• Figure 2: This figure provides a comparison of the new sub-bottom profiler Kongsberg Maritime SBP120 for three different configurations for the same transect over a muddy sediment deep water site in the Bay of Biscay (France). This systems uses a multibeam technology which allows to parametrize both emitting (Tx) and receiving (Rx) apertures. In this case, the best penetration is obtained with a 3 by 3 degrees Tx-Rx configuration (left) (the penetration is around 100 meters depth), while there is a constant degradation of the system performances with the widening of the Tx-Rx apertures (Middle: 6 by 6 degrees - Right: 12 by 12 degrees). The 3 by 3 degrees configuration provides a penetration that is twice that of the 12 by 12 degrees configuration.
• Figure 3: This figure provides a synthesis of the frequency-grazing angle domains that are involved in the available seabed characterization techniques. The techniques based on backscattering measurements are represented by filled areas, while those that are based on coherent reflection are represented by the unfilled areas. Ground truthing techniques (corings) are not represented here: they only involve measurements at very high frequencies (several MHz or hundreds of kHz). This figure obviously shows that each technique provide an image of the sea-bottom that is filtered by the technique itself.
• Figure 4: This figure provides a joint analysis of multibeam echosounder EM12 imagery andcorings. The picture (c) shows that there are two distinguishable areas: the backscatteredlevel is higher in the dark grey area that in the light grey one. In [11], theauthor mentions that both areas are governed by two different sedimentological depositprocesses that were characterized thanks to several corings (a)-(b)-(d) (4 corings in each areas). The superficial sediments are quite the same for both regions (mud).Below the superficial layer, the dark grey area is characterized by several silty thin and hard layers (whose compressional speed are higher than 1700 m/s) surrounded by muddy sediments. The light grey area sub-surface sediments are mixed muddy and silty sediments with very thin hard silty layers. The acoustical energy of the EM12 echosounder is known to reach depth penetrations of a few tens of meters [5].The seafloor imagery that EM12 give are therefore integrated images of the first superficial layers of the seafloor.
• Figure 5: This figure presents a coupled analysis of the imagery provided by an EM1002S (95kHz) multibeam echosounder and a sub-bottom profiler section on a sandy bank (ArmenBank) in a coastal area in the North-East Atlantic Ocean near Brest (France).The bank is around 30 meters high. On the left side of the bank (red underlined area),a vertical layering can be noticed. The layering is probably due to sedimentological dynamics. In the same area, it can be seen a smooth transition of the multibeamechosounder imagery. On the contrary, on the right side of the bank, there is no layering and the bound of the bank is rather abrupt on the imagery. Though still under investigation, the reasons of these different behaviours of the EM1002S imagery seem to be linked to the sediment processes.
• Figure 6: This figure presents the simulated transmission losses on a vertical line array that is 5.6 km distant from the acoustic source in the context of the INTIMATE’96 experiment[18]. The simulation shows that a semi-infinite fluid half space, a two layer bottom or a three layer one provide the same estimates of the propagation losses above 200 Hz. Some differences appear in the range 50 - 200 Hz due to the fact that the measurements were done for frequencies between 300 and 800 Hz.

• Figure 1: Location of the seabed reflection measurements in the Strait of Sicily: clay site(north), sand site (south).
• Figure 4: Analysis of data residuals from MAP inversion with ML standard-deviation estimates.Left column compares histograms of residuals (in units of data standard deviation)to theoretical Gaussian distributions (smooth curve). Right column shows residual autocorrelation functions. Panels (a) and (b) show results for clay-site data, and (c)and(d) for sand-site data.
• Figure 5: Marginal probability distributions from clay-site data. Histograms labeled 1 computed using ML estimate for σ (σ indicated by arrow); histograms labeled 2 computed with σ included in Bayesian inversion; histograms labeled 3 used standard deviations derived from experimental uncertainties.
• Figure 6: Marginal probability distributions from sand-site data: 1–ML standard deviation estimate,2–standard deviation included in Bayesian inversion, and 3–angle-dependent standard deviations from experimental errors.

• Figure 3: South Elba environment. (a) Acoustic pressure field generated with the wide-angle PE using NLBC calculated from measurement of the subbottom properties. The color scale represents the modulus of the complex-valued ψ, normalised to unity. (b) Range-average sound speed profile of the water column and sediment layer used for the modeling. The upper and lower scales refer to water and sediment, respectively.

• Figure 8: Acoustic signal received at 30 km during the PRIMER experiment and its timefrequency display.
• Figure 9: The location of the ASIAEX experiment in the East China Sea (Left panel). The right panel shows the bathymetry and locations of the Wide Band Source deployments.
• Figure 10: Acoustic signal and its time-frequency representation for one of the WBS deployed during the ASIAEX experiment. The signal shown here corresponds to shot A (Figure9) on the North-West side of the circle of shot deployments received at 30 km.
• Figure 11: Acoustic signal and its time-frequency representation for one of the WBS deployed during the ASIAEX experiment. The signal shown here corresponds to shot B (Figure9) on the North-East side of the circle of shot deployments received at 30 km.
• Figure 12: Surface sediment description at Primer location. Left panel shows the sediment tomography inversions [7]. The location of gravity cores (core 1, core 2 and core3)and the USGS data (1, 2 and 3) are also shown. The right panel shows the predictions from the ESME sediment model [8].
• Figure 14: Summary of the East China Sea sediment information. The contour lines are the sediment compressional speeds in excess of 1600 m/s obtained from tomography inversions. The two dashed lines indicate the grain size boundary (separating ø less than and greater than 4.3) and the Niino and Emery boundary separating sand and mud and sand (Miller et al. [11]).
• Figure 15: Sediment tomography inversions for the top 40 m (left panel) and top 3 m (rightpanel) of the East China Sea. Comparisons with other inversions are also provided(Miller et al. [11]).
• Figure 16: Sediment layer thicknesses estimated by water gun surveys. The surface layer thickness in maters is shown in the left panel whereas the sub-surface layer thickness in meters is shown in the right panel (Miller et al. [11]).

• Figure 1: Source/Receiver geometry for entire 50 Hz track. The data of interest was taken along Segment 2 (opening in range) and segment 3 (closing in range). Bathymetric contours are in meters.
• Figure 2: Spatial pressure field relative to source at 50 Hz. Upper is pressure magnitude, middle is phase, and lower is relative receiver position.
• Figure 3: Pressure magnitude and phase at 50 Hz for opening and closing segments of track near turnaround.
• Figure 4: Wavenumber estimation with range along segment 2 for 50 Hz data.
• Figure 5: Wavenumber estimation with range along segment 3 for 50 Hz data.
• Figure 6: Evolution of modes 1 and 2 with range illustrating bias in wavenumber estimates due to source motion along segments 2 and 3.
• Figure 7: Bias in inferred sound speed for the halfspace with range using uncorrected modal eigenvalue estimates for segments 2 and 3.
• Figure 8: Spectral estimates for single aperture of moving source data. Source speed was 2 m/s.

• Figure 1: Malta Plateau study area between Sicily (to the north) and Maltese island of Gozo (to the south). Seismic reflection survey, tracks are indicated by grey lines, depths are in meters. Seismic reflection survey lines of interest for this paper are indicated by the solid red line with a NE bearing (Figure 2) and the red line with a NW bearing (Figure7).
• Figure 6: Broadband reverberation (relative dB) from SUS in the Straits of Sicily. The island of Sicily is in the upper right hand corner. The data have a time varying gain. There is a left-right ambiguity in the figure around the axis of the towed array. The strong radial line to the northwest is from a ship (tanker) tending the Campo Vega oil rig (box).The reverberation return from two mud volcanoes (MVs) are indicated by the arrows.MVs are indicated by dots.

• Figure 1: Gas seepage in the Gulf of Gdansk, Baltic Sea.
• Figure 2: The 3.5 kHz subbottom profile indicating gas in deep bottom layers – Gulf of Gdansk,Baltic Sea.
• Figure 7: a) Area of gas detection in sediments in the Gulf of Gdansk. White’ circles indicate non-linear acoustic measurement at frequency 30.4/33.6 kHz, black crosses 105.5/115 kHz, dashed line the transect of 3.5 kHz subbotom profiling; b) Types of sediments: 1 – gravel, stones, 2 – sands, 3 – marine silty clay, 4 – marine clayey silt, glacial marine clay, 5 – till (from Pieczka, 1980).
• Figure 8: The echograms and amplitudes of linear and non-linear components of echo signals measured at the station where the top layer of sediment is marine silty sand laying on the deltaic silts. Frequency 30.4/33.6 kHz.
• Figure 9: The echograms and amplitudes of linear and non-linear components of echo signals measured at the station where the top layer of sediment is the thin layer of mud laying on the sand. Frequency 30.4/33.6 kHz.
• Figure 10: The spatial distribution of gas bubbles density along the transect A (see Fig.7.b) for consecutive stations. At each station the vessel was drifting during the measurements.Results were estimated for sum frequency component of the pair of incident waves at 30.4/33.6 kHz.
• Figure 11: The comparisons of spatial distributions of gas bubbles densities in transect B (see Fig.7.b) for consecutive stations. The top figures show results for sum frequency component of incident waves of 30.4/33.6 kHz, while the bottom figures for 105.5/115 kHz.

• Figure 5: Model-predicted field scattered from a box-shaped fluid body in homogeneous water,BIE (solid), Kirchhoff (dotted). Left: All 7 receivers. Right: Closeup of receivers 1 and 7.
• Figure 6: Model-predicted field scattered from a buried box-shaped fluid body. Above: BIE(solid), Ray-Kirchhoff (dotted). Below: Kirchhoff (solid), Ray-Kirchhoff (dotted).Left: All 7 receivers. Right: Closeup of receivers 1 and 7.
• Figure 7: Left: Emitted pulse by calibration measurements (dashed) and smooth band-limited approximation (solid). Right: Fourier spectra.
• Figure 8: Average over 84 pings of signals recorded by the middle five hydrophones of the vertical array.
• Figure 9: Signals in the middle five hydrophones of the vertical array. Left: Experimental data.Right: Predicted by the RK method (solid) and experimental data (dashed).
• Figure 10: Model-predicted (solid) and experimentally observed (dotted) signals in the vertical array. Signals are re-scaled to equal amplitude. Left: RK. Right: BIE.
• Figure 11: Fitness function Φ in eqn. (12) as function of roll- and yaw angle.
• Figure 13: Convergence history. Left: GA method, Right: DE method.
• Figure 14: Received signals. Dotted: Experimetal data. Solid: Modelled, with parameters by acoustic inversion shown in the right column of Table 3. Left: RK. Right: BIE.

• Figure 2: Calibration of the acoustic projector was performed in 2-D (left, in the XY plane) and in 3-D (right, in the XZ plane, with scaling circles every 10°). Note the narrow beamwidth (10°), the small sidelobes, and the asymmetry for the angles further away from the axis of the transducer.
• Figure 3: Top: details of a typical rotation scan, for bistatic angles from 140° to 220°. The hydrophone is mounted at the bottom of the stainless steel tube, and measurements at different scattering angles are obtained by vertically moving the tube. Note the positions of targets T₁–T₄ at the surface of the silt tray. Bottom: diagram of a typical line scan. The hydrophone (RX) starts at position TA1 and scans with 1-cm steps to position TA111, giving a range of scattering angles θ_s. The transmitter (TX) is tilted at 45. in this case.
• Figure 4: Scattered waveforms for bare silt (left) and with target T₁ proud in the Z direction,i.e. placed vertically on the seabed (right). In this case, the incidence angle is 45° and the bistatic angle is 180° (in-plane scattering). The normalised amplitudes of the scattered waveforms are plotted as a function of time and the scattering angle at which they were measured.
• Figure 5: Contour maps of the normalised amplitude as a function of the scattering angle for abistatic angle of 180°, for target T₁ alone (left) and targets T₁+T₂ (right).
• Figure 6: Contour maps of the normalised amplitude as a function of the scattering angle, for a bistatic angle of 160°, for targets T₁+T₂ (left) and targets T₁+T₂+T₃ (right).
• Figure 7: Contour maps of the normalised amplitude, now represented as a function of the bistatic angle, for a scattering angle of 67.4°, for targets T₁+T₂+T₃(left), and targets T₁+T₂+T₃+T₄ (right).
• Figure 8: (a) an arbitrary input signal, with high noise, shows the difference between nonrecursive Wiener filters (b) and the adaptive implementation optimising the relative weights (c). Note how the noise is considerably reduced in (c), and how the main peaks are correctly detected
• Figure 9: The scattering points, detected and localised using the algorithms presented in the text, are colour-coded according to their respective strengths (from blue, low, to red,high). They are presented next to the anticipated target positions for target T₁ (A) and for target T₂ (B). The offsets from the expected positions are very small, and can be explained by the uncertainty in positioning the targets by hand from the top of the water-filled tank.
• Figure 10: During the SITAR sea trials (late 2003), the ROV was flown above targets of interest,doing both line scans and rotation scans. It was accurately positioned through a transponder net (yellow, upright cylinders in this diagram). The targets were imaged with a narrow-beam parametric sonar, and the scattered signals were received on a hydrophone chain. Preliminary results are consistent with the conclusions from the tank experiments. Image ©FOI, Sweden.

• Figure 1: Left: Photograph of the FFCPt during deployments in St. Margaret’s Bay in 2002.Right: Diagram indicating the contents of the different modules and sensor locations in the present configuration of the FFCPt. Note the change from the brass rings to point electrodes and the addition of the sound velocity and pressure sensor in the tail.
• Figure 2: Time series of FFCPt sensors for a deployment at location 7 (Table 1) in St. Margaret’s Bay, Nova Scotia. Top: Uncalibrated optical backscatter that is used to detect the watersediment interface. In raw data, the signal is delayed because the sensor is located 16.8 cm from the tip of the nose cone. Middle: Measured acceleration and computed velocity. Bottom: Dynamic pore pressure measured in the nose cone and hydrostatic pressure measured on the tail of the FFCPt.
• Figure 3: An acceleration sensor time series is integrated in order to create profiles of the FFCPt measurements as a function of depth.
• Figure 5: Left: Photograph of the electrical resistivity module used during the June 2002 seatrial. Right: Photograph of two other electrode configurations that have been tested insubsequent experimentation.
• Figure 6: FFCPt and sediment core locations in St. Margaret’s Bay, Nova Scotia, superimposed on the Atlas Hydrosweep MD50 swath bathymetry map [6].
• Figure 7: Left: Sediment behaviour type of FFCPt drop 71 at a clay/silt site in St. Margaret’s Bay, Nova Scotia. Right: Grain size analysis of Core 7, collected by the NRV Alliance.
• Figure 8: Left: Sediment behaviour type of FFCPt drop 31 at a sandy site in St. Margaret’s Bay,Nova Scotia. Right: Grain size analysis of Core 3, collected by the NRV Alliance.
• Figure 9: Left: Sediment behaviour type of FFCPt drop 62 at a site in St. Margaret’s Bay, Nova Scotia. Right: Grain size analysis of Core 6, collected by the NRV Alliance. The seabed has a surficial layer of sand underlain by finer grained material.
• Figure 10: FFCPt and sediment core porosities at locations 3, 7, and 6 (from left to right).

• Figure 1: Posidonia oceanica leaves. (a) Adult and intermediate leaves covered by epiphytes and encrustation. (b) Juvenile leaves and rhizome. (Photographs taken during USTICA’99 experiments at the end of the summer).
• Figure 3: Oxygen concentration vs. geotime at different water depths. The concentration is given in mg O2 / kg water (ppm). The profiles were measured over a Posidonia prairie of the Tuscan Archipelago on February 24 and 25, 1997. Sunrise and sunset were at 0706 and 1800 hours, local time.
• Figure 4: USTICA’99 test site. (a) Sediments, biocenosis, and stratigraphy at the test location.The thick black line shows the position of the acoustic section. S: source. R: receivers.Legend from top to bottom: Posidonia prairies; SuV: volcanic substratum; Gh: gravel;SaG: gravelly sand; Sa: sand; corded pahoehoe; pillow lava. (b) View of the bed of Posidonia oceanica (L.) Delile and fish school. Ustica island, off Sicily. September 1999.
• Figure 6: Time series of water oxygen content. The small blue circles are the raw data of oxygen concentration obtained from an oceanographic probe with the oxygen sensor located just above the foliage (left scale). The dashed red line is the variation of received energy for the set of acoustic paths with 7 surface-bottom reflections (right scale).See also Fig.9.1999.
• Figure 8: Geotime variability of the medium acoustic-impulse response (log envelope) in the time-delay range 1 to 290 ms. Each thin line is the difference between a 0.5-hour and the 3-night median averages. Blue lines: nighttime 07001930 hours, 4 days. Orange lines: day hours, 4 days. Green lines: most active photosynthesis 1300-1600 hours, 1 day. The vertical lines indicate the time window that corresponds to grazing angles in the range 40°-70°. Top hydrophone data. September 22-25, 1999.

• Figure 1: INTIFANTE’00 sea trial site bathymetry with XBT locations and track for Event 2.
• Figure 3: Temperature profiles measured during the INTIMATE’00 sea trial using XBTs, and empirical orthogonal functions obtained from the XBT temperature profiles (b).
• Figure 4: Focalization results obtained for Event 2.
• Figure 5: Eigenspectra computed from the cross-frequency correlation matrix (a); ratio between the two greatest eigenvalues e1/e2(b - top) and e2/e1(b-bottom).
• Figure 7: Estimated noise level as a function of time and frequency (a); eigenspectrum of a full rank noise covariance matrix as a function of time (b); average noise level estimated using all hydrophone with absolute tide force value over time (c); correlation of estimate noise level with absolute tide force (d).

• Figure 1: Schematic of Tank Used in Experiment.
• Figure 2: Experimental array geometry and raw data time series.
• Figure 3: Bottom of tank after smoothing.
• Figure 4: The Weiner filtered signal at 45 degrees grazing.
• Figure 5: Normalizing measurement from the air/water interface. The error bars indicate one standard deviation.
• Figure 6: Normalized reflection loss in dB from a sand/water interface as a function of frequency and angle.
• Figure 7: Results from inversion based on the fluid model for (a) sound speed, (b) attenuation,and (c) density. Dotted lines are the inversion results and dashed lines are the documented values.
• Figure 8: Comparison of measured reflection coefficient in black to values modeled with the inversion estimates based on the fluid model.
• Figure 9: Changes in reflection when (a) sound speed, (b) attenuation, and (c) density are varied individually. The solid line is the measured data. The dashed line is modeled with the expected parameter values. The dotted and dash-dot lines are the resulting reflection when significantly lower and higher values of each parameter are used in the modeling.
• Figure 10: Results from inversion based on the elastic model for (a) sound speed, (b) attenuation,(c) density, (d) shear sound speed, and (e) shear attenuation. Dotted lines are the inversion results and dashed lines are the documented values.
• Figure 11: Comparison of measured reflection coefficient in black to values modeled with the inversion estimates based on the elastic model.
• Figure 12: Results from inversion based on the Biot-Stoll model for (a) porosity, (b) grain density,(c) fluid density, (d) grain bullk modulus, and (e) uid bulk modulus. Dash-dot lines indicate the values determined by the inversion. Dashed lines indicate the expected values. 23
• Figure 13: Results from inversion based on the Biot-Stoll model for (a) viscosity, (b) permeability,(c) pore size, and (d) tortuosity. Dash-dot lines indicate the values determined by the inversion. Dashed lines indicate the expected values.
• Figure 14: Results from inversion based on the Biot-Stoll model for (a) real and (b) imaginary part of the shear bulk modulus, and (c) real and (d) imaginary parts of the frame bulk modulus. Dash-dot lines indicate the values determined by the inversion. Dashed lines indicate the expected values.
• Figure 15: Comparison of measured reflection coefficient in black to values modeled with the inversion estimates based on the Biot-Stoll model.
• Figure 16: Comparison of measured reflection coefficient (solid) to values modeled with the inversion estimates based on the fluid model (dotted), the elastic model (dash-dot),and the Biot/Stoll model (dashed).
• Figure 17: Comparison of measured reflection coefficient as a function of frequency and angle to the Biot/Stoll model inversion results. 24

• Figure 2: Marginal scatter diagrams of the SAGA search for the model parameters. The vertical axis represents the attained misfit on a linear scale. The thick line is the sensitivity curve of the multi-frequency misfit function using the best-fit model as a baseline.
• Figure 3: Marginal probabilities for water depth and bottom sound speed. The contour plot shows the 2D distribution. The red vertical arrow indicates the ML solution (sagabest fit). Note, that the maximum in the marginal distribution might not correspond to the ML maximum.
• Figure 4: Prior (top) and posterior (bottom) probability distributions for transmission loss versus range at 50-m depth.
• Figure 5: Posterior (solid) and prior (dashed) probabilities at 830-m range and 50-m depth.These correspond to a cut (white lines) though the contours in Fig. 4.
• Figure 6: Median transmission loss (dB) based on prior (top) and posterior (bottom) distributions of environmental parameters (water depth and sediment sound speed).
• Figure 7: Range (dB) between 5th and 95th percentiles of the transmission loss for prior (top)and posterior (bottom) probabilities.
• Figure 8: The median TL (red) at 50-m depth (dashed) based on all (a) prior samples and (b)posterior samples. The gray area indicates the range between the 5th and 95th percentiles.25

• Figure 1: The PlaneRay model computes the received field from a source at receivers located on a horizontal line. The sound speed profile is only function of depth. The bottom can be a fluid sedimentary layer over an elastic half space and both can be range dependent.
• Figure 2: Recording of the ranges to the given receiver depth intersection as function of the rays initial angle resulting from the initial ray tracing.
• Figure 3: The eigenray to the receiver at range r₀ is found by interpolating between the two rays arriving at the same receiver depth at ranges r₁ and r₂.
• Figure 4: Bottom reflection loss for a sediment layer with thickness D = 5 m and with density 1500 kg/m3 and sound speed 1700 m/s over a homogenous half space with density 2500 kg/m3 and compressional sound speed 4700 m/s and shear speed 2200 m/s.
• Figure 5: Frequency and time response of a Pekeris’ wave guide. Left: Transmission loss as function of range and frequency. Right: Time response for a number of receivers with distances from 100 meter to 10 km from the source. The source signal is a short transient (Ricker wavelet).
• Figure 6: Comparison of the transmission loss as function of range for different frequencies by PlaneRay (red line) and OASES (blue line) for Pekeris’ waveguide with homogeneous bottom in range independent environment.
• Figure 7:Transmission loss as function of range for the frequencies of 50, 100, 150 and 200 Hz for CASE = 1 in Table 2. The red curves are from the ray trace model, the blue curves are the OASES results.
• Figure 8: The bottom structure for the CASE 1, 2 and 3. The parameter values given in the figure are the same for all cases, the other parameters mare given in Table 2.
• Figure 9:Transmission loss as function of range for the frequencies of 50, 100, 150 and 200 Hz for CASE = 1 in Table 2. The red curves are from the ray trace model, the blue curves are the OASES results.
• Figure 10: Transmission loss as function of range for the frequencies of 100 and 200 Hz. The red curves are from the ray trace model, the blue curves are the OASES results.
• Figure 11: Transmission loss as function of range for the frequencies of 100 and 200 Hz. The red curves are for the high shear speed of 2200 m/s and the blue curves are for he lower shear speed of 2000 m/s. All results are from the ray trace model.
• Figure 12: Sound speed profile (left), and (right) bottom topography and the rays that contribute to the sound field at a receiver located at range of 2000 m and depth of 25 m.
• Figure 13: Simulated time responses as function of range for the sloping bottom case of Figure12. The source signal is a short transient (Ricker wavelet).
• Figure 14: Transmission loss in dB as function of range for the selected frequencies 50, 100, 200 and 400 Hz. The red curves are calculated by PlaneRay, the blue by the propagation model RAM using the parabolic approximation.
• Figure 15: Comparison between the modeled results from PlaneRay (left panel) and the measured results from ARL East Coast Data for 52 receivers spanning the range from 650 m to 1200 m (right panel).

• Figure 5: (a)Measured bottom loss (BL).
• Figure 5: (b) Modelled BL using geo-acoustic parameters from [1].
• Figure 5: (c) Deviation of modelled BL (b) from measured BL.
• Figure 5: (d) Modelled BL using DE-inverted geo-acoustic parameters (single sediment).
• Figure 5: (e) Deviation of modelled BL (d) from measured BL.
• Figure 5: (f) Modelled BL using DE-inverted geo-acoustic parameters (two sediments).
• Figure 5: (g) Deviation of modelled BL (f) from measured BL.
• Figure 6:Two-sediment inversions of the synthetic data: histogram of the final energy function values obtained for the 171 successful inversions.
• Figure 7:Two-sediment inversions of the synthetic data: magnitude of the correlation coefficient for all parameter combinations. A star is plotted if the correlation is statistically significant.
• Figure 10: Two-sediment inversions of the North Elba data: histograms of the 114 estimates for each parameter. Note that we included the histogram for total sediment thickness.
• Figure 11: Two-sediment inversions of the North Elba data: The 114 estimates for h₂ plotted against those for h₃. The selected solutions are indicated by the red symbols.
• Figure 12: Two-sediment inversions of the North Elba data: histograms of the parameter estimates,which were selected on total sediment thickness.
• Figure 13: Two-sediment inversions of the North Elba data: magnitude of the correlation coefficient for the selected solutions. A star is plotted if the correlation is statistically significant.

• Figure 1: Modelled reflection loss for two sediment layers (thicknesses 1.5 and 2m) in between two half spaces.
• Figure 2: (a) The inverse Fourier transform of the complex reflection coefficient (i.e. the impulse response) from Fig.1. Two-way travel time is converted to an equivalent depth at 1500 m/s.
• Figure 2: (b) The inverse Fourier transform of the modulus-square of reflection coefficient, which is the autocorrelation of the impulse response.
• Figure 2: (c) The result of spectral factorization applied to the modulus of the reflection coefficient.
• Figure 3:Experimental impulse response derived by spectral factorization from ambient noise measured on a moored VLA at a mud site east of Elba.
• Figure 4: Drift tracks in the vicinity of the Malta Plateau and the Ragusa Ridge during BOUNDARY 2002, 2003, 2004.
• Figure 5: A selection of reflection loss plots during the drift experiment. Times are: 20:03,00:33, 03:01, 06:04, corresponding approximately to -4, 0, 3, 6 (hr) in Fig.6.
• Figure 6: (a) Sub-bottom profile derived, via reflection loss, from ambient noise received by a drifting VLA. (b) Zero-referenced and aligned boomer sub-bottom profile for comparison.30

• Figure 3: The position of ship (red circle) and array (yellow circle) on the experimental track.The range at this position was 3.3 km.
• Figure 4: A sample of ship noise data from the experimental track. The low frequency band,73–113 Hz, is highlighted in red.
• Figure 6: Ambiguity surface windows showing the focus of back-propagated signal near the source location for the high frequency band.
• Figure 7: Focal function for the first stage of the inversion to estimate water depth and sediment sound speed. The estimated values corresponding to the minimum at search index 18 are 400 m and 1550 m/s for the water depth and sound speed, respectively.
• Figure 8: Focal function for the second stage of the inversion to estimate the thickness of the first sediment layer using the high frequency band. The dynamic range is small, and the minimum value is ambiguous between about 80 and 130 m.
• Figure 9: Ambiguity surface windows showing the focus of back-propagated signal near the source location for the second stage of the inversion using the low frequency band.
• Figure 10: Focal function for the second stage of the inversion to estimate the sediment layer thickness using the ship noise in the low frequency band. The estimated value corresponding to search index 6 is 100 m.
• Figure 11: The ambiguity surface for a 45-Hz tonal, calculated using the estimated geoacoustic profile. The maximum value occurs at 3.6 km range and 30 m depth.

• Figure 1: Concept of MFP.
• Figure 2: Spectrogram of humpback whale call used in MFP inversion. Note broadband energy content between 50-800 Hz.
• Figure 3: Autonomous recorder used in vertical array deployment.
• Figure 4: Plot of correlation between measured data and modeled acoustic field, as a function of range and depth, using 12 frequencies spaced between 100 and 1000 Hz.

• Figure 2: Comparison of measured and modeled time series in 50-500 Hz band.
• Figure 3: Comparison of measured and modeled TL.
• Figure 5: Comparison of measured and modeled time series for upslope propagation.
• Figure 6: Comparison of measured and modeled time series for downslope propagation.
• Figure 7: Comparison of analytic solution at 10 Hz and 3-D coupled-mode solution for infinite wedge.
• Figure 8: Demonstration of the effects of surface roughness added to the infinite wedge problem.34