The formation of reservoir architecture is the result of the comprehensive effects of structure, sedimentation, and diagenetic transformation. Since the concept of reservoir architecture was proposed^[1,2,3,4], foreign and domestic researchers have made a lot of research on the architecture patterns and architecture characterization methods of clastic rock reservoirs^[5,6,7,8], which has played a positive role in guiding development and production. Due to the complex diagenesis of carbonate rocks, slow progress has been made on research of carbonate reservoir architecture. Some studies on architecture of porous reef flat facies reservoirs have been carried out^[9], but the architecture of reservoirs suffering strong reformation primarily controlled by diagenesis has been less investigated.

Fault-controlled karsts^[10] are a special type of carbonate fracture and cave reservoir formed by fault-controlled karstification. With large burial depth, complex spatial structure, and strong heterogeneity, fault-controlled karsts are the main targets for increasing reserves and production in the Tahe oilfield in recent years. The present researches on fault-controlled karst primarily focus on pattern discussion, spatial delineation, and connectivity analysis^{[11,12,13,14,15,16,17,18,19,20,21,22,23,24,25]}, but few researches have covered the architecture classification of fault-controlled karsts. In addition, the architecture characterization of fault-controlled karst is not systematic and hierarchical. The relationships between the dominant factors affecting of fault-controlled karst need to be examined more closely to form a new set of characterization idea and method.

A typical fault-controlled karst reservoir is developed in the Tuoputai area of the Tahe oilfield. Through comprehensive analysis of multiple data sets, the fault-controlled karst architecture classification plan has been initially established based on architecture concept. According to “hierarchical constraint” approach, the well-seismic combination method is used to explore effective underground fault-controlled karst architecture characterization technique, in the hope to provide practical technology and methods for the efficient development of such reservoirs.

The Tahe oilfield is located in the southern slope area of the Akekule bulge of the Shaya uplift in the northern Tarim Basin. The Ordovician fracture and cave reservoir there is the most special large carbonate reservoir discovered in China so far. The Tahe oilfield has experienced multistage tectonic movements in the Middle-Late Caledonian, Early Hercynian, Late Hercynian, and Indosinian-Yanshanian, giving rise to a series of superimposed fault systems of multiple levels and multi-periods^[26,27,28]. The Tuofutai area is located in the southwest wing of the Tahe oilfield, where the main target layers are the Middle-Lower Ordovician Yijianfang Formation and the Yingshan Formation at a burial depth of 5500-6000 m. There are a series of northeast and northwest trending deep large fault zones in this area, of which, TP12CX is the largest northeast strike-slip fault in the region. The fault-controlled karst reservoir is extremely developed along this fault strike, and the fault-controlled karst element is representative (Fig. 1).

The development of the TP12CX fault zone started earlier, with a number of wells drilled. According to the segmentation and development dynamics of the fault zone, the fault zone is divided into multiple relatively independent fault-controlled karst elements. The adjacent elements TP101 and T739 with abundant dynamic information and more wells drilled were investigated in this work. There are about 40 wells in the two elements. The wells are generally 500-600 m apart, some wells less than 400 m apart. Since most of the wells had drilling break and drilling fluid leakage when encountering caves, it was difficult to obtain mud logging data. Therefore, it is difficult to tell the scale of the fault-controlled karst architecture element solely by drilling data. The seismic data of the study area has good quality. The seismic data of the target layers has a frequency bandwidth of 5-70 Hz and center frequency of nearly 30 Hz, and thus can roughly distinguish strata 30 m thick and can predict larger-scale architecture elements quite well, but can’t recognize small-scale architecture elements.

Hierarchy classification is the basis of architecture research, and small hierarchical architecture elements are controlled by large hierarchical architecture elements. Tectonic movement and dissolution are the dominant factors affecting the formation of fault-controlled karst reservoirs^[29]. Classification of architecture hierarchy should consider the process of reservoir dissolution, and the scale of different architecture elements and their mutual constraints, namely "formation constraint and scale control.”

"Formation constraint" emphasizes geological constraints on architecture classification, including the constraint of strike-slip fault zone to fault-controlled karst, the constraint of fault-controlled karst to internal fracture and cave combination, and the constraint of large caves to internal filling characteristics. "Scale control" emphasizes the controlling relationship between the various scales in the architecture hierarchy. The scale of the fault controls the scale of the fault-controlled karst, the scale of fault-controlled karst controls the scale of the internal fracture and cave zone, and the large cave controls the internal filling characteristics.

Based on the above-mentioned classification principle, the fault-controlled karst in Tuofutai area is divided into four hierarchies: strike-slip fault impact zone, fault-controlled karst, fracture and cave zone inside fault-controlled karst, and filling inside cavity. The architecture elements at various levels have constraint and control relationships between each other.

2.2.1. Strike-slip fault impact zone

The strike-slip fault impact zone is the first level of the fault-controlled karst architecture. It is a large-scale fracture envelope with a certain range of influence from the bottom to the top of the fracture formed by torsional or shear stress during formation of strike-slip fault. Its nature is essentially the comprehensive impacting range of strike-slip fault. The outer boundary of the fracture impact zone is the architecture interface of this hierarchy, and the internal comprehensive impact zone is the architecture element of this hierarchy.

Affected by difference in stress, the strike-slip fault impact zone shows a certain segmentation along the strike-slip direction on the fault plane^[30,31], including the extrusion section (pressure torsional stress), tensile section (tensile torsional stress), and shear section (shear stress). These sections show different characteristics and usually occur alternately. From the profile, under the action of torsional stress, as the stress gradually releases from the root to the top, the fracture shows a "Y-shape" or "flower-shape,” commonly known as "flower-like structure,” which can be further subdivided into "half-flower- like," "positive flower-like," and "negative flower-like" structures; the stress fracture zone on the profile is "narrow at the bottom and wide upward". Under the action of shear stress, the fracture doesn’t change much in width from the root to the top, showing "columnar" or "single-branch like" feature. From the scale, the strike- slip fault impact zones are large in vertical extension height, even cut through to the top of the Precambrian, ranging up to several kilometers. The extension width of the profile is hundreds to thousands of meters from the root to the top, depending on the location of the segmentation. The extension scale of the plane strike is commonly greater than tens of kilometers, and the segmentation characteristics appear alternately.

2.2.2. Fault-controlled karst

The fault-controlled karst is the second hierarchy of fault-controlled karst architecture. It refers to the range in the strike-slip fault impact zone with higher degree of fracturing, higher fracture density, and relatively concentrated dissolution intensity. The internal stress differences in the strike-slip fault zone lead to differences in rock fragmentation degree, and the degree of fracture development is closely related to the main fault. Therefore, the morphology of fault-controlled karst is affected by the segmentation of the strike-slip fault and the combination of fault patterns. Previous studies have summarized the shapes of fault-controlled karst traps from the perspective of fault patterns and karstification differences, including strip, flat, and sandwich shapes. The strip- shaped fault-controlled karst is primarily formed in the torsional stress section of the dominant fault, features strong dissolution and caves in dominance, "V" or "funnel" shape on profile, and is wide strip-like along strike on the fault plane. This type is also the typical representative of the fault-controlled karst architecture anatomy in this article. The flat fault-controlled karst mostly occurs in the shear stress section of the dominant fracture, features stronger dissolution and dissolution pores in dominance, and is "line" or "column” shape on profile and narrow strip on the plane. The sandwich shape fault-controlled karst is controlled by secondary associated faults parallel or in echelon on the plane, has dissolution only concentrated along the fracture surface in a narrow range, and appears as a cluster of narrow strips along the fault strike.

With external geometry constrained by the strike-slip fault impact zone on the whole, fault-controlled karsts are smaller than strike-slip fault impact zones in longitudinal extension and profile width. They are tens of meters to kilometers wide, and hundreds of meters to kilometers in extension, depending on spatial partition.

2.2.3. Fracture and cave zone inside fault-controlled karst

The fracture and cave zone inside fault-controlled karst is the third hierarchy of fault-controlled karst architecture. According to diagenetic architecture, it refers to the element further subdivided according to the dissolution strength under the constraints of the external geometry of fault-controlled karst. Its architecture elements include dissolution caves, dissolution pores, and dense fracture zones.

Dissolution caves have the strongest dissolution and are the most favorable third-hierarchy architecture element for oil and gas charging. The architecture interface is the boundary of large cave with large difference in scale. The caves are from decimeters to ten meters, on average greater than 5 m in diameter, and irregular in shape. The dissolution pores have weaker dissolution than caves, are also favorable oil and gas storage space, smaller than larger caves, from centimeters to decimeters in diameter, and extremely irregular in shape. Since dissolution pores are commonly formed by the expansion and dissolution of small and medium scale fractures, larger dissolution pores are generally distributed near fault zones and large caves, while smaller dissolution pores are similar to those in conventional pore-type carbonate reservoirs. In terms of scale, it is very difficult to describe the architecture of individual dissolution pores. Therefore, the dissolution pore zone is taken as the architecture element in this paper. The dissolution pore zone represents a zone with denser dissolution pores of different sizes and high probability of dissolution pores. The boundary of the dissolution pore zone is the architecture boundary, and the diameter of the dissolution pore zone ranges from several meters to tens of meters. The fracture dense zone has no or very weak dissolution. The fractures only serve as pathways for deep fluid or meteoric water, and there is no significant interactive dissolution. The fractures basically maintain the state formed initially by fault activity, and most of the fractures are distributed outside the dissolution pores or may intertwine with dissolution pores. Although limited in storage space, the fracture dense zone can act as important channels for oil and gas migration. It is also difficult to characterize the architecture of individual small fractures. In this paper, the fracture dense zone represents the architecture element of this hierarchy. The fracture dense zone is composed of small fractures of various groups and scales. The meaning of the architecture element is that fracture has higher probability of development in this zone. The boundary of the fracture dense zone is the architecture boundary, which ranges from several meters to tens of meters in size.

From the perspective of the combination of the three kinds of architecture elements in the fault-controlled karst, large caves, dissolution pore zones, and fracture-dense zones present a roughly “tripartite” structure, and there is a certain combination law between the elements. Usually, there are more dissolution pore zones around large caves, and fracture-dense areas are also primarily distributed around pore zones. From geologic connotation, the three kinds of elements have some boundaries in genesis; while from the geophysical meaning, the boundaries may have some transition zones. Therefore, the three elements need to be considered together in architecture characterization.

2.2.4. Cave filling

Cave filling is the fourth hierarchy of fault-controlled karst architecture. From the perspective of the formation process of fault-controlled karst, large caves, dissolution pore zones, or fracture zones all have fillings of different lithologic rocks or fluids. The differences in filling degree and filling lithology directly affect the quality of fracture- cave reservoirs. In consideration of its importance, the cave filling is classified as a separate type of architecture hierarchy. Moreover, as large caves are the most important type of fractured and cave reservoirs, study on their filling conditions is of greater significance and relatively easier to operate due to their large scale, the filling of large caves is taken as the typical representative of architecture element of this hierarchy in this work.

The filling condition inside large caves is divided into unfilled, partially filled, and fully filled based on filling degree. According to lithology, the filling is divided into clastic rock sedimentation filling, carbonate rock cementation filling, collapsed breccia filling, and mixed filling of multiple lithologies. Unfilled caves refer to caves with no collapse or sedimentation after formation which may be filled by fluid in later reservoir forming process. As the best type of reservoir space, unfilled caves can be identified by serious loss of drilling fluid, unloaded drilling tool, and expanded well diameter during the drilling process. For fully-filled or partially-filled caves, the filling may be composed of one lithology or mixed rocks of several lithologies. When the filling is composed of clastic sediment, the typical feature is sedimentary bedding, which may be formed by gradual sedimentation of surface sedimentary rock collapsing into the cave or of deposits carried by the underground river flowing through the cave. The physical properties of the filling are related to the specific properties of the sedimentary rock. If the filling is mainly sandstone, it is a better reservoir. Ce-mented carbonate rock filling is generally the result of re-diagenesis after the cave is formed, which may be further reformed by deep hydrothermal fluids or formation fluids. The carbonate filling is usually calcite, which is tighter and poorer in reservoir quality. The breccia filling is complex in composition, including surface clastic rock breccia and carbonate surrounding rock collapse, which is significantly affected by later tectonic activities and differs widely in physical properties. If argillaceous or calcareous cementation occurs between the breccias, the filling would be poor in physical properties are poor. The mixed filling is the combination of all the above lithologies, and is more complicated in reservoir properties. Cave filling element also involves the issue of filling ratio, including the volume ratio of the filled part to the entire cave, and the ratio of various lithological components in the filled part, which both have important impacts on subsequent reservoir property evaluation.

A lot of research on the filling of fracture and cave reservoirs in the Tahe Oilfield has been done before, especially for underground river-type reservoirs, some filling lithology sequence characteristics have been summarized^[32]. However, main factors affecting fault- controlled karst reservoirs and underground river-type reservoirs have some differences. No matter from the analysis of formation process, or outcrop and drilling data, the filling combination of the fault-controlled karst large caves is relatively simple, primarily dominated by pure calcite or breccia with a small amount of sand and mud, but has hardly erosion bedding like that in underground river-type. In this study, the filling combination architecture elements primarily include different lithological bodies in the cave. The architecture interfaces are the contact surfaces between different filling lithologies. The filling architecture elements are restricted by the cave in shape and scale, and vary significantly in morphological characteristics. The elements are commonly at decimeter-meter level, and very difficult to characterize in three-dimension.

There is a certain composition relationship between architecture elements of different levels. The strike-slip fault impact zone contains fault-controlled karsts of different geometric forms, and a fault-controlled karst contains multiple caves, dissolution pore zones, and fracture zones of various scales, which is called the fracture and cave zone. The large caves have different filling structures inside. Different architecture elements have different scales and geological meanings and require different data and technologies to identify (Table 1). The strike-slip fault impact zone, large in scale, is primarily identified and characterized by seismic data. The key technology is the automatic interpretation technology of fault based on seismic coherent body, in which the faults are reasonably combined through manual intervention. The cave filling element, small in scale, is primarily characterized by using core and logging interpretation (imaging) under the guidance of outcrop patterns and is predicted tentatively by using well and seismic data. The key technology includes core calibrated logging interpretation and reference to seismic lithology inversion prediction.

Based on the fault-controlled karst architecture classification plan as well as comprehensive interpretation of single wells and well-seismic calibration, taking the typical fault-controlled karst of the TP12CX fault in Tuofutai area as an example, the applicability of the fault-controlled karst architecture characterization technology is illustrated.

3.2.1. Strike-slip fault impact zone characterization technology

The formation basis of strike-slip fault impact zone is the regional large strike-slip fault. With tectonic movements as the main affecting factor, strike-slip fault impact zones are large in transverse extension, cut multiple layers longitudinally, and extend tens of kilometers on the plane, their characterization includes two aspects, characterization on plane and on profile.

The post-stack 3D seismic data was smoothed, coherent data bodies that can reflect the distribution characteristics of faults were extracted, slices along the T74 interface at different locations were extracted, and some wells were used to calibrate the region to analyze the plane distribution characteristics of the strike-slip fault impact zone. The strike-slip fault impact zone is in northeast-southwest strike, and is 15 km long and 1000 to 3000 m wide in the target zone (Fig. 2a). Affected by different stresses in different sections, the strike-slip fault impact zone varies in width and fracture development intensity.

The strike-slip fault can be divided into pull-apart section, extrusion section, and translation section. The geometry of the fault impact zone and the structure of the internal fracture zone are both controlled by strike-slip stress. The seismic reflection shows that the strike-slip fault impact zone has clear "flower-like" structures in pull-apart section or the extrusion section in the longitudinal direction, and large width overall due to overlapping. Fracture distribution relationship in the impact zone is reflected well by the coherent body. The impact zone cut down to the base and extends up above the T₇⁴ interface (Fig. 2b). The translation section is in "chimney" shape longitudinally with little change in width, but also has fractures inside. This section is similar in longitudinal extension (generally 2000-3000 m) to the extrusion and pull- apart sections (Fig. 2c).

3.2.2. Fault-controlled karst characterization technology

The development of fault-controlled karst is controlled by the strike-slip fault impact zone, and the architecture characterization is primarily conducted from the profile and three-dimension. The delineation of external geometry of fault-controlled karst has been extensively studied from the perspective of geophysical attributes and inversion^[33,34]. The most widely used attribute at present is structure tensor, which has achieved fairly good effect in the application of Tahe oilfield. But due to the much greater depth of this strike-slip fault than the current target layer (T₇ ⁴-T₇⁶) in the Tahe oilfield, it is difficult to distinguish the deep sections with weaker dissolution from the undisturbed sedimentary strata, and thus difficult to predict the fault-controlled karst in this strike-slip fault impact zone. In view of this, the fault likelihood (FL) attribute more sensitive to structure information was used to describe the external geometry of the fault-controlled karsts.

The FL attribute is a fault imaging algorithm based on sample point processing^[35], which is more sensitive to seismic anomaly information and has a prominent effect on the local differences caused by fractures. FL is more suitable for the extraction and carving of large anomalies. A similarity attribute is used to express the possibility of fracture presence. A value ranging between 0 and 1 is used to indicate the possibility of fracture presence, which reflects the size and range of the possibility of fault-controlled karsts to a certain extent. Another core problem in predicting fault-controlled karst's external morphology is to determine the seismic attribute threshold. Horizontal well or vertical well drilling time curves or logs can be used as calibration data. When there is no well data available for calibration, automatic threshold segmentation technology was used to automatically recognize the external contour range^[36]. The principle of this technology is similar to the currently popular image identification. The threshold segmentation algorithm can automatically calculate the segmentation threshold between different data types (i.e., different seismic attribute values) in a certain data set and can guarantee the maximum statistical variance between the different types of data to achieve a statistically significant classification. From the implementation process, based on threshold segmentation, attribute extraction and carving, and then through appropriate manual modification, the fault-controlled karst characterization can be completed.

The results of three-dimensional characterization show the fault-controlled karst is a stripe in north-east strike restricted by the strike-slip fault impact zone (Fig. 3a). It has some difference in spatial continuity, which is closely related to the segmentation of the strike-slip fault. The fault-controlled karst is more developed in places with good spatial continuity (TP186 well area in Fig. 3a), large in lateral width and in wide stripe shape. Structurally, it is primarily located in the pull-apart section of the strike-slip fault. The fault-controlled karst is poorly developed in areas with poor spatial continuity (TK1058 well area in Fig. 3a). In these sections, the fault-controlled karst is smaller in lateral width, in narrow stripe shape, and is primarily located in the shear stress section structurally. The profile shows that the hierarchical constraint relationship is relatively clear, showing a "narrow down and wide up" form. The external geometric morphology of the fault-controlled karst is 800-2600 m wide, slightly smaller than the strike-slip fault impact zone (Fig. 3b).

3.2.3. Fault-controlled karst internal fracture and cave zone characterization technology

The fracture and cave zone inside the fault-controlled karst is an important element of fault-controlled karst reservoir development and primarily includes large cave, dissolution pore zone, and fracture-dense zone. The characterization of fracture and cavity zone in the fault- controlled karst is critical for reserve estimation and recovery enhancement. The architecture elements of fracture and cave zone belong to small-scale elements, most of which are lower than the seismic resolution, which can be detected by seismic survey but are difficult to accurately distinguish. The fracture and cave zone can be identified from drilling, but due to rapid variations of fault-controlled karst, it is difficult to figure out the fracture and cave zone by conventional well spacing. In view of this, in this study, a characterization method of fracture and cavity structure inside fault-controlled karst based on well-seismic pattern fitting, that is, under the guidance of the fault-controlled karst architecture pattern, dense well network areas were chosen or horizontal wells were made use to carry out well control-seismic control join research, fully fit the existing architecture pattern with well-seismic data, and complete the architecture characterization of the reservoir element.

For the prediction of fracture and cave zone in fault-controlled karst with seismic data, previous researchers primarily used wave impedance data nested with the structure tensor, but the seismic wave impedance is still not completely free from the influence of the layered strata, so this method has certain error in describing the internal structure of fault-controlled karst. In this study, the seismic texture attribute was used to define the fracture and cave zone. Texture is a seismic attribute that considers differences in waveform^[37]. It primarily combines and strengthens the information of similar spatial waveform structures through waveform clustering, and is more sensitive to reflections of abnormal geological bodies. The essential difference between the internal structural elements of fault-controlled karsts is dissolution strength, which can be reflected well by the Texture attribute. The texture attribute value indicates the possibility of cave development in the form of probability (ranging from 0 to 1). To a certain extent, this method can eliminate the interference of the original layered strata. Through frequency-increasing processing of seismic post-stack data, the texture attribute was extracted under the constraints of fault-controlled karst’s external geometry, and the attribute value was further calibrated through single-well interpretation, and the architecture characterization of fracture and cave zone in fault-controlled karst was gradually completed from the plane, profile, and in three dimensions.

3.2.3.1. Architecture plane anatomy

Taking the single-well interpretation result of the architecture elements as basic data and the fault-controlled karst development pattern as the guide, the single well data was used to calibrate seismic attributes. Also, well-seismic combination and pattern fitting were used to characterize the planar distribution of fault-controlled karsts. The slices of Texture attribute along the layer were extracted (Fig. 4a), and high texture attribute values in the figure indicate higher degree of dissolution and more developed fractured and cave reservoir. The threshold values of different internal structural elements were calibrated using single well data to obtain the distribution of large caves, dissolved pores and fracture-dense zones. As individual dissolution pores or fractures are small, and seismic data is limited in resolution, the characterization results can only show in the form of “zones”. The corresponding plane distribution of the fault-controlled karst internal structure slice along bedding shows that bounded by the fault-controlled karst’s external contour boundary, the large-scale caves take a smaller proportion than the dissolution pore zone, and the three kinds of architecture elements have gradually changing contact relationships (Fig. 4b). The most important types of reservoirs are large caves and dissolution pores. The fracture-dense zones were primarily characterized by seismic ant body technology.

3.2.3.2. Architecture profile anatomy

With logging interpretation as basic data and planar architecture anatomy results as guide, the profile architecture anatomy was completed in a plane-profile interactive way. A cross-well profile perpendicular to the strike of the fault zone with seismic texture of larger longitudinal extension (cutting through to Precambrian) was selected. The strength of the texture attribute represents the dissolution intensity. The fault-controlled karst on the profile has clear contour along the fault zone from the bottom to the top, especially it can be distinguished well from the surrounding layered strata (Fig. 5a). In the seismic texture reflection profile of the Yijianfang Formation to the upper part of Yingshan Formation (T₇-T₇⁶), two wells encountered fault-controlled karst reservoirs. The logging curves of them can distinguish different internal structures, and the internal structures revealed by logging correspond to texture intensity changes well (Fig. 5b). The corresponding anatomy results of fracture and cave zone in the fault-controlled karst show that (Fig. 5c, 5d) the external geometric shape restricts the internal fracture and cave structure in both overall form and local details. The closer to the top of the strike-slip fault zone, the higher the proportion of caves, indicating that the effect of fresh water leaching and dissolution became stronger near the surface (T₇⁴ is unconformity). The current development and drilling of Tahe oilfield only reveal the fault-controlled karsts developed in the upper part of the strike-slip fault zone (Fig. 5d). There is no drilling information to confirm whether there are deep fault-controlled karsts or how the development degree is, but the geological basis for the development of fault-controlled karsts exists.

Although the well-seismic combination was fully utilized in the architecture anatomy, the spatial shapes of large caves and dissolution pores change rapidly, and it is difficult to completely use pattern fitting to deduce their shapes. Moreover, the burial depth of the research target exceeds 6000 m, and the resolution of current seismic data can only identify distribution of large caves, while the distribution of dissolution pores can only be predicted in complex zone.

3.2.3.3. Three-dimensional architecture anatomy

The three-dimensional architecture anatomical results show that the distribution of fractures and caves in the fault-controlled karst is consistent with the strike-slip fault segmentation and fault-controlled karst external geometry (Fig. 6a). The longitudinal structure of fault- controlled karst on the through well profile extracted from the three-dimensional characterization results shows significant changes and clear signs of the control by the strike-slip fault, and the dissolution near the T₇⁴ interface is stronger obviously (Fig. 6b). From the results of seismic characterization, the large caves are 20-200 m wide and 35-550 m high, the dissolution pore zones are 150-450 m wide and 50-650 m high. But affected by seismic reflection information, the fracture and cave reservoirs characterized by seismic data are commonly larger than the actual fracture and cave reservoirs. This problem of “slimming” fracture and cave is a current hot spot of study for geophysical professionals. Therefore, errors should be appropriately considered in the current fracture and cave reservoir scale statistics.

3.2.4. Cave internal filling characterization technology

The cave filling element, with small scale and shape hard to identify, is difficult to characterize. A lot of research on the filling of epigenetic karst caves in the main area of Tahe oilfield has been done before ^{[38,39,40,41,42,43]}. These researches covered logging response characteristics and logging interpretation plan of the karst caves by natural gamma ray, well caliper, deep and shallow lateral resistivity, acoustic time difference, neutron density, and imaging logging data; and statistics on the proportion of filling materials and the distribution characteristics of the filling physical properties. The above research results have important guidance for research on the filling in large caves of fault-controlled karst.

Outcrops are the most intuitive data for analyzing cave filling and play an important guiding role in underground cave filling research. From the present field outcrop survey, filled large caves in the fault-controlled karst are rare. Most caves are unfilled (Fig. 7a), or if filled, the filling is mostly calcite or asphalt in simple filling pattern. In order to illustrate the complexity of cave filling in more detail, a Liuhuanggou outcrop was examined (Fig. 7b). Two large filled caves can be seen in this outcrop, and the upper cave is smaller and filled by breccia completely. The lower cave is larger and filled by collapsed breccia and cave sediments. Well TH10421 is a typical well drilled into a cave in the study area. In this well, the Yijianfang Formation and Yingshan Formation are primarily composed of tight limestone; and the cave fillings include breccia and clastic rock. The clastic rock is characterized by low density, low resistivity, and low natural gamma. The breccia filling, affected by cementation, features lower values and tooth shape on density and resistivity log curves (Fig. 7c). Generally speaking, the caves containing these two filling types are the primary reservoirs in this area. The cave filling is controlled by the cave size and 0.2-50.0 m high from preliminary statistics.

Individual wells can only show the filling situation of caves in one dimension. In this study, we attempted to find out the three-dimensional distribution of cave fillings by wave impedance inversion under logging constraint. According to the analysis of cave filling elements, sandstone filling usually is the best in physical properties (that is, higher porosity), breccia filling is the second, and carbonate cement filling is the worst. By predicting the porosity distribution and combining with logging interpretation results, the cave filling pattern can be roughly figured out. There is a strong correlation between wave impedance and reservoir porosity in general. The wave impedance inversion was carried out under the constraint of the fracture and cave zone in the fault-controlled karst (Fig. 8b). The high wave impedance area is the area with low porosity. The filling structure inside the cave was further characterized by logging interpretation (Fig. 8c). The characterization results preliminarily reflect the complexity of the cave filling pattern. The filling elements are complicated in contact relationship, extremely irregular in shape, and equivalent to the cave in width. Since the seismic resolution can’t reach the meter level and below, the spatial characterization of the filling pattern isn’t fine enough, and further geophysical research is needed.

Considering differences in the formation process and scale of fault-controlled karsts, according to the hierarchy interpretation plan of architecture, the fault-controlled karst is divided into four hierarchical architecture elements, strike-slip fault impact zone, fault-controlled karst, fracture and cave zone in fault-controlled karst, and cave filling. Characterization technologies for different architecture elements have been selected according to their features, a fault-controlled karst architecture characterization flow based on well-seismic fitting and stepwise anatomy has been set up. The strike-slip fault impact zone, largest in scale, is primarily characterized by seismic coherence combined with manual interpretation. The fault-controlled karsts, larger in scale, are calibrated by combining well-seismic data under strike-slip fault impact zone constraint, and characterized based on FL attribute. The fracture and cave zones in fault-controlled karst are medium in scale. Under fault-controlled karst constraint, the seismic texture attribute is calibrated using multiple wells, and the fracture and cave zone is characterized on plane and profile. The cave filling, small in scale, is characterized by well-seismic pattern fitting under the guidance of the outcrop pattern preliminarily. The semi-quantitative analysis of fault-controlled karst architecture elements of different scales has been realized preliminarily, which has a positive effect on geological modeling of fault-controlled karst.