A spatiotemporal prediction method and system for railway geological disasters

Abstract

The present application relates to a kind of railway geological disaster sensitivity space-time prediction method and system.The method includes obtaining the relevant data of China railway line and the geological disaster historical event of China railway line;China railway line geological disaster influencing factor is divided into dynamic influencing factor and static influencing factor;The geological disaster historical event of China railway line is converted from spatialization into point data, and 2 times sample point of not occurring geological disaster is randomly generated, then according to point data and sample point, event sample point is generated, and corresponding dynamic influencing factor is extracted to event sample point, and synthesis feature matrix;Random forest model is trained and tested using feature matrix, and the geological disaster sensitivity evaluation model is determined;Using geological disaster sensitivity evaluation model, real-time data and precipitation prediction data obtained according to GEE platform are used to predict the sensitivity of China railway geological disaster.The present application can improve the accuracy of railway geological disaster sensitivity space-time prediction.

Description

Technical Field

[0001] This invention relates to the field of disaster sensitivity prediction, and in particular to a spatiotemporal prediction method and system for railway geological disaster sensitivity. Background Technology

[0002] Railways are a major form of transportation in China and a vital part of the country's infrastructure, holding a backbone position in China's comprehensive transportation system. Due to the inherent characteristics of railways, damage in one area can disrupt the entire line, and the indirect economic losses from railway interruptions far outweigh the value of repairing the interrupted lines. Extreme rainfall is often the primary trigger for geological disasters. Currently, the China Meteorological Administration and the US GFS forecasting products can provide quantitative rainfall forecast data. Combining this data with static or dynamic influencing factors such as topography and environment, daily qualitative forecasts can be presented to indicate the sensitivity of railway lines to geological disasters caused by rainfall. For example, sensitivity levels can be classified as low, medium, and high. Furthermore, using geographic information systems to create thematic maps of disaster sensitivity can provide scientific guidance for disaster mitigation and emergency management, thereby minimizing casualties and property damage caused by geological disasters along railway lines based on published forecast information.

[0003] Data-driven approaches primarily rely on statistical knowledge and methods such as data mining to analyze the relationship between landslide occurrence and influencing factors. Among data-driven approaches, machine learning methods generally outperform other traditional methods. Data mining and machine learning algorithms are increasingly used in disaster sensitivity assessment because they perform well enough to handle linear or nonlinear relationships between disaster influencing factors at different scales and from different sources and the occurrence of disasters. In disaster sensitivity research, the random forest model has demonstrated outstanding performance in disaster sensitivity assessment due to its high performance and strong stability. It introduces sample and feature randomization, making it less prone to overfitting, possessing noise resistance, capable of handling high-dimensional data, requiring no feature selection, and offering fast training speeds.

[0004] Current research on mapping the spatial distribution of geological hazard sensitivity based on static variables is abundant, but the time scale is often ambiguous, and most studies do not consider the characteristics of influencing factors changing over time. Research on constructing geological hazard sensitivity prediction models considering dynamic variables is severely lacking. In reality, rainfall is a major trigger for geological hazards, and the occurrence of such hazards is largely related to the amount of rainfall in the preceding period. Therefore, considering the dynamic changes of influencing factors can improve the accuracy of modeling and prediction to some extent. Furthermore, the results of geological hazard sensitivity analysis are constrained by the number and spatial patterns of historical events; spatial extrapolation inevitably reduces the confidence level of the prediction results. Collecting sufficient data for modeling and prediction in specific applicable areas is more scientifically sound.

[0005] When considering the dynamic changes of influencing factors, the amount of time-series data is enormous, making the collection of massive amounts of data to extract dynamic influencing factors a significant challenge. Therefore, how to identify potential railway geological hazards in advance on a large regional scale in China and enhance the resilience of railway infrastructure remains an urgent technical problem to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for predicting the spatiotemporal sensitivity of railway geological hazards, which can improve the accuracy of predicting the spatiotemporal sensitivity of railway geological hazards, thereby enabling early identification of potential railway geological hazards and enhancing the resilience of railway infrastructure.

[0007] To achieve the above objectives, the present invention provides the following solution:

[0008] A spatiotemporal prediction method for railway geological hazard sensitivity includes:

[0009] This involves acquiring relevant data on Chinese railway lines and historical geological disaster events along those lines. These historical events include the time and location of the events. The relevant data includes basic information about Chinese railway lines and factors influencing geological disasters along those lines. These factors include topographical factors, soil factors, geological factors, environmental factors, and rainfall data corresponding to different locations.

[0010] The factors influencing geological hazards along China's railway lines are classified into dynamic and static influencing factors.

[0011] The historical geological disaster events along China's railway lines were spatialized and converted into point data. Two times the number of sample points that had not experienced geological disasters were randomly generated. Event sample points were then generated based on the point data and sample points. The corresponding dynamic influencing factors were extracted into the event sample points and a feature matrix was synthesized.

[0012] A random forest model was trained and tested using feature matrices to determine a geological hazard sensitivity assessment model;

[0013] A geological hazard sensitivity assessment model was used to predict the geological hazard sensitivity of China's railways based on real-time data obtained from the GEE platform and precipitation forecast data.

[0014] Optionally, the process of spatializing historical geological disaster events along Chinese railway lines into point data, randomly generating twice the number of sample points that have not experienced geological disasters, generating event sample points based on the point data and sample points, and extracting corresponding dynamic influencing factors into the event sample points to synthesize a feature matrix, specifically includes:

[0015] The point value extraction tool from the Python geemap library extracts dynamic influencing factors onto event sample points based on historical event time.

[0016] Optionally, the step of using the feature matrix to train and test the random forest model to determine the geological hazard sensitivity assessment model specifically includes:

[0017] The area under the receiver operating characteristic curve (ROC) was used as the evaluation index.

[0018] Optionally, the adoption of a geological hazard sensitivity assessment model to predict the geological hazard sensitivity of China's railways based on real-time data obtained from the GEE platform and precipitation forecast data specifically includes:

[0019] Use GPM data to determine real-time data;

[0020] Precipitation forecast data is obtained from The Global Forecast System weather forecast products;

[0021] The feature matrix to be predicted is determined based on real-time data and precipitation forecast data.

[0022] Optionally, the step of employing a geological hazard sensitivity assessment model to predict the geological hazard sensitivity of China's railways based on real-time data obtained from the GEE platform and precipitation forecast data further includes:

[0023] A geological hazard sensitivity prediction map is generated based on the prediction results.

[0024] A spatiotemporal prediction system for railway geological hazard sensitivity includes:

[0025] The historical data acquisition module is used to acquire relevant data on Chinese railway lines and historical geological disaster events along Chinese railway lines. The historical geological disaster events along Chinese railway lines include: the time and spatial location information of the events. The relevant data on Chinese railway lines includes: basic information on Chinese railway lines and influencing factors of geological disasters along Chinese railway lines. The influencing factors of geological disasters along Chinese railway lines include: topographic factors, soil factors, geological factors, environmental factors, and rainfall data corresponding to different locations.

[0026] The influencing factor classification module is used to classify the influencing factors of geological disasters along Chinese railways into dynamic influencing factors and static influencing factors.

[0027] The feature matrix synthesis module is used to convert historical geological disaster events along Chinese railways from spatial data into point data, and randomly generate twice the number of sample points that have not experienced geological disasters. Then, it generates event sample points based on the point data and sample points, extracts the corresponding dynamic influencing factors into the event sample points, and synthesizes the feature matrix.

[0028] The geological hazard sensitivity assessment model determination module is used to train and test a random forest model using a feature matrix to determine the geological hazard sensitivity assessment model.

[0029] The disaster sensitivity prediction module is used to predict the geological disaster sensitivity of China's railways based on real-time data obtained from the GEE platform and precipitation prediction data, using a geological disaster sensitivity assessment model.

[0030] Optionally, the feature matrix synthesis module specifically includes:

[0031] The dynamic influencing factor extraction unit is used to extract dynamic influencing factors onto event sample points based on historical event time using the point value extraction tool of the Python geemap library.

[0032] Optionally, the disaster sensitivity prediction module specifically includes:

[0033] Real-time data acquisition unit, used to determine real-time data using GPM data;

[0034] The precipitation forecast data acquisition unit is used to acquire precipitation forecast data based on the meteorological forecast products of The Global Forecast System.

[0035] The feature matrix determination unit is used to determine the feature matrix to be predicted based on real-time data and precipitation prediction data.

[0036] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0037] This invention provides a method and system for spatiotemporal prediction of railway geological hazard sensitivity. It collects historical events and dynamic and static influencing factors data of geological hazards along China's railways, and randomly generates time-sensitive locations of geological hazards that have not yet occurred. Using the Random Forest algorithm, which performs exceptionally well in sensitivity mapping, a model suitable for spatiotemporal dynamic prediction of railway geological hazards in China is constructed based on GEE. Combining dynamic influencing factor modeling not only improves model performance but also allows for relatively accurate predictions of railway geological hazard sensitivity several days in advance within GEE, incorporating near-real-time rainfall data and quantitative rainfall forecasting products. The spatiotemporal distribution of railway geological hazard sensitivity is dynamically published and updated in real-time according to the update date of rainfall forecasting products. The prediction results can provide scientific and technological support for the management, emergency response, and disaster reduction of railway geological hazards in China. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic diagram of a spatiotemporal prediction method for railway geological disaster sensitivity provided by the present invention;

[0040] Figure 2 A detailed technical flowchart of a railway geological disaster sensitivity spatiotemporal prediction method provided by the present invention. Detailed Implementation

[0041] 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.

[0042] The purpose of this invention is to provide a method and system for predicting the spatiotemporal sensitivity of railway geological hazards, which can improve the accuracy of predicting the spatiotemporal sensitivity of railway geological hazards, thereby enabling early identification of potential railway geological hazards and enhancing the resilience of railway infrastructure.

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] Figure 1 This is a schematic diagram of the spatiotemporal prediction method for railway geological disaster sensitivity provided by the present invention. Figure 2 A detailed technical flowchart of a railway geological hazard sensitivity spatiotemporal prediction method provided by the present invention is shown below. Figure 1 and Figure 2 As shown, the present invention provides a method for spatiotemporal prediction of railway geological hazard sensitivity, comprising:

[0045] S101, Obtain relevant data on Chinese railway lines and historical events of geological disasters along Chinese railway lines; the historical events of geological disasters along Chinese railway lines include: the time and spatial location information of the events; the relevant data on Chinese railway lines include: basic information on Chinese railway lines and influencing factors of geological disasters along Chinese railway lines; the influencing factors of geological disasters along Chinese railway lines include: topographic factors, soil factors, geological factors, environmental factors and rainfall data corresponding to different locations.

[0046] Data related to Chinese railway lines comes from OpenStreetMap; historical data on railway interruptions caused by geological disasters are collected from social media and news reports. Each piece of valid information includes the clear start and end locations of the roadbed damage, the length of the damage, and the time of the accident. Because spatial extrapolation can bring great uncertainty to model evaluation, the data collection process not only requires the data to be as detailed as possible, but also strictly limits each record to roadbed damage caused by geological disasters, and the location is spatialized based on Chinese railway lines.

[0047] As a specific example, 326 data points of railway roadbed damage and interruption caused by geological disasters in China from 2000 to 2017 were collected from social media. Each data point has clear information such as the time of the accident, the start and end locations, and the length of the damage. The data is very targeted for the analysis of the sensitivity of China's railways to geological disasters. Subsequently, China's railway data was downloaded from OpenStreetMap, and the collected information was spatially located using a geographic information system.

[0048] Various geological hazard influencing factors were collected. The topographic DEM was resampled to a 250m spatial resolution in ArcGIS Pro; slope, curvature, and topographic humidity index were calculated in ArcGIS Pro; and the topographic intensity index was calculated using QGIS. All topographic factor indices were calculated in a projected coordinate system.

[0049] For the topographic moisture index, the specific operational technical details are as follows: 1. Convert the DEM data to a projected coordinate system; 2. Use spatial analysis tools to calculate the flow direction; 3. Use spatial analysis tools to calculate the flow rate (FA); 4. Use spatial analysis tools to calculate the slope; 5. Use a raster calculator to convert the slope scale, the formula is ("slope"). 1.570796) / 90 and Con("slope">0, Tan("slope"), 0.001) to obtain tanslope; 6. Copy the raster data spatial resolution res, and use the raster calculator to calculate the flow FAscaled after the scale change, the formula is ("FA"+1) res; 7. Calculate the terrain humidity index using the formula Ln("FAscaled" / "tanslope").

[0050] Among the environmental factors, land cover data and surface cover percentage were loaded from GEE (Google Earth Engine). Surface cover percentage (FVC) was calculated using NDVI, with the following formula:

[0051] ;

[0052] The distance to rivers was calculated using ArcGIS Pro's Euclidean distance tool, which measures the shortest distance from the center point of a 250m raster in the Chinese region to rivers of various classifications.

[0053] Soil factors were selected from soil type maps, with OpenLandMap as the data source, and loaded using GEE.

[0054] In the geological factors, lithology data is stored in vector format and converted to 250m raster data based on the lithology field, then used as one of the explanatory variables. The distance to faults was calculated using ArcGIS Pro's Euclidean distance tool, determining the shortest distance from the center point of the 250m raster in China to the major faults in China.

[0055] In the rainfall factor, the historical rainfall data used were all loaded from GEE, and the daily rainfall and cumulative rainfall were calculated based on the roadbed damage date of each point.

[0056] S102 classifies the factors influencing geological hazards along Chinese railway lines into dynamic and static influencing factors.

[0057] Table 1 shows the influencing factors of geological hazard sensitivity and their corresponding spatial resolution. Among the precipitation factors, the cumulative rainfall includes the amount on the day the geological hazard occurs. The dynamic influencing factors include land cover type, surface cover, and precipitation; the rest are static influencing factors.

[0058] Table 1

[0059]

[0060] S103 converts historical geological disaster events along Chinese railway lines from spatial data into point data, randomly generates twice the number of sample points that have not experienced geological disasters, then generates event sample points based on the point data and sample points, extracts the corresponding dynamic influencing factors into the event sample points, and synthesizes a feature matrix.

[0061] The collected data was organized into raster data as shown in Table 1 according to requirements. Damaged roadbed sections were converted into points at 250m intervals, and twice the number of points without disasters were randomly generated. The corresponding dynamic and static influencing factors were extracted and compiled into a feature matrix. Then, 80% of the feature matrix was randomly selected as the training set, and 20% as the test set. Considering the extraction of dynamic influencing factors, the generated sample points without geological disasters needed to be randomly assigned a date attribute within the time range of the collected historical landslide locations. Furthermore, it was ensured that non-landslide points and landslide points did not overlap spatially and temporally simultaneously to avoid introducing human error due to sample point selection.

[0062] S103 specifically includes:

[0063] The point value extraction tool from the Python geemap library extracts dynamic influencing factors onto event sample points based on historical event time.

[0064] S104. A random forest model was trained and tested using the feature matrix to determine the geological hazard sensitivity assessment model; the area under the receiver operating characteristic curve (AUC) was used as the evaluation index.

[0065] The AUC is used to evaluate the modeling performance of dynamic prediction models. AUC is widely used in the performance evaluation of disaster sensitivity mapping. The model is evaluated according to the following criteria: excellent (0.9-1), very good (0.8-0.9), good (0.7-0.8), average (0.6-0.7) and poor (0.5-0.6).

[0066] Historical geological disaster data was converted from polyline data to point data using ArcGIS Pro at 250m intervals. Then, twice the number of points that did not experience geological disasters were randomly generated, as follows: Railway line data was converted from polyline data to point data using ArcGIS Pro at 250m intervals. A specific date was randomly generated between 2000 and 2017. If the randomly generated date was not recorded in the historical event set, a point was randomly selected from the railway line point data; otherwise, a point not within 0.1° of the corresponding disaster point was randomly generated to avoid spatiotemporal overlap. This process was repeated until the required number of points was reached. Further, the numerical values ​​of geological disaster influencing factors corresponding to both disaster-affected and disaster-free points needed to be extracted. For static influencing factors, these were downloaded locally and extracted to points using ArcGIS Pro's multi-value extraction tool. For dynamic influencing factors, the GEEMap value extraction tool was used to extract corresponding values ​​for each point on the GEE platform based on time and space, solving the difficulty of downloading time-series data. Historical rainfall data was obtained using the GPM product, a near-real-time rainfall product with a time resolution of 0.5 hours and a time span from 2000 to the present. When calculating cumulative rainfall characteristics, the following rules were followed: all precipitation data from the start time to the end time were selected, multiplied by a scaling factor of 0.5, and then summed to obtain the cumulative precipitation. After all influencing factor data were extracted to the point level, they were integrated to form a feature matrix, and 80% of the records were randomly assigned to the training set and 20% to the test set.

[0067] The model was built using the random forest algorithm with 100 base learners. The random forest algorithm was implemented using the pythonscikit-learn library. The model was built based on the training set partitioned in the previous steps, and then tested on the test set. The final AUC was 0.96, indicating that the dynamic prediction model has excellent performance.

[0068] S105 uses a geological hazard sensitivity assessment model to predict the geological hazard sensitivity of China's railways based on real-time data obtained from the GEE platform and precipitation forecast data.

[0069] S105 specifically includes:

[0070] Use GPM data to determine real-time data.

[0071] Precipitation forecast data is obtained from The Global Forecast System (GFS) weather forecast products, with a maximum forecast period of 384 hours. Land cover for the forecast month and corresponding land cover type for the year can be obtained from MODIS remote sensing products. All dynamic imagery data is loaded from GEE. Near-real-time and forecast data are used separately because GPM data has a 3-day data delay on the GEE platform; therefore, both forecast and data within the delayed period use GFS data.

[0072] The feature matrix to be predicted is determined based on real-time data and precipitation forecast data.

[0073] First, the acquired railway vector data is converted into point data at 250m intervals. Then, all explanatory variables corresponding to the dates to be predicted are extracted onto the road network points, and the constructed dynamic prediction model is used to predict the geological hazard sensitivity of each road network point.

[0074] As a specific example, on June 28, 2022, the geological hazard sensitivity of China's railways from June 29, 2022 to July 2, 2022 was predicted. After converting China's railway line data into point data using ArcGIS Pro, all near-real-time or predicted data were extracted from GEE using the geemap point value extraction function into the China railway line point data. Specifically, land cover data was loaded with the most recent 2020 product, NDVI data used the June 2022 median composite product, and precipitation products were loaded with daily cumulative precipitation from GPM (June 23-26, 2022) and daily cumulative precipitation from GFS (June 27-22, 2022), and statistics were compiled according to the corresponding prediction dates. The predicted feature matrix was constructed based on the daily rainfall, three-day cumulative rainfall, and seven-day cumulative rainfall. Specifically, two different data sources, near-real-time and predicted data, were used for rainfall data because GPM data has a three-day data delay on the GEE platform. Therefore, both the predicted data and data within the delayed period used GFS data. The GPM cumulative precipitation calculation method is the same as described in step two. GFS precipitation data provides a 6-hour cumulative precipitation prediction within the predicted time period, with a 6-hour step size. When calculating cumulative precipitation, only the GFS river water files within the corresponding dates need to be summed. The predicted precipitation data from June 26, 2022, for June 27 and June 28, and the predicted precipitation data from June 28 for June 29 to July 2, 2022, were used from the GFS dataset.

[0075] The sensitivity of China's railways to geological hazards in the next few days is predicted, and the sensitivity prediction results are generated according to the prescribed classification rules: very dangerous >99%, extremely high 90%-99%, high 66%-90%, medium 33%-66%, low 10%-33%, extremely low 1%-10%, and basically no occurrence <1%.

[0076] Following S105 are:

[0077] A geological hazard sensitivity prediction map is generated based on the prediction results. Finally, hierarchical color rendering is used to complete the GIS geological hazard sensitivity prediction and export the results.

[0078] The prediction results were categorized in ArcGIS Pro according to the following rules: Very dangerous >99%, Extremely high 90%-99%, High 66%-90%, Medium 33%-66%, Low 10%-33%, Extremely low 1%-10%, and Virtually no occurrence <1%. Finally, graded color rendering was used to complete the GIS geological hazard sensitivity prediction, and the final results were presented in the form of a thematic map of China's railway infrastructure.

[0079] As another specific embodiment, the present invention also provides a spatiotemporal prediction system for railway geological disaster sensitivity, comprising:

[0080] The historical data acquisition module is used to acquire relevant data on Chinese railway lines and historical geological disaster events along Chinese railway lines. The historical geological disaster events along Chinese railway lines include: the time and spatial location information of the events. The relevant data on Chinese railway lines includes: basic information on Chinese railway lines and influencing factors of geological disasters along Chinese railway lines. The influencing factors of geological disasters along Chinese railway lines include: topographic factors, soil factors, geological factors, environmental factors, and rainfall data corresponding to different locations.

[0081] The influencing factor classification module is used to classify the influencing factors of geological disasters along Chinese railways into dynamic influencing factors and static influencing factors.

[0082] The feature matrix synthesis module is used to convert historical geological disaster events along Chinese railway lines from spatial data into point data, and randomly generate twice the number of sample points that have not experienced geological disasters. Then, it generates event sample points based on the point data and sample points, extracts the corresponding dynamic influencing factors into the event sample points, and synthesizes a feature matrix.

[0083] The geological hazard sensitivity assessment model determination module is used to train and test a random forest model using a feature matrix to determine the geological hazard sensitivity assessment model.

[0084] The disaster sensitivity prediction module is used to predict the geological disaster sensitivity of China's railways based on real-time data obtained from the GEE platform and precipitation prediction data, using a geological disaster sensitivity assessment model.

[0085] The feature matrix synthesis module specifically includes:

[0086] The dynamic influencing factor extraction unit is used to extract dynamic influencing factors onto event sample points based on historical event time using the point value extraction tool of the Python geemap library.

[0087] The disaster sensitivity prediction module specifically includes:

[0088] The real-time data acquisition unit is used to determine real-time data using GPM data.

[0089] The precipitation forecast data acquisition unit is used to acquire precipitation forecast data based on the Global Forecast System meteorological forecast products.

[0090] The feature matrix determination unit is used to determine the feature matrix to be predicted based on real-time data and precipitation prediction data.

[0091] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0092] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A railway geological disaster sensitivity spatio-temporal prediction method, characterized in that, The application relates to a method for predicting the sensitivity of a geological disaster of a Chinese railway line. The method comprises the following steps: acquiring Chinese railway line related data and Chinese railway line geological disaster historical events; the Chinese railway line geological disaster historical events comprise event occurrence time and spatial position information; the Chinese railway line related data comprise Chinese railway line information and Chinese railway line geological disaster influencing factors; the Chinese railway line geological disaster influencing factors comprise topographic factors, soil factors, geological factors, environmental factors and rainfall data corresponding to different positions; dividing the Chinese railway line geological disaster influencing factors into dynamic influencing factors and static influencing factors; converting the Chinese railway line geological disaster historical events into point data after spatialization, randomly generating 2 times of sample points without geological disasters, generating event sample points according to the point data and the sample points, and extracting corresponding dynamic influencing factors to the event sample points to synthesize a feature matrix; training and testing a random forest model by using the feature matrix to determine a geological disaster sensitivity evaluation model; 2. The spatio-temporal prediction method for railway geological disaster sensitivity according to claim 1, characterized in that, adopting the geological disaster sensitivity evaluation model to predict the sensitivity of a Chinese railway line geological disaster according to real-time data and rainfall prediction data acquired from a GEE platform. The method for converting the Chinese railway line geological disaster historical events into point data after spatialization, randomly generating 2 times of sample points without geological disasters, generating event sample points according to the point data and the sample points, and extracting corresponding dynamic influencing factors to the event sample points to synthesize a feature matrix specifically comprises the following steps:

3. The method according to claim 1, wherein, extracting the dynamic influencing factors to the event sample points according to the historical event time by using a point value extraction tool of a python geemap library. The method for training and testing the random forest model by using the feature matrix to determine the geological disaster sensitivity evaluation model specifically comprises the following steps:

4. The spatio-temporal prediction method for railway geological disaster sensitivity according to claim 1, characterized in that, adopting the area under the receiver operating characteristic curve as an evaluation index. The method for predicting the sensitivity of a Chinese railway line geological disaster by adopting the geological disaster sensitivity evaluation model according to the real-time data and the rainfall prediction data acquired from the GEE platform specifically comprises the following steps: determining the real-time data by using GPM data; acquiring the rainfall prediction data according to a The Global Forecast System meteorological prediction product; 5. The spatio-temporal prediction method for railway geological disaster sensitivity according to claim 1, characterized in that, determining a feature matrix to be predicted according to the real-time data and the rainfall prediction data. The method for predicting the sensitivity of a Chinese railway line geological disaster by adopting the geological disaster sensitivity evaluation model according to the real-time data and the rainfall prediction data acquired from the GEE platform further comprises the following steps after the above steps:

6. A railway geohazard susceptibility spatio-temporal prediction system characterized in that, generating a geological disaster sensitivity prediction map according to a prediction result. The application relates to a method for predicting the sensitivity of a geological disaster of a Chinese railway line. The method comprises the following steps: acquiring Chinese railway line related data and Chinese railway line geological disaster historical events; the Chinese railway line geological disaster historical events comprise event occurrence time and spatial position information; the Chinese railway line related data comprise Chinese railway line information and Chinese railway line geological disaster influencing factors; the Chinese railway line geological disaster influencing factors comprise topographic factors, soil factors, geological factors, environmental factors and rainfall data corresponding to different positions; dividing the Chinese railway line geological disaster influencing factors into dynamic influencing factors and static influencing factors; The feature matrix synthesis module is configured to convert the spatialized historical events of the geological disasters of the Chinese railway lines into point data, randomly generate sample points twice the number of the events of the geological disasters, generate event sample points according to the point data and the sample points, and extract corresponding dynamic influence factors to the event sample points to synthesize the feature matrix. The geological disaster sensitivity evaluation model determination module is configured to train and test a random forest model by using the feature matrix to determine the geological disaster sensitivity evaluation model. The disaster sensitivity prediction module is configured to use the geological disaster sensitivity evaluation model to predict the sensitivity of the geological disasters of the Chinese railways according to real-time data obtained from the GEE platform and precipitation prediction data.

7. The system for spatio-temporal prediction of railway geological hazard susceptibility according to claim 6, wherein, The feature matrix synthesis module specifically includes: The dynamic influence factor extraction unit is configured to extract the dynamic influence factors to the event sample points according to the time of the historical events by using a point value extraction tool of a python geemap library.

8. The system for spatio-temporal prediction of railway geological hazard susceptibility according to claim 6, wherein, The disaster sensitivity prediction module specifically includes: The real-time data acquisition unit is configured to determine the real-time data by using GPM data. The precipitation prediction data acquisition unit is configured to obtain the precipitation prediction data according to The Global Forecast System meteorological prediction products. The feature matrix to be predicted determination unit is configured to determine the feature matrix to be predicted according to the real-time data and the precipitation prediction data.

Images