From Park to Present: The 25-Year Evolution of MASW Inversion Algorithms

Multichannel Analysis of Surface Waves (MASW) represents a significant methodology in the field of near-surface geophysics, utilized primarily for estimating shear-wave velocity (Vs) profiles. The technique was formally introduced to the geophysical community in 1999 through a seminal paper published in the journalGeophysicsBy Choon B. Park, Richard D. Miller, and Jianghai Xia of the Kansas Geological Survey (KGS). Over the subsequent 25 years, the method has evolved from a fundamental mode approximation used in site characterization to a complex, multimodal inversion process capable of resolving complex geological stratigraphy.

This evolution has been driven by improvements in computational power, the refinement of signal processing algorithms, and the integration of global optimization techniques. Today, MASW is a standard tool in geotechnical engineering, earthquake hazard assessment, and infrastructure monitoring. The method relies on the dispersive nature of Rayleigh waves in layered media, where different frequencies of vibration penetrate to different depths and travel at different phase velocities depending on the stiffness of the subsurface materials.

What changed

The transition from early surface wave methods to modern MASW involved several technical shifts in data acquisition and mathematical processing. The following list highlights the primary advancements since the late 1990s:

• Acquisition geometry:The move from two-receiver Spectral Analysis of Surface Waves (SASW) to multi-receiver (typically 12 to 48 or more geophones) arrays significantly improved the signal-to-noise ratio and allowed for the differentiation of signal from source-generated noise.
• Domain transformation:Early methods often struggled with temporal aliasing and manual phase picking; modern MASW utilizes the wavefield transformation method (e.g., phase velocity-frequency transform) to visually separate wave energy into distinct dispersion curves.
• Inversion strategy:Initial algorithms focused almost exclusively on the fundamental mode of Rayleigh waves. Contemporary inversion routines frequently incorporate higher-order modes, which provide greater resolution at depth and more stability in cases of velocity reversals (e.g., a soft layer beneath a stiff crust).
• Processing speed:Real-time or near real-time processing has become feasible, allowing field crews to assess data quality and subsurface variability during the acquisition phase rather than relying entirely on post-field office analysis.

Background

Surface waves, specifically Rayleigh waves, account for approximately two-thirds of the total seismic energy generated by a vertical impact on the ground surface. Historically, these waves were viewed as "ground roll" noise that obscured deeper reflections in oil and gas exploration. However, in near-surface applications, the dispersive property of these waves—whereby phase velocity is a function of frequency—contains critical information about the elastic properties of the soil and rock layers through which they travel.

Before MASW, the Spectral Analysis of Surface Waves (SASW) was the predominant technique. Developed in the 1980s, SASW used a pair of receivers and a source. By varying the distance between the receivers and the source, engineers could build a composite dispersion curve. The limitation of SASW was its susceptibility to noise and the difficulty in distinguishing between different modes of propagation. In complex urban environments or sites with significant lateral heterogeneity, the two-receiver approach often produced ambiguous results that required significant expert interpretation to resolve.

The 1999 Park, Miller, and Xia Milestone

The 1999 paper, "Multichannel analysis of surface waves," fundamentally altered the approach to surface wave processing. By utilizing a multichannel record, similar to those used in seismic reflection surveys, the authors demonstrated that it was possible to apply a wavefield transformation that mapped the data from the time-space (t-x) domain into the frequency-phase velocity (f-c) domain. This transformation creates a visual image of the energy distribution, where the dispersion curve appears as a high-intensity trend.

"The multichannel approach allows for the identification and isolation of various types of seismic waves, including reflected waves, refracted waves, and different modes of surface waves, which is practically impossible with the two-receiver SASW method."

This breakthrough allowed researchers to effectively ignore unwanted noise by simply masking out the regions of the f-c plot that did not correspond to the surface wave energy. It provided a more strong and objective way to extract dispersion curves, which serve as the input data for the subsequent inversion process.

The Mathematical Shift: From Fundamental to Multimodal

The core of MASW is the inversion algorithm, a mathematical process that seeks to find a subsurface shear-wave velocity model whose theoretical dispersion curve matches the observed (picked) curve. In the first decade of MASW application, practitioners focused primarily on the fundamental mode ($A_0$). The fundamental mode is typically the most energetic and easiest to identify in simple geological settings where velocity increases monotonically with depth.

However, as the application of MASW expanded to more complex environments—such as sites with pavement, buried concrete slabs, or alternating layers of clay and sand—it became clear that the fundamental mode alone was often insufficient. In these environments, seismic energy often "jumps" between the fundamental mode and higher modes (the first, second, or third overtones). This phenomenon can lead to significant errors in the estimated depth and velocity of subsurface layers if the algorithm only considers the fundamental mode.

Table 1: Comparison of Inversion Paradigms

Modern inversion algorithms now use a "full-wavefield" approach or multimodal optimization. Instead of picking a single line, the algorithm attempts to match the entire energy distribution in the f-c domain. This reduces the subjectivity of manual picking and allows for a more accurate representation of the subsurface, particularly in the detection of low-velocity zones or thin, high-stiffness lenses.

Commercialization and Software Proliferation

The transition of MASW from a research topic to an industrial standard was facilitated by the development of commercial and open-source software. The Kansas Geological Survey developedSurfSeis, which remains one of the benchmark tools for MASW processing. SurfSeis implemented the original Park et al. Algorithms and has continued to integrate newer features, such as 2D imaging and H/V (Horizontal-to-Vertical) ratio analysis.

Parallel to commercial efforts, the open-source community contributed significantly through projects likeGeopsy, developed largely by Marc Wathelet and the SESAME project team. Geopsy provided a suite of tools for both active-source MASW and passive-source Microtremor Array Measurements (MAM). The availability of these tools democratized the technology, allowing small-scale engineering firms to perform sophisticated seismic site characterization that was previously reserved for large research institutions or specialized geophysical contractors.

Inversion Algorithms and Global Optimization

As MASW matured, the focus shifted toward the robustness of the inversion process. Early inversion routines used local optimization methods, such as the Least Squares approach, which require an initial "best guess" model. If the initial model is too far from reality, the algorithm may get stuck in a "local minimum," producing a mathematically valid but physically incorrect result.

To combat this, the last 15 years have seen the rise of global optimization algorithms, including:

• Genetic Algorithms (GA):These mimic the process of natural selection to evolve a population of models toward an optimal solution.
• Simulated Annealing:A probabilistic technique that explores the model space by gradually "cooling" the search area, allowing the algorithm to escape local minima.
• Monte Carlo Methods:These involve running thousands of random models to identify the range of possible solutions, providing a statistical measure of uncertainty in the final Vs profile.

These advanced algorithms have made MASW more reliable in "blind" sites where little or no borehole data is available to constrain the initial model.

Current Applications and Future Directions

In the present day, MASW is no longer limited to 1D vertical profiling. Through the use of "rolling" geophone arrays—where the entire array is moved along a survey line—practitioners can generate 2D cross-sections of shear-wave velocity. This is particularly useful for detecting buried utilities, identifying the boundaries of landfills, or mapping the top of bedrock for foundation design. The high-frequency application of MASW, often using 4.5 Hz to 10 Hz geophones, is also being applied to non-destructive testing of bridge decks and pavements to identify internal delamination or voids.

Looking forward, the field is moving toward the integration of Machine Learning (ML). Deep learning models are being trained on massive synthetic datasets to recognize dispersion patterns and perform inversions in milliseconds. Furthermore, the use of Distributed Acoustic Sensing (DAS) using fiber-optic cables is beginning to replace traditional geophones in some large-scale surveys, allowing for kilometers of continuous surface wave monitoring.

After 25 years, the MASW method has moved from a novel research paper to a cornerstone of modern geophysics. Its evolution reflects a broader trend in the geosciences: the movement away from simple, idealized models toward a more detailed, data-driven understanding of the complex, heterogeneous ground beneath our feet.