A method for calculating a refined rainfall threshold based on landslide rainfall sensitivity correction

Abstract

This invention provides a refined rainfall threshold calculation method based on landslide rainfall sensitivity correction, comprising the following steps: S1, acquiring historical landslide and rainfall data for the study area and constructing a rainfall event database; S2, obtaining the total rainfall threshold T for the study area based on a statistical model. S S3. Analyze the characteristics of rainfall-induced landslides in the study area, and select at least one rainfall-sensitive factor for landslides from each of the three aspects of infiltration, runoff, and evaporation to conduct rainfall sensitivity assessment, and obtain the rainfall sensitivity index (RSSI) for a local area in the study area; S4. Based on the rainfall sensitivity index, correct the baseline threshold T for the entire area. S Obtain the refined rainfall threshold T for this local area, where T = T S (1+μ-RSSI), where μ is the baseline threshold coefficient. The beneficial effects of this invention are: it improves the accuracy of rainfall thresholds and solves the problem of converting rainfall thresholds from a region-wide threshold to a localized, finely defined rainfall threshold, thus obtaining a more reliable, refined rainfall threshold.

Description

Technical Field

[0001] This invention relates to the field of geological disaster prediction and forecasting technology, and in particular to a refined rainfall threshold calculation method based on landslide rainfall sensitivity correction. Background Technology

[0002] Rainfall threshold refers to the critical rainfall amount that triggers geological disasters and is crucial for establishing meteorological early warning systems for geological disasters. In regions with different geological environments, reaching or exceeding a certain rainfall condition can easily induce geological disasters. Scholars both domestically and internationally have conducted extensive research on the relationship between rainfall and geological disasters, and this research has been widely applied in the study and practice of geological disaster risk prevention and control.

[0003] Rainfall thresholds, derived from historical disaster and rainfall data, have rapidly developed, and different statistical methods are being applied to regions with varying geological structures. Threshold curves mainly fall into four categories: rainfall intensity-duration threshold (ID); cumulative rainfall threshold (E); cumulative rainfall-duration threshold (ED); and cumulative rainfall-intensity threshold (EI). These thresholds are widely used in geological hazard analysis as criteria for geological hazard risk prediction.

[0004] Currently, the same rainfall threshold is widely used, typically at the provincial or municipal level. However, applying provincial or municipal rainfall thresholds to small, localized areas such as villages, towns, or even individual slopes leads to problems like significant differences in rainfall thresholds, inapplicability, and low accuracy. Implementing township-level rainfall thresholds in every township would consume enormous human and material resources, making the work difficult to carry out. Therefore, how to transform provincial or municipal thresholds into more refined rainfall thresholds for villages, towns, or even individual slopes has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, in order to solve the problems mentioned in the background art, such as large differences between the single rainfall threshold for large areas and the rainfall threshold for local small areas, inapplicability, and low accuracy, the embodiments of the present invention provide a refined rainfall threshold calculation method based on landslide rainfall sensitivity correction.

[0006] Embodiments of the present invention provide a refined rainfall threshold calculation method based on landslide rainfall sensitivity correction, comprising the following steps:

[0007] S1. Obtain historical landslide and rainfall data for the study area and construct a rainfall event database;

[0008] S2. Obtain the total rainfall threshold T for the study area based on the statistical model. S ;

[0009] S3. Analyze the characteristics of landslides induced by rainfall in the study area, select at least one landslide rainfall sensitivity factor from each of the three aspects of infiltration, runoff, and evaporation to carry out rainfall sensitivity evaluation, and obtain the rainfall sensitivity index RSSI of a local area in the study area.

[0010] S4. Correcting the regional baseline threshold T based on the rainfall sensitivity index. S Obtain the refined rainfall threshold T for this local area, where , This is the baseline threshold coefficient.

[0011] Furthermore, the calculation method for the local rainfall sensitivity index (RSSI) in step S3 is as follows:

[0012] S301. Input each landslide rainfall sensitivity factor into the prediction model to obtain the sensitivity index distribution map of the entire study area;

[0013] S302. Based on the regional sensitivity index distribution map, calculate the average rainfall sensitivity value of the local area as the rainfall sensitivity index RSSI.

[0014] Furthermore, the prediction model is a random forest model, and the classifier of the random forest model is...

[0015]

[0016] Where k represents the number of nodes, t i It is the number of decisions. I and I are used for the output variable and characteristic function, respectively;

[0017] The marginal function is:

[0018]

[0019] The classification principle is as follows:

[0020]

[0021] in Let P be the probability space, and let P represent the feature variable.

[0022] Furthermore, the sensitivity index distribution map for the entire region is a raster map.

[0023] Furthermore, the prediction model is a data-driven model.

[0024] Furthermore, the statistical model in step S2 is a cumulative rainfall model, and the formula for calculating cumulative rainfall is:

[0025]

[0026] Where Re is the effective rainfall; Rn is the rainfall of the previous n days; The effective rainfall coefficient is denoted by n, where n is the duration of rainfall.

[0027] Furthermore, the statistical model in step S2 is a cumulative rainfall-duration model, and the formula for calculating cumulative rainfall is:

[0028]

[0029] Where E represents the cumulative rainfall; D represents the duration of the rainfall; , c are statistical parameters.

[0030] Furthermore, the statistical model in step S2 is a rainfall intensity-duration model, and the formula for calculating rainfall intensity is:

[0031]

[0032] Where I represents rainfall intensity; D represents rainfall duration; , c are statistical parameters.

[0033] Furthermore, the statistical model in step S2 is cumulative rainfall minus rainfall intensity.

[0034]

[0035] Where I represents rainfall intensity; E represents cumulative rainfall; , c are statistical parameters.

[0036] Furthermore, landslide rainfall sensitivity factors related to infiltration include lithology, topographic humidity index, distance from river, and groundwater level; landslide rainfall sensitivity factors related to runoff include slope, undulation, and slope variation coefficient; and landslide rainfall sensitivity factors related to evaporation include vegetation cover index, light index, and land use type.

[0037] The beneficial effects of the technical solution provided by the embodiments of the present invention are as follows: The present invention provides a refined rainfall threshold calculation method based on landslide rainfall sensitivity correction, which innovatively proposes landslide rainfall sensitivity evaluation. It selects landslide rainfall sensitivity factors from infiltration, runoff, evaporation, etc. to carry out rainfall sensitivity evaluation, and obtains a refined rainfall threshold for a small area by correcting the benchmark threshold for the whole area based on the rainfall sensitivity index. This improves the accuracy of the rainfall threshold and solves the problem of conversion from the rainfall threshold for the whole area to the refined rainfall threshold for a local small area, thereby obtaining a more reliable refined rainfall threshold, which provides a basis for emergency management departments to make more accurate early warning and forecast. Attached Figure Description

[0038] Figure 1 This example illustrates the relationship between cumulative rainfall and the cumulative frequency of geological disaster events.

[0039] Figure 2 This is a diagram of landslide rainfall sensitivity evaluation factors in the embodiments;

[0040] Figure 3 This is a distribution map of rainfall sensitivity index for the entire region, as shown in the example.

[0041] Figure 4 This is a fine-grained rainfall threshold distribution map of the slope unit over 1 hour in the embodiment;

[0042] Figure 5 This is a refined rainfall threshold distribution map of the slope unit over 6 hours in the embodiment;

[0043] Figure 6 This is a 24-hour refined rainfall threshold distribution map of the slope unit in the embodiment;

[0044] Figure 7 This is a map showing the distribution of geological disasters induced by "catfish" in the example.

[0045] Figure 8 This is a rainfall distribution map for the 24 hours prior to the geological disaster induced by the "catfish" in the example;

[0046] Figure 9 The bar chart shows the success rate of early warning at the baseline threshold and the refined thresholds of 1, 6, and 24 hours in the example. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described below in conjunction with the accompanying drawings. The following description presents a preferred embodiment of the various possible embodiments of the present invention, intended to provide a basic understanding of the invention, but not intended to identify key or decisive elements of the invention or to limit the scope of protection sought.

[0048] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0049] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0050] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures. Also, it should be understood that, for ease of description, the dimensions of the various parts shown in the figures are not drawn to actual scale.

[0051] In the description of this invention, it should be noted that the circuits, electronic components, and modules involved in this invention are all prior art, which can be fully implemented by those skilled in the art, and need not be elaborated upon. The content protected by this invention does not involve improvements to the internal structure and methods.

[0052] It should be further noted that, unless otherwise explicitly specified and limited, the terms "installation" and "connection" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0053] Please refer to Figure 1 The embodiments of the present invention provide a refined rainfall threshold calculation method based on landslide rainfall sensitivity correction, including the following steps S1-S4.

[0054] S1. Obtain historical landslide and rainfall data for the study area and construct a rainfall event database. The study area is generally a large region, such as a province, city, or county. The landslide and rainfall data mainly include landslide type, landslide instability triggering factors, and rainfall events in the 24 hours prior to the landslide disaster, which will be used for subsequent statistical analysis of rainfall thresholds for the entire study area.

[0055] S2. Obtain the total rainfall threshold T for the study area based on the statistical model. S .

[0056] Based on the hourly rainfall events before and after historical disaster occurrences, the correlation between historical disasters and rainfall is analyzed, and the rainfall threshold for the entire region is calculated using statistical models. These statistical models primarily include the cumulative rainfall model, the cumulative rainfall-duration (ED) model, the rainfall intensity-duration (ID) model, and the cumulative rainfall-intensity (EI) model. The cumulative rainfall can be either actual rainfall or effective rainfall. The four statistical models are explained in detail below.

[0057] Cumulative rainfall model: A single rainfall event typically does not necessarily lead to a landslide, and only a portion of the rainfall in each event contributes to landslide occurrence. Therefore, simple cumulative rainfall is unsuitable as a threshold parameter for the model. Thus, the role of effective rainfall needs to be further considered: that is, the amount of rainfall that actually infiltrates and acts on the landslide body during a single rainfall event, excluding surface runoff and evaporation. Here, cumulative rainfall is considered effective rainfall, and its calculation formula is:

[0058]

[0059] Where Re is the effective rainfall; Rn is the rainfall in the previous n days; α is the effective rainfall coefficient; and n is the duration of rainfall.

[0060] Cumulative Rainfall-Duration (ED) Model: This model considers the time factor in rainfall events. Short-duration and long-duration rainfall have different impacts on landslides. Generally, a strong rainfall event is considered necessary for a short-duration landslide, while a long-duration landslide does not require such stringent conditions. It is usually expressed using a linear or power-law function. The formula for calculating cumulative rainfall is:

[0061]

[0062] Where E represents the cumulative rainfall; D represents the duration of the rainfall; , c are statistical parameters.

[0063] Rainfall Intensity-Duration (ID) Model: The rainfall intensity-duration (ID) model is a modification of the cumulative rainfall-duration (ED) relationship threshold model. Rainfall intensity (I) is the ratio of cumulative rainfall to rainfall duration, which eliminates the influence of rainfall duration to some extent. However, it can be observed that rainfall intensity is still related to the temporal distribution. The trend of the relationship between the two is that rainfall intensity decreases as rainfall duration increases, which is essentially the same as the cumulative rainfall-duration model. Commonly used rainfall intensity-duration relationship thresholds are usually expressed in power-law form. The formula for calculating rainfall intensity is:

[0064] .

[0065] The cumulative rainfall-intensity (EI) model links the cumulative rainfall-duration (ED) threshold with the rainfall intensity-duration (ID) threshold model, using both rainfall amount and intensity to describe the rainfall threshold. It is typically expressed using linear and power-law equations. The formula for calculating rainfall intensity is:

[0066] .

[0067] S3. Analyze the characteristics of landslides induced by rainfall in the study area, select at least one landslide rainfall sensitivity factor from each of the three aspects of infiltration, runoff, and evaporation to carry out rainfall sensitivity evaluation, and obtain the rainfall sensitivity index RSSI of a local area in the study area.

[0068] Landslide rainfall sensitivity factors related to infiltration include lithology, topographic humidity index, distance from river, and groundwater level; landslide rainfall sensitivity factors related to runoff include slope, undulation, and slope variation coefficient; and landslide rainfall sensitivity factors related to evaporation include vegetation cover index, light index, and land use type.

[0069] The method for calculating the local rainfall sensitivity index (RSSI) is as follows:

[0070] S301. Input the rainfall sensitivity factors of each landslide into the prediction model to obtain the overall sensitivity index distribution map of the study area. The prediction model is a data-driven model, such as a machine learning model. In this embodiment, the machine learning model is specifically a random forest model.

[0071] When using a random forest model, the classifier of the random forest model is:

[0072]

[0073] Where k represents the number of nodes, t i It is the number of decisions. I and I are used for the output variable and characteristic function, respectively;

[0074] The marginal function is:

[0075]

[0076] The classification principle is as follows:

[0077]

[0078] in Let P be the probability space, and let P represent the feature variable.

[0079] By inputting each landslide rainfall sensitivity factor into the random forest model, the representative feature variable P obtained is the rainfall sensitivity value, and then the sensitivity index distribution map of the whole area is obtained. Generally, the sensitivity index distribution map of the whole area is a raster map.

[0080] S302. Based on the overall sensitivity index distribution map, calculate the average value of the rainfall sensitivity value of the local area as the rainfall sensitivity index RSSI, where RSSI is the average rainfall sensitivity of all grid cells in the local area.

[0081] S4. Based on the rainfall sensitivity index, the baseline threshold TS for the entire region is corrected to obtain the refined rainfall threshold T for this local area, where... The rainfall sensitivity index (RSSI) ranges from 0 to 1. This is the baseline threshold coefficient, ranging from 0 to 1. The probability of a red alert based on the baseline rainfall threshold for the entire study area is taken from the statistical data, specifically from... Figure 1 The curves showing the relationship between cumulative rainfall and cumulative frequency of geological disaster events over a certain period before the landslide are shown. The baseline threshold coefficient μ corresponding to the 1-hour cumulative rainfall is 0.78, the baseline threshold coefficient μ corresponding to the 6-hour cumulative rainfall is 0.82, and the baseline threshold coefficient μ corresponding to the 24-hour cumulative rainfall is 0.6.

[0082] The landslide rainfall threshold is the amount of rainfall required to induce a landslide within a specific area, thus requiring a precise and detailed rainfall threshold. Therefore, accurate and refined rainfall thresholds are crucial for early warning of landslides. Timely warnings and forecasts before a disaster occurs provide protection for risk management, disaster prevention and mitigation efforts, and the safety of people in areas prone to rainfall-induced landslides. Currently, a single rainfall threshold is typically used for a county or province. However, the rainfall threshold for landslides can vary significantly within a town, village, or even a single slope. Using the same threshold across the entire province leads to excessive errors, making it difficult to implement rainfall warnings at the township or village level. To obtain more reliable and refined rainfall thresholds for localized areas, this invention proposes a refined rainfall threshold calculation method based on landslide rainfall sensitivity correction. This method uses a random forest model to evaluate landslide rainfall sensitivity and then corrects the overall baseline threshold based on the rainfall sensitivity index to obtain a locally refined rainfall threshold.

[0083] The refined rainfall threshold calculation method based on landslide rainfall sensitivity correction of the present invention also verifies the reliability of the refined rainfall threshold through the calculation of geological disaster thresholds occurring during Typhoon Megi in 2016. Specifically, as follows:

[0084] The study area in this embodiment is located in Pingyang County, Wenzhou City, Zhejiang Province. Situated in southeastern Zhejiang and eastern Wenzhou, it borders the East China Sea to the east, Wencheng and Taishun counties to the west, Pingyang County and Longgang City to the south, and Ruian City to the north. The county's land area, between 27°21′ and 27°46′ north latitude and 120°24′ and 121°08′ east longitude, covers 1051 square kilometers, with a sea area of ​​1300 square kilometers. It has a subtropical monsoon climate with an average annual temperature of 17.9℃, characterized by mild temperatures and lush vegetation. The county comprises mountains (above 1000m), low mountains (500-1000m), and hills (below 500m), covering a total area of ​​588.9 km². 2The area accounts for 56.1% of the county's total area, with plains comprising approximately 43.9%. The overall terrain slopes from west to east, with higher elevations around the western perimeter and lower elevations in the center. Based on historical data such as the Pingyang County Geological Disaster Prevention and Control "14th Five-Year Plan" results, the Pingyang County 1:50,000 Geological Disaster Risk Survey Report, and the Pingyang County Emergency Investigation Report, a total of 642 geological disaster sites were collected and compiled. Among these, 269 sites were suitable for statistical models. Based on existing data, a total of 642 geological disasters were identified in Pingyang County, and information on their types, occurrence times, and scale was statistically analyzed. According to statistical analysis, among the 642 historical geological disasters recorded in Pingyang County, landslides accounted for 514 sites (86% of the total), collapses for 38 sites (6.8%), and debris flows for 42 sites (7.5%). Landslides are mainly composed of cohesive soil containing gravel, with a loose structure, making them highly susceptible to sliding under rainfall. Medium and small landslides are distributed throughout Pingyang County, but small landslides are the majority. There are 9 medium-sized landslides with a volume of 100,000 to 1 million cubic meters, mainly distributed in Aojiang Town and Shuitou Town, accounting for 2% of the total number of landslide disasters. There are 505 small landslides with a volume of less than 100,000 cubic meters, distributed in all towns of the county, accounting for 98% of the total number of landslide disasters.

[0085] This study uses a cumulative rainfall model to analyze the rainfall threshold that induces landslides in the study area. Figure 1 (a) The graph showing the relationship between rainfall in the hour before the landslide and the cumulative frequency of geological hazards shows that the maximum hourly rainfall was nearly 90 mm. As the hourly rainfall increased, the number of geological hazards also increased. When the hourly rainfall was 20 mm, nearly 40% of geological hazards occurred. When the hourly rainfall was 30 mm, the number of geological hazards reached 50%. Between 30 mm and 40 mm, the number of geological hazards increased rapidly, reaching 78% when the hourly rainfall was 40 mm.

[0086] Depend on Figure 1 (b) The graph showing the relationship between rainfall in the 6 hours prior to a landslide and the cumulative frequency of geological disasters indicates that the maximum 6-hour rainfall was approximately 240 mm, with an even lower average rainfall intensity. This suggests that the rainfall process was initially light but gradually increased, with relatively low initial rainfall followed by concentrated short-duration heavy rainfall. When the 6-hour rainfall was 40 mm, 80 mm, or 115 mm, which were at the turning points of the curve, these values ​​could be used as macro-level thresholds for geological disaster early warning.

[0087] Depend on Figure 1(c) The graph showing the relationship between rainfall in the 24 hours before the landslide and the cumulative frequency of geological disasters indicates that the maximum 24-hour rainfall exceeded 400 mm. There are two obvious inflection points at rainfall of 110 mm and 250 mm, where the proportion of geological disasters is 37% and 60%, respectively. At 190 mm, a significant accelerating upward trend in the cumulative frequency of geological disasters can be observed.

[0088] Based on the intensity analysis of short-term heavy rainfall in different time periods (1 / 6 / 24h) within 24 hours of the above-mentioned geological disasters, and taking into account the frequency of geological disasters and the inflection point brought about by rainfall, a rainfall threshold system for Pingyang County was established. The specific benchmark thresholds are shown in the table below.

[0089] Table 1 Rainfall Thresholds in Pingyang County (Baseline Thresholds)

[0090]

[0091] After rainfall, rainwater mainly consists of three parts: one part infiltrates into the slope, another part is discharged through surface runoff, and the last part evaporates. Based on previous research experience and field investigations, eight landslide rainfall sensitivity factors were selected from infiltration, runoff, and evaporation aspects to conduct rainfall sensitivity assessment. Figure 2 ).

[0092] Infiltration factors: (1) Lithology: Different rock strata have vastly different physical and chemical properties, and porosity and water absorption directly affect the infiltration amount. (2) Distance from the river: The distance from the river mainly affects the groundwater level and water content, and indirectly affects infiltration. Runoff factors: (3) Slope: The slope directly affects the runoff dynamics of water flow from high to low on the slope. (4) Undulation: Undulation, which is the height difference within a small area, directly determines the runoff hydrodynamics. Evaporation factors: (5) Vegetation cover index (NDVI): The surface evaporation of rainfall is weaker in areas with lush vegetation. (6) Sunlight index: Sunlight intensity is directly proportional to the natural evaporation of rainfall. (7) Building normalization index (NDBI): The effects of artificial buildings and natural slopes on rainfall evaporation are different. (8) Distance from the road: Highways or roads generally contain chemicals such as asphalt, which can accelerate or reduce rainfall evaporation.

[0093] All factor layers are represented using a raster with a resolution of 30x30m, such as Figure 2 As shown, the corresponding factors are (a) undulation; (b) slope; (c) NDBI; (d) NDVI; (e) light index; (f) distance from road; (g) distance from river; and (h) lithology.

[0094] By inputting each factor into an arbitrary prediction model, the distribution map of the sensitivity index of the entire region can be obtained. Figure 3This study selected the random forest model. The random forest model combines multiple decision trees for classification and prediction. Its classifier is a recursive process from the root node to the child nodes. Starting from a node in the tree, a branch is determined based on the best feature between nodes, and the branching continues until the result of the tree is obtained. Finally, the subset of categories with the most votes is selected as the final output. By inputting various landslide rainfall sensitivity factors into the random forest model, the representative feature variable P obtained is the rainfall sensitivity value.

[0095] The entire study area was divided into 1851 slope units based on mountain shadows and lithological layers. Each slope unit is a local area. Then, based on the slope unit boundary, the average rainfall sensitivity of all grid cells within the slope unit was taken as the rainfall sensitivity value of that slope unit. Figure 4 Then, based on the rainfall sensitivity value, the refined rainfall threshold of each slope unit is calculated. The specific calculation formula is shown in equation (4), which realizes the refined transformation of the rainfall threshold from the whole to the local.

[0096]

[0097] In the formula, T is the local fine-tuned rainfall threshold, T S The baseline threshold is used for the entire region, and RSSI is the rainfall sensitivity index, ranging from (0-1). This is the baseline threshold coefficient, ranging from (0-1). The baseline threshold coefficient is taken from the statistical probability of a red alert for the baseline rainfall threshold across the entire study area, specifically determined by... Figure 1 The curves showing the relationship between cumulative rainfall and cumulative frequency of geological disaster events over a certain period before the landslide are shown. The baseline threshold coefficient μ corresponding to the 1-hour cumulative rainfall is 0.78, the baseline threshold coefficient μ corresponding to the 6-hour cumulative rainfall is 0.82, and the baseline threshold coefficient μ corresponding to the 24-hour cumulative rainfall is 0.6.

[0098] Finally, the reliability of the refined threshold was verified through historical typical disasters. From September 27th to 29th, 2016, affected by Typhoon Megi (No. 17), Pingyang County experienced widespread torrential rain, with a total rainfall of 461.7 mm. The maximum rainfall during the entire period (from 08:00 on September 27th to 14:00 on September 29th) reached 643.5 mm, with a maximum hourly rainfall intensity of 91.1 mm, a maximum 6-hour rainfall of 234.2 mm, and a maximum 24-hour rainfall of 380.4 mm. As many as 41 geological disasters and potential hazards occurred within Pingyang County, mainly distributed in Shuitou Town and Tengjiao Town. Figure 7 The refined rainfall threshold for each disaster is the refined rainfall threshold of the slope unit where the disaster point is located.

[0099] Of the geological disasters triggered by Typhoon "Catfish," 38 had recorded rainfall in the 24 hours preceding the disaster. Figure 8The rainfall events related to the disasters were recorded by the rain gauges closest to the disaster sites. The 38 disaster sites were distributed around 11 rain gauges (k3021, k3022, k3023, k3024, k3033, k3135, k3152, k3199, k3203, k3704, k3707). Ten disaster sites occurred near k3135, 24 near k3152, and 3 near k3199. The remaining rain gauges had only one disaster near each of their respective locations.

[0100] The actual rainfall at 38 disaster sites (H1-H38) was statistically analyzed for 1 hour, 6 hours, and 24 hours prior to the disaster (Table 2). A successful warning was considered issued when the actual rainfall exceeded the red warning threshold. Based on this, the success rates of the baseline and refined thresholds for 1 hour, 6 hours, and 24 hours were compared. The bolded portions in the table represent the successful warning thresholds. Figure 9 As shown, the prediction success rates of the baseline thresholds for 1h, 6h, and 24h were 47%, 63%, and 66%, respectively; the prediction success rates of the refined thresholds for 1h, 6h, and 24h were 55%, 79%, and 74%, respectively. The results indicate that using refined thresholds can effectively improve the early warning success rate, with an improvement of about 10%.

[0101] Table 2. Statistics on actual rainfall, baseline threshold, and refined threshold for geological disasters induced by "catfish" (a type of meteorological phenomenon).

[0102]

[0103] In this document, the directional terms such as front, back, top, and bottom are defined based on the position of the components in the accompanying drawings and their relative positions to each other, solely for the purpose of clarity and convenience in expressing the technical solution. It should be understood that these are relative concepts and can vary depending on different methods of use and placement; the use of these directional terms should not limit the scope of protection claimed in this application.

[0104] Where there is no conflict, the above embodiments and features described herein can be combined with each other.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A refined rainfall threshold calculation method based on landslide rainfall sensitivity correction, characterized in that, Includes the following steps: S1. Obtain historical landslide and rainfall data for the study area and construct a rainfall event database; S2. Obtain the total rainfall threshold T for the study area based on the statistical model. S ; S3. Analyze the characteristics of landslides induced by rainfall in the study area, select at least one landslide rainfall sensitivity factor from each of the three aspects of infiltration, runoff, and evaporation to carry out rainfall sensitivity evaluation, and obtain the rainfall sensitivity index RSSI of a local area in the study area. The method for calculating the local rainfall sensitivity index (RSSI) is as follows: S301. Input each landslide rainfall sensitivity factor into the prediction model to obtain the sensitivity index distribution map of the entire study area; S302. Based on the regional sensitivity index distribution map, calculate the average rainfall sensitivity value of the local area as the rainfall sensitivity index RSSI; The prediction model is a random forest model, and the classifier of the random forest model is... Where k represents the number of nodes, t i It is the number of decisions. I and I are the output variable and characteristic function, respectively; The marginal function is: The classification principle is as follows: in Let P be the probability space, representing the feature variable; S4. Correcting the regional baseline threshold T based on the rainfall sensitivity index. S Obtain the refined rainfall threshold T for this local area, where , This is the baseline threshold coefficient.

2. The refined rainfall threshold calculation method based on landslide rainfall sensitivity correction as described in claim 1, characterized in that: The overall sensitivity index distribution map is a raster map, and the rainfall sensitivity index RSSI is the average rainfall sensitivity of all raster cells in the local area.

3. The refined rainfall threshold calculation method based on landslide rainfall sensitivity correction as described in claim 1, characterized in that: The prediction model is a data-driven model.

4. The refined rainfall threshold calculation method based on landslide rainfall sensitivity correction as described in claim 1, characterized in that: The statistical model in step S2 is a cumulative rainfall model, and the formula for calculating cumulative rainfall is: Where Re is the effective rainfall; Rn is the rainfall in the previous n days; α is the effective rainfall coefficient; and n is the duration of rainfall.

5. The refined rainfall threshold calculation method based on landslide rainfall sensitivity correction as described in claim 1, characterized in that: The statistical model in step S2 is a cumulative rainfall-duration model, and the formula for calculating cumulative rainfall is: Where E represents the cumulative rainfall; D represents the duration of the rainfall; , c are statistical parameters.

6. The refined rainfall threshold calculation method based on landslide rainfall sensitivity correction as described in claim 1, characterized in that: The statistical model in step S2 is a rainfall intensity-duration model, and the formula for calculating rainfall intensity is: Where I represents rainfall intensity; D represents rainfall duration; , c are statistical parameters.

7. The refined rainfall threshold calculation method based on landslide rainfall sensitivity correction as described in claim 1, characterized in that: The statistical model in step S2 is cumulative rainfall minus rainfall intensity. Where I represents rainfall intensity; E represents cumulative rainfall; , c are statistical parameters.

8. The refined rainfall threshold calculation method based on landslide rainfall sensitivity correction as described in claim 1, characterized in that: Landslide rainfall sensitivity factors related to infiltration include lithology, topographic humidity index, distance from river, and groundwater level; landslide rainfall sensitivity factors related to runoff include slope, undulation, and slope variation coefficient; and landslide rainfall sensitivity factors related to evaporation include vegetation cover index, light index, and land use type.

Images