Isostasy & Mountain Roots

Why mountains have deep roots: Archimedes' buoyancy principle applied to floating continental crust.

Worked example — Isostasy & Mountain Roots

Airy (Roots)	49.55 km
Root depth (below T₀)	49.55
Root factor	5.600
Submerged fraction	84.85%
Total crust thickness	88.40
Effective column density	2,800
Post-glacial rebound uplift	7.507

These are the values the calculator shows when the page opens. Change any input above to recompute them.

Loading interactive simulation...

A mountain is mostly underground, though less of it than an iceberg 🖖

Crust floats on mantle by the same buoyancy equation that floats an iceberg on water — this is Airy isostasy: root = h · ρ_crust / (ρ_mantle − ρ_crust). With typical densities (crust 2800 kg/m³, mantle 3300 kg/m³), the root factor is 5.600. Every 1 km of mountain above sea level is supported by 5.6 km of deep crustal root below — putting 84.85% of the block underground, compared to 89.46% of an iceberg underwater. Because the density difference between crust and mantle is only about 15%, the buoyant margin is thin and demands an enormous root. Dragging the mantle density toward the crust density demonstrates a sharp hyperbola: as the denominator approaches zero, the required root runs away toward infinity.

Airy over-predicts Everest, and the error is the assumption talking 🖖

Applied to Mount Everest (elevation 8.848 km), Airy isostasy predicts a massive crustal root of 49.55 km below reference crust, implying a total crustal thickness of nearly 88.40 km. However, seismic refractions measured across the Tibetan Plateau reveal crustal thicknesses closer to 65–70 km. This discrepancy is not sloppy math; it highlights Airy's assumption of local compensation, where each crustal block floats independently like separate logs in a raft. In reality, the continental lithosphere possesses flexural rigidity — acting like a rigid elastic sheet that bends and distributes narrow mountain loads sideways across hundreds of kilometers.

The surface cannot tell you which subsurface model is right 🖖

Pratt isostasy offers an alternative model: identical surface topography with no crustal roots at all, supported instead by columns of lower density (2572 kg/m³ for Everest) extending to a uniform compensation depth of 100 km. Both models reproduce observed elevations at the surface equally well. Nothing measured on a topographic map alone can distinguish between Airy and Pratt compensation; resolving the subsurface structure requires gravity anomaly surveys and seismic velocity measurements. Underdetermination in geophysics is fundamental: both models remain in textbooks because Airy fits mountain belts well, while Pratt better describes oceanic lithosphere and mid-ocean ridges.

Problems solved in full

Erosion strips a kilometre off the top of Everest 6 steps
Erosion strips a kilometre off the top of Everest. How much of it grows back, and how much rock has to leave before the summit is genuinely a kilometre lower?
1. Airy compensation fixes the root by flotation: r = h ρ_c/(ρ_m − ρ_c) = 8.848 × 2800/500 = 49.55 km, the Root depth card. Per kilometre of elevation that is 5.600 km of root, which is the Root factor card.
  
  $r = h \frac{ρ _{c}}{ρ _{m} - ρ _{c}} = 8.848 \times \frac{2800}{500} = 49.55 km$
2. Erode a slab of thickness e off the top. The column is floating, so it rises until the mantle it displaces makes up the missing mass: ρ_m u = ρ_c e, giving u = e × 2800/3300 = 0.8485 e. Set the whole 8.848 km as the load and the Post-glacial rebound uplift card reads 7.507 km.
  
  $ρ_{m} u = ρ_{c} e ⟹ u = e \frac{ρ _{c}}{ρ _{m}} = 0.8485 e$
3. So the summit only drops by e − u = e (ρ_m − ρ_c)/ρ_m = 0.1515 e. The 84.85% that comes back is the same number the Submerged fraction card shows, and for the same reason: root/(root + h) reduces to ρ_c/ρ_m.
  
  $e - u = e \frac{ρ _{m} - ρ _{c}}{ρ _{m}} = 0.1515 e, \frac{r}{r + h} = \frac{ρ _{c}}{ρ _{m}} = 0.8485$
4. Invert it. Lowering the summit by one kilometre costs ρ_m/(ρ_m − ρ_c) = 3300/500 = 6.600 km of rock, and flattening all 8.848 km costs 8.848 × 6.600 = 58.397 km. That is the elevation plus the root, 8.848 + 49.55.
  
  $e_{tot} = h \frac{ρ _{m}}{ρ _{m} - ρ _{c}} = 8.848 \times 6.600 = 58.397 km = h + r$
5. Over that history the column rises by 58.397 × 0.8485 = 49.55 km, which is exactly the root it started with, and it ends at the 30 km reference thickness. Every kilometre of root is paid back as uplift.
  
  $u_{tot} = 58.397 \times 0.8485 = 49.55 km = r$
6. Which means the rock now lying at the flattened surface began 49.55 km below sea level, carrying 58.397 km of crust above it. Its pressure was ρ_c g d = 2800 × 9.80665 × 58,397 = 1.60 × 10⁹ Pa.
  
  $P = ρ_{c} g d = 2800 \times 9.80665 \times 58, 397 = 1.60 \times 1 0^{9} Pa$
Answer

Removing an 8.848 km mountain takes 58.4 km of rock, and the ground you are left standing on spent its early life at 1.60 GPa. Sixteen kilobars is granulite and eclogite country. It is why the eroded cores of old mountain belts, the Scottish Highlands and the Norwegian Caledonides among them, expose rock that crystallised deeper than the Kola borehole ever drilled at 12.262 km, and why that rock is at the surface at all: the mountain that buried it also lifted it back out. This page solves the balance once, for the column as it stands. Running it forward is the step it does not take.

$1.60 \times 1 0^{9} Pa = 1.60 GPa = 16.0 kbar$
A root of 168.00 km, and how many of those digits mean anything 6 steps
Load Thin density contrast: a 3 km rise, crust at 2800 kg/m³, mantle at 2850. The Root depth card reads 168.00 km and the total crust thickness 201.00. Work out how far the answer moves if the mantle density is 1% out, then decide how many figures of it you would quote.
1. The balance first. Every kilometre of elevation costs 56 km of root at this contrast, so 3 km of mountain sits on 168 km of it, and the column runs 201 km from summit to root base against a 30 km reference. Ninety-eight percent of the block is below the reference surface.
  
  $r = h \frac{ρ _{c}}{ρ _{m} - ρ _{c}} = 3 \times \frac{2800}{50} = 168.00 km, T_{0} + h + r = 201.00 km, \frac{r}{r + h} = 98.25%$
2. Take logarithms and differentiate, and the sensitivity comes out with a single coefficient in front of it — the same coefficient for both densities, with opposite signs.
  
  $ln r = ln h + ln ρ_{c} - ln Δ ρ ⟹ \frac{δ r}{r} = \frac{ρ _{m}}{Δ ρ} (\frac{δ ρ _{c}}{ρ _{c}} - \frac{δ ρ _{m}}{ρ _{m}})$
3. That coefficient is the whole story. At Everest’s contrast it is 6.6, so a 1% density error is a 7% root error. At this one it is 57.
  
  $\frac{ρ _{m}}{Δ ρ} : Everest \frac{3300}{500} = 6.60, here \frac{2850}{50} = 57.0$
4. Which is far enough out of the linear regime that the two directions no longer match. One percent off the mantle density, each way.
  
  $ρ_{m} = 2821.5 (- 1%) : r = 390.70 km (+ 132.6%); ρ_{m} = 2878.5 (+ 1%) : r = 107.01 km (- 36.3%)$
5. One percent off each in opposite directions ends the calculation outright: the crust is now denser than the mantle it is supposed to float on, the guard in the model returns nothing, and the card shows a dash.
  
  $ρ_{c} = 2828, ρ_{m} = 2821.5 (1% each, opposite ways) : ρ_{m} \leq ρ_{c} ⟹ the card reads —$
6. Everest is well behaved by comparison, and still not precise: the same one percent spans thirteen kilometres of root.
  
  $Everest, the same \pm 1% : 43.72 \leq r \leq 57.00 km about 49.55; \pm 1 km needs Δ ρ to \pm 10 kg m^{- 3}$
Answer

One figure. Possibly two. The card prints 168.00, and the input it is most sensitive to is a mantle density nobody has measured directly — 1% either way puts the root between 107 and 391 km. Two decimal places on that is a formatter’s habit rather than a result. Even Everest, at a comfortable contrast of 500 kg/m³, needs the contrast known to ±10 kg/m³ before its root can be quoted to the nearest kilometre.

There is a second reason to distrust this particular column, and the tool hands it to you: 201 km of crust to support a 3 km rise, where seismic refraction under Tibet finds 65 to 70 km beneath the highest ground on Earth. The formula answered the question it was asked. A contrast of 50 kg/m³ between continental crust and mantle is not a number anyone measures, and a sensitivity of 57 says so before any geology is consulted.

$\frac{δ r}{r} = \frac{ρ _{m}}{Δ ρ} δ_{rel} : 6.60 at Everest, 57.0 here$

References (2)

Airy and Pratt isostasy equations and standard crustal and mantle densities Turcotte, D. L. and Schubert, G. (2014). Geodynamics (3rd ed.). Cambridge University Press.
flexural rigidity of continental lithosphere and Tibetan crustal thickness measurements Watts, A. B. (2001). Isostasy and Flexure of the Lithosphere. Cambridge University Press.

By the 42Math team · Updated 2026-08-23

Reviewed against published sources. Corrections welcome. Editorial policy · Corrections

All Bio, Geo, Chem lessons →

Example problems

Iceberg (Ice in water) - An iceberg floats with 89.46% of its body submerged, requiring 8.491 km of draft per 1 km above water.
Mount Everest (Airy) - Airy predicts a 49.55 km root under Everest (8.848 km peak), with 84.85% submerged and 88.40 km total crust.
Everest (Pratt model) - Pratt reproduces Everest's 8.848 km peak with zero root by lowering column density to 2572 kg/m³ down to 100 km.
Thin density contrast - When mantle density (2850 kg/m³) approaches crust density (2800 kg/m³), a 3 km peak demands a 168.00 km root.