Cold region groundwater environment early warning method fusing digital twin and freeze-thaw process

Abstract

The application provides a cold region groundwater environment early warning method fusing digital twinning and freezing and thawing process, and the method comprises the following steps: constructing a digital twin corresponding to the state of a cold region groundwater system; collecting and assimilating multi-dimensional observation data representing the state of a cold region groundwater environment in real time through a multi-source data perception layer; in a water-heat coupling model layer, based on the multi-dimensional observation data, performing strong coupling numerical simulation of the freezing and thawing process and groundwater flow, and dynamically deducing the influence of the freezing and thawing front migration on the groundwater system; using a data assimilation algorithm, fusing the multi-dimensional observation data and the simulation output of the water-heat coupling model layer, and updating the state variables and model parameters of the digital twin in real time; in a dynamic early warning decision layer, based on the updated state of the digital twin, evaluating a groundwater environment risk index and issuing graded early warning information; through the above technical solution, the water-heat dynamics of the cold region groundwater system can be simulated and mapped in real time.

Description

Technical Field

[0001] This application relates to the field of environmental monitoring technology, and in particular to a method for early warning of groundwater environment in cold regions that integrates digital twins and freeze-thaw processes. Background Technology

[0002] Groundwater environment monitoring and early warning in cold regions are crucial for ensuring ecological security, infrastructure stability, and sustainable water resource utilization in high-latitude or high-altitude areas. In cold environments where freeze-thaw cycles are frequent, the dynamic changes in groundwater are influenced not only by climate, geology, and hydrology but also closely related to the formation and thawing of the surface frozen layer. In recent years, with the development of remote sensing, the Internet of Things (IoT), and numerical simulation technologies, groundwater environment monitoring has gradually evolved towards multi-source sensing and model-driven approaches. Among these, digital twin technology, due to its ability to construct real-time mapping relationships between the physical world and virtual models, has been widely applied in environmental system simulation and risk prediction.

[0003] Among these, groundwater environmental early warning methods that integrate freeze-thaw process modeling have become a key direction in cold region hydrological research. These methods aim to dynamically characterize the impact mechanism of the permafrost layer on groundwater recharge, runoff, and discharge by coupling soil heat conduction, water migration, and phase change processes, and to predict key indicators such as water level, water quality, and frost depth using observational data. However, existing technologies still face significant challenges in achieving high-precision and timely groundwater early warning in cold regions.

[0004] Existing technologies generally suffer from problems such as low model coupling, weak dynamic response capability, and insufficient multi-source data coordination: First, most freeze-thaw-groundwater coupling models adopt static or quasi-steady-state assumptions, making it difficult to accurately reflect the nonlinear characteristics of hydrothermal coupling during the rapid migration of the freeze-thaw interface; second, traditional monitoring systems rely on discrete site data, resulting in limited spatial coverage and failing to support real-time reconstruction of the entire domain state by digital twins; third, existing early warning systems lack the ability to dynamically extrapolate the migration paths of groundwater pollution under freeze-thaw disturbances, leading to a lag in response to sudden environmental risks; finally, the model update mechanism is rigid, failing to establish a closed-loop feedback between observation data and the twin model, resulting in the accumulation of prediction biases during long-term operation. Summary of the Invention

[0005] This application provides a groundwater environment early warning method for cold regions that integrates digital twins and freeze-thaw processes. It solves the problem that existing technologies are difficult to accurately describe and warn of groundwater systems in complex cold environments. It achieves high-fidelity simulation and real-time mapping of hydrothermal dynamics of groundwater systems in cold regions, significantly improving prediction accuracy and dynamic response capabilities.

[0006] To achieve the above objectives, this application provides a method for early warning of groundwater environment in cold regions that integrates digital twins and freeze-thaw processes, the method comprising:

[0007] Based on multi-source heterogeneous monitoring data and geological and hydrological parameters, a three-dimensional high-resolution geological structure model including soil layer, freeze-thaw activity layer and aquifer is established, and a water-heat transfer equation and ice-water phase change dynamic model are coupled to construct a digital twin corresponding to the state of groundwater system in cold region; the digital twin includes a multi-source data sensing layer, a water-heat coupling model layer and a dynamic early warning decision layer.

[0008] Through the multi-source data sensing layer, multi-dimensional observation data characterizing the groundwater environment in cold regions are collected and assimilated in real time; the multi-dimensional observation data includes groundwater level, water quality parameters, soil temperature, soil moisture, and surface freeze-thaw status.

[0009] In the hydrothermal coupling model layer, based on the multidimensional observation data, a strongly coupled numerical simulation of the freeze-thaw process and groundwater flow is performed to dynamically deduce the impact of freeze-thaw front migration on the groundwater system recharge, runoff and discharge processes.

[0010] Using a data assimilation algorithm, the multidimensional observation data is fused with the simulation output of the hydrothermal coupling model layer to update the state variables and model parameters of the digital twin in real time, so as to correct the prediction bias.

[0011] In the dynamic early warning decision-making layer, groundwater environmental risk indicators are assessed based on the updated digital twin status. The risk indicators include the abnormal fluctuation range of groundwater level, the migration rate of pollutants in the freeze-thaw activity layer, and the probability of water quality indicators exceeding the standard. When the risk indicators exceed the preset threshold, graded early warning information is generated and released. The early warning information includes the early warning area, early warning type, occurrence time, cause analysis, risk level, and recommended measures.

[0012] In one possible implementation, the step of establishing a three-dimensional high-resolution geological structure model including soil layers, freeze-thaw activity layers, and aquifers based on multi-source heterogeneous monitoring data and geological and hydrological parameters, and coupling the water and heat transport equation with the ice-water phase transition dynamics model to construct a digital twin corresponding to the state of the groundwater system in cold regions, includes:

[0013] Define the three-dimensional spatial domain and temporal discretization scheme of the digital twin;

[0014] The governing equations in the hydrothermal coupling model layer are initialized. The governing equations include a set of partial differential equations describing soil moisture movement and heat conduction. The soil moisture movement equations consider the release and absorption of latent heat caused by the phase change of ice and water, while the heat conduction equations are coupled with convective heat transport caused by water migration.

[0015] A risk threshold matrix is ​​set for the dynamic early warning decision layer, which is associated with the abnormal fluctuation range of groundwater level, the multiple of exceedance of specific pollutant concentration, and the rate of change of the thickness of the freeze-thaw activity layer.

[0016] In one possible implementation, the real-time acquisition and assimilation of multidimensional observation data characterizing the groundwater environment in cold regions through the multi-source data sensing layer includes:

[0017] Deploy a ground-based IoT sensor network to acquire in-situ data on groundwater level, water quality parameters, soil temperature, and soil moisture with high spatiotemporal resolution;

[0018] Remote sensing satellite data is accessed synchronously to extract surface freeze-thaw status information within the target area; the surface freeze-thaw status information is obtained through passive microwave radiation brightness temperature data inversion;

[0019] An ensemble Kalman filter algorithm is used to perform spatiotemporal scale matching and error correction on the in-situ data and freeze-thaw state information to form a spatiotemporally consistent twin input data field.

[0020] In one possible implementation, the step of performing a strongly coupled numerical simulation of the freeze-thaw process and groundwater flow based on the multidimensional observation data in the hydrothermal coupling model layer, and dynamically extrapolating the impact of freeze-thaw front migration on groundwater system recharge, runoff, and discharge processes, includes:

[0021] Solve the coupled hydrothermal control equations; the hydrothermal control equations are discretized in space using the finite element method and advanced in time using the implicit Euler scheme;

[0022] Within each time step, the soil temperature field and liquid water content field are iteratively solved, and the phase state of pore water is determined based on the soil temperature. When the soil temperature is lower than the freezing temperature, some liquid water is converted into ice, causing a sudden change in soil porosity and hydraulic conductivity.

[0023] The dynamic migration process of freeze-thaw fronts was simulated, and their time-varying effects on the water infiltration capacity of the vadose zone and the boundary conditions of groundwater flow in the saturated zone were calculated.

[0024] In one possible implementation, the step of using a data assimilation algorithm to fuse the multidimensional observation data with the simulation output of the hydrothermal coupling model layer, and updating the state variables and model parameters of the digital twin in real time to correct prediction bias, includes:

[0025] Construct a state vector; the state vector contains all key state variables and parameters to be measured in the hydrothermal coupling model layer;

[0026] Define observation operators to map the model state space to the observation data space;

[0027] The ensemble Kalman filter update loop is executed to calculate the optimal analysis value of the state vector by minimizing the difference between the observed data and the model prediction value, and the initial conditions and parameters of the model are adjusted accordingly to achieve dynamic synchronization between the state of the digital twin and the physical entity.

[0028] In one possible implementation, the dynamic early warning decision-making layer assesses groundwater environmental risk indicators based on the updated digital twin status; and when the risk indicators exceed a preset threshold, generates and releases tiered early warning information, including:

[0029] Based on the updated digital twin status, determine risk indicators for specific future periods;

[0030] A blue warning is generated if any one of the risk indicators slightly exceeds the threshold; a yellow warning is generated if any one or two of the risk indicators significantly exceed the threshold; an orange warning is generated if any two of the risk indicators significantly exceed the threshold, or if a water quality indicator exceeds the standard and the probability of exceeding the standard is greater than a preset probability threshold; a red warning is generated if any three or more of the risk indicators slightly exceed the threshold, or if an actual pollution event has occurred. A slight exceedance of the threshold is defined as an indicator value exceeding the threshold corresponding to the warning trigger condition but not reaching 1.5 times the threshold value; a significant exceedance of the threshold is defined as an indicator value exceeding the threshold corresponding to the warning trigger condition by 1.5 times the threshold value.

[0031] In one possible implementation, the blue warning information represents a risk of concern, the yellow warning information represents a risk of alert, the orange warning information represents a serious risk, and the red warning information represents a particularly serious risk; the warning information is automatically pushed to relevant management terminals through a preset communication protocol.

[0032] In one possible implementation, the warning trigger condition for the abnormal fluctuation amplitude of the groundwater level is that the water level deviation exceeds 0.3m for 24 consecutive hours; the warning trigger condition for the pollutant migration rate is that it exceeds 0.1m / day; and the warning trigger condition for the probability of exceeding the water quality standard is that the chloride ion concentration exceeds 250mg / L or the nitrate nitrogen concentration exceeds 10mg / L.

[0033] In one possible implementation, the warning triggering conditions are adaptively adjusted based on seasonal changes, freeze-thaw conditions, and historical extreme event statistics; during the spring snowmelt season, the warning triggering condition for abnormal fluctuations in groundwater level is a water level deviation exceeding 0.5m for 24 consecutive hours, and during the frozen stable period, the warning triggering condition for abnormal fluctuations in groundwater level is a water level deviation exceeding 0.2m for 24 consecutive hours; the freeze-thaw conditions are determined by freezing depth data.

[0034] In one possible implementation, the method further includes:

[0035] Regularly compare early warning information with actual environmental events to determine the accuracy rate, underreporting rate, and false alarm rate of early warnings.

[0036] The early warning accuracy, false alarm rate, and false alarm rate are transmitted to the digital twin, and the model parameters or structure of the digital twin are iteratively optimized.

[0037] The technical solutions provided in this application embodiment have at least the following technical effects or advantages:

[0038] (1) By constructing a high-fidelity, physical mechanism-driven digital twin, the hydrothermal coupling mechanism of the freeze-thaw process and multi-source real-time monitoring data are deeply integrated, realizing high-precision synchronous mapping and forward-looking simulation of hydrothermal dynamics and pollutant migration behavior of groundwater system in cold regions. This effectively overcomes the defects of loose model coupling, poor dynamic adaptability and long-term prediction error accumulation in traditional methods.

[0039] (2) By relying on the ensemble Kalman filter data assimilation technology, closed-loop correction of model state and parameters is achieved, which significantly improves the accuracy of short-term prediction and the reliability of early warning;

[0040] (3) By establishing a multi-parameter, adaptive threshold early warning indicator system and a four-level graded release mechanism, multiple environmental risks such as abnormal groundwater level, water quality deterioration and pollution migration can be identified, accurately assessed and promptly warned, providing scientific, efficient and intelligent technical support for the ecological protection of groundwater in cold regions, the safety of water sources and emergency management. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments of this application or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 A flowchart illustrating a method for early warning of groundwater environment in cold regions that integrates digital twin and freeze-thaw processes, provided as an embodiment of this application;

[0043] Figure 2 A flowchart of another method for early warning of groundwater environment in cold regions that integrates digital twin and freeze-thaw process provided in an embodiment of this application;

[0044] Figure 3 A flowchart of another method for early warning of groundwater environment in cold regions that integrates digital twin and freeze-thaw process provided in an embodiment of this application;

[0045] Figure 4A schematic diagram illustrating a digital twin determination process provided in an embodiment of this application;

[0046] Figure 5 is a schematic diagram of the process for determining a multi-source heterogeneous real-time monitoring network according to an embodiment of this application;

[0047] Figure 6 This is a schematic diagram illustrating the process of determining dynamic early warning indicators and early warning information provided in an embodiment of this application. Detailed Implementation

[0048] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0049] In the description of the embodiments of this application, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the embodiments of this application and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. The terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; a mechanical connection or an electrical connection; a direct connection or an indirect connection through an intermediate medium; or a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this application according to the specific circumstances.

[0050] Figure 1 A flowchart illustrating a method for early warning of groundwater environment in cold regions that integrates digital twin and freeze-thaw processes, provided as an embodiment of this application. Figure 1 As shown, the method may include the following steps.

[0051] S101. Based on multi-source heterogeneous monitoring data and geological and hydrological parameters, a three-dimensional high-resolution geological structure model including soil layer, freeze-thaw active layer and aquifer is established, and the water and heat transfer equation and ice-water phase change dynamic model are coupled to construct a digital twin corresponding to the state of the groundwater system in cold regions.

[0052] The digital twin includes a multi-source data perception layer, a water-thermal coupling model layer, and a dynamic early warning decision layer.

[0053] For example, this three-dimensional high-resolution geological structure model can use grid cells generated based on ground-penetrating radar exploration and borehole data interpolation, with a horizontal resolution of no less than 10m and a vertical resolution of no less than 0.5m, and accurately depict the typical cold-region stratigraphic structure, including silty clay, gravel layers and bedrock.

[0054] For example, the hydrothermal transport equation employs an improved coupled control equation for heat conduction and moisture transport that considers the latent heat of the ice-water phase change. The moisture transport equation introduces an unsaturated hydraulic conductivity function that considers the effects of freeze-thaw cycles, and the improved heat conduction equation includes a phase change term to accurately describe the energy balance during freezing and thawing.

[0055] For example, the digital twin can be deployed on a high-performance computing cluster, using parallel computing technology to accelerate the solution of large-scale hydrothermal coupling equations, ensuring that model data assimilation and scenario prediction for the next 72 hours are completed within 2 hours of receiving new monitoring data.

[0056] S102. Through this multi-source data sensing layer, multi-dimensional observation data that characterizes the groundwater environment in cold regions are collected and assimilated in real time.

[0057] The multidimensional observation data may include groundwater level, water quality parameters, soil temperature, soil moisture, and surface freeze-thaw status.

[0058] S103. In this hydrothermal coupling model layer, based on the multidimensional observation data, a strongly coupled numerical simulation of the freeze-thaw process and groundwater flow is performed to dynamically deduce the impact of the freeze-thaw front migration on the groundwater system recharge, runoff and discharge processes.

[0059] S104. Using a data assimilation algorithm, the multidimensional observation data is fused with the simulation output of the hydrothermal coupling model layer to update the state variables and model parameters of the digital twin in real time, so as to correct the prediction bias.

[0060] S105. In this dynamic early warning decision-making layer, groundwater environmental risk indicators are assessed based on the updated digital twin status; and when the risk indicators exceed the preset threshold, graded early warning information is generated and released.

[0061] The risk indicators can include the abnormal fluctuation range of groundwater level, the migration rate of pollutants in the freeze-thaw activity layer, and the probability of exceeding water quality standards; the early warning information can include the warning area, warning type, occurrence time, cause analysis, risk level, and recommended measures.

[0062] According to the above technical solution, a high-fidelity, physical mechanism-driven digital twin can be constructed to deeply integrate the hydrothermal coupling mechanism of the freeze-thaw process with multi-source real-time monitoring data. This enables high-precision synchronous mapping and forward-looking simulation of hydrothermal dynamics and pollutant migration behavior in cold regions' groundwater systems, effectively overcoming the shortcomings of traditional methods such as loose model coupling, poor dynamic adaptability, and long-term prediction error accumulation.

[0063] In one possible implementation, S101 may include: defining the three-dimensional spatial domain and time discretization scheme of the digital twin; initializing the governing equations in the hydrothermal coupling model layer, the governing equations including a set of partial differential equations describing soil moisture movement and heat conduction, wherein the soil moisture movement equations consider the release and absorption of latent heat caused by the phase change of ice and water, and the heat conduction equations are coupled with convective heat transport caused by water migration; setting the risk threshold matrix of the dynamic early warning decision layer, the risk threshold matrix being associated with the abnormal fluctuation amplitude of groundwater level, the multiple of exceedance of specific pollutant concentration, and the rate of change of the thickness of the freeze-thaw activity layer.

[0064] In one possible implementation, S102 may include: deploying a ground-based Internet of Things (IoT) sensor network to acquire in-situ data of groundwater level, water quality parameters, soil temperature, and soil moisture with high spatiotemporal resolution; synchronously accessing remote sensing satellite data to extract surface freeze-thaw status information within the target area; obtaining the surface freeze-thaw status information through passive microwave radiation brightness temperature data inversion; and using an ensemble Kalman filter algorithm to perform spatiotemporal scale matching and error correction on the in-situ data and freeze-thaw status information to form a spatiotemporally consistent twin input data field.

[0065] For example, this ground-based IoT sensor network can be used for real-time monitoring of groundwater systems in target cold regions. It may include groundwater level sensors deployed in typical geomorphic units with a measurement accuracy of ±1 cm; soil temperature and moisture content probes buried at depths covering a range of 0 to 3 m with a vertical spacing of 0.2 m; a frost depth monitor employing the time-domain reflectometry principle with a monitoring accuracy of ±2 cm; and a water quality parameter detection unit periodically detecting chloride ion concentration, nitrate nitrogen content, and pH value in the groundwater. This target range can be set by the user and is not limited here.

[0066] For example, this ensemble Kalman filter algorithm uses a set of 50 members to characterize the uncertainty of the model state, with an assimilation period of 6 hours. The assimilated observation variables include soil temperature, volumetric water content, and groundwater level. The model's state variables and parameters are updated by minimizing the difference between the observed and simulated values.

[0067] In one possible implementation, S103 may include: solving the coupled hydrothermal control equations; the hydrothermal control equations are discretized spatially using the finite element method and advanced temporally using an implicit Euler scheme; within each time step, the soil temperature field and liquid water content field are iteratively solved, and the phase state of pore water is determined based on the soil temperature. When the soil temperature is lower than the freezing temperature, some liquid water is converted into ice, causing a sudden change in soil porosity and hydraulic conductivity; the dynamic migration process of the freeze-thaw front is simulated, and its time-varying influence on the water infiltration capacity of the vadose zone and the boundary conditions of groundwater flow in the saturated zone is calculated.

[0068] In one possible implementation, S104 may include: constructing a state vector; the state vector contains all key state variables and parameters to be measured in the hydrothermal coupling model layer; defining an observation operator to map the model state space to the observation data space; executing an ensemble Kalman filter update loop to calculate the optimal analysis value of the state vector by minimizing the difference between the observed data and the model prediction value, and adjusting the initial conditions and parameters of the model accordingly to achieve dynamic synchronization between the digital twin and the physical entity state.

[0069] In one possible implementation, S105 may include: determining risk indicators for a specific future period based on the updated digital twin state; generating a blue warning message if any one of the risk indicators slightly exceeds the threshold; generating a yellow warning message if any one of the risk indicators significantly exceeds the threshold or if any two indicators slightly exceed the threshold; generating an orange warning message if any two of the risk indicators significantly exceed the threshold, or if a water quality indicator exceeds the standard and the probability of exceeding the standard is greater than a preset probability threshold; and generating a red warning message if any three or more of the risk indicators slightly exceed the threshold or if an actual pollution event has occurred. The slight exceedance threshold is defined as the indicator value exceeding the threshold corresponding to the warning trigger condition but not reaching 1.5 times, and the significant exceedance threshold is defined as the indicator value exceeding the threshold corresponding to the warning trigger condition by 1.5 times. The range of the preset probability threshold can be set by the user and is not limited here.

[0070] In one possible implementation, the blue warning message indicates a risk of concern, the yellow warning message indicates a risk of alert, the orange warning message indicates a serious risk, and the red warning message indicates a particularly serious risk; the warning message is automatically pushed to the relevant management terminal through a preset communication protocol.

[0071] For example, the relevant management terminal can be a computer, server, mobile phone, etc., and there is no limitation here. Furthermore, the warning information can be spatially visualized using a geographic information system.

[0072] In one possible implementation, the warning trigger condition for the abnormal fluctuation amplitude of the groundwater level is a water level deviation exceeding 0.3m for 24 consecutive hours; the warning trigger condition for the pollutant migration rate is exceeding 0.1m per day; and the warning trigger condition for the probability of exceeding the water quality standard is a chloride ion concentration exceeding 250mg / L or a nitrate nitrogen concentration exceeding 10mg / L.

[0073] In one possible implementation, the warning triggering condition is adaptively adjusted based on seasonal changes, freeze-thaw conditions, and historical extreme event statistics; during the spring snowmelt season, the warning triggering condition for abnormal fluctuations in the groundwater level is a water level deviation exceeding 0.5m for 24 consecutive hours, and during the frozen stable period, the warning triggering condition for abnormal fluctuations in the groundwater level is a water level deviation exceeding 0.2m for 24 consecutive hours; the freeze-thaw condition is determined by the freezing depth data.

[0074] For example, the frost depth threshold can be adjusted based on different freezing periods: during the freezing deepening period, the daily frost depth increment is >2 cm; during the stable freezing period, the daily frost depth increment is ≤2 cm and the frost depth is >0.5 m; during the initial thawing period, the daily frost depth decrease is >2 cm; and during the complete thawing period, the frost depth is 0.

[0075] Figure 2 A flowchart illustrating another method for early warning of groundwater environment in cold regions that integrates digital twins and freeze-thaw processes, provided as an embodiment of this application. Figure 2 As shown, the method may also include the following steps.

[0076] S106. Regularly compare early warning information with actual environmental events to determine the accuracy rate, underreporting rate, and false alarm rate of early warnings.

[0077] S107. Transmit the early warning accuracy, false alarm rate and false alarm rate to the digital twin, and iteratively optimize the model parameters or structure of the digital twin.

[0078] In one possible implementation, this application also provides a cold region groundwater environment early warning system that integrates digital twins and freeze-thaw processes. This system is used to execute the above-mentioned cold region groundwater environment early warning method that integrates digital twins and freeze-thaw processes. The system may include the following modules.

[0079] The data acquisition module is used to deploy and manage ground sensor nodes and remote sensing data receiving equipment to realize the automated acquisition and preprocessing of multi-source heterogeneous monitoring data and geological and hydrological parameters.

[0080] The digital twin construction module is used to establish a three-dimensional high-resolution geological structure model, including soil layers, freeze-thaw activity layers, and aquifers, based on multi-source heterogeneous monitoring data and geological and hydrological parameters. It also couples the hydrothermal transport equation with the ice-water phase transition dynamics model to construct a digital twin corresponding to the state of the groundwater system in cold regions. This digital twin includes a multi-source data sensing layer, a hydrothermal coupling model layer, and a dynamic early warning decision-making layer.

[0081] The data assimilation module is used to collect and assimilate multidimensional observation data characterizing the groundwater environment in cold regions in real time through the multi-source data sensing layer. In the hydrothermal coupling model layer, based on the multidimensional observation data, a strongly coupled numerical simulation of the freeze-thaw process and groundwater flow is performed to dynamically deduce the impact of freeze-thaw front migration on the groundwater system's recharge, runoff, and discharge processes. Using the data assimilation algorithm, the multidimensional observation data is fused with the simulation output of the hydrothermal coupling model layer to update the state variables and model parameters of the digital twin in real time to correct prediction biases.

[0082] The multidimensional observation data may include groundwater level, water quality parameters, soil temperature, soil moisture, and surface freeze-thaw status.

[0083] The risk warning module is used in this dynamic warning decision-making layer to assess groundwater environmental risk indicators based on the updated digital twin status; and to generate and release graded warning information when the risk indicators exceed preset thresholds.

[0084] The risk indicators can include the abnormal fluctuation range of groundwater level, the migration rate of pollutants in the freeze-thaw activity layer, and the probability of exceeding water quality standards; the early warning information can include the warning area, warning type, occurrence time, cause analysis, risk level, and recommended measures.

[0085] According to the above technical solution, a high-fidelity, physical mechanism-driven digital twin can be constructed to deeply integrate the hydrothermal coupling mechanism of the freeze-thaw process with multi-source real-time monitoring data. This enables high-precision synchronous mapping and forward-looking simulation of hydrothermal dynamics and pollutant migration behavior in cold regions' groundwater systems, effectively overcoming the shortcomings of traditional methods such as loose model coupling, poor dynamic adaptability, and long-term prediction error accumulation.

[0086] The following is combined with Figure 3 The workflow of Embodiment 1 of this application will be described. The method provided in this application may include the following steps.

[0087] S301. Construct a high-fidelity digital twin of the groundwater system in cold regions.

[0088] The aforementioned S301 aims to establish a virtual entity that can accurately reflect the underground structure, hydrothermal physical properties, and freeze-thaw dynamics of the target area, providing a physical basis for subsequent dynamic simulation and early warning.

[0089] (1) Collect basic geological data of the target area, including borehole columnar sections, core descriptions, stratigraphic chronology, hydrogeological test data, and geophysical exploration results such as high-density electrical resistivity tomography and ground-penetrating radar, to ensure a detailed depiction of the underground structure. After standardization, these raw data are uniformly imported into the geographic information system platform for spatial registration and quality control, eliminating abnormal records caused by equipment failure or human operation, and retaining valid data for subsequent modeling.

[0090] (2) Integrate multi-source spatial data such as remote sensing images, digital elevation models, land use type maps and historical frost depth distribution maps to establish a unified spatial reference framework and provide a consistent geographical benchmark for subsequent modeling.

[0091] For example, all multi-source spatial data are generated using the WGS-84 coordinate system and UTM projection (Universal Transverse Mercator). Control point corrections are used to eliminate geometric deviations between data from different sources, ensuring that the spatial consistency error is less than 0.5 pixels. Based on this multi-source spatial data, Kriging interpolation, inverse distance weighting, or machine learning-assisted spatial interpolation methods are employed to construct a three-dimensional structured grid model with a horizontal resolution of at least 10 m and a vertical resolution of at least 0.5 m, fully reflecting the spatial variability of complex strata in cold regions.

[0092] The grid model employs hexahedral elements, with the total number of elements dynamically adjusted based on the area. A typical demonstration area (e.g., 10 km²) can generate over 2 million computational elements, sufficient to capture local heterogeneity during the advance of freeze-thaw fronts. Next, key hydrothermal parameters such as soil texture, porosity, saturated hydraulic conductivity, thermal conductivity, and specific heat capacity are assigned layer by layer to the grid model. This accurately characterizes typical cold-region stratigraphic units and their interface features, including silty clay, gravel layers, and bedrock, ensuring the model accurately represents sensitive layers within the freeze-thaw activity zone.

[0093] The parameter assignment process employs a stratified statistical method. Specifically, for each geological unit, laboratory test results from all borehole samples within the region are collected, and the mean, standard deviation, and coefficient of variation of each parameter are calculated. If the number of samples in a geological unit is less than 5, spatial extrapolation is performed using parameters from neighboring similar units, and Monte Carlo random perturbation is introduced to characterize parameter uncertainty. For example, the porosity of the silty clay layer is assigned in the range of 0.35–0.45, and the saturated hydraulic conductivity is 1×10⁻⁶. -6 –5×10 -6 The thermal conductivity is 1.2–1.8 W / (m·K), and the specific heat capacity is 1.8–2.2 MJ / (m³·K).

[0094] (3) Coupled Improved Hydrothermal Coupling Control Equation. The Richards equation considering the freezing retardation effect is adopted for water transport, and the unsaturated hydraulic conductivity function under freeze-thaw conditions is introduced. ,in This refers to the volumetric water content. For temperature; the heat conduction equation includes the latent heat of the ice-water phase transition, expressed as: ;in, The latent heat of phase transition of ice (J / kg). The density of water, The volumetric water content of the ice phase is (m³ / m³). The apparent density of the medium is (kg / m³). It is the volumetric heat capacity (J / (m³·K)). Where is the effective thermal conductivity (W / (m·K)), T is the temperature, and t is the time. This equation system strictly follows the principles of energy conservation and mass conservation, and can realistically reproduce the complex coupled behavior of water migration obstruction, heat release / absorption, and ice-water phase transition during freeze-thaw processes.

[0095] Specifically, the unsaturated hydraulic conductivity function Using the freezing correction form of the Mualem-van Genuchten model, when the temperature is below 0°C, the liquid water content is limited to the unfrozen water content, and the hydraulic conductivity decreases exponentially with increasing ice content; effective thermal conductivity... The series-parallel hybrid model is used to calculate the phase transition based on the volume fractions of ice, water, and solid particles, ensuring a continuous transition of thermophysical properties during the phase transition.

[0096] (4) The above physical model is embedded into a high-performance numerical simulation platform to construct a high-fidelity virtual entity with real-time response capabilities, namely a digital twin of the cold region groundwater system, which serves as the core computing engine and decision support foundation for the entire early warning system. The numerical simulation platform uses the finite element method to discretize the control equations, the implicit Crank-Nicolson scheme for time integration, and the second-order Lagrange element for spatial discretization. To improve computational efficiency, the model is deployed on a GPU-accelerated server cluster, and the time taken for a single 72-hour simulation is controlled within 15 minutes, meeting the time requirements for real-time early warning.

[0097] Reference Figure 4 The schematic diagram of the digital twin determination process shows the process of constructing a high-fidelity virtual entity through geological structure, hydrothermal parameters, governing equations, and numerical solvers. The above process constitutes the basis of virtual mapping driven by physical mechanisms.

[0098] S302. Deploy and integrate a multi-source heterogeneous real-time monitoring network.

[0099] The above-mentioned step S302 aims to establish a perception system that covers the entire domain, updates frequently, and synchronizes multiple parameters, providing continuous and reliable state input for the digital twin.

[0100] (1) Representative monitoring points are set up according to the geomorphological zones, and each monitoring point is equipped with an integrated intelligent monitoring station to ensure the representativeness of spatial coverage and the comparability of monitoring data.

[0101] For example, the geomorphological zoning is based on digital elevation models and automatic identification of remote sensing images, and is divided into four typical units: river valley plains, slopes, frost heave hills, and marsh wetlands. At least three monitoring points are set up in each unit, with a total of no less than 12 points, forming a spatially balanced monitoring network.

[0102] The groundwater level sensor is installed in a dedicated observation well, covering the local historical lowest to highest water levels, with a measurement accuracy of ±1 cm and a sampling frequency of once per hour, ensuring sensitive detection of even minor fluctuations in water level. The observation well penetrates the active layer to reach the stable aquifer. The well casing is made of PVC, and the filter section is wrapped with a stainless steel mesh with a pore size of 0.1 mm to prevent siltation. The sensor is a pressure-type water level gauge with a built-in temperature compensation module, ensuring long-term stability in environments ranging from -30℃ to +50℃.

[0103] Soil temperature and volumetric moisture content probes are deployed in a multi-depth array at depths of 0.2 m, 0.4 m, 0.6 m, 0.8 m, 1.0 m, 1.5 m, 2.0 m, 2.5 m, and 3.0 m, with a vertical spacing of 0.2–0.5 m. They support simultaneous measurement of temperature (accuracy ±0.1℃) and volumetric moisture content (accuracy ±2%), comprehensively reflecting the vertical migration process of freeze-thaw fronts. The probes employ the TDR principle, transmitting pulses at a frequency of 500 MHz and a signal sampling rate of 1 GHz. The dielectric constant is retrieved through waveform analysis, and the moisture content is then calculated. Nine probes are installed at each monitoring point, fixed along the same vertical profile within a PVC sleeve. The sleeve is sealed with bentonite to reduce lateral moisture interference.

[0104] The freeze depth monitor, based on the principles of time-domain or frequency-domain reflectometry, continuously tracks the position of the freezing front along a vertical profile, with a monitoring accuracy of ±2 cm and an update cycle of 30 minutes, enabling high-frequency capture of freeze depth dynamics. This instrument shares the same hardware platform as the soil moisture probe, determining freeze depth by identifying abrupt changes in dielectric constant (typically corresponding to the ice-water phase transition interface). The data is processed in real-time by a local microcontroller before being output.

[0105] The water quality parameter detection unit integrates an electrochemical sensor and a micro-spectral module, automatically collecting groundwater samples every 6 hours to detect chloride ions, nitrate nitrogen, and pH value. The detection limits are 1 mg / L, 0.1 mg / L, and ±0.05 pH units, respectively, meeting the requirements of the "Technical Specification for Groundwater Environmental Monitoring". A peristaltic pump is used for sampling, drawing 50 mL of water sample each time. The sample is filtered through a 0.45 μm filter before entering the detection chamber. The chloride ion sensor is a silver / silver chloride electrode, the nitrate nitrogen sensor is a UV absorption spectrometer (220 nm and 275 nm dual-wavelength differential), and the pH sensor is a glass composite electrode. All sensors are equipped with automatic cleaning and calibration functions, and standard solution calibration is performed weekly.

[0106] All sensors are connected to the local IoT gateway via LoRaWAN or NB-IoT wireless communication modules. After data cleaning, timestamp alignment and outlier removal are completed by edge computing devices, the data is uploaded to the cloud digital twin platform in real time in JSON format via HTTPS protocol. This enables millisecond-level synchronization and two-way interaction between monitoring data and the virtual model, forming a closed-loop data flow of "perception-feedback-optimization".

[0107] (2) Edge computing devices run lightweight data processing algorithms.

[0108] The raw data is smoothed using a sliding window (window size of 5 sampling points), and outliers are removed based on the 3σ principle. Data from different sensors are aligned according to UTC timestamps, packaged, and uploaded to a cloud server for centralized processing.

[0109] Referring to Figure 5, the schematic diagram of the determination process of this multi-source heterogeneous real-time monitoring network illustrates the complete data flow from sensor deployment, data acquisition, edge processing to cloud integration.

[0110] S303. Implement dynamic simulation and data assimilation of the freeze-thaw process.

[0111] The aforementioned S303 aims to continuously correct the state of the digital twin through deep fusion of physical models and observation data, ensuring its high-fidelity mapping to the real system.

[0112] (1) Initialize the soil temperature field, moisture content field, ice content field and groundwater level field at the current moment in the digital twin to ensure that the initial state is highly consistent with the actual observation. The initialization process adopts the nearest neighbor interpolation method, which assigns the measured values ​​of each monitoring point to the nearest grid cell, and fills the unmonitored area through linear interpolation to form a complete initial field.

[0113] Subsequently, a coupled hydrothermal model was run with a time step of 6 hours to simulate the spatiotemporal evolution of the freeze-thaw front, water redistribution, and heat flux changes over the next 72 hours, predicting the development trends of key hydrological processes. The time step was chosen based on a balance between CFL stability conditions and computational efficiency, ensuring the stability of the numerical solution while meeting real-time requirements. The simulation output includes variables such as temperature, liquid water content, ice content, head, and heat flux for each grid cell, with a time resolution of 1 hour.

[0114] (2) Data assimilation is performed using an ensemble Kalman filter algorithm. A model state set containing 50 members is constructed, each member representing a perturbed initial field or key parameter combination to quantify model uncertainty. The initial perturbation is achieved by adding Gaussian white noise to the initial field, with the noise standard deviation being 1.5 times the standard deviation of the measured values; parameter perturbations are performed on key parameters such as saturated hydraulic conductivity and thermal diffusivity, randomly sampled within their prior distribution range. The ensemble members run independently and in parallel, making full use of the computing cluster resources.

[0115] In each assimilation cycle (6 hours), the measured soil temperature, volumetric water content, and groundwater level at each depth are used as input observation vectors to enhance the model's ability to track the actual state. The dimension of the observation vector is equal to the total number of all valid monitoring data, typically approximately 120 dimensions (12 monitoring points × 9 depth layer temperatures + 12 monitoring points × 9 depth layer water content + 12 water levels). The model state is mapped to the observation space using observation operators, and the residuals between simulated and measured values ​​are calculated to identify the sources of model bias. The observation operators are simple interpolation functions that map the simulated values ​​of the grid cells to the sensor locations.

[0116] (3) Based on the minimum variance estimation criterion, update the state variables (temperature, water content, head) and key parameters (such as saturated hydraulic conductivity and thermal diffusivity) of each set member to generate the corrected optimal state estimate and improve the reliability of model prediction. The update formula follows the standard EnKF algorithm: ;in, Forecast state set, To analyze the set of states, For the observation vector, For the observation operator, The Kalman gain matrix is ​​calculated from the prediction error covariance and the observation error covariance. The observation error covariance matrix is ​​a diagonal matrix, with diagonal elements representing the square of the sensor's accuracy.

[0117] (4) The correction results are used as the initial conditions for the next simulation cycle to form a closed-loop optimization mechanism, ensuring that the simulation error of the freeze-thaw process is controlled within 10%, which is significantly better than the prediction performance of the traditional static model. The root mean square error (RMSE) index is used for error assessment, and it is calculated separately for temperature, moisture content and water level. The RMSE is required to be less than 0.5℃, 0.03 m³ / m³ and 0.05m respectively.

[0118] S304. Determine dynamic early warning indicators for multiple parameters of the groundwater environment.

[0119] The aforementioned S304 aims to transform complex model outputs into actionable risk quantification indicators, providing a direct basis for early warning decisions.

[0120] (1) Amplitude of abnormal fluctuations in groundwater level. Based on the hourly water level sequence output by the digital twin, the moving average deviation between the deviation and the daily average water level of the same period in the past 5 years is calculated. If the absolute value of the deviation exceeds the dynamic threshold (0.2 m during the freezing period, 0.5 m during the snowmelt period, and 0.3 m during the transition period) for 24 consecutive hours, it is judged as abnormal, reflecting the hydrological imbalance caused by the freeze-thaw cycle or extreme climate. The moving average window is 30 days, and the same period refers to 7 days before and after the same date. The dynamic threshold is determined according to the current season and freeze-thaw stage.

[0121] (2) Pollutant migration rate within the freeze-thaw activity layer. Virtual tracer particles were deployed within the freeze-thaw activity layer, and the average daily migration distance of typical pollutants such as nitrates was calculated by combining the simulated Darcy velocity and longitudinal dispersion coefficient. A risk warning was triggered when the migration rate exceeded 0.1 m / d, indicating that pollution may be rapidly approaching water sources or ecologically sensitive areas. The number of virtual tracer particles was 1000, evenly distributed in potential pollution source areas (such as farmland and landfills). Their trajectories were simulated using the Lagrange particle tracking method, and the migration rate was taken as the 90th percentile of the average daily displacement of all particles to characterize high-risk migration scenarios.

[0122] (3) Probability of exceeding water quality standards. By integrating real-time monitoring data and model predictions, a Bayesian probability model is used to estimate the probability of exceeding water quality standards within the next 24 hours. Concentration >250 mg / L or The probability of occurrence is calculated; if the probability is ≥70%, it is considered high-risk, providing early warning of potential water quality deterioration events. The Bayesian model uses monitoring data as the likelihood function, with the model's predicted values ​​being a prior distribution. The posterior distribution is sampled using the Markov chain Monte Carlo method to calculate the probability of exceeding the standard. The weight of monitoring data decays over time; the weight of data from the most recent 6 hours is 1.0, and the weight of data from 6–24 hours linearly decays to 0.5.

[0123] (4) The dynamic threshold is adaptively adjusted based on the season (winter, spring, summer, autumn), freeze-thaw stage (deepening freeze, stable freeze, early thaw, and complete thaw) and the statistical characteristics of extreme events in the past 10 years (such as snowmelt floods and frost heave damage). It is automatically updated by the system's built-in rule engine to ensure that the early warning standard always conforms to the actual hydrological and climatic patterns of the region. The freeze-thaw stage is automatically identified by real-time freeze depth data: the deepening freeze stage is defined as a daily increase in freeze depth > 2 cm, the stable freeze stage is defined as a daily increase in freeze depth ≤ 2 cm and freeze depth > 0.5 m, the early thaw stage is defined as a daily decrease in freeze depth > 2 cm, and the complete thaw stage is defined as freeze depth = 0.

[0124] S305. Generate and publish graded early warning information.

[0125] The aforementioned S305 aims to transform risk assessment results into intuitive and actionable early warning instructions and to efficiently communicate them to relevant parties.

[0126] (1) Set four levels of early warning risk level: blue risk level (any indicator slightly exceeds the threshold), yellow risk level (any indicator significantly exceeds the threshold or any two indicators slightly exceed the threshold), orange risk level (any two indicators significantly exceed the threshold, or water quality indicators exceed the standard and the probability of exceeding the standard is greater than the preset probability threshold), and red risk level (any three or more indicators slightly exceed the threshold or an actual pollution event has occurred), so as to achieve refined classification of risk level.

[0127] Among them, a slight exceedance is defined as an indicator value exceeding the threshold corresponding to the warning trigger condition but not reaching 1.5 times, while a significant exceedance is defined as an indicator value exceeding the threshold corresponding to the warning trigger condition by 1.5 times. When at least one warning indicator exceeds the threshold corresponding to its warning trigger condition, the warning information generation process corresponding to the level is automatically triggered without manual intervention, ensuring timely response.

[0128] (2) Generate and publish early warning information.

[0129] Based on different risk levels, corresponding color-coded warning messages are generated. A blue risk level corresponds to a blue warning message (risk status: pay attention), a yellow risk level corresponds to a yellow warning message (risk status: alert), an orange risk level corresponds to an orange warning message (severe risk status), and a red risk level corresponds to a red warning message (extremely severe risk status).

[0130] The early warning information includes the warning type (abnormal water level / pollution migration / water quality deterioration), occurrence time, warning area (GIS vector boundary), cause analysis (e.g., rapid warming leading to a surge in snowmelt infiltration), and recommended measures (e.g., strengthening monitoring, restricting water withdrawal, initiating emergency response), providing decision-makers with a complete basis for action. The impact range of the warning area is determined using a spatial clustering algorithm: connectivity analysis is performed on out-of-threshold grid cells, adjacent cells are merged to form continuous risk zones, and polygonal vector boundaries are generated.

[0131] Early warning information is spatially visualized through a WebGIS platform, supporting the overlay of basic geographic information layers such as terrain, water systems, settlements, and water sources to intuitively present the spatial pattern of risks. WebGIS is developed using the OpenLayers framework and supports interactive operations such as zooming, panning, and layer switching. Risk areas are rendered using color gradations, with darker colors indicating higher risks.

[0132] Warning information can be pushed to relevant departments, institutions, communities, and the public through multiple channels, including SMS, email, mobile apps, and public information display screens, with a push delay of no more than 10 minutes to ensure efficient delivery of information to relevant parties. Push strategies are customized based on user roles; for example, relevant departments receive complete technical reports, relevant institutions receive operational suggestions, communities receive concise risk warnings, and the public receives safety guidelines. All push content is encrypted during transmission to ensure information security.

[0133] See Figure 6 The schematic diagram of the dynamic early warning indicator and early warning information determination process shows the workflow of determining groundwater environmental risk indicators (abnormal fluctuation range of groundwater level, migration rate of pollutants in the freeze-thaw activity layer and probability of exceeding water quality standards) based on the assimilated and updated digital twin, judging the risk level based on the risk indicators, generating early warning information and releasing it to relevant management terminals.

[0134] Based on the above technical solution, a closed-loop, adaptive, and highly timely early warning system is constructed. The digital twin serves as the core computing engine, the monitoring network provides real-time input, data assimilation ensures model accuracy, early warning indicators quantify risks, and hierarchical release enables decision support.

[0135] The workflow of the technical solution of this application will be described below with reference to Embodiment 2.

[0136] In the scenario of centralized water source protection in cold-region cities, refined monitoring of groundwater pollution risks is implemented. The digital twin spatial domain focuses on a 3km radius around the water source, using a higher resolution grid of 20m for discretization.

[0137] The multi-source data sensing layer has added 16 sets of online water quality monitoring instruments, focusing on detecting volatile organic compounds and heavy metals, with the sampling frequency increased to once every 5 minutes. The hydrothermal coupling model layer has been particularly enhanced to simulate the migration process of pollutants in the unsaturated zone, adding convection-diffusion terms to the governing equations, where the retardation factor is dynamically adjusted based on soil organic matter content and pore ice saturation. A localized ensemble Kalman filter technique has been introduced to limit the assimilation radius to 1 kilometer, avoiding excessive influence of remote observation data on parameter estimation near the water source. The dynamic early warning decision layer has added a pollutant arrival time prediction function, calculating the shortest migration path time from the potential pollution source to the water intake well using an inverse particle tracking algorithm.

[0138] When it is predicted that the benzene concentration in the groundwater around a chemical plant will exceed the limit of 0.01 mg / L within 48 hours, a red alert will be automatically triggered and an emergency response plan will be activated, including specific measures such as shutting down the affected production wells and activating backup water sources. The alert information will be pushed to relevant management terminals simultaneously.

[0139] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, please refer to each other. Each embodiment focuses on describing the differences from other embodiments.

[0140] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of this application.

Claims

1. A method for early warning of groundwater environment in cold regions that integrates digital twin and freeze-thaw processes, characterized in that, The method includes: Based on multi-source heterogeneous monitoring data and geological and hydrological parameters, a three-dimensional high-resolution geological structure model including soil layer, freeze-thaw activity layer and aquifer is established, and a water-heat transfer equation and ice-water phase change dynamic model are coupled to construct a digital twin corresponding to the state of groundwater system in cold region; the digital twin includes a multi-source data sensing layer, a water-heat coupling model layer and a dynamic early warning decision layer. Through the multi-source data sensing layer, multi-dimensional observation data characterizing the groundwater environment in cold regions are collected and assimilated in real time; the multi-dimensional observation data includes groundwater level, water quality parameters, soil temperature, soil moisture, and surface freeze-thaw status. In the hydrothermal coupling model layer, based on the multidimensional observation data, a strongly coupled numerical simulation of the freeze-thaw process and groundwater flow is performed to dynamically deduce the impact of freeze-thaw front migration on the groundwater system recharge, runoff and discharge processes. Using a data assimilation algorithm, the multidimensional observation data is fused with the simulation output of the hydrothermal coupling model layer to update the state variables and model parameters of the digital twin in real time, so as to correct the prediction bias. In the dynamic early warning decision-making layer, groundwater environmental risk indicators are assessed based on the updated digital twin status. The risk indicators include the abnormal fluctuation range of groundwater level, the migration rate of pollutants in the freeze-thaw activity layer, and the probability of water quality indicators exceeding the standard. When the risk indicators exceed the preset threshold, graded early warning information is generated and released. The early warning information includes the early warning area, early warning type, occurrence time, cause analysis, risk level, and recommended measures.

2. The method according to claim 1, characterized in that, Based on multi-source heterogeneous monitoring data and geological and hydrological parameters, a three-dimensional high-resolution geological structure model including soil layer, freeze-thaw active layer, and aquifer is established. This model is then coupled with the water and heat transport equation and the ice-water phase transition dynamics model to construct a digital twin corresponding to the state of the groundwater system in cold regions, including: Define the three-dimensional spatial domain and temporal discretization scheme of the digital twin; The governing equations in the hydrothermal coupling model layer are initialized. The governing equations include a set of partial differential equations describing soil moisture movement and heat conduction. The soil moisture movement equations consider the release and absorption of latent heat caused by the phase change of ice and water, while the heat conduction equations are coupled with convective heat transport caused by water migration. A risk threshold matrix is ​​set for the dynamic early warning decision layer, which is associated with the abnormal fluctuation range of groundwater level, the multiple of exceedance of specific pollutant concentration, and the rate of change of the thickness of the freeze-thaw activity layer.

3. The method according to claim 1, characterized in that, The multi-dimensional observation data characterizing the groundwater environment in cold regions, collected and assimilated in real time through the multi-source data sensing layer, includes: Deploy a ground-based IoT sensor network to acquire in-situ data on groundwater level, water quality parameters, soil temperature, and soil moisture with high spatiotemporal resolution; Remote sensing satellite data is accessed synchronously to extract surface freeze-thaw status information within the target area; the surface freeze-thaw status information is obtained through passive microwave radiation brightness temperature data inversion; An ensemble Kalman filter algorithm is used to perform spatiotemporal scale matching and error correction on the in-situ data and freeze-thaw state information to form a spatiotemporally consistent twin input data field.

4. The method according to claim 1, characterized in that, In the hydrothermal coupling model layer, based on the multidimensional observation data, a strongly coupled numerical simulation of the freeze-thaw process and groundwater flow is performed to dynamically deduce the impact of freeze-thaw front migration on groundwater system recharge, runoff, and discharge processes, including: Solve the coupled hydrothermal control equations; the hydrothermal control equations are discretized in space using the finite element method and advanced in time using the implicit Euler scheme; Within each time step, the soil temperature field and liquid water content field are iteratively solved, and the phase state of pore water is determined based on the soil temperature. When the soil temperature is lower than the freezing temperature, some liquid water is converted into ice, causing a sudden change in soil porosity and hydraulic conductivity. The dynamic migration process of freeze-thaw fronts was simulated, and their time-varying effects on the water infiltration capacity of the vadose zone and the boundary conditions of groundwater flow in the saturated zone were calculated.

5. The method according to claim 1, characterized in that, The process of using a data assimilation algorithm to fuse the multidimensional observation data with the simulation output of the hydrothermal coupling model layer, and updating the state variables and model parameters of the digital twin in real time to correct prediction bias, includes: Construct a state vector; the state vector contains all key state variables and parameters to be measured in the hydrothermal coupling model layer; Define observation operators to map the model state space to the observation data space; The ensemble Kalman filter update loop is executed to calculate the optimal analysis value of the state vector by minimizing the difference between the observed data and the model prediction value, and the initial conditions and parameters of the model are adjusted accordingly to achieve dynamic synchronization between the state of the digital twin and the physical entity.

6. The method according to claim 1, characterized in that, In the dynamic early warning decision-making layer, groundwater environmental risk indicators are assessed based on the updated digital twin status. When risk indicators exceed preset thresholds, tiered early warning information is generated and issued, including: Based on the updated digital twin status, determine risk indicators for specific future periods; A blue warning is generated if any one of the risk indicators slightly exceeds the threshold; a yellow warning is generated if any one or two of the risk indicators significantly exceed the threshold; an orange warning is generated if any two of the risk indicators significantly exceed the threshold, or if a water quality indicator exceeds the standard and the probability of exceeding the standard is greater than a preset probability threshold; a red warning is generated if any three or more of the risk indicators slightly exceed the threshold, or if an actual pollution event has occurred. A slight exceedance of the threshold is defined as an indicator value exceeding the threshold corresponding to the warning trigger condition but not reaching 1.5 times the threshold value; a significant exceedance of the threshold is defined as an indicator value exceeding the threshold corresponding to the warning trigger condition by 1.5 times the threshold value.

7. The method according to claim 6, characterized in that, The blue warning information indicates a risk level of concern, the yellow warning information indicates a risk level of alert, the orange warning information indicates a serious risk level, and the red warning information indicates a particularly serious risk level. The warning information is automatically pushed to the relevant management terminal through a preset communication protocol.

8. The method according to claim 7, characterized in that, The warning trigger condition for the abnormal fluctuation amplitude of the groundwater level is that the water level deviation exceeds 0.3m for 24 consecutive hours; the warning trigger condition for the pollutant migration rate is that it exceeds 0.1m / day; the warning trigger condition for the probability of exceeding the water quality standard is that the chloride ion concentration exceeds 250mg / L or the nitrate nitrogen concentration exceeds 10mg / L.

9. The method according to claim 8, characterized in that, The warning triggering conditions are adaptively adjusted based on seasonal changes, freeze-thaw conditions, and historical extreme event statistics; during the spring snowmelt season, the warning triggering condition for abnormal fluctuations in groundwater level is a water level deviation exceeding 0.5m for 24 consecutive hours, and during the frozen stable period, the warning triggering condition for abnormal fluctuations in groundwater level is a water level deviation exceeding 0.2m for 24 consecutive hours; the freeze-thaw condition is determined by the freezing depth data.

10. The method according to any one of claims 1 to 9, characterized in that, The method further includes: Regularly compare early warning information with actual environmental events to determine the accuracy rate, underreporting rate, and false alarm rate of early warnings. The early warning accuracy, false alarm rate, and false alarm rate are transmitted to the digital twin, and the model parameters or structure of the digital twin are iteratively optimized.

Images