Regional rainfall-slope cutting composite threshold landslide dynamic probability calculation method

Abstract

The invention discloses a regional rainfall-slope cutting composite threshold landslide dynamic probability calculation method. The method comprises the steps that firstly, landslide record data, slope cutting samples, high-resolution remote sensing images, slope cutting strength evaluation factors and rainfall site data are collected; then constructing a critical rainfall threshold curve based on a landslide event and rainfall parameters (early effective rainfall EE and rainfall duration D), and fitting a dynamic probability rainfall threshold equation by adopting nonlinear logistic regression; quantitative evaluation is carried out based on multi-source data and a random forest model to obtain regional slope cutting intensity spatial distribution, and slope cutting and landslide relevance is analyzed in combination with a frequency ratio; finally, coupling rainfall and slope cutting factors, and fitting a dynamic probability rainfall-slope cutting composite threshold model by adopting a multi-class composite equation; and real-time calculation of the dynamic probability of the regional landslide is realized. According to the method, a rainfall and slope cutting joint induced landslide mechanism is fused, a rainfall-slope cutting composite threshold mechanism is innovatively provided, and the problem of landslide dynamic probability calculation under the joint action of rainfall and manual slope cutting is solved.

Description

Technical Field

[0001] This invention relates to the field of geological disaster early warning technology, specifically a method for calculating the dynamic probability of landslides using a regional rainfall-slope cutting composite threshold. Background Technology

[0002] Landslides are among the most destructive natural disasters globally, severely threatening human socio-economic development and the safety of people's lives and property. Landslide hazard assessment is a crucial link in geological disaster risk early warning. Landslide hazard is the spatiotemporal probability of a landslide occurring in a specific time and area, simultaneously considering both environmental factors and external triggering factors such as rainfall and slope cutting. Heavy rainfall and persistent light rainfall can significantly alter the physical properties of the landslide slope and the groundwater level, reducing the shear strength of the soil and rock mass, thus leading to slope instability and severe consequences. Slope cutting typically refers to human-induced excavation activities at the toe of the slope, such as road construction or building construction. Improper slope cutting alters the original mechanical equilibrium of the natural slope, creating a steep, unprotected surface that reduces the resistance to sliding and exacerbates the risk of slope instability.

[0003] Landslide hazard assessment has always been a focus of landslide risk research. Landslide hazard assessment is obtained by multiplying landslide susceptibility by the probability of landslide occurrence over a continuous time period. However, current landslide dynamic probability calculations suffer from the problem of considering only one inducing factor, primarily relying on critical rainfall thresholds to calculate temporal probabilities. Conventional temporal probability models, however, focus more on rainfall and less on the distribution of slope cutting intensity across different slopes within a region. In fact, rainfall-induced deposit landslides in the mountainous and hilly areas of central and eastern China often occur under the combined effects of rainfall and slope cutting, with one significant limiting factor being undesirable human-induced slope cutting. Therefore, there is an urgent need to explore a regional rainfall-slope cutting composite threshold method for calculating landslide dynamic probabilities. Summary of the Invention

[0004] The purpose of this invention is to provide a method for calculating the dynamic probability of landslides by combining regional rainfall and slope cutting composite thresholds, based on practical needs. By integrating the landslide-inducing mechanisms of rainfall and slope cutting, this method solves the problem of calculating the dynamic probability of landslides caused by the combined effects of rainfall and artificial slope cutting.

[0005] To achieve the above objectives, the present invention adopts the following technical solution.

[0006] A method for calculating the dynamic probability of landslides using a combined threshold of regional rainfall and slope cutting includes the following steps: Step S1: Basic data preparation; Based on high-resolution remote sensing images, positive samples marked with "1" and negative samples marked with "0" were selected in the study area to construct a positive-negative sample set of slope cutting, and landslide logging data and rainfall station data in the study area were collected. Step S2: Extraction of slope evaluation factors; Nine index factors were collected, including slope, road density, topographic relief, stratigraphy, NDBI, elevation, population density, radiation, and valley depth. Step S3: Regional slope cutting strength assessment; By combining the positive and negative sample sets of slope cutting in step S1 and the nine index factors in step S2, a spatial dataset is constructed and imported into a random forest model to predict the slope cutting intensity value of the study area. The nonlinear relationship between slope cutting and landslide is analyzed using the frequency ratio. Step S4: Rainfall data collection; Rainfall data for the corresponding dates of landslide samples were collected by daily rainfall station data collection, and statistical analysis was performed to obtain the effective rainfall amount EE and rainfall duration D. Step S5: Construction of the composite threshold model; By coupling the slope cutting intensity predicted in step S3 with the effective rainfall EE and rainfall duration D obtained in step S4, a composite threshold model is established to realize the dynamic probability calculation of landslides in the study area.

[0007] Specifically, the positive samples in step S1 include areas with poor slope cutting, such as houses and highways, and the negative samples include control areas with no slope cutting activity and stable geology; the landslide logging data includes at least the spatial location and occurrence time information of the landslide point, and the rainfall station data includes at least the daily rainfall data within the date range of the landslide occurrence.

[0008] Specifically, the method for obtaining the nine indicator factors in step S2 is as follows: The slope was extracted from the digital elevation model (DEM) using ArcGIS 10.8 software. The highway density was obtained based on road vector data through line density analysis using ArcGIS 10.8 software; The terrain relief is calculated based on the DEM, which is obtained through GIS spatial analysis to obtain the maximum value of the pixel DEM within the search range of the pixel fish. max and minimum DEM min The formula for calculating terrain relief Rel is: Rel = DEM max - DEM min ; The stratigraphic lithology was obtained based on the geological map and lithology classification functions of the 91 Satellite Map software; The NDBI calculates the normalized building index using multispectral remote sensing imagery. The elevation data was obtained by downloading the raster data of the Digital Elevation Model (DEM) using the 91 Satellite Map software. The population density was obtained based on statistical yearbook data and nighttime light remote sensing data analysis; The radiation data was obtained by downloading surface radiation grid data using 91 Satellite Map software. The valley depth is obtained through GIS spatial analysis based on the Digital Elevation Model (DEM). The formula for calculating the valley depth VD is: VD = DEM rid – DEM, where DEM rid The elevation is the ridge elevation after horizontal interpolation, while the DEM is the original elevation.

[0009] Specifically, in step S3, a spatial dataset is constructed and imported into a random forest model to predict the slope cutting intensity value of the study area. The nonlinear relationship between slope cutting and landslide is analyzed using the frequency ratio. The process is as follows: Step S31: Divide the constructed spatial dataset into a training set and a test set in a 7:3 ratio; Step S32: Import the training set and test set into the random forest model, model by setting decision tree and maximum depth hyperparameter, optimize the hyperparameters through grid search to obtain the optimal parameter combination, and use it to predict the slope cutting intensity value of the study area. Step S33: The area under the ROC curve (AUC) based on the test set's receiver operating characteristic (ROC) is used as the evaluation metric for model performance. The ROC curve is plotted with the false positive rate (FPR) on the horizontal axis and the true positive rate (TPR) on the vertical axis, as shown below. ; ; In the above formula, TP, FP, FN, and TN represent the number of true positives, false positives, false negatives, and true negatives, respectively. The AUC value is the area enclosed by the ROC curve and the horizontal axis, ranging from 0 to 1. The closer the AUC value is to 1, the stronger the model's classification ability and the better it can distinguish between positive and negative samples. When the AUC value is 0.5, the model has no discriminative ability and is equivalent to random guessing. When the AUC value is below 0.5, the model performs worse than random classification. Step S34: Divide the slope cutting intensity into 32 equally spaced intervals in ascending order, and analyze the nonlinear relationship between slope cutting and landslide using the frequency ratio method: ; In the above formula, Indicates the slope cutting strength in the th... The corresponding landslide area within the interval; This represents the total area of ​​known landslides within the study area; For the first The area of ​​the slope cutting strength in the interval; The total area of ​​the study region is represented by FR, which is the frequency ratio. A larger FR value indicates a stronger correlation between slope cutting and landslides.

[0010] Specifically, the calculation process for the effective rainfall EE and rainfall duration D in step S4 is as follows: The formula for calculating the effective rainfall amount EE in the preceding period is as follows: ; In the above formula, EE represents the effective rainfall in the preceding period; This represents the rainfall on the day the landslide occurred; Before the landslide Rainfall; This is the rainfall infiltration coefficient; The rainfall duration D is calculated as follows: if the cumulative rainfall in a day is greater than 5 mm, then that day is recorded in the rainfall duration; if the rainfall is less than 5 mm for two consecutive days, then the rainfall event is considered to have ended.

[0011] Specifically, the composite threshold model in step S5 includes a slope-corrected infiltration rainfall-duration threshold model, a rainfall-slope-cutting synergistic landslide threshold model, and a rainfall-duration-slope-cutting composite threshold model. The construction process of the three rainfall-slope-cutting composite threshold models is as follows: The modified infiltration rainfall-duration threshold model is derived by considering the difference in infiltration capacity between natural slopes and cut slopes, based on the traditional critical rainfall threshold model. Assuming the slope cutting strength Q is related to the rainfall infiltration coefficient It conforms to the Weibull distribution function: ; In the above formula, This represents the slope cutting strength value. and These are the shape and scale parameters of the Weibull distribution, respectively. and These represent the maximum and minimum values ​​of the rainfall infiltration coefficient, respectively. The corrected rainfall infiltration coefficient for each slope is calculated using the above formula, yielding the spatial distribution of the rainfall infiltration coefficient in the study area. This allows for the calculation of the corrected anterior effective rainfall. A new critical rainfall threshold model is established based on the corrected previous effective rainfall, and the formula for correcting the previous effective rainfall is expressed as follows: ; In the above formula, D represents the duration of rainfall; and All are statistical parameters; Corresponding to different rainfall durations (D) and slope cutting intensities Corrected effective rainfall under combination The equation for the landslide dynamic probability P1 based on the slope-corrected infiltration rainfall-duration threshold model is expressed as: ; In the above formula, P1 is the dynamic probability value of landslide based on the slope-corrected infiltration rainfall-duration threshold model; D is the corrected effective rainfall amount from the previous period; Q is the rainfall duration; and Q is the slope cutting intensity. The intercept; and These are the regression coefficients of the equation; The rainfall-slope cutting co-induced landslide threshold model is coupled with rainfall-induced probability. With slope induction probability Construct a composite equation for the landslide dynamic probability P2 based on a rainfall-slope-cutting synergistic landslide threshold model: ; In the above formula, Calculated using the dynamic probabilistic rainfall threshold equation: ; By analyzing the statistical relationship between slope cutting intensity and landslide occurrence, and using nonlinear equation fitting, the relationship between slope cutting intensity and landslide failure probability is obtained, thus yielding the probability of slope cutting-induced landslides: ; The rainfall-duration-slope cutting composite threshold model is an improvement on the traditional critical rainfall threshold model. By coupling the effective rainfall amount EE, rainfall duration D, and slope cutting intensity Q, a composite threshold model under the synergistic effect of multiple factors is constructed. The model equation is expressed as: ; In the above formula, EE represents the effective rainfall in the preceding period; D represents the duration of rainfall. For slope cutting strength; and All are statistical parameters; Under the synergistic effect of multiple factors, corresponding to different rainfall durations (D) and slope cutting intensities... The equation for the landslide dynamic probability P3, based on the combined effective rainfall EE and the rainfall-duration-slope cutting composite threshold model, is expressed as follows: .

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. The slope-cutting modified infiltration rainfall-duration threshold model provided by the method of this invention introduces the macroscopic law between slope cutting and infiltration coefficient, which can finely characterize how human engineering activities affect slope stability by changing rainwater infiltration conditions. It is applicable to low-to-moderate rainfall areas where human engineering activities (such as engineering slope cutting) are relatively dense and significantly change the properties of soil and rock. It can successfully explain the phenomenon that local hilly areas show a high probability of landslides due to changes in infiltration conditions caused by human slope cutting activities under low-risk background, providing a scientific basis for accurate investigation. 2. The rainfall-slope cutting synergistic landslide threshold model provided by the method of this invention realizes the quantitative characterization of the synergistic effect of rainfall and slope cutting, two major disaster factors, and on this basis, it shows high early warning economy; it can control the range of high probability early warning areas to a relatively concentrated level and provide early warning of historical landslide points that occurred on the same day. This feature makes the model particularly suitable for areas with limited disaster prevention resources. It can be used as the first choice for regional geological disaster risk management and priority prevention and control in areas with relatively dispersed populations, and provide a scientific basis for decision-makers to optimize the allocation of disaster prevention resources efficiently. 3. The rainfall-duration-slope cutting composite threshold model provided by the method of this invention, by adopting a more stringent composite threshold standard, provides early warning for a large area of ​​high probability. Although it makes some compromises in the accuracy of early warning, it significantly improves the coverage of disaster identification and minimizes the risk of missed reports. This "safety first" design concept makes the model suitable for areas with extremely high safety requirements, such as densely populated areas and areas around critical infrastructure, and can provide more reliable dynamic probability early warning support for geological disaster prevention and control in major projects and key areas. Attached Figure Description

[0013] Figure 1 This is a flowchart of a method for calculating the dynamic probability of landslides using a combined threshold of regional rainfall and slope cutting, according to the present invention. Figure 2 This is a schematic diagram showing the location distribution of landslide and slope cutting samples within the study area in this embodiment of the invention; Figure 3 This is a schematic diagram of the slope cutting strength evaluation index factors in the study area in this embodiment of the invention; Figure 4 This is a spatial distribution diagram of slope cutting intensity in the study area according to an embodiment of the present invention; Figure 5 This is a graph showing the landslide frequency ratio analysis results of the slope cutting intensity in the study area in this embodiment of the invention; Figure 6 This is a dynamic probability result diagram of landslides calculated using the slope-cutting modified rainfall threshold model in the study area of ​​this invention embodiment; Figure 7This is a map showing the dynamic probability results of landslides calculated using a rainfall-slope cutting probability coupling model in the study area of ​​this invention. Figure 8 This is a dynamic probability result diagram of the study area using the rainfall-duration-slope cutting composite threshold scheme in the embodiment of the present invention. Detailed Implementation

[0014] To facilitate understanding and implementation of the present invention by those skilled in the art, the various steps of the method proposed in this invention are described in detail below. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various modifications or alterations to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0015] Example like Figure 1 As shown, this invention discloses a method for calculating the dynamic probability of landslides using a regional rainfall-slope cutting composite threshold, comprising the following steps: Step S1: Basic data preparation; like Figure 2 As shown, based on high-resolution remote sensing images, positive samples marked with "1" and negative samples marked with "0" were selected in the study area to construct a positive-negative sample set of slope cutting, and landslide logging data and rainfall station data were collected. Step S2: Extraction of slope evaluation factors; like Figure 3 As shown, nine index factors were collected, including slope, road density, topographic relief, stratigraphic lithology, NDBI, elevation, population density, radiation, and valley depth. Step S3: Regional slope cutting strength assessment; By combining the positive and negative sample sets of slope cutting in step S1 and the nine index factors in step S2, a spatial dataset is constructed and imported into a random forest model to predict the slope cutting intensity value of the study area. The nonlinear relationship between slope cutting and landslide is analyzed using the frequency ratio. Step S4: Rainfall data collection; Rainfall data for the corresponding dates of landslide samples were collected by daily rainfall station data collection, and statistical analysis was performed to obtain the effective rainfall amount EE and rainfall duration D. Step S5: Construction of the composite threshold model; By coupling the slope cutting intensity predicted in step S3 with the effective rainfall EE and rainfall duration D obtained in step S4, a composite threshold model is established to realize the dynamic probability calculation of landslides in the study area.

[0016] Specifically, the positive samples in step S1 include areas with poor slope cutting, such as houses and highways, and the negative samples include control areas with no slope cutting activity and stable geology; the landslide logging data includes at least the spatial location and occurrence time information of the landslide point, and the rainfall station data includes at least the daily rainfall data within the date range of the landslide occurrence.

[0017] Specifically, the method for obtaining the nine indicator factors in step S2 is as follows: The slope was extracted from the digital elevation model (DEM) using ArcGIS 10.8 software. The highway density was obtained based on road vector data through line density analysis using ArcGIS 10.8 software; The terrain relief is calculated based on the DEM, which is obtained through GIS spatial analysis to obtain the maximum value of the pixel DEM within the search range of the pixel fish. max and minimum DEM min The formula for calculating terrain relief Rel is: Rel = DEM max - DEM min ; The stratigraphic lithology was obtained based on the geological map and lithology classification functions of the 91 Satellite Map software; The NDBI calculates the normalized building index using multispectral remote sensing imagery. The elevation data was obtained by downloading the raster data of the Digital Elevation Model (DEM) using the 91 Satellite Map software. The population density was obtained based on statistical yearbook data and nighttime light remote sensing data analysis; The radiation data was obtained by downloading surface radiation grid data using 91 Satellite Map software. The valley depth is obtained through GIS spatial analysis based on the Digital Elevation Model (DEM). The formula for calculating the valley depth VD is: VD = DEM rid - DEM, where DEM rid The elevation is the ridge elevation after horizontal interpolation, while the DEM is the original elevation.

[0018] Specifically, in step S3, a spatial dataset is constructed and imported into a random forest model to predict the slope cutting intensity value of the study area. The nonlinear relationship between slope cutting and landslide is analyzed using the frequency ratio. The process is as follows: Step S31: Divide the constructed spatial dataset into a training set and a test set in a 7:3 ratio; Step S32: Import the training and test sets into a random forest model. Model the model by setting decision trees and maximum depth hyperparameters. Optimize the hyperparameters using grid search to obtain the optimal parameter combination, and then use it to predict the slope cutting intensity value of the study area. The spatial distribution results of the slope cutting intensity are as follows: Figure 4 As shown; Step S33: The area under the ROC curve (AUC) based on the test set's receiver operating characteristic (ROC) is used as the evaluation metric for model performance. The ROC curve can intuitively reflect the comprehensive performance of the classifier at different thresholds. The ROC curve is plotted with the false positive rate (FPR) on the horizontal axis and the true positive rate (TPR) on the vertical axis, as shown below: ; ; In the above formula, TP, FP, FN, and TN represent the number of true positives, false positives, false negatives, and true negatives, respectively. The AUC value is the area enclosed by the ROC curve and the horizontal axis, ranging from 0 to 1. It is used to quantitatively measure the overall performance of a classification model: the closer the AUC value is to 1, the stronger the model's classification ability and the better it can distinguish between positive and negative samples; when the AUC value is 0.5, it means that the model has no discriminative ability and is equivalent to random guessing; when the AUC value is below 0.5, it means that the model performs worse than random classification. Step S34: Divide the slope cutting intensity into 32 equally spaced intervals in ascending order, and analyze the nonlinear relationship between slope cutting and landslide using the frequency ratio method: ; In the above formula, Indicates the slope cutting strength in the th... The corresponding landslide area within the interval; This represents the total area of ​​known landslides within the study area; For the first The area of ​​the slope cutting strength in the interval; The total area of ​​the study region is represented by FR, which is the frequency ratio. A larger FR value indicates a stronger correlation between slope cutting and landslides.

[0019] Specifically, the calculation process for the effective rainfall EE and rainfall duration D in step S4 is as follows: The formula for calculating the effective rainfall amount EE in the preceding period is as follows: ; In the above formula, EE represents the effective rainfall in the preceding period; This refers to the rainfall on the day the landslide occurred; Before the landslide Rainfall; This is the rainfall infiltration coefficient; The rainfall duration D is calculated as follows: if the cumulative rainfall in a day is greater than 5 mm, then that day is recorded in the rainfall duration; if the rainfall is less than 5 mm for two consecutive days, then the rainfall event is considered to have ended.

[0020] Specifically, the composite threshold model in step S5 includes a slope-corrected infiltration rainfall-duration threshold model, a rainfall-slope-cutting synergistic landslide threshold model, and a rainfall-duration-slope-cutting composite threshold model. The construction process of the three rainfall-slope-cutting composite threshold models is as follows: The modified infiltration rainfall-duration threshold model is derived by considering the difference in infiltration capacity between natural slopes and cut slopes, based on the traditional critical rainfall threshold model. Assuming the slope cutting strength Q is related to the rainfall infiltration coefficient It conforms to the Weibull distribution function: ; In the above formula, This represents the slope cutting strength value. and These are the shape and scale parameters of the Weibull distribution, respectively. and These represent the maximum and minimum values ​​of the rainfall infiltration coefficient, respectively. The corrected rainfall infiltration coefficient for each slope is calculated using the above formula, yielding the spatial distribution of the rainfall infiltration coefficient in the study area. This allows for the calculation of the corrected anterior effective rainfall. A new critical rainfall threshold model is established based on the corrected previous effective rainfall, and the formula for correcting the previous effective rainfall is expressed as follows: ; In the above formula, D represents the duration of rainfall; and All are statistical parameters; Corresponding to different rainfall durations (D) and slope cutting intensities Corrected effective rainfall under combination The equation for the landslide dynamic probability P1 based on the slope-corrected infiltration rainfall-duration threshold model is expressed as: ; In the above formula, P1 is the dynamic probability value of landslide based on the slope-corrected infiltration rainfall-duration threshold model; D is the corrected effective rainfall amount from the previous period; Q is the rainfall duration; and Q is the slope cutting intensity. The intercept; and These are the regression coefficients of the equation; like Figure 6 As shown, based on the constructed slope-corrected infiltration rainfall-duration threshold model, the dynamic probability spatial distribution of landslide disasters in the target area was calculated. Figure 6The results show that the probability of landslides generally decreases from north to south. Notably, while the overall probability level in the northern region is low, some hilly areas exhibit a significantly higher probability of landslides than flat areas. This is mainly attributed to the alteration of surface infiltration conditions by human engineering activities (such as slope cutting), resulting in hilly areas having a stronger rainwater infiltration capacity under the same rainfall conditions, thus exacerbating the risk of slope instability. Compared with traditional models that only consider rainfall factors, the core improvement of the slope-cut modified infiltration rainfall-duration threshold model lies in the introduction of slope cutting disturbance as a key parameter. This effectively characterizes how, under different slope cutting intensities, even with consistent rainfall conditions, the spatial heterogeneity of infiltration capacity leads to significant differences in landslide probabilities. The validation results show that all four landslide points that actually occurred on the day were located within the high-probability warning zone defined by the model, further confirming the effectiveness and early warning capability of the model in real-world scenarios.

[0021] The above results indicate that the core improvement of the slope cutting correction infiltration rainfall-duration threshold model provided by this invention lies in the introduction of macroscopic laws between slope cutting and infiltration coefficient. This model can precisely characterize how human engineering activities affect slope stability by changing rainwater infiltration conditions. It is applicable to low-to-moderate rainfall areas where human engineering activities (such as engineering slope cutting) are relatively concentrated and significantly alter the properties of soil and rock. It can successfully explain the phenomenon that local hilly areas exhibit a high probability of landslides due to changes in infiltration conditions caused by human slope cutting activities under low-risk backgrounds, providing a scientific basis for accurate investigation.

[0022] The rainfall-slope cutting co-induced landslide threshold model is coupled with rainfall-induced probability. With slope induction probability Construct a composite equation for the landslide dynamic probability P2 based on a rainfall-slope-cutting synergistic landslide threshold model: ; In the above formula, Calculated using the dynamic probabilistic rainfall threshold equation: ; By analyzing the statistical relationship between slope cutting intensity and landslide occurrence, and using nonlinear equation fitting, the relationship between slope cutting intensity and landslide failure probability is obtained, thus yielding the probability of slope cutting-induced landslides: ; like Figure 7 As shown, based on the constructed rainfall-slope cutting synergistic landslide threshold model, a dynamic probability spatial distribution map of landslide disasters in the target area on that day was generated. Figure 7The results show that the probability of landslides generally increases from north to south. Notably, even in the southern region with higher rainfall, the probability of landslides remains low in flat areas due to less artificial slope cutting. Conversely, in hilly areas and at the foot of slopes, the rainfall-slope cutting co-induced landslide threshold model effectively identifies and highlights the high probability of landslides caused by slope cutting. This indicates that all four landslide cases that actually occurred on that day fell precisely within the high-probability warning zone defined by the model, confirming the model's excellent warning reliability. More importantly, the rainfall-slope cutting co-induced landslide threshold model demonstrates high economy and efficiency: the high-probability warning zone it generates is relatively concentrated, enabling accurate identification with a smaller warning area while ensuring effective coverage of key areas. This will optimize the allocation of warning resources and provide a decision-making basis for prioritizing resource allocation in geological disaster risk management.

[0023] The above results demonstrate that the core advantage of the rainfall-slope cutting synergistic landslide threshold model provided by this invention lies in its ability to quantitatively characterize the synergistic effect of rainfall and slope cutting, two major disaster factors, and on this basis, it exhibits high early warning economic efficiency. It can control the range of high-probability early warning areas to a relatively concentrated level and provide early warning of historical landslide points that occurred on the same day. This feature makes the model particularly suitable for areas with limited disaster prevention resources and can serve as the preferred solution for regional geological disaster risk management and priority prevention in areas with relatively dispersed populations, providing a scientific basis for decision-makers to optimize the allocation of disaster prevention resources efficiently.

[0024] The rainfall-duration-slope cutting composite threshold model is an improvement on the traditional critical rainfall threshold model. By coupling the effective rainfall amount EE, rainfall duration D, and slope cutting intensity Q, a composite threshold model under the synergistic effect of multiple factors is constructed. The model equation is expressed as: ; In the above formula, EE represents the effective rainfall in the preceding period; D represents the duration of rainfall. For slope cutting strength; and All are statistical parameters; Under the synergistic effect of multiple factors, corresponding to different rainfall durations (D) and slope cutting intensities... The equation for the landslide dynamic probability P3, based on the combined effective rainfall EE and the rainfall-duration-slope cutting composite threshold model, is expressed as follows: ; like Figure 8 As shown, based on the constructed rainfall-duration-slope cutting composite threshold model, a dynamic probability spatial distribution map of landslide disasters in the target area was generated. Figure 8The results show that there is a significant spatial coupling relationship between the probability of landslides and topographic features (such as flat areas and hilly areas). Specifically, under the same rainfall conditions, the probability of landslides in hilly areas and at the foot of slopes is significantly higher than that in flat areas, highlighting the spatial differentiation of disasters under the joint control of topography and human engineering activities. Figure 8 The data shows that all four landslide cases that actually occurred on that day were located within the high-probability warning area identified by the model, confirming that the rainfall-duration-slope cutting composite threshold model has good early warning reliability. However, from the perspective of overall spatial distribution, the warning threshold set by the model is relatively strict, and the range of the high-probability warning area is relatively large. This reflects that the model's judgment criteria when quantifying the combined effects of rainfall and slope cutting are relatively conservative, and it tends to increase the early warning coverage to reduce the risk of missed reports.

[0025] The above results demonstrate that the core advantage of the rainfall-duration-slope cutting composite threshold model provided by this invention lies in the conservatism and security of its early warning results. By adopting a more stringent composite threshold standard, the model provides early warnings for a wider range of high-probability areas. Although some compromises have been made in terms of early warning accuracy, it significantly improves the coverage of disaster identification and minimizes the risk of missed reports. This "safety first" design philosophy makes the model suitable for areas with extremely high safety requirements, such as densely populated areas and areas surrounding critical infrastructure. It can provide more reliable dynamic probability early warning support for landslides in the prevention and control of geological disasters in major projects and key areas.

[0026] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for calculating the dynamic probability of landslides using a regional rainfall-slope cutting composite threshold, characterized in that, Includes the following steps: Step S1: Basic data preparation; Based on high-resolution remote sensing images, positive samples marked with "1" and negative samples marked with "0" were selected in the study area to construct a positive-negative sample set of slope cutting, and landslide logging data and rainfall station data in the study area were collected. Step S2: Extraction of slope evaluation factors; Nine index factors were collected, including slope, road density, topographic relief, stratigraphy, normalized building index (NDBI), elevation, population density, radiation, and valley depth. Step S3: Regional slope cutting strength assessment; By combining the positive and negative sample sets of slope cutting in step S1 and the nine index factors in step S2, a spatial dataset is constructed and imported into a random forest model to predict the slope cutting intensity value of the study area. The nonlinear relationship between slope cutting and landslide is analyzed using the frequency ratio. Step S4: Rainfall data collection; Rainfall data for the corresponding dates of landslide samples were collected by daily rainfall station data collection, and statistical analysis was performed to obtain the effective rainfall amount EE and rainfall duration D. Step S5: Construction of the composite threshold model; By coupling the slope cutting intensity predicted in step S3 with the effective rainfall EE and rainfall duration D obtained in step S4, a composite threshold model is established to realize the dynamic probability calculation of landslides in the study area.

2. The method for calculating the dynamic probability of landslides using a regional rainfall-slope cutting composite threshold as described in claim 1, characterized in that, The positive samples in step S1 include areas with poor slope cutting, such as houses and roads, and the negative samples include control areas with no slope cutting activity and stable geology; the landslide logging data includes at least the spatial location and occurrence time information of the landslide point, and the rainfall station data includes at least daily rainfall data within the date range of the landslide occurrence.

3. The method for calculating the dynamic probability of landslides using a regional rainfall-slope cutting composite threshold as described in claim 1, characterized in that, The method for obtaining the nine indicator factors in step S2 is as follows: The slope was extracted from the digital elevation model (DEM) using ArcGIS 10.8 software. The highway density was obtained based on road vector data through line density analysis using ArcGIS 10.8 software; The terrain relief is calculated based on the DEM, which is obtained through GIS spatial analysis to obtain the maximum value of the pixel DEM within the search range of the pixel fish. max and minimum DEM min The formula for calculating terrain relief Rel is: Rel = DEM max - DEM min ; The stratigraphic lithology was obtained based on the geological map and lithology classification functions of the 91 Satellite Map software; The NDBI calculates the normalized building index using multispectral remote sensing imagery. The elevation data was obtained by downloading the raster data of the Digital Elevation Model (DEM) using the 91 Satellite Map software. The population density was obtained based on statistical yearbook data and nighttime light remote sensing data analysis; The radiation data was obtained by downloading surface radiation grid data using 91 Satellite Map software. The valley depth is obtained through GIS spatial analysis based on the Digital Elevation Model (DEM). The formula for calculating the valley depth VD is: VD = DEM rid – DEM, where DEM rid The elevation is the ridge elevation after horizontal interpolation, while the DEM is the original elevation.

4. The method for calculating the dynamic probability of landslides using a combined threshold of regional rainfall and slope cutting, as described in claim 1, is characterized in that... In step S3, the spatial dataset is constructed and imported into the random forest model to predict the slope cutting intensity value of the study area. The nonlinear relationship between slope cutting and landslide is analyzed using the frequency ratio. The process is as follows: Step S31: Divide the constructed spatial dataset into a training set and a test set in a 7:3 ratio; Step S32: Import the training set and test set into the random forest model, model by setting decision tree and maximum depth hyperparameter, optimize the hyperparameters through grid search to obtain the optimal parameter combination, and use it to predict the slope cutting intensity value of the study area. Step S33: The area under the ROC curve (AUC) based on the test set's receiver operating characteristic (ROC) is used as the evaluation metric for model performance. The ROC curve is plotted with the false positive rate (FPR) on the horizontal axis and the true positive rate (TPR) on the vertical axis, as shown below: ； ； In the above formula, TP, FP, FN, and TN represent the number of true positives, false positives, false negatives, and true negatives, respectively. The AUC value is the area enclosed by the ROC curve and the horizontal axis, ranging from 0 to 1. The closer the AUC value is to 1, the stronger the model's classification ability and the better it can distinguish between positive and negative samples. When the AUC value is 0.5, the model has no discriminative ability and is equivalent to random guessing. When the AUC value is below 0.5, the model performs worse than random classification. Step S34: Divide the slope cutting intensity into 32 equally spaced intervals in ascending order, and analyze the nonlinear relationship between slope cutting and landslide using the frequency ratio method: ； In the above formula, Indicates the slope cutting strength in the th... The corresponding landslide area within the interval; This represents the total area of ​​known landslides within the study area; For the first The area of ​​the slope cutting strength in the interval; The total area of ​​the study region is represented by FR, which is the frequency ratio. A larger FR value indicates a stronger correlation between slope cutting and landslides.

5. The method for calculating the dynamic probability of landslides using a regional rainfall-slope cutting composite threshold as described in claim 1, characterized in that, The calculation process for the effective rainfall EE and rainfall duration D in step S4 is as follows: The formula for calculating the effective rainfall amount EE in the preceding period is as follows: ； In the above formula, EE represents the effective rainfall in the preceding period; This represents the rainfall on the day the landslide occurred; Before the landslide Rainfall; This is the rainfall infiltration coefficient; The rainfall duration D is calculated as follows: if the cumulative rainfall in a day is greater than 5 mm, then that day is recorded in the rainfall duration; if the rainfall is less than 5 mm for two consecutive days, then the rainfall event is considered to have ended.

6. The method for calculating the dynamic probability of landslides using a regional rainfall-slope cutting composite threshold as described in claim 1, characterized in that, Step S5 involves three composite threshold models: a slope-cutting modified infiltration rainfall-duration threshold model, a rainfall-slope-cutting synergistic landslide threshold model, and a rainfall-duration-slope-cutting composite threshold model. The construction process of these three rainfall-slope-cutting composite threshold models is as follows: The modified infiltration rainfall-duration threshold model is derived by considering the difference in infiltration capacity between natural slopes and cut slopes, based on the traditional critical rainfall threshold model. Assuming the slope cutting strength Q is related to the rainfall infiltration coefficient It conforms to the Weibull distribution function: ； In the above formula, This represents the slope cutting strength value. and These are the shape and scale parameters of the Weibull distribution, respectively. and These represent the maximum and minimum values ​​of the rainfall infiltration coefficient, respectively. The corrected rainfall infiltration coefficient for each slope is calculated using the above formula, yielding the spatial distribution of the rainfall infiltration coefficient in the study area. This allows for the calculation of the corrected anterior effective rainfall. A new critical rainfall threshold model is established based on the corrected previous effective rainfall, and the formula for correcting the previous effective rainfall is expressed as follows: ； In the above formula, D represents the duration of rainfall; and All are statistical parameters; Corresponding to different rainfall durations (D) and slope cutting intensities Corrected effective rainfall under combination The equation for the landslide dynamic probability P1 based on the slope-corrected infiltration rainfall-duration threshold model is expressed as: ； In the above formula, P1 is the dynamic probability value of landslide based on the slope-corrected infiltration rainfall-duration threshold model; D is the corrected effective rainfall amount from the previous period; Q is the rainfall duration; and Q is the slope cutting intensity. The intercept; and These are the regression coefficients of the equation; The rainfall-slope cutting co-induced landslide threshold model is coupled with rainfall-induced probability. With slope induction probability Construct a composite equation for the landslide dynamic probability P2 based on a rainfall-slope-cutting synergistic landslide threshold model: ； In the above formula, Calculated using the dynamic probabilistic rainfall threshold equation: ; By analyzing the statistical relationship between slope cutting intensity and landslide occurrence, and using nonlinear equation fitting, the relationship between slope cutting intensity and landslide failure probability is obtained, thus yielding the probability of slope cutting-induced landslides: ； The rainfall-duration-slope cutting composite threshold model is an improvement on the traditional critical rainfall threshold model. By coupling the effective rainfall amount EE, rainfall duration D, and slope cutting intensity Q, a composite threshold model under the synergistic effect of multiple factors is constructed. The model equation is expressed as: ； In the above formula, EE represents the effective rainfall in the preceding period; D represents the duration of rainfall. For slope cutting strength; and All are statistical parameters; Under the synergistic effect of multiple factors, corresponding to different rainfall durations (D) and slope cutting intensities... The equation for the landslide dynamic probability P3, based on the combined effective rainfall EE and the rainfall-duration-slope cutting composite threshold model, is expressed as follows: 。

Images