In the typical outcrop profiles of the Jiantianba area, a variety of lithologies may be identified, including:

• Algal-bound reef limestones primarily composed of 65% blue-green algae (Figures 4A,B);

Different stratigraphic positions are dominated by different species, with blue-green algae dominating at the bottom of the section and sponges at the top. The structure of the skeletal reef limestone is mainly composed of string sponges and fiber sponges reef as reef-building organisms (Figure 4F); in the lateral extension of the bottom of the Jiantianba section, a large amount of skeletal reef limestone was found. Limestone breccia, which was formed due to the collapse of the reef (Figure 4G); dark grey, thickly bedded dolomite, with visible cracks and dissolution vugs (Figures 4H,I) and caves that exceeded 10 m were also observed.

Different types of lithologies corresponded to different sedimentary microfacies in the study area. Bioclastic reef limestone, shoal facies and sand debris constituted the reef-based microfacies, while sponge skeletal and algal-bound rocks comprised the block reef core. Dolomite was interbedded with echinoderm limestone, micritic dolostone and stratified dolomite, in which there were abundant epigenetic dissolution pores in a layered distribution. Some of these pores have been filled by asphaltene and calcite. The reef top microfacies, which undulates and protrudes upward, is a favourable reservoir.

Several outcrop sections were surveyed and analysed in this study. The A–A’ section of Jiantianba was in the eastern part of the study area (Figure 1B). The section is ∼300-m-long from north to south, 400-m-long from east to west and has a strike of N70°E. The exposed part of the profile corresponds to the reef core and reef base microfacies and to the PSS2 and PSS1 quasi-sequence groups in the stratigraphic sequence (Figure 5A). At the bottom of the section, there is an ∼75-m-thick, dark grey, thin–medium bedded micritic limestone, in which single layers are ∼0.15-m-thick and ∼0.1 m at a minimum. The micritic limestone transitions upward into bioclast–micritic limestone that is ∼150-m-thick, with layers of ∼0.2 m and at least 0.15 m in thickness. The bioclast content gradually increases from the bottom to the top and there are local occurrences of grey, medium-bedded bioclast–micritic limestone. There are only a few types of organisms represented, mainly brachiopods, echinoderms and algae, none of which are abundant. The chrysanthemum reef is visible at the bottom, indicating a deep-water low-energy environment at the time of its deposition. In the middle of the section, the lithology transition from a micritic limestone to an algal-bound reef limestone that is ∼43-m-thick; the mean thickness of individual layers is ∼2–3 m.

On the algal-bound limestone, a set of sponge reef limestones is ∼35-m-thick and exhibits poor layering. A set of sponge skeletal bafflestones also occur at the top of the section, in which the main reef-building organisms are string sponges and fiber sponge reef. These rocks contain abundant algae and sponge but extremely poor layering, exhibiting a massive structure, with a small amount of visible asphaltene (Figure 5B). The profile worsens in layering from the bottom to the top and the fine, horizontal strata are mostly concentrated in the bottom of the outcrop area, with good continuity and characteristics typical of reef base sediments (Figure 5C).

The B–B’ section in Jiantianba (Figure 1B) is located in the southeast of a typical outcrop area, Yupi Village, with a strike of S60°E. The vertical sequence mainly corresponds to the PSS1 quasi-sequence group, on which the reef base microfacies in the sedimentary facies belt crops out, and the bioherm core microfacies is visible at the top. Except for the top, which is covered by vegetation, the rock mass is fully exposed (Figure 6A). At the bottom of the section is an ∼40-m-thick micritic limestone, which appears in evenly distributed layers of ∼0.25 m and at least 0.15 m in thickness. This limestone has localised silicate nodules and is lithological stable overall and large in its lateral extent. Above this layer is am ∼200-m-thick bioclast–micritic limestone. This rock comes contains layers of ∼0.3 m in thickness (minimum: ∼0.2 m) and has local silicate nodules. At the top of the section is a small amount of algal-bound limestone (∼25-m-thick), with poor layering, but a blocky structure and clustered distribution (Figure 6B). From the bottom to the top of the profile, micritic limestone, bioclast–micritic limestone and algal-bound limestones appear sequentially. The lithological transition between the bioclast–micritic limestone and algal-bound limestone corresponds to the boundary between the quasi-sequence groups PSS1 and PSS2. Having been subjected to later geological reworking, the top of the reef limestone varies in thickness at different locations (Figure 6C).

The C–C’ section in Jiantianba (Figure 1B) is 500 m southeast of the typical outcrop area in Village. The section exhibits a “V” shape in top view, stretching from the middle to both sides, with the rock walls on two sides striking N60°W and N55°E, respectively. The section has a middle bulge and is small in scale. The reef base, bioherm and reef core microfacies in the reef facies are completely exposed and the vertical sequence corresponds to three quasi-sequence groups (PS S3, PSS2, and PSS1) (Figure 7A). At the bottom of the section is am ∼15-m-thick micritic limestone distributed in even layers of ∼0.2 m and at minimum 0.1 m in thickness. Above the micritic limestone is a layer of bioclast–micritic limestone that is ∼25-m-thick, with layers of ∼0.3 m and a minimum of 0.2 m in thickness. Overlying the bioclast–micritic limestone lies a set of algal-bound reef limestones that are generally ∼15-m-thick and poorly layered at the top (Figure 7B). From the bottom to the top of the section, micritic limestone, bioclast–micritic limestone and algal-bound limestone occur in succession. Mudrocks are laterally consistent in content and stable. The lithologic transition between the bioclast–micritic limestone and algal-bound limestone corresponds to the boundary between the PSS1 and PSS2 quasi-sequence groups. Having been affected by later geological reworking, the top reef limestone is large in scale in the middle of the section and smaller in on the two flanks. The reef varies somewhat in thickness at different locations (Figure 7C). Based on the field survey and comparison of the sections, the quantitative parameters of the stratigraphic architecture of the aforementioned five sections were analysed (Table 1).

The UAV, combined with OPT, allowed for the rectangular spatial coordinates of any point on the image to be accurately obtained. In this survey, an eight-rotor UAV (Tiantu, China) was flown in parallel according to the planned cruising route. The aerial folding range was ∼40%. During flight, a Share-200 (Motoar Sky, China) oblique camera carried by the UAV continuously acquired image information from five angles according to the set exposure mode (Figure 2A, Figure 8). The UAV was also equipped with a positioning and orientation system (POS) that accurately recorded the real-time rectangular spatial coordinates of each point during flight. The POS and image information were continuously accepted by the ground sensor and imported into the ground terminal operating platform. After aerial photography, the ground terminal platform was used to guide the vehicle to return safely and data collection in the study area was completed.

The data were integrated using Context Capture v. 2.0 (Bentley System, Inc., United States) to construct a 3-D model of the Jiantianba outcrop area with coordinates (Figure 1B, Figure 5), which can reflect the real geological conditions in the field very well. The coordinates of any point can be obtained directly and the spatial distributions of the geological bodies in the model were well expressed. The scanning model built can satisfy 360° observation and can zoom in and display local parts of an image. The maximum resolution of a local image was ∼10–20 cm. By changing camera or adjusting the flying height, the resolution can be further improved (Figure 8). The model had a large range, covering many remote and dangerous areas that cannot be safely reached by humans; this kind of data acquisition is much more efficient than manual inspection. The data acquired also provides a solid geological foundation for quantitative modelling.

Virtual well data. Traditional outcrop investigations cannot connect the information from continuous points, limiting the interpretability of the research to some extent, especially when the geographical coordinates of the research block cannot be continuously obtained. The UAV OPT accurately matches the image and geographic coordinate information of any point in the study area. In this study, we selected the x-coordinates, y-coordinates and elevation, z (Table 2), of 17 wells in the outcrop field, which laid the foundation for 3-D geological modelling.

Lithological data. After determining the coordinates of well positions, by analysing the collected rock samples, combined with the aforementioned field analysis and lithological calibration results, the lithologies of different depth sections of atypical outcrop of Jiantianba was calibrated. In consideration of the subsequent modelling, lithotypes, including crystal–fine-grained dolomite, bioclast–dolomitic limestone, sponge skeletal reef limestone, algal-bound sponge reef limestone, algal-bound reef limestone, bioclast–micritic limestone and mudstone, were identified and represented by corresponding codes. The lithotypes and corresponding codes for different well depths were determined and taken as input data during modelling (Table 3).

Well top data. According to the field research, there are typical reef base, bioherm core, reef core and reef top sedimentary microfacies from the bottom to the top of the Jiantianba outcrop, with obvious lithological interfaces. Therefore, in the subsequent 3-D geological model, the section was divided into five single layers (PSS0, PSS1, PSS2, PSS3 and PSS4) based on the top bedding plane. The stratigraphic divisions for the five layers in the corresponding wells were obtained from UAV elevation data to lay a foundation for 3-D reservoir modelling.

After inputting the converted geological information (i.e., wells, seven lithologies and well top data) into Petrel V.2009 (Schlumberger, United States), the 3-D gridding model (Figure 9A) and stratigraphic model of the study area were built using Pillar gridding, horizon and layering modules (Figure 9B). The model was composed of 637,880 grids sized 1 m × 1 m × 0.25 m.

Based on the field research and UAV OPT data, including the thickness, spatial relationships and extensional direction of various lithofacies, and using lithological, well and well top data as constraints (Table 4), geological modelling was carried out using the “Data Analysis” and “Facies Modeling” modules in Petrel V.2009 (Schlumberger, United States). An SIS algorithm based on a variogram function was used to run 50 random simulations of the lithology of the Jiantianba outcrop area. The geological models obtained were compared and the reasonable model was finally selected (Figure 9C). This model was in line with the known geological understanding of the area. Based on the SIS model of the lithofacies, a map of the planar distribution of lithofacies was obtained using the “properties” function in Petrel (Figure 9D). Analysing the facies distribution, the sedimentary units generally show a lithological transition from micritic limestone to reef limestone to limy dolomite from the bottom to the top.

The contour map showing the lithological distribution from the SIS was combined with the geological outcrop profile analyses as the prototype model for the 3-D training image and MPS was carried out. The geometric shape of the training image in the study area was obtained from the lithological contour map. A darker colour (purple) and larger mean numerical code indicated that the lithology of the region was closer to that of the reef base. Conversely, a lighter (orange) colour and smaller numerical code indicated that the lithology of the area was closer to that of the reef top, if at the top of the reef, a higher degree of dolomite development was observed.

The lithofacies distribution shows that the sedimentary units evolved from a micritic limestone to a reef limestone to a limy dolomite from the bottom to the top. The dolomitic limestone in the reef top transitioned from a continuous distribution to an isolated, point-like distribution. In contrast, the reef limestone exhibited good overall continuity and a local point-like distribution (Figures 10A,B). Observing the reef core microfacies reveals that the reef-forming facies mainly occur in isolated points, and the algal-bound reef limestone is more abundant in content in the southwest (Figure 10C). Observation of the bioherm microfacies further reveals that the reef-forming facies were in the northwest region. Algal-bound and bioclast–micritic limestones occurred in layers, with high content locally (Figure 10D). By observing the reef base microfacies, it was found that the algal-bound reef limestone and bioclast–micritic limestone were continuous. The reef core microfacies, which are mainly algal-bound and bioclast–micritic limestones exhibit isolated, point-like distributions. The main reef base bodies (yellow) are in the northeast and southwest of the area (Figures 10E,F).

In order to improve the accuracy and quantitative capacity of the training images, the four single layers controlled by the five lithological interfaces were processed in detail. According to the vertical and planar facies sequences, the functions of the “Facies” module in Petrel were used to re-adjust the spatial distributions and quantitative relationships of various lithologies, and to obtain 3-D models of different reef shoal microfacies, including reef base, bioherm, reef core and reef top (Figures 11A–D). The results of the first depositional SIS were edited in Petrel. It can be seen from the 3-D grid diagram that the spatial contacts between the facies belts maintained the results of the original deposit and the mesh size of the training image matched that in the subsequent MPS, ensuring the accuracy and precision of the 3-D spatial search. The 3-D training image model reflects the variations of various lithologies in the vertical and horizontal directions quite well.

Multi-point geostatistical simulations were based on the Petrel software platform, using the fairly mature single normal equation simulation (SNESIM) algorithm. Based on the well lithofacies data, MPSs under the constraint of lithofacies was conducted to simulate the distribution of six lithologies in different layers of the reef base, bioherm, reef core and reef top microfacies. The SIS method was used to simulate the 3-D sedimentary microfacies of the reef base (Figure 12B), which was composed of argillaceous limestone, bioclast–micritic limestone and a small amount of algal-bound limestone. The results of the primary SIS showed that the micritic limestone and bioclastic limestone were unevenly distributed, while also existing as continuous pieces, with no geometric regularity. After MPS under the constraint of the training image (Figure 12A), it was found that the micritic limestone and bioclast–micritic limestone alternated. That is, the bioclasts exhibited multi-period distributions that roughly paralleled the direction of the coastline. The geological connotation of the training image is well expressed and the lithological relationships are more in line with the understanding of the reef base microfacies (Figures 12C,D).

The bioherm was mainly composed of bioclast–micritic limestone, algal-bound limestone and a small amount of sponge skeletal rock with an algal-bound structure. The results of primary SIS revealed that the algal-bound limestone and sponge skeletal rock were in a continuous piece with no obvious geometric distribution (Figure 13B). In the secondary MPS, under the constraint of the training image (Figure 13A), the planar trend and vertical variation in lithologies were strongly regulated. The results of MPS showed that the bioclastic limestone and algal-bound limestone were interbedded, largely in layers, but also as discrete points. The stratigraphic sequence was also more in line with our understand of bioherm core microfacies (Figures 13C, D).

The main lithologies of the reef core microfacies included a sponge reef limestone with an algal-bound structure, sponge skeletal bafflestone, bioclastic limestone and a small amount of fine-grained dolomite. The results of primary SIS showed that the contacts between the different lithologies were irregular, and the fine-grained dolomite at the reef top was continuously distributed (Figure 14B). After the secondary MPS, under the constraint of the training image (Figure 14A), it was found the bioclast–dolomitic limestone and the sponge skeletal reef limestone were heterogeneous in distribution. Overall, the algal-bound limestone was scattered and evenly distributed, while the sponge skeletal limestone appeared in layers in striking northeast–southwest, which is more consistent with the geospatial distribution of a linear reef. Locally, the content of the reef increased, which is presumed to represent single-point reefs on the seaward side. The lithologic relationships are also more in line with the understanding of reef core microfacies (Figures 14C, D).

The main lithologies of the reef top microfacies included a sponge reef limestone with an algal-bound structure, sponge skeletal bafflestone, bioclast–dolomitic limestone and a small amount of fine-grained dolomite. The SIS results showed that the fine-grained dolomite was concentrated in the eastern part of the model and was continuously distributed, while the bioclast–dolomitic limestone was concentrated in the west (Figure 15B). The planar distribution differs from the lithologic distribution of the reef–shoal facies considerably. The MPS, under the constraints of the training image, showed that the fine-grained dolomite and bioclastic limy dolomite were well regulated (Figure 15A). The amount of fine-grained dolomite was greatly reduced and shifted to the northwest. The sponge skeletal reef limestone appeared in layers approximately parallel to the coastline. The stratigraphic relationships observed are more consistent with the understanding of the reef top microfacies (Figures 15C,D).

Through comparing the morphologies and lithological distributions of different microfacies in the models, we found that in the MPS under the constraint of lithofacies, the training image controlled the spatial distribution of the lithofacies quite well. Compared with the SIS, the lithologies of the MPS displayed better regularity in both vertical and lateral distributions, which is more in line with the current geological understanding of the reef–shoal facies in the Jiantianba outcrop area. Combining field research and modelling, the parameters of the SIS and MPS were statistically compared. It was found that the geological knowledge effectively constrained the data for each lithofacies in the MPS. The difference between the results of the MPS parameters and the statistical results from the geological knowledge based on field surveys was small (Table 5), with the difference in lithology reaching a maximum of 15.2% and a minimum of 0%, which indicates that the model is generally reliable.