Rebecca L. Saltzer

Comparing P and S wave heterogeneity in the mantle

From the reprocessed data set of Engdahl and co-workers we have carefully selected matching P and S data for tomographic imaging. We assess data and model error and conclude that our S model uncertainty is twice that of the P model. We account for this in our comparison of the perturbations in P and S-wavespeed. In accord with previous studies we find that P and S perturbations are positively correlated at all depths. However, in the deep mantle systematic differences occur between regions that have undergone subduction in the last 120 million years and those that have not. In particular, below 1500 km depth ∂ln Vs/∂ln Vp is significantly larger in mantle regions away from subduction than in mantle beneath convergent margins. This inference is substantiated by wavespeed analyses with random realizations of the slab/non-slab distribution. Through much of the mantle there is no significant correlation between bulk sound and S-wave perturbations, but they appear to be negatively correlated between 1700 and 2100 km depth, which is also where the largest differences in ∂ln Vs/∂ln Vp occur. This finding supports convection models with compositional heterogeneity in the lowermost mantle.

Predicting Vshale and porosity using cascaded seismic and rock physics inversion

In characterizing hydrocarbon reservoirs, estimating reserves, and developing models for how to best extract the hydrocarbons, it is useful to know the lithology and associated porosity of the rocks in the target interval. It is particularly important to have accurate estimates of volumes and flow characteristics early in the development of deepwater reservoirs, as sizing the facilities to the resource is a significant component of an economically successful project. Accurate characterization of the reservoir can be a challenge in this situation as well control is typically very limited.

Earthquakes—a naturally occurring source of low-frequency data

Seismic reflection data contain information at two very different length scales. The long wavelength or lowfrequency information is derived primarily from move-out velocity analysis and provides information on the order of kilometers (i.e., ~1 to 2 km). The short wavelength (highfrequency) information comes from reflection amplitudes and/or changes in those amplitudes with offset (AVO) and provides information at scales of tens of meters (i.e., ~10 to 250m). Between these two length scales is a gap. This missing information (between 1 and 6 Hz) causes a general uncertainty in all seismic-reflection inversions and creates a challenge for full-waveform inversion methods to properly converge without getting stuck in local mathematical minima. This problem is further compounded in regions with complex structure (e.g., basalt flows, sills and dykes, salt bodies, fold and thrust belts, etc.) where it can be difficult to obtain the long-wavelength information reliably from the data, thereby leading to poor imaging. One possible solution for both of these problems is to use local, regional and teleseismic (far-away) earthquakes as a source of low-frequency energy, exploiting the Earth’s naturally occurring seismicity. This paper describes a field experiment that was designed to test this idea.

Using linear combinations of angle stacks to predict band‐limited porosity and vshale

We introduce a new type of weighted stack that can be applied to seismic data to produce band-limited estimates of porosity and shale volume (vshale). This weighted stack combines linearized approximations of the seismic reflectivity with linear rock property relationships into a single linear operator that can be applied directly to angle stacks. In order to account for the effects of fluid property variations in the rock property relationships, we specify the location of the fluid contacts within a reservoir and apply rock property relationships appropriate to the fluid at each point in the seismic volume. This method can be useful after an exploration well has been drilled and well logs are available for developing rock property relationships and fluid contact information is available. The method can then be applied to define the distribution of reservoir quality (sands) as part of development planning.