Layered Velocity Model Research Articles

Monte Carlo approach to assess the uncertainty of wide-angle layered models: Application to the Santos Basin, Brazil

In the Santos Basin (Brazil), two parallel wide-angle refraction profiles show different crustal structures. One shows moderate crustal velocity gradient, and a clear Moho with topography. The other has an anomalous velocity zone, and no clear Moho reflections. This has large implications on the geological and geodynamical interpretation of the basin. Model uncertainties must be excluded as a source of these differences.We developed VMONTECARLO, a tool to assess model uncertainty of layered velocity models using a Monte Carlo approach and simultaneous parameter perturbation using all picked refracted and reflected arrivals. It gives insights into the acceptable geological interpretations allowed by data and model uncertainty through velocity-depth plots that provide: a) the velocity-depth profile range that is consistent with the travel times; b) the random model that provides the best fit, keeping most of the observations covered by ray-tracing; c) insight into valid models dispersion; d) main model features unequivocally required by the travel times, e.g., first-order versus second-order discontinuities, and velocity gradient magnitudes; e) parameter value probability distribution histograms.VMONTECARLO is seamlessly integrated into a RAYINVR-based modelling work-flow, and can be used to assess final models or sound the solution space for alternate models, and is also capable of evaluating forward models without the need for inversion, thus avoiding local minima that may trap the inversion algorithms and providing information for models still not well-parametrised.Results for the Brazilian models show that the imaged structures are indeed geologically different and are not due to different interpretations of the same features within the model uncertainty bounds. These differences highlight the strong heterogeneity of the crust in the middle of the Santos Basin, where the rift is supposed to have failed.

Microseismic event location using the first and reflected arrivals

Locating microseismic events accurately is essential to characterizing hydrofracked reservoirs. Several location approaches have been developed to solve this problem. We have developed a global grid search method based on the Bayesian probabilistic approach. In addition to retrieving the global minimum of the cost function, it enabled the computation of the uncertainty that was essential to quantify the quality of the results. The location uncertainty was tightly related to the amount and the quality of observed data used to perform the location. Using multiple arrivals increased the quantity of information collected in a cost-effective manner (no additional receivers were required). In this location algorithm, we have used the first and reflected arrivals to constrain the locations. To locate the events, it was essential to model the first and reflected arrivals given a velocity model and acquisition geometry. We have developed a ray-tracing algorithm capable of computing traveltimes and polarization of direct, refracted, and reflected waves in layered velocity models. We have applied this methodology for several examples. The synthetic data set was generated using full-waveform finite-difference modeling and recorded in a single vertical well. The first and reflected traveltimes were picked using the 3C seismograms. In the last example, we used the Cotton Valley real data set in which reflections could be seen for strong events due to the high velocity contrasts. The applications revealed that the location uncertainty was significantly reduced when using both arrivals compared with the case when only first arrivals were used. Combining the first and reflected arrivals even allowed us to improve the location accuracy when the velocity model was inaccurate.

Independent estimate of velocity structure of Earth's lowermost outer core beneath the northeast Pacific from PKiKP − PKPbc differential traveltime and dispersion in PKPbc

AbstractThe presence of a low‐velocity layer at the base of Earth's outer core has been proposed. However, the seismic profile of the basal layer indeed has been poorly constrained. In previous seismic studies the model parameters of the layer are substantially nonunique and there are tradeoffs between the seismic velocity of the layer and the properties of the mantle and inner core. A more tightly constrained profile of the layer helps further examine the composition and dynamics of the layer. In this study we obtained the P wave velocity profile of the basal layer beneath the northeast Pacific based on two new seismic observations by analyzing seismograms of the Hi‐net in Japan. The new observations are particularly sensitive to the layer structure and are relatively insensitive to the structure of the other parts of the Earth: (1) the frequency dispersion in P waves that graze or are diffracted at the inner core boundary (PKPbc and PKPc‐diff) and (2) differential traveltimes between the P waves reflected from the inner core boundary (PKiKP) and those that turn above the boundary (PKPbc). The resulting velocity model of the lowermost outer core (called “F layer velocity model for the Western Hemisphere” (FVW)) has P wave velocities that lie between those of AK135 and the preliminary reference Earth model (PREM), and a velocity gradient that is slightly gentler than that of PREM. Models with a uniform P wave velocity value within the layer are not supported by the observations for the region investigated, which appears not to support the presence of a thick basal layer that is Fe rich and dense there.

Significance of Geological Units of the Bohemian Massif, Czech Republic, as Seen by Ambient Noise Interferometry

Broadband recordings of 88 seismic stations distributed in the Bohemian Massif, Czech Republic, and covering the time period of up to 12 years were processed by a cross-correlation technique. All correlograms were analyzed by a novel approach to get both group and phase dispersion of Rayleigh and Love waves. Individual dispersion curves were averaged in five distinct geological units which constitute the Bohemian Massif (Saxothuringian, Tepla-Barrandean, Sudetes, Moravo-Silesian, and Moldanubian). Estimated error of the averaged dispersion curves are by an order smaller than the inherent variability due to the 3D distribution of seismic velocities within the units. The averaged dispersion data were inverted for 1D layered velocity models including their uncertainty, which are characteristic for each of the geological unit. We found that, overall, the differences between the inverted velocity models are of similar order as the variability inside the geological units, suggesting that the geological specification of the units is not fully reflected into the S-wave propagation velocities on a regional scale. Nevertheless, careful treatment of the dispersion data allowed us to identify some robust characteristics of the area. The vp to vs ratio is anomalously low (~1.6) for all the units. The Moldanubian is the most rigid and most homogeneous part of the Bohemian Massif. Middle crust in the depth range of ~3–15 km is relatively homogeneous across the investigated region, while both uppermost horizon (0–3 km) and lower crust (>15 km) exhibit lower degree of homogeneity.

Improvements on particle swarm optimization algorithm for velocity calibration in microseismic monitoring

In this paper, we apply particle swarm optimization (PSO), an artificial intelligence technique, to velocity calibration in microseismic monitoring. We ran simulations with four 1-D layered velocity models and three different initial model ranges. The results using the basic PSO algorithm were reliable and accurate for simple models, but unsuccessful for complex models. We propose the staged shrinkage strategy (SSS) for the PSO algorithm. The SSS-PSO algorithm produced robust inversion results and had a fast convergence rate. We investigated the effects of PSO’s velocity clamping factor in terms of the algorithm reliability and computational efficiency. The velocity clamping factor had little impact on the reliability and efficiency of basic PSO, whereas it had a large effect on the efficiency of SSS-PSO. Reassuringly, SSS-PSO exhibits marginal reliability fluctuations, which suggests that it can be confidently implemented.

Six-degree-of-freedom near-source seismic motions I: rotation-to-translation relations and synthetic examples

The paper deals with theoretical aspects of rotation-to-translation relations in six-degree-of-freedom short-period seismic records at close hypocentral distances. Rotation-to-translation ratios are introduced as the ratios relating peak amplitudes of the relevant rotational and translational components. Their frequency dependence is analyzed in simple models. The relations between translations and rotations are expressed by equations derived under the assumption of a spherical S wave radiated from a shallow point source in a homogeneous medium. A set of numerical experiments is performed to examine these relations for a double-couple source buried in simplified structure models. These experiments indicate that at local distances (up to several km), at lower frequencies (up to a few Hz), and at locations with rapid amplitude changes due to the radiation pattern (e.g., in the vicinity of nodal planes), the rotational components are a linear combination of terms proportional to translational velocity and acceleration, and none of the terms can be, in general, neglected. We also focus on the possibility to retrieve the S-wave phase velocity along the wavepath. The applicability of the equations is tested also in a layered velocity model. It has been found that, under certain conditions, the equations allow us to find correct values of wavepath velocity even in vertically inhomogeneous structures. The depth-range sensitivity is examined for a specific 1D model containing a thin surficial low-velocity layer. Based on these experiments, we have concluded that the retrieved velocity is representative down to the depth not exceeding one wavelength.

Mapping Spontaneous Ground Fissures using Seismic Refraction, Reflection, and Ground Penetrating Radar in the Letlhakeng Area, Kweneng District, Botswana

Infrastructure in the Letlhakeng area is currently being damaged by ground fissures. Geophysical profiles of seismic refraction and reflection, and ground penetrating radar (GPR) were recorded at the areas affected by the ground fissures to map their depth extent and subsurface configuration. The GPR survey used 50 MHz antennas with an inline transmitter and receiver 2 m apart and a sample spacing of 30 cm. A reversed seismic refraction profile with two offset shots was recorded. Twenty four geophones were placed 2 m apart to make a total profile length of 46 m. The seismic reflection data were recorded with an equal source and receiver interval of 1 m and a total spread length of 23 m across a ground fissure. The GPR profile shows that the ground fissures are sub-vertical and extend to a vertical depth of ~5.0 m. The seismic refraction profile shows a two layer velocity model comprising overburden with a seismic velocity of 400-600 m/s lying over a substratum with a seismic velocity of 1200-1400 m/s. The top layer consists of the loose sand which is imaged to a depth of 5 m. The seismic reflection profile shows that the ground consists of subhorizontal layers from 0-15 m depth, which have all been broken by the ground fissure observed at the surface. The geophysical results show that the ground fissures persist from the surface to depths of 15 m.

H/V spectral ratio analysis and Rayleigh modelization in Eastern Thuringia, Germany

We use records from the East Thuringian Seismic Network (OTSN, Ostthüringer Seismisches Netzwerk) to characterize the site response for each station, and to analyze the scope and limits of the Rayleigh modeling for H/V spectral ratio. The stations considered in this work can be classified by their seismic response as hard rock sites or as sites with some site effect. From these results we propose velocity models based on Rayleigh modeling (theoretical Rayleigh wave ellipticity).Our results show that for locations affected by site effects the H/V spectral ratio can be modeled by the theoretical ellipticity of layered velocity models. For hard rock sites the spectral ratio is rather flat and the modeling with the theoretical ellipticity was not very clear. This may be explained by the fact that for hard rock sites the conditions for a clear fundamental frequency associated with S-wave resonance, and therefore with Rayleigh wave ellipticity, are not fulfilled.

Accurate Location of the Lushan, Sichuan M7.0 Earthquake on 20 April 2013 and Its Aftershocks and Analysis of the Seismogenic Structure

AbstractThe M7.0 Lushan earthquake and its aftershocks in 9 days were relocated by the double difference algorithm and a layered velocity model, using the seismic phases recorded at stations within 150 km including the fixed stations of the Sichuan earthquake network center, temporary stations, and reservoir stations. The events analyzed include the main shock and its 3324 aftershocks. The results show that the origin time of the main shock is 8:02:46.8 on 20 April 2013, hypocenter location 30.278°N, 102.989°E and the focal depth 16.67 km. The main rupture is about 40 km long, 20 km wide with an apparent area 800 km2. It strikes in southwest and dips about 40°. Most aftershocks occurred at depths 10∼22 km, concentrating on the hanging wall of the Dayi‐Mingshan fault, forming a southwest‐trending belt on the surface. Based on these and other data, this paper makes a preliminary analysis of the seismogenic structure of the Lushan event.

Modeling Geometrical Spreading and the Relative Amplitudes of Vertical and Horizontal High-Frequency Ground Motions in Eastern North America

Abstract Horizontally layered velocity models were used with point‐source and finite‐fault sources to investigate geometrical spreading and the relative amplitudes of vertical and horizontal ground acceleration within 120 km of the source. Full‐wave‐field simulations were done for a range of focal depths and for strike‐slip and reverse focal mechanisms. The attenuation of the geometric mean of randomly oriented horizontal‐component maximum acceleration amplitudes, averaged over all azimuths, significantly exceeds the theoretical geometrical spreading for far‐field body waves in a homogeneous whole space for hypocentral distances less than approximately 60 km. The behavior of the vertical component is different from the horizontal: vertical attenuation near the epicenter is greater and is more dependent on source mechanism and depth. Because of the rapid near‐source decay of the direct S wave, reflections from the mid‐lower crust and Moho control the maximum amplitude of the vertical‐component acceleration in the 60–120‐km hypocenter distance range, resulting in a flattening of the vertical amplitude‐distance relation. Near‐source vertical maximum amplitudes averaged over all source–receiver azimuths tend to be less than the geometric mean horizontal amplitude for strike‐slip focal mechanisms, but, near the source for reverse faults, the azimuthally averaged vertical‐component amplitude exceeds that of the geometric mean horizontal. The modeling indicates that similar vertical‐ and horizontal‐component geometrical spreading and approximately constant horizontal/vertical amplitude ratios observed in connection with the Lg phase at distances greater than approximately 100 km in eastern North America may not hold at smaller distances. Ground‐motion prediction models for the vertical component near the source may need to incorporate strong geometrical spreading and dependence on radiation pattern.

Study on Estimation of Subsurface Velocity Structure Model using Waveform Inversion

The estimation of layered velocity model by waveform inversion using observed ground motion records is one of ways for modeling deep subsurface structure for the numerical simulation of long-period ground motion. The effectiveness and problem of the estimation by waveform inversion are examined. The numerical test of the waveform inversion using simulated waveform gives good estimation of velocity model assumed beforehand, without constraints of number of layers and velocity of each layers. The case study of application to strong ground motion records at OJIYA, K-NET is conducted. The estimated velocity model by waveform inversion corresponds to the estimated velocity models based on geophysical exploration, geomorphology and geology in previous research. The effect of spatial irregularity of subsurface structure is examined using FDM simulation and waveform inversion. In waveform inversion, the waveform affected by spatial irregularity of subsurface structure gives distorted estimations of the velocity structure. In the region around epicenter, the waveform is not affected by spatial irregularity of subsurface structure so much. In the estimation of velocity model, therefore, the effect of spatial irregularity is small in the region.

High accuracy wave simulation – Revised derivation, numerical analysis and testing of a nearly analytic integration discrete method for solving acoustic wave equation

The nearly analytic integration discrete (NAID) method for solving the two-dimensional acoustic wave equation has been fully mathematically revised, analyzed and tested. The NAID method is an alternative numerical modeling method for generating synthetic seismograms. The acoustic wave equation is first transformed into a system of first-order ordinary differential equations (ODEs) with respect to time variable t, and then directly integrated at a small time interval of [tn,tn+1] to obtain semi-discrete ordinary differential equations. The integral kernel is expanded into a truncated Taylor series, to which the integration operator is explicitly applied. The high-order temporal derivatives involved in the integral kernel are replaced by high-order spatial derivatives, which then are approximately calculated as a weighted linear combination of the displacement, the particle-velocity, and their spatial gradients. In this article, we investigate the theoretical properties of the revised NAID method, including the discrete error and the stability criteria. Numerical results for constant and layered velocity models show that, comparing to the Lax–Wendroff correction (LWC) scheme and the staggered-grid finite difference method, the NAID method can effectively suppress the numerical dispersion and source-noises caused by the discretization of the acoustic wave equation with too-coarse spatial grids or when models have strong velocity contrasts between adjacent layers. The proposed NAID method has been applied in computing the acoustic wavefields for two heterogeneous models – the corner edge model and the Marmousi model. Promising numerical results illustrate that the NAID method provides an encouraging tool for large-scale and complex wave simulation and inversion problems based on the acoustic equation.

Reliability of the automatic procedures for locating earthquakes in southwestern Alps and northern Apennines (Italy)

Reliable automatic procedure for locating earthquake in quasi-real time is strongly needed for seismic warning system, earthquake preparedness, and producing shaking maps. The reliability of an automatic location algorithm is influenced by several factors such as errors in picking seismic phases, network geometry, and velocity model uncertainties. The main purpose of this work is to investigate the performances of different automatic procedures to choose the most suitable one to be applied for the quasi-real-time earthquake locations in northwestern Italy. The reliability of two automatic-picking algorithms (one based on the Characteristic Function (CF) analysis, CF picker, and the other one based on the Akaike’s information criterion (AIC), AIC picker) and two location methods (“Hypoellipse” and “NonLinLoc” codes) is analysed by comparing the automatically determined hypocentral coordinates with reference ones. Reference locations are computed by the “Hypoellipse” code considering manually revised data and tested using quarry blasts. The comparison is made on a dataset composed by 575 seismic events for the period 2000–2007 as recorded by the Regional Seismic network of Northwestern Italy. For P phases, similar results, in terms of both amount of detected picks and magnitude of travel time differences with respect to manual picks, are obtained applying the AIC and the CF picker; on the contrary, for S phases, the AIC picker seems to provide a significant greater number of readings than the CF picker. Furthermore, the “NonLinLoc” software (applied to a 3D velocity model) is proved to be more reliable than the “Hypoellipse” code (applied to layered 1D velocity models), leading to more reliable automatic locations also when outliers (wrong picks) are present.

Reflection seismic waveform tomography

In seismic waveform tomography, if using reflection data with limited source‐receiver offsets, it is difficult to reconstruct the deep part of the subsurface velocity model. We present two approaches to tackle this problem: layer stripping and weighted updating. In a layer‐stripping procedure, we replace the top portion of seismic data with synthetics generated from the previous‐layer inversion and make the current inversion focus on the minimization of the data misfit corresponding to the deep part of the model. To improve efficiency, we use only sparsely sampled frequency data in the deeper‐layer inversions, unlike the first‐layer inversion where we use densely sampled frequency data as usual. The sparsely sampled frequencies together have the full wave number coverage for effective imaging. Combined use of dense and sparse sampling in frequency is a compromise between resolution and efficiency, as it reduces the number of iterations needed in layer‐stripping inversion while still producing a good image. In the second scheme, we apply depth‐dependent weights to model updates in order to improve the convergence in an iterative solution. The weighting is inversely proportional to the ray density variation along the depth and is mathematically equivalent to the application of an inverse Hessian matrix which sharpens the gradient vector for model updating. For real seismic data, we transfer point source shot records to line source records, by partial amplitude compensation and phase adjusting, before inputting it to the waveform tomography. We perform traveltime inversion to generate a reliable layered velocity model and then waveform tomography to produce a high‐resolution image of the subsurface model through frequency domain iteration.

Kinematically equivalent velocity distributions

For a layered medium, the seismic velocity model can be vertically heterogeneous within the layers. The traveltime parameters estimated from each reflection must be converted into layer traveltime parameters by using the layer-stripping method. The layer traveltime parameters must be inverted into layer velocity model parameters. Interpretation or inversion of layer traveltime parameters depends on the chosen velocity model within the layer. Different or kinematically equivalent velocity distributions can result in the same traveltime parameters. The inversion problem for traveltime parameters is strongly nonunique even if they are estimated accurately. To evaluate the accuracy of a velocity model, one can choose the phase for the two-way propagator. The discrepancy in this phase factor between the kinematically equivalent velocity models depends on the number of traveltime parameters estimated and increases with spatial frequency. By estimating two traveltime parameters, we approximately preserve the average velocity, regardless of the complexity of the vertically heterogeneous model. By estimating three traveltime parameters, we approximately preserve the average velocity gradient.

Nonlinear 3D tomographic least-squares inversion of residual moveout in Kirchhoff prestack-depth-migration common-image gathers

Velocity-model estimation with seismic reflection tomography is a nonlinear inverse problem. We present a new method for solving the nonlinear tomographic inverse problem using 3D prestack-depth-migrated reflections as the input data, i.e., our method requires that prestack depth migration (PSDM) be performed before tomography. The method is applicable to any type of seismic data acquisition that permits seismic imaging with Kirchhoff PSDM. A fundamental concept of the method is that we dissociate the possibly incorrect initial migration velocity model from the tomographic velocity model. We take the initial migration velocity model and the residual moveout in the associated PSDM common-image gathers as the reference data. This allows us to consider the migrated depth of the initial PSDM as the invariant observation for the tomographic inverse problem. We can therefore formulate the inverse problem within the general framework of inverse theory as a nonlinear least-squares data fitting between observed and modeled migrated depth. The modeled migrated depth is calculated by ray tracing in the tomographic model, followed by a finite-offset map migration in the initial migration model. The inverse problem is solved iteratively with a Gauss-Newton algorithm. We applied the method to a North Sea data set to build an anisotropic layer velocity model.

Distribution of local earthquakes in the crust beneath central Lake Baikal

We have located earthquake hypocenters in the area of central Lake Baikal using the Hypoellipse inversion of direct P traveltimes collected by a local seismological network of seven stations in 2001 through 2005. The location accuracy depends on the agreement between the reference velocity model and the real subsurface. Traveltime residuals have no bearing on the accuracy of depth estimates due to origin time errors. Hypocenter location with reference to a DSS-derived layered velocity model predicts that the crust is seismogenic between 10 and 22 km. The hypocenters cluster densely in belts trending in NE and SE directions along and across Lake Baikal. Earthquakes become deeper from 10 to 22 km toward the Selenga basin.

Anomalous Pn Waves Observed in Eastern Taiwan: Implications of a Thin Crust and Elevated Oceanic Upper Mantle beneath the Active Collision-Zone Suture

Normal Pn waves are commonly observed in Taiwan from shallow regional earthquakes at epicentral distances larger than 120 km, similar to the observations in many other continental regions. However, the critical distances to observe Pn waves for shallow eastern Taiwan earthquakes vary with azimuth corresponding to a significant variation of crustal thickness. In particular, anomalous Pn waves are commonly observed for shallow eastern Taiwan earthquakes recorded on seismic stations at epicentral distances as small as 60 km along the collision zone suture, the Longitudinal Valley. For the same event, normal Pn waves are observed at other seismic stations elsewhere on the island. The apparent velocity of the anomalous and normal Pn waves from the same event is 7.8 ± 0.15 km/sec, which is consistent with the average Pn velocity in the Taiwan area. Thus, the unusually short critical distance for Pn waves in eastern Taiwan suggests that the crust beneath the collision zone suture must be very thin and the upper mantle beneath the Longitudinal Valley must be relatively elevated compared with that beneath the other parts of Taiwan. Assuming a simple 1D layered velocity model, the Moho depth beneath the suture zone can thus be estimated at ∼23 ± 2 km. This observation is consistent with the recent report from a high-resolution 3D tomographic inversion that a narrowly confined, anomalously elevated, and north-northeast–south-southwest elongated oceanic upper mantle was imaged beneath the Longitudinal Valley from Hualien in the north to Taitung in the south (Kim et al. , 2005, 2006). Furthermore, the preceding observations may also support the interpretation that the conduction of excess heat supply from the elevated hot oceanic upper mantle into the adjacent mid- to-lower continental crust over a long period of geological time may play an important role in the crustal deformation beneath the continent, including metamorphism, thickening, and uplifting.

Plus�minus method to solve large�amplitude near�surface static corrections

ABSTRACTThe surface conditions in Saudi Arabia demonstrate significant variability ranging from flat gravel plains to sand�dunes, sabkhas (salt flats) and desert wadis (washouts). Saudi Aramco has used a single�layer velocity model from surface to datum for most of the prospective areas for both 2D and 3D seismic data. Source gathers have been interpreted by picking the first arrivals to obtain a linear fit for the direct and refractor arrivals using the intercept�time method. This creates depth and velocity control input for a multiple‐layer model of the near‐surface, from which static corrections to datum are calculated. On 2D lines with very large amplitude near‐surface anomalies, the plusminus technique has been applied to solve local areas. The implementation of the plus‐minus technique utilizes a variable near‐surface velocity derived from either upholes or the direct‐arrival velocity determined from source gathers. With this method, multiple source‐pairs build up effective spreads. Then by compiling high fold of delay‐times at a location, a statistical average is taken where the high fold reduces the significance of picking errors. The model is built utilizing multiple refraction layers and is referenced to the datum, thus tying it to the regional model.The example 2D seismic line has a static model calculated from the single‐layer velocity model. The intercept‐time method was applied locally over a near‐surface anomaly caused by unconsoli‐dated material at the foot of a cliff. Further analysis of the near‐surface involved the plus‐minus method to solve the high amplitude time shifts.

Seismic characterization of a gas hydrate system in the Gulf of Mexico using wide-aperture data

SUMMARY Gas hydrates were discovered in a mud mound in the lease block Mississippi Canyon 798, Gulf of Mexico, through piston coring. Subsequently, a seismic experiment was carried out to investigate the dynamics behind the hydrate formation. During the experiment, high-resolution multichannel seismic reflection data using a 24-channel, 240 m long streamer and wide-aperture data using six ocean bottom seismometers were collected along five lines. High-reflectivity zones (HRZs) are present in the reflection data along all lines. To better constrain the interpretation of the reflection data, the traveltimes from the multichannel and wide-aperture data sets were jointly inverted to estimate a 2-D P-wave layered velocity model for each line. A minimum-parameter/minimum-structure modelling approach yielded simple models and a comparison of the models at their intersection points shows they are consistent to within ±10 m s −1 in velocity and ±20 m in depth. In the final P-wave velocity models, the HRZs are associated with a lowering of velocity. In the reflection data, the top of the HRZs show a polarity reversal with respect to the seafloor. Presence of free gas in the HRZs best explains the velocity lowering and polarity reversal. It is speculated that the gas has deeper sources and migrates upwards through conduits formed by salt movement in the vicinity. The upward migrating gas accumulates in the axis of a channel complex and manifests itself as HRZs in the reflection data. The fluids circulating along the conduits push the base of the hydrate stability zone close to the seafloor. From the channel axis, the free gas migrates further upwards and close to the seafloor, and as it comes within the gas hydrate stability zone, it forms hydrates.