How Thick Is The Earth S Mantle

So You Want To Know The Thickness Of The Earth's Mantle

The average is roughly 2,900 kilometers, though that single number obscures a lot of variation depending on where you measure. If you're a student and that's the answer your textbook wants, you'll pass the test. If you're actually trying to understand what that number means for anything practical, it's worth looking at how we get there and why the real value shifts across different regions. The mantle extends from the Mohorovičić discontinuity — commonly called the Moho — down to the core-mantle boundary. At a global average, that's about 2,900 kilometers of rock between the crust and the outer core. The crust itself is the variable piece. Oceanic crust sits at roughly 5 to 10 kilometers thick, while continental crust ranges from about 30 to 90 kilometers. This means the Moho depth isn't fixed. Under the Himalayas, for instance, it dips to around 75 kilometers. Under the Pacific Ocean, it can be as shallow as 7 kilometers. So the mantle's actual thickness at any given point is 2,900 kilometers minus however thick the crust happens to be at that location. Subdividing the mantle is where things get messier. Seismologists traditionally split it into three zones based on phase transitions in mantle minerals. The uppermost mantle, below the crust, goes down to about 410 kilometers. At that depth, olivine transforms into wadsleyite, a denser crystal structure, and this discontinuity is relatively sharp in most places. Between 410 and 660 kilometers lies the mantle transition zone, where ringwoodite becomes the stable phase. Then from 660 kilometers down to the core-mantle boundary at 2,890 kilometers, you have the lower mantle, dominated by bridgmanite and ferropericlase. This 660-kilometer boundary is the deep marker most people reference, though it's not a hard wall — it's a chemical and physical transition that behaves differently depending on local temperature and composition.

How We Measure It

Seismic tomography is the primary tool. When earthquakes happen, the waves they generate travel through the Earth and are recorded by seismometer networks around the world. P-waves and S-waves change speed depending on the material they pass through. By analyzing arrival times, deflections, and reflections across thousands of seismic events, researchers reconstruct velocity models of the interior. The discontinuities at 410 and 660 kilometers were identified this way — they show up as clear changes in wave velocity because the mineral phase transitions absorb and refract seismic energy. The core-mantle boundary was mapped using a combination of reflected and diffracted seismic phases, particularly PKP and PKiKP waves that interact with the boundary between the silicate mantle and the liquid iron outer core. The density contrast here is enormous, which makes it a very distinct seismic signal. Modern models like PREM (Preliminary Reference Earth Model) give us a one-dimensional average profile of the Earth, and from PREM, the mantle thickness comes out to about 2,891 kilometers from the base of the crust to the core-mantle boundary. But PREM is an average. The real Earth is nowhere near that uniform. Lateral variations in seismic velocity across the mantle can be several percent, which translates into significant depth variations in the discontinuity boundaries themselves. The 660-kilometer discontinuity, for example, can be depressed by more than 100 kilometers beneath cold subducting slabs and elevated by similar amounts beneath hot mantle plumes. This means the "660 kilometer mark" is not a flat surface. It's a wavy, complex boundary shaped by convection patterns that have been operating for billions of years.

What Happens When You Try To Use This In Practice

I spent several months building a geodynamic model that required precise boundary conditions at the 660-kilometer discontinuity. The standard assumption in most modeling software is a flat, uniform 660 km depth everywhere. That works fine if you're doing a first-order simulation, but it breaks down when you're trying to model how a subducting slab interacts with the transition zone in a specific region. The slab I was looking at — part of a western Pacific subduction system — was penetrating the 660-kilometer boundary, and the timing and dynamics of that penetration depend heavily on the local depth of the discontinuity. If your model assumes 660 km but the real value is 750 km due to the cold slab already depressing it, your simulation will predict the wrong behavior entirely. The workaround is to use a regional seismic dataset to define a depth-corrected boundary rather than the global average. I pulled teleseismic receiver function data from nearby stations, calculated the local 660-kilometer depth from the delay times of converted phases, and imported that corrected depth into the model geometry. It added maybe two days of work and completely changed the results. A flat-boundary model showed the slab floating above the 660-km discontinuity for millions of years. The corrected model showed it penetrating and stagnating at a shallower effective depth, which matched what the regional seismology literature was reporting. The difference mattered.

Common Misunderstandings

People often treat the mantle as a single homogeneous layer. It's not. The rheology changes dramatically with depth. The upper mantle, including the lithosphere and the asthenosphere beneath it, behaves quite differently from the lower mantle. The asthenosphere is weaker and partially molten in places, which is why it flows and allows tectonic plates to move. The lower mantle is under such extreme pressure that even though temperatures are high, the material behaves more like a solid on short timescales. It flows, but over millions of years, not seconds. Another misconception is that the 410 and 660 kilometer discontinuities are sharp chemical boundaries. They're not. They're phase transitions — the same rock, just rearranged into denser crystal structures under pressure. The compositional chemistry of the mantle below the crust is largely uniform in terms of major elements. The differences are in how those elements are packaged at different pressures and temperatures. There's also a persistent myth that the mantle is mostly molten. It's not. The vast majority of the mantle is solid. The only significant melt in the mantle exists in localized zones — the asthenosphere beneath mid-ocean ridges, and the sources of mantle plumes. Magma genesis in the mantle happens primarily through decompression melting, where hot rock rises and crosses its solidus without additional heat input. The mantle itself is not a ocean of magma. That image belongs to science fiction, not geophysics.

The Limitations

Seismic resolution is the main constraint. Even with dense station networks, the spatial resolution degrades with depth. At 660 kilometers, you're getting decent resolution under well-surveyed regions like North America and Europe, but in the Pacific Ocean, where most of the Earth's surface sits, the resolution is coarser. Receiver function studies can resolve individual discontinuity depths to within about 5 to 10 kilometers under good conditions, but that requires a network of seismic stations spaced closely enough together. Remote oceanic regions still rely on sparse data or shipborne deployments that cover limited areas for short periods. The 2,900-kilometer figure you'll find everywhere is a global average derived from PREM, which smooths over all lateral variation. If you need accuracy for a specific location, you should not use that number directly. You should look up regional seismic studies, check whether the local crustal thickness is known, and adjust accordingly. A rough rule of thumb is that the mantle thickness varies by roughly plus or minus 50 kilometers globally due to crustal thickness differences alone, and an additional plus or minus 100 kilometers or so due to dynamic variations in the transition zone and lower mantle boundaries. If you're doing something that requires high precision — a tomography study, a geodynamic model, a plate reconstruction — you'll want to use newer models beyond PREM. Models like SAS4A and SAW12A incorporate more recent seismic data and provide higher-resolution laterally varying structures. The difference in mantle thickness estimates between these models and PREM can be significant at regional scales, though the global averages remain similar.

The Bottom Line

The Earth's mantle is approximately 2,900 kilometers thick on average, measured from the base of the crust to the core-mantle boundary. The exact value at any point depends on local crustal thickness, which varies from about 5 kilometers under oceans to 90 kilometers under major mountain belts. The internal structure is divided into the upper mantle, the transition zone between 410 and 660 kilometers, and the lower mantle from 660 kilometers down to 2,890 kilometers. These boundaries are mineralogical phase transitions, not chemical interfaces, and they shift in response to temperature and flow patterns within the mantle itself. For most purposes, the average figure is sufficient. For anything requiring regional accuracy, you need regional data.