A data-driven method for predicting geological hazards potential in typical granite weathering soils

Abstract

This invention discloses a data-driven method for predicting the potential of geological hazards in typical granite weathered land, comprising: acquiring geological hazard data, rainfall observation data, and rainfall forecast data; constructing an I-D fitting model based on the geological hazard data and rainfall observation data; constructing a geological hazard probability prediction model based on the I-D fitting model; inputting the rainfall forecast data into the geological hazard probability prediction model to predict the probability of geological hazard occurrence; and using more refined hourly automatic weather station precipitation data to assess the previous effective rainfall and the current day's forecast rainfall, thereby establishing a two-parameter geological hazard prediction model applicable to the Wuzhou City area.

Description

Technical Field

[0001] The invention belongs to the technical field of geological disaster prediction, and in particular relates to a data-driven typical granite weathering soil geological disaster potential prediction method. Background Art

[0002] . There are many factors that affect and induce the occurrence of geological disasters, mainly including two categories: internal factors and external factors. Internal factors refer to the density of geological disaster development and the basic geological environment, such as landform type, slope height difference, geological structure, hydrological conditions, etc.; external factors refer to precipitation, human engineering activities, geological movements, etc. Among them, the internal factors are highly stable and can be easily determined through regional or historical surveys, but the external factors are highly volatile, especially precipitation. Once the atmospheric circulation conditions are mature, regional and local precipitation is almost changing rapidly, so the impact of precipitation on geological disasters has always been the research focus of scholars in the field of geological disasters. Precipitation is the main cause of geological disasters. Most types of geological disasters occur from April to October, which coincides with the flood season (rainy season) in various places. Therefore, more and more studies are focusing on how to determine the precipitation threshold that triggers geological disasters, focusing on the determination of the threshold of daily precipitation or previous effective precipitation in different regions and different geological types, and designing a geological disaster meteorological early warning model based on this, and have achieved good results. For example, Wang Kun et al. (2022) pointed out that there are differences in the precipitation thresholds for geological disasters in the Ruoergai area of ​​Sichuan through reverse query and correlation analysis. The daily precipitation and the precipitation in the previous 7 days were analyzed respectively, and it was pointed out that when the daily precipitation is greater than 35 mm, the probability of a geological disaster occurring the next day is 90%, and the cumulative precipitation in the previous 7 days has the best correlation with the occurrence of geological disasters; Yang Senlin et al. (2012) analyzed the precipitation and geological disaster data of Guizhou Province, and used the sum of the daily rainfall and the cumulative rainfall in the previous 5 days as the triggering threshold for geological disasters; Li Minghua (1985) used the data of tens of thousands of landslides triggered by heavy rains in Sichuan Province, and proposed a combination model of regional critical daily rainfall and previous rainfall for landslides triggered by heavy rains through cluster analysis. Brand et al. (1988) analyzed rainfall data and pointed out that landslides almost always occurred within 4 hours after the peak of rainfall, which was almost the same as the peak of heavy rain; the previous rainfall was not the dominant factor in most landslides. Lü Da (2016) pointed out that the relationship between the heavy rainfall on that day in Fanchang County, Anhui Province and geological disasters was the closest, and the effective rainfall of the previous rainfall type was an important factor in meteorological warning of geological disasters.

[0003] However, since the disaster-pregnancy conditions and triggering mechanisms of geological disasters in different regions are different, the research results from other regions cannot be directly applied to other regions.

[0004] Given that most of the current studies on the critical daily rainfall in the region that triggers landslides and other geological disasters do not distinguish between geological types and are based on daily precipitation data, with the continuous enrichment of hourly observation data and the increasing updating of analysis methods, the original geological meteorological early warning model cannot meet the current requirements for the precise monitoring, early warning and prevention of geological disasters in different time periods, regions and types. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes a data-driven typical granite weathering soil geological disaster potential forecasting method, which uses refined hourly meteorological automatic station precipitation data to evaluate the previous effective rainfall and the forecast precipitation for the day.

[0006] To achieve the above purpose, the present invention proposes a data-driven typical granite weathering soil geological disaster potential prediction method, comprising:

[0007] Obtain geological disaster data, rainfall observation data and rainfall forecast data;

[0008] Based on the geological disaster data and rainfall observation data, an ID fitting model is constructed;

[0009] Constructing a geological disaster probability forecasting model based on the ID fitting model;

[0010] The rainfall forecast data is input into the geological disaster probability forecast model to predict the probability of geological disasters.

[0011] Optionally, obtaining rainfall observation data includes:

[0012] Based on the information of the meteorological station closest to the disaster site, obtain the previous rainfall observation data;

[0013] Obtain effective rainfall data based on geological disaster data and previous precipitation index model;

[0014] Based on the previous rainfall observation data and the effective rainfall data, rainfall observation data is acquired.

[0015] Optionally, the rainfall observation data includes: rainfall duration, rainfall cumulative amount, and average rainfall intensity.

[0016] Optionally, the previous precipitation index model is:

[0017]

[0018] In the formula, E r is the effective rainfall, n is the total number of days with precipitation before the geological disaster occurs, and the value of n can be determined by combining the local precipitation characteristics, r k is the daily rainfall, where r1, r2, …, r nThe daily precipitation amount one day, two days, ..., and n days before the geological disaster occurs

[0019] Optionally, constructing the ID fitting model includes:

[0020] The precipitation data are sorted according to the average rainfall intensity, a preset number of the rainfall observation data are eliminated, and the processed rainfall observation data are obtained;

[0021] Sorting the geological disaster data according to the effective rainfall, taking the preset percentile value of the effective rainfall as the effective rainfall demarcation value;

[0022] Based on the demarcation value, dividing the processed rainfall observation data;

[0023] Based on the divided new rainfall observation data, the ID fitting model is constructed respectively.

[0024] Optionally, the ID fitting model is:

[0025] I=αD -β ,

[0026] Wherein, D is the duration of this rainfall, I is the average rainfall intensity of this rainfall, and α and β are model fitting parameters. Optionally, the geological disaster probability forecast model is:

[0027] p i =S iD / C D

[0028] S iD =I iD -I minD

[0029]

[0030] Where: p i is the probability of occurrence of geological disasters at the i-th geological disaster point to be predicted; S iD is the average rainfall intensity I of the i-th geological disaster point to be predicted when the precipitation duration is D iD To the threshold model curve (lower bound) I minD Distance I iD is the average rainfall intensity predicted for the i-th geological disaster point when the rainfall duration is D; C D :Threshold model curve (upper bound) I when precipitation duration is D maxD To the threshold model curve (lower bound) I minD Distance I maxD is the upper limit of the threshold model curve; I minD is the lower bound of the threshold model curve.

[0031] Here, the effective precipitation value E is calculated based on the previous measured precipitation when the forecast is issued. r , by comparing with the demarcation threshold R C The size of the threshold model used determines the upper bound I max and lower bound I min , where I max =αD -β +c,I min =αD -β , α, β, and c are model fitting parameters.

[0032] Compared with the prior art, the present invention has the following advantages and technical effects:

[0033] Based on geological disaster data, the present invention utilizes more refined hourly precipitation data from automatic meteorological stations to evaluate the previous effective rainfall and the forecast precipitation for the day, thereby establishing a dual-parameter geological disaster prediction model suitable for the Wuzhou city area. Case tests show that the ID fitting model has a higher TS score than the traditional effective rainfall disaster probability model and can effectively reduce the false alarm rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0035] Figure 1 A schematic diagram of a flow chart of the probability of occurrence of geological disasters according to an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of rainfall division in an embodiment of the present invention;

[0037] Figure 3 It is a schematic diagram of the average precipitation curve of the regional geological disaster sample in the embodiment of the present invention;

[0038] Figure 4 Schematic diagram of an ID threshold curve according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0040] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0041] This application selects the precipitation threshold of a typical granite weathering rock and soil geological disaster area (Wuzhou) as the research focus, attempts to use hourly precipitation to roll over and calculate the local precipitation, replace the daily precipitation, and use correlation to calculate the rainy day value with the largest correlation with the previous effective precipitation, and reconstruct the meteorological potential (probability) forecast model of geological disasters of typical geological types.

[0042] The embodiment of the present invention proposes a data-driven typical granite weathering soil geological disaster potential prediction method, including:

[0043] Obtain geological disaster data, rainfall observation data and rainfall forecast data;

[0044] Based on geological disaster data and rainfall observation data, an ID fitting model is constructed;

[0045] Construct a geological disaster probability forecast model based on the ID fitting model;

[0046] Input rainfall forecast data into the geological disaster probability forecast model to predict the probability of geological disasters.

[0047] Furthermore, obtaining geological disaster data and precipitation observation data includes:

[0048] Based on the information of the meteorological station closest to the disaster site, obtain the previous rainfall observation data;

[0049] Obtain effective rainfall data based on geological disaster data and previous precipitation index model;

[0050] Obtain rainfall observation data based on previous rainfall observation data and effective rainfall data.

[0051] Furthermore, the rainfall observation data include: rainfall duration, rainfall accumulation, and average rainfall intensity.

[0052] Furthermore, constructing the ID fitting model includes:

[0053] The rainfall observation data are sorted according to the average rainfall intensity, a preset number of rainfall observation data are eliminated, and the processed rainfall observation data are obtained;

[0054] The geological disaster data are sorted according to the effective rainfall, and the preset percentile value of the effective rainfall is taken as the effective rainfall cutoff value.

[0055] Based on the cutoff value, the processed rainfall observation data are divided;

[0056] Based on the divided rainfall observation data, ID fitting models are constructed respectively.

[0057] Example

[0058] The information on geological disaster events and potential hazards in Wuzhou City comes from the Wuzhou Geological Monitoring Station. A total of 155 landslide geological disaster samples triggered by precipitation from 2016 to 2022 were collected in Wuzhou City (i.e., Wanxiu District, Longxu District, and Changzhou District, excluding the three counties).

[0059] The precipitation data are taken from the Wuzhou Station basic station (since its establishment) included in the quality control assessment of the Guangxi Meteorological Information Center and the 30 automatic meteorological observation stations in Wuzhou City from 2012 to 2022 Beijing time hourly (20:00 to 20:00 the next day, the precipitation in the hour before this hour). The average distance between adjacent observation stations of rainfall elements is 5.2km, the data time series is unified, the quality is stable, it is true and reliable, the density is large, and the representativeness is strong.

[0060] According to the latitude and longitude of the geological disaster sample and the time of occurrence, the nearest meteorological station is found using the spherical distance method (1), and the precipitation data of this meteorological station represents the precipitation conditions at the geological disaster occurrence point.

[0061] d=R*arcos[cos(Y1)*cos(Y2)*cos(X1-X2)+sin(Y1)*sin(Y2)] (1)

[0062] Where: R is the radius of the earth, X1, Y1 are the longitude and latitude of point 1, X2, Y2 are the longitude and latitude of point 2.

[0063] According to formula (1), the proportion of the distance between the disaster sample site and the surrounding representative automatic meteorological stations is less than 5 km, reaching 86%. The rainfall data can truly reflect the precipitation conditions at the disaster site.

[0064] According to geological disaster investigation data, the geological occurrence time can be accurate to the hour. If the occurrence time is not accurate to the hour, the time of the maximum hourly rainfall on that day will be used as the occurrence time of the geological disaster.

[0065] Wuzhou City is located between the eastern end of Dayao Mountain, the northern end of Yunkai Mountain and the southern end of Dagui Mountain. The overall terrain is high around and low in the middle. The annual precipitation in Wuzhou City in the past 30 years (1978-2020) is between 904.2 mm (2000) and 2006.4 mm (2013). March to September is the concentrated period of precipitation in Wuzhou, with an average cumulative precipitation of 1235.8 mm, accounting for 83.4% of the annual precipitation. It belongs to the heavy rain area in eastern Guidong, with a maximum daily rainfall of 300.4 mm and a maximum 1-hour rainfall of 135.2 mm. According to the business regulations of the local meteorological department on the precipitation process: the day when any station in the jurisdiction begins to have a daily rainfall of ≥25 mm and above is recorded as the starting day of the precipitation process, and the end day when no station in the jurisdiction has heavy rain or above. Statistics show that the precipitation process in Wuzhou basically lasts for 1-3 days (the precipitation process lasts no more than 72 hours), and the contribution rate of 6-hour continuous and 12-hour continuous precipitation at each station to the daily precipitation is 69.24% and 78.52% respectively, indicating that the weather influencing system in Wuzhou area is mainly a mesoscale weather system with a life history of 6-12 hours.

[0066] Affected by natural conditions such as complex topographic and geological conditions, developed geological structures, uneven temporal and spatial distribution of precipitation, and the destructive effects of human activities, there are many geological disasters and hidden dangers, the main types of which are collapse and landslide. Geological disasters in some local areas have caused certain damage to the lives and property of local people and vegetation landscapes.

[0067] According to the geological disaster investigation and evaluation results of the geological environment monitoring department, the entire city of Wuzhou is prone to geological disasters. There are 371 geological disaster risk points announced by the Wuzhou Natural Resources Bureau at the end of 2021. The geological disaster risk points are numerous and widespread, with a high risk of disasters and great difficulty in prevention. Among them, the high and medium geological disaster prone areas (dangerous areas) account for as high as 73.8%, and the disaster situation of geological disaster risk points ranges from small to extra-large, with small ones being the main ones, accounting for as high as 86.74%. According to the statistical data of geological disasters that have occurred in Wuzhou City, there are two main types of geological disasters in Wuzhou City: collapse and landslide. The strata exposed in the areas with the highest density of geological disaster development are mainly granite and clastic rocks. The surface is generally severely weathered, with a gravel content of about 10-30%, a loose structure, and the core is easy to disperse by hand. It is easy to soften and disintegrate when encountering water (precipitation). It can be seen that the weathered rock and soil layers of granite and clastic rocks are the most important strata prone to collapse and sliding in Wuzhou City.

[0068] The precipitation factors that induce geological disasters are mainly the local precipitation and the previous effective precipitation. Based on the precipitation characteristics of the study area, individual geological disasters, and hourly precipitation at automatic stations in high-density areas, such as Figure 1As shown, in this embodiment, the precipitation information of this field is calculated, the cumulative precipitation of the previous n days with a better fitting relationship is selected as the previous effective rainfall, the previous effective rainfall boundary threshold is defined, and the geological disaster meteorological probability model is established respectively. The short-term precipitation grid forecast data of the study area is used as the current rainfall to substitute into the precipitation geological disaster meteorological probability model, and the probability of occurrence of geological disasters at the grid points in the future is predicted.

[0069] This rainfall is the closest rainfall to the disaster. The important parameters are the duration of this rainfall and the average rainfall intensity of this rainfall. Based on the precipitation intensity-duration ID model (2) proposed by Caine (1980), the hourly precipitation is used to calculate the information of this rainfall.

[0070] I=αD -β (2)

[0071] In the formula, the duration of this rainfall D is calculated by determining the start and end time of this rainfall. The duration of 12-hour cumulative rainfall ≥ 10 mm is calculated by sliding forward hourly (within 72 hours) from the time when the disaster occurs. If it is assumed that there is a continuous rainfall process, that is, the cumulative rainfall for multiple consecutive 12 hours is ≥ 10 mm, then the start time of the first 12-hour cumulative rainfall ≥ 10 mm is the end time of this rainfall; the end time of the last 12-hour cumulative rainfall ≥ 10 mm is the start time of this rainfall ( Figure 2 ). That is: the duration of this rainfall D = end time - start time (unit: h)

[0072] The accumulated rainfall SR of this rainfall: the accumulated rainfall of this rainfall is obtained by adding up the hourly rainfall within the duration of this rainfall (unit: mm);

[0073] Average rainfall intensity of this rainfall I: the ratio of the accumulated rainfall to the duration of this rainfall, that is, I = SR / D (unit: mm / h);

[0074] The previous effective precipitation information is determined based on the previous precipitation index (KPI) model formula:

[0075]

[0076] In the formula, E r --Effective rainfall, in millimeters,

[0077] n--The total number of days with precipitation before the geological disaster occurs. The value of n can be determined by combining the local precipitation characteristics.

[0078] r k --Daily rainfall, in millimeters, r1, r2, ... r n They are the day before, two days before, ... n days before the geological disaster occurs.

[0079] Most domestic scholars do not distinguish between precipitation areas when taking the value of the number of days with previous rainfall. They uniformly use the daily precipitation 15 days (n=15) before the disaster to calculate E. r Considering that the study area belongs to the rainy area in South China, the precipitation during the period of frequent geological disasters from April to September has high-frequency and high-value spatiotemporal characteristics. It is necessary to focus on its precipitation characteristics and redefine the value range of effective precipitation in the early stage.

[0080] The daily accumulated rainfall values ​​in the 15 days before the geological disaster samples in this region were normalized using the max-min (Formula 3).

[0081] r * =(rr min ) / (r max -r min ) (4)

[0082] In the formula, r is the original data, r * is the standardized data, r max and r min They are respectively the maximum and minimum values ​​in the daily precipitation sequence 15 days before a geological disaster sample.

[0083] Get the average state curve of the daily precipitation sequence of all geological disaster samples before 15 days ( Figure 3 ), the figure shows that about 10 days before the disaster, the daily precipitation average curve reaches a minimum deviation value, and then maintains a low value. Considering the cumulative impact of the previous precipitation in this study area, n = 11 in formula (3)

[0084] Based on the ID model of effective rainfall threshold, in order to reduce the influence of outliers, the rainfall sample data of this field is sorted by average rainfall intensity, and the first and last 5% samples of the sequence are removed. r Sort by the 50th percentile value (i.e. 64.6 mm) as the effective rainfall value cutoff value R C , press E r ≥R C and E r <R C The rainfall sample data of this field is divided into two groups, and the relationship curve of ID is constructed for each group of samples ( Figure 4 ), where there will be multiple average rainfall intensities for the same rainfall duration. The average rainfall intensity of the geological disaster samples with the same rainfall duration is calculated as the average rainfall intensity corresponding to the duration. Figure 4 The middle and upper profiles are the highest and lowest average rainfall intensities that may cause geological disasters in any precipitation period in the region. While ensuring that the upper and lower threshold curves include as many geological samples as possible, the following ID threshold curve equations are obtained for each group of samples: Figure 4 ):

[0085] When Er ≥64.6mm:

[0086] Threshold curve 1 (upper limit): I max =13.391D -0.513 +1.196 (5)

[0087] Threshold curve 2 (lower bound) I min =13.391D -0.513 (6)

[0088] When E r <64.6mm:

[0089] Threshold curve (upper limit): I max =16.315D -0.773 +1.781 (7)

[0090] Threshold curve (lower bound): I min =16.315D -0.773 (8)

[0091] Furthermore, it is characterized in that the geological disaster probability forecasting model is:

[0092] p i =S iD / C D (9)

[0093] S iD =I iD -I minD (10)

[0094]

[0095] Where:

[0096] p i : The probability of occurrence of geological disasters at the i-th geological disaster point to be predicted;

[0097] S iD : The average rainfall intensity I of the i-th geological disaster point to be predicted when the precipitation duration is D iD To the threshold model curve (lower bound) I minD Distance I iD is the average rainfall intensity predicted for the i-th geological disaster point when the rainfall duration is D

[0098] C D :Threshold model curve (upper bound) I when precipitation duration is D maxD To the threshold model curve (lower bound) I minD Here, the effective precipitation value E is calculated based on the previous measured precipitation when the forecast is issued. r, by comparing with the demarcation threshold R C The size of the threshold model used to determine the calculation of I maxD and I minD ;

[0099] According to formulas (5)-(11), combined with the current geological disaster warning business of the geological environment monitoring department, the early warning (QPF) of the grid precipitation forecast for the next 72 hours issued by the meteorological department at 20:00 every day is substituted into formula (9), and the objective probability forecast for the next 72 hours can be obtained. i The forecast value is divided into five levels, which respectively represent the extremely low possibility, low possibility, relatively high possibility, high possibility and extremely high possibility warning levels of geological disasters in the study area.

[0100] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A data-driven method for predicting the potential geological hazards of typical granite weathering soils, characterized in that: include: Obtain geological disaster data, rainfall observation data and rainfall forecast data; Acquiring the rainfall observation data includes: Based on the information of the meteorological station closest to the disaster site, obtain the previous rainfall observation data; According to the latitude and longitude of the geological disaster occurrence point and the time of occurrence, the nearest weather station is found using the spherical distance method, and the precipitation data of the weather station represents the precipitation conditions at the geological disaster occurrence point: Where: R is the radius of the earth, X1, Y1 are the longitude and latitude of point 1, X2, Y2 are the longitude and latitude of point 2; Based on the geological disaster data and the previous precipitation index model, obtaining effective rainfall data; Based on the previous rainfall observation data and the effective rainfall data, obtaining rainfall observation data; Based on the geological disaster data and the rainfall observation data, constructing an ID fitting model; Constructing the ID fitting model includes: The precipitation data are sorted according to the average rainfall intensity, a preset number of the rainfall observation data are eliminated, and the processed rainfall observation data are obtained; Sorting the geological disaster data according to the effective rainfall, taking the preset percentile value of the effective rainfall as the effective rainfall demarcation value; Based on the demarcation value, dividing the processed rainfall observation data; Based on the divided rainfall observation data, construct the ID fitting models respectively; Based on the ID fitting model, a geological disaster probability forecasting model is constructed; Inputting the rainfall forecast data into the geological disaster probability forecast model to predict the probability of geological disasters; The rainfall observation data include: rainfall duration, rainfall accumulation, and average rainfall intensity; The previous precipitation index model is: Among them, E r is the effective rainfall, n is the total number of days with precipitation before the geological disaster occurs, r k is the daily rainfall, k is a variable; The ID fitting model is: I=αD -β , Among them, D is the duration of this rainfall, I is the average rainfall intensity of this rainfall, and α and β are model fitting parameters; The geological disaster probability forecast model is: p i =S iD / C D S iD =I iD -I minD C D =I maxD -I minD Where: p i is the probability of occurrence of geological disasters at the i-th geological disaster point to be predicted; S iD is the average rainfall intensity I of the i-th geological disaster point to be predicted when the precipitation duration is D iD To the threshold model curve I minD Distance I iD is the average rainfall intensity predicted for the i-th geological disaster point when the rainfall duration is D; C D :Threshold model curve I when precipitation duration is D maxD To the threshold model curve I minD Distance I maxD is the upper limit of the threshold model curve; I minD is the lower limit of the threshold model curve; the upper and lower limits of the model curve are the highest and lowest average rainfall intensities that may lead to geological disasters in any precipitation duration in a certain area.

Images