Three-Dimensional Geological Modelling in Earth Science Research: An In-Depth Review and Perspective Analysis

Abstract

This study examines the development trajectory and current trends of three-dimensional (3D) geological modelling. In recent years, due to the rising global energy demand and the increasing frequency of regional geological disasters, significant progress has been made in this field. The purpose of this study is to clarify the potential complexity of 3D geological modelling, identify persistent challenges, and propose potential avenues for improvement. The main objectives include simplifying the modelling process, improving model accuracy, integrating different data sources, and quantitatively evaluating model parameters. This study integrates global research in this field, focusing on the latest breakthroughs and applications in mineral exploration, engineering geology, geological disaster assessment, and military geosciences. For example, unmanned aerial vehicle (UAV) tilt photography technology, multisource data fusion, 3D geological modelling method based on machine learning, etc. By identifying areas for improvement and making recommendations, this work aims to provide valuable insights to guide the future development of geological modelling toward a more comprehensive and accurate “Transparent Earth”. This review underscores the global applications of 3D geological modelling, highlighting its crucial role across various sectors such as mineral exploration, the oil and gas industry, urban planning, geological hazard assessment, and geoscientific research. The review emphasizes the sector-specific importance of this technology in enhancing modelling accuracy and efficiency, optimizing resource management, driving technological innovation, and improving disaster response capabilities. These insights provide a comprehensive understanding of how 3D geological modelling can significantly impact and benefit multiple industries worldwide.

1. Introduction

The evolving global economy and the escalating national energy demand have highlighted the challenges posed by resource scarcity [1,2,3,4]. Consequently, nations are increasingly relying on geospatial data for the effective management of resources and the environment, thereby propelling the progress of geospatial science [5,6,7,8]. Geological modelling harnesses spatial technologies such as GPS, GIS, and RS, underpinned by computing and communication technologies, to realize the “Digital Earth” vision [9,10,11]. Geological modelling integrates multidomain data to construct spatial information models [12,13]. Three-dimensional (3D) geological modelling techniques synergize geology and computing, enabling the extraction of valuable insights from data and facilitating research and resource prediction.

The pinnacle of 3D geological visualization has been researched using GoCAD software, an innovative development from France’s University of Nancy [14,15]. The rapid evolution of 3D geological modelling is facilitated by advances in software and principles [16,17,18], focusing on addressing the topological correlation and attribute-mapping challenges inherent in geology. In the field of regional geological structure research, the primary focus is on establishing unified workflows for constructing multisource models and meticulously delineating complex geological fabrics [19,20,21]. Researchers have employed cutting-edge 3D scanning techniques and unmanned aerial vehicle (UAV) oblique photography to acquire the original geological data [22,23,24,25]. This approach effectively solves the data limitations encountered in complex geological modelling and uses scenarios. Additionally, the integration of the Petrel platform elevates the visualization capabilities of hydrocarbon reservoirs [12,26,27], and 3D digital outcrop modelling also offers invaluable insights to enhance reservoir modelling and development [28,29,30]. These advancements have had significant impacts across various domains.

I. Solid mineral field: In this field, 3D geological modelling technology plays a crucial role in mineral exploration and mining by facilitating mineral body emulation, resource assessment, mining operation planning, and the elucidation of metallogenic sources and migration [31,32,33,34,35,36]. In particular, its applications in mineral body emulation have been extensively acknowledged, which enables the simulation and visualized exploration of mineral deposits. This approach enhances the comprehension of ore body distribution and improves operational efficiency while mitigating extraction risks and reducing costs.

II. Oil and gas field: Notably, 3D geological modelling plays a pivotal role in the oil and gas industry by facilitating fracture prediction, reservoir modelling, reservoir heterogeneity analysis, reserve estimation, and development planning [37,38,39,40]. It enables the precise assessment of reservoir properties and reserves, supporting optimal well placement, spacing, and injection strategies. These capabilities enhance the efficiency and safety of oil and gas exploration and extraction processes.

III. Urban planning field: In addition, 3D geological modelling is vital for urban planning, particularly for geological hazard risk assessments and construction planning [41,42,43]. This technology enables advanced geological hazard risk evaluations and appropriate mitigation measures [44]. Moreover, 3D modelling enhances urban geological comprehension, contributing to sustainable and dependable urban development and construction planning.

IV. Geological hazard field: In the realm of geological hazards, 3D geological modelling primarily provides predictions, early warnings, and responses to geological hazards [45,46,47,48]. It models geological hazards for accurate prediction and assessment, facilitating timely early warning and effective responses.

V. Integration of geological modelling and the time domain: Integrating the temporal dimension into 3D geological modelling can enhance the simulation and understanding of geological processes, elevating both the efficiency and precision of exploration and development [49].

Recent advancements in geological modelling have outstripped the achievements of the late 20th century achievements, with broad uptake, gaining widespread adoption across sectors. This study investigated the evolution of 3D modelling, identified its current limitations, and outlined future directions for propelling the field forward.

2. Technological Development and Application Hotspots

2.1. Development

As is well known, 3D geological modelling involves a comprehensive analysis of geological data through the utilization of spatial information processing, geological interpretation, and geostatistical tools, within a virtual 3D environment. In 1993, Houlding pioneered the concept of 3D geological modelling [50], and subsequent scholars introduced groundbreaking ideas such as “digital mining” [51] and “Glass Earth” [52], which have played a pivotal role in the evolution of geological modelling techniques and the research efforts of geologists. These concepts have become essential tools for diverse applications, including oil and gas exploration, mineral resource assessment, and hazard prediction. Go-CAD software created by J. Mallet introduced discrete smooth interpolation techniques. During the 1990s, global research on this technology further refined the theory and technology of geological modelling software, propelling its rapid development (

Table 1. Globally prevalent 3D geological modelling software.

Product Name	Developer and Collaborating Institutions	Main Application Areas and Features
MapGIS K9 (V.SP2)	Wuhan Zhongdi Digital Group, China University of Geosciences, Wuhan, China	Possesses strong geological modelling capabilities
Longruan GIS (V.3.2)	Beijing SuperMap Software Co., Ltd., Peking University, Beijing, China	Aimed at the field of digital mining, with the highest market share in the domestic coal industry
3D Mine (V.2017.5)	Beijing Sandiman Technology Co., Ltd., Beijing, China	Software system for mining geological, surveying, mining, and technical management work
DeepInsight (V.5.0)	Beijing Grid Technology Co., Ltd., Beijing, China	Targeting the field of geophysics and reservoir characterization, leading in fault and complex structure representation domestically
Direct (V.3.0)	Telon Corporation, Beijing, China	An integrated geological research software platform developed based on terrestrial deposition characteristics and Chinese research ideas. It serves as a geological research platform for different stages of oilfield exploration and development
Visual Geo (V.1.0)	State Key Laboratory of Hydraulic Engineering Simulation and Safety, Tianjin University, Tianjing, China	Mainly used for geological modelling of water conservancy and hydropower dam engineering, as well as simulation of water conservancy project construction
GoCAD (V.22.0)	INPNancy, Nancy, France	Recognized as the best geological modelling software, featuring strong functionality and a user-friendly interface
Petrel (V.22.0)	Schlumberger, New York, NY, USA	Integrated reservoir modelling developed for oil and gas exploration and development
RMS (V.11.0)	ROXAR Corporation, St. Louis, MO, USA	Widely used petroleum industry modelling software for predicting performance and production in oilfield development. It supports the analysis of geological data, design of reservoir development plans, production optimization, and decision-making
Earth Vision (V. 13.0)	DGI (Danamic Graphic) Company, Lafayette, LA, USA	Software system for three-dimensional geological modelling, analysis, and visualization. It can be used for the rapid creation of complex 3D models, reservoir characterization, reserves analysis, model validation, etc. Its structural modelling and complex fault-handling techniques are world-class. It is widely applied in oilfield geological research.

). The evolution of 3D geological modelling technology has exhibited a continuous commitment to innovation, progressing through distinct stages. These stages involved the early era of profile-based modelling, followed by software-driven modelling, and currently, the integration of machine learning and artificial intelligence-based modelling. Each stage has its own unique characteristics, advantages, and limitations. This study provided a comprehensive overview of the developmental phases of 3D geological modelling technology (

Table 2. Development stages of 3D geological modelling technology.

Development Stage	Characteristics	Advantages	Disadvantages	Technological Stage
Based on 2D profiles	Building a 3D model using geological profiles	Fast modelling speed	Insufficient modelling authenticity and accuracy	Deterministic modelling techniques Two-point geostatistics modelling technology Multipoint geostatistics modelling technology Numerical simulation technology based on geological processes
Based on 3D modelling software	Build a 3D model using professional 3D modelling software	High modelling accuracy, multiangle analysis, and display	Modelling speed is slow and requires a lot of time and effort
Based on machine learning and artificial intelligence technology	Combining machine learning and artificial intelligence technology to automate geological data processing	Fast modelling speed, more accurate and reliable modelling results	Requires significant consumption of computing resources and algorithm optimization	Intelligent geological modelling technology

) and its current applications (Table 3).

I. Deterministic modelling: This approach employs precise estimation techniques, including spline functions, geological profiles, and direct interpolation, each with distinct advantages and limitations. For instance, direct interpolation is recognized for its high speed and suitability for data-rich areas. However, it may produce overly smoothed results that lack fidelity in capturing geological complexities and variations. In addition, it is not well suited for regions with sparse or irregular data distributions.

II. Bivariate geological statistical modelling: The semivariogram function model quantifies decreasing variability with a distance between sample points, facilitating the determination of spatial correlations. Bivariate geological statistics are a fundamental component of traditional geological analysis. They guided deterministic models by employing various kriging methods and facilitated the development of equiprobable geological models through stochastic simulations [53]. Although this approach effectively captures spatial correlations, it requires substantial computational resources and relies on high-quality spatially distributed data. Furthermore, it encounters challenges in accurately representing complex spatial structures and maintaining the precision of the joint properties.

III. Multipoint geological statistical modelling: Multipoint geostatistical modelling overcomes single-point limitations by using bivariate statistics to create complex geological models. Both the two-point statistical method filled with multipoint criteria [54] and the multipoint Kriging method have significant advantages, such as simulating the unstable characteristics of geological objects and effectively describing and reconstructing the complex geometric shapes of nonlinear geological bodies [55]. This method also encountered specific challenges. Although multipoint statistical modelling can utilize nonstationary training images to reconstruct complex 3D geological models, it still faces challenges in practical applications. The selection of training images needs to ensure representativeness, as uneven distribution of sample points may affect simulation accuracy. For extreme or abnormal geological conditions, it is necessary to combine other geological information and professional knowledge. When applying this method, it is necessary to comprehensively consider data, model complexity, and actual geological conditions to ensure the accuracy of simulation results. Global scholars have performed extensive research on how to integrate multisource data in the modelling process to improve the accuracy and reliability of geological modelling. For example, Figueiredo et al. (2021) proposed a direct simulation technique for multivariate geostatistical simulation. They improved the efficiency of the algorithm applied to multivariate problems with more than three variables [56]. Liu et al. (2004) showed how to integrate well data, 3D seismic data, and geological information into multipoint simulation [57]. In addition, Wang et al. (2022) reconstructed a three-dimensional reservoir model from a two-dimensional image using seismic constraints based on multipoint statistics (MPS) [58]. Hansen et al. (2018) proposed two improved methods for how to integrate multisource data. The first method uses a priority simulation path to preferentially access model parameters with richer information. The second method involves the use of noncollocated uncertain information [59]. It is undeniable that multisource data fusion still faces challenges such as data quality, accuracy, and integrity [60]. Meanwhile, most MPS methods cannot consider the objects involved in the reconstruction process of the global spatial structural features of the model [55].

IV. Numerical simulation techniques based on geological processes: This stage relies on advancements in computer technology and incorporates numerical simulation techniques rooted in physical processes, including the finite element, finite difference, and lattice Boltzmann methods. These approaches conceptualize geological systems as complex assemblies comprising multiphase, multicomponent fluids, rocks, and soils, employing numerical simulations to reconstruct geological models.

V. Intelligent geological modelling technology: With the support of artificial intelligence and other technologies, intelligent geological modelling has become increasingly prominent in the geological field. These techniques use neural networks (such as GANs [61] and CNNs [62]) and deep learning [63] to automate the modelling of large volumes of geological data [62] and are expected to improve accuracy and efficiency. In the field of urban geological modelling, Hou [64] et al. (2023) successfully reconstructed the three-dimensional geological structure of a subway station in Guangzhou by combining a multipoint statistical method (MPS) and a deep neural network. In the field of reservoir geological modelling, the existing interdisciplinary research workflow includes traditional bivariate geostatistical modelling, multivariate geostatistical modelling, and sedimentary process-based modelling. It is worth noting that some scholars use the Gans neural network to generate geologically reasonable 3D reservoir facies models, which can capture more complex sedimentary facies structure and correlation than traditional methods [65]. However, artificial intelligence-based geological modelling techniques are needed to study the complex interactions between multiple disciplines. The essence of this method is to construct a multilayer neural network framework that can adapt to multidisciplinary, multidimensional, and multiscale data to solve the generation challenges related to complex geological objects.

VI. Geodynamics in 3D geological modelling: By applying the principles of geodynamics, 3D geological modelling can simulate the behaviour of materials inside the Earth to construct accurate models. This process visualizes the laws of motion and deformation, reflecting both static and dynamic geological processes. This improves the accuracy of exploration, resource assessment, and disaster prediction. However, complex simulation and data requirements pose challenges. Accuracy and reliability are crucial, which requires detailed modelling and simulation analysis based on principles of geodynamics. Dynamic geological modelling technology can simulate the changes in geological processes over time, such as plate movement, seismic activity, groundwater flow, etc. In recent years, with the improvement in computing power and the development of numerical simulation technology, dynamic geological modelling software such as MODLOW (V.6) [66] and Leapfrog Works have continuously emerged, providing powerful tools for the study of geological dynamic processes. Dynamic modelling technology has broad application prospects in fields such as local dynamic updates of models [67], dynamic simulation of geological hazards [68], and dynamic assessment of resources [69]. The current research hotspots mainly focus on how to improve the accuracy, efficiency, and visual expression of dynamic models. The future trend will focus on data fusion and dynamic process simulation to promote progress in geological research.

VII. Considering 3D geological modelling in exploration geochemistry: The application of 3D geological modelling in exploration geochemistry is gradually demonstrating its value. It integrates geological, geophysical, and geochemical data to construct a 3D geological structure model of the underground space, revealing the spatial distribution relationship between geological bodies and geochemical elements. This technology visualizes and quantifies complex geological information, enabling researchers to intuitively understand geological phenomena and predict geochemical anomalies. Its advantages lie in improving exploration accuracy and optimizing exploration plans, but it also faces challenges such as high data requirements and technical thresholds. With technological advancements, 3D geological modelling is expected to integrate with advanced technologies like big data and cloud computing, enabling stronger intelligent analysis and becoming a hotspot application tool in mineral resource exploration and geological disaster warning.

To summarize, 3D geological modelling has undergone significant development, progressing through various stages, from deterministic modelling techniques in the nascent years to the current era of intelligent geological modelling techniques. Each stage has its own distinct advantages and limitations. Furthermore, these technologies are not mutually exclusive but can be integrated to address specific scenarios and conditions, effectively overcoming the limitations associated with individual methods [64].

Table 3. Applications and comparative outcomes of 3D geological modelling worldwide.

Application Field	Specific Direction	District	Method	Effect		Reference
Mineral exploration	Implicit modelling	Brazil	3D implicit geological modelling	Determine the lithology and ore body model of the deposit and the complex tectonic framework	The geological characteristics and tectonic evolution history of the Kwatebala Cu-Co deposit are clarified.	[70]
	Evaluation of deep mineral resources	China	Based on a 3D model of the ore-controlling structure, the ore content ratio and similarity index are optimized to support the evaluation of gold resources.	Establish an accurate similarity index and accurately estimate the number of resources.	The quantitative evaluation of hydrothermal vein-type gold deposits is realized by improving the volume method and 3D geological modelling.	[71]
	Mineral resources evaluation	China	The process and update convenience of different 3D geological modelling are discussed.	The geological characteristics of Pb-Zn deposits in the Huayuan–Malichang area (MVT) of Hunan Province are clarified.	Determine data requirements, multiprocess modelling, and compare renewability.	[72]
Energy exploration	Geothermal exploration and utilization	France	A high-precision model is constructed based on borehole data, and 2D and 3D maps of temperature, salinity, porosity, permeability, and conductivity are used.	Reduce geological risks and optimize the location selection of future geothermal operations.	This study deepens the understanding of underground reservoirs and contributes to geothermal exploration and new well-planning.	[73]
	Geothermal Energy Potential Assessment	Germany	Using 3D seismic data, a new reservoir model is established to provide a basis for further development.	The main reflectors and stratigraphic units are drawn, and the 256 km3 underground structure is shown.	The geothermal potential and feasibility of studying deeply buried Permian reservoir rocks in the GroßSchönebeck geothermal study area.	[74]
	Analysis of the direction and size of ground stress in oil and gas development	China	Using borehole and acoustic emission measurement data, a three-dimensional geological model is constructed, and the ground stress field is predicted by finite element simulation.	Fault is the main factor of reservoir stress distribution in complex fault blocks.	It provides support for the efficient exploration and development of low-permeability reservoirs in complex fault blocks.	[75]
Structure	Three-dimensional modelling of fault geological structure	China	Using the generalized triangular prism model, a three-dimensional formation fracture modelling method based on GTP element reconstruction is proposed based on fault data.	The reconstruction of GTP elements solves the problem of faults and is suitable for complex geological modelling, which has obvious advantages.	The GTP model and reconstruction method are used to accurately model the fault distribution and improve the spatial analysis ability of the three-dimensional geological information model.	[76]
	Geological modelling and geophysical inversion	China	Three-dimensional geological structure inversion based on convolutional neural network.	Quickly predict geological structure parameters and restore geological structure with high accuracy.	The accuracy of synthetic and measured data is high, and the geological structure is effectively restored.	[77]
	Crustal scale structural interpretation	South Africa	The three-dimensional structure around the Vredefort Dome in the Witwatersrand Basin was evaluated by integrating geophysical and traditional geological data.	A new stratigraphic–structural relationship, first-order scale structure, and basement rock structure model are established to identify multiple stratigraphic–structural characteristics.	Integrating geophysical and geological data and using three-dimensional geological modelling technology to evaluate large-scale geological structures provide an important reference for geological research.	[78]
Modelling method	Automatic modelling of lens-containing and pinch-out strata	China	Determine the stratigraphic distribution and occupation, interpolate, intersect, correct the geological surface, and construct the geological body model.	Accurately reproduce the shape and distribution of geological bodies and significantly improve the modelling efficiency.	Develop multiscale drilling control methods, build 3D geological models, and support geological analysis.	[79]
	Construction of probability model for uncertain geological interpretation	Denmark	A three-dimensional hydro stratigraphic model is generated by geological interpretation and statistical simulation.	Reveal the uncertainty of the cognitive structure model and transform it into a probability model to convey to decision-makers.	Update the model with geological interpretation data, combined with geological intuition, to achieve model update and integration.	[80]
Hydrogeology	Basin model	Spain	MOVE 2017 software was used to develop a three-dimensional geological model to show underground geology and expand geological maps.	In-depth understanding of groundwater flow lays the foundation for future hydrogeological models	The 3D geological model updates the geological information of the Gallocanta Basin and highlights the effectiveness of 3D analysis.	[81]
	Study on groundwater management and flow	Canada	A multisource database is used to establish a three-dimensional hydro stratigraphic model to evaluate the uncertainty of geological characteristics, hydrodynamics, and groundwater storage.	The SRBS method calculates the optimal storage value and generates the formation storage map.	The regional hydrogeological model is established, and the new statistical method is used to spread the uncertainty.	[82]
	Quantification of groundwater resources	Norway	Integrate hydrogeological and oil and gas exploration data, integrate multisource modelling platforms and underground geological software.	The deep geological model is constructed for the first time in the Horn of Africa, revealing the deep aquifer system and providing new insights for the study of transboundary aquifers and the quantification of water resources.		[83]
Urban geology	Urban underground engineering decision-making	China	By integrating the buffer and inverse distance weighting algorithm, a new 3D modelling workflow is constructed.	It can accurately display the complex spatial structure of urban shallow strata. Provide strong support for urban underground engineering decision-making.		[84]
	Geological research and engineering geology	Singapore	A 3D geological model of the Jialeng River Basin was created using 161 boreholes and analyzed in combination with sediment samples and dating data.	The study constructs Singapore’s Quaternary stratigraphic framework to constrain regional geomorphological evolution and sea level changes. This study provides important information for the geological evolution and engineering geological work on the southern coast of Singapore.		[85]
	Three-dimensional modelling of urban geology	China	Combining analysis and layout, information is extracted from borehole records, and 3D geological modelling is performed based on knowledge-driven dynamic profiles.	Paper drilling records were converted into structured data for direct analysis. This method has been proved to be feasible and provides a new way for three-dimensional modelling of urban geology.		[86]
Multisource data fusion	Mineral exploration and resource evaluation	China	A 3D geological model is constructed by using geological data, and multilevel exploration standards are extracted, which are integrated into a 3D posterior probability model, and the posterior probability thresholds of 2D and 3D targets are defined.	The application of the evidence weight method and interpolation method to integrate exploration standards reveals high potential copper ( gold ) targets, provides an effective way for blind deposit location, and significantly improves the efficiency and accuracy of mineral exploration.		[87]
Visual interaction	Virtual reality	China	Using the borehole data in the GIS database to generate independently displayed models, supporting functions such as rotation, flight, and scaling, without the need for specialized modelling software.	The method is simple, with less manual intervention, and is suitable for engineering investigation decision support.	Combined with GIS and VRML, a real-time 3D geological modelling method is proposed, which is suitable for engineering survey and geological modelling, provides visual decision support, and improves the efficiency and accuracy of survey.	[88]
Machine learning	Geological data analysis	China	The NSF-McNURBS method uses the NSF model to learn complex high-dimensional joint distribution to achieve accurate estimation and joint sampling of three-dimensional geological parameters.	Compared with the traditional NURBS method, the fitting error is reduced by 52.4 %, flexibly and automatically considering the constraints of geological point coordinates and strike and dip angles.	In this study, the neural spline flow and McNURBS method are introduced to automatically construct a reliable three-dimensional geological model. Considering the constraints of geological points, the accuracy of geological modelling is improved.	[89]
	Geological data processing and intelligent generation	China	The cross-section intelligent generation based on CGAN is realized by combining CGAN with multisource geological data.	The accuracy of this method is higher than that of GAN and VAE models ( 87 % and 68 %, respectively ), and the generated 3D geological model is consistent with the manually created model.	This study solves the problem of inefficient data processing based on CGAN through intelligently generated cross-section methods, and provides an important method for regional 3D geological modelling.	[90]
	Training and verification of large labeled data	Australia	The Noddy platform was used to generate 1 million 3D geological models and corresponding gravity and magnetic data, forming a large-scale set.	The generation model solves the problem of lacking datasets in geoscience and provides a test for machine learning and geophysical inversion. It is valuable for geological problems.		[91]
Machine learning	Mineral prospect evaluation based on machine learning	China	Based on the known location of the deposit, the feature is extracted; trained with k nearest neighbor, neural network, and support vector machine; and applied to the 3D data of the whole region to obtain multiple sets of 3D mineral prospect models.	The mineral prospect model based on machine learning combined with the Earth model obtained from a variety of geophysical datasets and 3D geological modelling can promote 3D mineral prediction, thus greatly improving the efficiency of discovering further deposits.		[92]
Model uncertainty	Uncertainty of formation attributes	China	Markov chain is used to generate probability matrix, Monte Carlo simulation is used to generate formation state, and triangular irregular network (TIN) is used to build probability model.	The formation configuration is visualized effectively and the formation uncertainty is quantified.	This method can effectively generate a three-dimensional geological model and consider the uncertainty of formation attributes, which is helpful in optimizing resource development and management.	[93]
	Uncertainty assessment framework of geological model	China	Spatial diffusion and merging models are used to deal with the uncertainty in geological modelling. A quantitative method of geological uncertainty based on conditional information entropy is proposed, which takes more fully into account the constraints of geological laws.	In this study, the spatial model, combined model, and conditional information entropy method are used to establish the uncertainty evaluation framework of the geological model.		[94]
	Mineral exploration and resource assessment	China	Based on the 3D geological model and exploration data, three-dimensional models of multiple geological objects are constructed using GOCAD to analyze mineralization trends and carry out uncertainty analysis.	In the Shanggong gold deposit, a 3D geological model and attribute modelling are used to identify targets and assess resources. The high-grade area is related to intrusion and fault structure. Interpolation shows the trend of attributes, and simulation and uncertainty analysis reveal the uncertainty of the model.		[95]

Table 3. Applications and comparative outcomes of 3D geological modelling worldwide.

Application Field	Specific Direction	District	Method	Effect		Reference
Mineral exploration	Implicit modelling	Brazil	3D implicit geological modelling	Determine the lithology and ore body model of the deposit and the complex tectonic framework	The geological characteristics and tectonic evolution history of the Kwatebala Cu-Co deposit are clarified.	[70]
	Evaluation of deep mineral resources	China	Based on a 3D model of the ore-controlling structure, the ore content ratio and similarity index are optimized to support the evaluation of gold resources.	Establish an accurate similarity index and accurately estimate the number of resources.	The quantitative evaluation of hydrothermal vein-type gold deposits is realized by improving the volume method and 3D geological modelling.	[71]
	Mineral resources evaluation	China	The process and update convenience of different 3D geological modelling are discussed.	The geological characteristics of Pb-Zn deposits in the Huayuan–Malichang area (MVT) of Hunan Province are clarified.	Determine data requirements, multiprocess modelling, and compare renewability.	[72]
Energy exploration	Geothermal exploration and utilization	France	A high-precision model is constructed based on borehole data, and 2D and 3D maps of temperature, salinity, porosity, permeability, and conductivity are used.	Reduce geological risks and optimize the location selection of future geothermal operations.	This study deepens the understanding of underground reservoirs and contributes to geothermal exploration and new well-planning.	[73]
	Geothermal Energy Potential Assessment	Germany	Using 3D seismic data, a new reservoir model is established to provide a basis for further development.	The main reflectors and stratigraphic units are drawn, and the 256 km3 underground structure is shown.	The geothermal potential and feasibility of studying deeply buried Permian reservoir rocks in the GroßSchönebeck geothermal study area.	[74]
	Analysis of the direction and size of ground stress in oil and gas development	China	Using borehole and acoustic emission measurement data, a three-dimensional geological model is constructed, and the ground stress field is predicted by finite element simulation.	Fault is the main factor of reservoir stress distribution in complex fault blocks.	It provides support for the efficient exploration and development of low-permeability reservoirs in complex fault blocks.	[75]
Structure	Three-dimensional modelling of fault geological structure	China	Using the generalized triangular prism model, a three-dimensional formation fracture modelling method based on GTP element reconstruction is proposed based on fault data.	The reconstruction of GTP elements solves the problem of faults and is suitable for complex geological modelling, which has obvious advantages.	The GTP model and reconstruction method are used to accurately model the fault distribution and improve the spatial analysis ability of the three-dimensional geological information model.	[76]
	Geological modelling and geophysical inversion	China	Three-dimensional geological structure inversion based on convolutional neural network.	Quickly predict geological structure parameters and restore geological structure with high accuracy.	The accuracy of synthetic and measured data is high, and the geological structure is effectively restored.	[77]
	Crustal scale structural interpretation	South Africa	The three-dimensional structure around the Vredefort Dome in the Witwatersrand Basin was evaluated by integrating geophysical and traditional geological data.	A new stratigraphic–structural relationship, first-order scale structure, and basement rock structure model are established to identify multiple stratigraphic–structural characteristics.	Integrating geophysical and geological data and using three-dimensional geological modelling technology to evaluate large-scale geological structures provide an important reference for geological research.	[78]
Modelling method	Automatic modelling of lens-containing and pinch-out strata	China	Determine the stratigraphic distribution and occupation, interpolate, intersect, correct the geological surface, and construct the geological body model.	Accurately reproduce the shape and distribution of geological bodies and significantly improve the modelling efficiency.	Develop multiscale drilling control methods, build 3D geological models, and support geological analysis.	[79]
	Construction of probability model for uncertain geological interpretation	Denmark	A three-dimensional hydro stratigraphic model is generated by geological interpretation and statistical simulation.	Reveal the uncertainty of the cognitive structure model and transform it into a probability model to convey to decision-makers.	Update the model with geological interpretation data, combined with geological intuition, to achieve model update and integration.	[80]
Hydrogeology	Basin model	Spain	MOVE 2017 software was used to develop a three-dimensional geological model to show underground geology and expand geological maps.	In-depth understanding of groundwater flow lays the foundation for future hydrogeological models	The 3D geological model updates the geological information of the Gallocanta Basin and highlights the effectiveness of 3D analysis.	[81]
	Study on groundwater management and flow	Canada	A multisource database is used to establish a three-dimensional hydro stratigraphic model to evaluate the uncertainty of geological characteristics, hydrodynamics, and groundwater storage.	The SRBS method calculates the optimal storage value and generates the formation storage map.	The regional hydrogeological model is established, and the new statistical method is used to spread the uncertainty.	[82]
	Quantification of groundwater resources	Norway	Integrate hydrogeological and oil and gas exploration data, integrate multisource modelling platforms and underground geological software.	The deep geological model is constructed for the first time in the Horn of Africa, revealing the deep aquifer system and providing new insights for the study of transboundary aquifers and the quantification of water resources.		[83]
Urban geology	Urban underground engineering decision-making	China	By integrating the buffer and inverse distance weighting algorithm, a new 3D modelling workflow is constructed.	It can accurately display the complex spatial structure of urban shallow strata. Provide strong support for urban underground engineering decision-making.		[84]
	Geological research and engineering geology	Singapore	A 3D geological model of the Jialeng River Basin was created using 161 boreholes and analyzed in combination with sediment samples and dating data.	The study constructs Singapore’s Quaternary stratigraphic framework to constrain regional geomorphological evolution and sea level changes. This study provides important information for the geological evolution and engineering geological work on the southern coast of Singapore.		[85]
	Three-dimensional modelling of urban geology	China	Combining analysis and layout, information is extracted from borehole records, and 3D geological modelling is performed based on knowledge-driven dynamic profiles.	Paper drilling records were converted into structured data for direct analysis. This method has been proved to be feasible and provides a new way for three-dimensional modelling of urban geology.		[86]
Multisource data fusion	Mineral exploration and resource evaluation	China	A 3D geological model is constructed by using geological data, and multilevel exploration standards are extracted, which are integrated into a 3D posterior probability model, and the posterior probability thresholds of 2D and 3D targets are defined.	The application of the evidence weight method and interpolation method to integrate exploration standards reveals high potential copper ( gold ) targets, provides an effective way for blind deposit location, and significantly improves the efficiency and accuracy of mineral exploration.		[87]
Visual interaction	Virtual reality	China	Using the borehole data in the GIS database to generate independently displayed models, supporting functions such as rotation, flight, and scaling, without the need for specialized modelling software.	The method is simple, with less manual intervention, and is suitable for engineering investigation decision support.	Combined with GIS and VRML, a real-time 3D geological modelling method is proposed, which is suitable for engineering survey and geological modelling, provides visual decision support, and improves the efficiency and accuracy of survey.	[88]
Machine learning	Geological data analysis	China	The NSF-McNURBS method uses the NSF model to learn complex high-dimensional joint distribution to achieve accurate estimation and joint sampling of three-dimensional geological parameters.	Compared with the traditional NURBS method, the fitting error is reduced by 52.4 %, flexibly and automatically considering the constraints of geological point coordinates and strike and dip angles.	In this study, the neural spline flow and McNURBS method are introduced to automatically construct a reliable three-dimensional geological model. Considering the constraints of geological points, the accuracy of geological modelling is improved.	[89]
	Geological data processing and intelligent generation	China	The cross-section intelligent generation based on CGAN is realized by combining CGAN with multisource geological data.	The accuracy of this method is higher than that of GAN and VAE models ( 87 % and 68 %, respectively ), and the generated 3D geological model is consistent with the manually created model.	This study solves the problem of inefficient data processing based on CGAN through intelligently generated cross-section methods, and provides an important method for regional 3D geological modelling.	[90]
	Training and verification of large labeled data	Australia	The Noddy platform was used to generate 1 million 3D geological models and corresponding gravity and magnetic data, forming a large-scale set.	The generation model solves the problem of lacking datasets in geoscience and provides a test for machine learning and geophysical inversion. It is valuable for geological problems.		[91]
Machine learning	Mineral prospect evaluation based on machine learning	China	Based on the known location of the deposit, the feature is extracted; trained with k nearest neighbor, neural network, and support vector machine; and applied to the 3D data of the whole region to obtain multiple sets of 3D mineral prospect models.	The mineral prospect model based on machine learning combined with the Earth model obtained from a variety of geophysical datasets and 3D geological modelling can promote 3D mineral prediction, thus greatly improving the efficiency of discovering further deposits.		[92]
Model uncertainty	Uncertainty of formation attributes	China	Markov chain is used to generate probability matrix, Monte Carlo simulation is used to generate formation state, and triangular irregular network (TIN) is used to build probability model.	The formation configuration is visualized effectively and the formation uncertainty is quantified.	This method can effectively generate a three-dimensional geological model and consider the uncertainty of formation attributes, which is helpful in optimizing resource development and management.	[93]
	Uncertainty assessment framework of geological model	China	Spatial diffusion and merging models are used to deal with the uncertainty in geological modelling. A quantitative method of geological uncertainty based on conditional information entropy is proposed, which takes more fully into account the constraints of geological laws.	In this study, the spatial model, combined model, and conditional information entropy method are used to establish the uncertainty evaluation framework of the geological model.		[94]
	Mineral exploration and resource assessment	China	Based on the 3D geological model and exploration data, three-dimensional models of multiple geological objects are constructed using GOCAD to analyze mineralization trends and carry out uncertainty analysis.	In the Shanggong gold deposit, a 3D geological model and attribute modelling are used to identify targets and assess resources. The high-grade area is related to intrusion and fault structure. Interpolation shows the trend of attributes, and simulation and uncertainty analysis reveal the uncertainty of the model.		[95]

2.2. Application Hotspots

Notably, 3D geological modelling is a fundamental component of geological research, with wide-ranging applications in exploration, oil and gas prospecting, and mineral exploration. The continual advancement of technology and data processing has increased its significance in geological science and engineering. Current global developments are centered on refining modelling techniques, enhancing spatial databases, and improving application systems. In this context, this study provides an overview of the prevailing research hotspots in geological modelling (

Table 4. Hot research topics in the global 3D geological modelling field.

Research Hotspots	Research Hotspots
Data collection and processing field	Based on the technology of unmanned aerial vehicle (UAV) oblique photography. It combines high definition, precision, and wide coverage to comprehensively perceive complex geological scenes. Its agile global perspective and high quantification enable the collection of original geological data [96,97]. Compared to traditional methods, it greatly improves field reconnaissance efficiency while mitigating risks. Existing challenges: Data collection quality is influenced by unpredictable weather, affecting photorealistic model clarity and resolution. While models can often achieve centimeter-level precision, they may not meet the needs of scenes requiring finer analysis, e.g., distinguishing sandstone types [98,99]. Additionally, data processing and acquisition in rugged terrains pose challenges.	Multiscale and multisource data fusion. Methods using multiscale and multisource data fusion effectively process geological data for oil/gas exploration [100], mineral exploration/evaluation [101], geological hazard research [102], and urban sedimentary strata modelling [103]. They integrate data from various scales and sources, enhancing model accuracy, data application scope, and resource utilization efficiency. Existing challenges: The multisource and heterogeneous nature of geological data hinders effective fusion. Data redundancy and uncertainty lead to the presence of redundant or conflicting information in multiple data sources, making the removal of redundant information and the treatment of uncertainty an important issue.	Laser scanning technology. The method obtains point cloud data with 3D coordinates, reflectance, and other properties from scanned objects. These data can be used to create 3D models to simulate complex real-world forms. Applications include digital outcrop modelling [104,105], visualizing concealed cavities in goaf areas [106], developing small-scale regional structural models [107], assessing the stability of precarious rocks in engineering projects [108], and monitoring the dynamic reserves of open-pit mines [109]. Existing challenges: The handling of point cloud data confronts issues such as insufficient precision in point cloud matching, lackluster performance in automatic matching, partial loss of data features, and excessive automatic removal of point clouds, all posing significant obstacles to be overcome.
Modelling Method Domain	The three-dimensional geological modelling methodology based on machine learning. Data-driven methods like ANNs [110], logistic regression [111], PCA [112], and WoE [113] are commonly used in geospatial analysis and MPM [114,115]. By integrating 3D geological modelling with various geological predictive factors, WoE can analyze their spatial correlation with mineral occurrence. Appropriate criteria are then selected to quantitatively calculate posterior probabilities and accurately generate 3D MPMs [116,117,118]. Existing challenges: Acquiring abundant, high-precision training data is crucial. Noise or blurriness in seismic data can compromise model predictions. High-precision data, like field observations, satellite imagery, and high-resolution seismic data are costly. Challenges persist in feature selection and DNN self-learning.
	The method of implicit three-dimensional geological modelling adopts spatial interpolation techniques. Using sampled data to fit spatial surface functions generates 3D visual models, overcoming traditional explicit modelling’s limitations of manual labor, tediousness, inefficiency, and subjective errors. Coupled with machine learning, this offers a controlled, automated approach for 3D geological modelling based on sparse data [119,120]. Existing challenges: Choosing and optimizing modelling algorithms is crucial for accurate and efficient implicit 3D geological modelling. Algorithms vary in suitability, so selecting the right ones and optimizing them is challenging. Common algorithms include random fields, decision trees, and probabilistic neural networks. Data type, quantity, and modelling objectives must be considered. Each algorithm needs targeted optimization to improve efficiency and accuracy. Research by scholars, such as machine learning-based methods [121], HRBF [122], and VGP algorithms, offers valuable solutions for automated, smooth, continuous, and timely updated modelling with limited data [123].
Application areas	Geological modelling aims to predict deep-seated concealed mineral deposits. With rising mineral demand and dwindling shallow deposits, the hunt for deep, concealed minerals is urgent [53]. Scholars worldwide have intensively studied this, establishing a comprehensive modelling process [124,125]. Their efforts have significantly reduced exploration risks and workloads, making this a current and future exploration hotspot.	Research on geological models for the prevention of geological disasters is crucial. Frequent geological hazards are a severe threat to lives and impact economic development globally, with China being one of the affected regions. Three-dimensional laser scanning technology reproduces geological disaster models, providing technical assurance for future prevention. Prioritizing human well-being and strengthening research on geological models’ role in disaster prevention is crucial. The establishment of a three-dimensional visualization system for urban landscapes is imperative. As urbanization increases, the need for visual analysis of urban geology and underground info rises. Geological factors, water resources, and disasters constrain urban development. Current 3D visualization methods include Cesium.js, Three.js, Unity, etc., each with pros and cons. Integrating their strengths is crucial for future visualization technology.	Geological modelling has experienced rapid development in visualizing and analyzing groundwater and pollution in recent years. With only 0.5% of freshwater on Earth, waste and pollution make the situation dire. Scholars use hydrochemical data to visualize groundwater composition and aquifer structure [126,127]. Cesium.js technology enables spatial and temporal interpretations of geological formations and aquifers [128]. Accurate knowledge of aquifer structures prevents pollution and waste from urbanization and engineering. It also averts land subsidence due to excessive groundwater extraction and provides rational urban development advice.
	Application in oil and gas energy exploration. The rapidly expanding city’s energy dependency led to an urgent search for oil. Scholars have made significant progress in comprehensive multiscale reservoir modelling, particularly in fractured carbonate and clastic reservoirs, advancing modelling methods [129,130,131,132].	Research on quality control of three-dimensional geological models is underway. The quality of the models is influenced to varying degrees by factors such as raw data, interpolation methods, modelling techniques, computer and operator actions, and the complexity of geological phenomena. Currently, there are some theories (structural model control, attribute model control) that can partially address the issue of model quality. However, there are still areas that can be further optimized, such as how to ensure modelling accuracy with limited geological data, the lack of comprehensive and unified standards for evaluating model accuracy, and how to reduce the subjective impact on model quality by increasing constraint conditions.	The application of geological modelling combined with temporal attributes plays a significant role in interpreting the evolution of ancient structures, sedimentary environments, paleogeography, and the development of oil and gas fields. Using diverse data, structural restoration [133] and dynamic modelling of oil/gas fields are feasible. Potential applications include strain calculations to predict properties and fracture distribution in small-scale reservoirs, crucial for oil exploration [134,135,136,137,138,139,140,141,142].

).

3. Technical Content

3.1. Geological Modelling Data

3.1.1. Original Geological Data

Geological data form the foundational basis for 3D geological modelling and play a pivotal role in the creation, refinement, and evaluation of geological models. The utilization of data characterized by differing resolutions, origins, strata, and dimensions significantly enhances the accuracy and reliability of these models. Notable sources of geological data include geological cartography data, stratigraphic profiles, remote sensing data, seismic reflection profiles, geophysical records, well-logging data, geochemical data, and deep continental scientific drilling data, which extend to depths exceeding 13.4 km.

In current geological research, significant progress has been made in modelling basic geological structures, stratigraphic formations, and structural models. Conversely, the modelling of complex geological formations presents challenges owing to the acquisition of 3D spatial data. The model quality is adversely affected by insufficient geological knowledge, fragmented data sources, and suboptimal analysis techniques. To mitigate these challenges, we highlight two approaches: Firstly, data collection has been enhanced through advanced methods such as 3D laser scanning and aerial magnetometry. Secondly, Integrated technologies capable of effectively merging diverse geological data sources have been developed, with a particular focus on constraint modelling to enhance data fusion and constraints.

3.1.2. Spatial Data Model

The spatial data model is fundamental for spatial design, analysis, and geological model construction. It includes data structures, operations, and constraints that facilitate data storage, management, exchange, and utilization. In 1978, Hunter proposed the octree data model, followed by Pilouk’s presentation of the irregular tetrahedral data model, and Chinese scholars introduced the prism data model [143,144]. These models can be categorized into three types based on their inherent attributes: surface models, solid models, and hybrid models. Each model has distinct advantages and disadvantages that make it suitable for different contexts. For instance, surface models represent regular layered deposits and simple geological forms but struggle with complex geological structures. Solid models offer maneuverability and analytical power but have limitations in terms of accuracy and storage requirements. Hybrid models combine surface and solid elements to optimize modelling efficiency. Researchers have proposed solutions to overcome these limitations. Local virtual borehole modelling rectifies omissions in drilling control data and enhances fault zone modelling affected by stratigraphic discontinuities. Innovative approaches, such as the multiscale octree subdivision algorithm, efficiently organize data across various scales and tackle the complexities of modelling intricate geological structures. Overall, selecting the most appropriate model tailored to a specific geological context is critical for geological modelling. Building on previous research, this study provides a comprehensive overview of the strengths and limitations of the existing models (

Table 5. Advantages and disadvantages of common data models.

Model	Name	Disadvantages	Advantages
Face model	B-Rep	Large volume of data, resulting in low efficiency in handling complex geological structures	Simple operations, suitable for constructing regular surfaces or solids
	TIN	Inability to view object properties during modelling	Surface modelling is simple, easy to modify, and accurately represents changes in the surface.
	Wireframe	Insufficient spatial topology representation	Small data volume, simple data structure, easy to modify
	Quadrilateral Grid	Regular grid units may lead to data redundancy in areas where geological boundaries or geological phenomena vary dramatically. Rule-based grid cells lack the flexibility to adapt to complex and variable geological phenomena	Easy to generate high-quality 3D visualizations. Grid cells can carry a wealth of attribute information.
	Quadtree	Spatial data with irregular distribution or large changes in data density may cause uneven distribution of nodes and decrease query efficiency. Poor scalability for high-dimensional data (more than 3D)	Simple structure. The query efficiency is high when regular grid data or data are evenly distributed. The index structure is relatively compact.
Volume model	Octree	Large data volume, difficulty in displaying geological body boundaries, unable to preserve original observation data	Simple structure, easy conversion and representation of internal attribute information, fast retrieval
	Needle	Lower accuracy, susceptibility to speed issues in data processing	Space-saving
	GTP	Difficult visualization of complex objects	Complete topological information, convenient for entity querying and analysis
	TEN	Difficulties in modelling complex geological bodies, large data volume, insufficient representation of linear or planar features	Displaying internal attribute changes, complete topological relationships, easy decomposition, and data preservation
	Voxel	Large data volume, inadequate representation of entity shape, position, and topological relationships, unable to be manually corrected.	Simple operations
Hybrid model	TEN-Octree	Rapid increase in data volume, complex algorithms, difficulties in spatial object topology	Balancing overall description and local representation, higher modelling accuracy
	TIN-Octree	Mutual impact and pointer confusion, difficult data maintenance	Balancing surface and internal modelling, topological representation, and querying
	TIN-CSG	Difficulty in expressing complex geological objects, such as faults, folds, and fractures	Balancing surface and building modelling, separating data storage, and operating display
	Wire Frame-Block	Low practical efficiency, requiring further block segmentation and model modifications for every boundary change	Balancing target outline or boundary with internal expression, improving boundary model accuracy
	TIN-Grid	Implementation complexity. Data consistency maintenance is challenging.	Efficient expression of complex terrain; users can flexibly choose and use different data structures according to the actual situation, with improved computing efficiency.

).

3.2. Geological Modelling Objects

The primary objective of 3D geological modelling is to generate accurate and comprehensive models for geological and mineral exploration and surveying purposes. This includes capturing crucial aspects, such as topography, rock distribution, structural features, and mineral deposits. The objects to be modeled can be categorized into five distinct categories.

3.2.1. Reservoir Model

The 3D geological modelling process, which involves an interdisciplinary fusion of geology, geophysics, and computer science, plays a pivotal role in the oil and gas sector by translating data into accurate representations. This process enhances the comprehension of geological factors and reservoir distributions, informs decision-making processes, optimizes strategies, facilitates the selection of well-site locations, improves operational efficiency, mitigates the risk of overexploitation, and facilitates reservoir evaluation and resource estimation. Consequently, it contributes to the reliable and sustainable development of the oil and gas industry.

In recent years, the geological characterization and modelling of deep-to-ultra-deep carbonate reservoirs have gained prominence, focusing on intricate fine-scale analysis. Classically, research by Guo et al. (2021) [145] explored the intricate mechanisms governing the development and spatial distribution of high-quality reservoirs within carbonate formations. These efforts have yielded advanced technologies that are tailored to model the complexity of deep carbonate reservoirs. These technologies employ tailored modelling approaches for various sedimentary environments and reservoir types. The academic scientific community has put forth scholarly contributions proposing the establishment of comprehensive models that encompass the digital outcrop-based characterization of carbonate reservoirs, sedimentary microfacies, porosity, and permeability [146] (Figure 1). These models capture lithological variations, permeability distributions, porosity characteristics, and their interactions, thereby serving as a precise foundation for reservoir prediction and seismic interpretation. Gomes [147] presented a new open source carbonate reservoir case study, the COSTA Model, which uniquely considers the significant uncertainties of carbonate reservoirs to provide a more challenging and realistic benchmarking. Derived from the geological setting of the Upper Kharaib Member, the model covers large-scale geological settings and reservoir heterogeneity and can be used for geological modelling and reservoir simulation. The study produced semi-synthetic but geologically realistic carbonate reservoir models that match synthetic "real cases" and are available in an open source form to test the application of new numerical algorithms in geological modelling, uncertainty quantification, reservoir simulation, etc. Franc [148] introduced a novel multiscale finite volume method designed for addressing multiphase reservoir models. The method involves solving the pressure field on a coarse grid and addressing saturation fields on a finer reservoir grid, achieving resolution for dual-mesh scales. This adaptive refinement, based on Lagrange multiplier space defined on coarse faces, was validated through applications in heterogeneous porous media, including single-phase (well-testing) and multiphase flow scenarios.

In the future, meticulous small-scale geological modelling, deep to ultra-deep reservoir geological modelling, and artificial intelligence-driven 3D geological modelling hold significant potential in the field of petroleum geology [145,149].

3.2.2. Surface Model

The application of Earth’s surface modelling involves the emulation of its morphological features, primarily using digital elevation models (DEMs). These models represent spatial and topographical attributes and offer visual representations. The development of comprehensive 3D models is of paramount importance in geological endeavours and serves as an indispensable tool in modern digitalized geological practices. Surface models exhibit remarkable versatility, rendering them invaluable for conducting initial regional surveys by seamlessly integrating various types of geological data. This integration facilitates the development of optimal exploration strategies and pathways based on the terrain characteristics.

The construction process entails a series of meticulously planned operations including coordinate transformation, georeferencing, and cropping (Figure 2). These procedures were meticulously executed to enhance the visual quality of imagery within the geographical coordinates of longitude 90–91° E and latitude 45–46° N using ENVI software. Simultaneously, we conducted the digital processing of a geological map of Urumqi City at a 1:200,000 scale. The spatial coordinate systems of the digitized geological map, remote sensing images, and DEM elevations were harmonized. Ultimately, the ArcScene platform was employed to retrieve and analyze the spatial data, resulting in the establishment of a comprehensive surface model (Figure 3). Scholars have advocated the application of the constrained Delaunay meshing method to model intricate surface undulations and simulate gravity anomalies within intricate geological formations. Moreover, another group of researchers [149] have utilized digital geological mapping and modelling techniques to produce versatile cross-sectional representations, facilitate dynamic updates, and generate geological interfaces, thereby presenting substantial potential for practical applications (Figure 4).

In conclusion, seemingly straightforward surface modelling involves intricate processes with limited human intervention, posing challenges in achieving rapid, automated modelling for specific regions. The future trend involves the efficient utilization of geological exploration data to swiftly gain insights into regional geology and enhance work efficiency, necessitating further in-depth research in these domains.

3.2.3. Tectonic Models

Tectonic modelling is pivotal for numerous reasons. It enhances the 3D visualization of structural properties and their interplay with geological formations and streamlined data management, supporting the evaluation of geological hazards and guiding mineral exploration, particularly for fault and fold-related deposits.

Structural modelling represents a cutting-edge direction in contemporary structural research. Compared to their two-dimensional counterparts, 3D structural modelling offers superior advantages in terms of accuracy, applicability, and visualization. Oakley [150] demonstrated the use of ensemble Kalman inversion (EKI) for building three-dimensional, multifault, kinematically restorable structural geologic models, by means of a workflow in which fault geometry, the distribution of slip on a fault, and the geometry of folded horizons are all modeled. EKI can recover the true parameter values in the synthetic case and produce a solution consistent with the data in the real case, as well as quantify uncertainty in both cases. Wang [151] applied machine learning to structural modelling. The training network based on the proposed modelling framework can provide better fault interpretation results, and the proposed geological model has good generalization, which can effectively improve the applicability of machine learning.

The modelling of distinct geological structures, such as multiphase faults, faulted basins, and suture zones, remains limited and often produces suboptimal results. The primary challenges in structural modelling include the following:

(1) Uncertainties that impact model quality: variabilities stemming from data sources, geological interpretations, software inconsistencies, and human operations can substantially influence the model integrity;

(2) Establishing intricate geological surfaces and volumes;

(3) Model refinement: Contemporary global research indicates that implicit modelling techniques may provide solutions for intricate structural quandaries.

3.2.4. Deposit Model

Modelling mineral deposits constitutes both a centerpiece and a challenge in 3D modelling endeavours. This process not only captures the morphology of ore bodies but also assesses the potential of mineral resources. Typically, mineral deposit modelling is executed using three primary methodologies [152], which are the following:

(1) The morphology of the ore body was established by identifying contact boundaries with adjacent rock formations;

(2) A mineral deposit model was developed, which was derived from pre-established profile maps;

(3) Multiple data sources, including boreholes and geophysical exploration, were utilized to model mineral deposits (Figure 5A,D).

Researchers have employed original borehole data for the implicit modelling of ore bodies, facilitating rapid and accurate modelling, along with model updates [119]. Wu [153] presented a surface modelling method for interpolating orebody models based on a set of cross-contour polylines (geological polylines interpreted from the raw geological sampling data) using the bi-Coons surface interpolation method. The method is particularly applicable to geological data with cross-contour polylines acquired during the geological and exploration processes. The innovation of this method is that the proposed method can automatically divide the closed loops and automatically combine the sub-meshes. Moradpouri [154] used 2045 soil samples to analyze 36 elements and combined borehole data for 3D ore body modelling of Cu, Mo, Pb, and Zn elements to validate and demonstrate the results obtained. This is a classic case of going from shallow to deep.

3.3. Geological Modelling Methods

Since its introduction in 1993, the concept of 3D geological modelling has evolved over three decades. Although it has been developed internationally for decades, it has only been accepted in China for nearly 30 years since its introduction in the late 1980s. Researchers have developed various modelling techniques worldwide (

Table 6. Classification of different aspects of geological modelling.

Classification Perspective	Classification	Description
Scale	Microscopic Scale	Three-dimensional point cloud modelling based on microscopic characteristics of rock and mineral samples using instruments
	Macroscopic Scale	Regional geological models established based on outcrop data, borehole data, etc.
Geological Body Attribute Analysis	Attribute Modelling	Modelling of nonuniform attributes within geological bodies based on statistical analysis
	Structural Modelling	Modelling the geometric features of geological bodies in space
Data Sources for Modelling	Field Data Models	Based on remote sensing data, DEMs, etc.
	Geological Profiles	Geological modelling based on geological profiles
	Borehole Data	Geological models established based on borehole data to reflect the distribution of underground geological bodies
	Multisource Data	Geological modelling based on the fusion of geological, geophysical, and borehole data
Modelling Dynamism	Static Modelling	Models based on the characteristics of a geological body at a specific moment
	Dynamic Modelling	Models based on the characteristics of a geological body over a continuous period of time
Research Tools	System Integration	Collection of geological models from multiple modelling systems
	Database	Models stored and managed in databases
Need for Manual Involvement	Explicit Modelling	Defining the boundaries of geological bodies based on profiles and creating model surfaces using grid
	Implicit Modelling	Surface reconstruction without grids, utilizing specific isosurfaces extracted from volumetric functions to depict the surface of geological bodies

). Geological modelling methods vary by region and discipline, with mainstream methods including section-based, borehole data-driven, and multisource data modelling, accompanied by the emerging AI-based 3D geological modelling. In this paper, 3D geological modelling methods are divided into two categories based on the data source and method principles (

Table 7. Advantages and disadvantages of common modelling methods.

Classification Criteria	Modelling Methods	Disadvantages	Advantages
Based on data sources	Based on DEMs	Simple data are prone to geological information migration and distortion.	Fast modelling speed and relatively simple modelling process
	Based on drilling data	It is difficult to solve the modelling of complex geological phenomena such as faults and inverted strata, and the corresponding strata between boreholes, the location of strata pinching out, and the geological information between boreholes are prone to distortion and deviation.	Simple modelling process
	Based on profile data	Difficult to reflect accurate geological information, easily losing certain details and causing distortion and deviation of geological information.	Simple modelling process
	Based on 3D seismic data	The complexity of data interpretation, high cost, high requirements for computer performance, and complexity of data processing.	Notably, 3D seismic data can provide significant underground structure information, describe the underground structure in detail, and improve the accuracy and reliability of modelling.
	Based on multisource data	Easy to be affected by issues such as data consistency, overly integrating all data can easily lead to displacement and distortion of geological interfaces or attributes; The modelling effect for complex geological bodies is good, but the difficulty of data fusion is relatively high.	The quality and accuracy of the model have been greatly improved, and it can also model complex geological bodies well.
Based on technical principles	Based on cross-folded sections	Although cross-folded section modelling can deal with complex geological structures, there may still be limitations in the simulation of some specific geological phenomena (such as ore body morphology). Meanwhile, it is difficult to update the model in the later stage of cross-folded section modelling, which is basically equivalent to model reconstruction.	This method can easily solve the problem of corresponding geological boundaries between sections, improve the degree of automation of modelling, and better deal with complex geological structures.
	Based on geostatistical methods	Some geostatistical modelling algorithms, such as the Simpat algorithm, have low computational efficiency in practical simulation applications. The research object of multipoint geostatistics is the spatial multipoint combination information, which usually involves exponential problems, so the memory cost of the algorithm is an important consideration in the design of a multipoint modelling algorithm.	Considering the spatial correlation and the actual structure of geological bodies, the simulation results are more reasonable. The algorithm ideas are diversified. It is suitable for complex geologic body simulation. By introducing a "training image" and comprehensively considering the relationship between multiple points, it can fit the conditional data and better simulate the spatial morphology of complex geologic bodies.
	Based on Machine Learning	Demands high-quality, high-precision datasets for training purposes; entails intricate algorithms and models, necessitating expertise and proficiency for their application and refinement.	Reduced the need for manual intervention; enhanced efficiency mitigated the potential subjective errors inherent in conventional methods; and expedited the generation of three-dimensional geological models within a shorter timeframe, thereby conserving time and resource expenditures.

), each of which performs well in different modelling environments. Therefore, this section focuses on the challenges, recent progress, and future trajectory of the current mainstream modelling methods.

3.3.1. Geological Modelling Based on Profiles

Various geological profiles contain sufficient information. However, traditional two-dimensional representations map these data onto a flat plane. This can lead to inaccuracies and omission of certain details, hindering geologists from uncovering concealed geological information deep beneath the surface.

Currently, geologists employ geological profiles for modelling by generating profile diagrams on computers and subsequently constructing 3D geological bodies from these profiles, as illustrated in Figure 6. This technique simplifies the 3D challenge, thereby streamlining the modelling process. Therefore, this is the most straightforward and pragmatic method for addressing basic geological description tasks.

With the rapid development of geological modelling, notable breakthroughs in modelling theories have emerged. One study focused on the coupled modelling of a 50 m urban surface layer sequence [155]. This approach elucidates the properties and potential interconnections between layers, thereby reducing potential risks in geotechnical endeavours.

3.3.2. Geological Modelling Based on Drilling

Drilling data serve as a crucial component in geological modelling, not only as a calibration tool but also as a direct input for the construction of geological representations. For straightforward geological structures, these data are particularly effective, aiding in the observation of stratigraphic trends, 3D profile analysis, and data interpretation. However, in complex geological regions, sole reliance on drilling data can result in models that require refinement due to potential inaccuracies.

To tackle these modelling challenges, numerous methods have been explored. One such approach, automated modelling employing multipoint geostatistics of drilling data [156], significantly diminishes the complexity of manual adjustments, effectively capturing flat attitude layered geological bodies, and meeting the demands of engineering geological analysis. For more intricate geological formations, Zeng [157] introduced a recursive method leveraging the boundary representation approach to depict arbitrarily shaped geological entities. The proposed half-winged-edge-face structure topological framework provides a detailed description of the stratum model’s topological information for complex entities, adapting to intricate structures like terminations, overburden, and lens bodies. This approach not only resolves geological modelling issues related to intricate geometric and topological relations, but it also introduces a novel method and concept for 3D geological modelling. Additionally, techniques such as GA-Kriging interpolation [158] and machine learning-driven implicit approaches for drilling data [159] offer promising solutions for swift, precise, and automated modelling in such contexts.

3.3.3. Geological Modelling Based on Multisource Data

Due to the multifaceted nature of geological phenomena, models derived from a single data source frequently suffer from insufficient precision and reliability (Figure 7). Compared with other approaches, multisource data modelling can exhibit regional geological characteristics more accurately [160,161]. Although this method mitigates the limitations of single data sources, integrating multiple datasets presents considerable challenges. Furthermore, a comprehensive system for data fusion methods remains undeveloped, resulting in deficiencies in geological models based on multisource data.

To address the integration of data across varying scales, geological data fusion utilizes constraints and discrete smooth interpolation techniques within its data framework. Data preprocessing harmonizes and transforms the data from disparate scales to guarantee precise fusion and interpretation. Core data are reconfigured to align with well logging and seismic data, integrating geological data across scales using algorithms such as interpolation, inversion, and statistics. This process achieves a consistent spatial resolution and amalgamates multiscale data into a unified dataset.

Researchers have utilized various data sources to model basin structures, lithology, mineral deposits, and geological formations to facilitate mineral prediction and target area reduction [162,163]. Innovative geophysical modelling techniques that integrate and constrain geological data [164,165] have been proposed to effectively mitigate inversion ambiguity and improve the model quality and effectiveness of deep exploration. A novel 3D geological modelling method, termed the constraint-based modelling approach, integrates multisource data and prior geological knowledge [166]. This method refines human-computer interactions, simplifies intricate geological modelling, and enhances efficiency. It supports fundamental geological information and enables professionals to tackle challenges (Figure 8).

3.3.4. Geological Modelling Based on Artificial Intelligence

The core of artificial intelligence is machine learning, which enables computers to achieve a degree of intellectual capability. Deep learning methods for 3D geological modelling aim to create a learning dataset from existing control profiles or borehole data, partition the modelling region into grids, and employ deep learning techniques to infer the subsurface geological structures at each grid node. Using intricate geological structure data from each grid segment, a 3D geological model was established. The procedure for deep learning-based regional 3D geological automated modelling adheres to a specific workflow (Figure 9).

Artificial intelligence-based geological modelling, with deep learning as its foundation, predominantly employs generative adversarial networks (GANs). Introduced by Goodfellow in 2014 [167], GAN samples from the latent space of images to create entirely new images and is currently a preeminent generative artificial intelligence network. A GAN consists of a generator network (G) and a discriminator network (D) (Figure 10). The former aims to produce authentic image outputs, whereas the latter improves the ability to distinguish between real and fake images. The parameters of both networks were concurrently optimized using a loss function. As deep learning has progressed, researchers have introduced advanced neural networks. For example, a conditional GAN (CGAN) incorporates additional constraints on the foundational CGAN. Autoencoder conditional GAN (AECGAN) [90] integrates dimensionality reduction via autoencoder processing, substantially reducing the dimensionality of the training data and amplifying training efficiency. SinGAN can learn from and generate models rooted in a solitary natural image.

The application of machine learning in 3D mineral, oil, and gas prediction remains nascent. The rational construction of mineral prospecting models, automated extraction of mineralization information, and machine learning-driven automatic delineation of exploration target areas signify the future direction of this domain.

4. Discussion and Perspective

In the previous sections (Section 2.1, Section 3.1, and Section 3.2), we reviewed the development history of technology and models, summarized previous research results, and made speculations on future research trends based on these results. However, any advancement in technology or models cannot be achieved without in-depth analysis and resolution of existing problems. Therefore, in the following discussion section, we will further explore the key issues faced by the current model and discuss in detail the future research directions and strategies based on these issues.

4.1. Discussion about 3D Geological Modelling Software

Notably, 3D geological modelling spans multiple disciplines and can be applied in numerous fields, including mining, geology, environment, resource exploration, petroleum, and urban planning, offering versatility in numerous contexts. Despite global research progress, several challenges remain.

(1) Commercial 3D geological modelling software and research primarily focus on data model structure, visualization, and resource estimation [168,169]. Although the spatial morphology of geological objects is well represented [170,171], deficiency exists in 3D spatial analysis and the integration of multidimensional geoscientific data, limiting the efficacy of 3D geological models.

(2) The constraints required in the modelling process are relatively strict [172,173]. This can be an inherent limitation of a 3D model due to its complexity resulting from the transition from 2D to 3D space [174]. To construct a 3D model with high precision, it is essential to eliminate minor discrepancies such as "open lines" in 2D drawings.

(3) Geological modelling methods rely on historical professional data. As technology progresses and data volumes grow, manual analysis becomes increasingly challenging [175,176]. In the future, AI may supplement manual effort and boost efficiency [177,178,179,180].

4.2. Discussion about 3D Geological Modelling Technology

Despite the concrete results across numerous domains, we conducted comprehensive research, including all facets of the 3D geological modelling process, from data acquisition and processing to model construction, validation, and dissemination. Some of the challenges are as follows:

(1) Obtaining high-precision 3D geological model data from multiple sources presents a significant challenge [181]. Despite the availability of core samples at millimeter-to-centimeter scales and seismic data at meter-to-decameter scales, the high costs have inhibited model development. Emerging techniques, such as laser scanning and oblique drone photography, offer promise but are constrained by factors such as atmospheric conditions, occlusion effects, computational delays, and intricate postprocessing requirements [182,183].

(2) Improving the efficiency of processing large volumes of data is crucial. As previously described, the modelling process can yield datasets that reach terabyte scales [184,185]. Addressing the challenges of memory management and the computational hardware constraints caused by this vast volume of data remains a significant technical barrier.

(3) There are multiple interpretations and uncertainties of geological phenomena that are subject to interpretation and uncertainty [186,187,188,189]. Although the integration of multisource data can enhance model accuracy [190], it faces challenges related to data integration and transformation. Future efforts in multisource geological data fusion should focus on issues such as disparate data formats, fusion of heterogeneous information, and standardization.

(4) The overall modelling process should be further enhanced through intelligence and automation to alleviate manual intervention [191,192]. Although certain scholars have achieved automatic modelling of strata with clear boundaries and interlayer characteristics [193], the rapid modelling of intricate geological bodies requires further study [194].

(5) Topological–geological modelling algorithms should be developed to solve intricate stratigraphic challenges [156,157,195], including complex faults and cavern situations [196,197].

(6) Research on dynamic and real-time updates of geological models are required [198].

(7) A regional geological model should be developed that covers a wide range, which is essential for resource exploration and geological hazard assessment [199,200,201]. The absence of high-precision data compromises accuracy.

(8) Establishing shared service systems for 3D geological models remains inadequate [202]. Direct model transfers can result in inefficiencies, compromised user experience, and potential data security concerns.

4.3. Perspective on 3D Geological Modelling

Considering the aforementioned factors, we delineated the following predictions and directions for 3D geological modelling:

(1) Future geological modelling software should feature enhanced workflow, adaptability, and refined theories. However, addressing the key technical challenges in modelling methods, data fusion, and visualization requires interdisciplinary collaboration [203,204,205,206,207]. Improving R&D is pivotal for maximizing the potential of 3D geological modelling.

(2) Innovative modelling techniques should facilitate interdisciplinary collaboration, producing specialized software for applications including oil and gas exploration and structural analysis [208,209,210,211,212].

(3) Integrating geological modelling with multisource data, such as geophysics and drilling, is essential for cost-effectiveness and improved efficiency [213].

(4) There is an immediate need for process automation in geological modelling. The development of an automated geological modelling system [214,215], including data input, database creation, data preprocessing, method selection, and model optimization, remains crucial.

(5) AI-driven approaches in 3D geological modelling aim to transform the discipline by enabling rapid [216,217,218] and accurate modelling and analysis of actionable outcomes.

(6) Emerging technologies such as UAV oblique photography and laser scanning are promising [219,220,221,222,223], offering advantages including high-resolution imaging, comprehensive data collection, rapid processing, labor reduction, enhanced efficiency across terrains, and noninvasive operations.

(7) Creating a cooperative platform for geological data and model sharing can ensure the integrated visualization of diverse datasets [224,225,226], including combined models that reflect hydrogeology and groundwater pollution.

(8) The visual representation of geological models [227,228,229] allows users to focus on specific intervals and structures and present related data. Models should support any directional slicing, segmentation, and exploded views.

(9) Geological modelling will evolve from large to small scales, serving as foundational geological data such as maps with customizable model sections. With technological and data advancements, achieving global 3D modelling and realizing the "Glass Earth" concept is inevitable [230,231,232].

(10) Investigating implicit modelling techniques for deep and ultra-deep reservoirs remains a paramount avenue for future exploration and innovation [233,234,235,236].

5. Conclusions

We conducted a thorough analysis of the development, technical content, and prominent applications of 3D geological modelling over the past two decades, leading to the following conclusions:

Notably, 3D modelling has evolved from deterministic to intelligent methodologies, each presenting distinct advantages and constraints. The appropriate selection of data, techniques, and quality controls is critical to this process. Integrating multisource data substantially improves the model precision, with machine learning presenting significant promise. Addressing practical obstacles related to data integrity, interpretability, and uncertainty is vital for the broader acceptance of 3D models and positioning them as foundational data for exploration, environmental assessment, and analysis.

The integration of diverse data sources enhances the precisions of geological models. Automated geological modelling, with the potential for geological data processing promoted by artificial intelligence, signifies the future trajectory. Nevertheless, pivotal issues remain, particularly in enhancing methodologies for data fusion and managing geological modelling uncertainties. Additional challenges include data acquisition, integration, automation, and dynamic model revision.

Future research should focus on addressing these challenges, especially in refining modelling techniques for deep and ultra-deep reservoirs and deposits. Additionally, it is advisable to create an integrated platform for intelligent technology and large-scale geological data sharing to enhance collaboration and knowledge exchange within the discipline.

In conclusion, 3D geological modelling transcends mere techniques and embodies a mission of significant consequences. It is the cornerstone for decoding the Earth’s mysteries and acts as an indispensable instrument for advanced resource management, geological disaster forecasting, and environmental safeguarding.