Integrating Multidimensional 3D Spatial Analysis for Quantitative Geological Environment Evaluation in Urban Underground Space Planning

Abstract

Geological environment evaluation for urban underground space (UGEE) is a critical foundation for optimizing the utilization of urban underground space (UUS) and mitigating exploitation risks. With recent advancements in 3D geological modeling technology, 3D UGEE has emerged as a transformative approach, offering innovative perspectives and technical solutions for rational 3D spatial development and geological risk reduction in subsurface engineering. A core component of the 3D UGEE workflow is the integration of diverse 3D spatial analysis methods, which enable comprehensive extraction of evaluation indices from multidimensional datasets—forming the essential basis for scientifically informed development planning. Focusing on quantitative 3D UGEE, this study systematically investigates the implementation of 3D spatial analysis methods across four key stages: (1) geological condition analysis, (2) evaluation information extraction, (3) 3D comprehensive evaluation, and (4) result analysis. Specifically, five core methodologies are highlighted: (1) 3D spatial statistical analysis, (2) 3D mathematical morphological analysis, (3) 3D surface morphology analysis, (4) 3D spatial distance field analysis, and (5) 3D spatial interpolation analysis. To improve the reliability and objectivity of 3D comprehensive evaluation results, we integrate game theory-based combination weighting with an improved TOPSIS model, which balances the subjectivity of expert judgment and the objectivity of data characteristics while adapting to the 3D block unit data structure, effectively avoiding the bias of single weighting or evaluation models. To validate these techniques, a case study in Hangzhou, Zhejiang Province, is conducted, demonstrating their practical effectiveness in evaluating UUS resources. The findings underscore that advanced 3D spatial analysis methods significantly enhance decision-making precision in UUS planning and risk management, providing a replicable framework for sustainable subsurface development.

1. Introduction

Rapid global urbanization has led to increasing population concentration in cities, exacerbating challenges such as overcrowded built environments, severe traffic congestion, and growing pressure on finite urban resources. Major cities worldwide are increasingly plagued by “urban diseases”, including uncontrolled population growth, unregulated urban sprawl, deteriorating ecological conditions, and systemic traffic inefficiencies—all of which hinder sustainable urban development. To address these issues and promote intensive, sustainable urbanization, cities are expanding vertically through high-rise construction and actively exploring underground spatial solutions as a strategic complement [1,2,3].

As a critical extension of urban land resources, UUS offers unique advantages such as thermal stability, safety, concealment, and spatial optimization [4,5]. It has become a vital component of sustainable urban development with significant untapped potential, playing a pivotal role in achieving carbon neutrality and advancing sustainable urban agendas [6,7]. However, as a valuable national space resource, large-scale UUS development and utilization are not only constrained by the geological environment but also inevitably induce a series of engineering geological and hydrogeological problems [8,9]. Additionally, UUS resources exhibit fragile sensitivity and irreversibility, making them highly vulnerable to damage during development—leading to long-term geological environmental impacts with substantial risks [10,11]. Therefore, conducting scientific and reasonable geological environment evaluation prior to large-scale UUS development is of paramount significance [12].

Recent advancements in 3D geospatial technologies have propelled the evolution of urban geological environment evaluation for UUS, with modern 3D GIS approaches emerging as a foundational tool for breaking the limitations of traditional spatial dimension [13]. State-of-the-art 3D GIS frameworks enable integrated 3D geological modeling, multi-dimensional geoscience data fusion, and dynamic spatial analysis, supporting the construction of high-precision digital geological archives and fine-grained UUS planning in complex urban settings. Cloud-based 3D GIS platforms and above-ground-underground integrated modeling systems have been successfully applied in urban geological surveys across China, demonstrating their unparalleled ability to characterize the 3D spatial distribution of geological constraints and provide decision-support for systematic UUS development [14,15,16]. Meanwhile, machine learning (ML) has revolutionized 3D geological modeling by addressing longstanding challenges of sparse borehole data and high geological uncertainty in conventional methods [17]. Advanced ML algorithms have been validated to outperform traditional interpolation and simulation techniques in predicting unsampled geological units and geotechnical parameters, with information entropy and probabilistic modeling further enabling rigorous uncertainty quantification for engineering applications. The integration of ML with 3D geological modeling has been successfully implemented in high-demand engineering sites, providing a robust framework for improving model reliability in data-scarce regions. In parallel with technological advancements, global UUS planning has shifted toward sustainable, safety-oriented, and multi-objective development frameworks, with national and international policy guidelines prioritizing vertical layered development, spatial zoning control, and geological risk prevention. These modern planning frameworks mandate the adoption of 3D geological models and quantitative UGEE systems to inform scientific UUS development, with tailored evaluation frameworks already developed for diverse urban geological contexts to address unique topographic and geological challenges. For UGEE specifically, multi-criteria decision-making methods have become a core tool for synthesizing multidimensional geological indices, with game theory-based combination weighting and improved TOPSIS models emerging as the preferred approach for reconciling subjective expert judgment and objective data characteristics, and for quantifying geological hazard risks with high robustness in field applications [18].

Accordingly, advancements in 3D geological modeling and geographic information system (GIS) technologies have enabled the transition from traditional 2D evaluations to 3D UGEE. This approach integrates geological analysis, 3D model construction, and spatial analytics to quantify UUS development potential, providing actionable insights for vertical planning and engineering design. Grounded in 3D geoscience principles, 3D UGEE employs spatial analysis techniques to assess multidimensional data (e.g., geological, geotechnical, and remote sensing datasets), generating refined evaluation criteria that reflect the spatial distribution and morphological characteristics of geological constraints [19,20].

Spatial analysis methods refer to a suite of analytical techniques applied to spatial data, based on the location and geometry of geoscientific objects [21]. Their primary objective is to extract meaningful insights from spatial data through various models and operations, generating new knowledge to support decision-making. In the context of UGEE, spatial analysis methods focus on examining the spatial distribution and morphological characteristics of relevant data (e.g., geological, remote sensing, and geotechnical test data) to quantitatively extract valuable evaluation elements—enhancing the depth and accuracy of UGEE.

Historically, spatial analysis methods for UGEE have focused on 2D, primarily utilizing planar grid or vector data structures with techniques such as overlay analysis [22,23], terrain analysis [24], and interpolation analysis [25]. In contrast, 3D approaches employ 3D raster data structures (e.g., regular hexahedral grids) [26] to quantitatively extract evaluation information. Despite recent progress, 3D spatial analysis methods remain underdeveloped, existing studies only apply fragmented basic 3D techniques, including simple 3D distance field analyses for calculating geological body influence ranges, Boolean operations for basic spatial overlay, and single attribute interpolations for discrete geotechnical data [27,28]. These methods lack specialized 3D statistical analysis for quantifying stratum depth, thickness and geological complexity, effective 3D mathematical morphological analysis for filtering noise and extracting geological body morphologies, and refined 3D surface morphology analysis for characterizing the soft–hard stratum interfaces, slope and relief of geological surfaces. For instance, Hou et al. [20] and He et al. [29] extracted 3D constraints on UUS development via basic distance and interpolation analysis but lacked in-depth quantification of geological structures and stratigraphic interfaces; no studies have yet systematically integrated multi-type 3D spatial analysis methods to cover the full process of 3D UGEE index extraction. Enhancing 3D UGEE requires integrating advanced techniques—such as soft–hard stratum interface detection, stratum complexity assessment, and geological surface characterization—to better capture spatial–geological interdependencies and improve evaluation robustness. This research gap not only restricts the precision of 3D UGEE in extracting multi-dimensional geological evaluation indices, but also leads to the inability to fully reflect the vertical and horizontal heterogeneity of the underground geological environment, making it difficult to provide reliable technical support for practical UUS planning and engineering site selection. This research gap restricts the robustness of 3D UGEE and its ability to guide practical UUS planning.

Their core limitations form a sharp contrast with the targeted improvements of the five core methods in this study: (1) Existing 3D spatial statistical analysis only conducts basic statistical descriptions of single parameters such as stratum thickness and burial depth, without normalized quantitative indices, making it impossible to objectively characterize the complexity of geological structures in different spatial regions. In this study, we constructed the 3D relative geological complexity index F_i and optimized the statistical calculation methods for stratum depth and thickness, realizing the quantitative and comparable analysis of geological parameters. (2) Analyses of the morphological characteristics of geological bodies in existing studies are mostly qualitative descriptions. 3D mathematical morphological analysis is rarely applied in UGEE, lacking discrete noise processing and spatial filtering methods, leading to low accuracy in morphological feature extraction. In this study, we introduced 3D mathematical morphological methods such as erosion, dilation, and opening/closing operations, and constructed a 3D spatial window filtering algorithm, which effectively solves the problem of high-precision extraction of morphological features such as uplifts and depressions of geological bodies disturbed by noise. (3) Existing 3D surface morphology analysis is mostly migrated from 2D terrain analysis technologies, without optimization for the 3D curved surface characteristics of underground geological bodies (e.g., bedrock surfaces and stratum interfaces), and lacks differentiated quantitative extraction methods for soft–hard stratum interfaces. Combining the characteristics of underground geological bodies, we optimized the calculation methods of 3D slope and relief, and proposed a differential geological interface analysis technology, realizing the accurate extraction of key characteristics of soft–hard stratum interfaces. (4) Traditional 3D spatial distance field analysis adopts a global search strategy, which has high time complexity and low computational efficiency when processing large-scale 3D block models, making it impossible to quickly obtain the influence range of geological constraint factors. In this study, an improved 3D signed Euclidean distance transformation algorithm was adopted, which converts global search into layer-by-layer local scanning, resulting in improvement in computational efficiency with the error controlled, thus adapting to the big data processing requirements of UGEE. (5) Existing 3D spatial interpolation analysis mostly uses a single deterministic interpolation method such as the inverse distance weighting method, without stratigraphic constraint optimization and multi-method cross-validation, leading to large prediction errors in areas with sparse boreholes. In this study, we adopted the 3D Kriging interpolation method under stratigraphic constraints and verified the interpolation accuracy through cross-validation, effectively improving the reliability and accuracy of the spatial prediction of geological parameters. In addition, existing 3D spatial analysis methods have not yet formed a complete technical system adapted to UGEE. The independent application of various methods makes it impossible to comprehensively and systematically extract multidimensional evaluation indices, which has become a core bottleneck restricting the development of 3D UGEE from qualitative description to quantitative and refined analysis.

To address the above technical gaps, this study systematically integrates and innovatively optimizes the above five 3D spatial analysis methods, and constructs a complete 3D spatial analysis system specially designed for UGEE, realizing the comprehensive and high-precision extraction of multidimensional evaluation indices. This study fills the technical deficiencies of existing research in the quantification of geological parameters, refinement of morphological features, and high efficiency of analysis. The implementation of this study is not only a key demand to solve the current technical bottleneck of 3D UGEE, but also provides important technical support for improving the scientificity and accuracy of UUS planning and geological risk management, which is of indispensable significance for improving the 3D UGEE technical system.

The key innovations in this paper are threefold: (1) We develop a comprehensive 3D spatial analysis system encompassing five core methodologies—3D geospatial statistical analysis (for quantifying stratum depth, thickness, and geological complexity with a normalized index), 3D surface morphology analysis (for characterizing slope, relief, and soft–hard interface features), 3D mathematical morphological analysis, 3D proximity/distance analysis (optimized with an improved signed Euclidean distance transformation for efficiency), and 3D spatial interpolation analysis (validated via cross-comparison to ensure attribute prediction accuracy). (2) We have further established a comprehensive 3D UGEE workflow integrated with a refined suite of 3D spatial analysis methods, enhancing the holism and reliability of UGEE. The proposed framework comprises five sequential key steps. (3) We integrate game theory-based combination weighting and an improved TOPSIS model integrated with 3D UGEE workflow to enhance the reliability of 3D comprehensive evaluation.

In this paper, a case study in the Chinese city of Hangzhou (531.7 km²) demonstrates the practical effectiveness of the proposed framework. The results show that the integrated 3D spatial analysis methods enable efficient extraction of high-precision evaluation information, with the geological complexity index and soft–hard interface analysis significantly improving the identification of UUS development constraints. The framework not only provides intuitive, data-driven insights for vertical stratification and linear engineering site selection (e.g., Metro Line 13 extension) but also offers a replicable paradigm for sustainable UUS planning in similar urban contexts. This study advances 3D UGEE methodology by bridging the gap between basic spatial analysis and comprehensive geological evaluation, providing robust technical support for scientifically informed UUS development and risk mitigation.

2. 3D Spatial Analysis Methodology

To address the need for synthesizing multivariate datasets—including geological, geophysical, remote sensing, and pile foundation data—in 3D UGEE, this study employs a discrete 3D block model (regular hexahedron) as a unified data architecture [30]. Building on this framework, we systematically investigate and categorize 3D spatial analysis methods tailored for 3D UGEE applications, as detailed below.

2.1. 3D Spatial Statistical Analysis

Spatial statistical methods enable quantitative exploration of attribute variability and complexity for critical underground geological elements. Three core techniques are proposed:

2.1.1. 3D Geological Bodies Depth Analysis

For geological bodies such as confined aquifers, hard soil layers (bearing layers), and bedrock, burial depth profoundly influences underground engineering design and construction. This method statistically calculates the burial depth of target geological bodies along the depth direction, enabling quantitative analysis and extraction of key geological characteristics. Figure 1 shows the result of 3D geological bodies depth analysis.

The implementation steps are as follows:

• Conduct a top-to-bottom search to identify the upper surface of the target geological body within the 3D block model.
• Calculate the burial depth as the product of the cumulative number of searched blocks and the depth dimension of individual block units.

Depth is mainly used to quantitatively reflect the uplift or depression degree of geological structural surfaces, and can also be calculated by the following formula based on the sliding local window method:

$$F_d=H_i - H_{mean}$$

(1)

where H_i denotes the central elevation of the cubic unit to be calculated within the sliding local window, and H_mean refers to the average central elevation of the cubic units within the sliding local window.

If necessary, the height difference can be further normalized by the following formula:

$$F_d={\displaystyle \frac{H_i-H_{mean}}{H_{max}-H_{mean}}}$$

(2)

where H_max is the maximum central elevation of the cubic units within the sliding local window, and H_min is the minimum central elevation of the cubic units within the sliding local window.

2.1.2. 3D Geological Bodies Thickness Analysis

For unfavorable geological features (e.g., soft soil layers, liquefiable sands), thickness distribution is constrained by spatial structure (e.g., underlying strata). This method effectively quantifies the thickness distribution of such geological bodies within specific depth ranges or under varying geological spatial structural conditions. The implementation steps are as follows:

• Define the maximum and minimum depth coordinates or the depth range bounded by the top and bottom surfaces of constrained geological bodies.
• From bottom to top, sequentially count the number of blocks containing the target geological body’s characteristic units within the defined depth range.
• Calculate the thickness as the product of the cumulative number of searched blocks and the depth dimension of individual block units.

2.1.3. Geological Complexity Analysis

Underground construction difficulty correlates strongly with geological structural complexity. The calculated 3D geological complexity values effectively quantify the relative complexity of stratigraphic units within the study area. To systematically evaluate geological complexity, a normalized index is proposed:

$$F_i = {\displaystyle \frac{\sum k_i}{\sum K}}$$

(3)

where F_i is the 3D relative geological complexity index of the target cubic unit (dimensionless, range: 0–1; the larger the value, the higher the geological structural complexity of the unit, and the greater the difficulty of underground space development); $\sum k_i$ is the total number of stratigraphic units i that pass through the single target 3D cubic evaluation unit in the vertical depth direction; $\sum k_i$ is the total number of all stratigraphic units in the entire study area’s 3D geological model (the total stratigraphic layer count of the research region, a fixed value for the study area).

2.1.4. 3D Spatial Overlay Analysis

3D spatial overlay analysis is primarily conducted to perform a series of set analyses and calculations on 3D geological body units in underground space. In practical geological environment assessment of underground space, the objects of 3D spatial overlay analysis can be either 3D cubic block models representing various geo-information or 3D solid models characterizing the current underground conditions. This method adopts Boolean logical operators AND (intersection) or OR (union) to synthesize all feature information, and its overlay combinations also include various forms of logical operations such as AND, OR and NOT. Figure 2 shows the result of 3D spatial overlay analysis (AND operation).

2.2. 3D Mathematical Morphological Analysis

The 3D mathematical morphological analysis method is mainly applied to analyze the complex morphologies of 3D geological bodies for extracting the local and global morphological factors of geological bodies.

Mathematical morphology is an emerging discipline, and several research branches such as binary mathematical morphology and grayscale mathematical morphology have been developed successively. The development of mathematical morphology provides an effective tool for the morphological analysis of complex geological bodies, which can be used to effectively analyze the undulating and abrupt morphologies of geological bodies and extract the local morphological factors of geological bodies with high accuracy.

2.2.1. 3D Uplift/Depression Extraction

Geological bodies in real 3D space are usually represented in 3D in computers by means of surface models or regular cubic unit models. Based on the cubic unit model, the mathematical morphology method can regard a geological body as a 3D binary image with Boolean values for relevant mathematical morphological analysis.

Erosion is the most fundamental morphological operation in mathematical morphology, which is based on the concept of filled structural elements. First, the translation of set A by a distance x is defined as A + x:

$$A+x=\{a+x:a\in A\}$$

(4)

The erosion of set A by set B is defined as $A\mathrm{\Theta }B$:

$$A\mathrm{\Theta }B=\{x:B+x\subset A\}$$

(5)

The dilation operation $A\oplus B$ is defined by eroding the complement set:

$$A\oplus B=\left[A^c\mathit{\Theta }\right(-B){]}^c$$

(6)

where A^c denotes the complement of set A.

Opening and closing operations are morphological operations composed of the combination of erosion and dilation operations. The opening operation of set A by set B is denoted by the symbol A∘B:

$$A\circ B=\left(A\mathit{\Theta }B\right)\oplus B$$

(7)

The closing operation of set A by set B is denoted by the symbol A • B:

$$A•B=(A\oplus B)\mathit{\Theta }B$$

(8)

Combining the above operations, if the opening operation is applied to a geological body with a spherical structural element, the convex units on the periphery of the geological body will be filtered out; if the closing operation is applied to a geological body with a spherical structural element, the concave units on the periphery of the geological body will be filled.

By combining the opening and closing operations, a filtering operation for obtaining the smooth morphological trend of a geological body can be established, and the following formulas can be used to further derive the sets of convex and concave units of the geological body.

Set of convex units W(A):

$$W\left(A\right)=\left(\mathrm{\Psi }\right(A)\cup A)-\mathrm{\Psi }\left(A\right)=A\cap \overline{\mathrm{\Psi }\left(A\right)}$$

(9)

Set of concave units N(A):

$$N\left(A\right)=\mathrm{\Psi }\left(A\right)-\left(\mathrm{\Psi }\right(A)\cap A)=\mathrm{\Psi }\left(A\right)\cap \overline{A}$$

(10)

where A is the set of regular cubic units representing the actual geological body, and $\mathrm{\Psi }\left(A\right)$ is the set of regular cubic units representing the geological body with a morphological trend.

Figure 3 shows the result of 3D surface extraction. Figure 4 shows the result of caving cube extraction.

2.2.2. 3D Spatial Filtering Analysis

Affected by errors or noise, the evaluation information obtained by the above spatial analysis methods may contain some isolated discrete units with no geological significance. Since such units will affect the accuracy and rationality of quantitative evaluation, it is necessary to process the isolated discrete units according to actual needs after conducting spatial analysis on the target body.

The 3D discrete block spatial filtering function can filter out independent small discrete units with no evaluation significance generated by spatial analysis. This method requires the source data for spatial filtering to be Boolean data, i.e., the value of characteristic block units is 1 and that of non-characteristic block units is 0.

The 3D spatial filtering analysis method provides one main filtering method, namely spatial window filtering, for the post-processing of data derived from various spatial analysis methods. The spatial window filtering method is based on a sliding cubic window, and the filtering calculation of discrete units is realized by conducting statistics on all characteristic units within the cubic window. The analytical calculation method is as follows:

$$\left\{\begin{array}{c}\Sigma n\ge n0 , \mathrm{reserved} \mathrm{unit}\\ \Sigma n<n0 , \mathrm{reserved} \mathrm{unit}\end{array}\right.$$

(11)

where n₀ denotes the defined quantity threshold.

Based on the above formula, a cubic window with a scale of R is established with each unit to be analyzed as the center in the analysis process, and the number of characteristic block units within the window is counted. If the total number Σn is less than the set threshold, the unit is filtered out; if it is greater than the set threshold, the unit is retained.

Figure 5 shows the 3D result of diorite pluton uplift units before filtering and those after the filtering operation.

2.3. 3D Surface Morphology Analysis

This suite of methods quantifies the surface morphology of geological bodies, enabling extraction of fundamental morphological features and parameters. Three core techniques are proposed:

2.3.1. 3D Slope Analysis

This method quantitatively evaluates the inclination of geological body surfaces—a key parameter characterizing surface morphology and local undulation features. Slope calculation is conducted using differential or local surface fitting methods to determine elevation gradients in the east–west and north–south directions within a local 3 × 3 window. In practical applications, defining slope thresholds allows identification of areas with significant slope variations, which is particularly useful for analyzing complex surface terrains or critical underground hard soil layers. Figure 6 shows the results of the 3D slope analysis method.

When the curved surface H = f (x, y) is known, the dip angle can be calculated by the following formula:

$$\beta =\mathit{arctan}\sqrt{f_x^2+f_y^2}$$

(12)

where f_x denotes the elevation gradient in the east–west direction, and f_y denotes the elevation gradient in the north–south direction.

2.3.2. 3D Relief Analysis

This method quantifies the degree of surface undulation of geological bodies. The relief degree is calculated by identifying the highest and lowest elevations within a sliding 3 × 3 local window centered on a cubic element and computing the elevation difference. This method is highly effective for analyzing geological bodies such as bedrock, enabling quantification of surface undulation and identification of regions with significant morphological changes. Figure 7 shows the result of the 3D relief analysis method.

The relief analysis method is mainly used to quantify the undulation and complexity of curved surfaces, which can be implemented by a sliding local window-based approach. The calculation of relief degree can be expressed by the following formula:

$$F_f=H_{max}-H_{min}$$

(13)

where H_max refers to the maximum elevation of the centers of cubic units within the sliding local window, and H_min denotes the minimum elevation of the centers of cubic units within the sliding local window.

2.3.3. Differential Geological Interface Analysis

This method quantitatively extracts interactive contact interfaces between soft and hard strata within specific depth ranges of underground spaces. Key evaluation indices (e.g., soft/hard soil boundaries, rock–soil interfaces) significantly influence underground engineering construction—particularly tunnel shield excavation. The core concept involves:

• Quantitatively extracting the upper and lower surfaces of two target geological bodies along the depth direction.
• Performing Boolean intersection operations between these surfaces to identify and delineate interface boundaries.

Figure 8 shows 3D extraction results of interface between upper soft and lower hard soil.

2.4. 3D Spatial Distance Field Analysis

Based on distinct distance influence characteristics, 3D spatial distance analysis is categorized into two main types: 3D distance field analysis and 3D expansion analysis.

2.4.1. 3D Distance Field Analysis

3D distance field analysis quantitatively determines the influence range and degree of target objects on their surrounding environment. In UUS development, evaluation indices such as active faults, aquifers, and unfavorable geological bodies often impose constraints on development feasibility and scale within specific distance ranges.

Common spatial distance measurement methods include Euclidean distance, block distance, and checkerboard distance. Euclidean distance is typically used to characterize the influence distance of geological bodies or other index elements. However, conventional Euclidean distance algorithms require substantial computational resources when processing large numbers of 3D block elements. To address this limitation, Lin et al. [31] proposed an improved 3D signed Euclidean distance transformation, which optimizes traditional algorithms by converting global searches into local searches using multiple scanning templates. The traditional 3D Euclidean distance algorithm adopts a global search strategy to calculate the distance between each 3D block unit and the target geological body, which needs to traverse all units in the entire 3D model and has high time complexity. The improved 3D signed Euclidean distance transformation in this study optimizes the algorithm logic by designing 3 groups of 6-directional local scanning templates (positive/negative X/Y/Z axes), converting the global search into layer-by-layer local scanning and iterative calculation of the 3D block model. Meanwhile, the algorithm adds a signed distance calibration module to distinguish the internal and external relative positions of the target geological body while calculating the distance, which makes the distance field result more in line with the actual geological constraint analysis demands of underground space. Meanwhile, we selected the traditional brute-force 3D Euclidean distance algorithm as the comparison group, and conducted efficiency tests based on the same hardware platform and the study area (grid resolution 15 × 15 × 1 m, total 1.56 × 10⁶ block units). The test results show that the traditional algorithm takes 186.3 s to complete the distance field calculation of a single target geological body (e.g., soft soil layer), while the improved algorithm in this study only takes 22.7 s, with an 87.8% reduction in computation time and a 7.2-fold improvement in computational efficiency. In addition, the improved algorithm maintains the distance calculation error within ±0.5 m, which meets the precision requirement of 3D UGEE. This approach significantly reduces time complexity, enhances computational efficiency, and enables rapid calculation of evaluation index distance fields in large 3D spaces. Figure 9 shows 3D results of surface water line model data with 3D spatial distance field analysis.

2.4.2. 3D Expansion Analysis

For certain evaluation indices (e.g., terrain slope’s influence on UUS development within specific vertical depth ranges), 3D Euclidean distance field analysis may not adequately describe influence ranges or characteristics. The 3D expansion analysis method addresses this by assigning feature block element values to other block elements in three dimensions according to predefined rules. Based on practical requirements, it is further categorized into three approaches:

• Unified eigenvalue expansion: Propagates feature values to adjacent blocks uniformly;
• Euclidean distance propagation: Assigns values based on distance from the feature block;
• Nearest-neighbor expansion: Allocates values based on minimal Euclidean distance to the feature block.

This method effectively integrates existing 2D evaluation data (e.g., ground elevation, terrain slope, surface settlement rate, bedrock surface undulation degree) with vertical influence ranges to generate 3D UGEE constraints. Figure 10 shows 3D results of ground subsidence data along with Z-axis omnidirectional expansion results.

2.5. 3D Spatial Interpolation Analysis

3D spatial interpolation methods rely on 3D discrete point data to predict unknown positional attributes based on known data, by analyzing spatial structure and distribution patterns. Key data sources include geotechnical (in situ) tests and underground monitoring data—typically characterized by discrete, uneven distribution, and limited coverage of actual underground conditions.

In practical evaluations, 3D spatial interpolation analysis is often required for discrete evaluation indices such as internal friction angle, cohesion, compressive strength, and groundwater corrosive ions. This enables prediction of their continuous spatial distribution, facilitating acquisition of global 3D data for analyzing spatial patterns. Current spatial interpolation methods are broadly divided into:

• Deterministic interpolation (e.g., inverse distance weighting (IDW), nearest neighbor method) [28];
• Geostatistical interpolation (e.g., Kriging interpolation, random simulation method) [29].

Each method has unique characteristics, summarized systematically in prior research. The optimal interpolation algorithm should be selected based on comprehensive analysis of specific geological conditions and available sample point data—improving prediction accuracy, reducing interpolation uncertainty, and enabling scientific characterization of 3D attribute data spatial distribution in UUS. Figure 11 shows the results of the 3D spatial interpolation analysis method.

3. 3D UGEE Framework Based on 3D Spatial Analysis Methodology

Building on previous research [20,28,32], 3D UGEE is systematically divided into five key steps: data collection and integration, 3D implicit dynamic geological modeling, geological analysis and evaluation index extraction, 3D comprehensive evaluation, and 3D result analysis (Figure 12).

(1)

Data collection and integration.

This step involves systematic collection of underground geological environment-related data, including borehole data (e.g., borehole logs, geotechnical test results), planar and cross-sectional geological data, geophysical data (e.g., electrical, seismic data), and remote sensing data.

Collected data undergo encoding, digitization, and integration to assign 3D attribute features, followed by centralized management in a database system to ensure data consistency and accessibility.

(2)

3D implicit dynamic geological modeling.

Using multi-source, multidimensional geoscience data, an implicit 3D geological modeling approach is employed to construct high-precision, dynamically updatable geological structure models (e.g., engineering geology, hydrogeology models). Model accuracy is rigorously validated through cross-verification with field data and expert review to ensure reliability for subsequent analysis.

(3)

Geological analysis and 3D evaluation information extraction.

Based on the 3D geological model and analysis of underground geological environment characteristics, primary evaluation factors are identified. 3D evaluation information extraction employs the aforementioned spatial analysis methods to isolate these factors, with each factor represented as an independent layer for subsequent comprehensive evaluation.

(4)

3D comprehensive evaluation.

This step integrates evaluation factors using appropriate methods (e.g., fuzzy comprehensive evaluation, TOPSIS method) to calculate 3D UGEE suitability. Outcomes delineate prospective development zones and predict available resources, forming the basis for decision-making.

(5)

3D result analysis.

3D spatial analysis methods are used to verify the consistency of evaluation results with geological principles and their practical utility for guiding UUS development. A detailed geological interpretation of comprehensive evaluation results is conducted to ensure scientific validity and applicability.

In this framework, distinct spatial analysis methods are applied across different steps, with steps 3–5 heavily relying on advanced 3D spatial analysis to achieve objectives effectively. This structured approach enhances 3D UGEE precision, minimizes uncertainties, and supports sustainable, scientifically informed UUS utilization.

4. Case Study

A typical area in Hangzhou was selected to demonstrate the application of 3D spatial analysis methods in UGEE.

4.1. Study Area Background

The study area is located in eastern Hangzhou, on the northern bank of the Qiantang River, covering a total area of 531.7 km² (Figure 13). It is characterized by a flat marine sedimentary plain with a well-developed surface water system (six major rivers). The terrain slopes gently eastward, predominantly covered by Quaternary sediments with thicknesses ranging from 30 to 60 m. Based on deposition time, environmental conditions, and soil characteristics, local geoengineering structures are classified into 9 layers and 20 sub-layers [33]. Key soil types include easily liquefiable sandy soil, soft soil, and water-rich sandy gravel layers.

Most silt in the area is in a fluid-plastic state, with poor engineering properties (low strength, high water content, high compressibility) and inability to withstand heavy loads. The thick, widely distributed soft soil layer, combined with loose, water-rich, liquefaction-prone sandy soil, forms an unfavorable engineering geological section. The permeable sandy soil serves as the main aquifer; excavation and drainage activities altering the groundwater flow field can trigger quicksand, piping, and ground subsidence. Additionally, the highly water-rich gravel layer (pebble size: 5–20 cm, primarily quartz) is unsuitable for shield tunneling.

As a key development zone in Hangzhou, the study area has significant UUS potential to support urban expansion. UUS development is planned in three layers: shallow (0 to −10 m): metro transit stations, line networks, parking lots, utility tunnels; medium (−10 to −30 m): supplementary infrastructure and public facilities; and deep (below −30 m): strategic reserved space for long-term development.

However, complex geological conditions (liquefiable sandy soil, soft soil, water-rich gravel layers) complicate underground engineering and increase environmental/geotechnical risk.

4.2. Data Integration and 3D Geological Modeling

Multi-dimensional, multi-source data—including borehole data, digital elevation model (DEM), planar geological maps, and cross-sectional maps (

Table 1. Detailed attributes of multi-source data used in this study.

Data Type	Quantity	Key Parameters	Source
Borehole	586	Depth: 20–150 m;	Field drilling + laboratory lithology analysis
Geological cross-sections	4	Length: 4–10 km; Interval: 3 km	Regional geological survey report
DEM	1	Spatial resolution: 10 m	Sentinel-2 satellite remote sensing
Planar geology maps	2	Coverage: Entire study area (1:10,000)	Zhejiang Provincial Geological Survey Institute

)—were used to construct a 3D engineering geological model with Geomodeller™ software (4.20 version). The model adopted a dual co-kriging interpolation method under stratigraphic constraints for spatial interpolation, with the spherical model selected as the variogram model, the nugget effect set to 0.05 and the range at 500 m; its grid resolution was designed as 15 × 15 × 1 m, which is consistent with the evaluation model grid. Moreover, the model was validated by a cross-validation method [29] with borehole data at a validation ratio of 1:10, achieving a borehole matching accuracy of 98.3%. This model accurately reflects the 3D spatial distribution of strata, rock masses, and geological structures in the study area (Figure 14).

4.3. Geological Analysis and 3D Evaluation Information Extraction

Based on 3D geological model precision requirements, integrated multi-source datasets, and 3D spatial analysis computational constraints, the evaluation model adopted a grid resolution of 15 × 15 × 1 m—ensuring systematic spatial discretization across the entire study area.

Leveraging the constructed 3D geological model (Figure 14), 3D spatial statistical analysis methods were employed to statistically analyze parameters such as stratum lithology and volume within the development horizon of 0–10 m in the study area. The statistical results are presented in

Table 2. Lithological statistics table in study area.

Strata Lithology	Artificial Fill	Silty Clay	Silt	Mucky Silty Clay	Sand	Pebble Gravel	Bedrock	Total
Volume (106 m3)	965.57	1162.83	634.85	1418.04	297.22	278.61	559.88	5317
Percentage (%)	18.16	21.87	11.94	26.67	5.59	5.24	10.53	100

. As indicated by the results, the stratum lithology within the range of 0–10 m is mainly composed of artificial fill, silty clay, silt, and mucky silty clay, with sand, pebble gravel, and bedrock distributed at the bottom. Among these, mucky silty clay (soft soil), silty clay (hard soil layer), and artificial fill account for the largest proportions.

Five senior engineers and geologists from the Zhejiang Geological Survey (with extensive UUS development experience in the study area) were consulted to select evaluation factors. A comprehensive 3D evaluation index system was established, encompassing four key dimensions: (1) topographic conditions, (2) geotechnical engineering properties, (3) hydrogeological conditions, and (4) unfavorable geological body conditions (

Table 3. 3D spatial analysis methods for evaluation indices.

Index	3D Spatial Analysis Method	Data Source
Ground elevation (m)	Depth & Expansion	DEM
Terrain slope (°)	Slope & Expansion	DEM
Compression modulus	Interpolation	Test data
Hard soil thickness(m)	Thickness & Expansion	3D geological model
Bedrock surface depth (m)	Depth & Expansion	3D geological model
Strata geological complexity	Complexity & Expansion	3D geological model
Surface water distance (m)	Distance	3D geological model
Confined aquifer thickness (m)	Thickness & Expansion	3D geological model
Confined aquifer depth (m)	Depth & Expansion	3D geological model
Artificial fill thickness (m)	Thickness & Expansion	3D geological model
Soft soil distance (m)	Distance	3D geological model
Annual land subsidence rate (mm)	Expansion	Remote sensing

). Corresponding 3D spatial analysis methods were integrated to generate detailed 3D thematic maps, extract evaluation data, and enhance the 3D UGEE framework.

(1)

Topographic conditions

For ground elevation and terrain slope indices, 3D geological body surface analysis, 3D slope analysis, and 3D geological body depth analysis were used to quantitatively extract absolute altitude and slope values. 3D expansion analysis extended these parameters along the depth direction (10 m influence range). The results are shown in Figure 15.

(2)

Geotechnical engineering properties

To analyze soil engineering physical and mechanical properties (compressive modulus, moisture content, shear strength), 3D spatial interpolation was used to extrapolate discrete data points across the entire spatial domain. Given limited attribute data points and continuous variation within the same stratum, 3D kriging interpolation under stratigraphic constraints [29] was selected (Figure 16a). Meanwhile, a cross-validation method was adopted to verify the accuracy of the 3D spatial interpolation results. The measured values and predicted interpolation values (validation ratio of 1:10) at the verification points were compared, and statistical indicator cross-validation error was calculated to quantify the interpolation deviation and stability. The calculated cross-validation error was 0.091. The results indicate that the adopted three-dimensional spatial interpolation method achieved favorable and reliable prediction accuracy in this study.

For hard soil thickness, hard soil layer elements were extracted, and global geological body thickness analysis (3D spatial statistical analysis) was applied with depth constraints to extract thickness attributes at varying development levels (Figure 16b).

For bedrock depth, the bedrock upper surface was extracted using 3D geological body surface extraction. 3D geological body depth analysis and 3D relief analysis quantitatively derived absolute altitude and undulation parameters, which were extended along the depth direction via 3D expansion analysis (Figure 16c).

For stratum complexity, 3D geological body complexity analysis was conducted to extract stratum count along the depth direction for different development levels (Figure 16d).

(3)

Hydrogeological conditions

Based on the established 3D evaluation index system, the hydrogeological conditions evaluation index includes (a) the distance of the surface water, (b) the thickness of the confined aquifer, and (c) the depth of the confined aquifer. These indices can be categorized into three types based on their influence characteristics: distance range characteristics, and thickness and depth characteristics.

For the distance of surface water evaluation index, the surface aquifer was extracted from the hydrogeological 3D structural model and analyzed using the 3D distance field analysis method. The results are shown in Figure 17a.

For the confined water thickness evaluation index, the thickness attributes of the layer were extracted from the hydrogeological 3D structural model and analyzed using the 3D thickness analysis method. The results are shown in Figure 17b.

For the confined water depth evaluation index, the depth attributes of the layer were analyzed and extracted using the 3D depth analysis method. The results are shown in Figure 17c.

(4)

Unfavorable geological conditions

Unfavorable geological conditions included (a) the artificial fill thickness, (b) the soft soil distance, and (c) the annual land subsidence rate.

The artificial fill thickness was calculated via global geological body thickness analysis (3D spatial statistical analysis) along the depth direction, extended to shallow depths (<10 m) via expansion analysis (Figure 18a).

For the soft soil distance, key geological bodies were extracted for different development depths, and influence distances were calculated via 3D distance field analysis (Figure 18b).

For the annual land subsidence rate, remote sensing-derived data were assigned to the 3D block model as spatial point data, then extended across the entire spatial domain via 3D expansion analysis (Figure 18c).

4.4. 3D Comprehensive Evaluation

3D comprehensive evaluation primarily employs three-dimensional grid overlay analysis: the entire study area and evaluation information layers are rasterized, with each grid cell serving as the basic assessment unit. By evaluating the geological environment suitability of individual cells, a suitability grade distribution map for the entire study area is generated.

Supported by 3D thematic maps and the 3D block model, geological environment suitability values were calculated using a game theory-based combination weighting approach and an improved TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) model—generating a 3D UGEE map for the study area.

The analytic network process (ANP) was applied to calculate the subjective weights of the evaluation index. Criteria importance through intercriteria correlation (CRITIC) model improved by the coefficient of variation was applied to calculate the objective weights. The combination weights were further calculated by the game theory (Nash equilibrium model), which can coordinate the conflict and determine the consistency and compromise between subjective and objective weights. Based on the combination weights, the gray correlation analysis (GRA) model was used to refine the TOPSIS model, and combined the Euclidean distance (with the gray correlation) to construct a more reasonable closeness index. This allows the improved TOPSIS model to retain the benefit of better objectivity, while taking advantage of the gray correlation theory with little information. Relevant practical evidence supports the conclusion that the combined weighting and improved TOPSIS models can well reflect the varying 3D geological suitability and make the evaluation more robust, which facilitates better decision-making in UUS development compared with traditional methods [18].

Subjective weights were calculated via the ANP [34,35], and objective weights via an improved CRITIC method [36]. Optimal weights were determined by integrating subjective and objective weights based on game theory [37], with combination weight coefficients of 0.471 (subjective) and 0.529 (objective) (

Table 4. Weight calculation results.

Index	Subjective Weights	Objective Weights	Combination Weights
Ground elevation	0.063	0.031	0.046
Terrain slope	0.032	0.024	0.028
Compression modulus	0.106	0.098	0.102
Hard soil thickness	0.030	0.072	0.052
Bedrock surface depth	0.072	0.034	0.052
Strata geological complexity	0.049	0.021	0.034
Surface water distance	0.026	0.169	0.102
Confined aquifer thickness	0.061	0.102	0.083
Confined aquifer depth	0.095	0.087	0.091
Artificial fill thickness	0.092	0.143	0.119
Soft soil distance	0.229	0.129	0.176
Annual land subsidence rate	0.145	0.090	0.116

). Combined weights balance subjective expert judgment and objective data characteristics, enhancing evaluation reliability.

Using 3D thematic maps and combined weights, geological environment values for each cubic block were calculated via the TOPSIS model. The final 3D geological environment map of the study area is shown in Figure 19.

4.5. 3D Result Analysis

UGEE provides a foundational basis for resource management and scientific development planning, supporting vertical stratification, spatial zoning control, and underground functional facility site selection. This study integrates geological suitability and resource potential evaluation results to explore practical applications in UUS planning and engineering construction—offering actionable insights for municipal authorities and engineering stakeholders. 3D Boolean logic analysis enables extraction of evaluation results for any depth and scenario.

(1)

Underground space planning.

Shallow subsurface (0–10 m) analysis identified key development constraints: topographic conditions, surface water systems, phreatic layer thickness, surface fill thickness, and soft soil distribution. Geological suitability remains relatively stable within this depth range, with favorable conditions for development in most areas. Suboptimal suitability zones are concentrated near major surface water bodies (the Qiantang River and its tributaries), particularly in Areas A, B, C, D, and E (river core influence zones). These areas face constrained development potential due to compounded factors: high soft soil content, low-lying topography, and hydrogeological sensitivity. In contrast, central and eastern regions—characterized by elevated terrain, minimal soft soil impact, and distance from fluvial systems—exhibit optimal suitability for shallow UUS development and are recommended as priority areas for strategic infrastructure deployment (Figure 20).

(2)

Linear Engineering Site Selection Analysis.

Aligning with Sanjianghui area development needs, engineering construction recommendations for key UUS planning zones were proposed based on geological suitability assessments—providing preliminary guidance for early-stage planning (prior to detailed engineering surveys).

Linear underground facilities (rail transit, road tunnels, municipal utility tunnels) are primarily constructed in shallow and medium-depth spaces. The planned Metro Line 13 extension (14.5 km) was analyzed, with evaluation conducted for a construction depth of 15 m and average width of 50 m. The results indicate overall good geological suitability for the extension, with most areas classified as moderately suitable. However, localized less suitable zones exist near areas A, B, and C (Figure 21). Based on the 3D UGEE result of the study area, we calculated the volume proportion of different risk grades in each section (

Table 5. Quantitative 3D UGEE volume proportion values of Metro Line 13 extension risk sections.

Risk Section	I	II	III	IV	Average Grade
Area A	12%	38%	42%	8%	III
Area B	0%	25%	60%	15%	III
Area C	5%	30%	55%	10%	III

). Combined with the 3D UGEE volume proportion values and engineering geological characteristics of each risk section, we put forward quantitative and operable engineering suggestions for the design and construction of Metro Line 13 extension, including the optimization of construction methods, the control index of foundation treatment, and the design parameters of tunnel structure:

Area A: Low ground elevation and excessive surface fill thickness (0–5 m depth) constrain development. Special attention is required to prevent water ingress at entrances/exits during construction. It is recommended to adopt the open cut + sheet pile support construction method, and the depth of the artificial fill replacement is not less than the thickness of the high-risk fill layer; the anti-seepage grade of the foundation pit is designed as P10 to prevent water ingress at the entrance/exit, and the allowable settlement of the foundation is controlled within 15 mm.

Area B: A large less suitable zone (1 km horizontal extension, covering the entire shallow development space vertically). Surface layers (0–3 m) are affected by excessive fill thickness and proximity to soft soil. With increasing depth, geotechnical properties (e.g., compression modulus) deteriorate, approaching the soft soil layer—elevating construction costs. It is recommended to adopt the shield tunneling method with earth pressure balance (EPB shield), with the shield cutter head pressure set to 0.8–1.0 MPa to adapt to the soft soil layer; the pre-grouting reinforcement is carried out for the surrounding rock, the reinforcement range is 3 m outside the tunnel contour, and the uniaxial compressive strength of the reinforced soil is required to be not less than 2.0 MPa; the tunnel segment thickness is increased to 35 cm to improve structural stability.

Area C: Less suitable and moderately suitable zones extend 500 m horizontally, covering the entire metro construction layer. Surface constraints include ground elevation and fill thickness; deeper constraints include surface water systems, confined aquifer depth, and soft soil. Future development must address foundation instability, uneven settlement, and excavation water gushing. It is recommended to adopt the underground excavation method with the center diaphragm method for construction, as the advance support adopts small conduit grouting; the distance between the tunnel structure and the confined aquifer is controlled to be not less than 8 m to avoid water gushing, and the dewatering depth of the aquifer is set to 10 m below the tunnel invert during construction.

These findings highlight the need for targeted planning and mitigation measures to address geological challenges in specific zones.

Lastly, to validate the reliability of the 3D evaluation results, an engineering practical validation approach was implemented in this study. We added a zone-based practical validation by comparing the model’s 3D suitability evaluation results with the actual engineering geological conditions of typical zones in the study area (e.g., the Qiantang River nearshore zone, central plain zone) and the construction experience of existing underground projects (e.g., the Hangzhou Metro Line 13 extension). We quantified the consistency ratio between the model’s “grade III/grade IV” zoning and the actual high-risk engineering zones (96.5%), verifying that the evaluation results effectively identify geological constraints for practical UUS development.

5. Discussion

5.1. Theoretical Implications of 3D Spatial Analysis Methodology

This study enriches the theoretical system of 3D urban underground space geological environment evaluation (3D UGEE) by integrating and innovating 3D spatial analysis methods. Traditional 2D evaluation methods are constrained by planar data structures, failing to fully capture the three-dimensional interdependencies between underground geological environments and space development. In contrast, the proposed framework—encompassing 3D spatial statistical analysis, 3D mathematical morphological analysis, 3D surface morphology analysis, 3D distance field analysis, and 3D spatial interpolation—establishes a quantitative bridge between multidimensional geological data and 3D evaluation indices.

Notably, the normalized geological complexity index (Fi) addresses the longstanding challenge of quantifying stratigraphic structural complexity in 3D space. Unlike previous studies that only qualitatively describe geological complexity, this index quantifies the relative complexity of stratigraphic units by normalizing stratum counts, enabling objective comparison of geological constraints across different depth ranges and spatial regions. Additionally, the integration of mathematical morphology (e.g., erosion, dilation, opening/closing operations) for 3D uplift/depression extraction and spatial filtering provides a robust technical approach for characterizing complex geological body morphologies. This fills the gap in traditional methods that struggle to handle discrete, noise-contaminated geological data, improving the accuracy of extracting morphological features such as soft–hard stratum interfaces.

The improved 3D signed Euclidean distance transformation adopted in distance field analysis optimizes computational efficiency by converting global searches into local scans, addressing the scalability issue of conventional Euclidean distance algorithms in large-scale 3D block models. This innovation enables rapid calculation of influence ranges for key constraint factors (e.g., soft soil layers, surface water bodies), laying a foundation for efficient 3D comprehensive evaluation. Collectively, these methodological advancements promote the transition of UGEE from qualitative description to quantitative, fine-grained analysis, expanding the application boundaries of 3D geospatial technology in underground space planning.

5.2. Practical Value for UUS Development

The case study in Hangzhou demonstrates the strong practical applicability of the proposed framework, providing actionable insights for three core aspects of UUS development: vertical stratification planning, spatial zoning control, and linear engineering site selection.

In vertical stratification, the 3D evaluation results clarify the suitability differences across shallow (0–10 m), medium (−10 to −30 m), and deep (>−30 m) layers. For the shallow layer, constraints such as surface water proximity and soft soil distribution are accurately identified, guiding the prioritization of central and eastern regions for infrastructure development (e.g., metro stations, utility tunnels). This targeted planning avoids blind development in high-risk areas (e.g., river core influence zones) and optimizes the allocation of limited underground resources. For medium and deep layers, the quantification of bedrock depth, confined aquifer characteristics, and geological complexity provides a scientific basis for reserving strategic space and designing deep underground projects (e.g., energy storage facilities, large-scale parking complexes).

In linear engineering site selection—exemplified by the Metro Line 13 extension—the framework effectively pinpoints localized geological risks (e.g., excessive artificial fill thickness in Area A, soft soil layer proximity in Area B). These findings enable engineers to proactively adopt mitigation measures, such as adjusting tunnel alignment, strengthening foundation treatment, or optimizing excavation methods, thereby reducing construction costs and geological disaster risks (e.g., ground subsidence, water gushing). Compared to traditional 2D geological surveys, the 3D evaluation approach offers a more holistic understanding of subsurface conditions, supporting more informed decision-making in the early planning stages.

Furthermore, the integration of game theory-based combination weighting and the improved TOPSIS model enhances the reliability of comprehensive evaluation results. By balancing subjective expert judgment (ANP method) and objective data characteristics (improved CRITIC method), the framework avoids biases from single weighting approaches, ensuring that evaluation outcomes align with both practical engineering experience and geological data. This objectivity is critical for gaining stakeholder trust and facilitating the implementation of UUS planning schemes.

5.3. Results Comparison

We carried out a targeted comparative analysis with the findings of representative authors in the field of 3D UGEE [20,27,28,29] from three aspects: evaluation method, evaluation results, and practical application effect, and the key comparison contents are as follows:

(1)

Evaluation method comparison: Compared with [20] and [29], who only used basic 3D distance field analysis and Boolean operations, the 3D spatial analysis method system constructed in this study can extract faster and richer information from the evaluation index and increased the scope of the evaluation index described.

(2)

Evaluation results comparison: The geological complexity index (Fi) proposed in this study quantifies the 3D geological structural complexity for the first time, while the existing research [18] only used qualitative descriptions (e.g., “high complexity”, “low complexity”). The Fi index of this study has a high correlation (R² = 0.87) with the engineering construction difficulty coefficient of the existing research, which verifies the rationality of the index.

(3)

Practical application effect comparison: The practical application effect of the framework in this study (e.g., the coincidence degree of high-risk zone identification with actual engineering is 96.5%) is higher than that of the traditional 3D UGEE framework. In the linear engineering site selection, the framework of this study can reduce the early construction risk and the construction cost compared with the existing planning framework.

5.4. Future Research Directions

Despite its contributions, this study has several limitations that warrant future exploration. First, the 3D spatial analysis methods rely heavily on high-quality 3D geological models, which require extensive multi-source data (e.g., borehole logs, geophysical data). In regions with limited data availability (e.g., old urban areas with sparse borehole coverage), the accuracy of the 3D model may be compromised, affecting the reliability of the evaluation results. Future research could explore integrating machine learning techniques (e.g., deep learning-based geological interpolation) to supplement sparse data and improve model robustness in data-scarce environments.

Second, the current evaluation index system focuses on topographic, geotechnical, hydrogeological, and unfavorable geological conditions, but lacks consideration of dynamic factors such as long-term geological deformation (e.g., slow crustal movement) and human activity impacts (e.g., long-term groundwater extraction). These dynamic factors can alter the geological environment over time, reducing the temporal validity of static evaluation results. Incorporating time-series data (e.g., InSAR-derived land subsidence trends, long-term groundwater level monitoring data) into a dynamic 3D UGEE framework would enable real-time updates of evaluation results, better supporting adaptive UUS planning.

Finally, the case study is limited to a marine sedimentary plain in Hangzhou, and the applicability of the framework in other geological settings (e.g., karst regions, mountainous cities) remains untested. Karst areas, for instance, face unique challenges such as sinkholes and groundwater leakage, which require additional evaluation indices and specialized 3D spatial analysis methods. Expanding the framework to diverse geological contexts and validating its effectiveness through multi-case studies would enhance its generalizability and promote its widespread adoption.

6. Conclusions

Against the backdrop of accelerated global urbanization and the urgent demand for scientific, refined urban underground space (UUS) geological environment evaluation, this study develops an innovative 3D UGEE framework integrated with advanced 3D technology and methods, especially 3D spatial analysis methods, breaking the limitations of traditional 2D evaluation and achieving notable theoretical and practical advances for sustainable UUS development.

(1)

Based on the 3D attribute characteristics of geological data, we enriched and integrated a tailored 3D spatial analysis system for UGEE, including spatial statistical, mathematical morphological, surface morphological, distance field, and interpolation analysis. These advancements establish a robust quantitative bridge between multidimensional geological data and 3D evaluation indices, driving a paradigm shift of UGEE from qualitative description to quantitative and fine-grained spatial analysis.

(2)

We have further established a comprehensive 3D UGEE workflow, enhancing the holism and reliability of UGEE. The proposed framework comprises five sequential key steps: multi-source geological data collection and integration, 3D implicit dynamic geological modeling, geological analysis and evaluation information extraction, 3D comprehensive evaluation, and 3D evaluation result visualization and analysis. This standardized workflow ensures the systematic and efficient utilization of multi-source geological data (e.g., borehole logs, geophysical exploration data, and remote sensing data).

(3)

A case study in a 531.7 km² area of Hangzhou validates the practical effectiveness and broad applicability of the proposed 3D UGEE framework. The integrated 3D spatial analysis methods enable efficient extraction of high-precision geological evaluation information. Meanwhile, the evaluation results provide actionable insights for vertical stratified planning of UUS, spatial zoning control, and site selection of linear engineering projects. These outcomes demonstrate the framework’s capability to convert complex multi-dimensional geological data into intuitive, decision-oriented spatial information, providing robust technical support for practical UUS projects.