Toward Near Real-Time Tomography of the Upper Mantle

Wednesday, 17 December 2014: 10:35 AM
Eric Debayle and Fabien Dubuffet, LGLTPE : Laboratoire de Géologie de Lyon : Terre, Planètes et Environnement, Université Lyon 1, ENS Lyon and CNRS, Lyon, France
We added a layer of automation to the Debayle and Ricard (2012)’s waveform modeling scheme for fundamental and higher mode surface waves in the period range 50-160s.

We processed all the Rayleigh waveforms recorded on the LHZ channel by the virtual networks GSN_broadband, FDSN_all, and US_backbone between January 1996 and December 2013. Six millions of waveforms were obtained from IRIS DMC. We check that all the necessary information (instrument response, global CMT determination) is available and that each record includes a velocity window which encompasses the surface wave. Selected data must also have a signal-to-noise ratio greater than 3 in a range covering at least the periods between 50 and 100 s. About 3 millions of waveforms are selected (92% of the rejections are due to the signal to noise ratio criterion) and processed using Debayle and Ricard (2012)’s scheme, which allows the successful modeling of about 1.5 millions of waveforms. We complete this database with 60,000 waveforms recorded between 1976 and 1996 or after 1996 during various temporary experiments and with 161,730 Rayleigh waveforms analyzed at longer period, between 120 and 360 s. The whole data set is inverted using Debayle and Sambridge (2004)’s scheme to produce a 3D shear velocity model.

A simple shell command "update_tomo" can then update our seismic model in an entirely automated way. Currently, this command checks from the CMT catalog what are the potential data available at the GSN_broadband, FDSN_all, and US_backbone virtual networks, uses web services to request these data from IRIS DMC and applies the processing chain described above to update our seismic model. We plan to update our seismic model on a regular basis in a near future, and to make it available on the web.

Our most recent seismic model includes azimuthal anisotropy, achieves a lateral resolution of few hundred kilometers and a vertical resolution of a few tens of kilometers. The correlation with surface tectonics is very strong in the uppermost 200 km. Regions deeper than 400 km show no velocity contrasts larger than 1%, except for high velocity slabs which produce broad high velocity regions within the transition zone. The use of higher modes and long period surface waves allows us to extract the shear velocity structure down to about 1000 km depth.