Fine 3D Seismic Processing and Quantitative Interpretation of Tight Sandstone Gas Reservoirs—A Case Study of the Shaximiao Formation in the Yingshan Area, Sichuan Basin

Abstract

Targeting the thinly bedded and strongly heterogeneous tight sandstone gas reservoirs of the Shaximiao Formation in the Yingshan area of the Sichuan Basin, this study establishes an integrated workflow that combines high-fidelity 3D seismic processing with quantitative interpretation to address key challenges such as insufficient resolution of conventional seismic data under complex near-surface conditions and difficulty in depicting sand-body geometries. On the processing side, a 2D-3D integrated amplitude-preserving high-resolution strategy is applied. In contrast to conventional workflows that treat 2D and 3D datasets independently and often sacrifice true-amplitude characteristics during static correction and noise suppression, the proposed approach unifies first-break picking and static-correction parameters across 2D and 3D data while preserving relative amplitude fidelity. Techniques such as true-surface velocity modeling, coherent-noise suppression, and wavelet compression are introduced. As a result, the effective frequency bandwidth of the newly processed data is broadened by approximately 10–16 Hz relative to the legacy dataset, and the imaging of small faults and narrow river-channel boundaries is significantly enhanced. On the interpretation side, ten sublayers within the first member of the Shaximiao Formation are correlated with high precision, yielding the identification of 41 fourth-order local structural units and 122 stratigraphic traps. Through seismic forward modeling and attribute optimization, a set of sensitive attributes suitable for thin-sandstone detection is established. These attributes enable fine-scale characterization of sand-body distributions within the shallow-water delta system, where fluvial control is pronounced, leading to the identification of 364 multi-phase superimposed channels. Based on attribute fusion, rock-physics-constrained inversion, and integrated hydrocarbon-indicator analysis, 147 favorable “sweet spots” are predicted, and six well locations are proposed. The study builds a reservoir-forming model of “deep hydrocarbon generation–upward migration, fault-controlled charging, structural trapping, and microfacies-controlled enrichment,” achieving high-fidelity imaging and quantitative prediction of tight sandstone reservoirs in the Shaximiao Formation. The results provide robust technical support for favorable-zone evaluation and subsequent exploration deployment in the Yingshan area.

1. Introduction

The Jurassic Shaximiao Formation in the central–northeastern Sichuan Basin is widely developed in a foreland tectonic setting, and its reservoirs generally exhibit typical tight-sandstone characteristics, including thin bedding, low porosity and permeability, and strong heterogeneity [1,2]. This foreland setting is mainly reflected by the development of broad, gentle detachment-controlled folds, thin-skinned structural styles, and long-term compressional deformation related to peripheral orogenic belts, rather than by strongly developed syn-tectonic growth strata. The Yingshan area is located in the transitional zone between the gentle structural belt of central Sichuan and the fold belt of northeastern Sichuan [3]. It is jointly controlled by multi-phase compressional and detachment structures, forming a “dome–gentle slope” structural pattern dominated by the Yingshan anticline. In a shallow-water delta–fluvial depositional background, the Shaximiao Formation contains widely developed, multi-phase, superimposed channel sand bodies, which provide important storage and enrichment spaces for natural gas [4]. In this system, deltaic deposits are mainly represented by distributary channels and mouth-bar-related sand–mud alternations, characterized by relatively frequent mudstone interbeds and moderate lateral continuity, whereas fluvial components are dominated by erosional-based channel sand bodies with blocky to bell-shaped log responses and higher vertical stacking frequency. These differences exert direct control on sand-body geometry, continuity, and reservoir connectivity.

However, the tight sandstone gas reservoirs of the Shaximiao Formation in the Yingshan area exhibit strong heterogeneity [5]. The channel sand bodies are thin and vary laterally rapidly, and the complex near-surface conditions further degrade the imaging resolution and attribute sensitivity of conventional seismic data [6,7]. These limitations hinder the fine identification and reliable evaluation of favorable reservoirs. In practical exploration, this problem is particularly prominent: although drilling and test-production data indicate good overall oil and gas potential in the Shaximiao Formation—especially industrial gas flows widely obtained in the middle–lower part of the first member—the thinly bedded reservoirs, complex sand-body geometries, and frequent mudstone interlayers significantly reduce the stability and reliability of traditional reservoir prediction methods in delineating effective thick sands and major gas-bearing zones.

From the perspective of hydrocarbon accumulation mechanisms, the Yingshan area exhibits a clear “tri-element coupling” pattern [8]: deep hydrocarbon-generating faults provide continuous vertical migration pathways; the crest of the anticline serves as the primary structural background for hydrocarbon updip accumulation and preservation; and the multi-phase, relatively well-connected channel sand bodies determine the scale of reservoir enrichment and the distribution of high-quality intervals [9]. However, the superposition of these controlling factors further increases the complexity of sand-body distribution and imposes higher requirements on seismic resolution and quantitative interpretation.

In this study, we develop an integrated workflow that combines 2D-3D seismic processing with a quantitative interpretation for the Shaximiao Formation, emphasizing amplitude preservation and high resolution to achieve fine-scale three-dimensional characterization of tight sandstone reservoirs. Rather than introducing new geophysical methodologies, the primary scientific contribution of this work lies in the integrated, forward-modeling-constrained application of established seismic processing, multi-attribute analysis, and rock-physics-driven inversion techniques to systematically resolve the detectability, thickness quantification, and spatial distribution of thin, heterogeneous tight sandstone reservoirs. By quantitatively identifying the seismic responses of thin sand bodies and predicting the distribution of effective reservoir thickness, and by integrating interpretation results with structural position, hydrocarbon-supply pathways, and sedimentary microfacies, we establish a reservoir-forming model characterized by “deep hydrocarbon generation with upward migration, fault-controlled charging, structural trapping, and microfacies-controlled enrichment.” The results provide a reliable and transferable basis for predicting sweet spots and for subsequent exploration and deployment of tight sandstone gas reservoirs in the Yingshan area and comparable foreland-basin settings.

2. Regional Geological Background

2.1. Structural Characteristics

The Yingshan area lies within the transitional zone between the gentle fold belt of central–northeastern Sichuan Basin and the mild structural belt of northern Sichuan [10]. It belongs to the broad, gentle anticline–syncline structural system along the Yilong–Yingshan trend. Dominated by the Yingshan anticline, the regional structure exhibits a general NE–NW orientation. Spatially, it connects with the Longgang structure to the north and the Guang’an structure to the south, borders the Gongshan Temple structure to the west, and merges with the Huaying Mountain structural belt to the east. This configuration reflects the overall strike control of the central–northeastern Sichuan foreland fold system and the segmentation effects imposed by boundary structures, as shown in Figure 1. From a tectonic perspective, this structural configuration is consistent with a foreland fold system developed under sustained Mesozoic compression from surrounding orogenic belts (Longmenshan–Micangshan–Dabashan). The dominance of thin-skinned detachment folding, gentle anticline geometry, and limited fault displacement suggests that the Middle Jurassic strata, including the Shaximiao Formation, were deposited during a relatively stable compressional stage, with weak syn-depositional deformation but strong post-depositional folding, which is typical of the central Sichuan foreland basin. The region is characterized by a thick Jurassic–Triassic sedimentary cover, in which surface and shallow subsurface structures are mainly long-axis, gently dipping anticlines, displaying the typical “dome–gentle slope” morphology. It constitutes an integral part of the broad, gentle fold system of central and northern Sichuan.

From the perspective of geometric configuration and mechanical style, the Yingshan anticline and adjacent structures are predominantly thin-skinned folds controlled by detachment horizons such as the gypsum–salt layers of the Leikoupo Formation. The shallow structures exhibit obvious folding, while the deeper strata remain relatively gentle. Faults are generally sparse, dominated by small- to medium-scale thrusts and associated secondary faults. Both 2D/3D seismic interpretation and regional geological studies indicate that the Yingshan and Nanchong anticlines are detachment folds, with deep formations not significantly involved in strong deformation.

Topographically, this structural style results in low-relief hills at the surface, with smooth and gentle underground structures lacking major fault offsets. Strata generally dip at less than 5°, indicating a simple overall structural geometry. Such a gentle, positive anticline provides a favorable structural background for hydrocarbon accumulation, while also implying that subtle structural variations and lithologic traps mainly control reservoirs.

Regarding structural evolution, the Sichuan Basin has undergone multi-stage compressional reworking during the Indosinian, Yanshanian, and Himalayan periods [11]. Since the Mesozoic, the central–northeastern Sichuan foreland fold belt has been continuously affected by compressional stresses from peripheral orogenic belts (e.g., the Longmenshan–Micangshan–Dabashan), resulting in multiple phases of fold–thrust structures. Regionally, the superposition of “broad, gentle anticlines + localized thrust faults” constitutes the principal source of the present structural pattern in the Yingshan area [12,13]. Basin-scale studies further indicate that this region developed multi-level detachments and a combination of strike–slip/thrust influences under long-term Mesozoic–Cenozoic compression and slow uplift. These processes ultimately produced the modern differential structural framework of “three uplifts and three depressions” and the boundary effects of the parallel ridge–valley system in northeastern Sichuan, which collectively provide spatial pathways and structural highs favorable for vertical hydrocarbon migration and accumulation.

2.2. Stratigraphic Characteristics

The Yingshan area is located in the central–northeastern Sichuan Basin, where the Mesozoic strata are well preserved. The underlying Upper Triassic Xujiahe Formation, composed of marine-to-transitional continental deposits, serves as the primary hydrocarbon source rock [14]. Overlying it is the Jurassic succession, which consists predominantly of thick continental deposits, as shown in Figure 2.

Surface exposures in the Yingshan area are dominated by Jurassic strata, among which the Suining Formation and the Shaximiao Formation are widely distributed. Drilling data indicate that the interval from the Xujiahe Formation to the Shaximiao Formation constitutes a continuous sedimentary sequence, which is supported by consistent well-log motifs showing gradual lithologic and electrofacies transitions across the Triassic–Jurassic boundary, without evidence of abrupt truncation or missing sections. In addition, regional stratigraphic correlation and seismic profiles do not reveal major angular unconformities or regionally developed erosional surfaces within this interval, consistent with deposition under a relatively stable foreland basin setting. The Shaximiao Formation is Middle Jurassic in age and consists of interbedded sandstone and mudstone deposited in fluvial–lacustrine to deltaic environments. Regionally, it is commonly subdivided into a lower member (Shaximiao Member 1, or Sha-1) and an upper member (Shaximiao Member 2, or Sha-2). A prominent intraformational shale/muddy layer typically marks the boundary between the two members and represents a maximum lacustrine flooding surface. This surface serves as a key stratigraphic marker for well-to-seismic correlation. In this study, this flooding surface and other regionally stable marker beds were used as sequence-stratigraphic boundaries to subdivide the Shaximiao Formation into consistent sublayers, providing a stratigraphic framework for subsequent horizon-based seismic interpretation and attribute extraction. To avoid stratigraphic mixing, all attributes were extracted within horizon-constrained windows defined by interpreted sequence boundaries and marker beds calibrated from wells, rather than using constant-time slices. This approach ensures that the extracted seismic responses correspond to the same stratigraphic units across the entire 3D volume.

The Xujiahe Formation, a typical organic-rich continental source rock of the Upper Triassic [15], is characterized by “large thickness–high maturity–strong hydrocarbon generation” across the Sichuan Basin. It provides a continuous hydrocarbon supply and forms a regional basis for the accumulation of hydrocarbons in the overlying Middle Jurassic reservoirs. From the perspective of internal stratification and lithologic assemblages, Sha-1 generally represents a transitional system between shore-shallow lake and braided/meandering fluvial environments. It is characterized by variegated mudstone interbedded with gray-green fine–medium-grained sandstone, and can be further subdivided into several sub-members and “thin layers,” reflecting multi-phase channel-filling cycles. Sha-2 is dominated by fluvial to interdistributary bay deposits, with the development of purple-red mudstone interbedded with siltstone. The sand bodies in Sha-2 are generally less continuous than those in Sha-1.

Reservoirs are primarily composed of channel sand bodies, supplemented locally by sand bars and sheet sands. They exhibit geometric features such as “multiple episodes, narrow and discontinuous lateral extent, and thin single-layer thickness.” Core, log, and seismic interpretation indicate that individual fluvial channel sand bodies are typically 2–8 m thick, with widths on the order of several tens to a few hundred meters and lateral extents commonly ranging from several hundred meters to approximately 1–2 km along the channel direction. These geometric characteristics reflect the narrow, laterally variable nature of shallow-water delta–fluvial channel systems in the study area. Residual intergranular pores and microfractures dominate the reservoir space. Recent core and laboratory statistics indicate that the Middle Jurassic Shaximiao Formation sandstones exhibit low porosity and ultra-low permeability. Reported porosity values mostly range from 8% to 12%, with an average of approximately 10%, while permeability generally falls within the range of 10⁻⁴–10⁻¹ mD. According to commonly adopted criteria for tight sandstone gas reservoirs in China, including those applied to Jurassic tight sandstones in the Sichuan Basin, reservoirs with porosity below ~10–12% and permeability below 0.1–1 mD are classified as “tight.” Therefore, the petrophysical properties of the Shaximiao Formation reservoirs in the Yingshan area clearly meet the definition of tight sandstone reservoirs at both regional and national scales.

These stratigraphic, sedimentary, and petrophysical constraints establish quantitative boundaries for subsequent fine-scale sublayer division, well-to-seismic calibration, and channel sand-body prediction. First, sequence boundaries and marker beds provide stable anchors for sublayer window positioning and enhance the temporal consistency of attribute extraction. Second, the geometric characteristics of “thin and discontinuous layers” and the petrophysical background of “low porosity–ultra-low permeability” dictate that reservoir prediction in this area must follow an integrated framework combining amplitude-preserving high-resolution processing, forward-modeling-based multi-attribute fusion, and rock-physics-driven inversion.

3. Methods and Results

In this study, we developed the geological complexity of the Yingshan Shaximiao oil–gas reservoir by adopting an integrated research approach of “from macro to micro, from structure to reservoir, and from qualitative to quantitative.” A systematic workflow—combining structural interpretation, reservoir prediction, and hydrocarbon accumulation analysis—was established, as illustrated in Figure 3. In this workflow, key techniques such as coherent-noise suppression (to enhance seismic signal continuity), wavelet compression (to improve vertical resolution), first-break picking and static corrections (to ensure accurate time positioning), and attribute optimization (to highlight thin-sandstone responses) are implemented at appropriate stages to improve data quality and facilitate interpretation. The workflow aims to accurately delineate reservoir characteristics, identify the main controlling factors of hydrocarbon accumulation, and ultimately guide exploration and development decision-making. To ensure the reliability of amplitude-based attributes and quantitative interpretation, amplitude-preserving processing strategies—including surface-consistent corrections, true-amplitude recovery, Q-compensation, and controlled gain application—were explicitly applied throughout the seismic processing workflow.

First, a detailed structural framework interpretation and evolutionary analysis were conducted. The structural configuration exerts fundamental control over hydrocarbon migration and accumulation. Based on high-resolution 3D seismic data and well-seismic calibration, the major reflection interfaces within the Shaximiao Formation were precisely correlated and interpreted, leading to the establishment of a reliable structural–stratigraphic framework. Key faults that control hydrocarbon migration and accumulation were systematically characterized, including analyses of their geometric and kinematic features. On this basis, structural evolution was reconstructed to reveal the spatiotemporal development of the structural–fault system, clarify the evolution stages of hydrocarbon-source faults, and determine their coupling relationship with hydrocarbon-charging stages, thereby defining the macro-scale structural controls on hydrocarbon accumulation.

Second, fine-scale reservoir characterization and sedimentary microfacies analysis were conducted at the sublayer level. To achieve accurate reservoir prediction, high-resolution sequence stratigraphic division and correlation were performed within the structural framework, subdividing the target interval into multiple key sublayers. For these sublayers, the following approaches were integrated:

Forward modeling analysis: Geological models were built for seismic forward modeling to identify seismic response characteristics of different sand-body geometries, providing a theoretical basis for seismic interpretation.

Seismic attributes and inversion: Multiple seismic attributes (e.g., RMS amplitude, instantaneous frequency) were used for qualitative reservoir identification and boundary delineation. Subsequently, quantitative reservoir prediction—such as porosity and sandstone thickness—was achieved through acoustic-impedance inversion and related geophysical techniques.

Sedimentary microfacies characterization: By integrating inversion results, log facies, and core data, fluvial-channel-dominated sedimentary microfacies were analyzed to determine the spatial distribution patterns of favorable reservoirs. Single-well and well-to-well correlation further validated the reservoir-prediction results, leading to the establishment of a 3D geological model of reservoir distribution. To extrapolate sedimentary microfacies into interwell areas, an integrated, constraint-driven approach was adopted. Microfacies identified at well locations based on core observations and log facies served as calibration points. In interwell regions, seismic attribute responses were interpreted only when they were consistent with forward-modeled seismic signatures of fluvial sand bodies in terms of amplitude polarity, frequency content, and apparent thickness. Rock-physics–constrained seismic inversion results, particularly acoustic impedance and effective thickness volumes, were further used to distinguish sand-dominated from mud-dominated facies quantitatively. Microfacies interpretation was accepted only when seismic attributes, inversion responses, and stratigraphic trends were mutually consistent, thereby reducing subjectivity in interwell extrapolation. Consequently, prediction confidence is highest in areas characterized by strong well control and stable bipolar seismic responses, and decreases toward zones with thinner beds or weaker seismic illumination.

Finally, a comprehensive hydrocarbon accumulation analysis and exploration-target optimization were performed. Based on the structural and reservoir analysis results, the hydrocarbon accumulation characteristics of the study area were systematically summarized. By analyzing the key accumulation factors—including source, reservoir, seal, migration, trap, and preservation—the main controls on hydrocarbon enrichment were identified, and a hydrocarbon accumulation model consistent with the geological setting was established. Integrating favorable structural zones, reservoir-development trends, and accumulation dynamics, the exploration potential of different target zones was quantitatively evaluated and ranked. Ultimately, potential blocks and optimal well-location targets were selected, providing direct scientific support for subsequent exploration deployment.

3.1. Fine-Scale Sublayer Division and Correlation

To enable accurate correlation of thin interbeds and narrow channels, the Shaximiao Formation was subdivided into cross-well comparable sublayer units based on regional sequence stratigraphy. Stable sequence boundaries and marker shales (e.g., maximum lacustrine flooding surfaces) were used as references. Logging facies (GR, SP, sonic/density) and core-identified depositional cycles [16] were combined to recognize vertical rhythms such as coarsening-up or fining-up patterns, forming a three-tier framework of sublayer–sub-member–member. Such frameworks, tied to key stratigraphic surfaces and system tracts/subsequences, are standard practice in sequence stratigraphy to ensure cross-well connectivity and constrain attribute and inversion windows [17,18].

For well-to-seismic calibration, synthetic seismic traces were generated for each key well, as shown in Figure 4, to ensure phase consistency, time–depth accuracy, and waveform stability prior to inversion and attribute analysis. The main procedures are summarized below.

(1)

Reflection coefficient calculation

Based on QC-ed sonic transit time Δt and density ρ, the P-wave velocity is obtained as $V_p=1/\mathsf{\Delta }t$, and the acoustic impedance is then computed as $Z=\rho V_p$. The reflection coefficient at each interface is given by

$$R_i={\displaystyle \frac{Z_{i+1}-Z_i}{Z_{i+1}+Z_i}},$$

(1)

which forms the basis for subsequent synthetic seismogram construction.

(2)

Wavelet estimation near the well

Wavelet estimation was performed using commonly adopted approaches such as statistical matching, constant-phase rotation, and frequency-domain least-squares optimization [19,20,21]. In this study, a least-squares criterion was used to derive a wavelet compatible with the local seismic bandwidth and well-log reflectivity:

$$\underset{w\left(t\right)}{\min }\parallel S\left(t\right)-R\left(t\right)\times w\left(t\right){\parallel }^2,$$

(2)

with the frequency-domain solution

$$W\left(\omega \right)={\displaystyle \frac{R^{*}\left(\omega \right) S\left(\omega \right)}{R^{*}\left(\omega \right)R\left(\omega \right)+\lambda }},$$

(3)

where * denotes complex conjugation and λ is a stabilization factor. A zero-phase wavelet was selected to facilitate direct alignment between seismic peaks/troughs and impedance contrasts. The wavelet length was determined according to the dominant frequency range of the seismic data and the stability of the well tie, ensuring a balance between temporal resolution and noise suppression.

(3)

Time–depth relationship and velocity correction

Checkshot and/or VSP data were used to establish the depth–two-way-time relationship $T\left(z\right)$. Residual velocity drift was corrected to minimize cumulative timing errors:

$$T^{\prime }\left(z\right)=T\left(z\right)+\mathsf{\Delta }{T}_{\mathrm{corr}},$$

(4)

where $\mathsf{\Delta }{T}_{\mathrm{corr}}$ represents the drift-compensation term derived from well-to-seismic misfit analysis.

(4)

Synthetic seismogram generation

Synthetic seismic traces were generated using the convolutional model:

$$S_{\mathrm{syn}}\left(t\right)=R\left(t\right)\times w\left(t\right),$$

(5)

with localized phase and time-shift adjustments applied to optimize waveform similarity between synthetic and field seismic data in terms of amplitude, phase, and frequency content.

(5)

Crossline closure checking

The interpretability of key horizons is validated on strike/dip sections around the well, and the closure misfit is quantified as follows:

$$\epsilon ={\displaystyle \frac{1}{N}}{\displaystyle {{\displaystyle \sum }}_{i=1}^N}\mid T_i^{\mathrm{trace}}-T_i^{\mathrm{marker}}\mid .$$

(6)

Here, smaller misfit values indicate more reliable horizon tracking. In this study, only well ties satisfying a closure misfit of approximately 1–2 ms were accepted for subsequent inversion and attribute analysis.

Following single-well calibration, cross-well correlation was performed along survey lines within the established stratigraphic framework. Sublayer boundaries were used as control surfaces for horizon tracking in the 3D seismic volume, enabling the identification of systematic seismic response variations among different depositional units. In the Shaximiao Formation, fluvial-controlled sand bodies typically exhibit a strong-peak top and strong-trough base bipolar response under suitable frequency bands and velocity contrasts, whereas upward transitions commonly show weaker amplitudes, reduced continuity, and more chaotic reflection patterns. This characteristic “stronger at the base, weaker at the top” seismic expression reflects fluvial–interdistributary bay depositional cycles, but its detectability is strongly dependent on the local frequency content and effective bed thickness, which are quantitatively evaluated in subsequent forward-modeling analyses.

The fine-scale sublayer correlation is shown in Figure 5. Based on the fluvial-channel development characteristics of the study area, the first member of the Shaximiao Formation was subdivided into ten sublayers, corresponding to ten stages of major channel development. On the seismic profiles, this interval is characterized by weak amplitudes, poor continuity, short-axis reflections, and sub-parallel configurations with observable incisions and imbricated reflection patterns. These multi-phase channel-filling cycles and thin interbeds are interpreted to be predominantly controlled by autocyclic processes, such as lateral channel migration, avulsion, and repeated incision-filling under relatively stable climatic and tectonic conditions. Allocyclic controls, including regional base-level fluctuations, are considered secondary and mainly influence accommodation space and vertical stacking intensity. As a result, reservoir distribution is characterized by strong lateral heterogeneity and localized connectivity at the sublayer scale, which increases uncertainty in well-based prediction but allows effective delineation of channel belts and stacked sand-prone zones using high-resolution seismic attributes.

In the quality-control (QC) workflow, fine sublayer correlation was accompanied by three concurrent checks. First, wavelet phase and polarity consistency were verified to avoid peak–trough inversion caused by polarity reversal, using waveform coherence and wavelet spectral-ratio diagnostics. Second, horizon closure and residual statistics were examined, requiring crossline and cross-azimuth misfits to remain within approximately 1–2 ms; otherwise, time–depth calibration or wavelet estimation was re-adjusted. Third, geological plausibility was assessed by ensuring consistency with sedimentary facies interpretation, log-facies patterns, inter-well thickness trends, and channel-distribution geometry. Only sublayer boundaries satisfying all three criteria were retained, ensuring reliable inter-well correlation and a robust foundation for subsequent inversion and attribute-based analysis.

It should be emphasized that these sublayers represent stratigraphically consistent interpretation units rather than seismically independent beds. Their definition is primarily constrained by well-log cycles and sequence-stratigraphic markers, while seismic data provide support for lateral continuity and sand-prone trends within the limits of vertical resolution.

3.2. Detailed Structural and Fault Interpretation

In the Yingshan area and the Shaximiao Formation, where the overall structure is gentle and local minor faults develop, faults not only control sedimentary microtopography and sand-body distribution but also determine trap integrity and hydrocarbon migration pathways [22]. Therefore, it is essential to enhance the identification of subtle faults and minor discontinuities within the 3D seismic volume. To achieve this, a “hierarchical interpretation-integrated discrimination” workflow was adopted.

First, a first-order fault–fold framework was established on conventional pre-stack and post-stack sections. Subsequently, voxel-scale discontinuity and geometric attributes were computed, including coherence and its improved algorithms [23], curvature, multi-band 3D curvature [24], and ant/particle-swarm-based fault-enhancement volumes [25]. Structurally oriented filtering was applied to suppress noise and improve fault discrimination. Horizon and horizontal slices were then used to verify closure, connectivity, and displacement, resulting in a reliable fault system map.

The detailed procedure is as follows:

(1)

Coherence attributes and discontinuity characterization: Coherence, based on waveform similarity, is the fundamental approach for three-dimensional fault detection. Its most widely used form is the eigenvalue-based coherence coefficient (C3 coherence) derived from the cross-trace covariance matrix. For a covariance matrix $C$ within a voxel neighborhood and its ordered eigenvalues ${\lambda }_1\ge {\lambda }_2\ge {\lambda }_3$, the three-dimensional coherence is expressed as

$$\mathrm{Coh}={\displaystyle \frac{{\lambda }_1}{{\lambda }_1+{\lambda }_2+{\lambda }_3}}.$$

(7)

When faults or discontinuities occur, waveform dissimilarity among traces increases and ${\lambda }_1$ becomes significantly larger than the other eigenvalues, resulting in a reduction in Coh; in contrast, within continuous reflectors, the three eigenvalues are comparable, and Coh approaches 1. Coherence attributes displayed on time or horizon slices can directly highlight fault streaks, channels, and other geometric discontinuities, thereby providing a base map for automatic tracking and fault extraction.

(2)

Curvature and multi-scale curvature volumes: Curvature attributes are more sensitive to subtle bending, flexure, folding, and small-offset faults along reflectors, and therefore compensate for the limited resolution of coherence in weak-discontinuity domains. Using spatial variations in the reflector-normal vector field $n\left(x,y\right)$, the principal curvatures $k_{\max }$ and $k_{\min }$ can be derived, and the mean curvature is given by

$$k_{\mathrm{mean}}={\displaystyle \frac{k_{\max }+k_{\min }}{2}}.$$

(8)

High-curvature regions commonly correspond to fault-tip zones, drag folds, channel margins, or depositional terraces. Multi-scale curvature volumes, produced by filters of different wavelengths, enhance sensitivity to structural morphologies at multiple scales, providing a more complete depiction of fault strikes, variations in fault throw, and paleo-geomorphologic features.

(3)

Mathematical formulation of structure-oriented filtering: To enhance subtle fault breaks in seismic volumes—particularly in settings characterized by “continuous reflectors but broken faults” or elevated noise—we apply structure-oriented filtering within voxel neighborhoods. Its fundamental principle is to smooth along reflector dip/azimuth directions while preserving edges in the perpendicular direction:

$$u^{\prime }\left(x\right)={\displaystyle {{\displaystyle \sum }}_i}{w}_i\left(x\right) u\left(x_i\right),$$

(9)

where the weights

$$w_i\left(x\right)=\exp (-{\displaystyle \frac{\left(x_i-x{)}^2\right.}{{\sigma }_s^2}})\exp (-{\displaystyle \frac{{(u\left(x_i\right)-u\left(x\right))}^2}{{\sigma }_r^2}}).$$

(10)

The first exponential term enforces smoothing consistent with structural orientation (structural coherence), whereas the second term preserves amplitude contrasts at faults (edge preservation). Consequently, the filter suppresses random noise while retaining essential structural information such as fault offsets and phase discontinuities.

(4)

Ant-tracking fault-enhancement volumes: Furthermore, the ant-tracking algorithm constructs a composite cost field from coherence, gradients, curvature, and related attributes, and simulates swarm-intelligence behavior to search for optimal fracture/fault pathways. Its basic cost function is

$$C=\alpha \left(1-\mathrm{Coh}\right)+\beta \parallel \nabla A\parallel +\gamma \mid k_{\max }\mid ,$$

(11)

and the ants migrate along paths of minimal cumulative cost, progressively constructing the skeleton of faults. This approach enables automatic linkage of discontinuous fault segments, enhances the detectability of weak faults, and reduces subjectivity in interpretation.

(5)

Structural mapping and displacement-consistency verification: During structural mapping, fault picks on each interpreted horizon are merged into fault surfaces according to the criteria of “displacement consistency plus geometric continuity.” A crossline–inline closure test is performed using

$$\epsilon ={\displaystyle \frac{1}{N}}{\displaystyle {{\displaystyle \sum }}_{i=1}^N}\mid d_i^{\mathrm{inline}}-d_i^{\mathrm{xline}}\mid ,$$

(12)

and when $\epsilon$ is small (typically <2–4 ms), the assembled fault surface is considered reliable. For weak normal faults developed on structural flanks, connectivity is further validated by cross-examining curvature extrema, coherence low-value streaks, and ant-tracking enhancement volumes, thereby improving the robustness of structural interpretation.

The result is shown in Figure 6. The integrated interpretation shows that the Yingshan anticline is overall continuous and complete. Several small-scale normal faults are identifiable at the flanks, potentially connecting deep hydrocarbon source rocks with overlying reservoirs, serving as local vertical migration pathways and lateral sealing units. This fault framework provides spatial constraints for subsequent trap evaluation and favorable-zone delineation.

3.3. Seismic Forward Modeling Analysis

To quantify the detectability of typical sand bodies in the Shaximiao Formation under varying frequency, thickness, and lithology contrasts, and to determine thresholds for “bright spot/bipolar” responses, a convolutional seismic forward modeling approach was used. The dominant frequency of the amplitude-preserved 3D seismic data after high-resolution processing is approximately 50–55 Hz. Considering an average P-wave velocity of about 4.6–5.1 km/s for the Shaximiao sand–mud succession, the corresponding quarter-wavelength vertical resolution is on the order of 5–6 m. This vertical resolution is comparable to the thickness of the target fluvial channel sand bodies (typically 2–8 m), indicating that many of the effective reservoir sands lie near or above the tuning thickness. In the horizontal direction, the dense 3D seismic bin size and effective lateral resolution allow delineation of channel belts and sand-prone zones at scales of several tens of meters, which is sufficient to capture the width and lateral continuity of individual channel sand bodies and their stacked belts. Two-dimensional geological models of sand bodies, including semi-lenticular channel sands, bulbous sand bars, and sheet-like sands, were constructed using core and log-derived parameters. Reflection coefficients were calculated from sonic and density logs and then convolved with zero-phase Ricker wavelets [26] to generate synthetic seismic records, which were subsequently compared and calibrated against actual seismic sections, as shown in Figure 7.

It should be noted that the seismic forward modeling in this study adopts a convolutional approximation based on P-wave reflection theory, rather than full-waveform or elastic modeling. The forward modeling assumes horizontally layered, laterally homogeneous sand–mud successions with gentle structural relief, which is consistent with the geological characteristics of the Shaximiao Formation in the Yingshan area. Elastic parameters, including P-wave velocity and density, were directly derived from quality-controlled well logs and primarily represent brine-saturated sandstone and mudstone conditions. No explicit fluid-substitution modeling was conducted to simulate gas- and water-saturated scenarios separately, and shear-wave effects, mode conversions, and complex scattering were not explicitly considered.

The sensitivity of the modeled seismic responses was evaluated qualitatively by systematically varying key controlling parameters, including the dominant frequency, sand-body thickness, and sand–mud velocity contrast. The results indicate that seismic responses are most sensitive to the interaction between dominant frequency and bed thickness, particularly near the tuning thickness. At the same time, wavelet phase and bandwidth exert secondary but important control on amplitude stability and polarity. Variations in elastic parameters within the observed log-derived ranges mainly influence amplitude magnitude rather than response type. Because gas saturation would further reduce P-wave velocity and density of sandstones and enhance impedance contrast, the detectability thresholds derived from brine-saturated models should be regarded as conservative estimates for gas-bearing sands.

The modeling’s primary objective justifies this choice: to evaluate thin-bed detectability and tuning-controlled amplitude–polarity responses within the dominant seismic bandwidth of the amplitude-preserved 3D data. The Shaximiao Formation in the Yingshan area is characterized by gentle structural relief, small reflector dip angles, and relatively limited lateral velocity variations at the reservoir scale. Under such conditions, first-order P-wave reflection behavior dominates the seismic response, and the convolutional model provides a robust, physically meaningful approximation for analysis of the thickness–frequency–velocity relationship.

Elastic wave effects, such as mode conversions, transmission losses, and complex scattering, are not explicitly simulated in the forward models and are therefore treated as second-order effects for the purpose of thin-sand detectability analysis. Their cumulative influence is partially embedded in the real seismic data through amplitude-preserving processing and well–seismic calibration. Accordingly, the forward modeling results are not intended to reproduce full seismic wavefields, but rather to establish dataset-consistent thresholds and response patterns that can guide attribute interpretation and inversion within a constrained, interpretable framework.

At key well locations, the forward-modeled seismic responses were explicitly calibrated against well-log-derived sandstone thickness, lithology, and impedance contrasts, as well as available core descriptions of depositional facies. Particular attention was paid to the correspondence between modeled strong-peak/strong-trough bipolar reflections and log-identified sand layers of different thicknesses. This calibration establishes a quantitative relationship between thin-sand thickness and seismic amplitude–polarity expression, providing a physical basis for interpreting similar seismic responses away from wells.

The forward modeling analysis follows classical thin-bed tuning principles; however, the emphasis here is on their dataset-specific implications. Within the parameter ranges of this study—mudstone velocities of approximately 4.0–4.35 km/s, sandstone velocities of approximately 4.6–5.1 km/s, and a zero-phase 50–55 Hz Ricker wavelet—the detectability of thin sand bodies is strongly controlled by the interaction between dominant frequency, velocity contrast, and bed thickness.

It should be noted that the elastic parameters used in the forward modeling are primarily derived from well-log measurements that largely represent brine-saturated sandstone conditions. No explicit fluid-substitution modeling was performed to simulate gas-versus water-saturated scenarios separately. This approach was adopted to define conservative and internally consistent detectability thresholds under a unified reference state.

From a rock-physics perspective, gas saturation would generally reduce both P-wave velocity and bulk density of sandstones, leading to a larger impedance contrast relative to surrounding mudstones. Such changes would enhance reflection amplitude and polarity stability, thereby improving the detectability of thin gas-bearing sands and potentially lowering the effective thickness threshold compared to the brine-saturated case. Therefore, the modeled detectability limits in this study should be regarded as conservative estimates, while gas-charged sand bodies calibrated at producing wells are expected to exhibit stronger seismic responses under equivalent frequency conditions.

Modeling results indicate that sand layers thinner than approximately 2–4 m are dominated by waveform interference, producing unstable or ambiguous amplitude responses. It should be emphasized that the 2–4 m thickness threshold is not an absolute value, but a conditional detectability range controlled by seismic bandwidth, noise level, and wavelet phase. Forward modeling tests show that increasing the dominant frequency or effective bandwidth narrows the tuning zone and improves vertical resolution, allowing thinner sand bodies to generate more stable bipolar amplitude responses. Conversely, reduced bandwidth or elevated noise levels enhance waveform interference, resulting in less stable amplitude expression and a higher effective detectability threshold.

Wavelet phase also plays a critical role. The zero-phase wavelet adopted in this study provides optimal peak–trough symmetry and maximizes tuning sensitivity, thereby stabilizing the polarity of the amplitude near the tuning thickness. In contrast, non-zero-phase wavelets would broaden the interference zone and further reduce the reliability of amplitude anomalies associated with very thin beds. Therefore, under the dominant frequency (~50–55 Hz), signal-to-noise conditions, and zero-phase wavelet assumption of the present dataset, sand bodies thinner than approximately 2–4 m tend to exhibit unstable seismic responses, whereas thicker sands are more consistently detectable. This threshold should be regarded as dataset-specific and conservative, rather than universally applicable.

When the thickness exceeds 6 m, clear strong-peak/strong-trough bipolar reflections emerge, significantly enhancing continuity and interpretability. Increasing sand–mud velocity contrast further stabilizes polarity and amplitude expression, although the critical thickness remains constrained by dominant frequency and wavelet bandwidth.

These forward-modeling results provide quantitative constraints for attribute window selection and anomaly threshold definition, thereby establishing a direct link between seismic response characteristics and interpretable reservoir thickness in subsequent attribute and inversion analyses. On the one hand, attributes such as top-to-bottom peak separation, local dominant frequency, and instantaneous amplitude or “sweetness” can be aligned with forward-modeled responses to determine optimal extraction windows and anomaly thresholds, thereby avoiding misinterpretation of noise or topographic effects as fluvial sand bright spots. On the other hand, the relationship among detectable thickness, dominant frequency, and velocity contrast can be converted into a detectability chart, which allows rapid assessment of the geometric and petrophysical plausibility of attribute anomalies and serves as a prior constraint for subsequent rock-physics–driven inversion. By comparing with actual seismic sections, when zones of strong bipolar amplitude consistent with forward modeling predictions appear and satisfy the frequency–thickness relationship defined in the detectability chart, it can be reasonably inferred that the sand body thickness is approximately 6 m or more and has the potential to form an effective reservoir.

Therefore, seismic interpretation in this study does not attempt to resolve every thin bed individually. Instead, it focuses on identifying sand-prone sublayer packages and effective thickness intervals that are consistent with seismic resolution, forward-modeling constraints, and well calibration, thereby minimizing the risk of over-interpretation beyond seismic limits.

3.4. Seismic Attribute Analysis and Reservoir Prediction

Based on the refined stratigraphic framework and well–seismic calibrated horizons, a “broad-to-fine” attribute selection strategy was implemented for each sublayer of the Sha-1 section in the Shaximiao Formation. Candidate attributes from time-domain amplitude measures (e.g., peak amplitude, RMS amplitude), instantaneous attributes (e.g., envelope, instantaneous frequency, phase), and frequency-domain attributes were first screened using well-based correlation analysis with sandstone relative thickness and porosity. Subsequently, multicollinearity diagnostics were applied to evaluate redundancy among candidate attributes, ensuring that selected attributes exhibit both strong geological sensitivity and low mutual dependence. Attributes with stable correlations with reservoir parameters and minimal redundancy were retained for subsequent analysis.

The results indicate that amplitude-related attributes exhibit stronger sensitivity to vertical tuning responses of thick sandstone bodies (Figure 8). In particular, RMS amplitude and sweetness provide robust indicators of the relative thickness and connectivity of fluvial channel sandstones in the study area. Sweetness is defined as

$$S={\displaystyle \frac{\mathrm{Envelope}}{\sqrt{f_{\mathrm{inst}}}}}={\displaystyle \frac{\mid A+iH\left(A\right)\mid }{\sqrt{f_{\mathrm{inst}}}}},$$

(13)

where $A$ is the instantaneous amplitude, $H\left(A\right)$ is the Hilbert transform, and $f_{\mathrm{inst}}$ denotes the instantaneous frequency. High sweetness values typically correspond to low-frequency, high-energy reflections, often highlighting isolated sandstone bodies within shale-prone backgrounds. Compared with RMS amplitude and other conventional attributes, sweetness more effectively suppresses high-frequency noise and thin-bed interference because the normalization by instantaneous frequency reduces the impact of rapid vertical fluctuations, enhancing the detectability of narrow or discontinuous sand bodies. This property facilitates the identification of narrow channels, distributary channels, and channel-fill sequences that may be obscured in single-attribute maps.

It is well recognized that instantaneous frequency attributes derived from the Hilbert transform are sensitive to noise and phase instability, particularly in thin-bed interference zones. To stabilize instantaneous frequency estimates, several preprocessing and quality-control measures were applied prior to attribute extraction. These include amplitude-preserving processing to enhance signal-to-noise ratio, structure-oriented filtering to suppress random noise while preserving reflector phase continuity, and strict wavelet phase and polarity calibration through well-to-seismic ties. A zero-phase wavelet was consistently adopted to minimize phase-induced frequency artifacts.

Furthermore, instantaneous frequency was not interpreted in isolation. Instead, it was mainly incorporated into the sweetness attribute, where the square root of the instantaneous frequency normalizes amplitude. This formulation effectively suppresses spurious high-frequency fluctuations and reduces sensitivity to local noise, making it more robust for highlighting thin and discontinuous sand bodies. Attribute extraction was conducted within horizon-constrained, layer-parallel windows defined by sequence boundaries and marker beds, thereby avoiding stratigraphic mixing and artificial frequency variations associated with constant-time slicing. Consequently, instantaneous frequency anomalies were considered meaningful only when they were consistent with forward-modeling-predicted frequency–thickness behavior and supported by amplitude and inversion responses.

Within this framework, horizon-consistent attribute slices were extracted for each target sublayer using a layer-parallel window, which was slightly adjusted based on the sublayer thickness to align the tuning peak. Three-dimensional visualization reveals banded zones of positive-amplitude anomalies that occur widely in the core of the anticline and along its extension, consistent with the orientation of the channel system. Areas with high anomaly values typically correspond to predicted thicker sand bodies and may serve as priority candidates for potential sweet spots.

To reduce the non-uniqueness caused by brightness effects, local noise, and topographic interference in single-attribute analysis, multi-attribute integration was implemented under physically meaningful constraints. Sensitive attributes such as RMS amplitude, sweetness, and envelope dominant frequency were fused using fuzzy logic, unsupervised clustering, or statistically weighted learning. Attribute weighting and clustering were guided by their consistency with forward-modeling-derived frequency–thickness responses and their relative sensitivity to effective sandstone thickness, rather than by purely mathematical criteria. This integration strategy emphasizes attributes that jointly reflect geometric continuity and petrophysical contrasts of sand bodies, thereby improving the robustness and geological interpretability of the fused response volumes.

Based on the integrated interpretation of selected and composite attributes, planar maps of sand-body distribution and predicted effective thickness were generated for each sublayer, as shown in Figure 9. Overall, the sand bodies of the Sha-1 interval exhibit a northeast-trending belt-like distribution with relatively good continuity. In contrast, sand bodies in the Sha-2 interval are generally thinner and less laterally continuous. This difference is mainly attributed to depositional and accommodation controls: Sha-2 formed under relatively limited accommodation space and more variable fluvial energy conditions, leading to discontinuous, isolated channel sands, whereas Sha-1 represents a more stable, channel-dominated depositional system with enhanced accommodation, favoring laterally extensive, stacked sand bodies. Post-depositional modification is considered secondary in controlling the observed continuity contrast. Guided by the forward-modeling-derived detectability thresholds, attribute anomalies were interpreted only when their amplitude, frequency content, and spatial continuity were consistent with the modeled frequency–thickness relationships, thereby reducing uncertainty associated with thin-bed interference.

3.5. Rock-Physics Analysis and Seismic Inversion

Building on qualitative attribute prediction, an integrated rock-physics–seismic inversion approach was introduced to obtain quantifiable, comparable reservoir parameter volumes. First, multi-well log curves were quality-controlled and standardized to establish statistical relationships between lithology and properties. In this area, sandstones exhibit lower acoustic impedance than mudstones, which is controlled by both higher sonic transit times and lower densities [27,28]. The impedance contrast between sandstones and mudstones provides the physical basis for distinguishing sand from mud in inversion. Meanwhile, based on regional analogues and local well-log statistics, an empirical porosity–impedance relationship was established, which can be expressed using either an exponential or linear model:

$$\varphi =a{Z}^{-b}\mathrm{or} \varphi =c-dZ,$$

(14)

where $a$, $b$, $c$, and $d$ are derived from well-log regression. This relationship constrains the reasonable range and uncertainty of porosity inversion results, providing a reliable prior parameter for subsequent identification of effective thickness and gas-bearing potential.

For inversion, a constrained sparse-spike/sparse-point inversion (SSI/CSSI) approach [29] was employed, aiming to recover geologically meaningful sparse reflection series while maintaining a good fit to seismic data. The optimization objective is expressed as

$$\underset{r\left(t\right)}{\min }\parallel d\left(t\right)-w\left(t\right)\times r\left(t\right){\parallel }_2^2+\lambda \parallel r\left(t\right){\parallel }_1,$$

(15)

where $d\left(t\right)$ is the observed seismic trace, $w\left(t\right)$ is the estimated wavelet, $r\left(t\right)$ is the reflection coefficient series to be estimated, and $\lambda$ is the sparsity constraint parameter.

Absolute impedance volumes are obtained by integrating the reflection coefficient series and adding low-frequency trends derived from multiple wells:

$$Z\left(t\right)=Z_0\exp \left(\sum r\left(t\right)\right).$$

(16)

For tight sandstone gas reservoirs, sparse inversion effectively enhances vertical resolution under thin interbeds and limited bandwidth, and multi-trace constraints improve lateral continuity, yielding more reliable sand–mud differentiation and effective thickness estimation.

The inversion workflow includes ① preprocessing and depth–time alignment of well curves; ② estimation and stability verification of wellbore wavelets; ③ joint construction of low-frequency impedance trends from multiple wells; ④ constrained sparse spike inversion to obtain three-dimensional impedance volumes; ⑤ well-based calibration of effective thickness and threshold setting (e.g., integrating impedance upper/lower bounds with net sand identification criteria for net thickness calculation); and ⑥ cross-validation with attribute-fusion volumes to evaluate lateral consistency and uncertainty. Through these steps, the reliability of inversion-derived parameters is jointly constrained by well control, rock-physics relationships, and independent seismic-attribute responses. Uncertainty and sensitivity are mainly addressed by (i) forward-modeling-constrained thickness–frequency relationships that define the detectability limits of thin sand bodies, (ii) comparison and cross-validation between single-attribute, fused-attribute, and inversion-derived results to assess lateral stability, and (iii) well-based calibration using effective thickness and production test data to evaluate prediction consistency. Although a fully probabilistic uncertainty quantification was not performed, these integrated constraints provide a practical and geology-consistent assessment of uncertainty for the quantitative interpretation results.

The inversion results, as shown in Figure 10, demonstrate a high degree of consistency between the inverted section and the lithology encountered in wells. The sandstone bodies exhibit distinct inversion responses, with smooth lateral transitions and high reliability. Moreover, the channel sand bodies correspond to low-impedance facies, showing good agreement with the effective intervals revealed by the wells.

4. Discussion

4.1. Hydrocarbon Distribution Characteristics

The Shaoximiao Formation in the Yingshan area generally exhibits multi-layered stacking and overall oil- and gas-bearing characteristics. Integrated well–seismic data indicate that the middle to lower sections of the Sha-1 interval host the most concentrated gas layers. Test results primarily show industrial gas and oil flow, with stable water layers rarely observed, suggesting favorable sealing and preservation conditions in the area. Gas reservoirs with minor oil rims may dominate the accumulation style.

Based on well log data in this area and comparisons with neighboring regions, hydrocarbon-bearing and water-bearing layers were distinguished using criteria mainly based on resistivity and porosity. The hydrocarbon layers can be further subdivided into Class I and Class II according to quality. As shown in Figure 11, Class I hydrocarbon layers are predominantly found in the back-structure cores and along the axial extensions of channel sand bodies, with single-layer thicknesses typically ranging from 2 to 8 m and spatially distributed as patchy or belt-like accumulations. Class II hydrocarbon layers and dry layers are more commonly located in the flanks of the back-structure or in areas where sand bodies thin locally or pinch out.

It can be seen that thick, high-resistivity and high-porosity sand “sweet spots” serve as the main carriers for hydrocarbon accumulation [30]. In contrast, small, low-porosity–permeability sands and marginal sand bodies tend to exhibit low production or even appear as dry layers. On the plan view, hydrocarbon enrichment generally extends along the back-structure trend, forming multi-phase channel-stacked “preferred belts” at axial highs and minor structural highs, with a gradual decrease toward the flanks away from these high points. Areas near hydrocarbon-supplying faults often show richer hydrocarbon indications, and in some wells, oil and gas responses are observed even in the lower Sha-2 interval. This indicates that deep hydrocarbon supply combined with structural uplift promotes a concentration pattern along fault zones and back-structure highs, consistent with the regional enrichment characteristics of tight gas controlled by shallow-water delta–fluvial systems. Within this depositional framework, fluvial channel sand bodies play a dominant role in vertical connectivity and sweet-spot development, whereas deltaic distributary and mouth-bar deposits mainly contribute to lateral heterogeneity and local sand-body widening, jointly shaping the multi-phase stacked reservoir architecture.

4.2. Analysis of the Main Controlling Factors for Reservoir Formation

Based on the reservoir geometry, petrophysical properties, and gas distribution characteristics, hydrocarbon accumulation in the Shaximiao Formation at Yingshan is jointly controlled by three main factors. The first is deep hydrocarbon-supplying faults, which provide vertical migration pathways. The underlying Xujiuhe Formation and other source rocks generate sufficient hydrocarbons, which migrate upward along reverse faults and associated fracture systems, continuously charging the overlying Shaximiao Formation. The second factor is structural highs, represented by the Yingshan anticline, which provide the direction for hydrocarbon migration and the space for entrapment. The axial and high positions are most favorable for forming multi-layer stacked accumulations. The third factor is the fluvial sand bodies, which constitute the main storage units. Thick, well-connected stacked sands covered by mudstones form effective structural–lithological composite traps that control specific hydrocarbon enrichment zones and the main productive layers.

As shown in Figure 12, compared with typical conventional sandstone gas reservoirs, tight gas in this area exhibits a stronger coupling among “source-fault control—structural position control—high-quality microfacies control.” Based on this, a “deep source–upper reservoir” secondary accumulation model is established: gas generated in deep source rocks during the late stage migrates upward along source-root faults; the gas is effectively trapped in thick, sweet-spot sands at structural highs and along the anticline axis, with overlying and interbedded mudstones providing stable top and bottom seals, resulting in multi-layer stacked reservoirs with long-term preservation. This model aligns with the widely recognized “deep source–upper reservoir/composite accumulation” concept for Jurassic formations in the Sichuan Basin, emphasizing the critical role of source-root faults in secondary charging and in situ accumulation of tight gas. The model provides clear criteria for quickly delineating potential enrichment zones: within the overlay of predicted effective thickness maps and structural contour lines, priority should be given to thick sandstone stacked belts located near faults and at axial highs.

Compared with other tight gas basins, such as the Sulige area in the Ordos Basin, where controlling factors mainly rely on large-scale structural highs and local microfacies, the Yingshan area shows a stronger coupling among deep-source-supplying faults, structural highs, and fluvial-channel microfacies. This coupling results in more continuous multi-layer stacked sweet-spot sand bodies and a more predictable enrichment pattern, highlighting the practical advantage of the Yingshan reservoir-forming model for rapid identification of favorable zones. Meanwhile, the cap rock and preservation conditions are equally important for the long-term retention of tight gas. The mudstones within the Shaximiao Formation, combined with the overlying Suining Formation mudstones, form a stable sealing system. Coupled with regional gentle uplift, this helps maintain the pressure system and interface stability of the gas reservoirs. Differences in fault activity timing, charging sequence, and sealing integrity determine the local enrichment intensity and stratigraphic variations, which can be further quantified and refined in subsequent well control and inversion updates.

5. Conclusions

Based on a unified well–seismic calibration and high-fidelity amplitude-preserving processing, this study integrates forward-modeling-constrained attribute selection/fusion with rock-physics-driven inversion to achieve high-precision characterization and quantitative interpretation of tight sandstone gas reservoirs in the Shaximiao Formation at Yingshan. The study demonstrates the following: (1) The Shaximiao Formation, developed on a large gently tilted anticline, contains multi-phase stacked fluvial sand bodies, forming multi-layer structural–lithological composite gas reservoirs. Hydrocarbon enrichment is controlled synergistically by source-fault pathways, structural highs, and thick sweet-spot sands. (2) Fine 3D seismic interpretation significantly enhances thin-layer identification. When the dominant frequency, bandwidth, and sand–mud velocity contrast satisfy detectability conditions, thin sands of approximately ~6 m can exhibit stable “double-polarity bright spots” on seismic data, serving as direct geophysical indicators for preferred sweet spots. Rock-physics-constrained inversion provides 3D volumes of acoustic impedance and effective thickness, quantifying the spatial distribution of sweet-spot reservoirs. (3) Vertically, the main productive interval is concentrated in the middle-lower part of the Sha-1 section, with Class I sweet-spot reservoirs typically 2–8 m thick, predominantly distributed along the anticline axis and near hydrocarbon-supplying faults. Upper sand bodies locally thin or pinch out, with relatively weaker hydrocarbon-bearing capacity. The strengths of this study include the integration of high-resolution seismic processing, forward-modeling-guided attribute selection, and rock-physics inversion, which together allow reliable detection and quantitative characterization of thin, heterogeneous sand bodies. However, limitations exist: the inversion and attribute calibration rely heavily on well control, and areas with sparse well coverage may reduce the reliability of reservoir prediction. In addition, the detectability of thin sands is constrained by seismic frequency and the sand/mud velocity contrast, which may limit resolution for ultra-thin or low-contrast intervals. Future work may explore advanced techniques such as 4D monitoring and artificial intelligence-based methods to further enhance seismic interpretation and reservoir prediction.