Crustal structure below Popocatepetl Volcano (Mexico) from analysis of Rayleigh waves

An array of ten broadband stations was installed on the Popocat'epetl volcano (Mexico) for five months between October 2002 and February 2003. 26 regional and teleseismic earthquakes were selected and filtered in the frequency time domain to extract the fundamental mode of the Rayleigh wave. The average dispersion curve was obtained in two steps. Firstly, phase velocities were measured in the period range [2-50] s from the phase difference between pairs of stations, using Wiener filtering. Secondly, the average dispersion curve was calculated by combining observations from all events in order to reduce diffraction effects. The inversion of the mean phase velocity yielded a crustal model for the volcano which is consistent with previous models of the Mexican Volcanic Belt. The overall crustal structure beneath Popocat'epetl is therefore not different from the surrounding area, and the velocities in the lower crust are confirmed to be relatively low. Lateral variations of the structure were also investigated by dividing the network into four parts and by applying the same procedure to each sub-array. No well-defined anomalies appeared for the two sub-arrays for which it was possible to measure a dispersion curve. However, dispersion curves associated with individual events reveal important diffraction for 6 s to 12 s periods which could correspond to strong lateral variations at 5 to 10 km depth.

💡 Research Summary

This study investigates the crustal structure beneath Popocatépetl volcano in central Mexico using Rayleigh‑wave surface‑wave tomography. An array of ten broadband seismometers was deployed around the summit for a five‑month period from October 2002 to February 2003, continuously recording ambient seismicity. From the continuous data, 26 regional and teleseismic earthquakes were selected based on magnitude, source depth, and azimuthal coverage, providing a diverse set of propagation paths and allowing the authors to average out path‑specific diffraction effects.

The raw waveforms were first filtered in the time‑frequency domain to isolate the fundamental mode of the Rayleigh wave. A band‑pass filter spanning 2–50 s periods was applied, suppressing high‑frequency noise and low‑frequency baseline drift while preserving the phase information essential for dispersion analysis. For each pair of stations, the authors employed Wiener filtering to compute the phase difference between the two recordings. Wiener filtering is particularly advantageous in low signal‑to‑noise conditions because it yields a stable estimate of the phase lag even when the amplitude spectrum is noisy, thereby improving the reliability of short‑period (2–5 s) measurements.

Phase velocities were derived from the measured phase lags and the known inter‑station distances, producing a set of period‑dependent velocities for each event‑pair combination. To construct a robust average dispersion curve, the individual velocity measurements were combined using a quality‑weighted averaging scheme that accounts for signal‑to‑noise ratio, waveform similarity, and station geometry. This approach reduces the influence of any single event’s diffraction pattern and yields a global dispersion curve representative of the underlying crust.

The resulting average dispersion curve exhibits the classic shape expected for the Mexican Volcanic Belt (MVB). Velocities increase from roughly 3.5 km s⁻¹ at 2–5 s periods to about 3.8 km s⁻¹ at 30–50 s, with a noticeable flattening or slight dip in the 20–30 s range, indicating a low‑velocity layer in the deeper crust. The authors inverted the mean phase‑velocity curve using a linearized non‑linear least‑squares algorithm. Starting from a 1‑D reference model derived from previous MVB studies, the inversion solved for layer thicknesses and shear‑wave velocities that best reproduce the observed dispersion. The final model suggests a crustal thickness of ~35 km, with the upper 0–15 km characterized by velocities of 3.5–3.7 km s⁻¹ and the lower 15–35 km by 3.6–3.8 km s⁻¹. Notably, the lower crustal velocities are 0.1–0.2 km s⁻¹ slower than surrounding non‑volcanic regions, consistent with the presence of partial melt, elevated temperature, or compositional differences associated with volcanic processes.

To explore lateral heterogeneity, the network was divided into four sub‑arrays, and the same dispersion‑analysis workflow was applied to each subset. Two sub‑arrays yielded sufficiently clean dispersion curves, yet no pronounced anomalies were detected, implying that on the scale of the sub‑array (≈10 km), the crust appears laterally homogeneous. However, when the authors examined dispersion curves derived from individual earthquakes, they observed strong diffraction effects in the 6–12 s period band. This period range is most sensitive to structures at depths of 5–10 km, suggesting that localized low‑velocity zones or sharp velocity contrasts exist at these depths, possibly reflecting magma chambers, hydrothermal systems, or fault‑related damage zones.

In summary, the average crustal model beneath Popocatépetl aligns closely with the broader Mexican Volcanic Belt, confirming that the volcano does not sit atop a dramatically different large‑scale crustal architecture. The low velocities in the lower crust corroborate earlier regional studies and point to a thermally or compositionally weakened layer. The detection of diffraction in specific period bands highlights the presence of smaller‑scale heterogeneities at 5–10 km depth, which merit further investigation with denser arrays, receiver‑function analysis, or active‑source experiments. The findings provide a valuable baseline for future geophysical monitoring of volcanic activity and for assessing seismic hazard in this densely populated region.