Nima Riahi

The seismic traffic footprint: Tracking trains, aircraft, and cars seismically

Although naturally occurring vibrations have proven useful to probe the subsurface, the vibrations caused by traffic have not been explored much. Such data, however, are less sensitive to weather and low visibility compared to some common out-of-road traffic sensing systems. We study traffic-generated seismic noise measured by an array of 5200 geophones that covered a 7 × 10 km area in Long Beach (California, USA) with a receiver spacing of 100 m. This allows us to look into urban vibrations below the resolution of a typical city block. The spatiotemporal structure of the anthropogenic seismic noise intensity reveals the Blue Line Metro train activity, departing and landing aircraft in Long Beach Airport and their acceleration, and gives clues about traffic movement along the I-405 highway at night. As low-cost, stand-alone seismic sensors are becoming more common, these findings indicate that seismic data may be useful for traffic monitoring.

Using graph clustering to locate sources within a dense sensor array

We develop a model-free technique to identify weak sources within dense sensor arrays using graph clustering. No knowledge about the propagation medium is needed except that signal strengths decay to insignificant levels within a scale that is shorter than the aperture. We then reinterpret the spatial coherence matrix of a wave field as a matrix whose support is a connectivity matrix of a graph with sensors as vertices. In a dense network, well-separated sources induce clusters in this graph. The geographic spread of these clusters can serve to localize the sources. The support of the covariance matrix is estimated from limited-time data using a hypothesis test with a robust phase-only coherence test statistic combined with a physical distance criterion. The latter criterion ensures graph sparsity and thus prevents clusters from forming by chance. We verify the approach and quantify its reliability on a simulated dataset. The method is then applied to data from a dense 5200 element geophone array that blanketed 7km10km of the city of Long Beach (CA). The analysis exposes a helicopter traversing the array and oil production facilities.

A Statistical Strategy to Analyzing Passive Seismic Attributes

Attributes from surface-based spectroscopic studies of the ambient wave field are hypothesized to be at least partly controlled by subsurface hydrocarbon accumulations. Such attributes can be a useful de-risking tool for exploration and field development, but they are generally also sensitive to anthropogenic activity and other physical near-surface variables. This work introduces a statistical strategy to correlate the ambient wave field attributes as a function of frequency to both subsurface targets as well as potential surface confounders to reduce misinterpretations. A resulting unconfounded correlation and its uncertainty can then be used for quantitative risk assessments. The strategy is illustrated on a passive seismic data set acquired over the Jonah field in Wyoming, USA. The ambient wave field power correlates best to hydrocarbon pore thickness from 1.0-3.5 Hz. Anthropogenic activity, inferred from high frequency energy, and elevation are both found to be uncorrelated and do not require compensation. More confounders such as near-surface geology or trap structure can and should be tested using this methodology.

Rayleigh and Love wave anisotropy in Southern California using seismic noise

We study surface wave anisotropy using three-component frequency-wave number analysis of 1 year (2012) of ambient seismic noise measured by the Southern California Seismic Network. Significant 2θ and 4θ Rayleigh wave anisotropy is observed over most of the frequency range 15 to 100 mHz (Millihertz). The wide Rayleigh wave illumination and large data volume allow for relatively high precision and sensitivity: the estimation variability above 35 mHz as well as the magnitude of the weakest significant detections is about 0.1%. The estimates are consistent with previous anisotropy studies of the region. We also show preliminary results for Love waves, but ambient Love wave illumination in Southern California may not be sufficient for the approach.

Graph clustering for localization within a sensor array

We develop a model-free technique to identify weak sources within dense sensor arrays using graph clustering. No knowledge about the propagation medium is needed except that signal strengths decay to insignificant levels within a scale that is shorter than the aperture. We then reinterpret the spatial coherence matrix of a wave field as a matrix whose support is a connectivity matrix of a graph with sensors as vertices. In a dense network, well-separated sources induce clusters in this graph. The support of the covariance matrix is estimated from limited-time data using a hypothesis test with a robust phase-only coherence test statistic combined with a physical distance criterion. The method is applied to a dense 5200 element geophone array that blanketed 7 km × 10 km of the city of Long Beach (CA). The analysis exposes a helicopter traversing the array.

Locating sources in a dense array through network-based clustering

A non-parametric technique to identify weak sources within dense sensor arrays is developed using a network approach. No knowledge about the propagation medium is needed except that signal strengths decay to insignificant levels within a scale that is shorter than the aperture. We then reinterpret the spatial coherence matrix of a wave field as a matrix whose support is a connectivity matrix of a network of vertices (sensors) connected into communities. In the asymptotic case these communities correspond to sensor clusters associated with individual sources. The support of the coherence matrix is estimated from limited-time data using a robust hypothesis test combined with a physical distance criterion. The latter ensures sufficient network sparsity to prevent network communities from forming by chance. We verify the approach on simulated data and quantify its reliability. The method is then applied to data from a dense 5200 element geophone array that blanketed 7 χ 10 km of the city of Long Beach (CA). The analysis exposes a helicopter traversing the array and oil production facilities.

Resolving small events within a dense urban array

We use data from a large 5200 element geophone array that blanketed 70 km2 of the city of Long Beach (CA) to characterize very localized urban seismic and acoustic phenomena. Such small events are hard to detect and localize with conventional array processing techniques because they are only sensed by a tiny fraction of the array sensors. To circumvent this issue, we first identify significant entries in the large coherence matrix of the array (5200 × 5200 entries) and then use graph analysis to reveal spatially small and contiguous clusters of receivers with correlated signals. This procedure allows us to track local events over time and also characterize their frequency content. The analysis requires no prior medium information and can therefore be applied under conditions of relatively high scattering. We show how the technique exposes a helicopter traversing the array, several oil production facilities, and late night activity on a golf course.

Traffic monitoring with noise: Investigations on an urban seismic network

Traffic in urban areas generates not only acoustic noise but also much seismic noise. The latter is typically not perceptible by humans but could, in fact, offer an interesting data source for traffic information systems. To explore the potential for this, we study a 5300 geophone network, which covered an area of over 70 km2 in Long Beach, CA, and was deployed as part of a hydrocarbon industry survey. The sensors have a typical spacing of about 100 m, which presents a two-sided processing challenge here: signals beyond a few receiver spacings from the sources are often strongly attenuated and scattered whereas nearby receiver signals may contain complicated near-field effects. We illustrate how we address this issue and give three simple applications: counting cars on a highway section, classifying different types of vehicles passing along a road, and measuring time and take-off velocity of aircraft at Long Beach airport. We discuss future work toward traffic monitoring and also possible connections with...

Analyzing Passive Seismic Attributes: a Statistical Strategy

Attributes from surface-based spectral analysis of the ambient wave field are hypothesized to be at least partly affected by subsurface hydrocarbon accumulations. Such attributes can be a useful de-risking tool for exploration and field development, but they can also be sensitive to anthropogenic activity and laterally varying near-surface parameters. This work introduces a statistical strategy to correlate ambient wave field attributes as a function of frequency and relative amplitude to subsurface targets as well as surface and near-surface confounders. This allows the search for frequency bands and amplitude ranges where correlations to the subsurface are greater than to such confounders. The strategy is illustrated on a passive seismic data set acquired over the Jonah field in Wyoming, USA. The ambient wave field power correlates best to hydrocarbon pore thickness from 1.0-3.5 Hz during the least energetic periods of the day. Anthropogenic activity, proxied by high frequency and higher relative amplitudes, is uncorrelated while correlation to elevation is significantly weaker compared with the hydrocarbon pore thickness. More confounders such as nearsurface geology or trap structure can and should be tested using this methodology.