Seismic evidence for broad attenuation anomalies in the asthenosphere beneath the Pacific Ocean

Abstract

We present QADR17, a global model of Rayleigh-wave attenuation based on a massive surface wave data set (372 629 frequency-dependent attenuation curves in the period range 50-260 s). We correct for focusing-defocusing effects and geometrical spreading, and perform a stringent selection to only keep robust observations. Then, data with close epicentres recorded at the same station are clustered, as they sample the same Earth's structure. After this preselection, our data set consists of about 35 000 curves that constrain the Rayleigh-wave intrinsic attenuation in the upper mantle. The logarithms of the attenuation along the individual rays are then inverted to obtain global maps of the logarithm of the local attenuation. After a first inversion, outliers are rejected and a second inversion yields a variance reduction of about 45 per cent. Our attenuation maps present strong agreement with surface tectonics at periods lower than 200 s, with low attenuation under continents and high attenuation under oceans. Over oceans, attenuation decreases with increasing crustal ages, but at periods sensitive to the uppermost 150 km, mid-ocean ridges are not characterized by a very localized anomaly, in contrast to what is commonly observed for seismic velocity models. Attenuation is rather well correlated with hotspots, especially in the Pacific ocean, where a strong attenuating anomaly is observed in the long wavelength component of our signal at periods sampling the oceanic asthenosphere. We suggest that this anomaly results from the horizontal spreading of several thermal plumes within the asthenosphere. Strong velocity reductions associated with high attenuation anomalies of moderate amplitudes beneath the East Pacific Rise, the Red Sea and the eastern part of Asia may require additional mechanisms, such as partial melting.

Attenuation tomography of the upper mantle

“…The initial data set consists of 372,629 epicenter station path‐average attenuation curves measured in the period range 40–240 s for the fundamental mode and up to the fifth overtone. From these observations, fundamental mode attenuation maps were obtained by Adenis et al []. We supplement these maps with higher‐mode attenuation maps based on Debayle and Ricard 's [] data set.…”

“…Here we only summarize the main specificities of the regionalization which is described in details in Adenis et al []: A thorough data selection is applied to reject data likely to be biased by errors in the source or in the instrumental response. We use great circle ray theory and focusing effects are accounted for along each ray using the formalism of Woodhouse and Wong []. We discard paths longer than 110° where Woodhouse and Wong 's [] formalism overestimates the amplitudes [ Dalton et al , ]. Unlike earlier studies, we invert for rather than for Q or Q −1 . This brings the data close to a Gaussian distribution, avoids unphysical negative values, and allows for a much larger range of variations. All maps are obtained using a Gaussian correlation function having a standard deviation of 10° and an a priori error set to 20% of the a priori model.…”

“…However, despite recent improvements [ Dalton et al , ], the agreement among attenuation models remains generally low. While in some models the quality factor increases with seafloor age similarly to the velocity [ Ma et al , ; Dalton et al , ; Warren and Shearer , ], others suggest that large‐scale attenuating anomalies are rather correlated at asthenospheric depths with the hot spot distribution under the Pacific and the southern Indian Ocean [ Gung and Romanowicz , ; Romanowicz , ; Adenis et al , ]. Part of these differences may be due to different modeling techniques and to the way focusing‐defocusing terms are treated.…”

“…Part of these differences may be due to different modeling techniques and to the way focusing‐defocusing terms are treated. While some authors reject paths that might be strongly affected by focusing effects [ Romanowicz , ; Gung and Romanowicz , ], others account for focusing using great circle linearized ray theory [ Billien et al , ; Dalton and Ekström , ; Dalton et al , ; Adenis et al , ] or 2‐D finite‐frequency sensitivity kernels [ Ma et al , ].…”

“…The bulk attenuation is assumed to be zero in these inversions. In a previous paper [ Adenis et al , ], we introduced QADR17, a set of attenuations maps at different periods built from the fundamental mode of Rayleigh waves. QsADR17 is obtained after completing this data set with higher‐mode attenuation maps and inverting for a 3‐D model of the shear wave quality factor Q μ .…”