Identifying and early warning method for glacier jump

Abstract

The invention relates to the technical field of glacier jump automatic identification, in particular to a glacier jump identification and early warning method, which comprises the following steps of: 1, acquiring a glacier boundary database; 2, revising a glacier main stream line and converting the glacier main stream line into point vector data; step 3, constructing glacier surface flow velocity time sequence data; step 4, interpolating a vacancy value; 5, sensitivity research; step 6, calculating an F1 score and determining an optimal parameter combination; 7, automatically identifying the jumping glacier; and step 8, acquiring the characteristic threshold value of the jumping glacier flow velocity for early warning. According to the method, the jumping glacier can be identified quickly, automatically and uniformly.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic identification of glacier surges, and in particular to a method for identification and early warning of glacier surges. Background Art

[0002] Glacier surge refers to the special phenomenon of glaciers moving rapidly periodically. It is a type of glacial disaster. When glacier surge occurs, it may destroy roads and houses, engulf pastures, and threaten the lives of livestock and residents. It is highly catastrophic. In addition, glacier surge may cause secondary disasters such as barrier lakes, glacial lake outbursts and glacial debris flows, further threatening the livelihood and infrastructure safety of downstream communities.

[0003] At present, the method of identifying surging glaciers usually relies on manual judgment, mainly by observing the characteristics of glaciers when they surge. These characteristics include changes in the flow velocity on the glacier surface, changes in elevation, and characteristics of the surface morphology. Usually, this process completes the identification of surging glaciers by establishing a set of multi-standard judgment systems, checking whether the glaciers have these characteristics in turn. However, this method requires a lot of data support, is time-consuming and labor-intensive, and is also limited by factors such as image spatial coverage and temporal resolution. In addition, since the setting of the judgment system is highly subjective, different systems may lead to different results, resulting in significant differences.

[0004] In recent years, due to the complexity of glacier dynamics and the impact of global climate change, there is an urgent need for fast, automatic and unified methods to identify surging glaciers. Summary of the invention

[0005] The present invention provides a method for identifying and early warning of glacier surges, which can identify surged glaciers in a fast, automatic and unified manner.

[0006] The method for identifying and early warning glacier surge according to the present invention comprises the following steps:

[0007] Step 1: Obtain glacier boundary database;

[0008] Step 2: Revise the main glacier line and convert it into point vector data;

[0009] Step 3: Construct glacier surface velocity time series data;

[0010] Step 4: interpolation of missing values;

[0011] Step 5: Sensitivity study;

[0012] Step 6: Calculate the F1 score and determine the best parameter combination;

[0013] Step 7: Automatically identify leaping glaciers;

[0014] Step 8: Obtain the characteristic threshold of surging glacier flow velocity for early warning.

[0015] Preferably, in step 1, the specific steps are as follows:

[0016] 1.1) Using the vector range data of the study area as a mask, all glaciers in the area were extracted from the RGI data;

[0017] 1.2) Use the attribute table to further select glaciers with an area larger than 5 square kilometers;

[0018] 1.3) Glaciers with an area greater than or equal to 5 square kilometers are retained to form a glacier dataset in the study area. Each glacier has a unique identification ID.

[0019] Preferably, in step 2, the specific steps are as follows:

[0020] 2.1) Select and extract the main stream lines of the glacier by location, and only need to keep the main stream lines inside the glacier boundary;

[0021] 2.2) Check the glacier mainstream line data step by step, delete the redundant lines, and manually draw the mainstream line for the glaciers that lack mainstream lines;

[0022] 2.3) Generate a point every 100 meters along the main stream of the glacier to extract the glacier surface velocity value; assign a serial number FID to each point, and use the identification tool to assign the points inside the same glacier to the ID of the upper glacier.

[0023] Preferably, in step 3, the specific steps are as follows:

[0024] 3.1) Extracting glacier surface velocity values ​​using mainstream line point vector data;

[0025] 3.2) Export the attribute table of mainstream line points into txt text format, and then import it into Excel to merge it into a matrix; the matrix contains the point identification FID, the glacier identification ID, the time step and the corresponding glacier surface velocity value; in the matrix, the rows store all extractions of each time step, and the columns store the same extraction points of each mainstream line of all time steps.

[0026] Preferably, in step 4, the specific steps are as follows:

[0027] 4.1) Read time series data containing multiple glacier velocity records from a file and group the data by the unique identifier of the glacier;

[0028] 4.2) Apply a linear interpolation method to each glacier’s velocity time series;

[0029] 4.3) After completing the interpolation, check the data to ensure that all missing values ​​have been filled.

[0030] As a preferred method, in step 5, a sensitivity study of the case glacier is conducted through outlier detection, generalized extreme value studentized deviation test GESD and cluster analysis. The specific steps are as follows:

[0031] 5.1) Perform time series decomposition;

[0032] The seasonal trend decomposition method is used to decompose the time series of the case glacier, and the glacier velocity time series of each point is decomposed into three components: trend, seasonality and residual.

[0033] 5.2) Perform outlier detection:

[0034] Extract the residual sequence of each point in turn and perform multiple groups of outlier detection on it; this part involves two parameters, namely α and the maximum anomaly allowed to be detected in the time series; α is the width of the envelope function on the residue curve, where data points outside the envelope are classified as outliers;

[0035] The values ​​of α are 0.05 and 0.1; the maximum anomaly values ​​allowed to be detected are 20%, 40%, 60%, 80%, and 100%, respectively. Multiple groups of experiments are conducted to obtain whether the flow velocity of the points on the mainstream line of the glacier is abnormal in all time series. If it is abnormal, the output is yes, and if it is not abnormal, the output is no.

[0036] 5.3) Perform cluster analysis;

[0037] For each time series, a moving window of size 20 is used to count the outliers in the window; the clustering threshold sizes are selected as 10, 12, 14, 16, and 18 for sensitivity research; if the sum of the values ​​of yes in the window reaches the set threshold, the month is marked as an abnormal month; if 4 or more consecutive months meet the abnormal criteria, they are marked as jump events; abnormal months with less than 4 months are marked as speed fluctuations; the rest are marked as normal months;

[0038] 5.4) Keep the relevant information of glacier time step and judgment of month.

[0039] Preferably, in step 6, the specific steps are as follows:

[0040] 6.1) Construct the start and end time and duration matrix of the real surge of the case glacier;

[0041] 6.2) Calculate the F1 score and construct a table for easy comparison;

[0042] 6.3) Comprehensively consider and determine the best performance parameter combination.

[0043] Preferably, in step 7, the specific steps are as follows:

[0044] 7.1) Perform time series decomposition on all glaciers;

[0045] The seasonal trend decomposition method is used to decompose the time series, and the glacier velocity time series of each point inside each glacier is decomposed into three components: trend, seasonality and residual.

[0046] 7.2) Extract the residual sequence of each glacier and perform outlier detection on it using the best parameter combination; obtain information on whether the flow velocity of all points on the mainstream line of all glaciers is abnormal in all time series, where abnormal output is yes and non-abnormal output is no; the information of points on the same mainstream line of the glacier is stored in the same table, including the time step, the point number FID and whether it is abnormal related information, in long format;

[0047] 7.3) Perform cluster analysis;

[0048] Each glacier is indexed by the time step. A moving window of size 20 is applied to the abnormality column to count the number of abnormal points in the window. If the number exceeds the threshold of 12, the corresponding time step is returned as an abnormal month. The loop is used to complete the judgment of all time steps;

[0049] 7.4) If 4 or more consecutive months meet the abnormal criteria, it is marked as a surge glacier.

[0050] Preferably, in step 8, the specific steps are as follows:

[0051] The maximum flow velocity of the surging glacier and the average flow velocity in normal months are calculated, and the difference between the two is used as the multiple of the increase in glacier velocity. The minimum value is used as the threshold for the flow velocity increase.

[0052] The present invention takes the generalized extreme value studentized deviation test (GESD) as the main method for outlier detection, establishes a set of methods for outlier detection from glacier surface velocity time series, extracts surge glaciers, and lays a methodological foundation for automatically and quickly obtaining surge glaciers. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A schematic diagram of a flow chart of a method for identifying and early warning of glacier surge in an embodiment;

[0054] Figure 2 This is a comparison diagram of different parameter combinations and the actual surge time range of the Khurdopin Glacier when α=0.05 in the embodiment;

[0055] Figure 3This is a comparison diagram of different parameter combinations and the actual surge time range of the Khurdopin Glacier when α=0.1 in the embodiment;

[0056] Figure 4 This is a comparison diagram of different parameter combinations and the actual surge time range of Shispare Glacier when α=0.05 in the embodiment;

[0057] Figure 5 This is a comparison diagram of different parameter combinations and the actual surge time range of Shispare Glacier when α=0.1 in the embodiment;

[0058] Figure 6 It is a Gantt chart of the onset and duration of surge events of 49 surge glaciers in the China-Pakistan Economic Corridor from 2016 to 2021 with IDs 203 to 545 in the embodiment;

[0059] Figure 7 This is a Gantt chart of the start and duration of surge events of 49 surge glaciers in the China-Pakistan Economic Corridor from 2016 to 2021 with IDs 0 to 202 in the embodiment. DETAILED DESCRIPTION

[0060] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.

[0061] Example

[0062] like Figure 1 As shown, this embodiment provides a method for identifying and early warning of glacier surge, which includes the following steps:

[0063] Step 1: Obtain the glacier boundary database.

[0064] Data: Vector extent of the study area and Randolph Glacier Inventory Data (RGI).

[0065] The specific steps are as follows:

[0066] 1.1) Using the vector range data of the study area as a mask, all glaciers in the area were extracted from the RGI data;

[0067] 1.2) Use the attribute table to further filter out glaciers with an area larger than 5 square kilometers; this processing step is intended to ensure that the glaciers studied are large enough for more meaningful analysis;

[0068] 1.3) Glaciers with an area greater than or equal to 5 square kilometers are retained to form a glacier dataset in the study area. Each glacier has a unique identification ID.

[0069] Step 2: Revise the main glacier line and convert it into point vector data.

[0070] Data: Glacier data set and glacier main stream data in the study area. The glacier main stream data comes from the National Glacier, Frozen Soil and Desert Science Data Center (NCDC). Since there are many errors in the data, manual revision is required. In addition, if a single glacier or glacier system has multiple center lines, only the main stream line of the glacier trunk, that is, the longest main stream line, is retained to identify whether a surge event has occurred, and tributary surges are not considered.

[0071] The specific steps are as follows:

[0072] 2.1) Select and extract the main stream lines of the glacier by location, and only need to keep the main stream lines inside the glacier boundary;

[0073] 2.2) Check the glacier mainstream line data step by step, delete the redundant lines, and manually draw the mainstream line for the glaciers that lack mainstream lines; this step needs to ensure that each glacier contains only one mainstream line;

[0074] 2.3) Generate a point every 100 meters along the main stream of the glacier to extract the glacier surface velocity value; assign a serial number FID to each point, and the FIDs are arranged in sequence inside the glacier; use the identification tool to assign the points inside the same glacier to the ID of the upper glacier, so as to clearly understand the subordinate relationship between the points and the glacier.

[0075] Step 3: Construct glacier surface velocity time series data.

[0076] Data: glacier main flow line point vector data and Sentinel-1 glacier surface velocity RETREAT data. Glacier surface velocity data must be continuous and have a complete cycle.

[0077] The specific steps are as follows:

[0078] 3.1) Extracting glacier surface velocity values ​​using mainstream line point vector data;

[0079] 3.2) Export the attribute table of mainstream line points into txt text format, and then import it into Excel to merge it into a matrix; the matrix contains the point identification FID, the glacier identification ID, the time step and the corresponding glacier surface velocity value; in the matrix, the rows store all extractions of each time step, and the columns store the same extraction points of each mainstream line of all time steps.

[0080] Step 4: Interpolation of missing values.

[0081] When processing the time series data of glacier surface velocity, the missing values ​​in the original data will have a significant impact on the analysis results. In order to ensure the smooth progress of the experiment, this embodiment uses a standard interpolation technique to fill these missing values ​​to ensure the continuity and integrity of the data.

[0082] The specific steps are as follows:

[0083] 4.1) Read the time series data containing multiple glacier velocity records from the file and group the data by the unique identifier of the glacier; in this way, the velocity data of each glacier can be processed independently;

[0084] 4.2) Apply linear interpolation to each glacier velocity time series; Linear interpolation is a simple and efficient technique that estimates missing data points and restores the continuity of the time series by assuming that the changes between known data points are linear;

[0085] 4.3) After completing the interpolation, check the data to ensure that all missing values ​​have been filled.

[0086] Step 5: Sensitivity study.

[0087] Outlier detection Generalized Extreme Studentized Deviation Test (GESD) and cluster analysis allow more or less degrees of freedom. GESD is a detection algorithm for multiple outliers in a single variable. It can control the type and error, and can accurately remove seasonality and trend from time series. It includes two modules: time series decomposition and outlier detection. In order to obtain the best algorithm parameter combination to ensure that the surge glacier and the start and end time of the surge event can be correctly detected, a sensitivity study is essential. Glaciers with known start and end times of surge events are obtained from published literature and used as cases for sensitivity studies.

[0088] The specific steps are as follows:

[0089] 5.1) Perform time series decomposition;

[0090] The seasonal trend decomposition method is used to decompose the time series of the case glacier, and the glacier velocity time series of each point is decomposed into three components: trend, seasonality and residual.

[0091] 5.2) Perform outlier detection;

[0092] Extract the residual sequence of each point in turn and perform multiple groups of outlier detection on it; in the GESD algorithm, this part involves two parameters, namely Alpha (α) and the maximum anomaly allowed to be detected in the time series, expressed as a percentage; α is the width of the envelope function on the residue curve, where data points outside the envelope are classified as outliers;

[0093] α controls the width of the "normal" range. It can be seen as a sensitivity regulator that sets the standard for an observation to be classified as an anomaly. The lower α is set, the more conservative the classification. Higher α values ​​are more likely to misclassify "normal" observations as anomalies. Therefore, α can be understood as the width of the envelope function on the residue curve, where data points outside the envelope are classified as outliers. Outliers refer to data points that are not classified as normal values ​​after seasonal and trend decomposition. The percentage represents the maximum number of outliers allowed to be detected for each time series. In order to find the best performance parameters, α is taken as 0.05 and 0.1; the maximum anomaly allowed to be detected is 20%, 40%, 60%, 80%, and 100%, respectively, and multiple groups of experiments are conducted. This step can obtain whether the flow velocity of the point on the mainstream line of the glacier is abnormal in all time series. The output is yes for abnormalities and no for non-abnormalities.

[0094] 5.3) Perform cluster analysis;

[0095] Cluster analysis was used to handle falsely detected outliers to ensure that velocity outliers occurred within the glacier boundary.

[0096] For each time series, a moving window of size 20 is used to count the outliers in the window; the clustering threshold sizes are selected as 10, 12, 14, 16, and 18 for sensitivity research; if the sum of the yes values ​​in the window reaches the set threshold, the month is marked as an abnormal month; if 4 or more consecutive months meet the abnormal criteria, they are marked as jump events; abnormal months with less than 4 months are marked as speed fluctuations; the rest are marked as normal months.

[0097] 5.4) Keep the relevant information of glacier time step and judgment of month.

[0098] Step 6: Calculate the F1 score and determine the best parameter combination;

[0099] The example glacier has 25 experiments under the same α, and α has two values, so a total of 50 experiments. In order to quantify the performance of each parameter combination, this example uses the F1 score, which is one of the key indicators for evaluating model performance. The F1 score combines the two basic indicators of precision and recall, providing a single metric that balances the two.

[0100] The specific steps are as follows:

[0101] 6.1) Construct the start and end time and duration matrix of the real surge of the case glacier;

[0102] 6.2) Calculate the F1 score and construct a table for easy comparison;

[0103] 6.3) Comprehensively consider and determine the best performance parameter combination. Sensitivity studies show that the combination of α = 0.05, maximum anomaly 100% and clustering threshold 12 is based on a comprehensive consideration of algorithm stability, processing power and simplicity.

[0104] Step 7: Automatically identify leaping glaciers;

[0105] The specific steps are as follows:

[0106] 7.1) Perform time series decomposition on all glaciers;

[0107] The seasonal trend decomposition method is used to decompose the time series, and the glacier velocity time series of each point inside each glacier is decomposed into three components: trend, seasonality and residual.

[0108] 7.2) Extract the residual sequence of each glacier and perform outlier detection on it using the best parameter combination; use α=0.05 and the maximum anomaly of 100% in the GESD algorithm and apply it to all glaciers; obtain information on whether the flow velocity of points on the mainstream line of all glaciers is abnormal in all time series, where the abnormal output is yes and the non-abnormal output is no; the information of points on the mainstream line of the same glacier is stored in the same table, including the time step, the point number FID and whether it is abnormal related information, in long format;

[0109] 7.3) Perform cluster analysis;

[0110] Each glacier is indexed by the time step. A moving window of size 20 is applied to the "Is it abnormal?" column to count the number of abnormal points in the window. If the number exceeds the threshold of 12, the corresponding time step is returned as an abnormal month. A loop is used to complete the judgment of all time steps.

[0111] 7.4) If 4 or more consecutive months meet the abnormal criteria, it is marked as a surge glacier.

[0112] Step 8: Obtain the characteristic threshold of the surging glacier flow velocity for early warning;

[0113] The sharp increase in glacier flow velocity is the most significant feature of glacier surge. The velocity characteristic threshold can be used for early threshold. The specific steps are as follows:

[0114] The maximum flow velocity of the surging glacier and the average flow velocity in normal months are calculated, and the difference between the two is used as the multiple of the increase in glacier velocity. The minimum value is used as the threshold for the flow velocity increase.

[0115] This example takes the area along the China-Pakistan Economic Corridor as an example. The area in the area is larger than 5km 2There are 559 glaciers in total. In the sensitivity study, this embodiment selects Khurdopin Glacier and Shispare Glacier as experimental objects.

[0116] like Figure 2 As shown in Figure 1, the comparison of different parameter combinations and the true surge time range of the Khurdopin Glacier when α=0.05. The maximum anomaly ratios are 20%, 40%, 60%, 80% and 100%, the window size is 20, and the thresholds are 10, 12, 14, 16 and 18 respectively. Figure 3 The figure shows the comparison of different parameter combinations and the real surge time range of Khurdopin Glacier when α = 0.1. Table 1 shows the F1 score table of different parameter combinations of Khurdopin Glacier under the condition of α = 0.05, and Table 2 shows the F1 score table of different parameter combinations of Khurdopin Glacier under the condition of α = 0.1.

[0117] Table 1 F1 scores of different parameter combinations for Khurdopin Glacier under the condition of α = 0.05

[0118]

[0119]

[0120] Table 2 F1 scores of different parameter combinations for Khurdopin Glacier under the condition of α = 0.1

[0121]

[0122] like Figure 4 As shown in Figure 1, the comparison of different parameter combinations and the actual surge time range of Shispare Glacier when α = 0.05 is shown in Figure 1. Figure 5 The figure shows the comparison of different parameter combinations and the real surge time range of Shispare Glacier when α = 0.1. Table 3 shows the F1 score table of different parameter combinations of Shispare Glacier under the condition of α = 0.05, and Table 4 shows the F1 score table of different parameter combinations of Shispare Glacier under the condition of α = 0.1.

[0123] Table 3 F1 scores of different parameter combinations for Shispare Glacier under the condition of α = 0.05

[0124]

[0125] Table 4 F1 scores of different parameter combinations for Shispare Glacier under the condition of α = 0.1

[0126]

[0127] After comprehensively analyzing the F1 score performance of the Khurdopin and Shispare glaciers, this embodiment selects a parameter combination of α = 0.05, a maximum anomaly of 100%, and a clustering threshold of 12. This selection is based on multiple factors. First, when the clustering threshold is 12 and α = 0.05, the Khurdopin glacier always maintains a high F1 score of 0.96 under high anomalies (40% to 100%), demonstrating its stability and consistency under large-scale data and extreme conditions. For the Shispare glacier, although the overall F1 score is low, when the maximum anomaly is 100%, the F1 score with a clustering threshold of 12 is significantly improved to 0.80, showing its excellent adaptability when dealing with a large amount of abnormal data. Selecting a maximum anomaly of 100% ensures the reliability of the model under extreme conditions, while a lower clustering threshold of 12 helps to simplify the model while maintaining high performance. α = 0.05 has more stringent requirements on model performance, which enables this parameter combination to provide consistent and efficient results under different data sets and conditions while controlling errors.

[0128] The results of automatic detection by the algorithm show that there were 49 surge glaciers in the region between 2016 and 2021. In order to conveniently describe the distribution of surge glaciers in the China-Pakistan Economic Corridor, it is divided into five sub-regions, including Karakoram, Pamir, Hindu Kush, Western Himalaya and Western Kunlun. The Karakoram sub-region has the largest number of glaciers and the largest area, with a total of 353 glaciers covering an area of ​​12,146.79 square kilometers. Among them, there are 34 surge glaciers, accounting for 9.63% of the total number of glaciers; the area is 6,386.61 square kilometers, accounting for 52.58% of the total glacier area. Although the proportion of the number is not high, the proportion of the glacier area is very high, exceeding half. It can be seen that the surge glaciers in this sub-region are mainly large glaciers, indicating that the surge glaciers have a particularly significant impact on the glacier dynamics and environment of the Karakoram. The surge of large glaciers not only changes the glacier's morphology and flow rate, but may also have a profound impact on the water supply and flood risks downstream. The total area of ​​101 glaciers in the Hindu Kush sub-district is 1505.62 square kilometers, of which there are 9 surging glaciers with an area of ​​317.83 square kilometers. The number of surging glaciers accounts for 8.91% and the area accounts for 21.11%, which shows that the existence of surging glaciers in the Hindu Kush region cannot be ignored. There are 66 glaciers in the Pamir sub-district of the China-Pakistan Economic Corridor with a total area of ​​1029.17 square kilometers, of which 4 are surging glaciers with a total area of ​​232.15 square kilometers. The number of surging glaciers in the Pamirs accounts for 6.06% and the area accounts for 22.56%, showing that surging glaciers also have a relatively important impact in the region. The number of glaciers in the West Himalaya sub-district and the West Kunlun sub-district are relatively small, with 18 and 21 glaciers, respectively, with a total area of ​​232.35 square kilometers and 163.94 square kilometers. The West Himalayas sub-region has one surge glacier with an area of ​​8.28 square kilometers, accounting for 5.56% of the total number of glaciers and 3.56% of the total glacier area. The West Kunlun sub-region also has one surge glacier with an area of ​​10.65 square kilometers, accounting for 4.76% of the total number of glaciers and 6.50% of the total glacier area.

[0129] By using the Gantt chart to visualize the time information of different surge glaciers, we can clearly understand the onset and duration of glacier surge, and further understand the temporal characteristics and movement patterns of surge glaciers in the China-Pakistan Economic Corridor. Figure 6 and Figure 7 Shown is a Gantt chart of the onset and duration of surge events of 49 surge glaciers in the China-Pakistan Economic Corridor from 2016 to 2021.

[0130] The start time of surge events showed obvious seasonal characteristics. 19 surge glaciers, accounting for 39%, first began to experience surge events in winter and early spring months (December to March); 9 surge glaciers first began to experience surge events in summer (June to August). In general, surges occur throughout the China-Pakistan Economic Corridor throughout the year, which is an expected behavior of thermally controlled surge glaciers. However, 39% of the surge glaciers still tend to start surging in winter or early spring, which is consistent with the start time of hydrologically controlled surge glaciers. In terms of a single continuous surge event, the surge event of glacier ID 197 (RGI60-14.04404) lasted the longest, lasting 24 months, from July 2016 to June 2018. The surge events of glaciers with IDs 517 (RGI60-13.41792) and 260 (RGI60-14.05446) lasted 20 months and 11 months, respectively. The surge events of the remaining glaciers lasted for a shorter period of time, with 20 glaciers lasting only four months at most. This shows that the duration of the active phase of the surge glaciers in the study area is not long. In addition, the figure below also reveals that velocity anomalies frequently occur before and after surge events. Velocity anomalies are usually a precursor or consequence of surge events, indicating that glaciers experience short-term velocity fluctuations before and after surge events.

[0131] The sharp increase in glacier flow velocity is the most significant feature of glacier surges. The velocity characteristic threshold can be used for early thresholds. In the experiment, all 49 surge glaciers experienced a significant increase in velocity, with the peak velocity ranging from 4 to 118 times the velocity in normal months. In this embodiment, 4 times is selected as the characteristic threshold. If a 4-fold velocity increase is detected in remote sensing data, it will be taken as a key monitoring object, and early warning can be achieved.

[0132] The present invention and its embodiments are described schematically above, and the description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. Therefore, if a person skilled in the art is inspired by it and designs a structural method and an embodiment similar to the technical solution without creativity without departing from the purpose of the invention, they shall all fall within the protection scope of the present invention.

Claims

1. A method for identifying and early warning glacier surges, characterized by: The following steps are involved: Step 1: Obtain glacier boundary database; Step 2: Revise the main glacier line and convert it into point vector data; Step 3: Construct glacier surface velocity time series data; Step 4: interpolation of missing values; Step 5: Sensitivity study; Step 6: Calculate the F1 score and determine the best parameter combination; Step 7: Automatically identify leaping glaciers; Step 8: Obtain the characteristic threshold of surging glacier flow velocity for early warning.

2. The method for identifying and early warning glacier surge according to claim 1, characterized in that: In step 1, the specific steps are as follows: 1.1) Using the vector range data of the study area as a mask, all glaciers in the area were extracted from the RGI data; 1.2) Use the attribute table to further select glaciers with an area larger than 5 square kilometers; 1.3) Glaciers with an area greater than or equal to 5 square kilometers are retained to form a glacier dataset in the study area. Each glacier has a unique identification ID.

3. The method for identifying and early warning glacier surge according to claim 2, characterized in that: In step 2, the specific steps are as follows: 2.1) Select and extract the main stream lines of the glacier by location, and only need to keep the main stream lines inside the glacier boundary; 2.2) Check the glacier mainstream line data step by step, delete the redundant lines, and manually draw the mainstream line for the glaciers that lack mainstream lines; 2.3) Generate a point every 100 meters along the main stream of the glacier to extract the glacier surface velocity value; assign a serial number FID to each point, and use the identification tool to assign the points inside the same glacier to the ID of the upper glacier.

4. The method for identifying and early warning glacier surge according to claim 3 is characterized in that: In step 3, the specific steps are as follows: 3.1) Extracting glacier surface velocity values ​​using mainstream line point vector data; 3.2) Export the attribute table of the mainstream line points into txt format, and then import it into Excel to merge it into a matrix; the matrix contains the point identification FID, the glacier identification ID, the time step and the corresponding glacier surface velocity value; In the matrix, rows store all extractions at each time step, and columns store the same extraction points for each main flow line at all time steps.

5. The method for identifying and early warning glacier surge according to claim 4, characterized in that: In step 4, the specific steps are as follows: 4.1) Read time series data containing multiple glacier velocity records from a file and group the data by the unique identifier of the glacier; 4.2) Apply a linear interpolation method to each glacier’s velocity time series; 4.3) After completing the interpolation, check the data to ensure that all missing values ​​have been filled.

6. The method for identifying and early warning glacier surge according to claim 5, characterized in that: In step 5, the sensitivity study of the case glacier is carried out through outlier detection, generalized extreme value studentized deviation test GESD and cluster analysis. The specific steps are as follows: 5.1) Perform time series decomposition; The seasonal trend decomposition method is used to decompose the time series of the case glacier, and the glacier velocity time series of each point is decomposed into three components: trend, seasonality and residual. 5.2) Perform outlier detection: Extract the residual sequence of each point in turn and perform multiple groups of outlier detection on it; this part involves two parameters, namely α and the maximum anomaly allowed to be detected in the time series; α is the width of the envelope function on the residue curve, where data points outside the envelope are classified as outliers; The values ​​of α are 0.05 and 0.1; the maximum anomaly values ​​allowed to be detected are 20%, 40%, 60%, 80%, and 100%, respectively. Multiple groups of experiments are conducted to obtain whether the flow velocity of the points on the mainstream line of the glacier is abnormal in all time series. If it is abnormal, the output is yes, and if it is not abnormal, the output is no. 5.3) Perform cluster analysis; For each time series, a moving window of size 20 is used to count the outliers in the window; the clustering threshold sizes are selected as 10, 12, 14, 16, and 18 for sensitivity research; if the sum of the values ​​of yes in the window reaches the set threshold, the month is marked as an abnormal month; if 4 or more consecutive months meet the abnormal criteria, they are marked as jump events; abnormal months with less than 4 months are marked as speed fluctuations; the rest are marked as normal months; 5.4) Keep the relevant information of glacier time step and judgment of month.

7. The method for identifying and early warning glacier surge according to claim 6, characterized in that: In step 6, the specific steps are as follows: 6.1) Construct the start and end time and duration matrix of the real surge of the case glacier; 6.2) Calculate the F1 score and construct a table for easy comparison; 6.3) Comprehensively consider and determine the best performance parameter combination.

8. The method for identifying and early warning glacier surge according to claim 7, characterized in that: In step 7, the specific steps are as follows: 7.1) Perform time series decomposition on all glaciers; The seasonal trend decomposition method is used to decompose the time series, and the glacier velocity time series of each point inside each glacier is decomposed into three components: trend, seasonality and residual. 7.2) Extract the residual sequence of each glacier and perform outlier detection on it using the best parameter combination; obtain information on whether the flow velocity of all points on the mainstream line of all glaciers is abnormal in all time series, where abnormal output is yes and non-abnormal output is no; the information of points on the same mainstream line of the glacier is stored in the same table, including the time step, the point number FID and whether it is abnormal related information, in long format; 7.3) Perform cluster analysis; Each glacier is indexed by the time step, and a moving window of size 20 is applied to the abnormality column to count the number of abnormal points in the window. If the number exceeds the threshold of 12, the corresponding time step is returned as an abnormal month; Use the loop to complete the judgment of all time steps; 7.4) If 4 or more consecutive months meet the abnormal criteria, it is marked as a surge glacier.

9. The method for identifying and early warning glacier surge according to claim 8, characterized in that: In step 8, the specific steps are as follows: The maximum flow velocity of the surging glacier and the average flow velocity in normal months are calculated, and the difference between the two is used as the multiple of the increase in glacier velocity. The minimum value is used as the threshold for the flow velocity increase.

Images