Experiment geometry to measure reflection coefficients versus grazing angle . The direct and the bottom bounce path are shown as solid lines, the surface reflected paths as dotted lines. Source moves away from receiver which is mounted at height above the seabed and transmits at ranges .
Malta Plateau (Site 2) sound velocity profile.
Malta Plateau seismo-acoustic data in reduced time. The solid line indicates the upper limit of the layer stripping packets, and dashed lines represent the lower limits of the first and second packets.
Marginal probability distributions for the inversion of picked travel-time data for layer thickness , sound velocity , and bias parameters . Prior bounds correspond to plot widths.
Reflection coefficient data processed into two packets that extend to increasing depth (packets one and two indicated by solid and dashed lines, respectively).
Measured reflection coefficients and MAP model replica (solid line) for the first layer packet consisting of one layer over a half space.
Data residuals for the first packet. The residuals are strongly correlated.
Estimated running rms average standard deviation for the first packet. The variation with angle indicates nonstationary residuals.
Estimated non-Toeplitz covariance matrices for all frequencies, packet one.
Marginal probability distributions from reflection-coefficient inversion, packet one.
Measured reflection coefficient and MAP model replica (solid line) for the second layer packet consisting of three layers over a half space.
Marginal probability distributions recovered from reflection-coefficient inversion of layer packet two.
Upper diagonal of the parameter correlation matrix for packet two reflection-coefficient inversion.
Joint marginal probability distributions for selected parameters.
Autocorrelation function for raw (columns one and three) and standardized (columns two and four) residuals as a qualitative measure of data error covariances.
Runs test results: crosses and circles show values for the raw and standardized residuals, respectively. Dashed line shows the 0.05 level.
Histograms of raw (columns one and three) and standardized (columns two and four) residuals as a qualitative measure of Gaussianity of data error.
KS test results: Crosses and circles show values for raw residuals and standardized residuals, respectively. Dashed line shows the 0.05 level.
Inversion results compared to three shallow cores taken at the experiment site (piston core in blue, two gravity cores in red and green). The MAP result is given as the solid black line; profile marginals are given as colored areas with red regions corresponding to high probability. Note that the shifts to low velocity and density over the bottom meter of the blue core are likely extraction artifacts.
Numerical values of prior information and inversion results (MAP model and 100% HPD credibility intervals) for picked travel-time data.
Numerical values of prior bounds and final inversion results (MAP model and 95% HPD credibility intervals).
Article metrics loading...
Full text loading...