Challenges and Strategies in Modeling Thin-Bedded Carbonate Reservoirs Based on Horizontal Well Data: A Case Study of Oilfield A in the Middle East

Abstract

Thin-bedded carbonate reservoirs face significant challenges in characterization and development due to their thin formation thickness, strong interlayer heterogeneity, and rapid sedimentary transformation. In recent years, horizontal wells have played an increasingly important role in improving the productivity of thin-bedded carbonate reservoirs. However, building accurate geological models from horizontal well data is a major challenge for geoscientists. Using Middle East Oilfield A as a case study, this paper analyzes the specific challenges of horizontal well geomodeling and proposes a dedicated strategy for integrating horizontal well-derived constraints into the geological modeling workflow. To address the challenges of structural modeling constrained by horizontal well data, this study proposes three methodologies: stratigraphic layer iteration, virtual control point generation, and localized grid refinement. These techniques collectively enable the construction of a higher-fidelity structural framework that rigorously honors hard well data constraints while incorporating geological plausibility. To address the challenges posed by the spatial configuration of vertical and horizontal wells and the dominant trajectory patterns of horizontal wells, this study introduces two complementary approaches: the exclusion of horizontal well section data (relying solely on vertical wells) and the selective extraction of representative horizontal well section data for variogram derivation. These methods collectively enable the construction of a geologically realistic reservoir model that accurately captures the spatial distribution of reservoir properties. These methodologies not only effectively leverage the rich geological information from horizontal wells but also mitigate spatial clustering effects inherent to such data. Validation through development well production data confirms robust performance, providing transferable insights for reservoir characterization in analogous fields worldwide.

1. Introduction

According to the “BP Statistical Review of World Energy 2020”, Middle Eastern oil production reached 1.417 billion barrels in 2019, representing 31.61% of global output and consolidating its position as the world’s leading oil-producing region. Approximately 80% of Middle Eastern oil production originates from carbonate reservoirs. The region hosts the world’s largest concentration of carbonate-hosted hydrocarbon accumulations, where thin-bedded carbonate reservoirs—typically under 50 m thick with layered edge-water drive systems—constitute nearly 70% of these formations. The inherent complexity of these carbonate reservoirs, arising from intricate structural configurations, heterogeneous depositional systems, and multiscale pore networks, manifests as pronounced heterogeneity in reservoir properties and fluid flow behavior [1]. The predominant pore systems comprise intergranular pores and chalky microporosity, with subsidiary occurrences of vugs, intercrystalline pores, and isolated macroporosity, collectively exhibiting a heterogeneous and spatially disordered arrangement. Thin-bedded carbonate reservoirs in the Middle East are typically characterized by limited gross thickness, rapid vertical lithofacies transitions, pervasive high-permeability streaks, and pronounced heterogeneity—collectively resulting in complex reservoir architecture [2]. Prominent exemplars of thin-bedded, low-permeability carbonate reservoirs include the Kh2 reservoir in Oman’s Al Hed oil field, the laminated Upper Shuaiba Formation reservoirs in Oman, and the Thamama B (THB) zones within Abu Dhabi’s Bab and Asab fields. In the Middle East, thin-bedded carbonate reservoirs represent the predominant reservoir type, accounting for approximately 20% of the region’s carbonate reserves. However, despite decades of development, these reservoirs continue to exhibit persistent production declines, suboptimal utilization, and low recovery rates [3,4]. Consequently, conventional water-drive mechanisms have underperformed expectations. To address this deficiency, horizontal wells have emerged as the predominant development solution for such reservoirs. This approach necessitates robust geological characterization leveraging horizontal well data to optimize reservoir management strategies.

1.1. Characteristics of Horizontal Well Data

In reservoir development, horizontal wells serve as critical enabling technology for enhanced production efficiency, cost reduction, and improved recovery. Unlike vertical well data, horizontal well measurements exhibit directional sensitivity to in situ stress regimes, yielding azimuthally clustered data with superior spatial resolution along bedding planes—providing essential constraints for high-fidelity reservoir characterization. Given the unique characteristics of horizontal well data, effectively constructing high-fidelity reservoir models remains a critical research priority. However, directional clustering of well trajectories triggers spatial aggregation effects that introduce a significant bias in variogram analysis. This impedes the derivation of geologically meaningful variograms and consequently hinders the development of reservoir architectures consistent with geological principles [5,6,7,8].

Horizontal well data deliver exceptional information density and volumetric coverage, enabling superior interwell prediction and geological feature characterization. These wells provide extensive lateral penetration with directionally constrained trajectories, yielding concentrated, high-resolution geological datasets along preferential azimuths. Horizontal wells provide high-resolution constraints on the connectivity architecture and three-dimensional geometry of sub-seismic geological features (higher-resolution features than seismic waveforms). Compared to vertical wells, horizontal well data significantly enhance the accuracy of interwell predictions and enable more a robust characterization of reservoir heterogeneity elements, including microstructural features and reservoir properties in well sections [9,10].

1.2. Research Status of Horizontal Well Modeling

Structural modeling represents a fundamental phase in 3D geological workflows, where establishing an accurate structural framework underpins all subsequent reservoir characterization. This deterministic methodology integrates the digitization of formation top-structure maps with stratigraphic horizon constraints derived from well control points, culminating in kriging interpolation to generate precise structural surfaces for each stratigraphic unit. Horizontal wells exhibit unique spatial configurations relative to formation geometry. Consequently, compared to vertical or conventional directional wells, horizontal well data provide unparalleled advantages in characterizing lateral reservoir heterogeneity. These data offer critical constraints on reservoir microarchitecture [11,12,13,14], whereas vertical wells intersect the reservoir interface at a single point, providing only one stratigraphic control position per layer. Unlike vertical wells that intersect the reservoir interface at a single point, providing only one stratigraphic control position per layer, horizontal wells achieve high-resolution characterization along entire stratigraphic intervals. However, their complex trajectory layer intersections create intricate spatial relationships that challenge conventional modeling. Software-based stratigraphic correlation often oversimplifies these configurations through algorithmic smoothing, potentially generating layer geometries inconsistent with geological trends.

The variogram serves as a fundamental geostatistical tool for quantifying spatial correlation and heterogeneity of reservoir properties. In well log analysis, variogram range and azimuthal anisotropy characterize sediment transport directions and sand body geometry. During geostatistical reservoir modeling, variograms constrain spatial continuity models for petrophysical properties including porosity, permeability, and fluid saturation. However, the inherent directional bias of horizontal well trajectories combined with preferential targeting of high-permeability zones induces significant spatial sampling artifacts. This clustered data distribution creates statistical distortions in variogram analysis, yielding spatial continuity parameters inconsistent with actual geological architecture during reservoir modeling. The effective utilization of horizontal well data advantages, while preserving inherent information integrity, constitutes the core challenge in geomodeling. However, directional clustering of densely sampled well paths generates statistically biased spatial continuity estimates. This manifests as geologically implausible variogram structures [15,16,17], where azimuthal data accumulation distorts range and anisotropy parameters, ultimately propagating geological misunderstandings through erroneous reservoir models. However, the exclusion of horizontal well data substantially increases interwell prediction uncertainty and degrades geological model accuracy. To resolve this paradox, integrated modeling approaches combining horizontal and vertical well variograms under phased geological constraints and seismic guidance have been proposed. These methodologies remain impractical in areas with limited seismic resolution or insufficient vertical well coverage.

1.3. Horizontal Well Modeling Challenges in Thin Reservoir

1.3.1. Challenge 1: Stratigraphic Thickness Discrepancies

Discrepancies in the interpreted zone thickness between horizontal and adjacent vertical wells propagate through structural modeling workflows. Software extrapolation of conflicting well-point data generates mesh distortion anomalies, including zero-thickness cells (“dead grids”) along well trajectories due to unresolved stratigraphic conflicts.

1.3.2. Challenge 2: Trajectory–Layer Geometric Complexity

Complex spatial relationships between deviated well paths and formation layers create modeling artifacts. Automated smoothing algorithms applied post-intersection determination frequently produce stratigraphic surfaces inconsistent with geological frameworks [18,19,20], resulting in stratigraphic misalignment and compromised structural integrity.

1.3.3. Challenge 3: Data Density–Grid Resolution Mismatch

Sub-grid-scale clustering of horizontal well points (multiple formation tops within single grid cells) exceeds discretization capabilities. This spatial oversampling, particularly pronounced in high-curvature well segments, induces local structural artifacts through grid curvature effects.

1.3.4. Challenge 4: Sampling Bias in Variogram Derivation

Preferential drilling in high-quality reservoir intervals creates artifactually clustered data distributions. Resulting variograms exhibit statistically biased range and anisotropy parameters that misrepresent actual spatial continuity [21,22,23], propagating errors through subsequent property models.

2. Method

Using Middle East Oilfield A as a case study, this research addresses horizontal well structural modeling challenges through three novel methodologies: stratigraphic layer iteration for vertical consistency, virtual control point generation to constrain complex trajectories, and local grid refinement to resolve discretization conflicts. For reservoir property characterization, we propose complementary geostatistical approaches: selective exclusion of horizontal section data to mitigate sampling bias and strategic extraction of horizontal segment data for variogram derivation. This integrated framework leverages the rich geological information from horizontal wells while neutralizing spatial clustering artifacts, enabling accurate representation of reservoir architecture and providing transferable solutions for thin-bedded carbonate reservoirs globally [24,25,26,27,28]. The geological modeling workflow follows a multi-stage refinement process. This integrated approach iteratively updates the subzone model to enhance structural and stratigraphic accuracy (Figure 1).

2.1. Approach to Challenge 1: Stratigraphic Consistency Enforcement

Deterministic subzone modeling integrates seismic-calibrated structural surfaces as trend constraints, with hierarchical thickness statistics defining minimum z-value extrapolation thresholds. This workflow employs sequential horizon generation where each subzone undergoes iterative quality control against crossover violations before integration into the full structural framework, ensuring geologically consistent layer architectures while preventing grid collapse anomalies.

2.2. Approach to Challenge 2: Trajectory–Layer Reconciliation

Interactive virtual control point engineering supplements natural well tops at critical trajectory inflection points [29,30]. These synthetic constraints override automated smoothing algorithms during horizon generation through manual positioning along geologically plausible vectors, forcing layer convergence toward predetermined formation trends and eliminating stratigraphic misalignment artifacts in thin-bedded sequences.

2.3. Approach to Challenge 3: Grid–Data Resolution Optimization

Directional grid restructuring resolves sub-cell data clustering through anisotropic projection algorithms that redistribute formation tops along I/J vector fields (X-Y axis horizontal vector fields). By dynamically calibrating cell dimensions to wellbore influence radii and trajectory azimuths, this approach enforces strict one-to-one data-point/cell correspondence while preserving stratigraphic relationships in high-curvature well segments.

2.4. Approach to Challenge 4: Bias-Mitigated Variogram Derivation

Stratified hybrid variography combines vertical wells with systematically extracted horizontal segments through equal-interval trajectory sampling. Multiple realizations undergo principal/secondary range calculation with arithmetic aggregation of anisotropy parameters [21,22], generating representative spatial continuity models that balance data richness with geological objectivity across diverse well configurations.

3. Geological Setting

Oilfield A is situated in south-central Iraq, approximately 180 km southeast of Baghdad. Located within the passive continental margin of the northeastern Arabian Plate’s Persian Gulf Basin, the field exhibits a NW-SE trending anticlinal structure formed during the Late Cretaceous measuring 29 km × 8 km. This low-relief feature (regional dip < 2°) comprises three culminations: eastern Zone A and western Zone B (Figure 2). Development strategies differ by structural element: Zone A employs hybrid vertical–horizontal well patterns as the vertical well needs to simultaneously develop deeper oil reservoirs, while Zone B utilizes exclusively horizontal well configurations.

The Khasib Formation represents Oilfield A’s principal hydrocarbon-bearing reservoir, containing the field’s largest proven reserves across its four stratigraphic members (Kh1–Kh4), with the Kh2 member serving as the primary accumulation zone and development target. This unit is deposited on a low-angle carbonate ramp system characterized by limited lateral facies variability but pronounced vertical heterogeneity. Core analysis reveals a distinct shallowing-upward succession from distal mid-ramp deposits to proximal grain-dominated shoal complexes. These shoals comprise grainstone (1 mm grain size) with abundant moldic and intragranular dissolution pores and minor dissolution vugs, reflecting increasing hydraulic energy that drove significant vertical variations in sediment texture, allochem composition, and early diagenetic cementation (particularly anhydrite distribution). These depositional dynamics produced strongly heterogeneous pore architectures within high-frequency glacioeustatic sequences (3–5 m thick parasequences bounded by maximum flooding surfaces), resulting from Cenomanian–Turonian sea-level fluctuations that generated stacked shallowing cycles throughout the Kh2 interval [31].

Within the Kh2 carbonate system, high-frequency glacioeustatic fluctuations generated stacked thin parasequences bounded by prominent hardground horizons observable in core. These discontinuity surfaces formed through repeated phases of submarine cementation where aragonite and high-Mg calcite underwent neomorphic stabilization to low-Mg calcite microspar, creating isopachous calcite crusts that lithified the seafloor. This cyclic hardground development produced distinct shallowing-upward sequences exhibiting inverse cyclicity (fining-upward lithologies coarsening upward in sub-wavebase facies). The Kh2 interval displays widespread laterally continuous shoal facies with multistage vertical stacking, evidenced in core by meter-scale cycles capped by submarine-cemented horizons. These inverse cycles are recognized on wireline logs through decreasing gamma-ray trends overlain by abrupt increases at cycle boundaries (Figure 3). High-resolution stratigraphic correlation subdivides Kh2 into five parasequence sets (0–2 m thickness), reflecting short-duration depositional events punctuated by prolonged exposure surfaces.

4. Examples

4.1. Case Study

The Kh2 Formation in Middle East Oilfield A features thin-bedded carbonate reservoirs with minimal average thickness, stratified into two development zones: Zone A employs hybrid vertical–horizontal well patterns, while Zone B contains nearly 400 horizontal wells, presenting significant geomodeling challenges representative of thin carbonate reservoirs. Structural modeling (different subzones are displayed in different colors) reveals substantial stratigraphic thickness discrepancies between horizontal wells and adjacent vertical control points (Figure 4a), particularly between Kh2-1 and Kh2-2 subzones. Conventional interpolation propagates these anomalies, generating progressive gridline pinch-out toward Well 1’s periphery (Figure 4b) and culminating in zero-thickness “dead grids” (Figure 4c). To resolve this, deterministic subzone modeling implements (1) seismically derived top surfaces as stratigraphic trend constraints, (2) z-constrained extrapolation enforcing statistically calibrated minimum thickness thresholds, and (3) iterative quality control preventing layer crossover. The optimized framework (Figure 4d) demonstrates stabilized stratigraphic architecture with validated grid integrity (Figure 4e,f), eliminating non-physical cell geometries while preserving geological plausibility.

Horizontal well structural modeling leverages multiple formation intersections along the borehole trajectory to achieve superior microarchitectural control, yet encounters limitations when localized stratigraphic complexity exceeds the constraining capacity of well points. This insufficiency triggers automated smoothing algorithms during surface generation, causing trajectory-formational intersections to misalign with actual stratigraphic positions. Virtual control point insertion resolves this discrepancy by constraining layer geometry, reconciling wellbore trajectory-formational relationships, and establishing geologically plausible spatial contacts between horizons and well paths—collectively preserving stratigraphic fidelity while honoring trajectory-positioned constraints.

During structural modeling, insufficient stratigraphic control points may generate geologically implausible surfaces, exemplified in Figure 5a where horizontal well trajectories exit Zone A at Points B and A. Automated smoothing algorithms extrapolate layer geometry using limited well tops and formation strike orientations, creating excessive torsional distortion that misaligns the trajectory-formational intersections (actual entry/exit points A0 and A1 shown in Figure 5b). This results in significant well-to-model stratigraphic mismatches and structural framework errors. To mitigate this, engineered virtual control points (pseudo1, pseudo2) constrain the layer position at critical inflection Point A. As demonstrated in Figure 5c (dashed line), these synthetic constraints enforce geologically consistent layer orientations that reconcile borehole trajectories with stratigraphic architecture, yielding a validated structural framework.

Sub-grid clustering of horizontal well formation tops (inter-point spacing < cell dimensions) induces artificial smoothing that obscures reservoir heterogeneity and compromises data fidelity. This artifact is exemplified in Figure 6, where a 50 m × 50 m grid resolution forces multiple tops (e.g., Points 2–3 and 4–5) into single cells. Surface curvature constraints and automated smoothing algorithms then generate non-representative geometries inconsistent with the actual stratigraphy. To preserve high-frequency vertical heterogeneity, trajectory-aligned local grid refinement implements 25 m × 25 m cell dimensions, ensuring the discrete cellular isolation of each formation top. This approach ensures structural fidelity during model coarsening operations while resolving sub-seismic microstructural variations obscured by the original grid resolution. Critically, it provides high-resolution stratigraphic constraints essential for geologically grounded property modeling, as validated through discrete cellular isolation of all formation tops within trajectory-aligned refined grids.

4.2. Methods

To ensure data adequacy while mitigating spatial clustering artifacts in variogram analysis, well-type-specific declustering strategies are implemented to maximize horizontal well information utilization. These approaches adaptively weight or subset horizontal well data based on trajectory density and reservoir architecture, preserving geological representativeness in derived spatial continuity parameters.

4.2.1. For Vertical and Horizontal Well Patterns

To balance data integrity with clustering artifact mitigation in variogram analysis, trajectory-optimized declustering strategies adaptively weight or subset horizontal well data according to local trajectory density and reservoir architecture. This preserves geological fidelity in derived spatial continuity models while maximizing information extraction from high-density well paths (Figure 7).

Comparative analysis of directional data distribution reveals that horizontal well segments with inclinations <40° and <60° exhibit distributional characteristics statistically consistent with vertical well benchmarks. Given the superior sample density within the <60° cohort, this subset was preferentially retained for spatial continuity modeling to maximize data utility while preserving distributional integrity (Figure 8).

4.2.2. For Horizontal Well Patterns

To maximize data utility while minimizing uncertainty, strategic subsampling is implemented. Horizontal well trajectories are discretized into regularly spaced pseudo-vertical control points, achieving three objectives: (1) optimized spatial distribution uniformity, (2) preservation of lateral heterogeneity information, and (3) systematic repetition of sampling intervals. As illustrated in Figure 9, this workflow initiates by defining drainage-aligned sampling lines parallel to dominant well azimuths, spaced at characteristic horizontal well spacing intervals. Intersections between sampling lines and well trajectories generate initial control points, with spatial uniformity enforced through iterative grid-shifting. Each iteration shifts sampling lines to adjacent grid positions until full areal coverage is achieved, ensuring non-redundant point selection across cycles. Composite variograms derived from multiple realizations undergo arithmetic averaging to produce a final spatial continuity model representative of reservoir heterogeneity.

Implementation of this methodology significantly enhances reservoir variogram accuracy, validated through drainage well verification to demonstrate closer alignment with ground-truth formation parameters. By maximizing horizontal well information utility while effectively mitigating spatial sampling artifacts inherent to clustered data applications, the approach substantially reduces reservoir uncertainty. This advancement translates to markedly improved geological model prediction fidelity, establishing robust spatial continuity representations that faithfully capture subsurface heterogeneity.

Oriented parallel to dominant well trajectories, data extraction lines were established at 300 m intervals—corresponding to characteristic horizontal well spacing—to systematically sample reservoir properties. Intersections between extraction lines and well trajectories defined spatially optimized control points, ensuring uniform point distribution across the study area. Given the 50 m × 50 m grid resolution, six iterative sampling passes were executed, progressively shifting extraction lines through adjacent grid positions (Passes 1–6). For each pass, directional variograms were computed along principal (well-parallel) and secondary (orthogonal) anisotropy axes. Point-pair histograms and semi-variance curves were generated, with variogram range determined as the distance where semi-variance reached unity. Final reservoir variogram parameters were established through arithmetic averaging of range values across all six realizations (

Table 1. The variation function calculated after different data picks.

No.	Major Range/m	Minor Range/m	Direction/°
1	2364	1896	−65
2	2467	1946	−65
3	2533	1460	−65
4	2299	1739	−65
5	2653	2018	−65
6	2619	1920	−65
AVE.	2489	1830	−65

).

5. Results and Discussion

Porosity modeling comparisons using vertical well data, raw horizontal well data, and processed horizontal well data reveal distinct characteristics. The vertical well-derived model captures reservoir porosity’s overall distribution but fails to represent heterogeneity. The raw horizontal well model should better reflect heterogeneity while maintaining distribution trends, but exhibits clustering effects. As shown in Figure 10a, horizontal well data simulation produces a porosity field where overall simulated values appear higher (Figure 10b), with data distribution significantly elevated versus vertical wells (Figure 11b).

The extracted horizontal data model (Figure 10c) yields porosity values aligning with vertical well simulations (Figure 11c), while preserving the horizontal wells’ heterogeneity characterization. This approach better reflects actual reservoir property distributions by maintaining heterogeneity representation without clustering artifacts.

6. Conclusions

Horizontal well geological modeling in thin-bedded carbonate reservoirs confronts four critical challenges: (1) stratigraphic thickness discrepancies between horizontal wells and adjacent vertical wells due to differential zonal positioning, (2) complex geometric relationships between well trajectories and layer configurations, (3) sub-grid clustering of horizontal section data points causing structural artifacts, and (4) sampling bias in superior reservoir intervals generating non-representative variograms. These issues collectively compromise model accuracy and geological fidelity.

To address these challenges, this study developed an integrated workflow for Middle East Oilfield A featuring three structural methodologies: stratigraphic layer iteration resolves vertical inconsistencies, engineered virtual control points constrain complex trajectory–layer relationships, and trajectory-aligned grid refinement eliminates sub-grid clustering artifacts. Compared to traditional methods, this tripartite approach establishes more accurate structural frameworks that simultaneously honor well hard data and integrate geological plausibility.

For reservoir property characterization, we implemented two complementary strategies for different well patterns: selective horizontal data exclusion mitigates sampling bias, while directional data extraction derives representative variograms. These methods fully leverage horizontal wells’ rich geological information while neutralizing spatial clustering effects. The resulting models accurately characterize reservoir architecture and property distributions, establishing a globally applicable framework for thin-bedded carbonate reservoirs that balances data utilization with geological realism.