Method and system for catastrophic flood event identification and early warning

Abstract

This invention discloses a method and system for identifying and providing early warning of catastrophic flood events. It utilizes seismic wave signals acquired by a dense regional broadband seismic network to distinguish the seismic signal characteristics at different stages of catastrophic flood evolution. The initial stage of a landslide is reconstructed through single-source waveform inversion. Real-time tracking and positioning of the flood's migration trajectory are achieved by combining normalized weighted cross-correlation and amplitude-distance analysis. Simultaneously, signal effectiveness is judged and risk assessment is conducted by comprehensively considering environmental noise levels and station distribution characteristics. Ultimately, this completes the graded early warning, effect evaluation, and model optimization for catastrophic flood events.

Description

Technical Field

[0001] This invention belongs to the field of geological disaster monitoring technology, specifically relating to a method and system for identifying and providing early warning of catastrophic flood events. It is applicable to the rapid identification, real-time location and early warning of catastrophic floods (including debris flows, flash floods, etc.) caused by rock mass landslides, glacial landslides, glacial lake outbursts, and landslide dammed lake outbursts. Background Technology

[0002] Large-scale rockfalls, glacial landslides, and the resulting catastrophic floods (mudslides, flash floods) are among the most dangerous geological hazards in mountainous areas, seriously threatening human life and property. Influenced by global climate change, these disasters are becoming increasingly frequent and severe in the areas surrounding the Qinghai-Tibet Plateau and in mountainous regions worldwide, highlighting the critical need for corresponding early warning and disaster mitigation measures.

[0003] However, such disasters have the following significant characteristics that make it difficult for traditional methods to provide effective early warning: ① Rapid speed: The landslide body disintegrates, impacts, and is transported by floods extremely quickly, often taking only a few minutes to tens of minutes from the occurrence of the event to the formation of the disaster, leaving a very short window for early warning and emergency response. ② Large scale: The disaster is massive in size and extremely destructive, capable of destroying villages and infrastructure along its path. ③ High degree of concealment and suddenness: The disaster often has no obvious precursors, and traditional monitoring methods (such as rain gauges, water level gauges, video surveillance, etc.) are unable to detect anomalies in advance, resulting in insufficient early warning capabilities.

[0004] Existing flood disaster monitoring and early warning technologies mainly include the following categories: (1) Traditional hydrological monitoring methods: Rainfall and river water level changes are monitored by deploying rain gauges, water level gauges, flow meters and other equipment. This type of method has the following drawbacks: ① It can only monitor local point information and is difficult to cover the entire watershed; ② The response is delayed and disasters often occur when the water level rises; ③ It cannot monitor the dynamics of triggering sources such as upstream landslides.

[0005] (2) Remote sensing image interpretation method: Images before and after a disaster are acquired through satellite remote sensing, UAV aerial photography, etc., and then compared and analyzed. This type of method has the following drawbacks: ① Low temporal resolution, making it difficult to achieve real-time monitoring; ② Greatly affected by weather conditions (clouds, rain); ③ Can only be used for post-disaster assessment and cannot achieve early warning.

[0006] (3) Video surveillance method: Cameras are deployed at key locations to monitor the river conditions in real time. This method has the following drawbacks: ① It is greatly affected by lighting and weather conditions; ② It has limited nighttime monitoring capabilities; ③ It can only cover a limited area and cannot achieve large-scale monitoring.

[0007] (4) Traditional seismological methods: using seismic networks to record seismic signals generated by events such as landslides and floods. Existing studies have shown that the detachment and impact of rock masses / glaciers, as well as the flow processes of debris flows and flash floods, can generate corresponding seismic wave signals, providing the possibility for disaster early warning based on seismic network data. However, existing methods have the following technical bottlenecks: ① Difficulty in separating and identifying low-frequency and high-frequency signals: The initial stage of a landslide generates abundant low-frequency signals (0.08-0.15 Hz), which can be identified at far-field stations thousands of kilometers away; while subsequent debris flows and floods mainly generate high-frequency signals (>1 Hz), which can only be observed at near-field stations within hundreds of kilometers. Existing methods are difficult to process the two types of signals in a unified and effective manner. ② Low signal-to-noise ratio of high-frequency flood signals: Flood signals are weak and easily masked by environmental noise. Their detectability is significantly affected by diurnal and seasonal changes. They are easy to identify when the noise level is low at night, but difficult to identify when the noise level is high during the day. ③ Lack of automated identification and location methods: Traditional earthquake signal analysis relies on manual interpretation, which is inefficient, highly subjective, and difficult to meet the needs of real-time early warning. ④ Ineffective for disaster events without low-frequency triggering signals: For disaster events such as glacial lake outbursts and landslide dammed lake outbursts that lack landslide triggering mechanisms, there are no low-frequency signals available for early warning, and it is necessary to develop rapid identification and location methods for high-frequency signals. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention provides a method and system for identifying and providing early warning of catastrophic flood events. This method solves the problems of poor early warning effect, low signal identification and positioning accuracy, and great influence of environmental noise in traditional methods. It improves the efficiency and accuracy of flood event identification and early warning, and enables rapid identification and early warning of complex disaster events such as landslide-debris flow-flood.

[0009] To achieve the above objectives, the present invention provides the following solution: A method for identifying and providing early warning of catastrophic flood events, comprising: Step 1: Deploy a dense broadband seismic network in the study area and continuously acquire three-component seismic data; Step 2: Perform mean removal, tilt removal, instrument response correction, and bandpass filtering on the seismic data, and set up multi-band parallel processing channels; Step 3: Calculate the noise power spectral density of each station and frequency band in real time, and dynamically adjust the event detection threshold; Step 4: Use the long-time-short-time average ratio method to detect seismic events in real time across each frequency band channel; Step 5: Based on the frequency band characteristics, duration, and station distribution range of the triggering event, make a preliminary determination of the event type; Step 6: For events triggered by low-frequency channels, the landslide dismantling process is reconstructed using the single-force-source waveform inversion method, and dynamic parameters are extracted. Step 7: For events triggered by high-frequency channels, a combination of normalized weighted cross-correlation and amplitude-distance analysis is used for localization. Step 8: Calculate the flood flow velocity based on the positioning results, and predict the future path and arrival time by combining the digital elevation model and river network data; Step 9: Trigger tiered alerts based on event type, scale, location, and predicted arrival time.

[0010] As a preferred option, it also includes: step ten, continuously optimizing the detection threshold, location algorithm parameters, and early warning triggering conditions based on historical event data.

[0011] Preferably, in step two, the multi-band parallel processing channel includes: (1) Low-frequency channel: 0.05-0.5 Hz, used for landslide removal event identification; (2) Intermediate frequency channel: 0.5-2.0 Hz, used for debris flow event identification; (3) High frequency channel: 2.0-10 Hz, used for flood event identification.

[0012] Preferably, in step four, the method for detecting the long-term / short-term average ratio is as follows: in, For short-term average, For long-term averaging, when Event detection is triggered when the threshold is exceeded.

[0013] Preferably, in step seven, the normalized weighted cross-correlation localization method is as follows: For the two stations and Recorded signals and Define the normalized weighted cross-correlation coefficient: in, This is the weighting function.

[0014] The source location is determined by using a grid search to achieve the best consistency between the actual observed time difference and the theoretical time difference.

[0015] Preferably, in step seven, the amplitude-distance positioning method is as follows: seismic wave amplitude Source distance The relationship is represented as: in, The amplitude at the source, The geometric diffusion coefficient is 1. This is the attenuation coefficient.

[0016] The source location is determined by optimizing the fit between the observed amplitude and the theoretical amplitude through grid search.

[0017] The present invention also provides a system for identifying and providing early warning of catastrophic flood events, comprising: The first processing module is used to deploy a dense broadband seismic network in the study area and continuously acquire three-component seismic data. The second processing module is used to perform mean removal, tilt removal, instrument response correction, and bandpass filtering on seismic data, and sets up multi-band parallel processing channels. The third processing module is used to calculate the noise power spectral density of each station in each frequency band in real time and dynamically adjust the event detection threshold. The fourth processing module is used to detect seismic events in real time in each frequency band channel using the long-short time average ratio method; The fifth processing module is used to make a preliminary judgment on the event type based on the frequency band characteristics, duration, and station distribution range of the triggering event; The sixth processing module is used to reconstruct the landslide dismantling process and extract dynamic parameters for low-frequency channel-triggered events using a single-force-source waveform inversion method. The seventh processing module is used to locate events triggered by high-frequency channels using a combination of normalized weighted cross-correlation and amplitude-distance analysis. The eighth processing module is used to calculate the speed of the flood based on the positioning results, and to predict the future path and arrival time by combining the digital elevation model and river network data. The ninth processing module is used to trigger tiered early warnings based on event type, scale, location, and predicted arrival time.

[0018] As a preferred embodiment, it also includes: a tenth processing module, used to continuously optimize the detection threshold, location algorithm parameters and early warning triggering conditions based on historical event data.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes seismic wave signals acquired by a dense regional broadband seismic network to distinguish the characteristics of seismic signals at different stages of catastrophic floods. It reconstructs the initial stage of a landslide through single-source waveform inversion, and combines normalized weighted cross-correlation and amplitude-distance analysis to achieve real-time tracking and positioning of the flood's trajectory. Simultaneously, it comprehensively considers environmental noise levels and station distribution characteristics to determine signal effectiveness and conduct risk assessment, ultimately achieving graded early warning, effectiveness evaluation, and model optimization for catastrophic flood events. This invention enables continuous monitoring of the entire process of landslide-debris flow-flood complex disaster events, and can enhance and track high-frequency flood signals in real time, providing valuable early warning windows for downstream areas. This invention has advantages such as all-weather operation, wide coverage, strong early warning timeliness, high automation, and cost-effectiveness. It is applicable to various disaster types such as rockfalls, glacial landslides, and glacial lake outburst floods, providing a feasible technical solution for early warning of catastrophic flood disasters in mountainous areas. Attached Figure Description

[0020] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are 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.

[0021] Figure 1 A schematic diagram of the regional seismic network deployment and disaster monitoring; Figure 2 This is a flowchart of the catastrophic flood event identification and early warning method according to an embodiment of the present invention; Figure 3 Flowchart for multi-band parallel processing and time monitoring; Figure 4 Flowchart for single-force-source waveform inversion of low-frequency events; Figure 5 Flowchart for joint localization and trajectory tracking of high-frequency events; Figure 6 This is a flowchart for the triggering and release of tiered early warnings. Detailed Implementation

[0022] 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.

[0023] 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.

[0024] Example 1 This invention provides a method for identifying and providing early warning of catastrophic flood events. Based on the seismic signal characteristics of landslide-debris flow-flood composite disaster events, it reveals the signal patterns of events at different stages: (1) Landslide Deconstruction and Impact Stage ① Signal characteristics: mainly low-frequency signals, with the dominant frequency band being 0.08-0.15 Hz; ②Signal strength: The signal is strong and can be identified by far-field stations thousands of kilometers away; ③ Duration: From tens of seconds to several minutes; ④ Physical mechanism: During the acceleration and deceleration of the landslide as a whole, the movement of the huge mass generates long-period seismic waves; ⑤ Focal mechanism: It can be approximated as a single-source model.

[0025] (2) Debris flow transport stage ①Signal characteristics: Low-frequency components gradually decrease, while mid- and high-frequency components increase; ②Signal strength: The signal is moderate and can be identified by near-field stations within 100 kilometers; ③ Duration: From a few minutes to several tens of minutes; ④ Physical mechanism: The fractured rock and soil flow along the gully and interact with the gully bed to generate vibration.

[0026] (3) Flood transport stage ① Signal characteristics: dominated by high-frequency components, with a dominant frequency band >1 Hz; ②Signal strength: The signal is weak and can only be identified at near-field stations (<100 km) and when the ambient noise is low; ③ Duration: From tens of minutes to several hours; ④ Physical mechanism: During the water-solid two-phase flow process, high-frequency vibrations are generated by turbulence, particle collisions, and riverbed erosion; ⑤ Detectability: Significantly affected by diurnal and seasonal variations in environmental noise levels.

[0027] like Figure 2 As shown, the catastrophic flood event identification and early warning method of the present invention includes: Step 1: Deployment of Regional Seismic Network and Data Acquisition A dense broadband seismic network was deployed in the study area, such as Figure 1 As shown, the station spacing is determined to be 10-50 km based on terrain conditions and early warning requirements; the stations continuously record three-component seismic data and transmit it to the data processing center in real time.

[0028] Step 2: Real-time data preprocessing Preprocessing of real-time transmitted seismic data includes: mean removal, tilt removal, instrument response correction, and bandpass filtering, such as... Figure 3 As shown; based on the warning target, multiple frequency band parallel processing channels are set up: ① Low-frequency channel: 0.05-0.5 Hz, used for landslide removal event identification; ② Intermediate frequency channel: 0.5-2.0 Hz, used for debris flow event identification; ③ High-frequency channel: 2.0-10 Hz, used for flood event identification.

[0029] Step 3: Real-time monitoring of environmental noise levels The noise power spectral density of each station and frequency band is calculated in real time to establish a background noise model; the signal-to-noise ratio changes are dynamically monitored to identify periods when the noise level is abnormally low (at night) or high (during the day); and the event detection threshold is adaptively adjusted according to the real-time signal-to-noise ratio.

[0030] Step 4: Automatic Detection of Multi-Stage Events The long-time-short-time averaging ratio method is used to detect seismic events in real time across various frequency bands. in, It is a short-time average (reflecting the current signal strength). This is a long-term average (reflecting the background noise level). When When the preset threshold is exceeded, event detection is triggered.

[0031] Step 5: Initial Judgment of Event Type Based on the frequency band characteristics, duration, and station distribution range of the triggering event, the event type is preliminarily determined: ① If the low-frequency channel triggers significantly and multiple stations trigger at long distances, it is judged as a potential landslide removal event; ②If the high-frequency channel triggers significantly, and only near-field stations trigger, it is judged as a potential flood event; ③ If multiple frequency bands are triggered continuously and the frequency shifts from low to high over time, it is determined to be a landslide-debris flow-flood composite event.

[0032] Step Six: Low-Frequency Event Source Inversion and Parameter Extraction For events triggered by low-frequency channels, a single-force-source waveform inversion method is used, such as... Figure 4 As shown, the landslide demolition and impact process is reconstructed: ① Calculate the Green's function based on the regional velocity structure model; ② Use grid search or linear inversion methods to solve for the best-fit single-force-source time function; ③ Extract key parameters: force source direction, force source strength, total impulse, and duration; ④ Estimate landslide volume: Estimate the mass of the landslide based on the relationship between force source intensity and acceleration; ⑤ Trigger Level 1 Warning: When the estimated landslide volume exceeds the preset threshold, a potential disaster warning is issued.

[0033] Step 7: Real-time location and tracking of high-frequency events For events triggered by high-frequency channels, a combined approach of normalized weighted cross-correlation and amplitude-distance analysis is used for localization to achieve real-time tracking of flood migration trajectories, such as... Figure 5 As shown: 7.1 Normalized Weighted Cross-Correlation Positioning ① Select the 3-5 stations with the highest signal-to-noise ratio; ②Calculate the normalized weighted cross-correlation coefficient between station pairs using signal envelope or energy as weights; ③ Based on theoretical travel time difference, the optimal source location is determined through grid search; ④ Calculate continuously using a sliding time window (window length 30-60 seconds, sliding step 5-10 seconds) to obtain the trajectory of the flood front position changing over time.

[0034] 7.2 Amplitude-Distance Positioning ① Measure the root mean square amplitude of the signals at each station; ②Based on the regional attenuation model (geometric diffusion coefficient) attenuation coefficient The observed amplitude is optimally matched to the theoretical amplitude through grid search. ③ Obtain source location estimates independent of the cross-correlation method.

[0035] 7.3 Fusion of Positioning Results ① The positioning results obtained from the two methods are weighted and fused, and the weights are adaptively adjusted according to the real-time signal-to-noise ratio; ② Output the position, confidence interval, and movement speed of the fused torrent.

[0036] Step 8: Flood velocity estimation and path prediction Based on the continuous positioning results obtained in step seven, calculate the movement speed of the flood front: By combining digital elevation models and river network data, the future migration path of floodwaters and the time it takes to reach key downstream points are predicted. Let the path of the river centerline be a space curve parametric equation. , The path distance along the river channel; the flood propagation velocity field at point s along the river channel. Points are awarded based on the movement path of the Torrent Forward over time: Corresponding planar movement path coordinates: Step Nine: Triggering of Tiered Early Warnings Based on the event type, scale, location, and predicted arrival time, tiered alerts are triggered, such as... Figure 6 As shown: (1) Level 1 Warning (Attention Level): Triggering condition: Low-frequency event inversion shows landslide volume > 10 5 m 3 Or, a high-frequency event is detected by more than 3 stations; Response measures: Internal system flagging, enhanced monitoring, and notification of on-duty personnel.

[0037] (2) Level II Warning (Alert Level): Triggering conditions: Confirmation of a flood event, location results showing the flood entering the main river channel, and a movement speed >2 m / s; Response measures: Notify downstream communities and key facilities (hydropower stations, bridges, etc.) to strengthen vigilance and prepare for emergency response.

[0038] (3) Level III Warning (Alarm Level): Triggering condition: The flood is predicted to reach densely populated areas or critical infrastructure within 30 minutes; Response measures: Issue warnings to the public through emergency broadcasts, text messages, and other means, and organize evacuations.

[0039] (4) Level IV Warning (Emergency Level): Triggering conditions: The flood is predicted to arrive within 10 minutes, or the flood size has been detected to be far exceeding the threshold; Response measures: Forced evacuation, activation of emergency response plan, and mobilization of rescue forces.

[0040] Step 10: Early Warning Effectiveness Evaluation and Model Optimization Record the warning time, warning accuracy, false alarm rate, and false alarm rate for each event; optimize the detection threshold, location algorithm parameters, and warning trigger conditions based on historical event data; and introduce machine learning methods to continuously improve event recognition and location performance.

[0041] As one embodiment of the present invention, in step one, the following key locations are given priority consideration when deploying the regional seismic network: upstream potential landslide source areas, both sides of the main river channel, downstream densely populated areas, and the vicinity of hydropower facilities; broadband seismometers (frequency range 0.03-50 Hz) are used at the stations to ensure coverage of low-frequency landslide signals and high-frequency flood signals.

[0042] In one embodiment of the present invention, in step three, the detectability of the torrent signal is significantly affected by the level of environmental noise. The environmental noise mainly originates from: ①Human noise: Vibrations generated by daytime traffic, industrial activities, etc., with a frequency range of 1-10 Hz; ②Natural noise: Vibrations generated by wind, rivers, ocean waves, etc., with a frequency range of 0.1-1 Hz.

[0043] Environmental noise levels exhibit clear diurnal and seasonal variations. At night, reduced human activity lowers noise levels and improves the detectability of flood signals; during the day, higher noise levels may mask flood signals.

[0044] Signal-to-noise ratio (SNR) ) is defined as: in, For signal power, This represents noise power. When... When the threshold (usually 3-5) is reached, the signal can be effectively identified.

[0045] Furthermore, real-time monitoring of environmental noise levels includes: establishing day-night noise models and seasonal noise models, and calculating dynamic detection thresholds for each time period; for low-noise periods at night, lowering the detection threshold to improve sensitivity; and for high-noise periods during the day, raising the detection threshold to reduce false alarm rates.

[0046] As one embodiment of the present invention, in step five, the preliminary judgment of the event type also combines regional geological conditions and historical event characteristics to establish a localized event type discrimination rule library, thereby improving the accuracy of the preliminary judgment.

[0047] As one embodiment of the present invention, in step six, the single force source waveform inversion adopts a multi-scale inversion strategy: first, the overall force source time function is inverted using low-frequency data (0.05-0.2 Hz), and then higher-frequency data is gradually introduced to improve the resolution.

[0048] Furthermore, for mass motion events such as landslides, the resulting seismic waves can be approximated as single-source excitation. The basic principle of single-source inversion is as follows: earthquake records With single-source time function The relationship between them can be represented as: in, Let be the Green's function, describing the propagation effect of seismic waves from the source to the receiver. Through waveform inversion, the best-fit observation record can be solved. : The direction of the single force source is consistent with the direction of landslide movement, and the magnitude of the force is proportional to the acceleration of the landslide mass. Inversion can obtain the dynamic parameters of the landslide disintegration and impact processes, providing trigger source information for early warning.

[0049] As one embodiment of the present invention, in step seven, the normalized weighted cross-correlation method uses frequency domain calculation to improve computational efficiency; the weighting function uses the square of the signal envelope to highlight the energy concentration period.

[0050] Normalized weighted cross-correlation is a common method for tracing the location of moving sources using seismic network records. Its basic principle is as follows: For the two stations and Recorded signals and Define the normalized weighted cross-correlation coefficient: in, The weighting function typically uses the signal envelope or energy as the weights. This represents the time difference between the two stations.

[0051] For the assumed source location and earthquake time Theoretical time difference can be calculated. By using a grid search, the best agreement between the actual observed time difference and the theoretical time difference is achieved; the corresponding x is the source location. This method enables continuous tracking and positioning of moving sources, such as torrent fronts.

[0052] Furthermore, in step seven, the attenuation parameter of the amplitude-distance analysis method is dynamically corrected based on real-time monitoring data: the regional attenuation parameter is calibrated using events at known locations (such as quarry blasting or artificial seismic sources).

[0053] Amplitude-distance analysis utilizes the attenuation of seismic wave amplitude with distance for source location. (Seismic wave amplitude) Source distance The relationship can be represented as: in, The amplitude at the source, The geometric diffusion coefficient is 1. Let be the attenuation coefficient. Taking the logarithm of the above formula: For the amplitude of each station recorded by a given network The source location can be determined by minimizing the following expression through a grid search: This method has low requirements for signal waveform consistency and is suitable for high-frequency torrent signals with low signal-to-noise ratio and complex waveforms.

[0054] As one embodiment of the present invention, in step ten, the machine learning method includes: using a convolutional neural network to automatically extract seismic signal features and train an event detection and classification model; using a random forest or gradient boosting tree to establish an event type discrimination model; and using a recurrent neural network to predict the trend of flood velocity changes.

[0055] The present invention has the following technical effects: 1. Multi-stage event full-coverage monitoring: For the first time, continuous monitoring of the entire process of landslide-debris flow-flood complex disaster events has been achieved. From the low-frequency signals of landslide detachment to the high-frequency signals of flood transport, all can be effectively identified, filling the gap that traditional methods can only monitor a single stage.

[0056] 2. High-frequency flood signal enhancement processing: To address the technical challenges of low signal-to-noise ratio and difficulty in identifying flood signals, a method for real-time environmental noise monitoring and adaptive threshold adjustment is proposed. Combined with normalized weighted cross-correlation and amplitude-distance joint positioning, the identification capability and positioning accuracy of high-frequency signals are significantly improved.

[0057] 3. Real-time positioning and trajectory tracking: It enables continuous real-time positioning of the flood front, tracks the movement trajectory of the flood, calculates the movement speed, and predicts the time of arrival at key downstream points, providing accurate spatiotemporal information for emergency response.

[0058] 4. Tiered early warning mechanism: A four-tiered early warning mechanism has been established, which responds according to the scale, location and predicted time of the event. This avoids early warning fatigue caused by a "one-size-fits-all" approach and improves the effectiveness and relevance of early warnings.

[0059] 5. All-weather operation capability: The monitoring method based on the seismic network is not affected by light or weather conditions and can operate continuously around the clock, overcoming the limitations of optical remote sensing, video surveillance and other methods.

[0060] 6. Wide coverage: A single seismic station can monitor flood events within a range of tens to hundreds of kilometers. By rationally deploying a network of stations, full coverage monitoring of large river basins can be achieved, which has significant advantages over traditional point-based monitoring methods.

[0061] 7. Early warning time window: By monitoring upstream landslide triggering events, an early warning time window can be obtained before the flood forms; by tracking the movement of the flood in real time, a warning time of several minutes to tens of minutes can be provided to the downstream, buying valuable time for personnel evacuation and emergency response.

[0062] 8. Adaptability to events without landslide triggering: The high-frequency signal identification and location method of this invention is also applicable to disaster events without low-frequency triggering signals, such as glacial lake outbursts and landslide dammed lake outbursts, thus broadening the scope of application of earthquake network early warning.

[0063] 9. Intelligent Continuous Optimization: Introducing machine learning methods, the algorithm parameters are continuously optimized based on historical event data to improve the accuracy of event recognition and positioning, thereby enabling the early warning system to evolve itself.

[0064] 10. Economical and efficient: It utilizes existing regional seismic network resources, eliminating the need for large-scale construction of dedicated monitoring facilities, thus possessing high economic efficiency and scalability.

[0065] Implementation Case: Early Warning Simulation Application of a Landslide-Debris Flow-Flood Event on the Qinghai-Tibet Plateau 1. Overview of the study area and network deployment A catastrophic landslide-debris flow-flood event in a mountainous area of ​​the Qinghai-Tibet Plateau was selected as the simulation application area for this invention. This area is located on the southern slope of the Himalayas, with steep terrain, peaks exceeding 6000 m in altitude, and deep valleys, making it a high-risk area for landslide-flood disasters.

[0066] A dense broadband seismic network is deployed within the study area, with stations spaced approximately 20-50 km apart, totaling 15 broadband seismic stations. Each station is equipped with a broadband seismograph (bandwidth range 0.03-50 Hz), a data acquisition unit with a sampling rate of 100 Hz, and real-time data transmission to the data processing center.

[0067] 2. Simulation of the event occurrence process At 4:51 UTC in 2021, a large-scale glacial landslide occurred on the Qinghai-Tibet Plateau, with a landslide volume of approximately 2.7 × 10⁻⁶. 7 m 3 The landslide slid down at high speed, impacting the valley and transforming into a debris flow, which further eroded the riverbed material, creating a catastrophic flood. The floodwaters moved downstream along the valley, destroying several villages and hydroelectric facilities, resulting in more than 200 deaths or missing persons.

[0068] 3. Simulation of the application process of the method of the present invention 3.1 Real-time data acquisition and preprocessing The station continuously records three-component seismic data and transmits it to the data processing center in real time. The data preprocessing module removes the mean, tilt, and corrects the instrument response of the raw data, and performs bandpass filtering in three frequency bands: ① Low frequency channel: 0.05-0.5 Hz; ② Mid frequency channel: 0.5-2.0 Hz; ③ High frequency channel: 2.0-10 Hz.

[0069] 3.2 Environmental noise level monitoring The noise monitoring module calculates the noise power spectral density of each station at each frequency band in real time. The results show that the environmental noise in the study area exhibits a significant diurnal variation: daytime (6:00-20:00 local time) noise levels are high, mainly originating from traffic and industrial activities; nighttime noise levels decrease by approximately 10-15 dB. Based on this, the system adaptively adjusts the detection threshold: the nighttime threshold is reduced by 30% to improve sensitivity; the daytime threshold is increased by 20% to reduce the false alarm rate.

[0070] 3.3 Automatic Event Detection At 4:51:30 UTC, the low-frequency channel triggered STA / LTA detection simultaneously at multiple stations (STA=2s, LTA=60s, threshold=4.0). Triggering stations included Station 1 and Station 2, with the furthest triggering station exceeding 500 km, indicating a massive event energy. Subsequently (4:55-6:00), the mid-frequency and high-frequency channels were triggered successively, and the triggering station range gradually shrank to the near field (<100km), consistent with the evolution characteristics of a landslide-debris flow-flood complex event.

[0071] 3.4 Preliminary judgment of event type The event type identification module initially judges the event as a landslide-debris flow-flood composite event based on the following characteristics: ① Low-frequency channel triggers first, and multiple stations trigger remotely → landslide dismantling event; ② Medium and high-frequency channel triggers subsequently, and migrates from low frequency to high frequency over time → debris flow-flood transformation; ③ Long trigger duration (>1 hour) → flood migration process.

[0072] 3.5 Low-frequency event source inversion The low-frequency event inversion module initiates single-force-source waveform inversion: ① Calculate the Green's function based on the regional velocity structure model (CRUST1.0); ② Solve for the best-fit single-force-source time function using the grid search method; ③ Inversion frequency band: 0.08-0.15Hz.

[0073] The inversion results show that: ① the force source direction is nearly north-south, consistent with the landslide movement direction; ② the force source time function exhibits a bimodal characteristic, corresponding to landslide body disintegration (4:51:30) and impact valley floor (4:52:15); ③ the total impulse is approximately 1.2 × 10⁻⁶. 10 N·s; ④ The estimated mass of the landslide is approximately 2.5 × 10⁻⁶ N·s; 10 kg (corresponding to a body size of approximately 2.5 × 10⁻⁶ kg) 7 m 3 The system triggered a Level 1 warning (attention level) based on the results of the investigation, marking the event as a potential large landslide.

[0074] 3.6 Real-time location and tracking of high-frequency events The high-frequency event localization module initiates joint localization: (1) Normalized weighted cross-correlation method for localization: ① Select the 5 near-field stations with the highest signal-to-noise ratio (station 1, station 2, station 3, station 4, station 6); ② Calculate the normalized weighted cross-correlation coefficient between station pairs using the square of the signal envelope as the weight; ③ Time window length 40 seconds, sliding step size 10 seconds; ④ Grid search range: 80 km × 60 km in the study area, grid spacing 0.5 km.

[0075] (2) Amplitude-distance method for localization: ① Measure the root mean square amplitude of each station in the 2-10 Hz frequency band; ② Use the regional attenuation model: n=1.2, α=0.003 km -1 (Based on historical blasting events for calibration); ③ Grid search optimizes the fit between observed amplitude and theoretical amplitude.

[0076] (3) Fusion of positioning results: ① The positioning results of the two methods are weighted and fused, and the weights are adaptively adjusted according to the real-time signal-to-noise ratio (the cross-correlation method has a greater weight when the signal-to-noise ratio is high, and the amplitude method has a greater weight when the signal-to-noise ratio is low); ② Output the fused torrent position and confidence ellipse (95% confidence level).

[0077] Continuous location data shows that: ① at 4:55, the flood front was located in the landslide impact zone (approximately 2 km from the epicenter); ② at 5:10, the flood front had moved to approximately 8 km downstream; ③ at 5:30, the flood front had moved to approximately 15 km downstream; ④ at 5:50, the flood front had moved to approximately 22 km downstream (near the first village). The location data is highly consistent with the flood peak location shown in satellite imagery, with an error of less than 1.5 km.

[0078] 3.7 Flood velocity estimation and path prediction Based on continuous positioning results, the moving speed of the flood front was calculated as follows: ① 4:55-5:10 average speed: 6.7 m / s; ② 5:10-5:30 average speed: 5.8 m / s; ③ 5:30-5:50 average speed: 5.0 m / s. The overall speed showed a slow decreasing trend.

[0079] Combining digital elevation models and river network data, the future path of the flood is predicted as follows: ① Main downstream threat points: Village 1 (25 km from the epicenter), Village 2 (32 km from the epicenter), and hydropower station (35 km from the epicenter); ② Predicted arrival time: Village 1 approximately 6:00, Village 2 approximately 6:15, and hydropower station approximately 6:20.

[0080] 3.8 Triggering of Tiered Early Warnings (1) Level 1 Warning (4:52): Triggering condition: Low-frequency event inversion shows landslide volume > 10 5 m 3 ; Response measures: The system will internally flag the issue and notify on-duty personnel to strengthen monitoring.

[0081] (2) Level II Warning (5:00): Triggering conditions: Confirmation of a flood event, location results showing the flood entering the main river channel, and a movement speed >5 m / s; Response measures: Notify downstream villages and hydropower stations to strengthen vigilance and prepare for emergency response.

[0082] (3) Level III Warning (5:20): Trigger condition: The flood is predicted to reach village 1 within 40 minutes; Response measures: Residents of Village 1 were notified via emergency broadcast to prepare for evacuation.

[0083] (4) Level IV Warning (5:40): Trigger condition: The flood is predicted to reach village 1 within 20 minutes; Response measures: Force evacuation of 1 village resident and activation of emergency response plan.

[0084] 3.9 Evaluation of Early Warning Effectiveness ① Warning time: Only 1 minute from the occurrence of the event (4:51) to the first warning (4:52); ② Warning time for downstream villages: Village 1 approximately 20 minutes in advance, Village 2 approximately 35 minutes in advance, and hydropower station approximately 40 minutes in advance; ③ Early warning accuracy: Positioning error <1.5 km, speed estimation error <10%, and path prediction consistent with the actual situation; ④ Potential benefits: It can buy valuable evacuation time for downstream residents and significantly reduce the risk of casualties.

[0085] Example 2 The present invention also provides a system for identifying and providing early warning of catastrophic flood events, comprising: The first processing module is used to deploy a dense broadband seismic network in the study area and continuously acquire three-component seismic data. The second processing module is used to perform mean removal, tilt removal, instrument response correction, and bandpass filtering on seismic data, and sets up multi-band parallel processing channels. The third processing module is used to calculate the noise power spectral density of each station in each frequency band in real time and dynamically adjust the event detection threshold. The fourth processing module is used to detect seismic events in real time in each frequency band channel using the long-short time average ratio method; The fifth processing module is used to make a preliminary judgment on the event type based on the frequency band characteristics, duration, and station distribution range of the triggering event; The sixth processing module is used to reconstruct the landslide dismantling process and extract dynamic parameters for low-frequency channel-triggered events using a single-force-source waveform inversion method. The seventh processing module is used to locate events triggered by high-frequency channels using a combination of normalized weighted cross-correlation and amplitude-distance analysis. The eighth processing module is used to calculate the speed of the flood based on the positioning results, and to predict the future path and arrival time by combining the digital elevation model and river network data. The ninth processing module is used to trigger tiered early warnings based on event type, scale, location, and predicted arrival time.

[0086] As one embodiment of the present invention, it further includes: a tenth processing module, used to continuously optimize the detection threshold, positioning algorithm parameters and early warning triggering conditions based on historical event data.

[0087] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for identifying and providing early warning of catastrophic flood events, characterized in that, include: Step 1: Deploy a dense broadband seismic network in the study area and continuously acquire three-component seismic data; Step 2: Perform mean removal, tilt removal, instrument response correction, and bandpass filtering on the seismic data, and set up multi-band parallel processing channels; Step 3: Calculate the noise power spectral density of each station and frequency band in real time, and dynamically adjust the event detection threshold; Step 4: Use the long-time-short-time average ratio method to detect seismic events in real time across each frequency band channel; Step 5: Based on the frequency band characteristics, duration, and station distribution range of the triggering event, make a preliminary determination of the event type; Step 6: For events triggered by low-frequency channels, the landslide dismantling process is reconstructed using the single-force-source waveform inversion method, and dynamic parameters are extracted. Step 7: For events triggered by high-frequency channels, a combination of normalized weighted cross-correlation and amplitude-distance analysis is used for localization. Step 8: Calculate the flood flow velocity based on the positioning results, and predict the future path and arrival time by combining the digital elevation model and river network data; Step 9: Trigger tiered alerts based on event type, scale, location, and predicted arrival time.

2. The method for identifying and providing early warning of catastrophic flood events as described in claim 1, characterized in that, It also includes: Step 10, continuously optimizing the detection threshold, location algorithm parameters, and early warning triggering conditions based on historical event data.

3. The method for identifying and providing early warning of catastrophic flood events as described in claim 2, characterized in that, In step two, the multi-band parallel processing channel includes: (1) Low-frequency channel: 0.05-0.5 Hz, used for landslide removal event identification; (2) Intermediate frequency channel: 0.5-2.0 Hz, used for debris flow event identification; (3) High frequency channel: 2.0-10 Hz, used for flood event identification.

4. The method for identifying and providing early warning of catastrophic flood events as described in claim 3, characterized in that, In step four, the method for detecting the long-term / short-term average ratio is as follows: in, For short-term average, For long-term averaging, when Event detection is triggered when the threshold is exceeded.

5. The method for identifying and providing early warning of catastrophic flood events as described in claim 4, characterized in that, In step seven, the normalized weighted cross-correlation localization method is as follows: For the two stations and Recorded signals and Define the normalized weighted cross-correlation coefficient: in, This is the weighting function. The source location is determined by using a grid search to achieve the best consistency between the actual observed time difference and the theoretical time difference.

6. The method for identifying and providing early warning of catastrophic flood events as described in claim 5, characterized in that, In step seven, the amplitude-distance positioning method is as follows: seismic wave amplitude Source distance The relationship is represented as: in, The amplitude at the source, The geometric diffusion coefficient is... This is the attenuation coefficient. The source location is determined by optimizing the fit between the observed amplitude and the theoretical amplitude through grid search.

7. A system for identifying and providing early warning of catastrophic flood events, characterized in that, include: The first processing module is used to deploy a dense broadband seismic network in the study area and continuously acquire three-component seismic data. The second processing module is used to perform mean removal, tilt removal, instrument response correction, and bandpass filtering on seismic data, and sets up multi-band parallel processing channels. The third processing module is used to calculate the noise power spectral density of each station in each frequency band in real time and dynamically adjust the event detection threshold. The fourth processing module is used to detect seismic events in real time in each frequency band channel using the long-short time average ratio method; The fifth processing module is used to make a preliminary judgment on the event type based on the frequency band characteristics, duration, and station distribution range of the triggering event; The sixth processing module is used to reconstruct the landslide dismantling process and extract dynamic parameters for low-frequency channel-triggered events using a single-force-source waveform inversion method. The seventh processing module is used to locate events triggered by high-frequency channels using a combination of normalized weighted cross-correlation and amplitude-distance analysis. The eighth processing module is used to calculate the speed of the flood based on the positioning results, and to predict the future path and arrival time by combining the digital elevation model and river network data. The ninth processing module is used to trigger tiered early warnings based on event type, scale, location, and predicted arrival time.

8. The catastrophic flood event identification and early warning system as described in claim 7, characterized in that, Also includes: The tenth processing module is used to continuously optimize the detection threshold, location algorithm parameters, and early warning triggering conditions based on historical event data.

Images