Kabir Roy-Chowdhury

Rayleigh wave dispersion curves from seismological and engineering-geotechnical methods: a comparison at the Bornheim test site (Germany)

Active and passive procedures for estimating the local seismic response from surface-wave measurements are compared for a test site in the Bornheim area (Germany), where independent geophysical and geological information is available. Recording was done using geophones, as well as seismometers, in various configurations. Five popular and standardized techniques were used for analysing the data: multichannel analysis of surface waves (MASW), the refraction microtremor technique (ReMi), the extended spatial autocorrelation technique (ESAC) and frequency–wavenumber analysis (beam-forming and maximum likelihood methods). The resulting surface wave dispersion curves are largely consistent, but differ in their respective low-frequency ranges due to the resolving capabilities of the respective acquisition geometries. Two joint inversions of dispersion and H/V curves, one for the lower frequency range (2.3–9.2 Hz) and the other for the complete range (2.3–45 Hz) of the dispersion curves resulted in fairly similar S-wave profiles, but increasing the frequency range allowed better estimates for the lower velocities at shallow depths. The results also compare well with borehole information. The site responses obtained from the two S-wave profiles are very similar, even at higher frequencies. The use of combined procedures (geotechnical-engineering and seismological) allows a high quality estimation of the S-wave velocity structure to be obtained, both at shallow and large depth. However, if a combined approach is not possible, for site response estimation at sites with sedimentary cover thicker than 30 to 50 m and where knowledge of the average S-wave velocity is more important than higher resolution estimates at shallower depths, the use of passive seismological 2D arrays is strongly recommended.

An efficient, probabilistic neural network approach to solving inverse problems: Inverting surface wave velocities for Eurasian crustal thickness

Nonlinear inverse problems usually have no analytical solution and may be solved by Monte Carlo methods that create a set of samples, representative of the a posteriori distribution. We show how neural networks can be trained on these samples to give a continuous approximation to the inverse relation in a compact and computationally efficient form. We examine the strengths and weaknesses of this approach and use it to determine the full a posteriori distribution of crustal thickness from surface wave velocities. The solution to this inverse problem shows significant asymmetry and large uncertainties due to trade-off with shear velocity structure around the Moho. We produce maps of maximum likelihood crustal thickness across Eurasia which are in agreement with current knowledge about the crust; thus we provide an independent confirmation of these models. In this application, characterized by repeated inversion of similar data, the neural network algorithm proves to be very efficient.

Structural inversion of gravity data using linear programming

Structural inversion of gravity data — deriving robust images of the subsurface by delineating lithotype boundaries using density anomalies — is an important goal in a range of exploration settings (e.g., ore bodies, salt flanks). Application of conventional inversion techniques in such cases, using L2 -norms and regularization, produces smooth results and is thus suboptimal. We investigate an L1 -norm-based approach which yields structural images without the need for explicit regularization. The density distribution of the subsurface is modeled with a uniform grid of cells. The density of each cell is inverted by minimizing the L1 -norm of the data misfit using linear programming (LP) while satisfying a priori density constraints. The estimate of the noise level in a given data set is used to qualitatively determine an appropriate parameterization. The 2.5D and 3D synthetic tests adequately reconstruct the structure of the test models. The quality of the inversion depends upon a good prior estimation of ...

Multi-component wavefield separation applied to high-resolution surface seismic data

Abstract A detailed characterisation of the subsurface using high-resolution multi-component seismic data requires additional processing. We adapted a frequency-domain wavefield separation method and applied it in an iterative way to subsets of a nine-component CMP survey recorded in The Netherlands. The method is well suited to remove the Rayleigh-wave. Further decomposition of the data in the case of two-component wavefield separation made it possible to group P- and S-wave sections and attempt an interpretation of events based on zero-offset arrival times, computed from existing velocity–depth profiles. Three-component separation failed beyond the extraction of the Rayleigh wave because events were not sufficiently present on all receiver components to allow for a description in terms of all polarisation parameters. The absence of the Love wave due to subsurface conditions simplified the processing of the data subset, with source and receivers oriented in a crossline-horizontal direction, to a depth converted stack, which agrees with results found earlier. The detected dipping of a peat layer was later confirmed by cone penetration tests (CPT).

Ray field maps in position/angle domain: concepts and construction

Summary Modelling of transient wave fields using ray theory requires a mapping between medium coordinates and ray parameters. In heterogeneous media multi-pathing may occur, which makes the mapping singular and causes practical problems in imaging algorithms. It is shown that the mapping is regularised when a range of sources is considered rather than a single source, and the domain of medium coordinates is replaced by a combined space of position and angles. The resulting ray field map in position/angle domain provides all ray information required for angle-domain Kirchhoff-type imaging, in the form of single valued grids that may be interpolated unambiguously. In principle the ray field map may be calculated by shooting rays from each image point to the acquisition surface for a range of take-off angles. Using the ray field map concept, however, much more efficient algorithms may be designed, with a gain in speed roughly proportional to the average number of steps needed to trace a ray from each image point to the acquisition surface.