Skip to main content
earthscan.DEEP EARTH OBSERVATORY
INTERACTIVE RESEARCH
PHYSICS-INFORMED. AI-POWERED.

The Earth, from within.

A living atlas of structure, signals and the science beneath our feet.

FEATURED WHITEPAPERHow the deep Earth movesFlow → fabric → anisotropy → the signal
PLANETARY CROSS-SECTIONEARTH / 01
Analytical demonstration
6,371 km

MEAN EARTH RADIUS

Crust thickness exaggerated
to reveal the boundary.
A WORLD UNDER THE SURFACE
Click the Earth to extract a section
CrustMantleOuter coreInner core
DRAG TO ORBIT · SCROLL TO ZOOM
SCHEMATIC TIME
SYNTHETIC RECEIVER FUNCTIONRADIAL / TRANSVERSE
0s4s8s12s16sPsRT

Boundary depth becomes a delay in the recorded signal.

THICKNESS–VELOCITY TRADE-OFFH–κ LIKELIHOOD
20304050602.01.6km

Multiple structures can explain the same arrival time.

THE PUBLISHED COLLECTION

Explore the research. Then read the paper.

All 20 published studies in this collection. Choose any article to explore its experiments, methods and interpretation limits.

20 publications verified · 8 September 2026
ES-3501 / WHITEPAPER2026-04-30

How the Deep Earth Moves, and Why It Matters for Energy

The mantle flows a few centimetres a year, and that motion leaves a fabric in the rock that changes how seismic waves travel through it. This whitepaper follows the chain from deep flow to measurable signal, explains what a shear-wave splitting measurement does and does not constrain, and sets out where that constrains matters for an energy operator, from basin history to induced-seismicity risk.

Read the complete whitepaper
THE RESEARCH QUESTION

How does mantle motion become a measurable signal?

Follow flow, accumulated deformation, crystal alignment, directional wave speed and shear-wave splitting as one connected research problem.

Try this

Explore the five 3D chapters, then load the equal-delay examples to see why a splitting measurement does not identify a depth on its own.

THE CONNECTED OBSERVATORY EXPERIENCE

Five linked 3D chapters, an orientation distribution, three synthetic signal traces and equal-delay experiments. All respond to the shared whitepaper controls above.

Interpretation limits

Flow and fabric are schematic. Strain history, mineral composition, water, temperature and multiple anisotropic layers complicate the real inverse problem.

CONTINUE INTO THE FULL RESEARCHMethods, evidence and references
THREE DIMENSIONS. ONE CONNECTED RESEARCH STORY.Illustrative models · Geological depths are not to visual scale
RESEARCH & MODEL NOTES

From the deep Earth
to an observable signal.

Dr Tannistha Maiti

This observatory draws on Dr Tannistha Maiti’s research into continental Moho and lithosphere–asthenosphere boundaries, and EarthScan’s 54-part collection connecting those questions with modern AI methods.

Her 2018 University of Calgary PhD is titled Structure of the continental Moho and Lithosphere-Asthenosphere Boundary: Insights from receiver-function analysis and numerical modelling.

Author’s PhD presentation

What these models represent

The current scenes are analytical teaching models inspired by the research. They are not reconstructions of thesis datasets, outputs of RAYSUM, calibrated regional predictions or trained neural networks.

  • Earth: schematic shells; crust and upper-mantle thickness exaggerated for visibility. Core proportions provide context. Coastlines: Natural Earth, public domain.
  • Regional section: 300 km width, 250 km depth; a planar dipping Moho and illustrative LAB. Extracting it changes scale; it is not a sampled voxel of the globe.
  • Arrival time: one horizontal-layer central-column formula, tPs = H[√(1/Vs² − p²) − √(1/Vp² − p²)]. Dipping-interface waveform modelling requires a fuller solver.
  • Fabric: idealised fast and slow shear speeds about 4.5 km/s, 80 km path. Crystal display and directional ellipse are schematic, not a stiffness-tensor eigensolve.
  • Mantle motion: closed kinematic rolls; an illustrative strain-to-fabric mapping, A = 0.5 + 5.5(1 − exp(−strain)) percent. Flow speed animates motion separately from accumulated strain. Split delay uses the controlled path length and Vs = 4.5 km/s. The directional glyph exaggerates shape contrast 8×; it is not a tensor eigensolve.
  • Inference: a Gaussian synthetic Ps-pick likelihood on a finite H–κ grid. The interval fixes κ. Displayed surfaces are illustrative alternatives, not calibrated posterior samples.
  • Physics: u = sin(ξ − τ) + ε sin(3ξ + 1.7τ), ε = 0.35(1 − control). The normalised wave residual has amplitude 6.11ε. This controlled comparison explains the idea of a physical constraint; it does not train a PINN.

Next research layer

Verified thesis figures, station coordinates, receiver-function traces and numerical model exports will let this dashboard move from explanation to faithful, reproducible exploration.