Occurrence-Type-Constrained Well-Logging Evaluation of Natural Gas Hydrates in the South China Sea: Methods, Applicability

Abstract

Natural gas hydrate reservoirs in the South China Sea contain pore-filling, fracture-filling, and pore–fracture composite occurrences in fine-grained, clay-rich, and locally sandy sediments. Existing logging studies provide numerous resistivity, acoustic, nuclear magnetic resonance, image-log, and multimineral methods, but model assumptions are not always matched to occurrence type. This review organizes evidence from the Shenhu, Qiongdongnan, and Dongsha areas along an occurrence type–response mechanism–model selection–independent validation–uncertainty chain. Pore-filling reservoirs require mineral and clay constraints before integrated volumetric inversion using resistivity, acoustic, density/neutron, and Nuclear Magnetic Resonance (NMR) logs. Fracture-filling reservoirs should be evaluated primarily from electrical images, azimuthal resistivity, and electrical/acoustic anisotropy, with fracture and matrix porosities treated separately. Composite reservoirs require staged matrix inversion and fracture-component estimation, whereas hydrate–free-gas coexistence intervals require multiphase rock physics. We propose a workflow that classifies occurrence type before inversion, imposes model entry and exit criteria, validates results with independent evidence, and reports ranges rather than unsupported point estimates. This framework clarifies why no universal high-resistivity–high-velocity template or saturation equation is valid for all South China Sea hydrate reservoirs.

1. Introduction

Natural gas hydrate is a crystalline inclusion compound in which water molecules form cages around methane and other gas molecules under suitable pressure and temperature conditions. Successive drilling campaigns on the northern continental slope of the South China Sea (SCS) have recovered hydrate-bearing cores and acquired pore-water, logging-while-drilling, and wireline-log data. These investigations reveal fine-grained pore-filling, sandy pore-filling, fracture-filling, and pore–fracture composite reservoirs [1,2,3,4,5,6,7,8]. Shenhu studies established much of the regional basis for volumetric evaluation of pore-filling hydrate [1,2,3,5,9]. In contrast, drilling in the Qiongdongnan Basin demonstrated that gas chimneys, faults, and heterogeneous fluid migration can produce directional and multiscale fracture filling [8,10,11]. The expanded range of reservoir architectures exposes the limits of the familiar identification rule that equates high resistivity plus high velocity with hydrate.

Well logs provide continuous in situ measurements and connect point-scale core observations to seismic-scale interpretation. Studies have examined logging responses, rock-physics models, porosity, saturation, permeability, image-log classification, and machine-learning prediction [12,13,14,15,16,17,18,19,20,21,22]. More methods, however, do not necessarily produce a more reliable evaluation. A high-resistivity anomaly may reflect replacement of conductive pore water by hydrate, but it may also arise from high-angle hydrate-filled fractures, carbonate cement, a compact bed, invasion, or borehole conditions. A velocity increase depends not only on hydrate abundance but also on whether hydrate floats in the pore space, contributes to load-bearing contacts, or cements grains. In clay-rich fine sediment, surface conduction, bound water, and short-relaxation signals jointly alter resistivity and NMR responses [17,18,23]. An inversion can therefore converge mathematically while remaining geologically incorrect if storage space and response mechanism are not diagnosed first. Existing reviews and methodological studies are largely organized by tool type—resistivity, acoustic, NMR, or image logs—rather than by hydrate occurrence type. A method validated in pore-filling sands may be inappropriately applied to fracture-filling or composite reservoirs simply because its occurrence-specific assumptions are not explicitly stated. Consequently, the literature lacks a framework that (1) classifies occurrence type before inversion, (2) specifies when a given model enters or exits applicability, and (3) requires independent validation by different physical mechanisms. This review addresses that gap by organizing evidence along an occurrence-type–response-mechanism–model-selection chain, with explicit entry/exit criteria and uncertainty reporting.

This review shifts the organizing principle from a catalogue of tools to three questions: why a model is applicable, when it fails, and how its result can be independently tested. We compare the requirements of pore-filling, fracture-filling, and composite hydrate and propose technical routes for high-clay, fractured, composite, and hydrate–free-gas coexistence intervals. The central argument is that occurrence type controls the relationship between the measured response and volumetric hydrate content; classification must therefore precede inversion. This is a critical methodological review rather than a systematic review or a PRISMA study.

2. Review Scope and Evidence Organization

The review boundary comprises the SCS, natural gas hydrate, and well-logging evaluation. The evidence base includes journal articles, theses, and institutional reports available to the authors, supplemented by bibliographic checks in Web of Science, Scopus, CNKI, and publisher records. The search was updated through June 2026. Regional terms (South China Sea, Shenhu, Qiongdongnan, and Dongsha), target terms (gas hydrate and methane hydrate), and method terms (well logging, resistivity, acoustic, NMR, electrical image, porosity, saturation, and permeability) were combined. A source was retained as core evidence when it had a defined site or sample context and reported at least one of the following: a logging response, a model assumption, an input parameter, or an independent validation dataset. General descriptions without methodological information and records that could not be verified were not used as primary evidence. No preregistration, dual independent screening, or exhaustive search protocol was performed.

For each source, nine fields were extracted: region and well, lithology, occurrence type, logging suite, model assumptions, inputs, validation, main result, and applicability boundary. Evidence was graded by independence. Grade A denotes pressure-core or other quantitative core evidence consistent with logs and pore-water measurements. Grade B denotes conventional core or pore-water evidence consistent with at least two independent logging mechanisms. Grade C denotes agreement among multiple logs or between logs and seismic data without direct sample validation. Grade D denotes a single curve, one empirical model, or geological analogy. The grade applies to a claim rather than to an entire publication; evidence that establishes hydrate presence may be stronger than evidence that constrains volumetric saturation in the same study. Figure 1 summarizes this organization.

3. Hydrate Occurrence Types and Evaluation Requirements

3.1. Pore-Filling Hydrate

Hydrate in the Shenhu area commonly occupies intergranular pores in fine-grained slope sediment, although locally sandier intervals also occur. Accumulation is jointly controlled by depositional facies, grain size, gas supply, and the gas-hydrate stability zone [1,2,3,5]. Fine-grained pore-filling reservoirs commonly have high total porosity, low permeability, abundant clay minerals, and a substantial capillary-bound-water fraction. Hydrate formation reduces the connected conductive-water volume, increases resistivity, stiffens the effective frame, decreases acoustic slowness, and reduces NMR-visible mobile water. These responses have a meaningful volumetric average, so resistivity, acoustic, and NMR saturation models can be used after clay conduction, water salinity, borehole enlargement, and thin-bed resolution are addressed [9,17,18].

The objective is not to select the single most sensitive curve but to maintain material balance among mineral volumes, effective porosity, water, and hydrate. Gamma ray cannot automatically be converted to clay volume because volcanic ash, K-feldspar, or radioactive heavy minerals may contribute. Neutron porosity may overestimate effective porosity because it includes hydrogen in clay-bound water. Elemental capture spectroscopy (ECS), X-ray diffraction (XRD), or multimineral inversion should therefore constrain the mineral matrix before saturation is calculated.

3.2. Fracture-Filling Hydrate

Hydrate in the eastern Qiongdongnan Basin is closely associated with gas chimneys, faults, and high-flux fluid migration. GMGS5 data show filling of high-angle fractures, bedding-parallel partings, and fracture networks [8,10]. Tool orientation, fracture attitude, and borehole direction jointly control the measured response. A near-vertical resistive fracture can strongly increase one azimuthal resistivity while occupying only a small bulk volume. Applying an isotropic Archie equation to that response may yield saturation inconsistent with pressure-core or pore-water constraints [24].

Three distinct questions must be separated: whether hydrate-filled fractures exist, how fractures are oriented and connected, and how much bulk hydrate they contain. Electrical images are well suited to the first two questions, but bright-image area cannot be converted directly and robustly into volume. Azimuthal resistivity and acoustic anisotropy add directional constraints but require tool-response modeling and a matrix background. Fracture-porosity estimates depend on aperture, density, and image resolution. Consequently, fracture evaluation should be centered on imaging and anisotropy, with core and pore-water data used to bound volume. A large conventional-resistivity amplitude is not sufficient volumetric evidence.

3.3. Pore–Fracture Composite Hydrate

Matrix pores and fractures may both store hydrate in seep systems and sand–mud interbeds. Such systems may show local high saturation, rapid vertical variation, carbonate cement, and associated free gas [4,7]. Sandy beds may support relatively continuous pore filling, whereas adjacent mud-rich beds and structural pathways favor fracture filling [11]. Conventional resistivity then combines matrix-water saturation and directional fracture effects. Acoustic response combines grain support, cementation, fracture compliance, and free gas, so a single effective-medium model is rarely unique.

Composite evaluation is best divided into two stages. First, multimineral inversion, conventional porosity, and background trends define an unfractured matrix model and estimate pore-filling hydrate. Second, electrical images, azimuthal resistivity, and anisotropy identify the fracture component that departs from the matrix prediction. Matrix porosity, fracture porosity, and hydrate occupancy in each component should be reported separately, together with a range for total bulk content. Shenhu, Qiongdongnan, and Dongsha are representative settings rather than immutable type labels; more than one occurrence type can exist within one area, and classification must be made interval by interval. A combined P-wave velocity–density approach further demonstrates that neither velocity nor density alone reliably distinguishes pore from fracture filling [25].

Table 1. Diagnostic criteria and evaluation routes for hydrate occurrence types in the South China Sea.

Occurrence Type	Storage Space	Diagnostic Evidence	Main Non-Uniqueness	Recommended Route
Pore filling	Intergranular pores	Bed-scale high resistivity and acoustic stiffening; NMR-visible water deficit; weak anisotropy	Clay surface conduction, bound water, washout, and carbonate cement	Quality and mineral control → effective porosity → clay-aware electrical/acoustic inversion → core or pore-water validation
Fracture filling	High-angle fractures, bedding partings, or networks	Resistive image features; azimuthal resistivity and electrical/acoustic anisotropy consistent with fracture attitude	Induced fractures, tool orientation, sub-resolution fractures, and isotropic-Archie overestimation	Fracture detection and attitude statistics → anisotropic forward/inverse modeling → bounded fracture porosity and hydrate volume
Pore–fracture composite	Matrix pores and fractures	Superposed volumetric and directional responses with rapid vertical variability	Lithologic change, thin beds, free gas, and multiphase coexistence	Build an unfractured matrix model → isolate fracture residual → report matrix and fracture components with volume closure

and Figure 2 summarize the diagnostic routes and type–response–model relationships.

4. Logging Response Mechanisms and Non-Uniqueness

4.1. Resistivity: From Pore-Water Replacement to Directional Conduction

Pore water is the principal conductive phase in unconsolidated marine sediment. Hydrate is nearly insulating, so pore-filling hydrate generally increases resistivity by reducing connected water. The Archie relationship describes this effect using formation factor and water saturation and has a clear physical basis in clean, connected, nonconductive granular media [26]. Fine-grained SCS sediment, however, contains surface and double-layer conduction. An uncorrected Archie model then tends to underestimate water saturation and overestimate hydrate saturation. Simandoux-type parallel-conduction models or other clay-distribution corrections can reduce this bias only if clay volume, clay resistivity, and pore-water salinity are independently constrained [17,18].

For fracture-filling hydrate, the central issue changes from total conduction to conduction direction. Insulating material in steep fractures produces different current paths parallel and normal to the fracture set, resulting in macroscopic electrical anisotropy [24,27]. Measurements at different azimuths and depths of investigation may contain information on fracture attitude, aperture, density, and filling. A conventional axisymmetric resistivity curve compresses this directionality. Its high value cannot uniquely represent bulk saturation and may also be affected by invasion, bed thickness, eccentricity, and shoulder-bed response.

Free gas also decreases conductivity and can coexist with hydrate below the bottom-simulating reflector. A free-gas interpretation is strengthened when the high-resistivity interval is accompanied by a sharp velocity reduction, lower density, neutron–density separation, attenuation, and an appropriate stability-zone position [7]. Carbonate cement and compact beds can also be resistive, but they more commonly show simultaneously increased density and velocity and reduced porosity. Thus, high resistivity is an anomaly detector, not a unique hydrate indicator.

4.2. Acoustic Response: Hydrate Location Matters as Much as Abundance

The elastic effect of hydrate depends on its pore-scale location. Hydrate suspended in pores or replacing fluid adds less frame stiffness than hydrate at grain contacts or as cement. Effective-medium models distinguish pore filling, load-bearing behavior, and cementation [28], whereas three-phase Biot-type models treat sediment, hydrate, and pore fluid as interacting continua [29]. The same saturation can therefore produce different velocity increments, and a given velocity anomaly does not correspond to a unique saturation.

Shenhu hydrate-bearing intervals generally show increased P-wave velocity and reduced slowness [12,13,14], but compaction, effective stress, and porosity trends must first be removed. Using a fixed water-layer slowness as the sole baseline may incorrectly assign normal compaction to hydrate. Fractures add further complexity: velocity and attenuation differ for propagation parallel and normal to fractures; hydrate filling may reduce fracture compliance, whereas open fractures and free gas markedly reduce velocity. Dipole-sonic azimuths, image-log attitudes, and dispersion should therefore accompany P-wave slowness in fractured intervals.

Free gas is the strongest acoustic confounder. Even a small gas fraction reduces the bulk modulus and P-wave velocity. Where hydrate and free gas coexist, frame stiffening and fluid softening may partially cancel. Such intervals require a three- or four-phase model constrained by shear velocity, density, neutron response, resistivity, attenuation, and the stability-zone geometry.

4.3. NMR: Effective Pore Volume and Invisible Water

NMR logs detect relaxation of hydrogen-bearing fluids. Solid hydrate generally does not contribute the same measurable signal as pore water at routine echo spacing and wait time. Hydrate formation can therefore reduce NMR porosity and spectral amplitude and shift the observed distribution toward shorter relaxation times. The deficit relative to a calibrated background may constrain pore volume occupied by hydrate [9,16]. In clay-rich sediment, however, bound water has short T₂ and may fall below instrumental dead time or signal-to-noise limits. Assigning the entire NMR deficit to hydrate consequently overestimates saturation.

A hydrate-free background of comparable lithology and compaction is essential. Core low-field NMR, density porosity, clay volume, and nearby water-bearing intervals can calibrate this background. T₂ cutoffs established for conventional sandstone should not be transferred without testing because pore size, surface relaxivity, and clay mineralogy differ. Fracture volume may be too small for NMR resolution. The combination of a clear resistive image and only a minor NMR deficit can support fracture filling, but it cannot by itself provide a precise bulk volume.

4.4. Density, Neutron, Gamma-Ray, and Elemental Logs

Density logs constrain total porosity and mineral volumes, but the density contrast between hydrate and pore water is modest, and the measurement is sensitive to washout, mudcake, and heavy minerals. Neutron logs respond to hydrogen index; clay-bound water commonly makes neutron porosity too high. Their main roles are to identify gas through density–neutron separation, check porosity consistency, and constrain multimineral inversion.

Gamma ray reflects K, U, and Th rather than clay volume directly. ECS and XRD provide mineral constraints required by clay-aware electrical models. Caliper must act as a quality gate for all pad and shallow-investigation measurements because density, neutron, NMR, and image anomalies in enlarged holes may primarily reflect borehole conditions. Figure 3 shows a representative multi-log response, whereas Figure 4 illustrates why hydrate, free gas, and carbonate cannot be distinguished by one high-resistivity or low-slowness criterion. On the acoustic transit time–resistivity crossplot (a), the gas zone exhibits resistivity values of 1–4 Ω·m and acoustic transit times of 524–656 μs/m, with the gas zone located to the right of the red line. On the neutron porosity–density crossplot (b), the gas zone lies to the left of the red line, and it is evident that the gas zone has significantly lower neutron porosity, clearly distinguishing it from hydrates. Carbonate rocks and hydrates show similar responses in acoustic transit time (c), both exhibiting low acoustic transit times; however, carbonate rocks have lower resistivity than hydrates, generally ranging from 0.8 to 1.8 Ω·m, and their natural gamma-ray values (d) are also markedly lower, ranging from 10 to 35 API. By comprehensively considering the natural gamma-ray, acoustic transit time, and resistivity curves, carbonate rocks and hydrate reservoirs can be effectively distinguished.

The non-uniqueness mechanisms discussed above directly inform the decision rules in Section 5 and Section 6. Specifically: resistivity ambiguity (clay conduction vs. hydrate vs. fractures) dictates that resistivity-based saturation models require independent clay and fracture characterization before use (Section 5.2, Table 1). Acoustic ambiguity (hydrate location vs. free gas) dictates that P-wave velocity alone is never sufficient; shear velocity, density, and attenuation must be examined before accepting a saturation estimate (Section 6.3.2). NMR ambiguity (bound water vs. hydrate) dictates that a hydrate-free background of comparable lithology is mandatory, and NMR should not be used for fracture volume quantification (Section 6.3.3). These linkages are formalized in

Table 2. Selection and exit criteria for major logging-model families constrained by occurrence type.

Model Family	Entry Condition/Assumption	Minimum Inputs	Preferred Type	Exit Condition and Validation
Integrated porosity	Acceptable borehole and constrained mineral end members	Caliper, density/neutron, NMR, ECS/XRD	Pore filling; composite matrix	Stop when washout or end members are unresolved; validate with core porosity and comparable water-bearing beds
Archie	Nonconductive matrix, connected water, and approximate isotropy	Rt, Rw, porosity, a, m, n	Clean sandy pore filling	Reject for high clay or marked anisotropy; validate with pressure core, chloride, acoustic, or NMR
Clay-conduction model	Clay volume and electrical parameters independently constrained	Rt, Rw, porosity, Vcl, clay electrical parameters	Clay-rich pore filling	Report ranges if Vcl or clay conductivity is uncalibrated; constrain with XRD/ECS and water-bearing beds
Acoustic/joint model	Stable background velocity and diagnosed hydrate habit	Vp or slowness and background; optionally Rt	Pore filling and locally calibrated beds	Reject with unresolved gas or cement effects; test with Vs, density, core velocity, or chloride
Effective-medium/multiphase	Hydrate-location end members and mineral moduli defined	Vp, Vs, density, porosity, end-member moduli	Pore filling; hydrate–gas coexistence	Report multiple scenarios if habits fit equally well; use multi-attribute forward checks
Anisotropic/fracture	Directional response and fracture geometry observed	Azimuthal Rt, image log, attitude, matrix response	Fracture and composite	Do not perform unique volume inversion from conventional Rt alone; bound with images, core, and pore water
NMR/empirical permeability	Relaxation–pore-size relation locally calibrated	NMR porosity and T2, or porosity, Vcl, and Sh	Calibrated pore-filling reservoir	Use only for ranking without independent calibration; validate with core permeability and pressure/production data

as explicit entry and exit criteria for each model family.

5. Type-Constrained Qualitative Identification

5.1. Quality Control and Scale Harmonization

Identification begins with data reliability rather than anomaly magnitude. Caliper, bit size, mud properties, tool-quality flags, and repeat sections should classify intervals as reliable, correctable, or unusable. Curves from different runs should be depth-matched against gamma ray, density, and sharp boundaries. Measurements must then be compared at compatible vertical resolution; high-resolution images and lower-resolution nuclear logs should not be forced into a super-resolved quantitative result. Finally, hydrate-free trends within the same lithology are required to remove burial and compaction effects.

The following criteria are used to classify intervals (values are illustrative for LWD/wireline conditions and should be calibrated to local tool specifications):

Caliper: Reliable if borehole diameter variation < 10% of bit size; correctable if 10–20% with corrective models available; unusable if >20% or irregular washout.

Density/NMR pad logs: Reliable if standoff < 0.2 in and borehole rugosity < 5%; correctable if standoff 0.2–0.5 in with environmental correction; unusable if standoff > 0.5 in or pad lift-off is indicated.

Depth mismatch: Runs are depth-matched to within ±0.5 m against gamma ray and density sharp boundaries; mismatch > 1.0 m triggers re-matching; if mismatch cannot be resolved, the interval is flagged as correctable with uncertainty.

Repeatability: For resistivity and porosity curves, repeat sections should agree within ±5% (or ±0.02 for porosity); larger deviations indicate tool or borehole issues and the interval is downgraded.

These thresholds are not universal; they define decision gates that must be re-evaluated for each tool suite and borehole condition.

5.2. Operational Criteria and Evidence Combinations

Fixed numerical thresholds should not be transferred without local calibration. A useful rule consists of required evidence, supporting evidence, and exclusion conditions. For pore filling, required evidence is a bed-scale volumetric response: continuous increased resistivity and overall acoustic stiffening should coincide and should not be explained by lithology or compaction. Supporting evidence includes an NMR deficit relative to density porosity, layered rather than discrete fracture-like image responses, and weak azimuthal anisotropy.

For fracture filling, required evidence is a resistive sinusoid, bedding fracture, or network on the image together with azimuthal resistivity or dipole-sonic anisotropy consistent with fracture attitude. Very high conventional resistivity accompanied by limited density or NMR volumetric response is strong supporting evidence. Borehole breakouts, drilling-induced fractures, tool marks, and eccentricity must be excluded. Composite hydrate requires a matrix response that explains only part of the observations plus a reproducible directional residual associated with fractures. Hydrate–free-gas coexistence requires the electrical anomaly to be decomposed with acoustic attenuation, low velocity, density–neutron separation, and stability-zone constraints.

Classification output should communicate confidence: confirmed hydrate interval, high-probability interval, probable fracture-filling interval, three-phase coexistence interval, or ambiguous anomaly. “Confirmed” should be reserved for direct sample evidence or consistent independent physical mechanisms. Intervals supported only by resistivity remain high probability rather than confirmed.

5.3. Crossplots and Multivariate Identification

Crossplots convert single-curve ambiguity into relative patterns. Acoustic slowness–resistivity separates hydrate from free gas when gas produces acoustic softening despite increased resistivity. Neutron–density separation further supports free gas. Resistivity–slowness and gamma-ray–slowness can help distinguish carbonate from hydrate when carbonate increases density and stiffness and reduces apparent porosity. These boundaries are well- and tool-dependent; they define decision regions rather than universal thresholds. Clustering and supervised models can assist screening, but labels must be derived from independent evidence rather than from the same resistivity equation the model is meant to reproduce [21,22].

6. Quantitative Evaluation of Reservoir Parameters

6.1. Mineral and Clay Volumes

Gamma-ray clay estimates should be treated as a prior rather than a final value. Multimineral inversion combines density, neutron, sonic, resistivity, photoelectric, or elemental data with end-member response equations and volume closure. The physical requirements are accurate matrix properties, a parsimonious set of minerals, and explicit borehole corrections. Inputs include caliper, environmental corrections, log responses, and XRD or ECS constraints. The approach is most useful for pore-filling and the matrix component of composite reservoirs. It should be suspended where washout is severe or mineral end members are unidentifiable. Figure 5 presents a comparison between the actual drilling core data of Well SH-W19-2015 and the contents of illite, calcite, and quartz obtained from the theoretical calculation model based on XRD. The comparison results show that the contents of illite, calcite, and quartz derived from the XRD theoretical model exhibit good consistency in this well. However, due to the limited number of core samples, this model is only of certain reference value in this region.

6.2. Porosity

Density porosity follows from bulk and matrix density, but its uncertainty grows when matrix density varies, or washout affects the reading. Neutron porosity is sensitive to clay-bound water. Acoustic porosity requires a locally calibrated matrix and fluid baseline and is affected by hydrate cementation and free gas. NMR most directly measures fluid-filled effective porosity but misses short-T₂ water and has limited sensitivity to low-volume solid-filled fractures. Agreement among these methods is meaningful only after common depth and vertical support are established.

Fracture porosity cannot be equated with image bright-area fraction. A defensible estimate combines fracture frequency, apparent aperture, attitude correction, image coverage, and sub-resolution fracture bounds. The preferred output is a lower bound for resolved fractures and an upper scenario including unresolved fractures. Figure 6 compares conventional- and NMR-derived porosities with core observations for an internal consistency check. Because the available figure lacks a unified validation dataset, it should not be interpreted as demonstrating transferable cross-well accuracy.

For fractured natural gas hydrate reservoirs, the fractures are simplified as a parallel-plate model. It is assumed that the exposure of fractures on the borehole wall represents their average degree of development in the reservoir.

$${\mathrm{A}}_{\mathrm{f}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e}}={\mathrm{A}}_{\mathrm{a}\mathrm{p}\mathrm{p}}·{\displaystyle \frac{1}{\sin \mathrm{a}·\cos \mathrm{b}}}$$

$${\mathsf{\varphi }}_{\mathrm{f}}={\displaystyle \frac{{\mathrm{A}}_{\mathrm{f}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e}}}{{\mathrm{A}}_{\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}}}$$

$${\mathrm{A}}_{\mathrm{total}} = \mathsf{\pi }\mathrm{dL}$$

where $\mathrm{a}$ denotes the fracture dip angle, and $\mathrm{b}$ represents the angle between the fracture strike and the borehole axis direction. ${\mathrm{A}}_{\mathrm{a}\mathrm{p}\mathrm{p}}$ is the fracture area in the image, which was identified manually; ${\mathrm{A}}_{\mathrm{f}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e}}$ is the total area occupied by fractures on the unfolded borehole wall map; ${\mathrm{A}}_{\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}$ is the total borehole wall area corresponding to the statistical window length.

6.3. Hydrate Saturation

6.3.1. Resistivity Models

For clean pore systems, Archie’s equation is expressed as:

$${\mathrm{S}}_{\mathrm{w}} = {\left[{\displaystyle \frac{\mathrm{a}·{\mathrm{R}}_{\mathrm{w}}}{{\mathsf{\varphi }}^{\mathrm{m}}·{\mathrm{R}}_{\mathrm{t}}}}\right]}^{{\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$\mathrm{n}$}\right.}}, {\mathrm{S}}_{\mathrm{h}} = 1 - {\mathrm{S}}_{\mathrm{w}}$$

where S_w is water saturation (fraction), S_h is hydrate saturation (fraction), ${\mathrm{R}}_{\mathrm{t}}$ is true formation resistivity (Ω·m), ${\mathrm{R}}_{\mathrm{w}}$ is formation-water resistivity (Ω·m), ϕ is porosity (fraction), a is tortuosity factor (dimensionless), m is cementation exponent (dimensionless), and n is saturation exponent (dimensionless) [26].

For clay-rich pore-filling reservoirs, a Simandoux-type parallel-conduction model is employed:

$${\displaystyle \frac{1}{{\mathrm{R}}_{\mathrm{t}}}}={\displaystyle \frac{{\mathrm{V}}_{\mathrm{c}\mathrm{l}}}{{\mathrm{S}}_{\mathrm{w}}·{\mathrm{V}}_{\mathrm{c}\mathrm{l}}}}+{\displaystyle \frac{{\mathsf{\varphi }}^{\mathrm{m}}·{\mathrm{S}}_{\mathrm{w}}^{\mathrm{n}}}{\mathrm{a}·{\mathrm{R}}_{\mathrm{w}}}}$$

where ${\mathrm{R}}_{\mathrm{t}}$ is true formation resistivity, ${\mathrm{R}}_{\mathrm{w}}$ is formation-water resistivity, $\mathsf{\varphi }$ is porosity, ${\mathrm{V}}_{\mathrm{c}\mathrm{l}}$ is clay volume fraction, R_cl is effective clay resistivity, and $\mathrm{a}$, $\mathrm{m}$, and $\mathrm{n}$ are the tortuosity, cementation, and saturation exponents, respectively [30]. The following variants are considered depending on data availability: (1) Dispersed clay model—clay particles are distributed within the pore space, reducing the effective pore-throat radius; (2) Laminated clay model—clay exists as discrete thin layers, treated as parallel conductors; (3) Structural clay model—clay replaces framework grains. The dispersed model is preferred for Shenhu fine-grained sediments based on core petrography.

6.3.2. Acoustic and Effective-Medium Models

The simplest time-average relation is $\Delta \mathrm{t}=\sum \mathrm{V}\mathrm{i}·\mathrm{t}\mathrm{i}$, but three-phase time-average, effective-medium, contact-cement, and Biot-type formulations represent different hydrate locations [14,28,29]. Entry requires a hydrate-free velocity trend in the same lithology and sufficient P-wave, S-wave, density, or laboratory evidence to choose among pore-filling, load-bearing, and cementing end members. With only P-wave data and uncertain habit, several location-model scenarios should be reported. A single isotropic acoustic inversion should be rejected where free gas or fracture anisotropy remains unresolved. Validation requires simultaneous forward reproduction of density, P- and S-wave velocity, and their ratio rather than comparison with a second acoustic equation.

6.3.3. NMR, Sigma, and Joint Methods

An NMR porosity-deficit estimate can be written ${\mathrm{S}}_{\mathrm{h}}={\displaystyle \frac{\mathsf{\varphi }-{\mathsf{\varphi }}_{\mathrm{N}\mathrm{M}\mathrm{R}}}{\mathsf{\varphi }}}$, ${\mathsf{\varphi }}_{\mathrm{N}\mathrm{M}\mathrm{R}}$ is hydrate-free effective porosity for comparable lithology. Entry requires known acquisition timing and a correction for invisible short-T₂ water. An unstable background permits only a range, and NMR should not be used for unique fracture saturation when fracture volume is below resolution. Sigma provides an independent water-saturation estimate only where fluid salinity, borehole effects, and end-member capture cross sections are known. Coefficients in resistivity–acoustic joint models require local core or pore-water calibration. The available dual-parameter and Sigma results are broadly consistent with core estimates (Figure 7). Figure 7 presents a within-well consistency check between log-derived hydrate saturation estimates and core-derived saturations for Well SH-W19-2015 (pore-filling, high-clay reservoir). For the resistivity–acoustic dual-parameter method (Figure 7a, n = 44), the correlation coefficient is R² = 0.86; for the Sigma method (Figure 7b, n = 38), R² = 0.83. These values indicate internal consistency between the two log-based methods and core observations within the same well.

6.3.4. Fracture Anisotropy and Composite-Component Models

Directional high resistivity should not be inserted directly into isotropic Archie. An anisotropy ratio may be defined as ΔGR = ${\displaystyle \frac{\mathrm{G}\mathrm{R}-\mathrm{G}\mathrm{R}\mathrm{m}\mathrm{i}\mathrm{n}}{\mathrm{G}\mathrm{R}\mathrm{m}\mathrm{a}\mathrm{x}-\mathrm{G}\mathrm{R}\mathrm{m}\mathrm{i}\mathrm{n}}}$, where GRmin and GRmax represent the natural gamma-ray log values at clean sandstone and pure shale, respectively. Matrix resistivity, fracture attitude, density, and apparent aperture then support effective-conductivity models parallel and normal to fractures [24,27]. Entry requires co-located azimuthal resistivity or dipole sonic and a reliable image, with directional extrema consistent with fracture attitude. With conventional resistivity alone, only an upper scenario is defensible. Volumetric inversion should stop if induced fractures cannot be excluded, a matrix background cannot be established, or tool-response modeling is absent. Figure 8 explains why low fracture volume can generate strong directional resistivity and an isotropic-Archie overestimate. Composite reservoirs require separate matrix and fracture components subject to volume closure. In vertical wells, near-vertical fractures are roughly parallel to the borehole axis. Standard resistivity logs (such as induction or propagation resistivity) primarily measure horizontal currents or are mainly influenced by horizontal resistivity (R). If the fractures are filled with high-resistivity gas hydrate and are parallel to the borehole axis, they greatly impede vertical currents (or alter the magnetic field), leading to extremely high apparent resistivity (anisotropy effect), while the actual in-situ saturation is much lower. This resolves the issue of overestimation by Archie’s formula.

6.4. Permeability

Permeability depends on original pore throats, clay, hydrate saturation and habit, and fracture connectivity. Regional regressions may relate porosity, clay volume, and hydrate saturation to permeability; Shenhu data generally indicate decreasing permeability with increasing saturation [31]. Such relations are convenient but scale- and facies-dependent. The empirical relationship in Figure 9 should be regarded as conditional for comparable fine-grained Shenhu sediment, not as a universal SCS equation.

Coates- and SDR-type NMR relations can be represented as K = ${({\displaystyle \frac{\mathrm{\phi }}{\mathrm{C}}})}^4+{({\displaystyle \frac{\mathrm{F}\mathrm{F}\mathrm{I}}{\mathrm{B}\mathrm{V}\mathrm{I}}})}^2$ and K = ${\mathrm{\phi }}^4{\mathrm{T}}_{2\mathrm{G}\mathrm{M}}^2$, where FFI and BVI are free- and bound-fluid volumes and ${\mathrm{T}}_{2\mathrm{G}\mathrm{M}}$ is the geometric-mean relaxation time [32,33]. Entry requires paired core permeability and NMR data from the same facies and scale. Without hydrate-bearing core calibration, these equations are useful only for ranking. They should not be extrapolated when fracture connectivity dominates, or dissociation changes the pore network. Fractured reservoirs require image-derived networks, directional responses, and pressure or production data.

Table 2 compares model families through entry conditions, minimum inputs, exit criteria, and independent validation. No fixed ranking exists outside occurrence type. A simpler model is often more robust when its assumptions are met; a more complex model improves credibility only when its added parameters are independently constrained. Figure 10 visualizes this conditional applicability.

7. Type-Constrained Logging-Evaluation Framework

The proposed framework contains seven linked stages: quality control, lithology and storage-space diagnosis, occurrence-type classification, model selection, joint inversion, independent validation, and uncertainty reporting (Figure 11). Each stage has an exit condition. An interval with severe washout and no reliable deep measurement exits quantitative evaluation. If pore and fracture filling cannot be distinguished, two scenarios must be reported instead of one saturation.

7.1. Pore-Filling and High-Clay Routes

Multimineral inversion first establishes matrix density and clay volume. Density, neutron, sonic, and NMR then define a porosity ensemble. Saturation should include at least one electrical and one non-electrical model, calibrated against core, chloride anomaly, or pressure core. High-clay intervals require explicit surface-conduction and bound-water treatment, preferably through Simandoux-type or locally calibrated fine-grained models constrained by short-T₂ NMR and XRD. When models diverge beyond a predefined tolerance, the interpreter should revisit clay distribution, background porosity, and formation-water resistivity instead of averaging incompatible estimates.

7.2. Fracture-Filling Route

Electrical images and directional responses form the first evidence tier and define fracture attitude, density, connectivity, and image quality. An anisotropic resistivity or acoustic model then estimates directional properties after removal of the matrix background. Fracture porosity and bulk hydrate volume are estimated only when image resolution, borehole conditions, and core or pore-water constraints are adequate. The result should report a lower bound for resolved fractures and an upper model scenario that includes sub-resolution fractures. This converts fracture evaluation from unconstrained visual identification to bounded volumetric analysis without claiming false precision.

7.3. Composite and Hydrate–Free-Gas Routes

Composite reservoirs require component-wise inversion. Low-anisotropy intervals calibrate the matrix porosity and pore-filling model; directional residuals associated with image-log fractures then constrain the fracture component. Total volume is checked by forward modeling all relevant curves. Hydrate–free-gas coexistence requires an explicit gas fraction and a joint inversion of P- and S-wave velocity, density, neutron, and resistivity, using stability-zone geometry and attenuation to constrain gas. If the multiphase parameters are not identifiable, the output should comprise a no-free-gas upper scenario and a free-gas lower scenario.

7.4. Workflow Closure for the SH-W19 Pore-Filling Case

The SH-W19 case tests implementation in a high-clay fine-grained pore-filling setting; it is not a substitute for fracture-case validation. The sequence is: remove unreliable pad-log intervals using caliper and quality flags; depth-match runs; estimate clay and matrix density through ECS/conventional multimineral inversion and XRD checks (Figure 5); classify the interval using continuous resistivity, acoustic stiffening, NMR deficit, and weak directionality; compare density-, neutron-, acoustic-, and NMR-derived porosities against core (Figure 6); construct saturation ensembles from a clay-corrected resistivity–acoustic method and Sigma (Figure 7); and evaluate internal consistency with core, XRD, and independent logging physics.

The strength of this case is its complete sequence from mineral constraint to classification, multi-model calculation, and core comparison. Its limitation is that the raw numerical table is not presently available in the manuscript package. A fully independent calibration/validation split and unified n, R², RMSE, MAE, and bias cannot therefore be recomputed. Figure 3, Figure 5, Figure 6 and Figure 7 support workflow feasibility and internal consistency, not cross-well accuracy. These statistics, matching scale, outlier rules, and any same-well calibration optimism should be documented before submission.

It must be emphasized that the SH-W19 case demonstrates workflow feasibility only for high-clay fine-grained pore-filling reservoirs. The fracture-filling, composite, and hydrate–free-gas coexistence pathways presented in Section 7.2 and Section 7.3 are methodologically constructed from published evidence (GMGS5 and Dongsha data) but have not been validated with the same level of internal quantitative scrutiny as the SH-W19 case, because no disclosable in-house directional logs or pressure-core data are currently available for those settings. Validation of the full workflow for fractured and composite reservoirs remains a priority for future work.

7.5. Independent Validation and Uncertainty

The preferred validation hierarchy is pressure core or preserved core, pore-water chloride anomaly, conventional core and XRD, logs based on different physical mechanisms, and finally nearby-well or seismic constraints. Agreement between two equations derived from the same resistivity curve is not independent validation. The Shenhu example can use the authors’ combined logs, mineral/XRD comparison, and core properties; Qiongdongnan is evaluated from published evidence only because no disclosable in-house directional logs are available.

Table 3. Methodological implications and applicability boundaries of representative South China Sea cases.

Representative Case	Type and Evidence	Methodological Implication	Applicability Boundary
Shenhu GMGS1/GMGS3	Predominantly pore filling; resistivity, acoustic, core/chloride, and XRD evidence (Grades A–B)	Constrain clay and effective porosity before joint electrical, acoustic, or NMR volume evaluation	Short T2, clay, compaction, and thin-bed effects require in-well calibration; parameters are not directly portable
Qiongdongnan GMGS5 and related wells	Fracture and pore–fracture composite; image, directional, interbed, and gas-chimney evidence (Grades B–C)	Separate matrix and fracture contributions before selecting anisotropic or component models	Sub-resolution fractures and limited public raw data favor ranges over unique values
Dongsha and other seep systems	Composite, locally with free gas or carbonate; seismic, Bottom-Simulating Reflector(BSR), and downhole anomalies (Grade C)	Use seismic–log integration and three-way hydrate/free-gas/carbonate discrimination	Limited public well data support mechanism identification and risk grading, not unconstrained precision

compares evidence and applicability boundaries.

Uncertainty includes measurement error, end-member parameters, model structure, and occurrence-type classification. Reasonable ranges for porosity, formation-water resistivity, m, n, clay resistivity, and background velocity can be propagated through Monte Carlo or structured scenarios. Fracture evaluation must additionally vary attitude, aperture, and the sub-resolution fracture fraction. Regional resource and occurrence syntheses support the geological classification boundary [34,35]. Conventional reservoir-parameter evaluation in Shenhu and fracture-porosity research represent distinct local calibration paths [36,37]. Foundational hydrate-bearing-sediment models emphasize linked electrical, elastic, and permeability constraints [38]; Simandoux provides the classical basis for shaly-sand conduction correction [30]; and NMR porosity and permeability interpretation must remain tied to relaxation physics [32,33]. These sources reinforce that a model name does not guarantee transferability: every implementation must be recalibrated to SCS lithology and occurrence type.

8. Conclusions

• The first task in SCS hydrate evaluation is to identify storage space and occurrence type at the interval scale. Pore-filling, fracture-filling, and composite hydrate have different response mechanisms and cannot be evaluated using one high-resistivity–high-velocity template or one saturation equation.
• Pore-filling evaluation requires mineral and clay constraints followed by integrated resistivity, acoustic, density/neutron, and NMR analysis. Fracture filling requires image logs and anisotropy with separate fracture and matrix components. Composite reservoirs require staged matrix inversion, fracture identification, and component-wise volume closure.
• The proposed quality control–classification–model selection–joint inversion–independent validation–interval reporting workflow provides an actionable route for high-clay, fractured, composite, and hydrate–free-gas settings. Direct demonstration currently relies mainly on SH-W19 pore-filling data; fracture-volume quantification, unified error statistics, and independent cross-well testing remain priorities.