Prediction method of landslide risk level considering the influence of previous rainfall and soil moisture state

Abstract

This invention provides a landslide risk level prediction method that considers the impact of prior rainfall and soil moisture status. The method includes: acquiring soil moisture data from multiple soil layers and making it dimensionless; simultaneously acquiring rainfall data and constructing an input-response data structure by combining rainfall and multiple moisture level index vectors at a unified time scale; introducing a most unfavorable layer control mechanism to obtain representative soil moisture levels; and then constructing a soil moisture state function. S * ( t ); Calculate the impact of short-term and long-term rainfall separately, and then weight the two to determine the impact of previous rainfall. Pe(t) Construct state composite modulation coefficients α ( t ),Will Pe(t) Multiply α ( t ), to determine the effective impact of pre-rainfall after modulation; with S * ( t This involves constructing a comprehensive risk index to determine the landslide risk level. This enables quantitative identification of landslide risk levels, improving the accuracy and engineering applicability of landslide risk assessment.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring and early warning technology, specifically to a method for predicting landslide risk levels that considers the impact of previous rainfall and soil moisture status. This method comprehensively utilizes real-time rainfall data, the amount of previous rainfall, and soil moisture status information to classify and determine the landslide risk under different hydrological conditions, and is applicable to the risk assessment and early warning of rainfall-induced landslides. Background Technology

[0002] Landslides are one of the most common and destructive geological hazards in mountainous and hilly areas, and rainfall is a crucial external factor inducing natural landslides. Extensive engineering practice and statistical analysis show that landslides are not directly triggered by a single heavy rainfall event, but are often closely related to the evolution of slope moisture content caused by accumulated rainfall in the preceding period. Continuous infiltration of preceding rainfall leads to increased soil moisture content, decreased matrix suction, and accumulated pore water pressure, thereby reducing soil shear strength and increasing the probability of landslide instability. Therefore, in the early warning process for rainfall-induced landslides, how to simultaneously characterize the cumulative effects of preceding rainfall and the dynamic changes in soil moisture, and accordingly achieve a reasonable assessment of landslide risk levels, has become a critical issue.

[0003] In existing studies on rainfall-induced landslide early warning, a common approach is to construct empirical or statistical threshold models based on parameters such as rainfall intensity, duration, and cumulative rainfall, and to some extent, incorporate antecedent rainfall indicators to characterize the slope's moisture background. However, the effectiveness of rainfall in triggering landslides is not entirely determined by the amount of rainfall itself. The soil moisture state (i.e., soil moisture content or degree of moisture) at the time of the warning often has a significant controlling effect on the critical conditions for landslide occurrence. Abraham et al. (Usage of antecedent soil moisture for improving the performance of rainfall thresholds for landslide early warning) pointed out that when the soil moisture level is low in the early stages of a landslide, only strong rainfall events can trigger a landslide. However, when the landslide is moist or close to saturation, even relatively weak rainfall events can induce a landslide. This suggests that incorporating antecedent soil moisture information into the rainfall threshold system helps improve the sensitivity and reliability of the early warning criteria. Meanwhile, Han Zong et al. (Ensemble Predictions of Rainfall-Induced Landslide Risk under Climate Change in China Integrating Antecedent Soil-Wetness Factors) incorporated antecedent soil moisture levels into their risk assessment framework in their study on rainfall-induced landslide risk prediction. This approach more reasonably characterizes the spatiotemporal distribution and variation of landslide risk, further illustrating the crucial supporting role of soil moisture status in rainfall-induced landslide risk assessment and early warning modeling. Therefore, from the perspective of constructing early warning criteria, using only a fixed rainfall threshold or a single antecedent rainfall accumulation index is insufficient to fully reflect the differences in landslide triggering conditions under different initial moisture states. It is necessary to couple the "antecedent rainfall accumulation effect" with the "evolutionary characteristics of soil moisture status" in the prediction method to improve the adaptability and stability of the early warning model under different moisture backgrounds.

[0004] Therefore, it is necessary to propose a landslide risk level prediction method that can simultaneously characterize the impact of previous rainfall and soil moisture status. By establishing a modulation mechanism of soil moisture status on the effect of previous rainfall, a graded early warning output of landslide risk from low to high can be achieved. Furthermore, the representativeness of soil moisture at multiple depths should be taken into account to improve the stability and scalability of the early warning method under long-term engineering operation conditions. Summary of the Invention

[0005] To address the shortcomings of existing early warning methods in coupling the cumulative effect of prior rainfall with the soil moisture state, this invention aims to provide a landslide risk level prediction method that considers both prior rainfall and soil moisture state. By acquiring rainfall monitoring data and constructing a rainfall process, and combining multi-layer soil moisture data, the moisture level of the most unfavorable layer is extracted as a representative moisture index. A threshold classification is then used to identify the soil moisture state, characterizing the overall moisture conditions of the landslide. Furthermore, a time-decay weighted cumulative calculation of the prior rainfall impact is employed, and the moisture state is incorporated for hierarchical control and continuous correction to obtain the modulated effective rainfall impact, achieving a coupled expression between rainfall action and soil response. Finally, a risk assessment model is constructed based on the normalized rainfall index and the representative soil moisture level, enabling quantitative identification of landslide risk levels. This dynamically reflects the disaster-causing effect of rainfall under different moisture backgrounds, improving the accuracy and engineering applicability of landslide risk assessment.

[0006] The technical solution adopted by the present invention to solve the aforementioned technical problem is as follows: A method for predicting landslide risk levels that considers the impact of prior rainfall and soil moisture conditions includes the following: For the landslide area to be studied, soil moisture sensors are deployed at different depths to acquire soil moisture data of multiple soil layers. The data is then dimensionless to obtain a multi-layer moisture index vector. Through the deployment of multiple sensors, the range from the shallow layer to the potential sliding surface is covered, thereby reflecting the vertical distribution characteristics of soil water content during rainfall infiltration. Simultaneously acquire rainfall data, and then... R ( t ) and multi-layered humidity index vector S m ( t They form an input-response data structure on a unified time scale to characterize the external driving process and internal response state of the system. Based on the multi-layered wetting index, a most unfavorable layer control mechanism is introduced to extract representative wetting indices from the multi-layered wetting index, thereby obtaining the representative wetting degree of the soil. S * ( t ); To obtain representative soil moisture levels S * ( t Then, the continuously changing representative moisture level is transformed into a discrete state variable, and a soil moisture state function is constructed:

[0007] in, State ( t () indicates time tThe soil moisture state is represented by a finite set of discrete states; it is further divided into multiple levels based on the degree of moisture.

[0008] The corresponding division rule is: when When, it is defined as a low humidity state. S 1; when When, it is defined as a medium humidity state. S 2; when When, it is defined as a high humidity state. S 3; T 1 and T 2 is the threshold for classifying wet conditions, which satisfies... ; Short-term and long-term rainfall time windows are set, and the short-term rainfall impact is calculated using a time-decay weighted cumulative method. P short ( t ) and the impact of long-term rainfall P long ( t The weighted average of the two factors determines the impact of previous rainfall. Pe(t) ; Construct state composite modulation coefficients to incorporate the impact of the preceding rainfall. Pe(t) Multiply by the corresponding state composite modulation coefficient to determine the modulated effective anterior rainfall impact. The state composite modulation coefficient is determined by the following formula:

[0009] in, α ( t (time) t The state composite modulation coefficient; α State(t) Due to the soil moisture state State ( t The basic modulation coefficient is determined by λ; λ is the continuous modulation coefficient, used to control the influence of the representative soil moisture level on the modulation result. The modulated effective anterior rainfall impact Normalization was performed to obtain the normalized rainfall index. P ’ ( t Construct a comprehensive risk index:

[0010] in, I(t) For a moment t The comprehensive risk index; w 1> w2 is the weighting coefficient, which satisfies w 1+ w 2 = 1; Comprehensive risk index I(t) The landslide risk level is determined by comparing it with the preset risk classification threshold.

[0011] Furthermore, a time-decay weighted cumulative method is used to calculate the short-term rainfall impact of rainfall. P short ( t ) and the impact of long-term rainfall P long ( t The weighted average of the two factors determines the impact of previous rainfall. Pe(t) The process is; Set short-term and long-term rainfall time windows, and calculate the short-term rainfall impact using the following formula. P short ( t ) and the impact of long-term rainfall P long ( t ):

[0012] in ,R(ti) Let be the rainfall at the i-th time step before the current time t; n be the number of historical rainfall steps; ω(i) be the time decay weight function; i 0 This refers to the start time of either a short-term or long-term rainfall window. The weighting function uses an exponential decay form:

[0013] in, t The time decay coefficient controls the duration of the rainfall's impact. The impact of previous rainfall is calculated using the following formula:

[0014] in, β This is a weighting coefficient used to adjust the relative contributions of short-term and long-term rainfall.

[0015] Furthermore, β The value is 0.6–0.8; the correlation between the time decay coefficient and the rainfall impact time window L is as follows: .

[0016] Furthermore, S * ( t The value range is [0,1]; the basic modulation coefficientα State(t) The values ​​were set as increasing sequences according to different humidity levels: S1 for low humidity was 0.8, S2 for medium humidity was 1.0, and S3 for high humidity was 1.3. The continuous modulation coefficient λ was determined based on the hydrological characteristics of the soil in the study area or historical monitoring data, and its value ranged from 0.5 to 1.5.

[0017] Furthermore, the risk assessment function:

[0018] in, Risk ( t () indicates time t The landslide risk level; R 1. R 2 and R 3 represents low risk, medium risk, and high risk, respectively; I 1 and I 2 is the risk classification threshold, which meets the requirements. I 1< I 2.

[0019] Furthermore, if the comprehensive risk index determines the state to be high-risk, a wet state constraint is introduced to further assess the risk. The risk level will be downgraded from R3 to medium risk. R 2, of which T 1 represents the threshold for low humidity.

[0020] Furthermore, the process of obtaining multi-layered humidity index vectors through dimensionless transformation is as follows: Soil moisture content in multi-layered soil structures is obtained using soil moisture sensors deployed at different depths. i j ( t ),in, i j ( t ) indicates the first j Layered soil at time t Soil moisture, expressed in volume ratio (m³). 3 / m 3 , j This indicates the layer number representing the depth at which the sensor is buried. k To monitor the number of layers; If soil moisture i j ( t If there are excess values ​​or abrupt jumps, smoothing should be performed according to the following formula:

[0021] in It is the first after smoothing. jSoil moisture content in the layer; For the original first j Soil moisture content in the layer; c Smoothing coefficient, 0 < c <1; This is the smoothed result from the previous time step; Soil moisture value after smoothing treatment Normalize according to the following formula: in, Sm j ( t ) indicates the first j Layered soil at time t The degree of humidity; i min,j and i max,j These are the minimum and maximum moisture contents of the soil layer during the observation period, used to characterize the range of change of the soil layer from a dry state to a saturated state. The multi-layered humidity index is constructed into a vector form: ; in, This is a vector of multi-layered humidity indexes.

[0022] Compared with the prior art, the beneficial effects of the technical solution of the present invention are: Compared to existing landslide early warning methods that rely solely on rainfall amount, rainfall intensity-duration thresholds, or empirical antecedent rainfall indicators, this invention, in addition to rainfall input information, further incorporates multi-layered soil moisture states to modulate the effects of rainfall. This allows for a more complete characterization of the evolutionary process from rainfall infiltration to soil response to risk assessment. Existing domestic and international research generally agrees that while traditional rainfall threshold methods are widely used, they often fail to adequately reflect the differences in hydrological conditions within landslides, leading to high false alarm rates and insufficient spatial adaptability. This invention, by simultaneously incorporating soil moisture and effective antecedent rainfall into the early warning model, further enhances its identification performance. Specifically, (1) This invention introduces the soil moisture state into the calculation process of the impact of early rainfall. By constructing a state modulation mechanism, the effect of rainfall can be dynamically changed with the soil moisture level. This overcomes the shortcomings of the existing technology, which only makes judgments based on rainfall or empirical thresholds and ignores the internal moisture state of the soil. This allows for a more realistic reflection of the triggering mechanism of rainfall-induced landslides and improves the accuracy of risk identification.

[0023] (2) In actual landslide monitoring and early warning, the identification of wet conditions is not simply based on soil moisture thresholds. First, the water content of landslide soil varies significantly at different depths, and landslide triggering is often controlled by the water content near the potential sliding surface. If thresholds are defined based solely on a single depth or surface moisture, it can easily lead to distorted identification of wet conditions. Second, soil types and pore structures vary greatly in different regions, making it difficult to use absolute moisture content thresholds uniformly. Standardization is required; otherwise, the wet condition classification lacks regional adaptability. Furthermore, soil moisture changes are continuous. When moisture approaches the threshold boundary, simple interval division can easily cause abrupt changes in the state, leading to frequent fluctuations in risk levels, which is detrimental to the stable operation of the early warning system. To address this issue, this application first standardizes soil moisture content to obtain a dimensionless multi-layered moisture index vector, thereby improving the consistency of moisture state classification under different regions and soil types. Second, when multi-depth moisture monitoring data are available, the representative moisture level of the soil is determined by controlling the most unfavorable layer to improve the representativeness of the moisture state of potential sliding surfaces. This application does not simply treat moisture state as a simple classification variable, but introduces it into the calculation of the impact of previous rainfall through modulation coefficients, realizing the dynamic modulation of the moisture state on the triggering ability of rainfall, thereby forming a stable and physically meaningful landslide risk assessment mechanism.

[0024] (3) In actual landslide monitoring, soil moisture usually shows significant differences at different depths. In natural landslides, the surface soil is greatly affected by rainfall and changes rapidly; the deep soil is affected by rainfall more slowly and changes less rapidly. Landslides are often controlled by the water content near the potential sliding surface. Therefore, if only soil moisture at a single depth is used or if moisture at multiple depths is simply averaged, it is easy to lead to bias in the identification of the wet state, thus affecting the landslide risk assessment results. In addition, in actual engineering monitoring, the number and depth configuration of moisture sensors deployed in different landslide monitoring systems vary. If the model relies on fixed depth data, it is difficult to apply to real-time early warning applications under multi-source monitoring conditions. This invention analyzes the moisture content of multi-layered soil and adopts the most unfavorable layer control method. It selects the humidity value that has the greatest impact on landslide stability as the control index, that is, extracts the layer closest to saturation at each depth as the representative moisture content, and constructs discrete moisture state variables. This avoids the weakening of key hazard information by traditional multi-layer averaging or single-point measurement, and can more sensitively identify potential sliding control layers, thereby improving the ability to characterize key conditions for landslide instability and more reasonably reflecting the comprehensive influence of soil water content at different depths.

[0025] (4) This application does not directly change the rainfall threshold through the wet state, but constructs a wet state modulation mechanism, takes the soil wet state as a control variable, and dynamically adjusts the amount of previous rainfall through the state composite modulation coefficient, thereby forming the modulated effective amount of previous rainfall.

[0026] (5) The present invention constructs a dual-index coupled risk judgment model based on the modulated effective rainfall impact and the representative soil moisture degree. It can simultaneously consider the rainfall driving intensity and the soil internal moisture response. Compared with the technical solution that only uses a single rainfall threshold or a single humidity index, it can improve the sensitivity of identifying high-risk states when the soil is close to saturation and suppress false alarms caused by short-term heavy rainfall when the soil is relatively dry, thereby improving the accuracy, stability and engineering applicability of the risk classification results.

[0027] (6) Existing methods incorporate soil moisture indices into the threshold fitting process, mainly focusing on the statistical matching relationship between "deformation events and rainfall thresholds." A unified mechanism has not yet been established that uses soil moisture status as a control variable to drive changes in the mode of action of previous rainfall and further realizes dynamic risk level classification. This application constructs the impact of previous rainfall through a time-decay weighted cumulative method, and further achieves multi-timescale decomposition and expression by combining short-scale and long-scale impacts, replacing the direct use of fixed-time-window cumulative rainfall. This reflects the continuous memory characteristics of rainfall impact decaying over time, realizing multi-timescale decomposition and expression of the impact of previous rainfall and the modulation relationship of soil moisture status on the contribution of previous rainfall.

[0028] (7) This application constructs the modulation relationship of the contribution of wet state to previous rainfall by identifying the wet state of the slope and introducing a modulation coefficient. α State This approach uses soil moisture status as a control variable to hierarchically control and continuously correct the impact of previous rainfall. Under different moisture conditions, the contribution of previous rainfall to landslide risk is dynamically amplified or weakened. By coupling the modulated effective rainfall impact with representative moisture levels for risk grading, the dynamic changes in risk thresholds under different moisture backgrounds are achieved. This describes the decay process of rainfall impact over time, making historical rainfall information more consistent with actual hydrological processes. Through a moisture status modulation mechanism, the impact of previous rainfall can dynamically change with the slope moisture level, thus more accurately reflecting landslide triggering conditions under different hydrological backgrounds. The state-driven decision structure allows the risk assessment results to dynamically adjust with changes in slope hydrological status, improving prediction reliability. Attached Figure Description

[0029] Figure 1 A flowchart illustrating a method for predicting landslide risk levels that considers the impact of prior rainfall and soil moisture conditions.

[0030] Figure 2 This is a schematic diagram of the extraction and modulation mechanism in a moist state.

[0031] Figure 3This is a schematic diagram of the comprehensive landslide risk assessment based on rainfall and soil moisture in Example 1. Detailed Implementation

[0032] The present invention will be further explained below with reference to the embodiments and accompanying drawings, but this is not intended to limit the scope of protection of this application.

[0033] Soil moisture is a dynamic process, influenced by various factors such as rainfall and evaporation, and its evolution is quite complex. A key challenge is how to reflect the relationship between soil moisture status and the impact of previous rainfall in real time under different hydrological conditions. Without a reasonable mechanism, this dynamic change may lead to inaccurate estimations of the impact of previous rainfall, thus affecting landslide risk prediction. The impact of previous rainfall is not a simple linear relationship. The degree to which rainfall affects landslide stability changes significantly with changes in soil moisture. In particular, the triggering effect of the same amount of rainfall on landslides may vary considerably under different moisture conditions. Soil moisture status has a significant impact on landslide risk assessment; different moisture states lead to adjustments in risk assessment thresholds. Directly incorporating moisture status into rainfall threshold models presents certain technical difficulties. First, simply dividing different thresholds according to humidity ranges can easily lead to abrupt changes in risk levels near the humidity threshold, affecting the stability of early warnings. Second, the influence of soil moisture on rainfall infiltration and pore water pressure development has obvious nonlinear characteristics, and the triggering ability of the same rainfall conditions varies greatly under different humid backgrounds, making it difficult to express with a single fixed threshold. Third, if variables such as rainfall intensity, rainfall duration, previous rainfall, and soil moisture are included in the threshold model at the same time, it is easy to form a high-dimensional threshold structure, which increases the difficulty of parameter determination and reduces the feasibility of engineering applications.

[0034] This invention modulates the impact of antecedent rainfall by using soil moisture state as a control variable. Specifically, different soil moisture states (dry, wet, near saturation) affect the contribution of antecedent rainfall to landslide triggering. In a dry state, antecedent rainfall contributes less to the final landslide formation; while in a wet or near-saturated state, the contribution of antecedent rainfall is amplified. This invention introduces a multi-depth soil moisture control layer mechanism to improve the representativeness of moisture state identification. By selecting soil moisture data at different depths, it more comprehensively reflects the moisture state of each layer of the slope, and by comparison, it ensures that the most unfavorable moisture layer is selected as the representative moisture level of the soil, thereby more accurately adjusting the contribution of antecedent rainfall. This multi-depth moisture control layer mechanism greatly enhances the responsiveness to changes in soil moisture state and improves the stability and adaptability of the model. This invention dynamically switches with soil moisture state, solving the problems of poor adaptability and false alarms / missed alarms caused by fixed thresholds in traditional models. It can automatically adjust prediction results according to different moisture backgrounds, significantly improving the sensitivity and stability of early warning.

[0035] This application does not simply recalculate previous rainfall, but rather incorporates soil moisture status into the calculation and risk assessment process of previous rainfall impact, forming an integrated method of "cumulative rainfall effect + soil moisture status + dynamic modulation + dual-index assessment." Specifically, it first extracts representative soil moisture status from multiple moisture layers, then uses this moisture status to modulate the impact of previous rainfall, and finally combines the modulated rainfall effect with the representative soil moisture level for risk classification.

[0036] This invention relates to a landslide risk level prediction method that considers the impact of prior rainfall and soil moisture conditions. The steps of this method are as follows: Step 1: First, based on the complete monitoring data already obtained in the study area, construct a basic variable system to describe the rainfall input process and soil moisture response characteristics, providing a unified data expression basis for subsequent rainfall impact modeling and wetting state modulation.

[0037] Specifically, rainfall and soil moisture content are selected as the basic observed variables of the system and expressed on a unified time scale. For rainfall input, the time series function is defined as:

[0038] in, R ( t () indicates time t Rainfall amount, in millimeters (mm). t This indicates the discrete time step (which can be taken as an hour or a day depending on the actual monitoring conditions). N This represents the total observation time. R ( t Data can be derived from on-site rain gauges, regional weather stations, or gridded precipitation data products, and is used to characterize the instantaneous input intensity of external rainfall on the surface system.

[0039] Simultaneously, based on soil moisture sensors deployed at different depths, soil moisture data of multiple soil layers is acquired and defined as:

[0040] in, i j ( t ) indicates the first j Layered soil at time t Soil moisture, expressed as a volume ratio (m³). 3 / m 3 ), j This indicates the layer number representing the depth at which the sensor is buried. k To monitor the number of layers. By deploying multiple layers of sensors, the area can be covered from the shallow layer to the potential sliding surface, thereby reflecting the vertical distribution characteristics of soil water content during rainfall infiltration.

[0041] The aforementioned multi-layered soil moisture data can characterize the migration and accumulation of water within the soil during rainfall infiltration. The differences in soil moisture between different depth layers reflect the lag and non-uniformity of the soil's response to rainfall. Therefore, compared to a single-depth moisture index, multi-layered moisture information can more comprehensively describe the dynamic evolution of the water field within a landslide.

[0042] If soil moisture i j ( t If there are outliers or abrupt changes, smoothing should be performed:

[0043] in It is the first after smoothing. j The soil moisture value of the layer, after smoothing, can better reflect the actual trend of change. The smoothing can be achieved by filtering. For the original first j Soil moisture content in the layer. c Smoothing coefficient (0 < c <1), recommended to adopt c A value of ≈0.3-0.5 can smooth out noise and provide a rapid response to sudden rainfall. It is used to control the weight of the latest measurement value and the historical filtered value. The larger the value, the more emphasis is placed on the current measurement. This is the filtering result from the previous time step, used to smooth abrupt changes. This formula uses an exponentially weighted filtering method for smoothing, so that soil moisture data retains its trend while removing abrupt noise, ensuring continuity and stability. This is a key step in supporting the moisture state modulation mechanism.

[0044] Subsequently, the soil moisture values ​​obtained from soil moisture monitoring data were converted into a dimensionless moisture index. ， Eliminating dimensional differences between different sensor layers enhances the comparability of humidity data at different depths and facilitates subsequent multi-depth integration. in, Sm j ( t ) indicates the first j Layered soil at time t The humidity index has a value range of [0,1]. i min,j and i max,j These are the minimum and maximum soil moisture values ​​of the soil layer during the observation period (in this embodiment, volumetric water content is used as soil moisture), which are used to characterize the range of change of the soil layer from a dry state to a saturated state. imin,j and i max,j The quantiles of historical soil moisture in each monitoring layer (e.g., 5% and 95%) can be used to reduce the influence of extreme values ​​and ensure the accuracy of moisture level indicators. Sm j ( t It has stability and comparability in distinguishing wet states.

[0045] Through the above normalized expression, the moisture state of soil at each depth is mapped to a uniform scale, thus preserving its original physical meaning while also possessing the ability to directly participate in state determination and coupled calculations. Sm j ( t The value of ) reflects the relative position of the current moisture level of the soil layer within its historical range of variation, and can be used to characterize the process of soil development from unsaturated to saturated.

[0046] Furthermore, the multi-layered humidity levels are constructed as a vector form:

[0047] This vector describes the distribution characteristics of the instantaneous wet state of soil at different depths and is an important representation of the internal moisture structure of soil.

[0048] It should be noted that the rainfall R ( t ) and multi-layered humidity vector S m ( t These are used to characterize the external driving process and internal response state of the system, respectively. They constitute an input-response data structure on a unified time scale:

[0049] This structure provides basic data support for the subsequent construction of the impact of previous rainfall and the extraction of soil moisture state variables. This is the basic variable system used to describe the rainfall input process and the soil moisture response characteristics.

[0050] In addition, the output of step one S m ( t This serves only as the basic input for constructing the soil moisture state control variables in the subsequent step two, and is used to modulate the impact of previous rainfall. P e ( t By further extracting its features, state variables that characterize the overall soil wetting state can be formed and participate in the modulation calculation process of the impact of previous rainfall, thereby realizing the coupled expression between rainfall-driven and soil response.

[0051] Finally, step one outputs the rainfall. R ( t and multi-layered humidity vector S m ( t This provides a unified data foundation for the subsequent construction of rainfall driving factors, identification of wet conditions, and establishment of modulation mechanisms.

[0052] Step 2: In order to effectively characterize the overall soil moisture state and enable it to participate in the modulation process of subsequent rainfall impact, this step comprehensively extracts multi-layered moisture information to construct soil moisture state variables.

[0053] Specifically, considering the non-uniform vertical distribution of water within the soil during rainfall infiltration, the impact of different depth layers on landslide stability varies. Locally high-humidity areas tend to preferentially form potential sliding surfaces and play a controlling role in overall stability. Therefore, based on multi-layered wetting indexes, a most unfavorable layer control mechanism is introduced. Representative wetting indices are extracted from the multi-layered wetting indices, and the representative soil wetting degree is defined as:

[0054] in, S * ( t () indicates time t The representative value of the overall soil moisture state reflects the level of moisture closest to saturation at each depth.

[0055] The physical significance of this expression method lies in the fact that when a certain depth layer first reaches a high level of moisture, that layer may become the key layer controlling changes in the mechanical properties of the soil, thus dominating the formation and development of potential sliding surfaces. Therefore, maximum value extraction can effectively capture the most unfavorable moisture state inside the soil, avoiding the weakening of key risk information by multi-layer averaging or weighted processing, and improving the sensitivity and engineering rationality of the moisture state characterization.

[0056] To obtain representative humidity levels S * ( t To enable it to participate in subsequent calculations and play a regulatory role, the continuously changing representative moisture level is transformed into a discrete state variable, and a soil moisture state function is constructed:

[0057] in, State ( t () indicates time t The soil moisture state is represented by a finite set of discrete states. Specifically, it is divided into multiple levels according to the degree of moisture, for example:

[0058] The corresponding division rule is: when When, it is defined as a low humidity state. S 1; when When, it is defined as a medium humidity state. S 2; when When, it is defined as a high humidity state. S 3.

[0059] in, T 1 and T 2 is the threshold for classifying wet conditions, which satisfies... This is used to distinguish between different humidity levels. The humidity level grading threshold in this embodiment... T 1 = 0.4 and T 2=0.7 to ensure that the discrete state is representative for landslide sensitivity discrimination.

[0060] Through the above discretization process, the original continuously changing multi-layered humidity information is transformed into state variables with clear physical meaning, enabling them to participate in subsequent calculations in a unified form.

[0061] It should be emphasized that, in this embodiment, the soil is in a moist state. State ( t This variable is not used directly as a risk assessment indicator, but rather as a control variable introduced into the subsequent analysis process. It characterizes the soil's current moisture condition's responsiveness to rainfall, and influences the calculation results of the impact of previous rainfall through subsequent modulation mechanisms, thereby indirectly regulating the soil moisture state's influence on the rainfall-triggered effect.

[0062] Therefore, by constructing the soil moisture state, this step realizes the transformation from "multidimensional humidity data" to "single control variable", providing a key foundation for establishing the coupling relationship between rainfall-driven and soil response in the future.

[0063] Finally, this step outputs the soil moisture status. State ( t () is the control variable of the wet state modulation mechanism in step four, used to dynamically adjust the impact of previous rainfall. P e ( t This ensures that the differences in landslide triggering contributions under different humid backgrounds are accurately reflected.

[0064] Step 3: Constructing rainfall data in Step 1 R ( tBased on continuous rainfall input data established at a unified time scale, this step constructs an impact of previous rainfall to characterize the comprehensive influence of historical rainfall on the current soil moisture state, in order to depict the characteristics of the rainfall process on soil moisture accumulation and its continuous effect.

[0065] Specifically, the impact of rainfall on landslide stability depends not only on the current rainfall intensity but also on previous rainfall events. After infiltrating into the soil, rainfall remains in the pores and gradually migrates deeper, causing the soil moisture content to rise continuously, thereby reducing shear strength and increasing the risk of instability. However, the impact of early rainfall on the soil gradually weakens over time. Therefore, when constructing rainfall impact parameters, both the cumulative effect of rainfall and its time decay characteristics must be considered.

[0066] Based on the above mechanism, a time-weighted cumulative method is introduced to attenuate and weight historical rainfall, defining the impact of previous rainfall as:

[0067] in, P e ( t () indicates time t The impact of previous rainfall; R ( you ) is the number of times before the current moment. i Rainfall at each time step; n Historical rainfall steps; oh ( i ) is a time decay weighting function used to describe the process of historical rainfall impact decreasing over time.

[0068] In the formula P e ( t The term "aspect-related rainfall impact" is a general definition reflecting the cumulative effect of historical rainfall on the current soil moisture state. In actual calculations, to account for both the triggering effect of short-term heavy rainfall and the long-term background moisture accumulation, [the following is omitted as it is not directly related to the previous sentence]. P e ( t Decomposed into short-term timescale rainfall impact P short ( t ) and the impact of long-term rainfall P long ( t and by weight β Weighted synthesis.

[0069] To reflect the objective law that the impact of rainfall gradually diminishes over time, the weighting function adopts an exponential decay form:

[0070] in, t The time decay coefficient controls the duration of the rainfall's impact.

[0071] In practical applications, to ensure the model has clear physical meaning and operability, the historical rainfall steps are used. n and attenuation coefficient t It needs to be determined reasonably based on the research subjects.

[0072] Typically, the timing window for rainfall impact can be determined based on regional soil permeability, rainfall event characteristics, and historical landslide records. L Combined with the time step Δ t The calculated value of n is:

[0073] Where L represents the duration (in hours) during which rainfall has a significant impact on the soil, Δt is the time step (in hours), and n is the number of historical rainfall steps included in the cumulative calculation.

[0074] At the same time, a correlation was established between the time decay coefficient and the rainfall impact time window:

[0075] This setting method ensures that approximately 3 t The impact of rainfall within the area has been largely reduced to a low level, thus taking into account both the cumulative effect of rainfall and the characteristics of time decay.

[0076] The result obtained by the above method P e ( t The value of 100% can comprehensively reflect the contribution of historical rainfall to the current soil moisture state. The larger the value, the more significant the impact of previous rainfall on the soil and the higher the degree of internal moisture accumulation.

[0077] Furthermore, considering the varying impacts of rainfall events on soil at different time scales, short-term heavy rainfall often produces a rapid response in shallow soil layers, while long-term cumulative rainfall has a sustained impact on the wetting state of deep soil layers. Therefore, in constructing the impact of early rainfall, it can be decomposed into two parts: short-term and long-term impacts.

[0078] in, P short ( t This indicates the impact of rainfall on a short timescale, reflecting the rapid effect of recent rainfall on the soil. P long ( t () indicates the long-term rainfall impact, reflecting the cumulative background effect of rainfall;β This is a weighting coefficient used to adjust the relative contributions of the two.

[0079] Short-term rainfall impact P short ( t This reflects the rapid wetting effect of recent rainfall and the long-term impact of rainfall. P long ( t This reflects the wetness of the background. Weighting β This can be determined based on historical rainfall-landslide event matching or engineering experience to enhance the response capability to sudden rainfall. The short-term rainfall window is much smaller than the long-term rainfall window. β The value can be set between 0.6 and 0.8 based on the rainfall characteristics of the study area to enhance the response capability to sudden heavy rainfall.

[0080] It should be noted that the amount of pre-rainfall impact constructed in this step... P e ( t This variable is not used directly as an indicator for landslide risk assessment, but rather as a foundational variable in subsequent analyses. This is achieved by introducing the soil moisture state variable constructed in step two. State ( t ), can be used P e ( t The intensity of rainfall can be dynamically adjusted under different humid conditions by modulating the rainfall.

[0081] Therefore, the impact of previous rainfall P e ( t Essentially, it is a "basic driving force" that comprehensively characterizes the cumulative effect of rainfall. It will be coupled with the soil wetting state in subsequent steps to form a comprehensive index that can reflect the rainfall-soil interaction mechanism.

[0082] Finally, this step outputs the impact of previous rainfall. P e ( t ), used for subsequent modulation calculations and landslide risk analysis.

[0083] Step 4: The soil moisture status has been obtained in Step 2. State ( t and representative soil moisture level S * ( t In step three, the impact of previous rainfall is calculated. P e ( tBuilding upon the previous work, in order to further characterize the differences in the impact of rainfall on landslide stability under different soil wetting conditions, this step constructs a dynamic modulation mechanism based on soil wetting state to perform state-dependent correction on the impact of previous rainfall, thereby obtaining an effective rainfall impact that can reflect the interaction between rainfall and soil.

[0084] Specifically, the impact of previous rainfall obtained in step three P e ( t This primarily reflects the cumulative effect of historical rainfall events over a time scale, essentially describing the comprehensive intensity of external rainfall input. However, under actual engineering conditions, the effect of rainfall on landslide stability is not solely determined by the rainfall itself, but is also significantly influenced by the current soil moisture state. When the soil is dry, rainfall is primarily used to replenish soil moisture, with infiltration mainly involving adsorption and filling, contributing little to pore water pressure. Conversely, as the soil approaches saturation, rainfall more readily transforms into pore water pressure, thereby reducing effective soil stress and weakening shear strength. Therefore, under the same rainfall conditions, the actual landslide-causing effects differ significantly depending on the moisture state.

[0085] Based on the above analysis, the soil wetting state variables obtained in step two will be... State ( t As a hierarchical control variable, the corresponding basic modulation coefficient is introduced. α State(t) This serves as a benchmark to characterize the differences in rainfall intensity across different humidity levels. Simultaneously, to reflect the continuous variation in soil moisture levels within the same humidity level, a representative soil moisture level is introduced. S * ( t As a continuously adjusting variable, the basic modulation coefficients are further adjusted to construct the state composite modulation coefficients. Its expression is:

[0086] in, α ( t (time) t The state composite modulation coefficient; α State(t) Due to the soil moisture state State ( t The basic modulation coefficient is determined by the soil moisture level, which has different values ​​for different states and is used to reflect the graded influence of soil moisture level on rainfall. λ is the continuous modulation coefficient, which is used to control the influence of representative soil moisture level on the modulation result. S * ( t ) represents the representative soil moisture level obtained in step two, and its value ranges from [0,1].

[0087] The state composite modulation coefficient is obtained. α ( t After that, it is applied to the amount of previous rainfall impact constructed in step three. P e ( t Thus, the modulated effective anterior rainfall impact amount is obtained:

[0088] Right now:

[0089] The above expression realizes the coupled calculation of rainfall accumulation effect and soil wetting state, where P e ( t This reflects the historical cumulative effect of rainfall. α State(t) A graded control system reflecting the effect of soil moisture level on rainfall, λ· S * ( t This is used to describe the continuous changes within the same humidity level, thus constructing a two-layer modulation mechanism of "graded control - continuous correction".

[0090] Regarding parameter determination, to ensure the feasibility of the method, the basic modulation coefficients... α State(t) The modulation coefficient can be set as an increasing sequence according to different humidity conditions: S1 is 0.8 for low humidity, S2 is 1.0 for medium humidity, and S3 is 1.3 for high humidity. The continuous modulation coefficient λ can be determined based on the hydrological characteristics of the soil in the study area or historical monitoring data. Generally, the value range is 0.5 to 1.5, so as to ensure that the modulation coefficient increases reasonably with the degree of humidity.

[0091] Through the above construction, this step transforms soil moisture state from a single descriptive variable into a rainfall-regulated variable, enabling the impact of antecedent rainfall to dynamically change with soil moisture state. Compared with traditional methods, this method not only considers the cumulative effect of rainfall but also achieves a coupled expression between rainfall input and soil response by introducing hierarchical control and continuous correction of soil moisture state, thereby improving the ability to characterize the hydrological mechanisms during landslide formation.

[0092] Finally, this step outputs the modulated effective anterior rainfall impact. P* e ( t This provides basic indicators for subsequent risk assessment and early warning analysis.

[0093] Step 5: Obtain the modulated effective anterior rainfall impact amount after step 4. P* e ( tBased on this, in order to achieve a quantitative assessment of the stability of landslides and fully consider the synergistic effects of rainfall driving force and soil wetting state, this step constructs a landslide risk assessment method based on "dual index coupling".

[0094] Specifically, the result obtained in step four P* e ( t This comprehensively reflects the cumulative effect of rainfall and the moderating influence of soil moisture on the rainfall effect, and can characterize the actual intensity of the rainfall's effect on landslide stability; while the representative soil moisture level obtained in step two... S * ( t This reflects the degree to which the soil is currently close to saturation, and is an important internal factor affecting the development of pore water pressure and changes in shear strength. Therefore, introducing... P* e ( t )and S * ( t These two indicators are used together to construct a risk assessment model.

[0095] First, the effective antecedent rainfall impact is normalized to eliminate the influence of dimensions and enhance the comparability between different operating conditions. The normalized rainfall index is defined as follows:

[0096] in, P ’ ( t () is a normalized rainfall index; P 2 represents the rainfall threshold parameter corresponding to high risk, and its value can be determined based on historical landslide events or monitoring data statistics.

[0097] Based on this, a comprehensive risk index is constructed:

[0098] in, I(t) For a moment t The comprehensive risk index; w 1> w 2 is the weighting coefficient, which satisfies w 1+ w 2=1. In this embodiment, rainfall is considered as the primary triggering factor, and soil moisture status as an auxiliary amplification factor. w 1 = 0.7 w 2 = 0.3.

[0099] Furthermore, to achieve risk level classification, a comprehensive risk index will be used. I(t) A risk assessment function is constructed by comparing the risk assessment function with a preset threshold.

[0100] in, Risk ( t () indicates time t The landslide risk level, R 1. R 2 and R 3 represents low risk, medium risk, and high risk, respectively; I 1 and I 2 is the risk classification threshold, which meets the requirements. I 1< I 2. In this embodiment, it is acceptable to... I 1 = 0.8 I 2 = 1.2.

[0101] Furthermore, to avoid misjudgment of risk due to short-term heavy rainfall when the soil is dry, if the comprehensive risk index determines the state to be high-risk, a wet state constraint is introduced for further risk assessment. The upper limit of the risk level will be increased from R 3 downgraded to medium risk level R 2, of which T 1 represents the threshold for low humidity.

[0102] Through the above construction, this step achieves the coupled determination of effective rainfall impact and soil moisture status. Specifically, P ’ ( t Characterizes the intensity of rainfall driving force. S * ( t The two factors characterize the soil's internal wetting response, and are combined by weighting to form a comprehensive risk index, so that the risk assessment results can simultaneously reflect the cumulative effect of rainfall and the soil's moisture status.

[0103] Compared to methods that rely solely on rainfall or a single humidity index, this method can improve the sensitivity of risk identification when the soil is close to saturation and effectively suppress false alarms when the soil is dry, thereby improving the accuracy and engineering applicability of landslide risk assessment.

[0104] Finally, this step outputs the landslide risk level. Risk ( t It is used for landslide early warning and landslide (slope) safety management.

[0105] Example This embodiment selects the upstream slope of the debris flow gully in Baishe Valley Scenic Area, Jizhou District, Tianjin as the research object. The slope is about 18 m high and has a slope of 38°. The surface layer is silty clay mixed with gravel, and the shallow layer has moderate permeability. Under heavy rain conditions, shallow landslides or slope erosion instability are likely to occur.

[0106] To obtain rainfall and soil response data, automatic rain gauges and multi-layer soil moisture sensors were deployed in the study area. The soil moisture sensors were buried at depths of 20 cm, 50 cm, and 100 cm, corresponding to the shallow soil, the intermediate-shallow transition zone, and the potential sliding influence layer, respectively. The three-layer sensor system can reflect the entire process of water migration from the surface to the deeper layers during rainfall infiltration, thus comprehensively characterizing the vertical distribution of soil moisture.

[0107] S1: Rainfall data was collected by an automatic rain gauge in the scenic area, and soil moisture data was collected by soil moisture sensors buried within the landslide. Considering that shallow slope slippage usually occurs in the shallow to medium-shallow soil layers, this embodiment deploys three layers of soil moisture sensors on the slope at depths of 10 cm, 30 cm, and 60 cm to obtain the volumetric water content at the time of the warning. i 1( t ), i 2( t ), i 3( t (Unit: m) 3 / m 3 The 12 hours prior to the landslide (from t-12h to time t, where t is the time of the landslide) were used as the research time window.

[0108] To reflect the cumulative effect and short-term triggering effect of rainfall, this embodiment divides rainfall into short-term and long-term components. Short-term rainfall (Table 1) is selected from the 3 hours before the landslide (t-3 h to t), mainly used to reflect the rapid wetting of shallow soil and the triggering effect of heavy rainfall; long-term rainfall (Table 2) is selected from the 9 hours before the landslide (t-12 h to t-3 h), used to describe the background wetting accumulation of soil. The short-term and long-term rainfall data are weighted and combined to reflect their respective contributions when calculating the impact of previous rainfall.

[0109] Table 1: Short-term rainfall data

[0110] Table 2: Long-term rainfall data

[0111] Subsequently, three layers of soil moisture sensors were deployed in the potential landslide area at depths of 10 cm, 30 cm, and 60 cm, respectively. The volumetric water content of the soil was recorded one hour before the landslide and at the moment of the landslide (Table 3). These data were used to reflect the immediate soil moisture status under rainfall, providing a basis for subsequent moisture level calculations.

[0112] Table 3. Soil moisture data from multiple layers

[0113] To smooth the monitoring data and eliminate abrupt noise, an exponentially weighted filter is used, and a smoothing coefficient is selected. c ≈0.4: Calculate the filter values ​​for each layer: 1) 10 cm: 2) 30 cm: 3) 60 cm: Next, the filtered humidity is converted into a dimensionless humidity index. Sm j ( t ):

[0114] Minimum water content of soil at various depths i min,j With maximum moisture content i max,j It was not set arbitrarily, but determined comprehensively based on the typical moisture content variation range of silty clay slopes and engineering experience. Specifically, i min,j It represents the dry state of the soil layer, and its value is close to the low quantile (e.g., 5th percentile) of long-term monitoring or empirical observation. i max,j This indicates that the soil layer is close to saturation, and its value is close to the high quantile (e.g., 95th percentile) of long-term monitoring or empirical observation. In this embodiment, referring to the typical range of silty clay and combining engineering experience, the values ​​for each soil layer are set as follows:

[0115] Calculate separately: 1) 10 cm: 2) 30 cm: 3) 60 cm: This leads to the construction of a multi-layered humidity level index vector:

[0116] This indicates that the 20 cm shallow layer is close to saturation, the 50 cm layer is significantly humidified, and the 100 cm layer has a strong infiltration response.

[0117] S2: Rainfall data for the 12 hours prior to the landslide has been obtained. R ( t ) and the three-layer soil moisture index vector The soil moisture in each layer has been filtered and smoothed.

[0118] First, considering the non-uniform vertical distribution of soil moisture during rainfall infiltration and the different impacts of different depth layers on slope stability, locally high humidity areas typically preferentially form potential sliding surfaces, playing a dominant role in overall stability. Therefore, this step introduces a most unfavorable layer control mechanism, taking the maximum value of the multi-layer wetting index and vectorizing it from the multi-layer wetting index... Sm ( t Extract representative soil moisture content. S * ( t ):

[0119] Substitute the data:

[0120] This indicates that the most unfavorable wetting layer controlling the slope stability is the 20 cm shallow soil layer, which is an important input for subsequent landslide triggering analysis.

[0121] Subsequently, the representative soil moisture levels were continuously measured. S * ( t Transform into discrete state variables State ( t This allows for tiered control of the impact of previous rainfall in subsequent calculations. The discretization rules are as follows:

[0122] This embodiment sets a threshold for grading the humidity state. T 1 = 0.4 T 2=0.7, ensuring that the discrete state has significant engineering implications and can sensitively reflect the potential contribution of soil wetting to landslide triggering. The calculation yields: This indicates that the soil is in a highly humid state and is most sensitive to rainfall-triggered conditions.

[0123] S2: Using the obtained data, calculate the cumulative impact of rainfall prior to the landslide on the soil's moisture state, forming the amount of impact from prior rainfall. P e ( t This provides input for subsequent humidification modulation.

[0124] Considering the lag effect of rainfall on slope soil, meaning that the impact of rainfall infiltration on pore water pressure and soil moisture content gradually decreases over time, the impact of early rainfall is represented by a time-decay weighted cumulative form.

[0125] This embodiment uses an exponentially decaying weighting function:

[0126] Calculating the short-term rainfall impact using exponential decay weighting: Calculation process: 1) t-3 h: 2) t-2 h: 3) t-1 h:

[0127]

[0128] Calculating the long-term rainfall impact using exponential decay weighting:

[0129] here n The value of 12 is chosen because the short-term timeframe of 3 hours is included. i =4 is the starting value, and the short-term part is not included.

[0130] Partial calculation process: 1) t-4 h: 2) t-5 h: After adding up other factors, the long-term rainfall impact To reflect the difference between cumulative rainfall and sudden effects, the impact of short- and long-term rainfall scales is weighted according to the combined effects of short- and long-term rainfall. β= 0.7 combination: .Right now:

[0131] S3: In order to quantify soil moisture state as a modulating factor on the effect of rainfall, this embodiment uses discrete moisture state... State ( t Based on this, a basic modulation coefficient is introduced. α State(t) Meanwhile, to reflect the representative moisture level of the soil S * ( t To continuously correct the modulation results, a continuous modulation coefficient λ is introduced to construct a state composite modulation coefficient:

[0132] because Under high humidity conditions, the basic modulation coefficients are obtained. αState(t) =1.3, and taking the continuous correction coefficient λ=0.8, the state composite modulation coefficient is:

[0133] The impact of previous rainfall was also calculated as follows:

[0134] The modulated effective anterior rainfall impact was obtained:

[0135] That is, the modulated effective anterior rainfall impact is obtained:

[0136] Under high humidity conditions, the same amount of accumulated prior rainfall is modulated and amplified, reflecting the sensitivity of the soil wetting background to rainfall-triggered landslides. This value will be used as input for subsequent risk assessment models, reflecting the rainfall-soil coupling effect.

[0137] S4: After obtaining the modulated effective anterior rainfall impact amount P* e ( t and soil moisture status State ( t and representative soil moisture S * ( t These variables are introduced into the risk assessment model, and the risk level of landslides is calculated through comprehensive indicators, providing a quantitative basis for slope risk management.

[0138] First, the effective impact of prior rainfall. P* e ( t Normalization is performed:

[0139] in, P ’ ( t (This refers to the normalized rainfall index.) P 2. The historical high-risk reference rainfall is taken as 40 mm. Substituting the values ​​into the calculation, we get: .

[0140] Overall Risk Index I(t) A dual-index coupling method is used to combine the modulated effective antecedent rainfall impact with the representative soil moisture level:

[0141] in, w 1 = 0.7 w 2 = 0.3; P ’( t ) = 1.095; S * ( t Substituting 0.853 into the risk formula:

[0142] Comprehensive risk index I(t) A risk assessment function is constructed by comparing the risk assessment with a preset risk classification threshold.

[0143] Therefore, it is determined that: If the slope risk level is medium when the soil is moist due to rainfall in the 12 hours prior to the landslide, it indicates a potential landslide and requires monitoring or early warning.

[0144] Any aspects not covered in this invention are applicable to existing technologies.

Claims

1. A method for predicting landslide risk level considering the impact of prior rainfall and soil moisture status, characterized in that, The prediction method includes the following: For the landslide area to be studied, soil moisture sensors are deployed at different depths to acquire soil moisture data of multiple soil layers. The data is then dimensionless to obtain a multi-layer moisture index vector. Through the deployment of multiple sensors, the range from the shallow layer to the potential sliding surface is covered, thereby reflecting the vertical distribution characteristics of soil water content during rainfall infiltration. Simultaneously acquire rainfall data, and then... R ( t ) and multi-layered humidity index vector S m ( t They form an input-response data structure on a unified time scale to characterize the external driving process and internal response state of the system. Based on the multi-layered wetting index, a most unfavorable layer control mechanism is introduced to extract representative wetting indices from the multi-layered wetting index, thereby obtaining the representative wetting degree of the soil. S * ( t ); To obtain representative soil moisture levels S * ( t Then, the continuously changing representative moisture level is transformed into a discrete state variable, and a soil moisture state function is constructed: ， in, State ( t () indicates time t The soil moisture state is represented by a finite set of discrete states; it is further divided into multiple levels based on the degree of moisture. ， The corresponding division rule is: when When, it is defined as a low humidity state. S 1; when When, it is defined as a medium humidity state. S 2; when When, it is defined as a high humidity state. S 3; T 1 and T 2 is the threshold for classifying wet conditions, which satisfies... ; Set short-term and long-term rainfall time windows, and calculate the short-term rainfall impact for each. P short ( t ) and the impact of long-term rainfall P long ( t The weighted average of the two factors determines the impact of previous rainfall. Pe(t) ; Construct state composite modulation coefficients to incorporate the impact of the preceding rainfall. Pe(t) Multiply by the corresponding state composite modulation coefficient to determine the modulated effective anterior rainfall impact. The state composite modulation coefficient is determined by the following formula: ， in, α ( t (time) t The state composite modulation coefficient; α State(t) Due to the soil moisture state State ( t The basic modulation coefficient is determined by λ; λ is the continuous modulation coefficient, used to control the influence of the representative soil moisture level on the modulation result. The modulated effective anterior rainfall impact Normalization was performed to obtain the normalized rainfall index. P ’ ( t Construct a comprehensive risk index: ， in, I(t) For a moment t The comprehensive risk index; w 1> w 2 is the weighting coefficient, which satisfies w 1+ w 2 = 1; Comprehensive risk index I(t) The landslide risk level is determined by comparing it with the preset risk classification threshold.

2. The method according to claim 1, characterized in that, The short-term rainfall impact is calculated using a time-decay weighted cumulative method. P short ( t ) and the impact of long-term rainfall P long ( t The weighted average of the two factors determines the impact of previous rainfall. Pe(t) The specific process is as follows: Set short-term and long-term rainfall time windows, and calculate the short-term rainfall impact using the following formula. P short ( t ) and the impact of long-term rainfall P long ( t ): ， in , R(ti) Let be the rainfall at the i-th time step before the current time t; n be the number of historical rainfall steps; ω(i) be the time decay weight function; i 0 This refers to the start time of either a short-term or long-term rainfall window. The weighting function uses an exponential decay form: ， in, τ The time decay coefficient controls the duration of the rainfall's impact. The impact of previous rainfall is calculated using the following formula: ， in, β This is a weighting coefficient used to adjust the relative contributions of short-term and long-term rainfall.

3. The method according to claim 2, characterized in that, β The value is 0.6–0.8; the correlation between the time decay coefficient and the rainfall impact time window L is as follows: .

4. The method according to claim 1, characterized in that, S * ( t The value range is [0,1]; the basic modulation coefficient α State(t) The humidity levels were set as an increasing sequence: S1 for low humidity was 0.8, S2 for medium humidity was 1.0, and S3 for high humidity was 1.

3. The continuous modulation coefficient λ is determined based on the hydrological characteristics of the soil in the study area or historical monitoring data, and its value ranges from 0.5 to 1.

5.

5. The method according to claim 1, characterized in that, Risk assessment function: ， in, Risk ( t () indicates time t The landslide risk level; R 1. R 2 and R 3 represents low risk, medium risk, and high risk, respectively; I 1 and I 2 is the risk classification threshold, which meets the requirements. I 1< I 2.

6. The method according to claim 5, characterized in that, If the comprehensive risk index determines the state to be high-risk, a wet state constraint is introduced to further assess the risk. The risk level will be downgraded from R3 to medium risk. R 2, of which T 1 represents the threshold for low humidity.

7. The method according to claim 1, characterized in that, The process of obtaining a multi-layered humidity index vector through dimensionless transformation is as follows: Soil moisture content in multi-layered soil structures is obtained using soil moisture sensors deployed at different depths. θ j ( t ), in, θ j ( t ) indicates the first j Layered soil at time t Soil moisture, expressed in volume ratio (m³). 3 / m 3 , j This indicates the layer number representing the depth at which the sensor is buried. k To monitor the number of layers; If soil moisture θ j ( t If there are excess values ​​or abrupt jumps, smoothing should be performed according to the following formula: ， in It is the first after smoothing. j Soil moisture content in the layer; For the original first j Soil moisture content in the layer; γ is Smoothing coefficient, 0 < γ <1; This is the smoothed result from the previous time step; Soil moisture value after smoothing treatment Normalize according to the following formula: ， in, Sm j ( t ) indicates the first j Layered soil at time t The degree of humidity; θ min,j and θ max,j These are the minimum and maximum moisture contents of the soil layer during the observation period, used to characterize the range of change of the soil layer from a dry state to a saturated state. The multi-layered humidity index is constructed into a vector form: ； in, This is a vector of multi-layered humidity indexes.

Images