A Proactive Landslide Intervention Decision-Making Method Based on Stochastic Simulation and Dynamic Risk Mapping

Abstract

This invention relates to the field of geological disaster risk identification technology, specifically to a proactive landslide intervention decision-making method based on stochastic simulation and dynamic risk mapping. The method includes the following steps: collecting and fusing multi-source data from the landslide area to construct a landslide instability probability prediction model and a disaster-bearing body loss assessment model, both models containing inherent parameters and input parameters; determining the uncertainty parameters of input fluctuation class and model fluctuation class, and based on their probability distribution laws, using Monte Carlo random simulation sampling and adjusting the model parameters to generate a joint dataset of landslide failure probability and disaster-bearing body economic loss estimates; calculating risk values, drawing a dynamic risk map, and dividing risk level areas; constructing a structured contingency plan library and a rule matching engine, and generating the optimal proactive intervention decision-making scheme through intelligent comparison. This invention fully covers uncertain factors, accurately depicts the overall risk profile, improves the scientific nature and timeliness of decision-making, and is adaptable to complex scenarios such as high-altitude concealed landslides and multiple disaster-bearing bodies overlapping.

Description

Technical Field

[0001] This invention relates to the field of geological disaster risk identification technology, specifically to a proactive landslide intervention decision-making method based on stochastic simulation and dynamic risk maps. Background Technology

[0002] Landslides are a common type of geological disaster in the mountainous areas of southwest my country, which seriously threatens the safety of people's lives and property as well as the operational safety of major projects. With the upgrading of the concept of geological disaster risk management, the traditional "passive response" model can no longer meet the needs of refined and intelligent disaster prevention, and is now gradually transforming into a "proactive intervention" model.

[0003] Existing landslide risk assessment and decision-making technologies suffer from several shortcomings: First, risk assessments often rely on deterministic safety coefficients or single warning thresholds, resulting in an output that can only determine whether a landslide is "stable" or "unstable," failing to quantify the probability of a landslide and the specific losses it may cause. This makes it difficult to accurately depict the full picture of the risk and provide a core basis for decision-making regarding "how great the risk is and how severe the losses are." Second, landslide evolution is influenced by various random factors such as fluctuations in geological parameters, errors in meteorological forecasts, and inherent model biases. However, existing technologies mostly employ deterministic models, ignoring these uncertainties, leading to risk assessment results that are either overly optimistic or conservative, and failing to cover diverse risk scenarios in complex geological environments. Finally, there is a lack of a quantitative mapping relationship between risk levels and intervention measures. The decision-making process relies on human experience and judgment, resulting in problems such as delayed response and insufficient targeted measures. This is especially problematic when facing multiple potential landslide hazard points and various optional intervention measures, making it difficult to achieve scientific comparison and precise implementation.

[0004] Therefore, there is an urgent need for a decision-making method that can simulate and quantify uncertainty, dynamically characterize risk evolution, and accurately match intervention measures, so as to achieve the leap from "risk is knowable" to "risk is controllable" and enhance the initiative and scientific nature of landslide disaster risk prevention and control. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention aims to provide a proactive landslide intervention decision-making method based on stochastic simulation and dynamic risk mapping. This method solves the problems of existing technologies, such as the inability to quantify uncertainty, the disconnect between risk assessment and intervention decision-making, and poor dynamic adaptability. It can achieve accurate quantification of potential landslide risks, dynamic tracking, and intelligent recommendation of intervention measures.

[0006] The basic solution provided by this invention is a landslide proactive intervention decision-making method based on stochastic simulation and dynamic risk maps, comprising the following steps: S1. Collect multi-source data of the landslide area and perform fusion processing. Based on the fused data, construct a landslide instability probability prediction model and a disaster-bearing body loss assessment model. The landslide instability probability prediction model is used to output the landslide failure probability value. The disaster-bearing body loss assessment model is used to output the estimated economic loss of the disaster-bearing body. Both the landslide instability probability prediction model and the disaster-bearing body loss assessment model have their own inherent parameters and input parameters. S2. Determine the uncertainty parameters affecting the output results of the two models. The uncertainty parameters include input fluctuation parameters and model fluctuation parameters. Input fluctuation parameters characterize the fluctuation characteristics of the input parameters of the two models, and model fluctuation parameters characterize the fluctuation characteristics of the inherent parameters of the two models. Based on the probability distribution of each uncertainty parameter, a Monte Carlo random simulation method is used for sampling. Based on the parameter samples obtained from the sampling, the input parameters or inherent parameters of the corresponding landslide instability probability prediction model or disaster-bearing body loss assessment model are adjusted. The two models output landslide failure probability values ​​and disaster-bearing body economic loss estimates. The sampling and calculation process is repeated to obtain multiple sets of landslide failure probability values ​​and disaster-bearing body economic loss estimates, thereby constructing a joint dataset of landslide failure probability and disaster-bearing body economic loss estimates. S3. Based on multiple sets of landslide failure probability values ​​and estimated economic losses of the affected body in the joint dataset, multiple sets of risk values ​​are calculated. Based on the joint dataset and multiple sets of risk values, a dynamic risk map is drawn and risk level areas are divided. S4. Construct a structured contingency plan library and a rule matching engine. The rule matching engine will intelligently compare the real-time dynamic risk map with the structured contingency plan library to generate and output the optimal proactive intervention decision-making plan.

[0007] The principle of this invention is as follows: By collecting and fusing multi-source data from landslide areas, a comprehensive and standardized data foundation is provided for the construction of dual models, ensuring the reliability of model inputs; simultaneously, the inherent parameters and input parameters of the dual models are obtained; considering the randomness of landslide evolution, as well as the errors in the collected data and the differences between different collection points, the uncertain parameters are divided into input fluctuation categories and model fluctuation categories, corresponding to the model's input parameters and inherent parameters, respectively; using the Monte Carlo stochastic simulation method, large-scale sampling covers all possible parameter fluctuation scenarios, and after adjusting the model parameters, multiple sets of landslide failure probability and disaster-bearing body economic loss estimates are output to form a joint dataset, achieving a panoramic characterization of risk; based on the joint dataset, the risk value (the coupled result of multiple sets of landslide failure probabilities and disaster-bearing body economic loss estimates) is calculated, and the correlation between failure probability, economic loss, and risk value is intuitively displayed through a three-dimensional dynamic risk map, while risk level areas are divided, allowing decision-makers to quickly grasp the core risk information; by constructing a structured contingency plan library and a rule matching engine, an automatic comparison mechanism between risk and intervention measures is established, which can output the optimal decision plan without human experience intervention, achieving precise connection between risk quantification and plan matching.

[0008] The beneficial effects of this invention are as follows: By quantifying the fluctuation characteristics of input parameters and inherent model parameters, it comprehensively covers uncertainties such as geology, meteorology, and models. It employs Monte Carlo stochastic simulation to simulate these uncertainties, generating multiple parameter combinations. This allows for the calculation of multiple sets of landslide failure probability values ​​and estimated economic losses to affected bodies, thus comprehensively depicting the risk distribution and providing data support for dynamic risk mapping. This avoids overly optimistic or conservative risk assessments and accurately portrays the overall risk profile. By drawing a three-dimensional dynamic risk map, it clearly presents the probability distribution of risks and the correlation between losses. Compared to traditional static risk maps or single-threshold warnings, this makes it easier for decision-makers to quickly identify high-risk scenarios. The rule-matching engine automatically compares real-time risks with a structured contingency plan library, outputting the optimal intervention plan. This solves the problems of traditional decision-making relying on experience and delayed response, improving the scientific rigor and timeliness of proactive intervention. Furthermore, the solution can be flexibly applied to complex scenarios such as high-altitude concealed landslides and multiple affected bodies, demonstrating strong versatility.

[0009] Furthermore, S1 includes the following steps: S11. The collected multi-source data includes geomechanical data, dynamic monitoring data, disaster-bearing body data, and environmental data; geomechanical data includes shear strength of soil and rock mass; dynamic monitoring data includes pore water pressure, surface displacement rate, and rainfall intensity; disaster-bearing body data includes spatial distribution information, structural attribute information, and vulnerability foundation coefficient of disaster-bearing body; environmental data includes water content of soil and rock mass and topography. S12. Perform fusion processing on multi-source data. The processing operations include data cleaning, spatiotemporal alignment, format unification, outlier removal, and feature extraction to obtain a standardized fusion dataset. S13. Based on a standardized fusion dataset, a machine learning algorithm is used to construct a landslide instability probability prediction model. The input parameters of the model are the pore water pressure change rate, displacement acceleration, cumulative rainfall intensity, and soil moisture content. The pore water pressure change rate is extracted from the collected pore water pressure data, the displacement acceleration is extracted from the collected surface displacement rate, and the cumulative rainfall intensity is obtained by statistically analyzing the collected rainfall intensity. The inherent parameters of the landslide instability probability prediction model include the shear strength of the soil. The model output is the probability value of landslide failure within a preset future time period. S14. Based on a standardized fusion dataset, a disaster-bearing body loss assessment model is constructed. The input parameters of the model are landslide failure intensity, disaster-bearing body exposure, and loss coefficient. The landslide failure intensity is derived from the landslide failure probability value, the disaster-bearing body exposure is calculated by combining the spatial distribution information of the disaster-bearing body with the topography, and the loss coefficient is quantified through the structural attribute information of the disaster-bearing body. The inherent parameters of the model include the basic vulnerability coefficient of the disaster-bearing body. The model outputs the estimated economic loss of the disaster-bearing body through the coupling calculation of the input parameters and inherent parameters.

[0010] Furthermore, the input fluctuation parameters correspond one-to-one with the input parameters of the two models, including the fluctuation values ​​of pore water pressure change rate, displacement acceleration, cumulative rainfall intensity, soil moisture content, and loss coefficient; the model fluctuation parameters correspond one-to-one with the inherent parameters of the two models, including the coefficient of variation of soil shear strength and the vulnerability coefficient of the disaster-bearing body; the probability distribution of each uncertainty parameter is as follows: the coefficient of variation of soil shear strength, the fluctuation values ​​of pore water pressure change rate, displacement acceleration, soil moisture content, and loss coefficient all follow a normal distribution, the fluctuation value of cumulative rainfall intensity follows a uniform distribution, and the vulnerability coefficient of the disaster-bearing body follows a triangular distribution; S2 includes the following steps: S21. Set the total number of iterations for the Monte Carlo random simulation to a positive integer N; S22. Sampling is performed according to the probability distribution of each uncertainty parameter to generate parameter samples; S23. Adjust the input parameters or intrinsic parameters of the corresponding model based on the parameter samples. The input parameters are adjusted using the following formula:

[0011] in, The actual value of the input parameter. The collected values ​​of the input parameters, This refers to the input fluctuation class parameter corresponding to this input parameter; The inherent parameters are adjusted using the following formula:

[0012] in, The actual value of the inherent parameter. These are the collected values ​​of the inherent parameters. These are the model fluctuation class parameters corresponding to this inherent parameter; S24. Based on the actual values ​​of the adjusted input parameters and the actual values ​​of the inherent parameters, the landslide failure probability value and the estimated economic loss of the affected body are calculated through two models. S25. Repeat steps S22 to S24 for a total of N times to obtain a joint dataset containing N sets of landslide failure probability values ​​and estimated economic losses of the affected body.

[0013] By precisely matching the uncertain parameters with the dual-model parameters, the blind adjustment of parameters is avoided, ensuring the scientific nature of the simulation results; the probability distribution law of each uncertain parameter is clarified, and this distribution law is designed based on the physical characteristics of the parameters and industry statistical laws (such as using a log-normal distribution for non-negative parameters), ensuring the rationality of sampling; specific adjustment formulas for input parameters and inherent parameters are set to quantify the impact of parameter fluctuations on the model, making the simulation process quantifiable and verifiable.

[0014] Furthermore, S3 includes the following steps: S31. The risk value is calculated by multiplying the landslide failure probability value and the estimated economic loss of the affected body. Multiple risk values ​​are obtained by sequentially calculating the product of each group of landslide failure probability values ​​and the estimated economic loss of the affected body in the joint dataset. Based on the joint dataset and the risk values, a dynamic risk map is drawn on a three-dimensional coordinate plane. The horizontal axis of the three-dimensional coordinate plane is the landslide failure probability value, the vertical axis is the estimated economic loss of the affected body, and the vertical axis is the risk value. S32. Preset risk classification standards, which include four levels: low risk, medium risk, high risk and extremely high risk. Each level corresponds to a clear risk value range, a landslide damage probability value range and an estimated economic loss value range for the affected body, and the risk level areas are divided on the dynamic risk map.

[0015] The coupling relationship between the probability value of landslide damage and the estimated economic loss of the affected body is quantified to avoid ambiguity in the determination of risk level; the coordinate definition of the three-dimensional dynamic risk map is refined to make risk visualization more practical; at the same time, the four levels and corresponding intervals of the risk classification standard are clearly defined to ensure that the risk level classification is clear and unified, so that decision-makers can not only grasp the quantitative risk data, but also intuitively identify the risk level through the map, thereby improving decision-making efficiency.

[0016] Furthermore, S4 includes the following steps: S41. Construct a structured contingency plan library, which includes engineering mitigation plans, early warning release plans, and emergency evacuation plans. Engineering mitigation plans include the installation of anti-slide piles, the construction of drainage systems, and the setting of intercepting ditches. Early warning release plans determine different release scopes and notification targets based on risk level areas. Emergency evacuation plans include preset evacuation routes, assembly points, and evacuation sequences. S42. Construct a rule matching engine, and the matching rules include the rules for matching risk level with contingency plan level and the rules for matching impact range with evacuation route; S43. Input the real-time risk value, the risk level area in the dynamic risk map and the information of the affected disaster-bearing body into the rule matching engine, compare it with the structured contingency plan library, and generate the optimal proactive intervention decision plan. The proactive intervention decision plan includes at least one of the following: engineering governance plan, early warning release plan, and emergency evacuation plan.

[0017] The structured contingency plan library covers various contingency plans for engineering management, early warning issuance, and emergency evacuation, comprehensively covering all scenarios of landslide intervention and meeting the intervention needs of different risk levels; the optimal decision-making solution is obtained through matching, solving the problem of traditional early warning "only reporting but not making decisions". Attached Figure Description

[0018] Figure 1 This is a flowchart of an embodiment of the landslide proactive intervention decision-making method based on stochastic simulation and dynamic risk mapping of the present invention. Detailed Implementation

[0019] The following detailed description illustrates the specific implementation method: The basic implementation examples are as follows: Figure 1 As shown: A proactive landslide intervention decision-making method based on stochastic simulation and dynamic risk mapping includes the following steps: S1. Collect multi-source data of the landslide area and perform fusion processing. Based on the fused data, construct a landslide instability probability prediction model and a disaster-bearing body loss assessment model. The landslide instability probability prediction model is used to output the landslide failure probability value. The disaster-bearing body loss assessment model is used to output the estimated economic loss of the disaster-bearing body. Both the landslide instability probability prediction model and the disaster-bearing body loss assessment model have their own inherent parameters and input parameters. S2. Determine the uncertainty parameters affecting the output results of the two models. The uncertainty parameters include input fluctuation parameters and model fluctuation parameters. Input fluctuation parameters characterize the fluctuation characteristics of the input parameters of the two models, and model fluctuation parameters characterize the fluctuation characteristics of the inherent parameters of the two models. Based on the probability distribution of each uncertainty parameter, a Monte Carlo random simulation method is used for sampling. Based on the parameter samples obtained from the sampling, the input parameters or inherent parameters of the corresponding landslide instability probability prediction model or disaster-bearing body loss assessment model are adjusted. The two models output landslide failure probability values ​​and disaster-bearing body economic loss estimates. The sampling and calculation process is repeated to obtain multiple sets of landslide failure probability values ​​and disaster-bearing body economic loss estimates, thereby constructing a joint dataset of landslide failure probability and disaster-bearing body economic loss estimates. S3. Based on multiple sets of landslide failure probability values ​​and estimated economic losses of the affected body in the joint dataset, multiple sets of risk values ​​are calculated. Based on the joint dataset and multiple sets of risk values, a dynamic risk map is drawn and risk level areas are divided. S4. Construct a structured contingency plan library and a rule matching engine. The rule matching engine will intelligently compare the real-time dynamic risk map with the structured contingency plan library to generate and output the optimal proactive intervention decision-making plan.

[0020] In this embodiment, the specific process is as follows: S1. Multi-source data fusion and dual-model construction: S11. Data Collection from Multiple Sources: Geomechanical Data: Shear strength of the landslide soil and rock mass obtained through borehole sampling, serving as an inherent parameter for the landslide instability probability prediction model; Dynamic Monitoring Data: Pore water pressure collected by piezometers, surface displacement rate collected by GNSS monitoring stations (multiple monitoring points deployed, covering key areas of the landslide), and hourly rainfall intensity collected by meteorological stations (within 3km of the landslide center); Data on the Threat-Bearing Body: Data collected via UAV LiDAR (point cloud density not less than 50 points / m²). 2 The spatial distribution information of disaster-bearing bodies (such as residential buildings, rural roads, and irrigation pipelines) was collected, and the structural attribute information of disaster-bearing bodies was recorded by field survey (residential buildings are brick-concrete structures, roads are Class IV highways, and pipelines are made of PE material). The basic vulnerability coefficients of disaster-bearing bodies were determined by the "Specification for Vulnerability Assessment of Geological Hazard-Bearing Bodies" (residential buildings 1.0, roads 0.8, pipelines 0.6) as inherent parameters of the disaster-bearing body loss assessment model. Environmental data: Soil moisture content of rock and soil was collected by soil moisture sensors (deployed at the same depth as piezometers), and topography (including slope, slope orientation, etc.) was extracted by high-resolution remote sensing images.

[0021] S12. Data Fusion Processing: Linear interpolation is used to fill in missing values ​​in rainfall data, through 3... The criteria remove abnormal peaks caused by equipment failures in the displacement data, unify the time base as UTC+8 and the spatial base as WGS-84 coordinate system, and form a standardized fusion dataset (the data volume is 12 months of time series data, and the sampling frequency is 1 hour / time).

[0022] S13. Constructing a landslide instability probability prediction model: The random forest algorithm is used, with the training set and test set divided in a 7:3 ratio (the training set is used for model training, and the test set is used for accuracy verification). The model input parameters are the pore water pressure change rate, displacement acceleration, cumulative 72-hour rainfall intensity, and soil moisture content. The pore water pressure change rate is extracted by differential calculation of pore water pressure time series data (unit time interval is 1 hour), the displacement acceleration is extracted by differential calculation of surface displacement rate time series data (unit time interval is 24 hours), and the cumulative rainfall intensity is obtained by statistically summing hourly rainfall intensity. The inherent parameter of the model is the shear strength of the soil. The model output is the probability value of landslide failure in the next 72 hours (value range 0-1).

[0023] S14. Construct a disaster-bearing body loss assessment model: The model input parameters are landslide failure intensity, disaster-bearing body exposure, and loss coefficient; where landslide failure intensity is expressed as "Landslide failure intensity = Landslide failure probability value". The "maximum destructive intensity" was estimated (the maximum destructive intensity was set based on historical landslide disaster survey data for the region); the exposure of the affected body was determined by spatial distribution information combined with topography (e.g., the exposed area of ​​residential buildings is 0.6 km²). 2 The exposed length of the highway is 3.2 km, and the exposed length of the pipeline is 2.8 km. The loss coefficient is determined quantitatively through structural attribute information (e.g., 1.2 for brick-concrete residential buildings, 0.9 for Class IV highways, and 0.7 for PE pipelines; the worse the structural stability, the higher the loss coefficient). The inherent parameters of the model are the basic vulnerability coefficients of the disaster-bearing body (1.0 for residential buildings, 0.8 for highways, and 0.6 for pipelines). The model uses the formula "Estimated economic loss = Damage intensity". Exposure Loss coefficient The vulnerability baseline coefficient is coupled and calculated to ultimately output the estimated economic loss of the disaster-bearing body.

[0024] Due to inherent limitations in landslide monitoring data, on the one hand, the collected data can only reflect the local situation at the monitoring point and cannot fully represent the parameter distribution of the entire landslide body (such as the spatial heterogeneity of shear strength in soil and rock). On the other hand, the data itself contains certain errors due to the influence of equipment accuracy and environmental interference (such as sensor signal fluctuations caused by rainfall). Furthermore, the inherent parameters of the model, obtained through sampling or based on specifications, may also deviate from the actual scenario. Therefore, it is necessary to simulate these uncertainties by inputting fluctuation parameters and model fluctuation parameters, forming multiple parameter combinations, and calculating multiple sets of landslide failure probability values ​​and estimated economic losses to the affected body. This allows for a comprehensive characterization of the risk distribution and provides data support for the creation of dynamic risk maps.

[0025] S2. Uncertainty Simulation and Joint Dataset Construction: S21. Set the total number of iterations for the Monte Carlo random simulation to N=10000 (the number of iterations needs to meet the requirement of sampling coverage of more than 99% of parameter combination scenarios. Industry practice has verified that N greater than or equal to 8000 can balance accuracy and efficiency. This embodiment selects 10000 iterations to further improve reliability).

[0026] S22. Determine the uncertainty parameters and probability distribution, and obtain parameter samples: Input fluctuation parameters (corresponding one-to-one with the input parameters of the dual models) include: pore water pressure change rate fluctuation value, displacement acceleration fluctuation value, cumulative rainfall intensity fluctuation value, soil moisture content fluctuation value, and loss coefficient fluctuation value; model fluctuation parameters (corresponding one-to-one with the inherent parameters of the dual models) include: soil shear strength variation coefficient and disaster-bearing body vulnerability foundation coefficient. Probability distribution law: The coefficient of variation of shear strength of soil and rock mass follows a normal distribution N(1, 0.1). 2 The shear strength variation coefficient of the soil and rock mass has a normal distribution with a mean of 1, ensuring that subsequent adjustments are made around the baseline of the collected values, and the adjusted results are concentrated near the collected values. The fluctuation values ​​of pore water pressure change rate, displacement acceleration, soil and rock mass water content, and loss coefficient all follow a normal distribution N(0, 0.15). 2 The normal distribution naturally covers positive and negative fluctuations, fitting the actual scenario where "deviations can be high or low." The mean of the normal distribution for each fluctuation value is set to 0, ensuring that the fluctuation is near 0, meaning the final adjusted result is close to the collected value. Variance control keeps the fluctuation amplitude within a concentrated range to avoid extreme deviations. The cumulative rainfall intensity fluctuation value follows a uniform distribution U(-0.15, 0.15) (rainfall prediction error has no obvious bias, and the uniform distribution can fully cover the error range). The basic coefficient of vulnerability of disaster-bearing bodies follows a triangular distribution with a lower limit of 0.8, an upper limit of 1.2, and a peak value of 1.0 (the triangular distribution can highlight the probability of fluctuations near the benchmark value, while clearly defining the fluctuation boundary, which conforms to the fluctuation characteristics of the standard parameters). Sampling was performed according to the above distribution pattern to generate 10,000 sets of parameter samples (each set of samples contains 7 uncertain parameters, corresponding to 5 parameters of input fluctuation class and 2 parameters of model fluctuation class respectively).

[0027] S23, Parameter Adjustment: Input parameter adjustment: using formula Adjustments were made, including The actual value of the input parameter. The collected values ​​of the input parameters, This refers to the input fluctuation class parameter corresponding to this input parameter; inherent parameter adjustment: using the formula Adjustments were made, including The actual value of the inherent parameter. These are the collected values ​​of the inherent parameters. These are the model fluctuation parameters corresponding to this inherent parameter.

[0028] S24. Model Calculation: The actual values ​​of the input parameters after adjustment by a set of parameter samples and the actual values ​​of the inherent parameters are input into the landslide instability probability prediction model and the disaster-bearing body loss assessment model to obtain the corresponding landslide failure probability value and disaster-bearing body economic loss estimate data.

[0029] S25. Repeat steps S22 to S24 a total of 10,000 times to form a joint dataset of landslide failure probability values ​​and estimated economic losses of the affected body. The joint dataset is arranged in descending order of landslide failure probability values ​​to facilitate subsequent risk value calculation and map drawing.

[0030] S3. Dynamic Risk Mapping and Risk Classification: S31. Calculate the risk value: Assume "Risk value = Landslide failure probability value". The calculation of the "estimated economic loss" is performed; the risk values ​​of 10,000 sets of data in the joint dataset are calculated sequentially to obtain multiple risk values; a dynamic risk map is drawn: a three-dimensional coordinate plane is constructed using Python (horizontal axis: probability of destruction 0-1, vertical axis: estimated economic loss, vertical axis: risk value), and a probability density cloud map is drawn (using a scatter plot overlaid with a density heatmap, the darker the color, the higher the probability of occurrence of the combination of (probability of destruction, estimated economic loss, risk value), and the highest density area corresponds to the most likely risk scenario).

[0031] S32. Risk Classification: Preset four-level risk standards (based on the "Technical Specification for Geological Hazard Risk Assessment" and regional disaster prevention capacity settings): Low Risk: Risk Value 500, probability of destruction 0.2, estimated economic loss 2500: This risk level poses a minor threat to the affected area and requires no emergency intervention, only routine monitoring; Medium risk: 500 Risk Value 1500, 0.2 Probability of destruction 0.5, 2500 Estimated economic loss 2800: This range presents a risk that requires attention; monitoring should be strengthened and intervention measures prepared. High risk: 1500. Risk Value 2500, 0.5 Probability of destruction 0.8, 2800 Estimated economic loss 3000: This risk level requires immediate intervention to reduce the risk level; Extremely high risk: Risk value 2500, probability of destruction 0.8, estimated economic loss 3000 indicates an extremely high risk level, requiring emergency evacuation and remediation. The dynamic risk map uses solid lines of blue, yellow, orange, and red to delineate risk level zones, with the solid lines representing the critical risk levels, facilitating intuitive identification of the current risk zone.

[0032] S4. Intelligent Decision-Making Solution Generation: S41. Construct a structured contingency plan library: Engineering mitigation plan: This includes measures such as anti-slide pile installation (1.0m diameter, 12m length, 6m spacing), drainage system construction (0.6m pipe diameter, 3% slope), and intercepting ditch installation (0.8m width, 1.0m depth), with selection of mitigation measures based on different risk levels. Early warning release plan: The release scope and notification recipients are set according to regional differences in risk levels. For example, low risk corresponds to a blue warning (release scope within 1km of the landslide-affected area, notification recipients are village collectives and monitoring personnel), medium risk corresponds to a yellow warning (release scope within 2km of the landslide-affected area, notification recipients are affected residents, village collectives, and township emergency departments), and high risk... The corresponding orange alert (issued within 3km of the landslide-affected area, notifying affected residents, village collectives, townships, and county-level emergency departments) and the extremely high risk corresponding red alert (issued within 5km of the landslide-affected area, notifying affected residents, relevant units, and municipal and county-level emergency management departments); emergency evacuation plans: such as Route 1 (from the residential area to the assembly point on the flat open space to the north, route length 1.2km, evacuation priority is given to the elderly, weak, sick, and disabled, then ordinary residents), Route 2 (from residents along the highway to the highway service area assembly point, route length 1.5km, evacuation order is orderly evacuation in sections according to residential road sections).

[0033] S42. Constructing a rule matching engine: Risk level and contingency plan level correspondence rules: Match the appropriate contingency plan type according to the risk level. Low risk is matched only with early warning release contingency plan; medium risk is matched with early warning release contingency plan and light engineering treatment contingency plan (such as intercepting ditch setting); high risk is matched with early warning release contingency plan, medium engineering treatment contingency plan (such as drainage system construction) and emergency evacuation contingency plan; extremely high risk is matched with early warning release contingency plan, heavy engineering treatment contingency plan (such as anti-slide pile layout and drainage system construction) and emergency evacuation contingency plan. Matching rules between affected areas and evacuation routes: Evacuation routes are automatically matched based on the spatial distribution information of affected disaster-bearing bodies. Route 1 is prioritized for densely populated residential areas, Route 2 is prioritized for areas along highways, and both routes are activated simultaneously for mixed distribution areas to ensure targeted evacuation.

[0034] S43. Decision Matching and Solution Generation: Based on real-time monitoring data, the standardized fusion dataset is updated. Real-time risk values ​​are calculated through steps S2-S3. For example, if the dynamic risk map shows the current area is at a medium risk level, and the affected disaster-bearing area is a densely populated residential area (0.6 km²), then... 2 The data includes a 1.0km section of road and a road segment. This information is input into a rule-matching engine and intelligently compared with a structured contingency plan database: Based on the rule of matching risk level with contingency plan level, it matches the early warning release plan and the engineering treatment plan for intercepting ditches corresponding to medium-risk situations; based on the rule of matching impact range with evacuation routes, it matches Route 1 corresponding to densely populated residential areas and Route 2 corresponding to road segments. Finally, an optimal proactive intervention decision-making plan is generated, which includes a combination of three types of contingency plans: early warning release, engineering treatment, and emergency evacuation.

[0035] This embodiment is data-driven, using Monte Carlo simulation to accurately cover multiple uncertainties such as geological parameters, monitoring data, and model biases. It leverages dynamic risk maps to achieve a panoramic and intuitive presentation of risks, and then establishes a precise mapping between risks and intervention measures through a rule-matching engine. This eliminates the reliance on human experience in traditional decision-making, forming a scientific and closed-loop proactive landslide intervention system. This technical solution is not only suitable for complex scenarios such as high-altitude, concealed landslides and multiple overlapping disaster-bearing bodies in the southwestern mountainous areas, but can also be flexibly transferred to risk management of other regions and other types of landslides by adjusting data collection types, parameter distribution patterns, risk classification standards, and contingency plan database content. It can effectively improve the accuracy of landslide risk assessment, the timeliness of decision-making response, reduce the blindness and cost of disaster prevention and mitigation, and provide solid technical support for proactive intervention in geological disasters.

[0036] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A landslide proactive intervention decision-making method based on stochastic simulation and dynamic risk mapping, characterized in that, Includes the following steps: S1. Collect multi-source data of the landslide area and perform fusion processing. Based on the fused data, construct a landslide instability probability prediction model and a disaster-bearing body loss assessment model. The landslide instability probability prediction model is used to output the landslide failure probability value. The disaster-bearing body loss assessment model is used to output the estimated economic loss of the disaster-bearing body. Both the landslide instability probability prediction model and the disaster-bearing body loss assessment model have their own inherent parameters and input parameters. S2. Determine the uncertainty parameters affecting the output results of the two models. The uncertainty parameters include input fluctuation parameters and model fluctuation parameters. Input fluctuation parameters characterize the fluctuation characteristics of the input parameters of the two models, and model fluctuation parameters characterize the fluctuation characteristics of the inherent parameters of the two models. Based on the probability distribution of each uncertainty parameter, a Monte Carlo random simulation method is used for sampling. Based on the parameter samples obtained from the sampling, the input parameters or inherent parameters of the corresponding landslide instability probability prediction model or disaster-bearing body loss assessment model are adjusted. The two models output landslide failure probability values ​​and disaster-bearing body economic loss estimates. The sampling and calculation process is repeated to obtain multiple sets of landslide failure probability values ​​and disaster-bearing body economic loss estimates, thereby constructing a joint dataset of landslide failure probability and disaster-bearing body economic loss estimates. S3. Based on multiple sets of landslide failure probability values ​​and estimated economic losses of the affected body in the joint dataset, multiple sets of risk values ​​are calculated. Based on the joint dataset and multiple sets of risk values, a dynamic risk map is drawn and risk level areas are divided. S4. Construct a structured contingency plan library and a rule matching engine. The rule matching engine will intelligently compare the real-time dynamic risk map with the structured contingency plan library to generate and output the optimal proactive intervention decision-making plan.

2. The landslide proactive intervention decision-making method based on stochastic simulation and dynamic risk mapping according to claim 1, characterized in that, S1 includes the following steps: S11. The collected multi-source data includes geomechanical data, dynamic monitoring data, disaster-bearing body data, and environmental data; geomechanical data includes shear strength of soil and rock mass; dynamic monitoring data includes pore water pressure, surface displacement rate, and rainfall intensity; disaster-bearing body data includes spatial distribution information, structural attribute information, and vulnerability foundation coefficient of disaster-bearing body; environmental data includes water content of soil and rock mass and topography. S12. Perform fusion processing on multi-source data. The processing operations include data cleaning, spatiotemporal alignment, format unification, outlier removal, and feature extraction to obtain a standardized fusion dataset. S13. Based on a standardized fusion dataset, a machine learning algorithm is used to construct a landslide instability probability prediction model. The input parameters of the model are the pore water pressure change rate, displacement acceleration, cumulative rainfall intensity, and soil moisture content. The pore water pressure change rate is extracted from the collected pore water pressure data, the displacement acceleration is extracted from the collected surface displacement rate, and the cumulative rainfall intensity is obtained by statistically analyzing the collected rainfall intensity. The inherent parameters of the landslide instability probability prediction model include the shear strength of the soil. The model output is the probability value of landslide failure within a preset future time period. S14. Based on a standardized fusion dataset, a disaster-bearing body loss assessment model is constructed. The input parameters of the model are landslide failure intensity, disaster-bearing body exposure, and loss coefficient. The landslide failure intensity is derived from the landslide failure probability value, the disaster-bearing body exposure is calculated by combining the spatial distribution information of the disaster-bearing body with the topography, and the loss coefficient is quantified through the structural attribute information of the disaster-bearing body. The inherent parameters of the model include the basic vulnerability coefficient of the disaster-bearing body. The model outputs the estimated economic loss of the disaster-bearing body through the coupling calculation of the input parameters and inherent parameters.

3. The landslide proactive intervention decision-making method based on stochastic simulation and dynamic risk mapping according to claim 2, characterized in that, The input fluctuation parameters correspond one-to-one with the input parameters of the two models, including the fluctuation values ​​of pore water pressure change rate, displacement acceleration, cumulative rainfall intensity, soil moisture content, and loss coefficient. The model fluctuation parameters correspond one-to-one with the inherent parameters of the two models, including the coefficient of variation of soil shear strength and the vulnerability coefficient of the disaster-bearing body. The probability distribution of each uncertainty parameter is as follows: the coefficient of variation of soil shear strength, the fluctuation values ​​of pore water pressure change rate, displacement acceleration, soil moisture content, and loss coefficient all follow a normal distribution; the fluctuation value of cumulative rainfall intensity follows a uniform distribution; and the vulnerability coefficient of the disaster-bearing body follows a triangular distribution. S2 includes the following steps: S21. Set the total number of iterations for the Monte Carlo random simulation to a positive integer N; S22. Sampling is performed according to the probability distribution of each uncertainty parameter to generate parameter samples; S23. Adjust the input parameters or intrinsic parameters of the corresponding model based on the parameter samples. The input parameters are adjusted using the following formula: in, The actual value of the input parameter. The collected values ​​of the input parameters, This refers to the input fluctuation class parameter corresponding to this input parameter; The inherent parameters are adjusted using the following formula: in, The actual value of the inherent parameter. These are the collected values ​​of the inherent parameters. These are the model fluctuation class parameters corresponding to this inherent parameter; S24. Based on the actual values ​​of the adjusted input parameters and the actual values ​​of the inherent parameters, the landslide failure probability value and the estimated economic loss of the affected body are calculated through two models. S25. Repeat steps S22 to S24 for a total of N times to obtain a joint dataset containing N sets of landslide failure probability values ​​and estimated economic losses of the affected body.

4. The landslide proactive intervention decision-making method based on stochastic simulation and dynamic risk mapping according to claim 3, characterized in that, S3 includes the following steps: S31. The risk value is calculated by multiplying the landslide failure probability value and the estimated economic loss of the affected body. Multiple risk values ​​are obtained by sequentially calculating the product of each group of landslide failure probability values ​​and the estimated economic loss of the affected body in the joint dataset. Based on the joint dataset and the risk values, a dynamic risk map is drawn on a three-dimensional coordinate plane. The horizontal axis of the three-dimensional coordinate plane is the landslide failure probability value, the vertical axis is the estimated economic loss of the affected body, and the vertical axis is the risk value. S32. Preset risk classification standards, which include four levels: low risk, medium risk, high risk and extremely high risk. Each level corresponds to a clear risk value range, a landslide damage probability value range and an estimated economic loss value range for the affected body, and the risk level areas are divided on the dynamic risk map.

5. The landslide proactive intervention decision-making method based on stochastic simulation and dynamic risk mapping according to claim 4, characterized in that, S4 includes the following steps: S41. Construct a structured contingency plan library, which includes engineering mitigation plans, early warning release plans, and emergency evacuation plans. Engineering mitigation plans include the installation of anti-slide piles, the construction of drainage systems, and the setting of intercepting ditches. Early warning release plans determine different release scopes and notification targets based on risk level areas. Emergency evacuation plans include preset evacuation routes, assembly points, and evacuation sequences. S42. Construct a rule matching engine, and the matching rules include the rules for matching risk level with contingency plan level and the rules for matching impact range with evacuation route; S43. Input the real-time risk value, the risk level area in the dynamic risk map and the information of the affected disaster-bearing body into the rule matching engine, compare it with the structured contingency plan library, and generate the optimal proactive intervention decision plan. The proactive intervention decision plan includes at least one of the following: engineering governance plan, early warning release plan, and emergency evacuation plan.

Images