Advanced geospatial monitoring of oil and gas infrastructure via satellite data

2.1. Study area

The study area was defined as the Kashkadarya region of Uzbekistan. Kashkadarya is one of the hydrocarbon-oriented regions of Uzbekistan and is therefore an appropriate case for evaluating an integrated framework based on satellite geodata, GIS analysis, three-dimensional geological reconstruction, and computational modeling for oil and gas infrastructure tasks.

From the geological point of view, the study area is considered as a structurally heterogeneous subsurface system composed of interconnected aquifer and hydrocarbon-bearing layers partly isolated by low-permeability barriers, such as clay- and silt-rich strata. This type of configuration is suitable for analyzing restricted fluid migration, local pressure accumulation, and the influence of semi-permeable barriers on subsurface flow behavior. In the context of digital oil and gas planning, such a regional case provides a meaningful basis for combining surface geospatial observations with a simplified but physically interpretable subsurface representation [2], [10], [12], [13].

In the present article, the Kashkadarya case is treated as a regional demonstration area used to evaluate the coherence of the proposed integrated workflow rather than as a fully field-calibrated reservoir simulation case.

For clarity, Fig. 1 shows the regional position of the Kashkadarya region within Uzbekistan and the extent of the analyzed local subarea used in the present study. In this work, Kashkadarya is treated as the broader regional case for evaluating the proposed integrated workflow, whereas the Shurtan gas field is used as the local example for the satellite-based spatial interpretation presented in the Results section.

Fig. 1Location of the study area in Uzbekistan and extent of the analyzed local subarea in the Kashkadarya region: regional location of Kashkadarya within Uzbekistan (left); local demonstration area used for the integrated geospatial analysis (right). Base map prepared by the authors using administrative boundary data and Sentinel-2 imagery, spatial resolution 10 m

2.2. Data sources

The methodological framework was based on the integration of open-access satellite, terrain, and geological information. The manuscript uses MODIS, Landsat, and Sentinel-2 data as the main satellite sources for the structural and surface-oriented representation of oil and gas fields. Landsat-8 OLI and Sentinel-2 were treated as the principal multispectral sources for geological interpretation due to their wide adoption in environmental and geoscientific applications, while MODIS was included as a large-area, high-temporal-coverage source [5-7], [16], [17].

In addition to satellite imagery, digital elevation models (DEMs) were used to describe terrain morphology and support the interpretation of structural features and possible fluid migration controls. The geospatial inputs were complemented by field lithological information, which was incorporated to reconstruct the architecture of aquifer and hydrocarbon-bearing layers. The manuscript also states that the applied data resources included the NASA Earthdata portal and the ESA Copernicus Open Access Hub, which ensured access to current and reliable Earth observation datasets.

The integrated dataset therefore included the following components: multispectral imagery from MODIS, Landsat, and Sentinel-2; DEM-derived terrain information; interpreted geological and lithological information; and geospatial layers used for GIS-based integration and 3D visualization [2], [13], [16], [17], [19]. The main characteristics of the satellite datasets used in the study are summarized in Table 1.

2.3. Geospatial preprocessing and GIS integration

All spatial datasets were harmonized within a GIS-oriented environment. The manuscript explicitly refers to ArcGIS, ERDAS Imagine, ENVI, and Google Earth Engine as the main software environments used for interpretation and integration of satellite observations. In the revised methodological formulation, GIS is considered not as a simple overlay tool, but as a data-fusion environment that enables joint analysis of topographic, lithological, environmental, and infrastructure-related layers [2], [10], [11].

At this stage, satellite imagery was interpreted together with DEM-derived terrain attributes and geological information to improve structural understanding of the study area. Multispectral products, including NDVI-related and thermal-band information, were used to detect spatial patterns relevant to geological interpretation, while terrain descriptors were used to support the identification of surface morphology and structurally controlled zones. Radar interferometry is also mentioned in the manuscript as an auxiliary method for spatial interpretation, especially in the context of deformation-oriented geospatial analysis [18], [19].

The purpose of this stage was to prepare an internally consistent geospatial base for subsequent subsurface reconstruction. Through GIS-based integration, surface observations and geological assumptions could be linked within a single analytical space, making it possible to proceed from satellite interpretation to physically meaningful 3D modeling [2], [8], [10], [18].

2.4. Three-dimensional geological reconstruction

The next stage of the workflow consisted of constructing a three-dimensional representation of the subsurface. In the manuscript, the formulated model uses DEMs, multispectral satellite imagery, including NDVI and thermal bands, and field lithological information to recreate the structure of aquifer and hydrocarbon-related layers. The modeled system is described as a set of linked aquifers partly isolated by low-permeability barriers, such as clay or silt strata, which is consistent with the intended analysis of constrained flow and pressure redistribution [10], [12], [13].

Each geological stratum was represented digitally within a structured model space. In this representation, the subsurface is discretized into units characterized by properties such as permeability, pressure, and porosity. The explicit inclusion of heterogeneity is essential because fluid movement in hydrocarbon-bearing systems is strongly affected by spatial contrasts in permeability and by the presence of semi-permeable barriers. This representation was used to improve the realism of subsurface interpretation and to support subsequent simulation of filtration behavior [10-12].

Three-dimensional visualization was performed using Python-based tools, specifically PyVista and Matplotlib, as stated in the manuscript. Such visualization improves the interpretability of geological architecture and allows geoscientists and engineers to inspect reservoir geometry, structural continuity, and potentially hydrocarbon-relevant zones in an intuitive and reproducible form [11], [21].

2.5. Mathematical model of subsurface flow

The mathematical core of the framework is based on Darcy’s law extended into three-dimensional space. The model is used to simulate groundwater and hydrocarbon filtration in a heterogeneous system of interconnected permeable and semi-permeable layers. In the revised formulation, the flow behavior is described through the relationship:

where $q$ is the volumetric flow vector, $k$ is permeability, $\mu$ is dynamic viscosity, and $\nabla P$ is the pressure gradient. This formulation is consistent with the manuscript’s description of a structured-grid representation in which geological strata are assigned permeability, pressure, and porosity values that vary spatially throughout the modeled formation [10], [13].

To ensure the conservation of mass within the simulated domain, Darcy’s law is coupled with the continuity equation for steady-state incompressible flow, formulated as:

where $\rho$ is the fluid density. The differential equations were solved numerically using the Finite Difference Method (FDM) on a regular 3D Cartesian grid. To accurately represent the physical system, specific boundary conditions were applied: Dirichlet boundary conditions (constant pressure) were assigned to the outer boundaries of the primary aquifer to simulate regional flow drivers, while Neumann boundary conditions (zero flux, $\partial P/\partial n=$ 0) were enforced across the identified low-permeability barriers to represent no-flow boundaries. The domain was discretized with a spatial resolution of 50×50 meters in the horizontal plane $\left(X,Y\right)$ and variable vertical steps ($Z$) aligned with the interpreted stratigraphic horizons.

The practical purpose of the model is to evaluate how low-permeability barriers modify fluid migration paths, reduce flow velocity, and generate local pressure accumulation. The explicit inclusion of such barriers improves the interpretation of heterogeneous underground systems compared with simplified field-based conceptual models that do not adequately represent structural compartmentalization. As a result, the model supports the identification of zones where hydrocarbon movement may be restricted, redirected, or locally intensified due to lithological contrasts [10], [12], [13].

2.6. AI-assisted image interpretation

Artificial intelligence and machine learning methods were incorporated as supportive tools for image interpretation. The manuscript explicitly mentions convolutional neural networks (CNNs) and Random Forest as the principal models used for object classification in satellite imagery, and also refers to supervised and unsupervised classification logic [14], [15].

In the revised methodological presentation, AI is positioned as an assisting component rather than as an isolated end in itself. The role of machine learning is to support the recognition of land-use patterns, geological properties, rock-type variations, and potential semi-permeable strata visible or indirectly reflected in satellite imagery. CNN-based interpretation is especially relevant for pattern recognition in multispectral data, whereas Random Forest and related classifiers can support spatial categorization tasks when labeled or semi-labeled image sets are available [14], [15].

At the same time, the final geological interpretation is not based solely on automated classification. Instead, AI-derived outputs are integrated with GIS layers, DEM-based interpretation, and geological reasoning. This makes the methodological structure more robust and avoids overstating the independent predictive role of machine learning in the current study [3], [4], [14], [15]. For the supervised classification tasks, the Random Forest (RF) algorithm was employed to map surface anomalies and vegetation stress indices, utilizing multispectral bands (Visible, NIR, SWIR) and DEM derivatives as input features. The RF model was trained using 100 decision trees (Nestimators = 100) with a maximum tree depth set to prevent overfitting. The dataset was split into training (70 %) and testing (30 %) subsets based on visually validated regions of interest (ROIs). Model performance was evaluated using standard quality metrics, achieving an overall classification accuracy of 86 % and an F1-score of 0.84 for target geological classes. For complex spatial feature extraction, a pre-trained Convolutional Neural Network (CNN) architecture was adapted. The multispectral imagery was preprocessed into 64×64 pixel patches to capture local textural contexts before feeding them into the network.

2.7. Validation strategy

To confirm the practical usefulness of the proposed framework, simulation testing was performed on a dataset representing an oil-bearing region. In the present study, the test case is associated with the Kashkadarya region. The validation logic is qualitative but geologically oriented: the performance of the modeling system was compared against known geological patterns and structural expectations for a hydrocarbon-bearing setting.

According to the manuscript, the validation demonstrated four practically relevant outcomes: a clear visual representation of aquifers and semi-impermeable barrier layers; identification of zones with abnormal pressure accumulation caused by flow restriction; support for more informed preliminary drilling-site screening; and identification of locations that may require additional geological or engineering attention due to possible upward gas migration. These results do not yet constitute a full quantitative benchmark, but they do provide a coherent proof-of-concept for the proposed integrated framework [10], [12], [13].

3.1. Three-dimensional structural interpretation

The first main result of the study is the generation of a three-dimensional visual representation of the reservoir-related subsurface architecture using satellite-derived geodata and geological interpretation. The resulting digital model allows the user to inspect the geometry of subsurface layers, the spatial arrangement of stratigraphic units, and the continuity or discontinuity of structurally relevant zones. This is important because oil and gas exploration decisions depend not only on general geological maps, but also on the ability to understand layer relationships in three dimensions [10-12].

The visual output improves the interpretability of the study area by linking surface-derived geospatial information with subsurface structure. In practical terms, the model supports the identification of prospective zones, structurally constrained pathways, and locations where additional subsurface attention may be required before planning drilling or infrastructure expansion. In this way, 3D structural visualization serves as more than an illustrative output; it becomes a decision-support instrument for exploration-oriented analysis [11], [13], [21].

Fig. 2 presents the generated 3D visualization of the modeled oil field, illustrating how satellite-derived geospatial inputs were transformed into a structured digital representation suitable for exploration-oriented interpretation.

Fig. 2Three-dimensional visualization of the modeled oil-field structure reconstructed from integrated satellite-derived geospatial and geological inputs. The subsurface model was discretized on a structured grid with a horizontal cell size of 50×50 m and variable vertical steps aligned with the interpreted stratigraphic horizons. Visualization prepared by the authors using PyVista and Matplotlib

3.2. Representation of heterogeneous layers and barrier-controlled flow

A second important result is the successful representation of heterogeneous geological layers, including low-permeability barriers that affect filtration dynamics. The model does not treat the reservoir as a uniform porous body. Instead, it represents a linked system of aquifer and hydrocarbon-bearing layers separated in part by clay- and silt-rich units. This heterogeneity is central to the analytical value of the model, because it controls how fluid pathways develop and where pressure buildup may occur [10], [12], [13].

Fig. 3Structural representation of the reservoir surface and distribution of permeable and low-permeability zones used in the flow analysis

The simulations show that semi-permeable barriers slow fluid migration, alter local flow direction, and promote pressure accumulation in upstream zones. From the standpoint of oil and gas infrastructure tasks, this result is important because such zones may affect drilling risk, reservoir accessibility, and the probability of abnormal migration behavior. Thus, the model supports a more realistic interpretation of reservoir behavior than simplified homogeneous conceptualizations [10], [12], [13].

Fig. 3 shows the 3D structural representation of the reservoir surface and the modeled arrangement of permeable and semi-permeable layers used for subsequent flow analysis.

3.3. Darcy-based simulation results and practical interpretation

The Darcy-based simulation framework allowed the study to evaluate fluid flow under heterogeneous underground conditions. The main analytical output is not merely visual; it is the interpretation of how pressure gradients, permeability contrasts, and barrier geometry interact to produce constrained or redirected migration pathways. This constitutes the central technical contribution of the study.

The simulation results indicate that the inclusion of low-permeability zones enhances the realism of the predicted flow field. In particular, the model identifies zones of restricted flow and abnormal pressure accumulation that are not easily recognized in simplified representations. These outputs are relevant for practical oil and gas decision-making because they can support preliminary well-placement screening, improve interpretation of structurally constrained zones, and highlight areas that may require additional validation before field decisions are made.

A quantitative comparison with a simplified homogeneous baseline model highlights the impact of this methodological integration. The introduction of semi-permeable barriers ($k<$ 0.1 mD) into the simulation reduced the calculated fluid migration velocity by approximately 65 % in the barrier-adjacent zones compared to the homogeneous baseline. Furthermore, the localized pressure buildup in constrained areas reached an excess of 15-20 % relative to the unconstrained flow scenario. Consequently, the adjustment of well coordinates by 349 m and 1238 m (as illustrated in Fig. 4) was mathematically justified by shifting the target zones out of the identified high-pressure gradient anomalies, thereby reducing potential drilling risks.

These findings should be interpreted cautiously. The model supports preliminary drilling-site screening and helps reduce interpretive uncertainty, but it is not intended to demonstrate fully quantified economic optimization in the absence of field-calibrated validation.

Fig. 4 illustrates the satellite-based spatial interpretation of adjusted well locations in the Shurtan gas field within the integrated geospatial analysis framework.

Fig. 4Satellite-based spatial interpretation of adjusted well locations in the Shurtan gas field within the broader Kashkadarya study area based on integrated geospatial analysis

3.4. Discussion of methodological value and limitations

The obtained results demonstrate the broader value of integrating satellite geodata, GIS, 3D modeling, and AI-assisted interpretation into a single workflow. The main strength of the framework is methodological integration: surface geospatial observations are not treated in isolation, but are transformed into a digital subsurface representation that can be explored visually and interpreted through Darcy-based filtration logic.

Another important strength is the use of open and widely accessible data sources and software environments. The reliance on MODIS, Landsat, Sentinel-2, GIS platforms, and Python-based visualization tools increases the practical reproducibility and affordability of the approach. For regions where high-cost proprietary subsurface datasets may be limited, such an integrated workflow can still provide meaningful exploratory support [16-19].

At the same time, the present study has limitations. First, the validation remains predominantly qualitative and is based on consistency with known geology rather than on a quantitative benchmark with formal error metrics. Second, the model simplifies real reservoir behavior by representing complex geological systems through a structured digital framework. Third, the AI component requires further validation on broader, multi-regional datasets to act as an independent predictive engine. These limitations should be acknowledged explicitly, because doing so will strengthen the scientific credibility of the paper rather than weaken it [10], [12], [13].

Finally, the results suggest that future work should focus on improving spatial resolution, integrating additional geological constraints, and developing more automated feature-recognition procedures for satellite imagery. This would allow the framework to evolve from a proof-of-concept modeling system toward a more quantitatively validated digital tool for oil and gas infrastructure analysis.

Therefore, the present results should be interpreted as a methodologically integrated and geologically plausible proof-of-concept rather than as a fully quantified operational forecasting system.