Method for constructing geological disaster meteorological risk early warning model based on rainfall pattern analysis

Abstract

The invention discloses a method for constructing a geological disaster meteorological risk early warning model based on rainfall pattern analysis. The method comprises the following steps: collecting basic geographic data and basic geological data of a research area; obtaining susceptibility grades of all slope units in the research area and obtaining effective landslide sample data induced by rainfall in the research area; establishing a rainfall pattern recognition module, and matching different types of rainfall threshold models for different rainfall patterns; and coupling the slope unit-based landslide susceptible partition and the rainfall threshold through the meteorological risk early warning criterion matrix so as to determine the geological disaster meteorological risk early warning of each slope unit and output the early warning result of the research area. The precision degree of the geological disaster meteorological risk early warning model can be improved, and the method has remarkable benefits in the field of geological disaster prevention and reduction.

Description

Technical Field

[0001] This invention relates to geological disaster early warning technology in the field of engineering geology, and in particular to a method for constructing a geological disaster meteorological risk early warning model based on rainfall pattern analysis. Background Technology

[0002] In recent years, with the development of disaster management and disaster prevention and mitigation decision-making services, it has been widely recognized that meteorological early warning of geological disasters has become a powerful means of risk prevention and control. Various regions are actively establishing meteorological early warning platforms for geological disasters. The meteorological risk early warning model for geological disasters is the core of the early warning system. This model uses rainfall data as input and dynamically outputs geological disaster early warning information, playing a crucial role in disaster prevention and mitigation.

[0003] Regional geological disaster early warning models have evolved from implicit statistical early warning models to explicit statistical early warning models. The former establishes an empirical statistical relationship between landslides and rainfall based on historical rainfall and landslide data, identifies critical rainfall intensities, and assumes that geological environmental information is implicitly included. The latter first determines the spatial susceptibility of landslides based on geological factors such as topography, stratigraphy, geological conditions, hydrology, and environment, and then couples this with a rainfall threshold model that induces landslides to establish early warning criteria.

[0004] In the prior art, 1) the method and apparatus for determining the critical rainfall threshold for landslide prediction disclosed in CN120297524B focuses on using machine learning to analyze the correlation between landslide density and rainfall parameters in extreme rainfall events, thereby determining a two-dimensional critical rainfall threshold. Although this improves the precision of early warning in extreme rainfall scenarios, its threshold model is mainly for the specific situation of extreme rainfall. When dealing with daily rainfall or multiple rainfall types, the universality of the threshold may be insufficient, and it does not fully consider the impact of the temporal distribution characteristics (i.e., rainfall pattern) of the rainfall process on the threshold; 2) the meteorological-based geological disaster risk early warning method, server, and storage medium disclosed in CN120580797A proposes a two-factor dynamic evaluation framework. By quantifying the geological background and the cumulative effect of recent rainfall, it generates a real-time dynamically adjusted current rainfall disaster threshold for each grid cell. This scheme aims to solve the problem that traditional early warning cannot adapt to the spatiotemporal differences of the geological environment. However, the core of this method lies in the coupling of geological background and rainfall effect, without involving the detailed identification of the temporal distribution (rain pattern) of the rainfall process itself, and it also lacks a mechanism for matching different threshold models for different rainfall patterns.

[0005] In summary, the currently used geological disaster early warning models have the following problems:

[0006] 1) The model does not fully consider the intrinsic geological environmental factors and external triggering factors that influence the development of geological disasters; the accuracy and refinement of the geological background data and geological disaster attribute data used in the model are low.

[0007] 2) The early warning model does not adequately consider rainfall data, neglecting rainfall patterns that significantly impact landslides, or it fails to use forecasted rainfall data to predict future risks. In cases of widespread rainfall or inaccurate rainfall forecasts, the early warning area may be too large, failing to meet the risk management needs for precise early warning.

[0008] 3) The early warning process has a low degree of automation. Technical processes such as the vectorization and import of rainfall data and the release of early warning products are not standardized and generally require manual operation by technical personnel, which cannot meet the response requirements of rapid and real-time early warning. Summary of the Invention

[0009] The purpose of this invention is to provide a method for constructing a meteorological risk early warning model for geological disasters based on rainfall pattern analysis in order to solve the problem of insufficient accuracy in current geological disaster early warning systems.

[0010] This invention achieves the above objective through the following technical solution: a method for constructing a geological disaster meteorological risk early warning model based on rainfall pattern analysis, the method comprising the following steps:

[0011] S1. Collect basic geographical and geological data of the study area;

[0012] S2. Through slope unit division and field investigation verification, landslide susceptibility zones based on slope units are identified, and the susceptibility distribution map is standardized to obtain the susceptibility level of all slope units in the study area. At the same time, by screening historical landslide logging data and historical geological disaster sample data, effective landslide sample data induced by rainfall in the study area are obtained.

[0013] S3. By using effective landslide sample data induced by rainfall and historical rainfall data, a rainfall pattern identification module is established, and different types of rainfall threshold models are matched for different rainfall patterns;

[0014] S4. By coupling the landslide susceptibility zone based on the slope unit and the rainfall threshold through the meteorological risk early warning criterion matrix, the meteorological risk early warning of geological disasters for each slope unit is determined, and the early warning results for the study area are output. At the same time, the actual effect of the early warning model is verified by the early warning inversion of the actual rainfall process.

[0015] As a further technical solution of the present invention, in S1:

[0016] Basic geographic data includes slope unit division, field survey verification, and historical landslide logging data;

[0017] Basic geological data include historical sample data of geological hazards and corresponding historical rainfall data, spatiotemporal distribution characteristics of rainfall data, real-time rainfall data, numerical rainfall forecast products, and spatial susceptibility distribution maps of geological hazards in the study area;

[0018] Among them, historical sample data of geological disasters includes specific time information of the occurrence of disasters; historical rainfall data includes meteorological satellite data, station rainfall data or similar data with spatial distribution characteristics, which must cover the time corresponding to the disaster, and the data time accuracy must reach daily rainfall or above; real-time rainfall data must completely cover the scope of the study area and be transmitted at a fixed time every day.

[0019] As a further technical solution of the present invention, in S2: the susceptibility distribution map includes at least four different susceptibility levels, and the susceptibility level with the largest area proportion within the slope unit is used to assign a value to the slope unit, thereby transforming the collected susceptibility distribution map into a susceptibility distribution result based on the slope unit.

[0020] As a further technical solution of the present invention, in S2: the screening of effective landslide sample data induced by rainfall includes removing non-rainfall-induced disaster points induced by human engineering activities, as well as invalid disaster sample points induced by extreme rainfall events.

[0021] As a further technical solution of the present invention, in S3, a rainfall pattern recognition module is established, and different types of rainfall threshold models are matched for different rainfall patterns, specifically including:

[0022] S31) The uniformity of rainfall temporal distribution is characterized by the Gini coefficient based on the Lorentz curve M. The mathematical expression for the Gini coefficient G is:

[0023] ;

[0024] ;

[0025] ;

[0026] ;

[0027] In the formula, S a S is the area enclosed by M and the diagonal y=x; b G is the area enclosed by M, the x-axis, and x = 100%; G ∈ [0, 1]. The closer G is to 1, the more concentrated the distribution; conversely, the closer G is to 1, the more uniform the distribution.

[0028] S32) The rainfall peak location coefficient is used to characterize the location of the rainfall peak, and its calculation formula is as follows:

[0029] ;

[0030] In the formula, r is the rain peak location coefficient; U max The unit time period where the rain peak occurs; U is the total duration of this rainfall event.

[0031] S33) Determine the Gini coefficient and rain peak location coefficient based on the real-time rainfall conditions in each slope unit, and divide the rainfall pattern identification module into early-stage, mid-stage, late-stage, and uniform types to achieve rapid identification of rainfall patterns;

[0032] S34) Based on rainfall pattern recognition, and based on the most relevant rainfall feature indicators, establish a zoned classification and grading landslide rainfall threshold based on rainfall pattern. For different rainfall patterns, use different rainfall thresholds to convert the input rainfall data into rainfall levels and extract them to each slope unit.

[0033] After determining the type of rainfall threshold (S35), the rainfall threshold curve is further fitted.

[0034] As a further technical solution of the present invention, in S34), after identifying the rainfall pattern through the rainfall pattern module, the most relevant rainfall characteristic index considered in the rainfall threshold is determined through Pearson correlation analysis, which includes:

[0035] S341) Rainfall intensity-duration threshold, calculated using the following formula:

[0036] I=α1×D β1 ;

[0037] In the formula, I is the rainfall intensity that triggers the landslide; D is the duration of the rainfall event that triggers the landslide; the threshold curve is obtained by fitting historical disaster samples and their rainfall characteristics, α1 and β1 are the fitting parameters of the threshold curve, where α1 is the intercept of the curve, β1 is the slope of the curve, and β1 < 0 in the ID threshold.

[0038] S342) Cumulative rainfall - duration threshold, calculated using the following formula:

[0039] E=α2×D β2 ;

[0040] In the formula, E is the cumulative rainfall that triggers the landslide; D is the duration of the rainfall event that triggers the landslide; α2 and β2 are the fitting parameters of the threshold curve, and β2>0 in the ED threshold.

[0041] S343) The formula for calculating the effective rainfall in the preceding period minus the rainfall intensity threshold is:

[0042] R e =α3×Iβ3 ;

[0043] In the formula, I represents the rainfall intensity that triggers the landslide event; α3 and β3 are the fitting parameters of the threshold curve; R e The effective rainfall in the preceding period can be determined using a power-law method, which is obtained by multiplying the daily rainfall over a certain period by the effective rainfall coefficient. The calculation formula is as follows:

[0044] R e =R0+θR1+θ 2 R2+…+θ n R n ;

[0045] In the formula: R e R0 represents the effective rainfall; R0 represents the daily rainfall; R n θ represents the rainfall on the nth day; θ is the effective rainfall coefficient.

[0046] As a further technical solution of the present invention, in S35, after determining the type of rainfall threshold, a rainfall threshold curve is further fitted, specifically including:

[0047] S351) A scatter plot is drawn using rainfall event data points as samples, and rainfall threshold curves are fitted according to the occurrence proportions of 5%, 30%, 50% and 80% of landslides in the region.

[0048] (S352) Rainfall levels are determined by rainfall thresholds for each zone and are divided into five levels: no risk, low risk, medium risk, high risk, and extremely high risk. These levels correspond to the threshold curves in the rainfall threshold curve diagram. No risk corresponds to a threshold range of <5%, low risk to a threshold range of 5% to 30%, medium risk to a threshold range of 30% to 50%, high risk to a threshold range of 50% to 80%, and extremely high risk to a threshold range of >80%.

[0049] As a further technical solution of the present invention, in S4, the meteorological risk warning criterion matrix uses the geological disaster susceptibility level and rainfall level as the horizontal and vertical axes, respectively, and serves as the basis for judging the warning level of each slope unit.

[0050] A geological disaster meteorological risk early warning model, which is capable of providing meteorological risk warnings; also includes:

[0051] The early warning model is equipped with a dynamic rainfall pattern recognition and rainfall threshold matching mechanism;

[0052] The early warning model uses real-time rainfall data and forecast rainfall products as input data;

[0053] The early warning model dynamically outputs early warning results of different levels based on slope units.

[0054] The beneficial effects of this invention are:

[0055] 1) This invention achieves accurate matching and dynamic calling of rainfall thresholds. By introducing a rainfall pattern identification module based on the Gini coefficient and the rain peak position coefficient, it for the first time uses the temporal distribution characteristics of the rainfall process as the key basis for threshold selection. Different rainfall threshold models are matched and used for different rainfall patterns such as early-stage, mid-stage, late-stage, and uniform types. This completely changes the problem of poor adaptability of single and static threshold models in the existing technology, enabling the early warning model to more realistically reflect the differences in the landslide triggering mechanism of different rainfall processes. Thus, under complex and ever-changing rainfall conditions, it can still output high-precision early warning results, significantly reducing missed and false alarms caused by threshold mismatch.

[0056] 2) This invention constructs a fully automated early warning mechanism from data input to result output: it integrates real-time rainfall data, forecast rainfall products, slope unit susceptibility zoning, and dynamic rainfall pattern identification and threshold matching, and establishes a standardized automatic processing flow. It overcomes the drawbacks of relying on manual intervention and fragmented processes, and can achieve rapid and real-time early warning without human intervention. This greatly improves the response speed and work efficiency of geological disaster early warning and meets the stringent timeliness requirements of emergency management departments.

[0057] 3) This invention not only considers the dynamic changes in rainfall, but also finely characterizes the spatial heterogeneity of the internal disaster-prone environment by dividing the slope unit into spatial susceptibility zones. Finally, it couples the two through an early warning matrix to determine the geological disaster meteorological risk warning level of each slope unit and output the meteorological risk warning results for the study area.

[0058] 4) This invention can automatically classify rainfall patterns and match different types of rainfall thresholds for different rainfall patterns using real-time rainfall data and forecast rainfall products as input. After calculation by the early warning model, it can dynamically output early warning results of different levels based on slope units, which can improve the accuracy of geological disaster meteorological risk early warning models and has significant benefits in the field of geological disaster prevention and mitigation. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the process of the present invention;

[0060] Figure 2 This is a geological environment-based study area and disaster point distribution map of Embodiment 3 of the present invention;

[0061] Figure 3 This is the slope unit division process based on hydrological analysis in Embodiment 3 of the present invention;

[0062] Figure 4 This is a susceptibility zoning map of the study area based on slope units in Embodiment 3 of the present invention;

[0063] Figure 5 This is a schematic diagram of the Lorentz curve and the generalized model of the rainfall pattern in Embodiment 3 of the present invention;

[0064] Figure 6 The rainfall threshold for geological disasters in the study area of ​​Embodiment 3 of this invention;

[0065] Figure 7 This is the geological disaster meteorological risk early warning matrix in Embodiment 3 of the present invention;

[0066] Figure 8 This is a daily rainfall distribution map of a typical rainfall process in Embodiment 3 of the present invention;

[0067] Figure 9 This is a geological disaster meteorological risk warning map for a typical rainfall process in Embodiment 3 of the present invention. Detailed Implementation

[0068] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] Example 1, as Figure 1 As shown in the figure, this embodiment provides a method for constructing a meteorological risk early warning model for geological disasters based on rainfall pattern analysis. The method specifically includes the following steps:

[0070] First, collect basic geographical and geological data of the study area.

[0071] Basic geographic data includes slope unit division, field survey verification, and historical landslide recording data.

[0072] Basic geological data includes historical sample data of geological hazards and corresponding historical rainfall data, spatiotemporal distribution characteristics of rainfall data, real-time rainfall data, numerical rainfall forecast products, and spatial susceptibility distribution maps of geological hazards in the study area. Among them, historical sample data of geological hazards includes specific time information of hazard occurrence, historical rainfall data includes meteorological satellite data, station rainfall data, or similar data with spatial distribution characteristics, which must cover the time corresponding to the hazard, and the data time accuracy must reach daily rainfall or above. Real-time rainfall data must completely cover the scope of the study area and be transmitted at a fixed time every day.

[0073] Second, through slope unit division and field investigation verification, landslide-prone areas based on slope units were identified, and the susceptibility distribution map was standardized to obtain the susceptibility level of all slope units in the study area. At the same time, by screening historical landslide logging data and historical geological disaster sample data, effective landslide sample data induced by rainfall in the study area were obtained.

[0074] The susceptibility distribution map includes at least four different susceptibility levels. The susceptibility distribution map can be drawn manually based on the topographic map or automatically through an auxiliary program. The susceptibility level of the unit is assigned based on the largest area proportion within the slope unit, and the collected susceptibility distribution map is transformed into a susceptibility distribution result based on the slope unit.

[0075] The screening of valid landslide sample data induced by rainfall includes removing non-rainfall-induced disaster points caused by human engineering activities such as road excavation and slope cutting for housing construction, as well as invalid disaster sample points induced by extreme rainfall events.

[0076] Third, by using effective landslide sample data induced by rainfall and historical rainfall data, a rainfall pattern identification module is established, and different types of rainfall threshold models are matched for different rainfall patterns.

[0077] A rainfall pattern recognition module is established, and different types of rainfall threshold models are matched for different rainfall patterns, including:

[0078] 1) The uniformity of rainfall temporal distribution is characterized by the Gini coefficient based on the Lorenz curve M, where the vertical and horizontal axes represent the cumulative rainfall percentage and cumulative time percentage, respectively, and the diagonal line is the absolute mean line, indicating that the rainfall intensity remains constant. The mathematical expression for the Gini coefficient G is:

[0079] ;

[0080] ;

[0081] ;

[0082] ;

[0083] In the formula, S a S is the area enclosed by M and the diagonal y=x; b G is the area enclosed by M, the x-axis, and x = 100%; G ∈ [0, 1]. The closer G is to 1, the more concentrated the distribution; conversely, the closer G is to 1, the more uniform the distribution.

[0084] 2) The peak location coefficient is used to characterize the location of the rainfall peak. The peak location coefficient is the ratio of the time period during which the peak occurs to the total duration of the rainfall event. The calculation formula is as follows:

[0085] ;

[0086] In the formula, r is the rain peak location coefficient; U max The unit time period where the rain peak occurs; U is the total duration of this rainfall event.

[0087] Based on the peak location coefficient, rainfall events can be qualitatively classified into three types: front (0 < r ≤ 0.33), middle front (0.33 < r ≤ 0.66), and rear front (0.66 < r ≤ 1).

[0088] 3) Based on the real-time rainfall conditions within each slope unit, the Gini coefficient and rain peak location coefficient are determined. The rainfall pattern identification module (rain pattern mode) can be divided into early-stage, mid-stage, late-stage, and uniform types to achieve rapid identification of rainfall patterns.

[0089] 4) Based on rainfall pattern recognition, and using multiple most relevant rainfall characteristic indicators such as previous effective rainfall, daily rainfall, peak rainfall intensity, and duration, establish a zoned classification and grading landslide rainfall threshold based on rainfall pattern. Different rainfall thresholds are used for different rainfall patterns to convert the input rainfall data into rainfall levels and extract them to each slope unit.

[0090] In step 4), after identifying the rainfall pattern using the rainfall pattern module, Pearson correlation analysis is used to determine the most relevant rainfall characteristic indicators considered in the rainfall threshold, which include:

[0091] 41) Rainfall intensity-duration (ID) threshold, calculated using the following formula:

[0092] I=α1×D β1 ;

[0093] In the formula, I is the rainfall intensity that triggers the landslide; D is the duration of the rainfall event that triggers the landslide; the threshold curve is obtained by fitting historical disaster samples and their rainfall characteristics, α1 and β1 are the fitting parameters of the threshold curve, where α1 is the intercept of the curve, β1 is the slope of the curve, and β1 < 0 in the ID threshold.

[0094] 42) The cumulative rainfall duration (ED) threshold is calculated using the following formula:

[0095] E=α2×D β2 ;

[0096] In the formula, E is the cumulative rainfall that triggers the landslide; D is the duration of the rainfall event that triggers the landslide; α2 and β2 are the fitting parameters of the threshold curve, and β2>0 in the ED threshold.

[0097] 43) The threshold for effective rainfall in the preceding period minus rainfall intensity (EI) is calculated using the following formula:

[0098] R e =α3×I β3 ;

[0099] In the formula, I represents the rainfall intensity that triggers the landslide event; α3 and β3 are the fitting parameters of the threshold curve; R e The effective rainfall in the preceding period can be determined using a power-law method, which is obtained by multiplying the daily rainfall over a certain period by the effective rainfall coefficient. The calculation formula is as follows:

[0100] R e =R0+θR1+θ 2 R2+…+θ n R n ;

[0101] In the formula: R e R0 represents the effective rainfall; R0 represents the daily rainfall; R n θ represents the rainfall on the nth day; θ is the effective rainfall coefficient.

[0102] 5) After determining the type of rainfall threshold, the rainfall threshold curve is further fitted.

[0103] Step S5 includes:

[0104] 51) Plot a scatter plot using rainfall event data points as samples, and fit rainfall threshold curves according to the occurrence proportions of 5%, 30%, 50%, and 80% of landslides in the region;

[0105] 52) Rainfall levels are determined by rainfall thresholds for each zone and are divided into five levels: no risk, low risk, medium risk, high risk, and extremely high risk. These correspond to the threshold curves in the rainfall threshold curve diagram. No risk corresponds to a threshold range of <5%, low risk to a threshold range of 5% to 30%, medium risk to a threshold range of 30% to 50%, high risk to a threshold range of 50% to 80%, and extremely high risk to a threshold range of >80%.

[0106] Fourth, by coupling the landslide susceptibility zones based on slope units and the rainfall threshold through the meteorological risk early warning criterion matrix, the geological disaster meteorological risk early warning for each slope unit is determined, and the early warning results for the study area are output. At the same time, the actual effect of the early warning model is verified by the early warning inversion of the actual rainfall process.

[0107] The meteorological risk warning criterion matrix uses the geological disaster susceptibility level and rainfall level as the horizontal and vertical axes, respectively, and serves as the basis for determining the warning level of each slope unit.

[0108] Example 2: This example provides a geological disaster meteorological risk early warning model, which can provide meteorological risk warnings and is constructed using the method described in Example 1.

[0109] The early warning model is equipped with a dynamic rainfall pattern recognition and rainfall threshold matching mechanism. The early warning model takes real-time rainfall data and forecast rainfall products as input data, and dynamically outputs early warning results of different levels based on slope units.

[0110] The slope unit is a unit type in the geological hazard susceptibility zoning. The early warning model simultaneously considers the regional landslide susceptibility, rainfall type characteristics and corresponding rainfall thresholds, and the impact of previous effective rainfall and future forecast rainfall on landslide occurrence, realizing the coupled response of rainfall type identification, threshold matching and result output.

[0111] Example 3: This example provides a method for constructing a geological disaster meteorological risk early warning model based on rainfall pattern analysis. This method is particularly suitable for early warning and forecasting of rainfall-induced landslides in the hilly areas of Southwest China, and specifically includes:

[0112] First, we collected historical sample data of geological disasters in a city in Southwest China, along with corresponding historical rainfall data, real-time updated rainfall data, numerical rainfall forecast products, and a regional spatial susceptibility distribution map of geological disasters.

[0113] The geological disaster historical sample data includes all geological disaster risk points that have occurred in a city in Southwest China over the past ten years, and includes specific information such as the location, time, and type of the disaster; the historical rainfall data includes daily rainfall data from more than 2,000 rain gauge stations in the city in Southwest China at the corresponding time; the numerical rainfall forecast product is the rainfall forecast data for the next 24 hours, which is updated in real time by connecting to the meteorological department's rainfall monitoring and rainfall forecasting system, and the forecast rainfall data for the next day is updated at 19:00 every day.

[0114] Taking a city in Southwest China as an example, the overall situation of the collected geological hazard susceptibility assessment data is as follows: Considering all possible influences of landslide triggering mechanisms and the contribution of corresponding influencing factors to regional disasters, based on detailed geological hazard survey data of the city in Southwest China, the disaster-inducing environment of the city in Southwest China was fully studied. After analysis, eight factors were selected as landslide susceptibility assessment factors for the city: slope structure type, density of disaster-inducing points and disaster-inducing bodies, engineering geological rock group, slope, soil layer thickness, distance from water system, rock layer dip angle, and geological structure. Based on information volume analysis, stratification analysis, and field survey results, the geological hazard susceptibility values ​​of 66,077 slope units in the city were obtained. Combined with the field survey results of each district and county in the city, the geological hazard susceptibility values ​​were divided into four levels: non-susceptible, low-susceptible, medium-susceptible, and high-susceptible. Of these, the area of ​​high-risk geological disaster zones is approximately 17,100 square kilometers, accounting for about 20.83% of the city's total area; the area of ​​medium-risk zones is approximately 34,700 square kilometers, accounting for about 42.19% of the city's total area; the area of ​​low-risk zones is approximately 26,400 square kilometers, accounting for about 32.01% of the city's total area; and the area of ​​non-risk zones is approximately 4,100 square kilometers, accounting for about 4.97% of the city's total area.

[0115] Second, the collected historical geological disaster sample data were screened to obtain valid disaster samples induced by rainfall. At the same time, the susceptibility distribution map was standardized to obtain the susceptibility level of all slope units in the region.

[0116] The accuracy of the effective disaster sample induced by rainfall has a decisive impact on the determination of rainfall thresholds. On the one hand, it is necessary to screen for geological disaster points induced by rainfall from all potential disaster sites, excluding those induced by human engineering activities such as road excavation and slope cutting for housing construction. On the other hand, rainfall events exceeding the rainfall threshold may induce landslides, but the probability of landslides induced by different rainfall events varies greatly: compared to ordinary rainfall events, extreme rainfall events undoubtedly have a much higher probability of inducing landslides. Historically, there are many records of extreme rainfall events inducing a large number of landslides. If landslide disasters induced by such events are included in the statistical analysis of rainfall thresholds, the value of the rainfall threshold will be greatly increased. However, the probability of extreme rainfall events is low. Applying a higher rainfall threshold obtained from low-probability events to daily geological disaster early warning management may lead to a lower warning level and increase the probability of missed geological disaster reports. Therefore, when calculating rainfall thresholds, it is necessary to reasonably exclude historical extreme heavy rainfall events to ensure that the constructed rainfall thresholds can better serve daily geological disaster early warning and control.

[0117] Taking the Three Gorges Reservoir area as an example, the stability analysis of landslide disasters in the Three Gorges Reservoir area considers the rainstorm conditions, defining rainfall events with a return period of more than fifty years as extreme rainfall events. There are three commonly used methods for studying the return period of rainfall: Pearson Type III distribution, exponential distribution, and Gumbel distribution. This embodiment uses the Gumbel distribution to determine rainfall extremes.

[0118] The Gumbel distribution can be written in the following common form:

[0119] (a>0,-∞ <u<∞);

[0120] To use it for extreme rainfall simulation, the parameters α and μ must first be estimated. Commonly used estimation methods include the method of moments, Gumbel's method, least squares method, and maximum likelihood method.

[0121] This embodiment provides parameter estimation for the Gumbel method:

[0122] ;

[0123] in, The standard deviation is the sample standard deviation.

[0124] ;

[0125] m=1,2,…,N;

[0126] Therefore, the probability of a random variable ξ greater than x is:

[0127] ;

[0128] Conversely, given a probability value P or a recurrence value T, the intensity R of a certain extreme rainfall event occurring with probability P can be calculated. P Or the intensity R of a certain extreme rainfall occurring with a certain return period T. T :

[0129] ;

[0130] ;

[0131] Due to differences in external factors such as meteorological conditions, hydrogeological environment, geomorphological factors, and human activities in the study area, as well as differences in infrastructure such as data accuracy and specific observation equipment, the threshold model parameters and expressions vary significantly across different regions. In this embodiment, the study area is divided into four sub-regions based on differences in geological environmental characteristics, such as... Figure 2As shown. After landslide sample screening, a total of 2142 data points were used for landslide rainfall threshold analysis in each region, including 271 data points in region 1, 411 data points in region 2, 241 data points in region 3, and 1219 data points in region 4.

[0132] For the four pre-defined study areas, representative rain gauges were selected and the rainfall in each area was analyzed using the Gumbel distribution under specific return period conditions. The rainfall with a 50-year return period was used as the upper limit of the rainfall threshold.

[0133] Standardization of geological hazard susceptibility involves two main steps: the method of dividing slope units and the transformation of susceptibility results. If there is no basic susceptibility data for the study area, a susceptibility assessment is required. This assessment necessitates the division of slope units, which can be manually drawn based on topographic maps or automatically using auxiliary programs. This embodiment utilizes the hydrological analysis function of ArcGIS software to extract slope units, and then manually modifies them based on field surveys and actual topography, adjusting unreasonable unit boundaries to obtain the final unit division result: the entire 82402 km² area of ​​a city in Southwest China is divided into units. 2 The area was divided into 66,077 slope elements. The slope element division process is as follows: Figure 3 As shown, large river systems such as the Yangtze River, the Wujiang River, their tributaries, reservoirs, and lakes were avoided during the field survey phase. The total area of ​​the slope units in the entire city of a certain city in Southwest China is 82,398.02 km². 2 The slope unit density reaches 1.25 units / km. 2 The slope units are numbered 1-66077. If the collected susceptibility distribution map is based on the evaluation results of other slope units, then the slope units are first divided, and then the susceptibility level of the slope unit is assigned using the largest area proportion within the slope unit. The geological hazard susceptibility zoning of the study area based on slope units is as follows: Figure 4 As shown.

[0134] Third, by analyzing the temporal distribution characteristics of historical rainfall data, a rainfall pattern identification module is established. Since the temporal distribution characteristics of the historical rainfall data used differ, rainfall pattern thresholds corresponding to different rainfall pattern identification modules can be constructed.

[0135] In this embodiment, the Gini coefficient based on the Lorenz curve M is used to represent the uniformity of the distribution characteristics of the rainfall data time history. Figure 5 a is a schematic diagram of the Lorenz curve, where the diagonal line is the absolute average line, and the vertical and horizontal axes are the cumulative rainfall percentage and the cumulative time percentage, respectively; Figure 5 b is a schematic diagram of the generalized model of rain patterns; Figure 5 c represents the cumulative rainfall distribution curve of the actual rainfall event.

[0136] In this embodiment, the correlation coefficients between multiple indicators, such as prior effective rainfall, peak rainfall intensity, and rainfall duration, and whether a landslide occurs are calculated to screen rainfall characteristics. Taking late-peak rainfall as an example, disaster points induced by late-peak rainfall are identified through screening. Rainfall sequences from the same region during non-landslide periods are randomly selected to create a control sample, and the correlation coefficients between rainfall characteristics of each late-peak rainfall event and whether a landslide occurs are calculated. The results show that for this type of rainfall, the correlation coefficient between prior effective rainfall and whether a landslide occurs is the highest. Therefore, this indicator is selected to construct a rainfall threshold model for late-peak rainfall. A scatter plot is created with the cumulative effective rainfall of the previous 9 days for the sample points as the horizontal axis and the daily rainfall as the vertical axis. Threshold curves are fitted according to the 5%, 30%, 50%, and 80% occurrence rates of landslides in the region, as shown below. Figure 6 As shown. Based on the obtained rainfall threshold, the input rainfall data is converted into rainfall levels and extracted to each slope unit, thus obtaining the rainfall level of each unit.

[0137] Fourth, by coupling the spatial susceptibility level of geological disasters and the rainfall threshold through the meteorological risk warning criterion matrix, the meteorological risk warning level of geological disasters for each slope unit is determined, the meteorological risk warning results of geological disasters in the study area are output, and the actual effect of the meteorological risk warning model of geological disasters is verified by the warning inversion of actual rainfall processes.

[0138] The meteorological risk warning criterion matrix uses rainfall level as the horizontal axis and the spatial susceptibility level of geological hazards as the vertical axis to determine the meteorological risk warning level for geological hazards of each slope unit under different working conditions. The meteorological risk warning criterion matrix used in this embodiment is as follows: Figure 7 As shown.

[0139] The analysis and early warning results for this embodiment were selected from September 1st and September 2nd of the "831" extreme rainfall event in 2014. The daily rainfall distribution of this rainfall event is as follows: Figure 8 Geological disaster meteorological risk warning results such as Figure 9 Nine districts and counties experienced torrential rain during this rainfall event, with some areas recording daily rainfall exceeding 385 millimeters, the highest rainfall ever recorded in the region. This triggered numerous geological disasters in the northeastern part of the city, resulting in casualties and significant economic losses.

[0140] Combining rainfall distribution and geological disaster meteorological risk early warning results ( Figure 9It can be seen that there is a good correlation between the concentrated rainfall areas and the red warning areas, and the occurrence of landslide disasters is mainly based on immediate response, with some delays. The geological disaster meteorological risk warning results show that the geological disaster meteorological risk warning model performs well for this event, with most landslides induced by heavy rain located within the warning area. The accuracy rate for immediate response warnings is over 95%. It is worth noting that on September 2nd, as the widespread heavy rainfall subsided, the main warning area also slightly decreased, but a large area of ​​slope units still maintained a high warning level, proving that many disasters occurred in the warning areas, validating the good warning capability of this model (geological disaster meteorological risk warning model).

[0141] Work Process: First, background data (basic geographic data) and existing geological hazard assessment results (basic geological data) of the study area were collected from two aspects: geological hazard-inducing conditions and landslide-inducing factors. Second, a generalized rainfall pattern model was established based on historical rainfall data pattern analysis, and the collected historical geological hazard sample data were screened and classified to obtain effective hazard samples induced by different rainfall patterns. Then, rainfall threshold models corresponding to different rainfall patterns were constructed using different characteristic indicators such as previous effective rainfall and current rainfall. Based on the rainfall threshold model, the input rainfall data was converted into rainfall levels and extracted to each slope unit. The susceptibility distribution map was standardized to obtain the susceptibility level of all slope units in the region. Finally, the spatial susceptibility of geological hazards and rainfall thresholds were coupled through a risk warning criterion matrix to determine the meteorological risk warning level of geological hazards for each unit.

[0142] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0143] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for constructing a meteorological risk warning model of geological disasters based on rainfall pattern analysis, characterized in that, The method comprises the following steps: S1, collecting basic geographic data and basic geological data of the study area; S2, by slope unit division and field investigation verification, landslide prone distribution area based on slope unit is carried out, the susceptibility distribution map is standardized, the susceptibility grade of all slope units in the study area is obtained, at the same time, by sample screening of historical landslide catalog data and geological disaster historical sample data, the effective landslide sample data induced by rainfall in the study area is obtained; S3, by the effective landslide sample data induced by rainfall and the historical rainfall data, a rainfall pattern recognition module is established, and different types of rainfall threshold models are matched for different rainfall patterns; S4, by meteorological risk early warning criterion matrix, the landslide prone distribution based on slope unit and the rainfall threshold are coupled, so as to determine the geological disaster meteorological risk early warning of each slope unit, and output the early warning result of the study area, at the same time, through the early warning inversion of the actual rainfall process, the actual effect of the early warning model is verified.

2. The method of claim 1, wherein, In the S1: The basic geographic data includes slope unit division, field investigation verification and historical landslide catalog data; The basic geological data includes geological disaster historical sample data and corresponding historical rainfall data, rainfall spatio-temporal distribution characteristic data, real-time rainfall data, numerical rainfall forecast product and susceptibility distribution map of the study area geological disaster space; The geological disaster historical sample data includes specific time information of disaster occurrence; the historical rainfall data includes meteorological satellite data, station rainfall data or similar data with spatial distribution characteristics, which needs to cover the corresponding time of disaster, and the data time accuracy reaches daily rainfall and above; the real-time rainfall data needs to completely cover the range of the study area and be transmitted at fixed time every day.

3. The method of claim 2, wherein, In the S2: The susceptibility distribution map includes at least four different susceptibility grades, and the susceptibility grade with the largest area ratio in the slope unit range is used to assign values to the slope unit, so as to convert the collected susceptibility distribution map into the susceptibility distribution result based on slope unit.

4. The method of claim 2, wherein, In the 2: The screening of the effective landslide sample data induced by rainfall includes eliminating non-rainfall-induced disaster points induced by human engineering activities and invalid disaster sample points induced by extreme rainfall events.

5. The method of claim 1, wherein, In the S3, the rainfall pattern recognition module is established, and different types of rainfall threshold models are matched for different rainfall patterns, which specifically includes: S31) the Gini coefficient based on Lorenz curve M is used to represent the uniformity of rainfall time distribution, and the mathematical expression of Gini coefficient G is: ； ； ； ； In the formula, S a is the area enclosed by M and the diagonal y = x; S b is the area enclosed by M and the x-axis and x = 100%; G ∈ [0, 1], the closer G is to 1, the more concentrated the distribution, and vice versa; S32) the rain peak position coefficient is used to represent the rain peak position, and the calculation formula is: ； where r is the rain peak position coefficient; U max is the unit time period of the rain peak occurrence; U is the total duration of the rainfall of the field S33) according to the real-time rainfall condition in each slope unit, the Gini coefficient and the rain peak position coefficient are determined, and the rainfall pattern recognition module is divided into early type, middle type, late type and uniform type, so as to realize the rapid identification of rainfall pattern; S34) on the basis of rainfall pattern recognition, based on the most relevant rainfall characteristic index, the zoning classification grading landslide rainfall threshold based on rainfall pattern is established, different rainfall thresholds are used for different rainfall patterns to convert the input rainfall data into rainfall grade, and are extracted to each slope unit; S35) After determining the type of rainfall threshold, further fitting the rainfall threshold curve is obtained.

6. The method of claim 5, wherein, In the S34), after identifying the rainfall pattern by the rainfall pattern module, the most relevant rainfall feature index considered in the rainfall threshold is determined by the Pearson correlation analysis, which includes: S341) Rainfall intensity-duration threshold, the calculation formula is: I = a1 x D β1 ; In the formula, I is the rainfall intensity of the landslide-inducing rainfall event; D is the duration of the landslide-inducing rainfall event; the threshold curve is fitted according to the historical disaster sample and its rainfall characteristics, and a1, b1 are the fitting parameters of the threshold curve, wherein a1 is the intercept of the curve, b1 is the slope of the curve, and b1<0 in the I-D threshold; S342) Accumulated rainfall-duration threshold, the calculation formula is: E = a2 x D β2 ; In the formula, E is the accumulated rainfall of the landslide-inducing rainfall event; D is the duration of the landslide-inducing rainfall event; a2, b2 are the fitting parameters of the threshold curve, and b2>0 in the E-D threshold; S343) Antecedent effective rainfall-rainfall intensity threshold, the calculation formula is: R e = α3x I β3 ; where I is the rainfall intensity of the landslide-inducing rainfall event; a3, b3 are the fitting parameters of the threshold curve; R e is the effective rainfall, which can be determined in the form of a power index. It is obtained by multiplying the daily rainfall in a period of time by the effective rainfall coefficient, and the calculation formula is: R e = R0+ θR1+ θ 2 R2+ … + θ n R n ; wherein R e is effective rainfall; R0is the rainfall on the day; R n is the rainfall on the nth day before; and θ is the effective rainfall coefficient.

7. The method of claim 5, wherein, In the S35, after determining the type of rainfall threshold, further fitting the rainfall threshold curve is obtained, which specifically includes: S351) Scatter plot is drawn with rainfall event data points as samples, and rainfall threshold curve is fitted according to 5%, 30%, 50% and 80% occurrence proportions of regional landslide occurrence; S352) Rainfall grade is determined by the rainfall threshold of each partition, which is divided into five grades of no danger, low danger, medium danger, high danger and extremely high danger, which correspond to the threshold curves in the rainfall threshold curve graph respectively, and the threshold ranges of no danger, low danger, medium danger, high danger and extremely high danger are <5%, 5%-30%, 30%-50%, 50%-80% and >80% respectively.

8. The method of claim 1, wherein: In the S4, the meteorological risk early warning criterion matrix takes the geological disaster susceptibility grade and the rainfall grade as the horizontal and vertical axes respectively, and serves as the basis for determining the early warning grade of each slope unit.

9. A geological disaster meteorological risk early warning model, characterized in that, The early warning model is constructed by the method of any one of claims 1-8, and the early warning model can perform meteorological risk early warning; Further comprising: The early warning model is provided with a rainfall pattern dynamic identification and rainfall threshold matching mechanism; The early warning model takes real-time rainfall data and forecast rainfall products as input data; The early warning model dynamically outputs early warning results of different grades based on slope units.

Images