Landslide disaster monitoring and early warning method for frozen soil area of Tibet Plateau
Through multi-source data collection and processing, feature extraction, risk assessment and dynamic updating methods, combined with satellite remote sensing and ground sensor networks, a landslide risk assessment model was constructed, which solved the problems of inaccurate landslide disaster monitoring and untimely early warning in the permafrost area of the Qinghai-Tibet Plateau, achieved accurate monitoring and timely early warning of landslide disasters, and improved the reliability and effectiveness of early warning.
Patent Information
- Application Number
- CN202510770747.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-10
- Publication Date
- 2025-09-12
AI Technical Summary
The existing landslide hazard monitoring methods in the permafrost areas of the Qinghai-Tibet Plateau have the problems of single data collection mode, traditional data processing, untimely warning and lack of dynamic update mechanism, resulting in inaccurate landslide risk assessment and inaccurate warning information.
Using multi-source data collection, data preprocessing, feature extraction, risk assessment, warning issuance and dynamic update methods, combined with satellite remote sensing technology and ground sensor networks, a landslide risk assessment model is constructed through machine learning algorithms, and warning information is issued through multiple channels. The model is updated regularly to adapt to environmental changes.
It has achieved accurate monitoring and timely early warning of landslide disasters in the permafrost areas of the Qinghai-Tibet Plateau, improved the reliability and effectiveness of early warning, and reduced the losses caused by landslide disasters.
Smart Images
Figure CN120636098A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of disaster monitoring and early warning methods, and in particular to a landslide disaster monitoring and early warning method in a permafrost region of the Qinghai-Tibet Plateau. Background Art
[0002] Due to the unique geological and climatic conditions of the permafrost region of the Qinghai-Tibet Plateau, landslides are frequent, posing a serious threat to local infrastructure construction and the safety of residents' lives and property. Currently, monitoring and early warning methods for landslides in this region have certain limitations.
[0003] Existing monitoring methods often rely on a single data collection method, such as relying solely on ground sensors or satellite remote sensing, which makes it difficult to obtain comprehensive and accurate landslide-related information. Data processing and analysis methods are also relatively traditional and cannot effectively handle multi-source and complex data, resulting in inaccurate landslide risk assessments. Furthermore, existing early warning methods often fail to issue warning information in a timely and accurate manner and lack a dynamic update mechanism, making them difficult to adapt to the complex and changing geological and climatic conditions of the permafrost regions of the Qinghai-Tibet Plateau. Therefore, it is necessary to develop a landslide hazard monitoring and early warning method for the permafrost regions of the Qinghai-Tibet Plateau to meet practical needs. Summary of the Invention
[0004] In view of this, the present invention addresses the deficiencies in the existing technology, and its main purpose is to provide a method for monitoring and early warning of landslide disasters in the permafrost area of the Qinghai-Tibet Plateau. Through multi-source data collection, data preprocessing, feature extraction, risk assessment, early warning issuance and dynamic updating, it can achieve accurate monitoring and timely early warning of landslide disasters in the permafrost area of the Qinghai-Tibet Plateau, improve the reliability and effectiveness of early warning, and reduce the losses caused by landslide disasters.
[0005] To achieve the above object, the present invention adopts the following technical solutions:
[0006] A method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau comprises the following steps:
[0007] S1. Multi-source data acquisition: Satellite remote sensing technology is used to obtain regional topography and vegetation cover data, and a ground sensor network is used to collect permafrost temperature, humidity, displacement, groundwater level and meteorological data;
[0008] S2. Data preprocessing: cleaning, filtering, and normalizing the collected data;
[0009] S3, feature extraction: extracting characteristic parameters related to landslide hazards from the preprocessed data;
[0010] S4. Risk assessment: Build a landslide risk assessment model based on machine learning algorithms and evaluate the landslide risk level based on characteristic parameters;
[0011] S5. Warning release: When the risk level reaches the set threshold, warning information will be released through multiple channels;
[0012] S6. Dynamic update: Continuously collect data and regularly update risk assessment models.
[0013] As a preferred solution: in the feature extraction step S3, the thickness change rate of the frozen soil active layer V is calculated. ALT The formula is V ALT =(ALT t -ALT t-1 ) / Δ t， Among them ALT t ALT is the thickness of the active frozen soil layer at the current moment. t-1 is the thickness of the frozen soil active layer at the previous moment, Δ t is the time interval.
[0014] As a preferred solution: in the multi-source data collection in step S1, satellite remote sensing technology adopts a combination of synthetic aperture radar and optical remote sensing, and the ground sensor network includes displacement sensors, strain sensors, temperature and humidity sensors, and groundwater level sensors.
[0015] As a preferred solution: in the data preprocessing step S2, median filtering is used to remove noise in the data, and the minimum-maximum normalization method is used to normalize the data to the interval [0,1].
[0016] As a preferred solution: the feature extraction step S3 further extracts the displacement rate, groundwater level change rate and temperature change rate as feature parameters.
[0017] As a preferred solution: in the risk assessment of step S4, the machine learning algorithm adopts the random forest algorithm, uses historical landslide data as training samples, characteristic parameters as input, and landslide risk level as output for model training.
[0018] As a preferred solution: in the step S5 of issuing the warning, the warning information includes the landslide risk level, the time and place of possible occurrence, and suggestions for response measures, and the release channels include text messages, broadcasts, online platforms, and on-site alarm devices.
[0019] As a preferred solution: in the dynamic update of step S6, new data is collected at regular intervals, and the risk assessment model is updated using an incremental learning method.
[0020] As a preferred solution: in the multi-source data collection step S1, a data quality monitoring unit is further provided to monitor the accuracy and stability of the sensor data in real time, and to promptly maintain or replace the sensor when the data is abnormal.
[0021] As a preferred solution: in the risk assessment of step S4, landslide risks are divided into levels according to human activity factors in the area, and corresponding risk thresholds are set for each level.
[0022] Compared with the existing technology, the present invention has obvious advantages and beneficial effects. Specifically, it can be seen from the above technical scheme that through multi-source data collection, data preprocessing, feature extraction, risk assessment, warning issuance and dynamic updating, accurate monitoring and timely warning of landslide disasters in the permafrost area of the Qinghai-Tibet Plateau can be achieved, the reliability and effectiveness of the warning can be improved, and the losses caused by landslide disasters can be reduced; by adopting satellite remote sensing technology and ground sensor networks, comprehensive collection of various aspects of information in the permafrost area of the Qinghai-Tibet Plateau can be achieved; the median filtering, minimum-maximum normalization method and random forest algorithm are adopted to improve the accuracy of data processing and risk assessment; the dynamic update mechanism ensures that the model can adapt to environmental changes in a timely manner, thereby improving the reliability of warnings; the data quality monitoring unit ensures the accuracy and stability of the collected data, further enhancing the performance of the entire monitoring and early warning system.
[0023] To more clearly illustrate the structural features and effects of the present invention, it is described in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 This is a flow chart of a method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to the present invention;
[0025] Figure 2 This is a flow chart of the multi-source data collection steps of the present invention. DETAILED DESCRIPTION
[0026] The present invention Figures 1 to 2 As shown, a method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau includes the following steps:
[0027] S1. Multi-source data acquisition: Satellite remote sensing technology is used to obtain regional topography and vegetation cover data, and a ground sensor network is used to collect permafrost temperature, humidity, displacement, groundwater level and meteorological data;
[0028] S2. Data preprocessing: cleaning, filtering, and normalizing the collected data;
[0029] S3, feature extraction: extracting characteristic parameters related to landslide hazards from the preprocessed data;
[0030] S4. Risk assessment: Build a landslide risk assessment model based on machine learning algorithms and evaluate the landslide risk level based on characteristic parameters;
[0031] S5. Warning release: When the risk level reaches the set threshold, warning information will be released through multiple channels;
[0032] S6. Dynamic update: Continuously collect data and regularly update risk assessment models.
[0033] In the feature extraction step S3, the thickness change rate of the frozen soil active layer V is calculated. ALT The formula is V ALT =(ALT t -ALT t-1 ) / Δ t， Among them ALT t ALT is the thickness of the active frozen soil layer at the current moment. t-1 is the thickness of the frozen soil active layer at the previous moment, Δ t is the time interval.
[0034] In the multi-source data collection in step S1, satellite remote sensing technology adopts a combination of synthetic aperture radar and optical remote sensing, and the ground sensor network includes displacement sensors, strain sensors, temperature and humidity sensors, and groundwater level sensors.
[0035] In the data preprocessing step S2, median filtering is used to remove noise in the data, and the minimum-maximum normalization method is used to normalize the data to the interval [0, 1].
[0036] The feature extraction step S3 further extracts the displacement rate, groundwater level change rate and temperature change rate as feature parameters.
[0037] In the risk assessment of step S4, the machine learning algorithm adopts the random forest algorithm, with historical landslide data as training samples, characteristic parameters as input, and landslide risk level as output for model training.
[0038] In step S5, the warning information includes the landslide risk level, possible time and location of occurrence, and response measures. The release channels include text messages, broadcasts, online platforms, and on-site alarm devices.
[0039] In the dynamic update of step S6, new data is collected at regular intervals and the risk assessment model is updated using an incremental learning method.
[0040] In the multi-source data collection step S1, a data quality monitoring unit is also set up to monitor the accuracy and stability of sensor data in real time, and to perform maintenance or replace sensors in a timely manner when data is abnormal.
[0041] In the risk assessment of step S4, landslide risks are classified into different levels according to human activities in the area, and corresponding risk thresholds are set for each level.
[0042] Multi-source data fusion: The comprehensive use of satellite remote sensing technology and ground sensor networks has enabled the collection of multi-source data, enabling comprehensive and accurate acquisition of geological, meteorological and other information in the permafrost areas of the Qinghai-Tibet Plateau, providing a rich data foundation for landslide risk assessment.
[0043] Data processing and analysis technology: Median filtering and minimum-maximum normalization methods are used for data preprocessing, combined with the random forest algorithm for risk assessment. This can effectively process multi-source and complex data and improve the accuracy of landslide risk assessment.
[0044] Dynamic update mechanism: By regularly collecting new data and using incremental learning methods to update the risk assessment model, it can promptly adapt to changes in geological and climatic conditions in the permafrost areas of the Qinghai-Tibet Plateau, ensuring the timeliness and reliability of early warnings.
[0045] Comprehensive early warning information release: Early warning information includes landslide risk level, possible time and location of occurrence, response measures and recommendations, and is released through multiple channels, enabling relevant personnel to understand the landslide disaster situation in a timely manner and take effective response measures to reduce disaster losses.
[0046] Data quality monitoring: A data quality monitoring unit is set up to monitor sensor data in real time, ensuring the accuracy and stability of collected data and further improving the reliability of the entire monitoring and early warning system.
[0047] Synthetic aperture radar (SAR) offers all-day, all-weather observation capabilities, capable of penetrating clouds and some vegetation to obtain highly accurate topographic and geomorphological information. It is unrestricted by weather conditions and can provide continuous monitoring at night and in inclement weather. Optical remote sensing provides rich spectral information, helping to accurately identify vegetation cover type and status. The combination of these two technologies can comprehensively and accurately obtain regional topography and vegetation cover data, providing fundamental geographic information and vegetation status information for subsequent landslide hazard analysis, as vegetation cover is closely related to landslide stability, and topography is a key factor influencing landslide occurrence.
[0048] Displacement sensors monitor soil displacement in real time, enabling the detection of even the smallest signs of displacement before a landslide occurs. Strain sensors measure stress and strain within the soil, reflecting the load on the soil. Temperature and humidity sensors capture the temperature and humidity of frozen soil, as its physical properties vary significantly with temperature and humidity, making them key factors influencing landslides. Groundwater level sensors monitor the dynamics of groundwater levels. Rising groundwater levels increase soil weight and pore water pressure, reducing soil shear strength and potentially triggering landslides. Working in concert, these multiple sensors can comprehensively collect data on frozen soil, improving the accuracy and reliability of landslide monitoring.
[0049] Median filtering is used to remove noise from data: Median filtering is a nonlinear filtering method that effectively removes impulse noise and salt-and-pepper noise while preserving edge information. When processing multi-source data, noise may be present due to factors such as sensor errors and external interference. This noise can affect subsequent feature extraction and risk assessment. Removing noise through median filtering improves data quality and makes subsequent analysis more accurate and reliable.
[0050] Use the min-max normalization method to normalize the data to the [0, 1] range: making the data comparable helps the machine learning algorithm better process and analyze the data, improving the model's training effect and prediction accuracy.
[0051] The displacement rate directly reflects the speed of soil movement; a sudden increase in displacement rate may signal an impending landslide. The groundwater level change rate reflects its dynamics; a rapidly rising groundwater level increases the risk of landslides. Temperature changes affect the physical properties of permafrost, and the temperature change rate reflects changes in the thermal state of permafrost, which has a significant impact on landslide occurrence. Taken together, these characteristic parameters can more comprehensively describe the potential risk of landslide hazards and improve the accuracy of risk assessments.
[0052] The machine learning algorithm uses the Random Forest algorithm: an ensemble learning method with high accuracy and stability. It constructs multiple decision trees and integrates their results for prediction, making it capable of handling high-dimensional data and complex nonlinear relationships. Using historical landslide data as training samples, characteristic parameters as input, and landslide risk levels as output for model training, the Random Forest algorithm automatically learns the complex relationship between characteristic parameters and landslide risk levels, thereby accurately assessing current landslide risk.
[0053] Landslide risk is categorized based on human activities within the region, with corresponding risk thresholds set for each tier. Human activities such as construction, road excavation, and mineral extraction can alter the soil's original state, increasing the likelihood of landslides. Considering human activities within a region allows for a more comprehensive assessment of landslide risk. This risk grading and setting corresponding thresholds make early warnings more scientific and reasonable. Different risk levels correspond to different early warning measures, enhancing their relevance and effectiveness.
[0054] Early warning information allows relevant personnel to promptly understand the severity of landslide hazards, the specific circumstances that may occur, and the appropriate response measures. The landslide risk level helps people determine the threat level; the time and location of possible occurrences allow people to prepare for prevention; and the response measures recommended provide specific guidance for action, helping to reduce losses caused by the disaster.
[0055] Multiple dissemination channels ensure that warning information reaches relevant personnel promptly and widely. Text messages can be sent directly to individual mobile phones, ensuring timely receipt of information; broadcasts can cover a large area, allowing more people to understand the warning; online platforms can provide detailed information and real-time updates; and on-site alarm devices can sound an alarm at the scene of a disaster, prompting nearby personnel to evacuate quickly.
[0056] The risk assessment model is updated using incremental learning methods to collect new data at regular intervals. The geological, meteorological, and human activities in the permafrost region of the Qinghai-Tibet Plateau are constantly changing, and old models may not accurately reflect current conditions. Regularly collecting new data and updating the model using incremental learning methods allows the model to adapt to these changes in a timely manner, maintaining the accuracy and effectiveness of landslide risk assessments. This incremental learning method allows for local adjustments to the model using new data without retraining the entire model, improving the efficiency of model updates.
[0057] The design focus of the present invention is to achieve accurate monitoring and timely warning of landslide disasters in the permafrost area of the Qinghai-Tibet Plateau through multi-source data collection, data preprocessing, feature extraction, risk assessment, warning issuance and dynamic updating, thereby improving the reliability and effectiveness of warnings and reducing the losses caused by landslide disasters; by adopting satellite remote sensing technology and ground sensor networks, it is possible to achieve comprehensive collection of various aspects of information in the permafrost area of the Qinghai-Tibet Plateau; by adopting median filtering, minimum-maximum normalization method and random forest algorithm, the accuracy of data processing and risk assessment is improved; the dynamic update mechanism ensures that the model can adapt to environmental changes in a timely manner, improving the reliability of warnings; the data quality monitoring unit ensures the accuracy and stability of the collected data, further enhancing the performance of the entire monitoring and early warning system.
[0058] The above description is merely a preferred embodiment of the present invention and does not limit the technical scope of the present invention. Therefore, any minor modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention are still within the scope of the technical solution of the present invention.
Claims
1. A method for monitoring and early warning landslide disasters in permafrost areas of the Qinghai-Tibet Plateau, characterized by: The steps include: S1. Multi-source data acquisition: Satellite remote sensing technology is used to obtain regional topography and vegetation cover data, and a ground sensor network is used to collect permafrost temperature, humidity, displacement, groundwater level and meteorological data; S2. Data preprocessing: cleaning, filtering, and normalizing the collected data; S3, feature extraction: extracting characteristic parameters related to landslide hazards from the preprocessed data; S4. Risk assessment: Build a landslide risk assessment model based on machine learning algorithms and evaluate the landslide risk level based on characteristic parameters; S5. Warning release: When the risk level reaches the set threshold, warning information will be released through multiple channels; S6. Dynamic update: Continuously collect data and regularly update risk assessment models.
2. The method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to claim 1, characterized in that: In the feature extraction step S3, the thickness change rate V of the frozen soil active layer is calculated. ALT The formula is V ALT =(ALT t -ALT t-1 ) / Δ t， Among them ALT t ALT is the thickness of the frozen soil active layer at the current moment. t-1 is the thickness of the frozen soil active layer at the previous moment, Δ t is the time interval.
3. The method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to claim 1, characterized in that: In the multi-source data collection step S1, satellite remote sensing technology adopts a combination of synthetic aperture radar and optical remote sensing, and the ground sensor network includes displacement sensors, strain sensors, temperature and humidity sensors, and groundwater level sensors.
4. The method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to claim 1, characterized in that: In the data preprocessing step S2, median filtering is used to remove noise in the data, and the minimum-maximum normalization method is used to normalize the data to the interval [0, 1].
5. The method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to claim 1, characterized in that: The feature extraction in step S3 also includes extracting displacement rate, groundwater level change rate and temperature change rate as feature parameters.
6. The method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to claim 1, characterized in that: In the risk assessment of step S4, the machine learning algorithm adopts the random forest algorithm, uses historical landslide data as training samples, characteristic parameters as input, and landslide risk level as output for model training.
7. The method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to claim 1, characterized in that: In the step S5 of issuing the warning, the warning information includes the landslide risk level, the time and location of possible occurrence, and suggestions for response measures. The release channels include text messages, broadcasts, online platforms, and on-site alarm devices.
8. The method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to claim 1, characterized in that: In the dynamic update of step S6, new data is collected at regular intervals and the risk assessment model is updated using an incremental learning method.
9. The method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to claim 1, characterized in that: In the multi-source data collection step S1, a data quality monitoring unit is also set up to monitor the accuracy and stability of sensor data in real time, and to perform maintenance or replace sensors in a timely manner when data is abnormal.
10. The method for monitoring and early warning of landslide disasters in permafrost areas of the Qinghai-Tibet Plateau according to claim 1, characterized in that: In the risk assessment of step S4, landslide risks are classified into different levels according to human activities in the area, and corresponding risk thresholds are set for each level.