Glacier jump prediction method and system based on intelligent algorithm, and terminal

Abstract

The invention relates to the technical field of glacier monitoring, in particular to a glacier jump prediction method, system and terminal based on an intelligent algorithm, and the method comprises the steps: data collection: collecting multi-source data related to glacier jump through a sensor and a remote sensing technology, the meteorological data comprises temperature, humidity, air pressure and water content in ice, and the glacier movement data comprises flow velocity, ice layer thickness and displacement; according to the method, a simple, convenient and efficient real-time prediction scheme is constructed by adopting a multi-source data and deep learning technology for medium and small-scale glaciers, so that the problem that glacier data cannot be predicted in real time in a multi-scale physical constraint method adopted by existing glacier research is solved, and the prediction precision and timeliness are ensured; meanwhile, a simple and universal model is constructed, and the process and engineering automation deployment are simplified; in addition, disaster early warning is supported in result visualization, and the method is easy to embed into a GIS (Geographic Information System).

Description

Technical Field

[0001] The present invention relates to the technical field of glacier monitoring, and in particular to a glacier surge prediction method, system and terminal based on an intelligent algorithm. Background Art

[0002] Glaciers are an important indicator of climate change, and their surge (rapid movement caused by meltwater, gravity, etc.) is increasingly threatening ecological and human safety. Traditional glacier monitoring relies on ground-based measurements and artificial remote sensing, such as: Chinese invention patent CN119357734A relates to a method for identifying and early warning of glacier surges, which includes the following steps: Step 1: Acquire a glacier boundary database; Step 2: Revise the glacier mainstream line and convert it into point vector data; Step 3: Construct glacier surface velocity time series data; Step 4: Interpolate missing values; Step 5: Sensitivity study; Step 6: Calculate the F1 score and determine the optimal parameter combination; Step 7: Automatically identify surged glaciers; Step 8: Obtain the surge glacier velocity characteristic threshold for early warning; The present invention can quickly, automatically and uniformly identify surged glaciers, but it only analyzes the velocity time series.

[0003] Chinese invention patent CN117952010B discloses a glacier surge dynamic potential perception system and storage medium based on data engineering. The system includes a data acquisition unit, a data preprocessing unit, a model matching unit, a dynamic monitoring unit, an evaluation and analysis unit, and a collaborative analysis unit. It realizes cascade analysis of multi-regional data features, supports the discovery of hidden correlation trends of multiple factors in the cryosphere and the characterization of the laws of multi-process coupling, and improves the accuracy of glacier surge prediction; however, it has the problems of multi-source monitoring without prediction function, poor timeliness and weak prediction ability.

[0004] Chinese invention patent CN119397262A discloses a method for glacier surge prediction and data sample enhancement management, which is used to solve the problem of accurately identifying and processing the local albedo anomaly effects of surface moraine-covered glaciers. It achieves accurate identification and processing of the local albedo anomaly effects of surface moraine-covered glaciers by comprehensively analyzing the coupling relationship between spectrum and temperature and the spatial sensitivity of temperature gradients. However, although its spectral analysis can improve data accuracy, it cannot provide real-time warning of surge risks.

[0005] Chinese invention patent CN119027830A discloses a glacier identification method based on remote sensing images. The method obtains multi-source remote sensing data with time series characteristics based on the target area, fuses and optimizes the high-resolution remote sensing image and multispectral data based on the digital elevation data to obtain a fused image; constructs a glacier identification cycle model based on the BoTNet network architecture; inputs the fused image into the glacier identification cycle model to obtain a glacier identification result; and achieves efficient and accurate identification of glaciers. However, although its static identification improves the accuracy of the data, it still has the problem of being unable to provide real-time warning of the risk of glacier oscillation.

[0006] While intelligent algorithms are making a name for themselves in glacier research, existing technologies struggle to meet engineering requirements. Compared to multi-scale, physical-constrained approaches that focus on long-term ice sheet evolution, this paper proposes a real-time prediction technology for small and medium-scale glaciers based on multi-source data (meteorological and motion) and deep learning (CNN + LSTM). This technology aims to overcome time lag and accuracy bottlenecks, providing an efficient and easily implemented solution for disaster prevention and control. Summary of the Invention

[0007] The present invention provides a glacier surge prediction method, system and terminal based on an intelligent algorithm, which overcomes the shortcomings of the above-mentioned existing technologies and can effectively solve the problem that the multi-scale physical constraint method used in existing glacier research cannot predict glacier data in real time.

[0008] To solve the above problems, one of the technical solutions of the present invention is achieved in the following way: a glacier surge prediction method based on an intelligent algorithm, characterized by comprising the following steps: Data collection: Multi-source data related to glacier surges is collected through sensors and remote sensing technology. The multi-source data includes meteorological data (temperature, humidity, air pressure, and ice moisture content) and glacier motion data (flow velocity, ice thickness, and displacement). Data preprocessing: denoising, standardization and missing value filling of collected multi-source data; Feature extraction: Convolutional neural networks and long short-term memory networks are used to extract spatiotemporal features from the preprocessed data to capture the spatial correlation and temporal dependence of glacier surges. Model construction and training: Build a dynamic prediction model based on the extracted spatiotemporal features, and optimize model parameters through historical data training; Real-time prediction and evaluation: Use the trained model to make real-time predictions on new data, output jump probabilities and trends, and evaluate prediction accuracy using root mean square error. Result visualization: The prediction results are converted into a jump probability distribution map and a future trend prediction curve, which are displayed through the user interface to support disaster warning decision-making.

[0009] The equipment used in the above data collection includes temperature sensors, humidity sensors, air pressure sensors, conductivity sensors, drones, satellite remote sensing equipment and ground observation stations. The collected data is transmitted in real time to the data preprocessing step for data processing to ensure the timeliness of the data.

[0010] The above data preprocessing includes the following steps: Use filtering or statistical methods to denoise the collected multi-source data to remove environmental interference and outliers; Use standardization technology to scale the denoised data features to a uniform scale to adapt to the model input; The missing values ​​of the standardized data are filled using the mean or interpolation method to maintain data integrity.

[0011] In the above feature extraction, the convolutional neural network (CNN) extracts spatial features, and the long short-term memory network (LSTM) analyzes time series features. The two are combined to form a spatiotemporal feature representation of glacier surges.

[0012] In the above model construction and training, an adaptive optimization algorithm is used to optimize the model parameters, where the adaptive optimization algorithm includes stochastic gradient descent (SGD) or Adam optimizer.

[0013] The above-mentioned real-time prediction and evaluation includes real-time display of prediction results through an interactive interface and supports users to dynamically adjust the prediction period, where the prediction period is at least 1 day to 1 year.

[0014] The above-mentioned results are visualized, including generating a jump probability distribution map to mark high-risk areas, and generating a future trend prediction curve to show the jump change trend, wherein the visualization results are output through an automated system; or / and, the disaster warning decision-making includes glacier disaster warning methods and glacier disaster management methods, wherein glacier disaster warning includes warning information generation, warning level classification, information transmission and visualization, real-time and dynamic adjustment; glacier disaster management methods include emergency response, monitoring and adjustment, regional control and long-term planning.

[0015] The second technical solution of the present invention is achieved by the following method: a glacier surge prediction system based on an intelligent algorithm, comprising: A data acquisition unit collects multi-source data related to glacier surges through sensors and remote sensing technology, wherein the multi-source data includes meteorological data and glacier movement data. The meteorological data includes temperature, humidity, air pressure, and ice water content, and the glacier movement data includes flow velocity, ice thickness, and displacement. Data preprocessing unit, which performs denoising, standardization, and missing value filling on the collected multi-source data to improve data quality and consistency; The feature extraction unit uses convolutional neural networks and long short-term memory networks to extract spatiotemporal features from the preprocessed data to capture the spatial correlation and temporal dependence of glacier surges; The model building and training unit builds a jump prediction model based on the extracted spatiotemporal features and optimizes the model parameters through historical data training; The real-time prediction and evaluation unit uses the trained model to make real-time predictions on new data, outputs jump probabilities and trends, and evaluates prediction accuracy using root mean square error. The result visualization unit converts the prediction results into a jump probability distribution map and a future trend prediction curve, which are displayed through the user interface to support disaster warning decision-making.

[0016] The data preprocessing unit includes a denoising module, a standardization module and a missing value filling module, wherein: The denoising module uses filtering or statistical methods to denoise the collected multi-source data to remove environmental interference and outliers; The normalization module uses normalization technology to scale the denoised data features to a uniform scale to adapt to the model input; The missing value filling module uses the mean or interpolation method to fill the missing values ​​of the standardized data to maintain data integrity.

[0017] The third technical solution of the present invention is achieved in the following way: a terminal, including a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the program includes instructions for executing steps in a glacier surge prediction method based on an intelligent algorithm.

[0018] Compared with the prior art, the present invention has the following advantages: This paper targets small and medium-scale glaciers and is implemented in Python. It uses multi-source data and deep learning technology to build a simple and efficient real-time prediction solution, which solves the problem that the multi-scale physical constraint method used in existing glacier research cannot predict glacier data in real time, ensuring the accuracy and timeliness of the prediction. At the same time, the present invention constructs a simple and universal model, simple process and engineering automation deployment. In addition, the visualization of the results of the present invention supports disaster warning and can be easily embedded in the GIS system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This is a flow chart of the glacier surge prediction method based on intelligent algorithm in Example 1 of the present invention.

[0021] Figure 2This is a flow chart of the glacier surge prediction method in Example 1 of the present invention.

[0022] Figure 3 This is a flow chart of the method for data collection and processing in Example 1 of the present invention.

[0023] Figure 4 This is the architecture diagram of the intelligent algorithm model in Example 1 of the present invention.

[0024] Figure 5 This is a block diagram of a glacier surge prediction system based on an intelligent algorithm according to embodiment 2 of the present invention.

[0025] Figure 6 This is a visualization diagram of the results in Example 3 of the present invention. DETAILED DESCRIPTION

[0026] The present invention is not limited to the following embodiments, and specific implementation methods can be determined based on the technical solutions of the present invention and actual conditions.

[0027] Example 1: Figure 1 、 2 As shown, the embodiment of the present invention discloses a glacier surge prediction method based on an intelligent algorithm, comprising the following steps: S101, Data Collection: Collect multi-source data related to glacier surges through sensors and remote sensing technology, where the multi-source data includes meteorological data and glacier movement data. The meteorological data includes temperature, humidity, air pressure, and ice water content, and the glacier movement data includes flow velocity, ice thickness, and displacement. S102, data preprocessing: denoising, standardizing and filling missing values ​​of collected multi-source data to improve data quality and consistency; S103, Feature Extraction: Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks are used to extract spatiotemporal features from the preprocessed data to capture the spatial correlation and temporal dependence of glacier surges; S104, model construction and training: constructing a jump prediction model based on the extracted spatiotemporal features, and optimizing the model parameters through historical data training; S105, Real-time Prediction and Evaluation: Use the trained model to make real-time predictions on new data, output jump probabilities and trends, and evaluate prediction accuracy using root mean square error (RMSE); S106, result visualization: convert the prediction results into a jump probability distribution diagram and a future trend prediction curve, and display them through the user interface to support disaster warning decision-making.

[0028] like Figure 3As shown, in step S101, the equipment used for data collection includes temperature sensors, humidity sensors, air pressure sensors, conductivity sensors, drones, satellite remote sensing equipment, and ground observation stations. The collected data is transmitted in real time to the data preprocessing step for data processing to ensure the timeliness of the data. The temperature sensor is used to collect temperature data; the humidity sensor is used to collect humidity data; the air pressure sensor is used to collect air pressure data; the conductivity sensor is used to collect ice water content; the drone is used to collect glacier cracks and surface structure, glacier area and boundary; the satellite remote sensing equipment is used to collect glacier surface elevation, glacier mass balance, and ice flow rate; and the ground observation station is used to observe ground ice thickness, precipitation, and snow depth. The data observed by the ground observation station can be obtained by manual on-site measurement. Thus, multi-source data is collected in real time through sensors and remote sensing equipment to ensure dynamic monitoring capabilities.

[0029] like Figure 3 As shown, in the above step S102, data preprocessing includes the following steps: Use filtering or statistical methods to denoise the collected multi-source data to remove environmental interference and outliers; Use standardization technology to scale the denoised data features to a uniform scale to adapt to the model input; Missing values ​​in the standardized data are filled using the mean or interpolation method to maintain data integrity. Thus, data preprocessing through denoising, standardization and filling processes can improve data quality and consistency.

[0030] like Figure 4 As shown, in the above step S103, in the feature extraction, the convolutional neural network (CNN) extracts spatial features, wherein the spatial features include feature maps, activation values, and dimensionality reduction feature representations; the long short-term memory network (LSTM) analyzes time series features, wherein the time series features include state outputs and sequence outputs, and the spatial features and time series features are combined to form a spatiotemporal feature representation of glacier surges.

[0031] like Figure 4 As shown, in the above step S104, during model construction and training, an adaptive optimization algorithm is used to optimize model parameters, wherein the adaptive optimization algorithm includes stochastic gradient descent (SGD) or Adam optimizer to improve training efficiency and prediction accuracy; wherein model construction includes constructing the input layer, hidden layer and output layer of the model; model training includes data segmentation, forward propagation, loss calculation and back propagation.

[0032] like Figure 4As shown, in step S105, real-time prediction and evaluation includes displaying prediction results in real time through an interactive interface, allowing users to dynamically adjust the prediction period, where the prediction period ranges from at least one day to one year. Thus, the accuracy of the model prediction is evaluated by outputting real-time jump probabilities and trends. Model evaluation includes verification and testing, as well as performance indicator analysis.

[0033] like Figure 4 As shown, in the above step S106, the result visualization includes generating a jump probability distribution map to mark high-risk areas, and generating a future trend prediction curve to show the jump change trend, wherein the visualization result is output through an automated system to support quick decision-making.

[0034] In the above step S106, the disaster warning decision includes glacier disaster warning methods and glacier disaster management methods. Among them, glacier disaster warning includes warning information generation, warning level classification, information transmission and visualization, real-time and dynamic adjustment; glacier disaster management methods include emergency response, monitoring and adjustment, regional control and long-term planning.

[0035] Warning information generation involves obtaining the jump probability and future trend forecast curve output by the model to determine the jump probability. Warning levels are categorized based on the jump probability: Low-level warning (probability <50%): routine monitoring recommended; Intermediate warning (probability 50%-70%): increased monitoring frequency and preparation of emergency supplies; High-level warning (probability >70%): immediate evacuation or restriction of high-risk areas. Information transmission and visualization involves displaying the warning level through the user interface and pushing warning information via text messages and dedicated applications. Real-time and dynamic adjustment involves adjusting the forecast period as needed.

[0036] Emergency responses include: Low-level alerts: routine inspections and updated public risk warnings; Intermediate alerts: deployment of temporary conductivity sensors and notification to downstream communities to prepare supplies; High-level alerts: evacuation of downstream residents, restriction of high-risk areas, and construction of temporary flood control dikes. Monitoring adjustments involve adjusting monitoring frequency based on emergency response. Regional control involves demarcating high-risk areas based on probability distribution maps, installing warning signs, restricting tourism activities, and publishing dynamic risk maps. Long-term planning involves predicting flow velocity data within a year based on trend curves and planning permanent flood control projects or resident relocation.

[0037] Example 2: Figure 5 As shown, the embodiment of the present invention discloses a glacier surge prediction system based on an intelligent algorithm, comprising: A data acquisition unit collects multi-source data related to glacier surges through sensors and remote sensing technology, wherein the multi-source data includes meteorological data and glacier movement data. The meteorological data includes temperature, humidity, air pressure, and ice water content, and the glacier movement data includes flow velocity, ice thickness, and displacement. Data preprocessing unit, which performs denoising, standardization, and missing value filling on the collected multi-source data to improve data quality and consistency; The feature extraction unit uses convolutional neural networks and long short-term memory networks to extract spatiotemporal features from the preprocessed data to capture the spatial correlation and temporal dependence of glacier surges; The model building and training unit builds a jump prediction model based on the extracted spatiotemporal features and optimizes the model parameters through historical data training; The real-time prediction and evaluation unit uses the trained model to make real-time predictions on new data, outputs jump probabilities and trends, and evaluates prediction accuracy using root mean square error. The result visualization unit converts the prediction results into a jump probability distribution map and a future trend prediction curve, which are displayed through the user interface to support disaster warning decision-making.

[0038] The data acquisition unit includes temperature sensors, humidity sensors, air pressure sensors, conductivity sensors, drones, satellite remote sensing equipment, and ground observation stations. The collected data is transmitted in real time to the data preprocessing step for data processing to ensure the timeliness of the data. The temperature sensor is used to collect temperature data; the humidity sensor is used to collect humidity data; the air pressure sensor is used to collect air pressure data; the conductivity sensor is used to collect ice water content; the drone is used to collect glacier cracks and surface structure, glacier area and boundaries; the satellite remote sensing equipment is used to collect glacier surface elevation, glacier mass balance, and ice flow rate; and the ground observation station is used to observe ground ice thickness, precipitation, and snow depth. The data observed by the ground observation station can be obtained by manual on-site measurement. Thus, through the real-time collection of multi-source data by sensors and remote sensing equipment, dynamic monitoring capabilities are ensured.

[0039] The data preprocessing unit includes a denoising module, a standardization module and a missing value filling module, wherein: The denoising module uses filtering or statistical methods to denoise the collected multi-source data to remove environmental interference and outliers; The normalization module uses normalization technology to scale the denoised data features to a uniform scale to adapt to the model input; The missing value filling module uses the mean or interpolation method to fill the missing values ​​of the standardized data to maintain data integrity.

[0040] Among them, the feature extraction unit includes a convolutional neural network (CNN) extraction module and a long short-term memory network (LSTM) analysis module. The convolutional neural network (CNN) extraction module extracts spatial features, wherein the spatial features include feature maps, activation values, and dimensionality reduction feature representations; the long short-term memory network (LSTM) analysis module analyzes time series features, wherein the time series features include state outputs and sequence outputs. The spatial features and time series features are combined to form the spatiotemporal feature representation of glacier surges.

[0041] Among them, the model construction and training unit uses an adaptive optimization algorithm to optimize the model parameters, among which the adaptive optimization algorithm includes stochastic gradient descent (SGD) or Adam optimizer to improve training efficiency and prediction accuracy; the model construction includes constructing the input layer, hidden layer and output layer of the model; the model training includes data segmentation, forward propagation, loss calculation and back propagation.

[0042] The real-time prediction and evaluation unit displays prediction results in real time through an interactive interface, allowing users to dynamically adjust the prediction period, which ranges from at least one day to one year. This outputs real-time jump probabilities and trends to evaluate the accuracy of model predictions. Model evaluation includes verification and testing, as well as performance indicator analysis.

[0043] Among them, the result visualization unit includes generating a jump probability distribution map to mark high-risk areas, and generating a future trend prediction curve to show the jump change trend. Among them, the visualization results are output through an automated system to support quick decision-making.

[0044] Among them, disaster warning decisions include glacier disaster warning methods and glacier disaster management methods. Among them, glacier disaster warning includes warning information generation, warning level division, information transmission and visualization, real-time and dynamic adjustment; glacier disaster management methods include emergency response, monitoring and adjustment, regional control and long-term planning.

[0045] From the above, it can be seen that the glacier surge prediction method based on the intelligent algorithm in Example 1 of the present invention can be implemented based on the above-mentioned system. The method data is embedded in the system of Example 2 through the real-time data stream interface to realize the full process automation of data collection, real-time prediction and result visualization, and support docking with the geographic information system (GIS) to generate disaster warning information including high-risk area markings for glacier disaster warning and management.

[0046] In addition, in an embodiment of the present invention, the prediction model is deployed on a device that supports high-performance computing and runs through cloud computing or a local server to achieve parallel processing of multi-source data and real-time dynamic prediction, ensuring the computational efficiency and low latency of the feature extraction and prediction process.

[0047] Example 3: Prediction of a high mountain glacier surge This example takes a mountain glacier (with a total area of ​​approximately 50 km²) as an example to describe in detail the implementation process of the glacier surge prediction method based on intelligent algorithms, covering data collection, preprocessing, feature extraction, model construction, prediction, visualization, early warning, and management steps. Figure 6 The output requirements include the jump probability distribution map (5m resolution), future trend prediction curve (7-day prediction), user interface and GIS docking.

[0048] 1. Data Collection As attached Figure 3 As shown, the following equipment is used to collect multi-source data in real time, covering meteorological data and glacier movement data, supporting Figure 6 Probability distribution diagram and trend prediction curve generation: 1) Meteorological Data: Temperature sensor (-50°C to 50°C, accuracy ±0.1°C): collects daily temperature. Humidity sensor (0-100%RH, accuracy ±2%): collects daily relative humidity. Pressure sensor (300-1100 hPa, accuracy ±1 hPa): collects daily air pressure. SG-410 gravimetric rain and snow gauge (0-1500 mm, equipped with heating function, accuracy ±0.01 mm): collects daily precipitation (including solid and liquid precipitation). Frequency: Data is collected daily and transmitted to a local server via 4G / satellite networks with a latency of <1 minute. The above meteorological data collection equipment was used to continuously collect relevant data for 7 days, resulting in the data table shown in Table 1.

[0049] 2) Glacier movement data: A. Ground Sensors: Conductivity sensor (0-10 mS / cm, accuracy ±0.01 mS / cm): measures water content in ice. ‌LTD-ID1 ice-penetrating radar (absolute error approximately 5%): measures ice thickness.

[0050] B. Satellite Remote Sensing: Sentinel-1 SAR (C-band, 5m resolution): Measure ice flow velocity and displacement using InSAR, with an accuracy of 0.1 m / day. ICESat-2 Laser Altimeter (532nm, 70m resolution): Measure glacier surface elevation changes with an accuracy of 10cm. Sentinel-2 Optical Imagery (10m resolution): Generate area and boundary data.

[0051] C. Drone: DJI Zenmuse P1 camera (45MP, 12mm lens, 84° FOV, -20°C to 50°C): Flying at an altitude of 50m and a speed of 6m / s, it produces 2 cm / pixel crevasse images and 5 cm / pixel orthophotos with 85% overlap. MicaSense RedEdge-MX (5-band, 1280x960): Distinguishes glacier areas from rock boundaries. FLIR Vue Pro R (640x480, 7-14μm): Detects meltwater in glacier crevasses with a temperature accuracy of ±0.1°C.

[0052] D. Data example: Spatial data (100m×100m grid, 5m resolution, support Figure 6 Probability distribution map: Flow velocity: 1.2-1.8 m / day, with a high velocity area of ​​1.8 m / day in the northeast corner (20% of the grid), and a low velocity area of ​​1.2 m / day in the southwest corner (50% of the grid). Cracks: 10-50 cm wide, 1-10 m deep, with a dense area (northeast corner) occupying 20% ​​of the grid, averaging 20 cm wide and 5 m deep. In-ice water content: 0.5-2 mS / cm, with a peak of 2 mS / cm in the northeast corner (high-risk area). Area: 50 km², with boundary error <1% (Sentinel-2 orthophoto). Elevation change: Average annual drop of 0.3 m, accuracy ±10 cm (ICESat-2).

[0053] Time series (7 days, support Figure 6 Trend curve): Flow rate: 1.2 m / day on day 1, 1.3 m / day on day 2, 1.8 m / day on day 3, 1.7 m / day on day 4, 1.6 m / day on day 5, 1.5 m / day on day 6, and 1.5 m / day on day 7. Displacement: Cumulative 7 m, accuracy ±0.1 m (Sentinel-1 InSAR). Meltwater temperature: Anomaly +1.5°C in the northeast corner fracture zone (FLIR Vue ProR). Transmission: Data is transmitted to a local server in real time via 4G / satellite networks (latency <1 minute).

[0054] Data processing: Pix4DMapper was used to generate 3D fracture models (depth error <5%) and orthophotos, SNAP was used to process Sentinel-1 InSAR data, and GEE was used to analyze Sentinel-2 imagery. Data volume: Approximately 1,000 grid points per day (5m resolution, 100m × 100m), representing a 7-day time series (168 hours).

[0055] Therefore, compared with the existing single data collection method, the present invention integrates sensor, satellite and drone data, improves collection diversity and efficiency, and supports "multi-source data".

[0056] 2. Data Preprocessing As attached Figure 2 With attached Figure 3 As shown, the processing steps are as follows: Denoising: Low-pass filtering (Python scipy library) was used to remove noise from the flow velocity and temperature data, improving the signal-to-noise ratio by 20%. The IQR method (upper and lower bounds Q1-1.5IQR to Q3+1.5IQR) was used to remove outliers (e.g., flow velocity > 3 m / day).

[0057] Standardization: Z-score standardization (mean 0, standard deviation 1) unifies the dimensions of data such as flow rate and temperature, making it suitable for CNN and LSTM inputs. Compared to existing simple data fusion methods, this method significantly improves data quality and consistency through denoising, padding, and standardization.

[0058] Filling: Missing values ​​(about 10% of elevation data) are filled with the mean (pandas library), with a filling error of <1%.

[0059] 3. Feature Extraction As attached Figure 4 As shown, CNN and LSTM are used to extract features: CNN: Input imagery (5m resolution, 64×64 pixels) and meteorological data (temperature, humidity, etc.) are used. A three-layer convolution (kernel 3×3, ReLU activation) is used to extract spatial features (such as flow velocity distribution and crack location). Classification accuracy is 95% (compared with actual imagery).

[0060] LSTM: Input 7 days of time series data (daily flow rate, water content, etc., 3 features), 50 units per layer, captures time dependencies (such as flow rate acceleration trends), and has a prediction error RMSE of 2.5 m / year.

[0061] Output: Combining spatial features (velocity grid) and time series features (7-day trend) to form a spatiotemporal feature representation of glacier surges; compared to the existing static BoTNet model, this paper uses CNN and LSTM to achieve dynamic spatiotemporal analysis.

[0062] 4. Model construction and training As attached Figure 4 As shown, build a CNN-LSTM model: Architecture: CNN (32 filters) followed by LSTM (50 units), with a fully connected output layer predicting jump probabilities (0%-100%).

[0063] Training: Use TensorFlow, 70% training data (historical 7-day flow rate and meteorological data), Adam optimizer (learning rate 0.001), 50 iterations, and MSE loss converges to 0.01.

[0064] Performance: The prediction RMSE is 2.5 m / year, outperforming the traditional method (10 m / year). The proposed model is concise and versatile, making it easy to deploy in projects.

[0065] 5. Real-time prediction and evaluation As attached Figure 2 As shown, real-time prediction and evaluation: Forecast: Input the latest 24 hours of data (flow rate 1.2-1.8 m / day, temperature -8°C to -5°C) and output the 7-day jump probability (peak 65%).

[0066] Evaluation: Calculate RMSE = 0.05 (target < 0.1). If it exceeds the target, adjust the dropout rate (0.2).

[0067] Performance: The prediction accuracy is 95%, which is better than the traditional method of 80%.

[0068] 6. Results Visualization As attached Figure 6 As shown, the output includes: Probability distribution map: 5m resolution heat map (100m×100m), with high-risk areas (probability >60%, 65% in the northeast corner) marked in red, generated by Matplotlib.

[0069] Trend curve: 7-day forecast (20% on the first day, 65% on the third day, 40% on the seventh day), blue line, confidence interval ±5%, Flask interface supports CSV export.

[0070] User Interface: Flask display box refreshes in real time (hourly), includes "Refresh" and "Export CSV" buttons, and a time slider to adjust the forecast period (1 day to 1 year).

[0071] GIS Integration: Vector maps of high-risk areas are transferred to ArcGIS daily, with flood zones 5 km downstream marked. Data is stored on a local hard drive for easy access.

[0072] 7. Glacier disaster early warning methods Based on the real-time forecast in Part 5 (7-day probability 65%) and the probability distribution map in Part 6 (5m resolution, Figure 6 ), the present invention generates high-precision disaster warning information to support timely prevention of glacial disasters. The specific implementation steps are as follows: Warning information generation: The model outputs a 7-day jump probability (0%-100%) and a future trend forecast curve; it uses spatial features extracted by CNN (such as high-risk areas with a flow rate of 1.8 m / day) and time series features extracted by LSTM (such as the flow rate acceleration trend) to generate a jump probability distribution map (5m resolution, Figure 6 ).

[0073] Example: For a certain area (100m×100m), the probability of jump on the third day is predicted to be 65%, and the northeast corner grid (20% area) is marked as a high-risk area (red, Figure 6 heatmap).

[0074] Warning levels are divided into low-level warning (probability <50%): routine monitoring is recommended; medium-level warning (probability 50%-70%): increase monitoring frequency and prepare emergency supplies; high-level warning (probability >70%): immediate evacuation or restriction of high-risk areas.

[0075] Example: On the third day, the flow rate is 1.8 m / day with a probability of 65%, triggering a medium-level warning. If it reaches 80%, it will be upgraded to a high-level warning.

[0076] Information transmission and visualization: through the Flask interactive interface ( Figure 6 ) displays real-time probability distribution maps (heat maps, red highlights probability >60%) and trend curves (7-day, blue line, confidence interval ±5%). GIS integration (ArcGIS): Outputs vector maps, annotating the latitude and longitude of high-risk areas (e.g., flood zones 5 km downstream). Warning information is pushed via SMS and a dedicated app, with a response time of 2 hours.

[0077] Example: The high-risk area (65%) in the northeast corner is exported as a vector diagram and exported in CSV format ( Figure 6 button).

[0078] Real-time and dynamic adjustment: forecast period adjustment (1 day to 1 year), hourly data update (flow rate, temperature, etc.).

[0079] For example, on the third day, the flow rate suddenly increased to 1.8 m / day, and the system upgraded its warning from a medium to a high level within six hours. Compared to traditional thresholds or manual analysis (error >10%), this method reduces prediction error to less than 5% using CNN and LSTM, and GIS integration enhances the targeted nature of warnings.

[0080] 8. Glacier disaster management methods Based on the probability distribution diagram in Section 6 ( Figure 6 Heatmap) and trend curve ( Figure 6 The present invention provides a scientific basis for glacier disaster management through high-precision early warning information. The specific implementation steps are as follows: Emergency response includes: Low-level warning: routine inspections and updated public risk warnings; Intermediate-level warning: deployment of temporary conductivity sensors and notification to downstream communities to prepare supplies; High-level warning: evacuation of downstream residents, restriction of high-risk areas, and construction of temporary flood control dikes.

[0081] Example: On the third day, the probability is 65%, and a medium-level warning is triggered in the high-risk area in the northeast corner. The management department deploys drones for daily patrols.

[0082] Monitoring adjustments: Based on early warnings, adjust the monitoring frequency (e.g., increase drone patrols from once a week to once a day), and install ground sensors (e.g., GPS, with an accuracy of 0.1 m).

[0083] Example: A flow rate of 1.8 m / day triggers a medium-level alert, and a new sensor covers the northeast corner grid.

[0084] Regional control: GIS probability distribution map ( Figure 6 ) Designate high-risk areas (e.g. 10 km downstream), set up warning signs, restrict tourist activities, and publish dynamic risk maps.

[0085] Example: The northeast corner grid (65% probability) is marked as a flood zone and is under control for 7 days.

[0086] Long-term planning: The trend curve predicts a 15% acceleration in flow rate within one year, so permanent flood control projects or resident relocation should be planned.

[0087] For example, a 0.3 m / year drop in elevation indicates ice avalanche risk, optimizing water resource management. Compared to traditional manual experience or static data, this method improves response efficiency (2 hours) and targeted response through automated prediction and GIS analysis.

[0088] This example uses a mountain glacier as an example. Multi-source data (flow rate 1.2-1.8 m / day, crack width 10-50 cm) is collected daily. The model predicts a 65% probability of jump within seven days. A 5-meter resolution probability distribution map and a seven-day trend curve are generated, and management departments are notified to take protective measures (such as intermediate warnings) to verify practicality. Compared with traditional multi-scale physical constraint methods (with an error of 10-15%), this method solves the real-time prediction problem through the CNN-LSTM model implemented in Python (RMSE <5%) and multi-source data integration, ensuring accuracy and timeliness. It is easy to embed into GIS systems and glacier monitoring systems.

[0089] In summary, the present invention targets small and medium-scale glaciers and is implemented in Python. It uses multi-source data and deep learning technology to construct a simple and efficient real-time prediction solution, which solves the problem that the multi-scale physical constraint method used in existing glacier research cannot predict glacier data in real time, ensuring the accuracy and timeliness of the prediction. At the same time, the present invention constructs a simple and universal model, simple process and engineering automation deployment. In addition, the visualization of the results of the present invention supports disaster warning and can be easily embedded in the GIS system.

[0090] Example 4: An embodiment of the present invention discloses a terminal, including a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs include instructions for executing steps in a glacier surge prediction method based on an intelligent algorithm.

[0091] Example 5: An embodiment of the present invention discloses a storage medium, on which a computer program that can be read by a computer is stored. The computer program is configured to execute a glacier surge prediction method based on an intelligent algorithm when running.

[0092] The above storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory, a mobile hard disk, a magnetic disk, or an optical disk.

[0093] Example 6: An embodiment of the present invention discloses an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the computer program is loaded and executed by the processor to implement a glacier surge prediction method based on an intelligent algorithm.

[0094] The electronic device further includes a transmission device and an input / output device, wherein the transmission device and the input / output device are both connected to the processor.

[0095] The processor may be a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an ASIC, an FPGA, or other programmable logic device, a transistor logic device, a hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. It may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and so on. Memory may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memories, removable hard drives, magnetic disks, or optical disks.

[0096] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiment of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal translation scripting language JavaScript, etc.

[0097] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0098] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0099] The present application also provides a computer program product comprising a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a computer to perform some or all of the steps of any of the methods described in the above method embodiments. The computer program product may be a software installation package, and the computer may comprise an electronic device.

[0100] The above technical features constitute the best embodiment of the present invention, which has strong adaptability and best implementation effect. Non-essential technical features can be added or removed according to actual needs to meet the requirements of different situations.

[0101]

Claims

1. A glacier surge prediction method based on an intelligent algorithm, characterized in that: The following steps are involved: Data collection: Multi-source data related to glacier surges is collected through sensors and remote sensing technology. The multi-source data includes meteorological data (temperature, humidity, air pressure, and ice moisture content) and glacier motion data (flow velocity, ice thickness, and displacement). Data preprocessing: denoising, standardization and missing value filling of collected multi-source data; Feature extraction: Convolutional neural networks and long short-term memory networks are used to extract spatiotemporal features from the preprocessed data to capture the spatial correlation and temporal dependence of glacier surges. Model construction and training: Build a dynamic prediction model based on the extracted spatiotemporal features, and optimize model parameters through historical data training; Real-time prediction and evaluation: Use the trained model to make real-time predictions on new data, output jump probabilities and trends, and evaluate prediction accuracy using root mean square error. Result visualization: The prediction results are converted into a jump probability distribution map and a future trend prediction curve, which are displayed through the user interface to support disaster warning decision-making.

2. The method for predicting glacier surge based on an intelligent algorithm according to claim 1, characterized in that: The equipment used in the data collection includes temperature sensors, humidity sensors, air pressure sensors, conductivity sensors, drones, satellite remote sensing equipment and ground observation stations. The collected data is transmitted in real time to the data preprocessing step for data processing to ensure the timeliness of the data.

3. The method for predicting glacier surge based on an intelligent algorithm according to claim 1, characterized in that: The data preprocessing comprises the following steps: Use filtering or statistical methods to denoise the collected multi-source data to remove environmental interference and outliers; Use standardization technology to scale the denoised data features to a uniform scale to adapt to the model input; The missing values ​​of the standardized data are filled using the mean or interpolation method to maintain data integrity.

4. The method for predicting glacier surge based on an intelligent algorithm according to claim 1, characterized in that: In the feature extraction, the convolutional neural network extracts spatial features, and the long short-term memory network analyzes time series features. The two are combined to form the spatiotemporal feature representation of glacier surges.

5. The method for predicting glacier surge based on intelligent algorithm according to claim 1, characterized in that: During the model construction and training, an adaptive optimization algorithm is used to optimize the model parameters, wherein the adaptive optimization algorithm includes stochastic gradient descent or Adam optimizer.

6. The method for predicting glacier surge based on an intelligent algorithm according to claim 1, characterized in that: The real-time prediction and evaluation includes displaying the prediction results in real time through an interactive interface and supporting users to dynamically adjust the prediction period, wherein the prediction period includes at least 1 day to 1 year.

7. The method for predicting glacier surge based on an intelligent algorithm according to claim 1, characterized in that: The result visualization includes generating a jump probability distribution map to mark high-risk areas, and generating a future trend prediction curve to show the jump change trend, wherein the visualization results are output through an automated system; or / and, the disaster warning decision includes glacier disaster warning methods and glacier disaster management methods, wherein glacier disaster warning includes warning information generation, warning level classification, information transmission and visualization, real-time and dynamic adjustment; glacier disaster management methods include emergency response, monitoring and adjustment, regional control and long-term planning.

8. A glacier surge prediction system based on intelligent algorithms, characterized by: include: A data acquisition unit collects multi-source data related to glacier surges through sensors and remote sensing technology, wherein the multi-source data includes meteorological data and glacier movement data. The meteorological data includes temperature, humidity, air pressure, and ice water content, and the glacier movement data includes flow velocity, ice thickness, and displacement. Data preprocessing unit, which performs denoising, standardization, and missing value filling on the collected multi-source data to improve data quality and consistency; The feature extraction unit uses convolutional neural networks and long short-term memory networks to extract spatiotemporal features from the preprocessed data to capture the spatial correlation and temporal dependence of glacier surges; The model building and training unit builds a jump prediction model based on the extracted spatiotemporal features and optimizes the model parameters through historical data training; The real-time prediction and evaluation unit uses the trained model to make real-time predictions on new data, outputs jump probabilities and trends, and evaluates prediction accuracy using root mean square error. The result visualization unit converts the prediction results into a jump probability distribution map and a future trend prediction curve, which are displayed through the user interface to support disaster warning decision-making.

9. The glacier surge prediction system based on intelligent algorithm according to claim 8, characterized in that: The data preprocessing unit includes a denoising module, a standardization module and a missing value filling module, wherein: The denoising module uses filtering or statistical methods to denoise the collected multi-source data to remove environmental interference and outliers; The normalization module uses normalization technology to scale the denoised data features to a uniform scale to adapt to the model input; The missing value filling module uses the mean or interpolation method to fill the missing values ​​of the standardized data to maintain data integrity.

10. A terminal, characterized in that: It includes a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the processor, and the programs include instructions for executing the steps in a glacier surge prediction method based on an intelligent algorithm as described in any one of claims 1 to 7.

Images