The Evolution of Microtremor Array Methods: From Kanai (1957) to Modern MAM

Microtremor analysis represents a fundamental shift in seismic site characterization, moving from active-source methods to the utilization of ambient vibrations. These vibrations, often termed ambient noise or microtremors, originate from various sources including atmospheric pressure changes, wind, ocean waves, and human-induced activities such as traffic and industrial machinery. By capturing these low-amplitude ground motions, geophysicists can derive the shear-wave velocity (Vs) structure of the subsurface without the need for explosive charges or mechanical impactors.

The evolution of these techniques has transitioned from simple period estimation to complex array processing and inversion. This historical trajectory involves the refinement of mathematical frameworks for wave propagation in heterogeneous media, the standardization of field protocols, and the development of sophisticated software capable of extracting dispersion characteristics from seemingly stochastic signals. Today, Microtremor Array Methods (MAM) are essential tools for seismic hazard assessment, urban planning, and the non-destructive evaluation of critical infrastructure.

Timeline

• 1957:Kiyoshi Kanai publishes foundational research on the determination of site periods using microtremor measurements, establishing the link between ambient vibration and local geology.
• 1970:Tatsuo Okada introduces the Spatial Autocorrelation (SPAC) method, providing a mathematical basis for extracting phase velocity from microtremor arrays.
• 1989:Yutaka Nakamura presents the Horizontal-to-Vertical Spectral Ratio (HVSR) technique, significantly simplifying the process of estimating the fundamental resonance frequency of soil layers.
• 2001–2004:The European SESAME (Site Effects Assessment Using Ambient Excitations) project standardizes microtremor measurement and processing, leading to the widely adopted 2004 guidelines.
• 2010s–Present:Integration of MASW (Multichannel Analysis of Surface Waves) and MAM, along with the development of high-resolution inversion algorithms and Bayesian statistical frameworks for subsurface imaging.

Background

The study of microtremors is rooted in the physics of surface waves, primarily Rayleigh and Love waves, which dominate the ambient noise field at certain frequencies. In a layered, heterogeneous solid-state medium, these waves exhibit dispersion, meaning their velocity is frequency-dependent. Shorter wavelengths (higher frequencies) travel through shallower materials, while longer wavelengths (lower frequencies) penetrate deeper into the Earth's crust. By measuring this dispersion, researchers can infer the physical properties of the subsurface stratigraphy.

Historically, seismic exploration relied heavily on body waves (P-waves and S-waves) generated by controlled sources. However, in urban environments where active sources are disruptive or logistically impossible, microtremors provide a continuous and non-invasive alternative. The challenge lies in the complex nature of the wavefield, which requires precise sensor calibration and rigorous statistical analysis to separate signal from random noise. The field of Surface Wave Hub research focuses on this empirical study, applying acoustic wave propagation characteristics to solve engineering and geological problems.

Kanai (1957) and the Roots of Site Period Estimation

Kiyoshi Kanai's early work focused on the observation that different geological settings amplified ground motion at specific frequencies. Using early electromagnetic seismometers, Kanai observed that the dominant period of microtremors at a given location closely matched the period of earthquake-induced ground motion. He proposed that microtremors could be used as a proxy to determine the resonant characteristics of a site, which is a critical factor in earthquake-resistant design.

Kanai's approach was primarily empirical. He categorized site conditions based on the shape of the microtremor amplitude spectra. While this method lacked the depth-profiling capabilities of modern techniques, it established the concept of "site effects" and demonstrated that the subsurface structure acts as a filter for seismic energy. His research proved that the impedance contrast between soft surface layers and the underlying bedrock could be identified through passive monitoring.

The Spatial Autocorrelation (SPAC) Method

In the 1970s, Tatsuo Okada and his colleagues developed the Spatial Autocorrelation (SPAC) method to move beyond simple period estimation. The SPAC method assumes that the microtremor wavefield is a stochastic process that is stationary in both time and space. By deploying an array of sensors—often in circular or triangular geometries—researchers can calculate the spatial coherence of the signals between different sensor pairs.

The mathematical core of SPAC involves the Bessel function of the first kind and zero order. By fitting the observed spatial autocorrelation coefficients to this function, the phase velocity of the surface waves can be determined for specific frequencies. This allowed for the first reliable passive estimation of shear-wave velocity profiles at depth. Okada's work transformed microtremor analysis from a qualitative observation of site periods into a quantitative geophysical tool for subsurface mapping.

Nakamura’s HVSR and the 1989 Breakthrough

Perhaps the most significant milestone in the popularization of microtremor studies was the introduction of the Horizontal-to-Vertical Spectral Ratio (HVSR) technique by Yutaka Nakamura in 1989. Nakamura proposed that the ratio of the horizontal components' Fourier spectra to the vertical component's spectrum could eliminate the source effects of microtremors, leaving only the site response. This method, often referred to as the "Nakamura technique," allowed for the identification of the fundamental resonance frequency (f0) using only a single three-component seismometer.

The HVSR method gained rapid global acceptance due to its logistical simplicity and low cost. It proved particularly effective in identifying the depth to bedrock in areas with high impedance contrasts. However, early applications were controversial, as the theoretical justification for why the H/V ratio accurately represented the S-wave transfer function was debated. Despite these debates, empirical evidence consistently showed that the peak of the HVSR curve correlated well with the fundamental resonance of the soil column.

The SESAME Project: Standardization and Guidelines

As the use of HVSR and array methods proliferated, the lack of standardized field and processing protocols led to inconsistent results. Between 2001 and 2004, the European Commission funded the SESAME project to address these issues. The project involved a consortium of researchers who conducted extensive field tests and numerical simulations to determine the reliability and limitations of microtremor methods.

The 2004 SESAME guidelines established rigorous criteria for:

• Field setup:Minimum recording durations, sensor coupling with the ground, and the avoidance of transient noise sources like heavy machinery.
• Data processing:Window selection, smoothing techniques, and the calculation of geometric means for the horizontal components.
• Interpretation:Criteria for the stability and significance of the HVSR peak, ensuring that researchers did not over-interpret artifacts in the data.

The SESAME project provided the scientific community with a common framework, ensuring that results obtained in different parts of the world could be compared and validated. It also clarified that while HVSR is excellent for finding resonance frequencies, it is less reliable for determining the full Vs profile without additional data, such as array measurements.

Contemporary Microtremor Array Methods (MAM)

Modern Microtremor Array Methods (MAM) represent the integration of earlier theories with high-speed computing and advanced signal processing. Contemporary MAM often uses linear or nested circular arrays to capture a broad range of wavelengths. These methods are frequently used in conjunction with active-source techniques like MASW to create a unified dispersion curve spanning from 1 Hz to over 100 Hz.

Current research focuses on the development of inversion algorithms. Once a dispersion curve is extracted from the array data, an inversion process is used to generate a 1D, 2D, or even 3D model of the subsurface shear-wave velocity. Advanced techniques use Global Optimization algorithms (like Genetic Algorithms or Neighborhood Algorithms) and Bayesian statistics to quantify the uncertainty in the resulting models. These profiles are used to determine the Vs30 value (the average shear-wave velocity in the top 30 meters), which is a key parameter in building codes worldwide.

Comparison of Passive and Active Methods

What sources disagree on

Despite the advancements in MAM, several areas of disagreement persist among practitioners and theoreticians. One primary point of contention is the composition of the microtremor wavefield. While most methods assume a dominance of Rayleigh waves, some researchers argue that Love waves and body waves contribute significantly to the signal, which can lead to inaccuracies in dispersion curve interpretation if not properly accounted for.

There is also ongoing debate regarding the inversion process. Because seismic inversion is inherently non-unique—meaning multiple different velocity models can produce the same dispersion curve—different researchers favor different regularization techniques. Some advocate for "smooth" models that favor gradual transitions between layers, while others prefer "sharp" models that allow for distinct lithological boundaries. Furthermore, the depth of penetration claimed by passive methods is often scrutinized, as it depends heavily on the presence of low-frequency energy in the ambient noise, which may not always be present in all environments.

Finally, the interpretation of the HVSR peak remains a subject of discussion. While most agree it represents the fundamental frequency, there is disagreement on whether the amplitude of the HVSR peak can be directly correlated with the actual amplification factor during an earthquake. Some studies suggest a strong correlation, while others indicate that the peak amplitude is influenced by the complex interaction of different wave types and does not purely reflect the S-wave amplification.

Practical Applications in Engineering

The practical application of these methods extends to various engineering disciplines. In the field of non-destructive testing, MAM is used to assess the integrity of large foundations and embankments. By monitoring changes in the dispersion curves over time, engineers can detect internal erosion or the development of voids within heterogeneous engineered structures. In urban settings, microtremor surveys are used to map buried utilities and identify abandoned tunnels or karst features that pose a risk to new construction.

The meticulous interpretation of microtremor data, as practiced by organizations like the Surface Wave Hub, allows for the characterization of shallow subsurface anomalies without the need for invasive drilling. This empirical approach to acoustic wave propagation provides a cost-effective and environmentally friendly solution for site characterization in the 21st century.