow controlled step-downs and split squats load the same tissue in the same direction. TrailMath biases the strength pool toward eccentric exercises when descent is set as a secondary focus.
Pace the first descent of the race deliberately. This is a race-strategy fix for a training gap, and it is a legitimate one. Quads you have not prepared will not survive being spent early.

Putting it together

A workable flat-terrain block looks like this:

Compute your race’s km-effort and note its total gain. The ITRA index calculator does the first part; the race effort estimator puts it in context against your own performance.
Multiply gain by three and divide across your available weeks to get a rough weekly vertical target.
Pick your highest-tier substitute and use it for the block’s designated hill session - treadmill first, then stairs, then weighted flat.
Grow at 10% duration and 15% elevation, maximum, with a deload every fourth week.
Protect one long run and grow it toward 70% of your estimated race duration in time, not distance.
Spend real terrain on descent. Every trip to the mountains should end with sore quads, not lungs.

None of this makes a flatlander’s preparation equal to someone who lives at the foot of the Alps. It makes the difference measurable and mostly closable, and it puts the residual gap somewhere you can plan around instead of somewhere that surprises you at kilometre 60.