Skip to content
Back to search results

Ambient Noise Tomography

Introduced by
Shapiro, Campillo, Stehly, and Ritzwoller
Year
2005
Type
Passive seismic imaging via correlation of ambient noise
Family
Seismic surface wave imaging
Font size

What does it do?

Ambient Noise Tomography (ANT) is a seismic imaging method that extracts surface wave information from long-term records of seismic background noise, enabling high-resolution imaging of crustal and upper mantle structure.

Developed by Shapiro, Campillo, and colleagues in 2005, ANT has revolutionized seismic imaging by enabling detailed crustal velocity maps at minimal cost without requiring earthquakes or active sources.

The intuition behind the method

The Earth is constantly shaking due to ocean waves, wind, and human activity (ambient seismic noise). When two seismometers are far apart, the cross-correlation of their ambient noise records converges (over weeks to months) to the Green's function—the impulse response between the two stations. The Green's function is dominated by surface waves (Rayleigh waves for vertical motion, Love waves for horizontal). By analyzing this emergent Rayleigh wave dispersion from many station pairs, a 3D velocity model of the crust and upper mantle is reconstructed. ANT works passively using background seismic noise, eliminating the need for earthquake waiting or expensive active sources.

History

Pioneered by Shapiro, Campillo, Stehly, and Ritzwoller in 2005, ambient noise tomography initially faced skepticism but rapidly became mainstream. Early applications focused on dense networks in Europe and North America. By 2010, ANT was applied globally using permanent networks. Recent advances include 3D ANT, joint inversion with teleseismic surface waves and receiver functions, and application to monitoring (detecting time-lapse changes in crustal properties).

When to use it

Use ANT for regional-scale crustal imaging where earthquake catalogs are sparse or where rapid deployment is essential. ANT excels at mapping shallow crustal structure (depths ~1–20 km) and detecting near-surface anomalies. For very-deep structure (mantle), ANT is less sensitive; combine with teleseismic surface waves. ANT is ideal for rapid hazard assessment (earthquake risk), geothermal exploration, and hydrocarbon prospecting.

Limitations and considerations

  • Directional dependence: noise sources (ocean, weather) are heterogeneously distributed, biasing wave directions and introducing anisotropy artifacts
  • Sensitivity mainly to shallow crust: energy content decreases at longer periods, limiting depth penetration compared to earthquake-based methods
  • Non-linear dependence on shear velocity: inversion can be non-unique; multiple velocity models may fit the dispersion data equally well
  • Computational expense: cross-correlations of months to years of continuous data at high sampling rates demand significant computational resources

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Passive and cheap: requires no active source, only broadband seismometers; deployment is faster than active-source surveys
  • High spatial resolution: dense station arrays and multiple frequencies enable detailed 2D/3D imaging
  • Rapid imaging: continuous noise processing means useful results appear within weeks to months of deployment, not years
  • Complements body-wave methods: surface waves are sensitive to shear velocity and damping; body waves sense P-velocity. Joint inversion is powerful.

Applications

  • Crustal structure mapping: creating detailed 3D shear-velocity models for tectonic and hazard studies in regions with sparse earthquakes
  • Geothermal resource exploration: detecting high-temperature zones via low-velocity anomalies in geothermal fields (Iceland, New Zealand, Kenya)
  • Earthquake hazard assessment: imaging fault zones and stress accumulation regions via velocity variations, informing seismic hazard maps
  • CO2 storage monitoring: time-lapse ANT detecting velocity changes from pressure and saturation changes in geologic carbon storage sites

How it works

How it works

1

Continuous seismic noise waveforms are recorded at broadband seismometer stations for extended periods (months to years). A dense array of stations (spacing 1–10 km for regional studies, 10–100 km for continental studies) is deployed.

2

Raw noise data are filtered to emphasize frequency bands of interest (typically 5–150 seconds period for crustal imaging). Data are quality-controlled to remove instrument artifacts and periods of transient seismicity.

3

Cross-correlations between all station pairs are computed over long time windows. Stack tens of days to years of cross-correlations to enhance the signal (Green's function) relative to noise.

4

Rayleigh wave dispersion (group velocity vs. period) is extracted from the stacked cross-correlations using frequency-time analysis (FTAN) or other methods. 2D or 3D tomographic inversion maps dispersion variations across the study region.

5

Dispersion maps are inverted for shear-velocity profiles at each location. 3D shear-velocity models reveal lithological variations, geothermal anomalies, and crustal structure.

Common pitfalls

  • Inadequate stacking: insufficient time window (< a few weeks) leads to poor Green's function recovery and biased dispersion estimates
  • Directional bias: if noise sources are concentrated (e.g., from ocean swell in one direction), tomographic results show azimuthal anisotropy artifacts
  • Over-interpreting high-frequency (short-period) variations; short-period Rayleigh waves are sensitive to very shallow structure and may reflect heterogeneous near-surface variability rather than meaningful geology

Frequently asked questions

Why does cross-correlation of noise converge to the Green's function?

Mathematical theory (equipartition) states that when two points are in a diffuse wave field (excited equally from all directions), cross-correlation converges to the Green's function. In reality, noise sources are not perfectly diffuse, but sufficient averaging over weeks to months and careful spatial sampling yields robust Green's functions.

What is the difference between ambient noise ANT and earthquake surface waves?

ANT uses continuous background noise passively; earthquake surface waves use energy from distant earthquakes. ANT requires long averaging but no earthquakes; earthquake surface waves are intermittent but can be rapid if a large earthquake occurs nearby. For crustal imaging, ANT provides higher frequency content (shorter periods, finer resolution) than most teleseismic surface waves.

How long must seismometers record for ANT?

Minimum useful stacking window is ~2–4 weeks; 3 months to 1 year gives much better signal. For high-frequency (short period) ANT targeting shallow structure, faster convergence (days to weeks) is possible. For deep mantle imaging, years of data may be needed.

Can ANT detect time-lapse changes in crustal properties?

Yes. Repeated ANT surveys (repeating the deployment over months/years) can detect velocity changes. Applications include CO2 sequestration monitoring, geothermal production changes, and earthquake-triggered stress changes. Changes on the order of 1–2% are detectable with careful processing.

Related methods

Sources

  1. Shapiro, N. M., Campillo, M., Stehly, L., & Ritzwoller, M. H. (2005). High-resolution surface-wave tomography from ambient seismic noise. Science, 307(5715), 1615-1618.

    Open DOI reference
    Copied.
  2. Bensen, G. D., Ritzwoller, M. H., Barmin, M. P., et al. (2008). Processing seismic ambient noise data to obtain reliable broad-band surface wave dispersion measurements. Geophysical Journal International, 169(3), 1239-1260.

    Open DOI reference
    Copied.