Natural gas hydrates, composed primarily of water andmethane, are solid, crystalline, ice-like substances found inpermafrost areas and deepwater basins around the world.As the search for oil and gas extends into ever-deeper waters,particularly within the northern Gulf of Mexico, gas hydratesare becoming more of a focus in terms of both safety and asa potential energy resource.

Locating likely areas of gas hydrates using remote seis-mic sensing is relatively straightforward in many parts ofthe world where bottom-simulating reflectors (BSR) arereadily evident. A BSR is a high-amplitude reflector thatapproximately parallels the seafloor, and which results fromthe strong acoustic impedance contrast between the gashydrate-bearing sediments above the reflector and theunderlying sediments containing free gas. Because the BSRfollows a thermobaric surface rather than a structural orstratigraphic interface, it is normally observed to crosscutother reflectors. Locating gas hydrates in the Gulf of Mexico,however, is much more challenging. Rarely has there beena documented case of a BSR in the Gulf of Mexico. Thereare many theories as to why this is the case. One thought isthat the GOM sediments are too chaotic and heterogeneousto observe a BSR. Others believe that BSRs do exist in theGOM, but are largely undetectable due to the inadequaciesof current seismic data.

In this article, we discuss the detection and estimationof gas hydrates using seismic data prior to drilling. Theabsence of well logs and other hard data presented a keychallenge in the study. This was overcome by using a five-step integrated multidisciplinary approach that included:(1) reprocessing conventional 3D seismic data at the higherresolution using an amplitude-preserving flow with prestacktime migration, (2) a detailed stratigraphic evaluation andinterpretation to identify potential hydrate zones, (3) seis-mic attribute analysis to further delineate anomalous zones,(4) full waveform prestack inversion to characterize acousticproperties of gas hydrates in 1D (Mallick, 1995) and map in3D using a hybrid inversion technique (Mallick et al., 2000),and (5) quantitative estimation of gas hydrate saturationusing rock property models (Figure 1). In this article, we willfocus on discussing rock physics modeling, seismic inver-sion, and gas hydrate quantification in the northern deep-water GOM.

Stratigraphic evaluation. Two OCS blocks, one in KeathleyCanyon and the other in Atwater Valley, were selected fordetailed stratigraphic evaluation. In the Keathley Canyonstudy area, a BSR is identifiable that obliquely cuts the strati-graphic reflections approximately 500 ms below the seafloor(Figure 2). This crosscutting feature can be seen using reflec-tion strength on a vertical seismic section and is definableby terminations of the high-amplitude gas sands below(Figure 3). In addition to the BSR, this area also shows ahydrate mound on the east side of the major fault ridge.Below this mound is evidence of free gas accumulation anda possible destabilized BSR near the surface. This mound isdirectly adjacent to one of the major faults, a likely conduitfor gas and fluids (Figure 2).

60 THE LEADING EDGE JANUARY 2004

Detection and estimation of gas hydrates using rock physics andseismic inversion: Examples from the northern deepwater Gulf ofMexicoJIANCHUN DAI, HAIBIN XU, FRED SNYDER, and NADER DUTTA, Schlumberger Reservoir Services/Data and Consulting Services, Houston, Texas, U.S.

Figure 1. Five-step process for gas hydrate detection and estimation usingseismic.

Figure 2. Seismic section with stratigraphic interpretation at KeathleyCanyon with BSR indicated by yellow dotted line.

Figure 3. BSR defined by the termination of bright sands as indicated bythe negative amplitude that is caused by the free gas in the pore space.

The Atwater Valley study area is within the MississippiValley channel complex and, therefore, has a thick clasticblanket above the salt. Sediments deposited during thisactive period are complex and chaotic with evidence ofmany channel levee and slope fan systems (Figure 4). Noregional BSR is evident in this area, although numerousseafloor mounds can be seen with amplitude wipeout zonesextending about 0.4 s below the mudline (Figure 5). Fromheat flow and thermal gradient estimates, the hydrate sta-bility zone is thought to extend approximately 500-1000 mbelow the mudline.

Rock physics of gas hydrates. The presence of a BSR,seafloor mounds, amplitude blanking, or other gas hydrateindicators cannot positively confirm the existence ofhydrates. To better determine the existence of gas hydrate,and to quantify actual saturation, elastic property inversionis first performed using high-quality seismic data. This isfollowed by rock physics inversion to further transform theelastic properties into gas hydrate saturation estimates.

Results of recent gas hydrate drillings worldwide, suchas the Mallik 2L-38 well in Northern Canada and ODP Leg164 wells at Blake Ridge on the Atlantic coast, have demon-strated a consistent relationship between the rock elasticproperties and gas hydrate saturations in the sediments.Higher gas hydrate concentrations create an increase in theelastic properties. There are a number of rock physics mod-els in the literature that attempt to quantify this effect (Figure6). The cementation models of Dvorkin and Nur (1996) treatthe grains as randomly packed spheres where the gashydrates occur at the contact point (model 1) or grow aroundthe grains (model 2). However, these models predict largeincreases in the elastic properties with only a small amountof gas hydrate but stay relatively flat as the concentrationof gas hydrate increases further. Models 3 and 4 are varia-tions of the cementation models, but consider the gas hydrateas either a component of the load-bearing matrix or fillingthe pores (Dvorkin et al., 1999; Helgerud et al., 1999). Models3 and 4 use the Hertz-Mindlin contact theory to calculatedry rock moduli at critical porosity (35-40%). A modifiedlower Hashin-Shtrikman (HS) bound is used for porositysmaller than critical porosity, and a modified upper HSbound is used for porosities larger than critical porosity. TheGassmann equation is then used to derive the composite rockvelocities. Model 5 is an inclusion-type model that treats gashydrate and grains as the matrix and inclusions respec-tively, solving for elastic moduli of the system by iterativelysolving either the inclusion-type or self-consistent type equa-tions. Models 1-5 all consider gas hydrate as homogeneouslydistributed in the sediments. However, evidence of gashydrate coring reveals that hydrates often exist as nodulesand fracture fillings in the shallow shaly sediments. Thisgeometry is illustrated in model 6. No quantitative treatmentof this geometric model exists in the literature. Not illus-trated in Figure 6 is a series of empirical relations to describethe acoustic properties of gas hydrates (e.g., the weightedaverage equation by Lee, 1996). The advantage of an empir-ical relation is that it is based upon real observations, andvery simple to implement. However, empirical relationshipsare not necessarily valid in geologic settings and for rockproperties different from where they were formulated.

Figures 7 and 8 show the P-wave and S-wave velocitiesfor all models. These were calculated using identical inputparameters and represent the average background proper-ties of the gas hydrate hosting rocks at the Mallik 2L-38 well.The models (colored lines) are compared to the actual welldata (blue triangles). Although large variations exist among

these model predictions, the solid green line (model 3)closely matches both the P-wave and S-wave data. As willbe shown later, this model also accurately predicts the gashydrate saturation at the Blake Ridge drill site. Hence, weadopted this model for our modeling work and gas hydrateestimation from the seismic data. It must be noted, however,

JANUARY 2004 THE LEADING EDGE 61

Figure 4. Seismic section of Atwater Valley study area showing hydratemound and general stratigraphy.

Figure 5. Seafloor gas hydrate features at Atwater Valley study area.

Figure 6. Existing microstructural models of gas hydrate bearing sedi-ments.

that this model tends to overestimate S-wave velocity at highgas hydrate saturations. It is also sensitive to the choice ofco-ordination number, critical porosity, and component elas-tic properties.

In deep ocean sediments, gas hydrates are only stable

at a very shallow interval below the seafloor. This gashydrate stability zone (GHSZ) is determined by water depth,pore pressure, seafloor temperature, thermal gradient, andgas and fluid composition. Figure 9 shows the phase curvesfor three cases: 100% methane (blue), gas from gas ventingareas one in Green Canyon (green), and another inMississippi Canyon (red). As the water depth (vertical axis)increases, the temperature threshold for gas hydrate alsoincreases. At a given water depth, the temperature thresh-olds for gas samples at GC and MC are higher than that ofpure methane because they possess larger gas molecules.Knowing the water depth and thermal gradient at a partic-ular location, the GHSZ can be estimated by finding theseafloor temperature (black curve in Figure 9) and drawinga straight line to the appropriate thermal gradient. Thedepth at which the thermal gradient line intersects the gashydrate phase curve represents the deepest depth at whichgas hydrate can be formed with that type of gas.

In Keathley Canyon, the GHSZ is predicted to be over500 m BML for 100% methane, and even larger for the MCand GC gas samples (Figure 9). For Keathley Canyon, thisis about 20% greater than what is observed from the BSR(Figure 2). Because the water depth and seafloor tempera-ture are well defined, and the thermal gradient used is atthe high end of typical GOM values, the overestimation issuspected to result from the nonlinear behavior of shallowthermal gradients. In other words, the thermal gradients ofshallow sediments may be substantially higher than theregional average values used (25°C/km). This results in aslightly thicker GHSZ than estimated from average thermalgradients. To match the observed BSR at Keathley Canyonto predicted values, an average shallow thermal gradient ofover 30°C/km is required. This is more than 50% higher thanthe typical thermal gradient in the GOM, which is slightlyless than 20°C/km.

From Figure 9, it should be noted that all predictedhydrate stability zones are within 1000 m below the mud-line (BML), and most are within the first 500 m.

At shallow depths, rock properties such as porosity, den-sity, P-wave, and S-wave velocities are extremely variable.Porosity of shales can vary from 80% at the seafloor to lessthan 40% within 500 m of the water bottom. Figures 10 and11 show the range of porosity and VP within the first 3000ft below the mudline. The variations in these values with

62 THE LEADING EDGE JANUARY 2004

Figure 7. P-wave velocity versus gas hydrate saturation for the rockphysics models shown in Figure 6. M3 is the best model (model 3 inset)that matches the gas hydrate saturation at Mallik 2L-38.

Figure 8. S-wave velocity versus gas hydrate saturation for the rockphysics models shown in Figure 6. M3 (model 3 inset) best matches thegas hydrate saturation at Mallik 2L-38.

Figure 9. Prediction of gas hydrate stability zone in the deepwater Gulf ofMexico. The thermal gradient used is 25°C/km.

Figure 10. GOM shales and sands (Gregory, 1977), rigid global sand(Paxton et al., 2002), and Hamilton’s data (1965) showing shale porosityversus depth within the first few thousand ft BML.

changing lithology illustrate the importance of under-standing the shallow rock properties for gas hydrates delin-eation and volume estimation.

By understanding the porosity depth trend at shallowdepths due to compaction, and by using an appropriatevelocity-porosity model, a background rock physics trendcan be constructed. Using the gas hydrate rock physicsmodel (model 3) discussed earlier, velocities for different gashydrate saturations can be predicted. This methodologywas applied to the ODP leg 164 hole 995B (Figure 12). Theblocky colored lines in the VP column are the replacementcurves for different gas hydrate saturations (0-50%) in stepsof 10%. The P-wave measurement falls mostly within 10%gas hydrate saturation line and between 10 to 20% at thebase of gas hydrate zone. This is in good agreement withestimations made through several other means that werereported in the literature. Panel 3 shows the S-wave esti-mation based on the P-wave measurement and estimationof VP/VS ratio. Panel 4 shows the bulk density variation withdifferent gas hydrates saturations. The bulk densitydecreases as the gas hydrate saturation increases due to thelower density of the hydrate compared to fluid. However,this effect is negligible.

Gas hydrate quantitative estimation. With no availabledrilling information, we designed a quantitative estimationprocedure that included elastic property inversion, rockphysical modeling, and quantitative gas hydrate saturationcalculation. Realistic gas hydrate quantitative estimationbased on seismic data relies on accurate elastic propertyestimation from seismic inversion and a practical gas hydraterock physical model. Full waveform prestack inversion wasapplied at numerous locations to estimate high-resolutionVP, VS, and density. The elastic volume used was then cre-ated from hybrid inversion combining the full waveformprestack inversion and conventional linear prestack inver-sion for robustness and efficiency. Technical details for theseinversion schemes are documented by Mallick (1995),Mallick et al. (2000), Dutta (2002), and Mallick and Dutta(2002). Figure 13 shows the high-resolution full waveformprestack inversion results at two locations in the KeathleyCanyon study area. Shallow sand shale sequences, BSR, andpossible gas hydrate anomalies were readily recognizablein the results. Figure 14 shows a portion of a stacked seis-mic section from the Atwater Valley area. Superimposed onthat figure we show slowness (inverse of VP in microsec-onds per foot) obtained from full waveform prestack inver-

JANUARY 2004 THE LEADING EDGE 63

Figure 11. GOM shales and sands (Gregory, 1977), rigid global sand(Paxton et al., 2002), and Hamilton’s data (1965, 1979) showing VPwithin the first few thousand ft BML.

Figure 12. Gas hydrate modeling and estimation of gas hydrate satura-tion from velocity data at ODP Leg 164-995B, Blake Ridge.

Figure 13. P-wave impedance estimation at two Keathley Canyon loca-tions, through full waveform prestack inversion (blue curve). Red curvesare the starting interval velocity (derived from stacking velocity) modelinput to the inversion. Green curves are velocities from spatially continu-ous velocity analysis (SCVA).

Figure 14. Matched slowness (inverse of VP) from seismic inversion atthe Atwater Valley area.

sion at indicated locations. Note the velocity reversal at theburied gas hydrate mound and the match between the seis-mic data (amplitudes) and the inversion-derived slownesses.A similar observation is made for other parts of Figure 14.A similar match can be observed for the Keathley Canyonarea in Figure 15.

From the results of the high-resolution impedance data,gas hydrate saturation was quantitatively estimated (Figure16). To achieve this, we first derived a monogram from therock model that relates gas hydrate saturation to P-imped-ance (product of velocity and bulk density). Then, P-imped-ance data obtained from seismic inversion as describedearlier is used to derive appropriate gas hydrate saturationvalues.

The above procedure has been used in 3D using thehybrid inversion procedure to derive 3D gas hydrate satu-ration volumes at Keathley Canyon and Atwater Valley.Figures 17 and 18 show the results at two arbitrary seismiclines. Gas hydrate saturation estimated at these two sectionsrange from 0 to a maximum of over 30% of pore space. Aword of caution is warranted for the reliability of the inver-sion results for gas hydrate saturation. There are several

sources of uncertainties: noise in the seismic data, ambigu-ities associated with the inversion results, and inadequaciesof the rock model and the parameters that dictate the pre-dictability of hydrate saturation are just a few.

In addition, lithology variations are also present that areignored in the current approach as we assumed an average-mix lithology of sands and shales for background (sedi-ments with no gas hydrate). Thus, relating all observedP-impedance variations above the background as due to gashydrate saturation only, is expected to yield upper boundsof gas hydrate concentration. Nonetheless, in frontier areaswhere no well data are available, and lithology hetero-geneities are poorly understood, this type of estimation willprovide valuable predrill information. This sort of infor-mation can be useful in selecting potential drill sites to fur-ther quantify gas hydrate saturations and properties. Forgeologic environments such as the Blake Ridge area wherethe host rock does not have distinctive layering structure,this type of estimation may indeed be close to the actual gashydrate saturation values.

How reliable are our predictions? There are numeroussources of ambiguities as discussed earlier. Gas hydrate sat-uration cubes such as those shown in Figures 17 and 18 mustbe calibrated. It should be noted that, despite the large num-ber of drilled hydrate wells worldwide, quality hydrate log-ging and coring data are scarce, especially in the Gulf ofMexico. Such data are urgently needed. This must also besupplemented by controlled laboratory measurements onthe properties of gas hydrates where parameters can be con-trolled. Until we devote resources to undertake such log-ging, coring, and laboratory measurements, currentestimates of possible gas that can be obtained from gashydrates must be questioned.

Discussion. Seismic surveys cover the shelf and much ofthe northern deepwater Gulf of Mexico (both 2D and 3D).Except for hazard surveys, the upper several hundred metersbelow the seafloor have been of little interest to the oil andgas industry. As a result, processing of the upper section wasperfunctory with conventional streamer data being used tovisualize deeper objectives at the expense of shallower zones.For the GOM in general, with its absence of consistent BSRs,conventional off-the-shelf seismic data must be reprocessedfor shallow targets. The current processing was designed toaddress the shallow objectives. This included 2-ms sam-pling, demultiple, and amplitude-preserving 3D Kirchhofftime migration.

The proposed workflow for gas hydrate detection andquantification (five-step process) is independent of whetherBSR is present or absent. It provides a framework for gashydrate characterization using an integrated geologic andgeophysical approach. Full waveform prestack inversionand detailed assessment of rock physics models for gashydrates are centerpieces of the proposed methodology.

We note that seismic technology—being a remote sens-ing tool—is appropriate for gas hydrate detection, with orwithout BSR. However, the data requirements are numer-ous: High S/N and wider frequency contents are just twoof the main prerequisites. Lately, the seismic industry hasprogressed to meet these requirements. An example is shownin Figure 19 using single-sensor data (Q data) in the EastBreaks area of the Gulf of Mexico. A subtle BSR crosscut-ting the strata in the shallow sediments is clearly revealedalong with several dewatering features. These may be relatedto shallow hazards as well. The high fidelity of the Q dataindeed helped in the identification of such features.

64 THE LEADING EDGE JANUARY 2004

Figure 15. Matched slowness (inverse of VP) from seismic inversion atthe Keathley Canyon area.

Figure 16. Gas hydrate saturation estimation at two Keathley Canyonlocations. Black curves are P-impedance values estimated from full wave-form prestack inversion. The group of smooth colored curves are gashydrate saturation estimated from rock model (M3). The blue curves onthe left of both panels indicate the estimated gas hydrate saturation.

Conclusions. Elevated P-wave and S-wave velocities arediagnostic features of shallow gas hydrate-bearing sedi-ments. Seismic detection and quantification of gas hydratesrely on qualitative processing, robust elastic inversion, and

practical gas hydrate rock physical model construction. Thefive-step integrated multidisciplinary approach proves to bean effective tool for gas hydrate characterization. Full wave-form prestack inversion and hybrid inversion generate

JANUARY 2004 THE LEADING EDGE 65

Figure 17. Gas hydrate saturation for Keathley Canyon line. Upper panel displays the P-impedance derived from hybrid inversion and the lower panelshows the estimated gas hydrate saturation.

Figure 18. Gas hydrate saturation for Atwater Valley line. Upper panel displays the P-impedance derived from hybrid inversion and the lower panelshows the estimated gas hydrate saturation.

robust elastic property volumes from seismic data. From this,and with our gas hydrate rock physics-based tool, gashydrate saturation volumes can be generated, thereby pro-viding guidance for the detection of gas hydrates and apotential quantitative resource estimation tool.

Suggested reading. “Deepwater geohazard prediction usingprestack inversion of large offset P-wave data and rock model”by Dutta (TLE, 2002). “Elasticity of high-porosity sandstones:Theory for two North Sea data sets” by Dvorkin and Nur(GEOPHYSICS, 1996). “Elasticity of marine sediments: rock physics

modeling” by Dvorkin et al. (GRL, 1999). “Rock physics of agas hydrate reservoir” by Dvorkin et al. (TLE, 2003). “Aspectsof rock physics from laboratory and log data that are impor-tant to seismic interpretation” by Gregory (AAPG Memoir 26,1977). “Sound speed and related physical properties of sedi-ments from experimental Mohole (Guadalupe site)” byHamilton (GEOPHYSICS, 1965). “VP/VS and Poisson’s ratio inmarine sediments and rocks” by Hamilton (Journal of AcousticSociety of America, 1979). “Elastic-wave velocity in marine sed-iments with gas hydrates: Effective medium modeling” byHelgerud et al. (GRL, 1999). “Gas hydrates-geological per-spective and global change” by Kvenvolden (Reviews ofGeophysics, 1993). “Seismic velocities for hydrate-bearing sed-iments using weighted equation” by Lee et al. (JGR, 1996).“Model-based inversion of amplitude-variation with offset datausing a genetic algorithm” by Mallick (GEOPHYSICS, 1995).“Hybrid seismic inversion: A reconnaissance tool for deepwa-ter exploration” by Mallick et al. (TLE, 2000). “Shallow waterflow prediction using prestack full waveform inversion of con-ventional 3D seismic data and rock modeling” by Mallick andDutta (TLE, 2002). “Construction of an intergranular volumecompaction curve for evaluating and predicting compaction andporosity loss in rigid-grain sandstone reservoirs” by Paxton etal. (AAPG Bulletin, 2002). TLE

Acknowledgments: We thank Diana Gillespie, Adam Koesoemadinata,Lecia Muller, Dianna Shelander, Randy Utech, Gary Wool, Tom Dittrich,Emil Nassif, and Chung-Chi Shih for help with the project. We also thankWesternGeco for permission to publish this article.

Corresponding author: ndutta@slb.com

66 THE LEADING EDGE JANUARY 2004

Figure 19. BSR shown on high-resolution seismic data (Q data) in theEast Breaks region, Gulf of Mexico.