The Anyue Gas Field is located in the Central Sichuan Paleohigh in the Sichuan Basin. Reservoirs in this field are mainly of shoal origin formed in large-scale shallow-water and high-energy environment in the Lower Cambrian Longwangmiao Formation. Early karstification is an important factor controlling the formation of reservoir^[1]. The heterogeneity of the reservoir is extremely strong, which is reflected by rapid changing of sedimentary micro-facies, porosity, permeability, and thickness^[2,3,4,5]. In recent years, many works have been done on lithofacies, sedimentary facies^{[3, 6-7]}, diagenesis^[8,9], sequence stratigraphic framework^[10], and seismic interpretation^[11,12], etc. Nevertheless, problems still exist in hydrocarbon exploration and development in the Longwangmiao Formation. One problem is that the current well spacing is too large to characterize detailed distribution of strata and reservoirs. The other is that the target zone is buried so deeply that the dominant frequency and resolution of seismic data are low. One major challenge is how to integrate different data under the guidance of geological model so as to improve the accuracy and reliability of interpretation.

Seismic sedimentology is a new discipline emerged after seismic stratigraphy and sequence stratigraphy. It is a science of studying sediment rocks and sedimentary processes by using seismic data^[13,14,15]. Seismic sedimentology can be viewed as a branch of sedimentology. Under the restriction of current technologies, seismic sedimentology is a discipline that studies lithology, sediment origin, sedimentary systems, and basin history by integrated analysis of seismic lithology (characteristics of lithology, thickness, petrophysical properties, and influence of fluids) and seismic geomorphology (sedimentary paleogeomorphology, erosional paleogeomorphology, and the relationship and the evolution of topographic units)^[16]. In recent years, many studies have been conducted by Chinese scholars on the methodology and application of seismic sedimentology. However, most of the applications are limited in siliciclastic strata. Few discussions are on carbonates, especially on very ancient marine carbonate strata^{[17,18,19,20]}. Compared to siliciclastics, carbonate strata are characterized by different basin filling mode, sedimentary systems, diagenesis, petrophysics and seismic reflection. Therefore, methodology and workflow of seismic sedimentology in carbonates should also be different. In other words, carbonates are a less-known domain for seismic sedimentology.

As for methodology, lacking of seismic-facies indicators (e.g. platform-margin progradational reflection) in large platform carbonate sequences may lead to failure of restoring paleogeomorphology and paleo bathymetry by conventional seismic geomorphology (i.e., plane-view geometry of sedimentary landforms). Therefore, traditional structural paleogeomorphology method should be explored to assist the restoration of paleogeomorphology. Methods of cast and isopach, which were first systematically discussed by Martin^[21] in 1966, are still useful. In application, however, the calibration method should be upgraded to improve the matching between the two. In seismic lithology, carbonate lithologies are poorly related to acoustic impedance (AI). Therefore, much attention should be paid to the relationship between porosity and lithology (lithofacies) association. Correlation of high-energy facies, porosity, AI or other seismic attributes should be established to quantitatively predict reservoir. In addition, despite of much more complicated diagenesis in carbonates than in siliciclastics, some special seismic attributes are still effective in characterizing diagenesis in carbonates given specific favorable conditions. The mode-driven seismic geomorphology and the data-driven seismic lithology have to be integrated to generate a new quantitative, more accurate and reliable method to predict carbonate reservoirs.

The Cambrian System in the Central Sichuan Paleohigh includes five formations/groups^[22] (Fig. 1). The lowermost Qiongzhusi Formation and Canglangpu Formation are marine siliciclastics, which are overlain by carbonates of Longwangmiao Formation and siliciclastics of Gaotai Formation. The uppermost of Cambrian is carbonates of Xixiangchi Group. Thickness of carbonates in Longwangmiao Formation on the Paleohigh is 60-150 m, which thickens to 150-300 m in the surrounding depressions. The shallow-water environment with high energy on the Paleohigh offered favorable conditions for the forming of shoal sub-facies^{[1,2,3,4,5,6,7]}. Top and bottom boundaries of Longwangmiao Formation are confined by siliciclastics, which generates high AI contrasts, enhancing the seismic reflections at the two boundaries and benefiting seismic interpretation.

The Gaoshiti-Moxi area is located in the middle of Central Sichuan Paleohigh (Fig. 1). The 3D seismic survey covers an area of 2 500 km² with 35 wells. The seismic survey acquired in 2011 possesses a frequency range of 10-60 Hz, with the dominant frequency of 30 Hz. The seismic resolution is about 50 m. Thanks to the high signal-to-noise ratio, detectable limit can reach 15 m if a horizontal display (stratal slice) is used.

Synthetic seismogram of MX 19 is shown in Fig. 2. The synthetic trace using -90°, 30-Hz Rick wavelet correlates well with the stacked and migrated seismic traces. Despite of mi-nor deviation, the major stratigraphic boundaries (e.g., top, middle, and base of Longwangmiao Formation, all of which are third-order sequence boundaries) are traceable on the seismic sections and can be correlated throughout the study area. The two main seismic reflection units in Longwangmiao Formation (i.e., the upper reservoir unit with high porosity and low AI and the lower non-reservoir unit with low porosity and high AI) correspond to a trough event (negative polarity) and a peak one (positive polarity), respectively. Seen from the actual comparison results, the -90° phased seismic trace correlates to the velocity and porosity logs, which effectively simplifies the well-seismic correlation, benefitting geologic analysis. In contrast, on the standard zero-phased seismic profile, the analysis of every AI unit involves two events, making seismic interpretation more challenging.

The structure in Gaoshiti-Moxi area is a large asymmetric anticline, which is complicated by faulting in different orientations. The longitudinal faults are inherited from the Sinian^[4]. The other faults form a NW-SE oriented en echelon fault system. To date, the distribution and origin of the later fault system have not been discussed systematically. The gas-water contact in gas reservoirs of Gaoshiti-Moxi area is lower than the spill point, and the field was classified as a structural-stratigraphic combination trap^[1].

According to the observation of conventional cores from MX 19 (Fig. 3), four main lithofacies associations were recognized from the Longwangmiao Formation in the study area, i.e., (1) grainstone (Fig. 4a) and grain-dominated packstone consisted of large quantity of ooids (with the diameter of 0.5-1.0 mm) and small amount of intraclasts, pellets, and bioclasts; (2) intraclast grainstone and grain-dominated packstone. Many of the intraclasts were found in grainstone with a diameter up to 7.0 mm; (3) pellet packstone and wackestone, which contain abundant pellets and minor bioclasts; (4) massive mudstone with some local laminations (Fig. 4b). All rocks have been dolomitized.

From the lithology associations and vertical trend (Fig. 3), three sub-facies can be interpreted in the carbonate platform facies: (1) Shoal sub-facies, which includes three micro-facies. The first one is shoal crest, which is mainly consisted of pellet oolitic grainstone, as well as intraclast grainstone and grain dominated packstone, with thin (0.5-1.0 m) interbed of mudstone and wackestone. Water energy in this micro-facies is referred to be the strongest. Each of the upward-shoaling cycles is thicker than 3.5 m. The second is shoal margin, which has similar lithofacies association with shoal crest, however thickness of grainstone and packstone decreases to 1.5-2.0 m, while mudstone and wackestone interbeds thicken to 1.0-2.0 m. Water energy in this micro-facies is referred to be strong to moderate. The third is intershoal, which is mainly consisted of thick (3.0-5.0 m) wackestone with minor mudstone and packstone. Water energy is interpreted to be moderate. (2) Lagoon sub-facies, which is dominated by wackestone with minor packstone. This sub-facies was mainly formed in lower and the uppermost of Longwangmiao Formation. Neither bioclast nor bioturbation is found in this low water energy environment. (3) Deep shelf sub-facies, which is composed of mudstone and wackestone in lower part of Longwangmiao Formation. Bioturbation is common in this sub-facies. The interbedded grainstone and packstone (0.1-0.5 m) can be interpreted as storm or density current sediments. Water energy of this sub-facies is probably the weakest.

According to lithofacies associations observed in conventional cores in MX 19, two third-order sequences^{[2-3, 10]} were recognized in Longwangmiao Formation (i.e., Sequences 1 and 2 in Fig. 3). The sequences start from transgressive systems tract (wackestone and mudstone) and end up to the top of grainstone-dominated shoal sub-facies (SB2) or restricted lagoon (SB3) formed in highstand systems tract. Maximum flooding surface (MFS) lies in mudstone formed in deep shelf sub-facies. The two third-order sequences can be tracked in the whole study area by using well and seismic data.

From porosity logs (Fig. 3), it was observed that reservoirs with high (4%-8%) porosity are mainly in shoal sub-facies (including shoal crest and shoal margin). In other words, shoal and reservoir rocks share the same locations. Reservoir thickness (RH, thickness of carbonate rocks with porosity larger than 4%) calculated by porosity logs, in turn can indicate the existence of shoal sub-facies. Therefore, high-energy sedimentary facies is the main controlling factor for the distribution of carbonate reservoir rocks.

Selective dissolution, dolomitization, and cementation were major types of diagenesis that control porosity. Cementation is the key factor responsible for low (<3%) matrix porosity in Longwangmiao Formation. Vugs are unevenly distributed in grainstone and packstone. Size of the vugs is usually larger than that of the intergranular pores or even the grains (Fig. 4a). Selective leaching of the fresh water is one probable origin. Vugs formed by dissolution may emerge before dolomitization, and remain intact afterward. Selective leaching is an important factor that controls the distribution and property of reservoir. The vugs tend to touch with each other when they are densely spaced, which may dramatically increase the permeability (Fig. 4c).

The term seismic geomorphology was first proposed by Posamentier^[23]. It is a mode-driven method by interpreting amplitude (or other seismic attributes) map guided by sequence stratigraphic or sedimentary facies model. Similar to traditional cast and isopach methods, interpretation of paleostructure and sedimentary rate is based on geologic model, leading to qualitative results.

Structure of Longwangmiao Formation was modified by later tectonic activities. Therefore, cast method was used to restore the paleogeomorphology. Being closest to the top of Longwangmiao Formation, the main seismic-stratigraphic boundary at the base of Permian was artificially flattened (Fig. 5). The resultant “cast” map (the thickness between the base of Permian and the top of Longwangmiao, Fig. 6) can be used to interpret relative relief of Longwangmiao paleotopography. Thanks to minor erosion (20-80 m) in the 3D seismic survey, this paleo-relief should represent the sedimentary paleogeomorphology of Longwangmiao Formation. Well data reveal that the paleohighs are indeed the favorable area for the development of shoal sub-facies with high porosity. The only exception is the deep slope area in southeast of the study area where the recovered paleostructural relief is larger than 240 m (dash line in Fig. 6). In this area, the relationship between paleohighs and porous shoal sub-facies is unclear. One possible reason is the differential tectonic deformation (i.e., differential subsidence and/or folding) during the uplifting after the deposition of Longwangmiao Formation.

Another well-seismic profile (B—B', Fig. 7), which was flattened to the base of Permian crosses some high-angle (50°-80°) and small (40-80 m) syndepositional faults. These faults separate thin horst sediment in paleohighs and thick graben sediment in paleolows in Longwangmiao Formation. This phenomenon indicates that the faults were formed during the sedimentation of Longwangmiao Formation. Controlled by the syndepositional faults, grabens subsided continuously. The accommodation space remained to be relatively large, resulting in weaker water energy. In addition, the areas of paleogeomorphologic low sustained less erosion after deposition, resulting thicker residual sediments. On the gross thickness (GH) map of Longwangmiao Formation (Fig. 8), these horsts and grabens distribute alternatively and form multiple local subsidence centers. Consequently, Fig. 8 may approximately reflect the sedimentary paleogeomorphology. Calibrated with well-measured reservoir thickness, it becomes clear that the thinner horsts are the favorable area for developing shoal sub-facies with high porosity.

An integrated interpretation of the results of cast and isopach methods (Figs. 7 and 8) revealed the relationship between reconstructed paleogeomorphology and present residual thickness for Longwangmiao Formation. Further, the sedimentary paleogeomorphology of Longwangmiao Formation was qualitatively determined. There is a negative correlation between gross thickness (GH) and reservoir thickness (RH) at well sites (Fig. 9). RH decreases with increasing GH. The sample points can be visually divided into two groups, which come from two sub-facies. The first is shoal sub-facies with larger RH and smaller GH at paleohighs. The second is lagoon and/or deep shelf sub-facies with smaller RH and larger GH at paleolows. The cast thickness at the top of Longwangmiao Formation, therefore, correlates to the isopach of residual Longwangmiao sediments, i.e., reservoir thickness decreases with increasing cast thickness, showing a negative correlation. Although there is no result in deep slope area where the cast thickness is over 240 m, relationship between paleohigh and paleolow agrees fairly well with that disclosed by isopach method at most well locations.

According to characteristics of isopach- and cast- derived paleogeomorphology, the sedimentary facies in Longwangmiao Formation was restored by combining the two factors of paleogeomorphology and well-derived lithofacies (Fig. 10), previously referred as “single factor method”^[24]. Perceived from Figs. 6, 8, and 10, the paleohighs and paleolows are mainly distributed along the syndepositional faults with a general latitudinal direction, except in Gaoshiti area, where the pattern of paleogeomorphology inherits features related to longitudinal deep faults. Oolitic and grain shoals were formed on paleohighs in shallow and high-energy environment. The high-energy shoal occurred in shoal crest, and extended to the surrounding area to form shoal margin. Sediments with higher mud content are common in moderate-to-low energy environment of paleolows in intershoal. Deep shelf facies are mainly located in the southeastern and southern downdip area. The low-energy mudstone is the dominated lithology, which is interrupted by several thin beds of short-term storm-related density current sediments.

Both the paleogeomorphology map (Fig. 6) and the gross thickness map (Fig. 8) of Longwangmiao Formation reflect a syndepositional en echelon fault system in addition to the inherited longitudinal deep fault in the study area. Most faults have a curved shape, which convex to the south, converge to the east, and diverge to the north, northwest, and west in an orderly form. The fault pattern implies that blocks on the opposite sides of each fault were sliding while depositing, which was controlled by dextral slip activities around the Central Sichuan Paleohigh in Longwangmiao time. In the meanwhile, the curved fault planes lead to uneven local compressional and extensional conditions. Uplifts were formed where compressional forces dominated, while small basins were initiated where extensional forces prevailed. This kind of tectonic geomorphic features is common in strike-slip structures. The deep-water basin in offshore Los Angeles is an ideal analogy. Controlled by San Andres strike-slip faults, alternative horsts and grabens were formed in the basin.

Seismic lithology is to directly transform large quantity of seismic attribute information generated along a geologic horizon between wells into a lithology-indicative map. Although the process involves use of petrophysical model, the results are mainly determined by data. This data-driven method is quantitative, which offers a supplement to the mode-driven methods.

Whether conventional poststack seismic attributes are useful in predicting reservoir depends on many factors. The most critical is if there is a significant AI contrast between reservoir and non-reservoir rocks. In MX 19 well, the wireline-log-interpreted porosity of Longwangmiao Formation decreases with increasing AI, revealing a negative correlation. Therefore, the porosity ranges can be used to define AI values of reservoir and non-reservoir (Fig. 11). Although multiple lithofacies were identified in the core description (Fig. 3), these lithofacies can be simply divided as reservoir (porous dolomitized grainstone and packstone) and non-reservoir (other lithologies), separated at around 4% porosity. The reflection coefficient calculated by average AI contrast between the two is about 0.03, capable to generate moderate reflection energy. Therefore, dynamic seismic attributes such as amplitude can be used to predict reservoir within the seismic detectable limit.

We firstly tried to predict reservoir thickness in Long- wangmiao Formation directly by the amplitude from original

data. The result is not satisfactory by showing a correlation coefficient of 0.6 with well-derived RH values. The main reason is that the seismic data are characterized by a resoluble limit of around 50 m, while the reservoir thickness in Longwangmiao Formation is up to 63 m. Since the linear correlation between amplitude and thickness only exists in seismically thin beds^[25,26], the poststack amplitude cannot be used to accurately calculate reservoir thickness.

PCA of frequency-decomposed amplitude can improve the prediction significantly. Firstly, the poststack seismic data were numerically separated into low-, moderate-, and high-frequency panels, aiming at expanding amplitude tuning range and improving the correlation between amplitude and thickness, as well as resolution^[27]. Secondly, a multiple attribute PCA^[28] was applied to the frequency-decomposed amplitude panels, which transformed the seven amplitude attributes into seven principle components. This process may extract low-, moderate-, and high-frequency information effectively. The first two principle components recovered most of the information in seismic attributes, and were used to fit measured reservoir thickness at well sites. The workflow reduced the number of independent arguments from seven to two, simplifying the fitting process without significant information loss. In other words, the process was optimized with improved prediction^[29,30]. Reservoir thickness between wells was predicted after the two components were input into the resulted fitting function (Fig. 12). The main trend of predicted distribution of reservoir (Fig. 12a) is similar to the result of seismic-geomorphology mapping (Figs. 6 and 9) with some local differences. In 23 wells used in the fitting, the correlation coefficient between the predicted values and actual values is 0.78 (Fig. 12b). The predicted results for other 14 blind-testing wells were all fallen within the general trend, indicating the reliability of the method. This quantitative method is clearly better than the prediction methods based solely on wireline-log facies interpolation^[22] or simple amplitude attributes^[10,11].

Seismic diagenetic facies is essentially within the scope of seismic lithology, i.e., to detect diagenesis-related variation of rock property by using seismic information^[31]. Dissolution- enhanced porosity and permeability (see the touching vugs in Fig. 4c) is partially responsible for high gas production of the Longwangmiao reservoir in Gaoshiti-Moxi area. The dissolution effects identified on thin sections include penecontemporaneous freshwater dissolution, burial dissolution, and hyper- gene dissolution^[8,9]. In most locations in the study area, reservoir rocks are mainly distributed within 80 m underneath the unconformity near the base of Permian. Most reservoir rocks are generally above the water table. Therefore, the vugs should be formed by near-surface karstification (or layered karst^[32]). In the study area, the karstification was in its early stage, far from forming large-scale collapsed caverns. The reflection events of carbonate rocks in Longwangmiao Formation are relatively continuous (Fig. 5). No vertically distinguishable seismic- event dislocations caused by the collapsing have been observed.

After many tests, similarity variance (SV), a geometric seismic attribute, was found to be effective in reflecting the lateral similarity of rocks. Residual similarity between the adjacent traces is calculated to disclose the lateral continuity of seismic events or rock property. The smaller the similarity is, the larger the SV value is, and the better the reservoir is. In the SV stratal slice near the top of Longwangmiao Formation (Fig. 13), the abnormally high SV values at MX 19 well are related to dissolution zone revealed by core, FMI log, and porosity log. Core data show that the vugs are mainly in Upper Longwangmiao Formation (Fig. 4a). According to the interpretation of FMI and porosity logs, zones with vugs are also common in Upper Longwangmiao Formation (Fig. 3). In the SV stratal slices at middle and lower Longwangmiao Formation, SV values decrease.

Despite of certain likenesses between SV map (Fig. 13) and distributions of facies (Fig. 10) and reservoir (Fig. 8), there are still three important differences.

(1) There are three (i.e., low, moderate, and high) levels of abnormalities. Low abnormality is of very small SV values, reflecting little alterations to the rocks by dissolution. Located mostly in paleolows, seismic events are typically continuous. Moderate abnormality is shown by mid-range SV values. Indicating a moderate alteration to the rocks, circular features on the map are similar to collapsed caverns in karstified formation, where the reservoir connectivity is poor. Moderate abnormality is common in the transition zone between paleohighs and paleolows. High abnormality is indicated by the largest SV values, where rocks were repeatedly altered, resulting in better reservoir connectivity. It occurs typically on paleohighs. This phenomenon confirms that locations with thick reservoir (Fig. 8) are indeed paleohighs. In the shallow-water, high energy environment, shoal prevailed. During contemporaneous to penecontemporaneous period, exposed sediments on paleohighs underwent freshwater dissolution to form reservoir with matrix pores. During later supergene karst dissolution period (late Caledonian to early Hercynian), the matrix pores in the shoal sub-facies were strongly dissolved and enlarged to form karst reservoir.

(2) Zones (1-2 km wide) with high SV values are common along faults, which indicate the strength of groundwater activity and rock alteration along the faults (blue arrows in Fig. 13). This phenomenon reveals that fault activity and the related fracture system may have contributed to the forming of large-scale reservoirs.

(3) From updip to downdip along the paleo-structure, the alteration of rocks expresses a trend from strong to weak. Top of the paleo-structure (present main body of the gas field) is dominated by high SV values. To the downdip direction, alternative zones of low and moderate SV abnormalities are apparent along strike (Fig. 13). This pattern may reflect the control of water table on the karst activities in slope area. The frequent changes of water table may have related to several periods of sea-level fluctuations during the exposure of the paleohighs after the deposition of Longwangmiao Formation. Should the boundary between low and moderate SV values be taken as the approximate locations of shoreline, four periods of shoreline (or zones of karst reservoir development) would be identified in the downdip direction in the study area.

The most important discovery is that gas production correlates well with SV distribution. Within the range of the gas field, wells with high gas production are mostly located in areas of high SV values, while wells with low gas production are almost all located in areas with low SV values (red circles in Fig. 13). SV abnormalities along faults and downdip areas of paleo- structure are potential targets to find remaining hydrocarbon reserves. In addition, relationship between the trends of SV abnormalities and the flowing directions of fluid in reservoir may affect gas development, which deserves further study.

In seismic sedimentology, the complementary relationship between seismic geomorphology and seismic lithology is emphasized. Seismic geomorphology is mode-driven, which can help identify geomorphology facies and their distribution. They are qualitatively correlated with sedimentary system and processes. In comparison, seismic lithology is data-driven, which can quantitatively determine lithology, i.e., calculate clay content in siliciclastics or identify grainstone, wackestone, etc. in carbonates. For hydrocarbon exploration in carbonates, reservoirs with high porosity and permeability, such as grainstone-dominated reservoirs in shoal sub-facies of Longwangmiao Formation, are the major target. By integrating seismic geomorphology and seismic lithology, the sedimentary environment and the quality of the reservoirs may be simultaneously defined with lithology-geomorphology facies. In this study, maps of paleogeomorphology and sedimentary environment derived from different methods (Figs. 6, 8, and 10) correlate well with the distributions of the reservoirs (Fig. 12a) and dissolution features (Fig. 13). The methodology discussed here can effectively reduce ambiguities caused by single data source and over-simplified method, which offers new reliable criteria for decision making in hydrocarbon exploration and development.

There are some limitations, however. In the complex carbonate reservoir system in Longwangmiao Formation, many factors were not considered in this study. They include possible interferences of the upper and lower rock formations, effect of fractures, etc. Accuracy of reservoir prediction can be further improved. Furthermore, restricted by the complex geologic setting and limited data, influences of multiple episode tectonic movements and differential compaction were not considered for the restoration of paleogeomorphology. Meanwhile, the discussions about the origin and distribution of diagenetic facies are still at the initial stage.

This study was the first application of seismic sedimentology to Paleozoic, marine carbonate strata in Sichuan Basin. A workflow was established for qualitatively restoring sedimentary facies and quantitatively predicting reservoir rocks. Lithofacies in Longwangmiao Formation ranges from grainstone to mudstone, forming such sub-facies as shoal, intershoal, lagoon, and deep shelf. High porosity (>4%) reservoirs are mainly distributed in the shoal sub-facies.

Following the principle of seismic geomorphology, paleogeomorphology of the study area was restored by integrating cast method and isopach method. A syndepositional en echelon fault system controlled the distribution of paleohighs (shoal) and paleolows (intershoal, lagoon, and deep shelf). The carbonate platform sedimentary facies in Longwangmiao time were restored.

Following the principle of seismic lithology, reservoir distribution was quantitatively predicted by PCA of multi-frequency amplitude panels. Thick reservoirs are mainly distributed in shoal sediment formed on paleohighs. Similarity variance seismic attribute can reflect internal lateral variation of carbonate rocks, revealing the distribution of dissolution diagenetic facies in carbonates with assistance of stratal slices.

Seismic geomorphology and seismic lithology are complementary, which is supported in this study by the good correlation between various maps derived by different methods (e.g., paleogeomorphology, sedimentary environment, reservoir thickness, and karst distribution). The ambiguity caused by single data source and oversimplified methodology is effectively reduced. The integrated workflow is both mode- driven (qualitatively) and data-driven (quantitatively), which provides a new tool for decision-making in hydrocarbon exploration and development.