Resilience-Based Anomaly Detection and Risk Assessment for Groundwater Systems During Tunnel Excavation

Abstract

Tunnel construction disturbs groundwater systems, threatening water resource sustainability and ecological stability, while insufficient drainage can pose serious safety risks. Balancing safety with groundwater protection is therefore a critical challenge. Using the Huafu and Huayan tunnels as case studies, this research integrates field observations with MODFLOW simulations to develop a resilience indicator system. The system incorporates stability, recovery capacity, and critical thresholds to quantify groundwater responses to construction disturbances. A mechanism for detecting groundwater anomalies during tunneling is also proposed. Results show that extreme disturbances cause sharp declines in system performance, but subsequent recovery and resilience gains reflect self-organizing and self-healing properties of groundwater systems. This study demonstrates a resilience-based framework for evaluating tunnel impacts on groundwater. The findings provide early warning insights for water-related hazards and guidance for groundwater protection and restoration in tunnel engineering.

1. Introduction

With the advancement of national urbanization and rapid economic development, large-scale tunnel construction has become an inevitable engineering task. Traditional tunnel construction has primarily focused on drainage, which has disrupted the flow paths, recharge patterns, and hydrological characteristics of groundwater. This disruption has triggered a range of secondary geotechnical hazards, such as damage to the surrounding ecological environment [1,2,3], ground fissures, and surface subsidence [4,5,6]. These issues have severely affected both the local ecosystems and the daily lives of nearby residents. Clearly, tunnel engineering and the water environment are not independent but interact closely. On one hand, the groundwater environment influences tunnel engineering; on the other hand, tunnel construction exerts a feedback effect on groundwater systems. The interaction chain is illustrated in Figure 1. Due to the complex coupling between tunnel engineering, the surrounding water environment, and geotechnical systems—and the limited research on this topic—it is crucial to consider tunnel projects and the regional geological environment as an integrated system. Investigating the spatiotemporal evolution of groundwater systems and evaluating their resilience levels are key to addressing the scientific challenges related to water environment degradation and secondary geotechnical hazards.

The evolution of groundwater systems under tunnel excavation disturbances is highly complex. It not only involves the unsteady characteristics of groundwater dynamics [7,8,9,10] but is also closely related to the mechanical response of the geotechnical medium [11,12,13], resulting in multi-field coupling effects among stress fields, seepage fields, and even temperature fields. To address these issues, scholars at home and abroad have conducted extensive research. The main approaches include numerical simulation, mathematical theoretical analysis, and comprehensive evaluation based on field monitoring data. In numerical simulation, researchers commonly use finite element methods [14,15,16] and finite difference methods [16,17,18,19] to develop three-dimensional groundwater–geotechnical coupling models, aiming to reveal the mechanisms of groundwater system evolution under excavation disturbance. Mathematical theoretical analysis simplifies the physical processes to build analytical or semi-analytical models for exploring seepage characteristics and pressure responses [20,21,22,23,24]. In addition, field monitoring data [25,26,27] are often used to validate model predictions and support the analysis of groundwater dynamics in actual projects. In summary, traditional methods are more often applied for retrospective analysis and post-event evaluation. They are insufficient for meeting the modern demands of proactive monitoring and intelligent early warning. Numerical simulation also has limitations related to initial conditions, boundary settings, and parameter selection, requiring substantial prior knowledge and post hoc data. This makes it difficult to achieve real-time assessment and dynamic response during early or critical stages of construction. Given the hidden and nonlinear nature of groundwater systems, introducing resilience theory provides a promising way to reflect the system’s capacity to resist and recover from excavation-induced disturbances.

Resilience, often defined as the ability of a system to recover to its original state after external disturbances, reflects its capacity to rebound [28,29,30,31]. The concept of resilience has been widely applied in fields such as ecology [32,33,34,35] and economics [36,37,38,39,40]. In recent years, it has also been extended to tunnel engineering, particularly in seismic resistance [41,42] and disaster risk assessment [43,44,45,46,47]. In the realm of hydrogeology, groundwater resilience has emerged as a pivotal metric for sustainable resource management under the dual pressures of anthropogenic interference and environmental fluctuations. Current scholarly discourse primarily unfolds across two methodological frontiers: (1) Temporal Dynamics and Recovery Metrics, which characterize the system’s intrinsic self-repair capacity by quantifying the inverse of drought duration [48], the lag time required to regain equilibrium [49], or post-depletion recovery percentages [50]. (2) Standardized Indices and Predictive Frameworks, which mitigate data scarcity through multi-dimensional indicators like the Groundwater Resilience Index [51], often integrated with advanced non-linear architectures such as Artificial Neural Networks [52] and Hilbert–Huang Transforms [53] to forecast discharge stability. Despite progress in macro-scale and statistical research, applying these methods to high-intensity transient disturbances like tunnel excavation remains a challenge. As noted in a recent review [54], current groundwater resilience assessments have limitations in dealing with rapidly evolving engineering environments. Traditional deterministic simulations often fail to track real-time, non-steady-state changes, leading to predictive errors. To address this, this study proposes a holistic resilience framework. By quantifying resistance, recoverability, and overall resilience, this approach offers a more robust and scientific way to evaluate the dynamic equilibrium and environmental recovery of groundwater systems during construction.

To accurately quantify the spatio-temporal evolution of the groundwater system under dual-tunnel construction in the Zhongliangshan area, this study employs MODFLOW as the core numerical simulation platform. As an internationally recognized mainstream computational tool in the field of hydrogeological research, MODFLOW integrates various advanced finite-difference algorithms, enabling the effective characterization of complex hydrogeological systems [55,56,57,58,59,60,61]. Its sophisticated framework excels in simulating highly heterogeneous subsurface media and transient boundary conditions, which is crucial for capturing the dynamic responses of aquifers under engineering disturbances. Building upon these simulation results, a resilience index system for tunnel-related groundwater environments was developed, and a quantitative evaluation method was proposed to reveal the resilience evolution patterns under excavation-induced disturbances. These findings provide a scientific theoretical foundation for assessing groundwater impacts in similar engineering projects and lay the groundwork for the future sustainable management and protection of groundwater resources.

The remainder of this paper is organized as follows: Section 2 describes the tunnel and its hydrogeological background. Section 3 develops the hydrogeological model. Section 4 presents the construction of resilience indicators. Section 5 uses engineering case to analyze the resilience of the groundwater system and the anomaly warning model. Section 6 provides conclusions and discussion.

2. Study Area

2.1. Project Overview

The study area (${106}^{\circ }{34}^{\prime }$∼${106}^{\circ }{46}^{\prime }$ E, ${29}^{\circ }{34}^{\prime }$∼${29}^{\circ }{47}^{\prime }$ N) is located within the Zhongliang Mountain region in southwestern Chongqing, China (Figure 2). Stretching approximately 100 km from north to south, they, together with the Tongluo Mountains, form a natural barrier that restricts east–west access between Chongqing’s central urban area and external regions. The elevation ranges from 480∼640 m a.s.l. The area is situated in a subtropical monsoon climate zone, with an average annual rainfall of 1213.50 mm. The average annual temperature is approximately 18.6 °C, and the average relative humidity is around 80%. Since 1990, driven by national strategies such as the “The Belt and Road Initiative” and the “The development of the western region in China”, a total of 14 tunnels have been completed, are under construction, or are planned to traverse the mountain range from east to west. These tunnels have become key transportation corridors for the development of the Sichuan–Chongqing region. However, they have also caused significant damage to the local groundwater environment. This has led to a series of serious ecological and geotechnical secondary disasters, including farmland drought, land subsidence, and ground collapse. According to incomplete statistics, tunnel drainage in the Chongqing region has caused the drying up of 363 surface wells and springs, leakage in 42 reservoirs (above type II small reservoirs), and interruption of flow in 16 streams. The area of paddy fields affected by drought exceeds 33.33 km², and more than 300 ground subsidence events have been triggered, as shown in Figure 3.

Huafu Tunnel and Huayan Tunnel are located in the Zhongliang Mountain range, separated by 4850 m and situated within the same hydrogeological unit. The basic information on the construction of the two tunnels and their impacts on the groundwater environment are listed in

Table 1. Basic information of Huafu tunnel and Huayan tunnel.

Parameter	Huafu Tunnel	Huayan Tunnel
Total Length	≈3562 m	≈4800 m
Maximum burial depth	≈247 m	≈348 m
Tunnel Configuration	Twin tunnels, 4 lanes	Twin tunnels, 6 lanes
Width × Clearance Height	≈10.5 m × 6.76 m	≈13.5 m × 9.0 m
Construction Start Date	October 2003	June 2013
Completion Date	March 2024	June 2017
Construction Duration	5 months	48 months
Initial Groundwater Level	354 m	387 m
Affected groundwater area	1100 m	4400 m

.

2.2. Hydrogeological Condition

Based on spatial distribution, hydrodynamic characteristics, and lithological properties, the main tunnel area of Zhongliang Mountain is divided into two aquifers and three aquitards. Detailed information is provided in

Table 2. Stratigraphic Composition and Hydrogeological Characteristics.

Main Strata	Stratigraphic Composition	Characteristics
Carbonate rocks (T2l + T1j + T1f + P2c)	Limestone, Dolomitic limestone, Dolomite	Aquifer
Clastic rocks (T3xj)	Thick-bedded sandstone	Aquifer
Jurassic formations	Purplish-red mudstone, Interbedded thin sandstone layers	Aquitard
Xujiahe Formation (T3xj)	Shale, Carbonaceous shale, Mudstone	Aquitard
Feixianguan Formation (T1f)	Mudstone, Interbedded argillaceous limestone, Calcareous shale	Aquitard

.

Based on the investigation results, key hydrogeological parameters such as groundwater level, permeability coefficient, porosity, compression coefficient, and hydraulic conductivity were selected. Specifically, the hydraulic conductivity is characterized by three components: $K_x$ (m/d), $K_y$ (m/d), and $K_z$ (m/d) to account for spatial anisotropy. Furthermore, the porosity for each stratum includes both effective porosity and total porosity, aiming to provide a comprehensive characterization of the media. A three-dimensional hydrogeological model of the study area was developed using Visual MODFLOW flex 6.1. The model parameters for each stratum were then calibrated using field measurement data and pumping/injection test results. Detailed data are shown in

Table 3. Hydrogeological Parameters of Model Formations.

Formation	${\mathit{K}}_{\mathit{x}}$ (m/d)	${\mathit{K}}_{\mathit{y}}$ (m/d)	${\mathit{K}}_{\mathit{z}}$ (m/d)	${\mathit{S}}_{\mathit{s}}$	${\mathit{S}}_{\mathit{y}}$	Effective Porosity	Total Porosity
J	0.0193	0.0171	0.0115	0.0004	0.015	0.031	0.034
T3xj	0.072	0.052	0.072	0.0011	0.023	0.048	0.051
T1j, T2l	0.255	1.181	1.178	0.0023	0.141	0.193	0.196
T1f2, T1f4	0.0211	0.0211	0.0211	0.0002	0.011	0.022	0.025
T1f1, T1f3	0.0193	0.0171	0.0115	0.0021	0.107	0.114	0.117
P2c	0.261	0.337	0.263	0.0023	0.136	0.177	0.18

.

The refined stratigraphic data in Table 3 improved the spatial and temporal resolution of the hydrogeological model, enhancing its accuracy and detail representation. This allows for a more precise assessment of the ecological impacts caused by the excavation of the twin tunnels in Zhongliang Mountain. The three-dimensional hydrogeological model of Zhongliang Mountain is shown in Figure 4.

3. Hydrogeological Model

3.1. Mathematical Model

This study selects Huafu Tunnel and Huayan Tunnel in Zhongliang Mountain, Chongqing, China, as the research objects. The groundwater system in the tunnel area exhibits characteristics of unsteady flow in a heterogeneous anisotropic porous medium. The three-dimensional finite difference mathematical model of the groundwater system can be expressed by the partial differential equation in Equation (1) [7]:

$${\displaystyle \frac{\mathsf{\partial }}{\mathsf{\partial }x}}\left(k_x{\displaystyle \frac{\mathsf{\partial }h}{\mathsf{\partial }x}}\right)+{\displaystyle \frac{\mathsf{\partial }}{\mathsf{\partial }y}}\left(k_y{\displaystyle \frac{\mathsf{\partial }h}{\mathsf{\partial }y}}\right)+{\displaystyle \frac{\mathsf{\partial }}{\mathsf{\partial }z}}\left(k_z{\displaystyle \frac{\mathsf{\partial }h}{\mathsf{\partial }z}}\right)+w=s_s{\displaystyle \frac{\mathsf{\partial }h}{\mathsf{\partial }t}}$$

(1)

where $k_x$, $k_y$, and $k_z$ are the permeability coefficients in the x, y, and z directions; h is the groundwater head; w is the volumetric flux; $S_s$ is the specific storage coefficient; and t represents time. To realistically describe the mathematical model of the groundwater system, initial conditions and boundary conditions are introduced.

$${\left.h(x,y,t)\right|}_{t=0}=h_0(x,y)\&(x,y)\in D$$

(2)

$$h(x,y,z,t)\Gamma 1=h_0(x,y,z,t),\&(x,y,z)\in {\Gamma }_1,t\ge 0$$

(3)

$$\left.k{\displaystyle \frac{\mathsf{\partial }h}{\mathsf{\partial }t}}\right|\Gamma 2=q(\mathrm{x},\mathrm{y},\mathrm{z},\mathrm{t}),\&(x,y,z)\in {\Gamma }_2,t>0$$

(4)

where ${\Gamma }_1$ is the boundary with a fixed hydraulic head, and ${\Gamma }_2$ is the boundary with a fixed flow rate; q denotes the flow rate. Equation (2) represents the initial condition (groundwater head distribution at the initial time), Equation (3) is the Dirichlet boundary condition (fixed head boundary), and Equation (4) is the Neumann boundary condition (specified flow boundary). Combining the above, the finite difference formula for calculating groundwater seepage is obtained:

$$\sum _{n\in N\left(i\right)}{\mathrm{C}}_n\left(h_n^m-h_i^m\right)P_i{h}_i^m+Q_i=S_{s,i}{V}_i{\displaystyle \frac{h_i^m-h_i^{m-1}}{\Delta t}}$$

(5)

where $N\left(i\right)$ denotes all cells adjacent to cell i, $C_n$ is the hydraulic conductance between adjacent cells, $P_i$ and $Q_i$ are external source coefficients, $S_{s,i}$ is the specific storage coefficient, and $V_i$ is the cell volume. Detailed stratigraphic data are listed in Table 2. Using Equation (5), the hydraulic head distribution at each time step can be calculated iteratively.

3.2. Model Calibration

In the hydrogeological model (Figure 4), the boundary conditions are defined based on the regional geological and hydraulic characteristics. The bottom boundary of the domain is set as a no-flow boundary, as the underlying strata exhibit low permeability and are located far from the tunnel excavation zone, making them minimally affected by construction activities. The top boundary is designated as an open boundary to account for groundwater recharge from atmospheric precipitation. For the lateral boundaries, the eastern and western edges are characterized as no-flow boundaries due to the low permeability of the strata. The northern boundary, serving as the watershed for both surface and groundwater, is defined as a Type II no-flow boundary (Neumann boundary). In contrast, the southern boundary, situated at the Yangtze River, is generalized as a constant-head boundary (Dirichlet boundary), as it remains relatively stable and is geographically distant from the Huayan and Huafu tunnels. As illustrated in Figure 7, the drawdown cone induced by tunnel excavation does not reach the distal no-flow boundaries, or its impact is negligible. This confirms that the application of approximate flow boundaries has a minimal effect on the simulated flow field and the overall modeling accuracy. The model is discretized using square grids with a side length of 100 m, resulting in 73 × 127 × 7 cells.

4. Construction of Groundwater System Resilience Indicators

4.1. Resilience Conceptual Model

To address the complex coupling of the dual groundwater systems during tunnel excavation, a resilience conceptual model of the tunnel groundwater system under excavation disturbance is first established to clarify the system’s resistance (robustness) and recovery capacity. Based on the actual construction requirements, groundwater drawdown is permitted within a reasonable range, and dynamic hierarchical monitoring of groundwater system resilience is conducted at varying distances from the disturbance center.

Figure 5 presents a conceptual model of the impact and recovery process of groundwater levels during extreme tunnel excavation events (drainage events).

$t_0<t<t_d$, the initial stable stage of the groundwater system, during which the system has not yet been disturbed by external factors. At time $t_d$, dewatering begins to affect the system.

$t_d<t<t_p$, the groundwater system is disturbed. In the early stage of tunnel excavation, dewatering is mainly used to ensure construction safety, resulting in a discharge rate far exceeding the natural recharge. Consequently, the groundwater level drops rapidly and reaches its lowest point at time $t_p$.

$t_p<t<t_r$, the system enters the recovery stage. As grouting is implemented during construction, the dewatering rate decreases significantly, and the groundwater level gradually recovers.

$t\ge t_r$, the groundwater system is considered to be basically recovered. Based on the calculated water level results, the deviation from the initial water level is less than 0.5%. Therefore, in this study, the recovered water level is assumed to be equal to the initial water level.

Taking Huayan Tunnel as an example, a sensitivity analysis was conducted on three main influencing variables—average annual precipitation, tunnel length ratio, composite hydraulic coefficient, and their effects on groundwater levels, as shown in Figure 6. The results indicate that the impact ranking on groundwater levels is: composite hydraulic coefficient > average annual precipitation > tunnel length ratio. Since the composite hydraulic coefficient is determined by the physical properties of the rock strata and is difficult to change in practice, this study focuses on the time-varying impact of average annual precipitation on groundwater levels. The average annual precipitation in Zhongliang Mountain is approximately 1213.50 mm. Numerical simulations were performed for Huafu and Huayan Tunnels under three different precipitation scenarios—800 mm, 1200 mm, and 1600 mm—to clarify the influence of average annual precipitation on the resilience of the groundwater system.

In Figure 7, it can be observed that the three average annual rainfalls significantly affect the shock resistance and recoverability of the groundwater level in the Huafu and Huayan tunnels, which is consistent with the generalised features in Figure 5. However, due to the discrete and non-linear characteristics of the data, it is difficult to efficiently realise the monitoring and early warning assessment of groundwater levels by multi-point high-frequency field measurement methods such as the traditional well drilling method. This makes it difficult to clarify the influence boundaries of the depth of groundwater decline and to realise a dynamic recovery strategy that focuses on strong recovery in high water loss zones and weak recovery in low water loss zones. Therefore, resilience indicators need to be introduced to carry out quantitative analyses.

4.2. Construction of Resilience Indicators

We refer to ref [37] regarding evaluating system resilience. Introducing robustness R to represent the concept of the lowest threshold, the ability of the groundwater system to maintain its fundamental stability under disturbances and the maximum extent of impact it can withstand before causing irreversible environmental damage. The rate of change $RO{C}_{DS}$ is used to quantify the rate at which the system’s performance declines when the groundwater is subjected to disturbances.

$$R=min\left\{V\left(t\right)\right\},t_d<t<t_{\mathrm{r}}$$

(6)

$$RO{C}_{DS}={\displaystyle \frac{V\left(t\right)-V\left(t-\Delta t\right)}{\Delta t}},t<t_{\mathrm{p}}$$

(7)

where $V\left(t\right)$ represents the discrete function of groundwater level varying with time; $t_p$ denotes the time when groundwater recovers to its initial stable state; and $\Delta t$ is the time increment. The system resilience loss, $LON{E}_{DS}$, is used to measure the system’s capacity to withstand damage during the disturbance phase. $LON{E}_{DS}$ can be quantified as the integral area under the discrete curve $V\left(t\right)$ from time $t_d$ to $t_p$:

$$LON{E}_{\mathrm{DS}}={\int }_{t_d}^{t_{\mathrm{p}}}\mathrm{V}\left(\mathrm{t}\right)\mathrm{dt}$$

(8)

To accurately quantify the resilience of different groundwater systems, the unit-time resilience loss, $UTLON{E}_{DS}$, is introduced during the disturbance phase as a risk measure of system resilience to evaluate system stability:

$$UTLON{E}_{\mathrm{DS}}={\displaystyle \frac{LON{E}_{\mathrm{DS}}}{\left(t_p-t_d\right)}}$$

(9)

Depending on the disturbance phase, the recovery phase introduces the system performance change rate $RO{C}_{RS}$, resilience loss $LON{E}_{RS}$, and unit-time resilience loss $UTLON{E}_{RS}$ to evaluate the recovery and adjustment capacity of the system:

$$RO{C}_{\mathrm{RS}}={\displaystyle \frac{V\left(t\right)-V\left(t-\Delta t\right)}{\Delta t}},t_{\mathrm{p}}<t<t_{\mathrm{r}}$$

(10)

$$LON{E}_{\mathrm{RS}}={\int }_{t_p}^{t_r}\mathrm{V}\left(\mathrm{t}\right)\mathrm{dt}$$

(11)

$$UTLON{E}_{\mathrm{RS}}={\displaystyle \frac{{\mathrm{LONE}}_{\mathrm{DS}}}{\left(t_{\mathrm{r}}-t_{\mathrm{p}}\right)}}$$

(12)

To further assess the comprehensive resilience of groundwater levels during disturbance and recovery processes, the composite resilience index W is introduced to quantify the resilience of the groundwater system:

$$W=UTLON{E}_{DS}+UTLON{E}_{RS}$$

(13)

4.3. Anomaly Data Recognition

During tunnel construction, the dynamic changes of the groundwater system are of great significance for project safety, construction organisation and environmental protection. Due to factors such as complex geological structure, non-linear hydrological response and construction disturbance, anomalies often appear in groundwater monitoring data. These anomalies may be indicative of drastic disturbances in the groundwater system such as water surges, mud surges, seepage, and pipe surges. Small anomalies can lead to system failure, and the traditional construction process is usually remedied after the fact. We use Equations (7) and (10) to construct a response plane that allows preventive detection of large anomalous deviations in the data once they occur, avoiding systematic failures and monitoring the recovery time.

5. Instance Analysis

5.1. Basic Characteristics and Evolution of the Groundwater System

The groundwater system during the disturbance and recovery phases of tunnel excavation is a dynamic hydrogeological system characterised by a non-stationary response, the evolution of which is influenced by a combination of natural rainfall factors and anthropogenic activities. Figure 8 shows the evolution of the normalised Figure 7 groundwater system performance. The main use of the normalisation formula: $K=V\left(t\right)/V\left(t_0\right)$, eliminates the irrelevant characteristics of the data and noise interference, but retains the key features of the data, which is transformed into the performance of the groundwater system and enhances the universality of the model study, the system resilience of the system is 1 when it is not perturbed, and the resilience of the system is 0 when it is running down, and the resilience of the system is in a better state when $k\ge 1$.

As can be seen from the figure, the groundwater system has significant hydrodynamic response characteristics:

(1) $t_d<t<t_p$ period, the groundwater level is significantly affected by the external construction disturbance, and the fluctuation of the water level is large under different initial conditions, indicating that the system is sensitive to recharge, discharge and medium infiltration.

(2) $t_p<t<t_r$ period, the hydraulic gradient adjustment drives the gradual recovery of the water level, and the recharge–discharge relationship tends to stabilise, but the adjustment process is characterised by non-steady state due to the influence of seepage resistance and boundary effect, indicating that the system is resistant.

(3) $t\ge t_r$ period, the system enters the steady-state flow stage, the water level change tends to zero, all curves tend to be close to the stable value of 1.00, the influence of different initial disturbance conditions is gradually eliminated, and the groundwater system finally reaches the dynamic equilibrium, which indicates that the groundwater system possesses the ability of steady-state adjustment.

5.2. Robustness and Recovery Regulation Capacity of the Groundwater System

As shown in Figure 9, the robustness of the groundwater systems of the two tunnels under construction disturbances with different rainfall amounts is shown as follows: the average robustness of the Huayan Tunnel is 0.97 and that of the Hualfu Tunnel is 0.99, which proves that the latter is more resistant to the construction disturbances of the tunnels. To quantify the spatial response, water level monitoring points were established at 100 m intervals along a transect extending from 0 M to 1000 M relative to the tunnel axis center, where “M” denotes the horizontal distance in meters. For Huafu Tunnel, the value is smaller under high rainfall of 0 M∼400 M, which indicates that high rainfall can enhance the resistance of the system, but at 400 M∼1000 M, high rainfall weakens its resistance, and the groundwater level is lower with higher rainfall, which is contrary to common sense. According to numerical simulation and field testing, the impact range is 0 M∼400 M in the case of excavation disturbance in the Wah Fu Tunnel. Within 400 M, negative pressure is formed due to groundwater discharge, and surface water is sucked in by siphoning to increase effective infiltration and replenish the groundwater level. As the distance from the excavation point, 0 M, decreases, the effective infiltration decreases, and the high rainfall is limited in its ability to resist disturbance. Beyond 400 M, there is no pressure difference in the groundwater system as it is virtually unaffected, while high rainfall values bring more surface runoff and less infiltration instead, accompanied by more vigorous vegetation that enhances evapotranspiration. This results in the problem of reduced groundwater resistance to disturbance under high rainfall. This suggests that the groundwater system under disturbance cannot be increased by simply increasing the recharge, but needs to be combined with the scope of influence of the disturbance, stratified management, enhanced recharge in places with small pressure differences within the scope of influence, and weakened recharge outside the scope of influence [62].

Based on Equations (9) and (12), Figure 10 illustrates the resistance capacity of the groundwater system within the main impact zone (0 m–400 m) during tunnel construction disturbances, while Figure 11 presents its regulation capacity during the recovery phase. As can be seen from the graphs, (1) the groundwater resistance capacity of both tunnels increases with the increase of the influence distance, and the recovery capacity shows a trend of decreasing and then increasing with the increase of the distance. (2) The disturbance resistance and recovery capacity of Huafu Tunnel are higher than that of Huayan Tunnel. Rainfall has almost no effect on the resistance of the groundwater system in Huayan Tunnel, mainly because the drainage outflow of the tunnel construction is too large, and the effect of rainfall on the recharge of the groundwater level is limited. However, rainfall has a significant positive effect on the groundwater system resilience. The recovery capacity of the groundwater system in the Huafu Tunnel is amplified by a positive trend with the change of rainfall values, and the recovery capacity of the groundwater system in the Huayan Tunnel is accelerated with the change of rainfall values. It can be seen that the tunnel disturbance phase is mainly based on monitoring, and the amount of rainfall recharge in the recovery phase can achieve rapid recovery of the tunnel groundwater system after damage.

5.3. Comprehensive Resilience of the Groundwater System

The evolution of the integrated resilience of the tunnel was plotted by Equation (13), as shown in Figure 12. It can be observed that the Huafu Tunnel exhibits a high and stable integrated resilience under all rainfall conditions. The main zone of influence of the groundwater system in the Huafu Tunnel is 0 M∼400 M, so only this section of change needs to be observed. In conclusion, both tunnels have significant potential for performance improvement under high rainfall scenarios.

5.4. Identification of Potential Anomalies

Using Equations (7) and (10), the $ROC$ values of the groundwater system during the resistance and recovery phases are plotted to form a hypersurface as shown in Figure 13. Deviations of the $ROC$ values exceeding 2% in absolute value from the hypersurface trigger detection and early warning, enabling reasonable groundwater drawdown during construction and orderly recovery during the restoration phase.

6. Discussion and Application

This study developed a quantitative resilience assessment framework for tunnel-groundwater systems. By integrating numerical simulation with staged resilience indicators, the underlying mechanisms and spatio-temporal patterns of groundwater response were analyzed, leading to the following key findings:

(1) Mechanisms of Resilience Distribution: Groundwater resilience is governed by the synergistic effects of seepage field disturbance and geological buffering. Resistance to disturbance (average robustness) is higher in Huafu Tunnel (0.99) compared to Huayan Tunnel (0.97). The underlying mechanism is related to the spatial distance from the recharge boundary and the hydraulic conductivity of the rock mass, which determines the system’s ability to maintain hydrostatic pressure during excavation-induced stress release.

(2) Nonlinear Patterns of Recovery: Precipitation acts as the primary external driving force for resilience recovery. A threshold effect was observed: for Huafu Tunnel, recovery capacity increases linearly with precipitation; however, Huayan Tunnel exhibits an accelerated, nonlinear growth trend. This pattern indicates that once precipitation exceeds a critical recharge threshold, the groundwater system shifts from a storage-depletion mode to an active-recharge mode, significantly enhancing recovery efficiency.

(3) Spatio-temporal Evolution Patterns: The study identifies a consistent spatial pattern where resistance increases with distance from the tunnel axis, while recovery capacity follows a “U-shaped” trend (first decreasing, then increasing). This occurs because the “disturbance center” experiences the highest hydraulic gradient, which limits initial recovery, whereas the distal areas benefit from the stability of the regional flow field.

(4) Quantitative Anomaly Identification: By mapping abnormal data onto a ROC numerical hypersurface, we transformed qualitative monitoring into a rigorous quantitative warning system. This provides a mathematical basis for identifying non-steady-state behaviors and potential risk points during construction.

In summary, this study clarifies that groundwater resilience in tunnel engineering is not a static property but a dynamic response governed by the interaction between construction-induced drainage and natural precipitation recharge. The proposed quantitative method provides a scientific basis for the sustainable management of groundwater resources in complex hydrogeological environments.