Glacier change evaluation method based on multi-source data fusion

Abstract

The invention relates to a glacier change assessment method based on multi-source data fusion, and the method comprises the steps: collecting radar interference measurement data, optical image data and GPS actual measurement data of a target glacier region, and constructing a multi-source data set; the multi-source data set is preprocessed and then fused, and a fused flow velocity field is obtained; performing spatial gradient analysis on the fused flow velocity field to obtain a local change rate of the flow velocity field, judging a sudden motion event based on the local change rate, and obtaining a motion event distribution diagram; constructing a dynamic flow velocity field for a sudden motion area in the motion event distribution diagram, calculating glacier surface stress distribution, and obtaining glacier motion state distribution; and carrying out time sequence analysis on the glacier motion state distribution, obtaining the change trend of the motion state, evaluating the change condition, and completing the change evaluation of the target glacier area. By integrating the multi-source remote sensing data, the temporal-spatial resolution and precision of glacier change monitoring and evaluation are improved.

Description

Technical Field

[0001] This invention relates to the field of glacier change monitoring and assessment technology, and in particular to a method for assessing glacier change based on multi-source data fusion. Background Technology

[0002] Glaciers are a sensitive indicator of global climate change, and their changes directly affect water supply, ecosystem stability, and geological disaster risks. Studying glacier dynamics is crucial for addressing climate change and preventing and mitigating disasters.

[0003] However, existing glacier monitoring methods mostly rely on single data sources, such as optical imagery or ground measurements, which are insufficient to fully capture the complexity of glacier movement. Furthermore, data acquisition is limited in harsh environments, resulting in insufficient accuracy and timeliness, leading to inaccurate assessments of accelerated glacier flow or instability. Accurate measurement of glacier flow velocity is a core challenge in assessing glacier changes. The limitations of single data sources make it difficult for glacier velocity models to accurately reflect their dynamic characteristics, especially under complex terrain or extreme weather conditions, where data coverage and resolution are limited, hindering comprehensive monitoring. The heterogeneity and incompleteness of data sources further exacerbate the problem. Different types of data (such as radar, optical imagery, and GPS measurements) differ significantly in spatiotemporal resolution and measurement principles, making the fusion of these data to construct a unified velocity model a technical bottleneck. Imperfect data fusion directly leads to a decrease in the accuracy of dynamic monitoring of glacier movement, particularly in identifying sudden movement events, where existing methods struggle to provide real-time and reliable evidence. Therefore, this invention proposes a glacier change assessment method based on multi-source data fusion. Summary of the Invention

[0004] The purpose of this invention is to provide a glacier change assessment method based on multi-source data fusion. By fusing radar interferometry data, optical image feature matching, and GPS measured data, a high-precision glacier flow velocity model is constructed to achieve dynamic monitoring and change assessment of glacier movement.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for assessing glacier change based on multi-source data fusion includes:

[0007] Collect radar interferometry data, optical image data, and GPS measured data of the target glacier area to construct a multi-source dataset;

[0008] The multi-source dataset is preprocessed and then fused to obtain the fused flow velocity field;

[0009] Spatial gradient analysis is performed on the fused velocity field to obtain the local rate of change of the velocity field, and sudden motion events are judged based on the local rate of change to obtain a motion event distribution map.

[0010] A dynamic velocity field is constructed for the sudden movement region in the motion event distribution map, and the stress distribution on the glacier surface is calculated to obtain the glacier motion state distribution;

[0011] A time series analysis was performed on the distribution of the glacier's motion state to obtain the changing trend of the motion state and assess the changes, thus completing the change assessment of the target glacier area.

[0012] Optionally, the multi-source dataset is preprocessed before fusion to obtain a fused velocity field, including:

[0013] Determine the spatiotemporal resolution and measurement error of each data source, and perform standardization and alignment processing on the multi-source dataset based on the spatiotemporal resolution and measurement error to obtain a standardized multi-source dataset;

[0014] The standardized multi-source dataset is fused using a weighted average algorithm to obtain a preliminary fused velocity field, which is then smoothed to obtain the final fused velocity field.

[0015] Optionally, the spatiotemporal resolution and measurement error of each data source are determined, and the multi-source dataset is standardized and aligned based on the spatiotemporal resolution and measurement error to obtain a standardized multi-source dataset, including:

[0016] The multi-source dataset is preprocessed by denoising, format unification, and missing value imputation. The metadata of the preprocessed multi-source dataset is analyzed to determine the spatiotemporal resolution of each data source and calculate the measurement error. A spatiotemporal resolution description table and measurement error distribution are constructed.

[0017] Based on the spatiotemporal resolution description table and measurement error distribution, the preprocessed multi-source dataset is normalized, and the resolution difference of each data source is calculated. If the resolution difference exceeds the preset difference, the normalized multi-source dataset is spatiotemporally aligned through data consistency verification to obtain the standardized multi-source dataset.

[0018] Optionally, spatiotemporal alignment of the normalized multi-source dataset is performed through data consistency verification, including:

[0019] S1. Use time-series interpolation to linearly interpolate the time series of radar interferometric measurement data to obtain a first intermediate dataset with consistent time resolution.

[0020] S2. The spatial resolution of the optical image data is processed by bicubic interpolation through spatial resampling to obtain a second intermediate dataset with consistent spatial resolution.

[0021] S3. Based on the spatiotemporal distribution characteristics of GPS measured data, the Kriging interpolation method is used to perform spatiotemporal joint interpolation on the first intermediate dataset and the second intermediate dataset to obtain the third intermediate dataset.

[0022] S4. Determine the spatiotemporal consistency parameter of the third intermediate dataset. If it reaches the preset parameter range, proceed to S5. If it does not reach the preset parameter range, adjust the interpolation parameter and return to S1.

[0023] S5. Based on the qualified third intermediate dataset with spatiotemporal consistency parameters, the features of radar interferometry data, optical image data and GPS measured data are weighted and fused to obtain the final dataset.

[0024] Optionally, a weighted average algorithm is used to fuse the standardized multi-source dataset to obtain a preliminary fused velocity field, which is then smoothed to obtain the final fused velocity field, including:

[0025] The weight values ​​of each data source are calculated based on the measurement error, and radar interferometric measurement data, optical image matching data and GPS measured data are fused according to the weight values ​​to obtain a preliminary fused velocity field.

[0026] Based on the preliminary fused velocity field, velocity data of local and neighboring regions are obtained, and the velocity difference is calculated to identify outliers.

[0027] The outliers are smoothed using a Gaussian process regression algorithm to obtain smoothed velocity values, which are then used to replace the outliers to construct a smoothed velocity field, i.e., the final fused velocity field.

[0028] Optionally, spatial gradient analysis is performed on the fused velocity field to obtain the local rate of change of the velocity field, and sudden motion events are determined based on the local rate of change to obtain a motion event distribution map, including:

[0029] The spatial gradient of the smoothed velocity field and the local rate of change at each spatial point are calculated using the finite difference method to obtain the rate of change distribution.

[0030] Based on the rate of change distribution, spatial points where sudden motion events exist are identified and marked, and an event mark set is obtained;

[0031] Spatial clustering of the event marker set is performed using a clustering algorithm to obtain the cluster distribution of sudden motion events;

[0032] A two-dimensional spatial distribution map of motion events is generated based on the cluster distribution, and the distribution map is smoothed to generate the final motion event distribution map.

[0033] Optionally, a dynamic velocity field is constructed for the sudden movement region in the motion event distribution map, and the stress distribution on the glacier surface is calculated to obtain the glacier motion state distribution, including:

[0034] Detect sudden motion regions in the motion event distribution map and extract the region boundaries to obtain the coordinates of the sudden motion regions;

[0035] Real-time radar interferometry data and optical image data corresponding to the coordinates of the sudden movement area are acquired, and the velocity field is updated in time based on the real-time radar interferometry data and optical image data to obtain a preliminary dynamic velocity field.

[0036] The preliminary dynamic velocity field is optimized using a particle filtering algorithm to obtain an optimized dynamic velocity field, and the velocity distribution on the glacier surface is constructed based on the optimized dynamic velocity field.

[0037] The finite difference method is used to divide the velocity distribution on the glacier surface into a grid. Based on the grid and boundary conditions, the stress distribution on the glacier surface is calculated, and the stress values ​​are obtained.

[0038] Based on the stress value, determine the stress concentration area, obtain the stress concentration distribution, analyze the glacier movement state, and obtain the glacier movement state distribution.

[0039] Optionally, a time series analysis is performed on the distribution of the glacier's motion state to obtain the changing trend of the motion state and assess the changes, including:

[0040] The periodic and trend components of the glacier's motion state distribution are extracted using the time series decomposition method to obtain the changing trend, and the local motion velocity of the changing trend is calculated using the sliding window method to determine the accelerated flow region.

[0041] Unstable regions are extracted based on the accelerated flow region, and the changes in unstable regions in adjacent time periods are compared through differential analysis to obtain the dynamic characteristics of the regions and determine the dynamic monitoring results.

[0042] The dynamic features of the region are input into a pre-trained time series prediction model to predict the future trend of glacier movement and to identify potential risk areas based on the future trend.

[0043] The dynamic monitoring results and potential risk areas are spatially mapped to output a distribution map of glacier movement status.

[0044] The beneficial effects of this invention are as follows:

[0045] This invention constructs a standardized multi-source dataset by collecting radar interferometry, optical imagery, and GPS measured data, and performs spatiotemporal alignment processing. A weighted average algorithm is used to fuse the multi-source data, and Gaussian process regression is employed to smooth outliers, thus building a high-precision glacier flow velocity model. Sudden motion events are identified through spatial gradient analysis, and Kalman filtering is used for dynamic updates. Finally, the stress distribution on the glacier surface is calculated, and the changing trends of the motion state are analyzed, enabling dynamic monitoring and assessment of accelerated glacier flow and unstable regions. This invention effectively integrates multi-source remote sensing data, improving the spatiotemporal resolution and accuracy of glacier monitoring, and providing reliable technical support for glacier change research and disaster early warning. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart of a glacier change assessment method based on multi-source data fusion, according to an embodiment of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] This embodiment provides a method for assessing glacier changes based on multi-source data fusion, such as... Figure 1 As shown, it includes:

[0051] Collect radar interferometry data, optical image data, and GPS measured data of the target glacier area to construct a multi-source dataset;

[0052] The multi-source dataset is preprocessed and then fused to obtain the fused flow velocity field;

[0053] Spatial gradient analysis is performed on the fused velocity field to obtain the local rate of change of the velocity field, and sudden motion events are judged based on the local rate of change to obtain a motion event distribution map.

[0054] A dynamic velocity field is constructed for the sudden movement region in the motion event distribution map, and the stress distribution on the glacier surface is calculated to obtain the glacier motion state distribution;

[0055] A time series analysis was performed on the distribution of the glacier's motion state to obtain the changing trend of the motion state and assess the changes, thus completing the change assessment of the target glacier area.

[0056] Specifically, this embodiment constructs a standardized multi-source dataset by collecting radar interferometry, optical imagery, and GPS measured data, and performs spatiotemporal alignment processing. A weighted average algorithm is used to fuse the multi-source data, and Gaussian process regression is used to smooth outouts, constructing a high-precision glacier flow velocity model. Sudden motion events are identified through spatial gradient analysis, and Kalman filtering is used for dynamic updates. Finally, the stress distribution on the glacier surface is calculated, and the changing trends of the motion state are analyzed, enabling dynamic monitoring and assessment of accelerated glacier flow and unstable regions. This effectively integrates multi-source remote sensing data, improves the spatiotemporal resolution and accuracy of glacier monitoring, and provides reliable technical support for glacier change research and disaster early warning.

[0057] Furthermore, the multi-source dataset is preprocessed before fusion to obtain the fused velocity field, including:

[0058] Determine the spatiotemporal resolution and measurement error of each data source, and perform standardization and alignment processing on the multi-source dataset based on the spatiotemporal resolution and measurement error to obtain a standardized multi-source dataset;

[0059] The standardized multi-source dataset is fused using a weighted average algorithm to obtain a preliminary fused velocity field, which is then smoothed to obtain the final fused velocity field.

[0060] As an optional implementation, the spatiotemporal resolution and measurement error of each data source are determined, and the multi-source dataset is standardized and aligned based on the spatiotemporal resolution and measurement error to obtain a standardized multi-source dataset, including:

[0061] The multi-source dataset is preprocessed by denoising, format unification, and missing value imputation. The metadata of the preprocessed multi-source dataset is analyzed to determine the spatiotemporal resolution of each data source and calculate the measurement error. A spatiotemporal resolution description table and measurement error distribution are constructed.

[0062] Based on the spatiotemporal resolution description table and measurement error distribution, the preprocessed multi-source dataset is normalized, and the resolution difference of each data source is calculated. If the resolution difference exceeds the preset difference, the normalized multi-source dataset is spatiotemporally aligned through data consistency verification to obtain the standardized multi-source dataset.

[0063] This includes spatiotemporal alignment of the normalized multi-source dataset through data consistency verification, including:

[0064] S1. Use time-series interpolation to linearly interpolate the time series of radar interferometric measurement data to obtain a first intermediate dataset with consistent time resolution.

[0065] S2. The spatial resolution of the optical image data is processed by bicubic interpolation through spatial resampling to obtain a second intermediate dataset with consistent spatial resolution.

[0066] S3. Based on the spatiotemporal distribution characteristics of GPS measured data, the Kriging interpolation method is used to perform spatiotemporal joint interpolation on the first intermediate dataset and the second intermediate dataset to obtain the third intermediate dataset.

[0067] S4. Determine the spatiotemporal consistency parameter of the third intermediate dataset. If it reaches the preset parameter range, proceed to S5. If it does not reach the preset parameter range, adjust the interpolation parameter and return to S1.

[0068] S5. Based on the qualified third intermediate dataset with spatiotemporal consistency parameters, the features of radar interferometry data, optical image data and GPS measured data are weighted and fused to obtain the final dataset.

[0069] Specifically, in this embodiment, radar interferometry data is acquired using synthetic aperture radar (SAR), such as C-band radar data from the Sentinel-1 satellite, with a spatial resolution of 5×20 meters and a temporal resolution of 12 days. The interferometric phase is processed using the DInSAR algorithm, and the measurement error of surface deformation can be controlled within ±3 mm. Simultaneously, optical image data, such as multispectral data from Landsat-8, is acquired, with a spatial resolution of 30 meters and a temporal resolution of 16 days. The NDVI algorithm is used to analyze vegetation cover changes, and the radiometric calibration error is less than 5%. Combined with GPS measured data and coordinate data collected through a network of reference stations, the planar accuracy reaches ±2 mm, and the elevation accuracy is ±5 mm. A Kalman filter algorithm is used to eliminate multipath effects and clock bias. Different data sources are unified to the WGS84 coordinate system. The optical images are resampled to a 5-meter resolution using a spatiotemporal registration algorithm (such as nearest neighbor interpolation) to align with the radar data. Linear interpolation is used to fill in missing dates in the time series. To address measurement errors, a weighted fusion algorithm is employed, which assigns weights inversely proportional to the errors of each data source. For example, GPS has a weight of 0.6, radar has a weight of 0.3, and optics has a weight of 0.1. This results in the generation of a standardized dataset, reducing the overall deformation monitoring error to ±1.5 mm.

[0070] When standardizing multi-source datasets, if the spatiotemporal resolution difference exceeds a preset threshold (e.g., time interval greater than 12 hours or spatial resolution difference greater than 10 meters), temporal interpolation and spatial resampling methods are required to align the data. Taking radar interferometry data (24-hour temporal resolution, 5-meter spatial resolution), optical imagery data (16-day temporal resolution, 10-meter spatial resolution), and GPS measured data (1-hour temporal resolution, 1-kilometer spatial resolution) as an example, firstly, linear interpolation is performed on the optical imagery data with the lowest temporal resolution to interpolate the 16-day time interval to 24 hours, consistent with the radar data. Next, bicubic convolutional resampling is performed on the GPS data with the lowest spatial resolution to improve the 1-kilometer resolution to 5 meters. After completing the spatiotemporal alignment, consistency is verified through root mean square error (RMSE) analysis, requiring the RMSE between each dataset to be less than 0.8 pixels. For the band differences between radar and optical data, a histogram matching algorithm is used to adjust the DN value distribution of the optical imagery to align it with the statistical characteristics (mean ± 2 standard deviations) of the radar data in the same ground feature area. The resulting spatiotemporally consistent dataset meets the requirements of subsequent deformation analysis, with time alignment error controlled within ±30 minutes and spatial registration accuracy better than 0.3 pixels.

[0071] As an optional implementation, a weighted average algorithm is used to fuse the standardized multi-source dataset to obtain a preliminary fused velocity field, which is then smoothed to obtain the final fused velocity field, including:

[0072] The weight values ​​of each data source are calculated based on the measurement error, and radar interferometric measurement data, optical image matching data and GPS measured data are fused according to the weight values ​​to obtain a preliminary fused velocity field.

[0073] Based on the preliminary fused velocity field, velocity data of local and neighboring regions are obtained, and the velocity difference is calculated to identify outliers.

[0074] The outliers are smoothed using a Gaussian process regression algorithm to obtain smoothed velocity values, which are then used to replace the outliers to construct a smoothed velocity field, i.e., the final fused velocity field.

[0075] Specifically, this embodiment first acquires surface deformation data using radar interferometry, assuming a measurement error of 0.5 mm. Then, it acquires surface displacement data using optical image matching, with a measurement error of 0.3 mm, and finally, surface displacement data using GPS measurements, with a measurement error of 0.2 mm. Weights are determined based on the reciprocals of the measurement errors of each data source: the weight for radar interferometry data is 1 / 0.5 = 2, the weight for optical image matching data is 1 / 0.3 ≈ 3.33, and the weight for GPS measurements is 1 / 0.2 = 5. A weighted average algorithm is then used for data fusion, with the specific formula: Fusion velocity field = (Radar data × 2 + Optical data × 3.33 + GPS data × 5) / (2 + 3.33 + 5). Assuming that the radar data at a certain point is 1.2 mm / year, the optical data is 1.0 mm / year, and the GPS data is 1.1 mm / year, then the preliminary fused velocity field = (1.2×2+1.0×3.33+1.1×5) / (2+3.33+5)≈1.09 mm / year.

[0076] In the initial fusion of the velocity field, if the velocity value in a local area differs from that in neighboring areas by more than a preset threshold of 0.5 m / s, it is identified as an outlier. A Gaussian process regression algorithm is then used to smooth the outlier. First, velocity data from neighboring areas are obtained using spatial interpolation. Assuming there are 10 sampling points in the neighboring area with velocity values ​​of 1.2 m / s, 1.3 m / s, 1.1 m / s, 1.4 m / s, 1.2 m / s, 1.3 m / s, 1.1 m / s, 1.4 m / s, 1.2 m / s, and 1.3 m / s, respectively. Then, the Gaussian process regression algorithm is used to model these data. A radial basis function is selected as the kernel function, and a scale parameter of length 0.1 is set. The model parameters are optimized using maximum likelihood estimation. Next, the velocity value from the outlier area, for example, 2.0 m / s, is input into the trained model, resulting in a smoothed velocity value of 1.3 m / s. Finally, the smoothed velocity values ​​are updated in the velocity field to ensure the continuity and consistency of the entire velocity field.

[0077] Furthermore, spatial gradient analysis is performed on the fused velocity field to obtain the local rate of change of the velocity field, and sudden motion events are determined based on the local rate of change to obtain a motion event distribution map, including:

[0078] The spatial gradient of the smoothed velocity field and the local rate of change at each spatial point are calculated using the finite difference method to obtain the rate of change distribution.

[0079] Based on the rate of change distribution, spatial points where sudden motion events exist are identified and marked, and an event mark set is obtained;

[0080] Spatial clustering of the event marker set is performed using a clustering algorithm to obtain the cluster distribution of sudden motion events;

[0081] A two-dimensional spatial distribution map of motion events is generated based on the cluster distribution, and the distribution map is smoothed to generate the final motion event distribution map.

[0082] Specifically, this embodiment collects data from a smoothed velocity field and uses a spatial gradient algorithm to calculate the local rate of change of the velocity field. For example, the central difference method is used to calculate the gradient of the smoothed velocity field in the x and y directions, obtaining the gradient value at each point. The gradient value is then used to determine whether a sudden motion event exists. A threshold is set; for example, a gradient magnitude greater than 1.0 m / s / m is considered a sudden motion event. By traversing all points, points that meet the condition are selected. For example, a point with a gradient magnitude of 1.2 m / s / m is marked as a sudden motion event point. Finally, all marked points are plotted on a motion event distribution map, using different colors to represent motion events of different intensities, such as red for high-intensity events and blue for low-intensity events. By analyzing the distribution map, the spatial distribution characteristics of sudden motion events in the velocity field can be intuitively understood. For example, a dense concentration of red dots in a certain area indicates the presence of many high-intensity motion events in that area.

[0083] Furthermore, a dynamic velocity field is constructed for the sudden movement region in the motion event distribution map, and the stress distribution on the glacier surface is calculated to obtain the glacier motion state distribution, including:

[0084] Detect sudden motion regions in the motion event distribution map and extract the region boundaries to obtain the coordinates of the sudden motion regions;

[0085] Real-time radar interferometry data and optical image data corresponding to the coordinates of the sudden movement area are acquired, and the velocity field is updated in time based on the real-time radar interferometry data and optical image data to obtain a preliminary dynamic velocity field.

[0086] The preliminary dynamic velocity field is optimized using a particle filtering algorithm to obtain an optimized dynamic velocity field, and the velocity distribution on the glacier surface is constructed based on the optimized dynamic velocity field.

[0087] The finite difference method is used to divide the velocity distribution on the glacier surface into a grid. Based on the grid and boundary conditions, the stress distribution on the glacier surface is calculated, and the stress values ​​are obtained.

[0088] Based on the stress value, determine the stress concentration area, obtain the stress concentration distribution, analyze the glacier movement state, and obtain the glacier movement state distribution.

[0089] Specifically, in this embodiment, if a sudden movement region is detected in the motion event distribution map, the velocity field of that region is first updated temporally using a Kalman filter algorithm. Next, the latest acquired radar interferometry data and optical image data are fused. The radar interferometry data provides the vertical component of the velocity field, and the optical image data provides the horizontal component. Assuming the radar interferometry data is [8 m / s, 3 m / s] and the optical image data is [12 m / s, 6 m / s], a weighted average method is used with weights of 0.6 and 0.4 respectively, resulting in a fused velocity field of [10.4 m / s, 4.2 m / s]. Finally, the dynamic velocity field is obtained through the Kalman filter algorithm and the fused data.

[0090] In calculating the stress distribution on the glacier surface, a finite difference grid was first constructed based on dynamic velocity field data, with a grid resolution of 50 m × 50 m. The momentum conservation equation was discretized using a central difference scheme, and Glen's flow law was applied to calculate the stress tensor. In a region with an ice thickness of 200 m, the maximum principal stress reached 150 kPa. Regions with a stress concentration factor exceeding 2.5 were identified as potential fracture development zones. The principal stress direction field was extracted through eigenvalue decomposition and classified into motion states based on the velocity vector field: when the angle between the principal stress direction and the velocity was less than 15°, it was marked as a tensile zone; when it was greater than 75°, it was marked as a shear zone. Within 300 m of the glacier terminus, the measured tensile zone accounted for 62%, with a spatial correlation of 0.78 with the surface fracture observation data. The Morris sensitivity analysis method was used to determine that the ice thickness measurement error had the greatest impact on the stress results, with a sensitivity index of 0.53. The thickness measurement accuracy needed to be controlled within ±5 m to ensure that the relative error of the stress calculation was less than 10%.

[0091] Furthermore, a time series analysis is performed on the distribution of the glacier's motion state to obtain the changing trend of the motion state and assess the changes, including:

[0092] The periodic and trend components of the glacier's motion state distribution are extracted using the time series decomposition method to obtain the changing trend, and the local motion velocity of the changing trend is calculated using the sliding window method to determine the accelerated flow region.

[0093] Unstable regions are extracted based on the accelerated flow region, and the changes in unstable regions in adjacent time periods are compared through differential analysis to obtain the dynamic characteristics of the regions and determine the dynamic monitoring results.

[0094] The dynamic features of the region are input into a pre-trained time series prediction model to predict the future trend of glacier movement and to identify potential risk areas based on the future trend.

[0095] The dynamic monitoring results and potential risk areas are spatially mapped to output a distribution map of glacier movement status.

[0096] Specifically, this embodiment analyzes the distribution of glacier movement over time. First, it uses satellite remote sensing data to obtain glacier surface displacement information. Then, it employs a pixel offset tracking (POT) algorithm to calculate the glacier's displacement at different time points. For example, glacier surface displacement data were acquired on January 1, 2020, and January 31, 2020. By comparing the two sets of data, the average displacement of the glacier over 30 days was calculated to be 12 meters, with a displacement accuracy of 0.1 meters. Next, time series analysis methods, such as the autoregressive integral moving average (ARIMA) model, are used to model the glacier displacement data and predict future movement trends. For instance, based on displacement data from January to December 2020, the ARIMA model predicts a glacier displacement of 15 meters in January 2021, which is highly consistent with the actual observed value of 14.8 meters. Furthermore, by calculating the acceleration of glacier displacement, regions of accelerated glacier flow were identified. For example, in one region, the glacier displacement was 6 meters from January to June 2020, while it was 9 meters from July to December 2020, with an acceleration of 0.5 m / month², indicating accelerated glacier flow in this area. Finally, combining glacier thickness and topographic data, the finite element method (FEM) was used to simulate glacier stress distribution and identify unstable regions. For instance, in one region, simulation results showed that the stress at the glacier base exceeded the critical value of 1.5 MPa, indicating a risk of instability in this area.

[0097] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for assessing glacier change based on multi-source data fusion, characterized in that, include: Collect radar interferometry data, optical image data, and GPS measured data of the target glacier area to construct a multi-source dataset; The multi-source dataset is preprocessed and then fused to obtain the fused flow velocity field; Spatial gradient analysis is performed on the fused velocity field to obtain the local rate of change of the velocity field, and sudden motion events are judged based on the local rate of change to obtain a motion event distribution map. A dynamic velocity field is constructed for the sudden movement region in the motion event distribution map, and the stress distribution on the glacier surface is calculated to obtain the glacier motion state distribution; A time series analysis was performed on the distribution of the glacier's motion state to obtain the changing trend of the motion state and assess the changes, thus completing the change assessment of the target glacier area.

2. The glacier change assessment method based on multi-source data fusion according to claim 1, characterized in that, The multi-source dataset is preprocessed and then fused to obtain a fused velocity field, including: Determine the spatiotemporal resolution and measurement error of each data source, and perform standardization and alignment processing on the multi-source dataset based on the spatiotemporal resolution and measurement error to obtain a standardized multi-source dataset; The standardized multi-source dataset is fused using a weighted average algorithm to obtain a preliminary fused velocity field, which is then smoothed to obtain the final fused velocity field.

3. The glacier change assessment method based on multi-source data fusion according to claim 2, characterized in that, Determine the spatiotemporal resolution and measurement error of each data source, and perform standardization and alignment processing on the multi-source dataset based on the spatiotemporal resolution and measurement error to obtain a standardized multi-source dataset, including: The multi-source dataset is preprocessed by denoising, format unification, and missing value imputation. The metadata of the preprocessed multi-source dataset is analyzed to determine the spatiotemporal resolution of each data source and calculate the measurement error. A spatiotemporal resolution description table and measurement error distribution are constructed. Based on the spatiotemporal resolution description table and measurement error distribution, the preprocessed multi-source dataset is normalized, and the resolution difference of each data source is calculated. If the resolution difference exceeds the preset difference, the normalized multi-source dataset is spatiotemporally aligned through data consistency verification to obtain the standardized multi-source dataset.

4. The glacier change assessment method based on multi-source data fusion according to claim 3, characterized in that, Spatiotemporal alignment of the normalized multi-source dataset is performed through data consistency verification, including: S1. Use time-series interpolation to linearly interpolate the time series of radar interferometric measurement data to obtain a first intermediate dataset with consistent time resolution. S2. The spatial resolution of the optical image data is processed by bicubic interpolation through spatial resampling to obtain a second intermediate dataset with consistent spatial resolution. S3. Based on the spatiotemporal distribution characteristics of GPS measured data, the Kriging interpolation method is used to perform spatiotemporal joint interpolation on the first intermediate dataset and the second intermediate dataset to obtain the third intermediate dataset. S4. Determine the spatiotemporal consistency parameter of the third intermediate dataset. If it reaches the preset parameter range, proceed to S5. If it does not reach the preset parameter range, adjust the interpolation parameter and return to S1. S5. Based on the qualified third intermediate dataset with spatiotemporal consistency parameters, the features of radar interferometry data, optical image data and GPS measured data are weighted and fused to obtain the final dataset.

5. The glacier change assessment method based on multi-source data fusion according to claim 2, characterized in that, A weighted average algorithm is used to fuse the standardized multi-source dataset to obtain a preliminary fused velocity field, which is then smoothed to obtain the final fused velocity field, including: The weight values ​​of each data source are calculated based on the measurement error, and radar interferometric measurement data, optical image matching data and GPS measured data are fused according to the weight values ​​to obtain a preliminary fused velocity field. Based on the preliminary fused velocity field, velocity data of local and neighboring regions are obtained, and the velocity difference is calculated to identify outliers. The outliers are smoothed using a Gaussian process regression algorithm to obtain smoothed velocity values, which are then used to replace the outliers to construct a smoothed velocity field, i.e., the final fused velocity field.

6. The glacier change assessment method based on multi-source data fusion according to claim 5, characterized in that, Spatial gradient analysis is performed on the fused velocity field to obtain the local rate of change of the velocity field. Based on the local rate of change, sudden motion events are identified, and a motion event distribution map is obtained, including: The spatial gradient of the smoothed velocity field and the local rate of change at each spatial point are calculated using the finite difference method to obtain the rate of change distribution. Based on the rate of change distribution, spatial points where sudden motion events exist are identified and marked, and an event mark set is obtained; Spatial clustering of the event marker set is performed using a clustering algorithm to obtain the cluster distribution of sudden motion events; A two-dimensional spatial distribution map of motion events is generated based on the cluster distribution, and the distribution map is smoothed to generate the final motion event distribution map.

7. The glacier change assessment method based on multi-source data fusion according to claim 1, characterized in that, A dynamic velocity field is constructed for the sudden movement region in the motion event distribution map, and the stress distribution on the glacier surface is calculated to obtain the glacier motion state distribution, including: Detect sudden motion regions in the motion event distribution map and extract the region boundaries to obtain the coordinates of the sudden motion regions; Real-time radar interferometry data and optical image data corresponding to the coordinates of the sudden movement area are acquired, and the velocity field is updated in time based on the real-time radar interferometry data and optical image data to obtain a preliminary dynamic velocity field. The preliminary dynamic velocity field is optimized using a particle filtering algorithm to obtain an optimized dynamic velocity field, and the velocity distribution on the glacier surface is constructed based on the optimized dynamic velocity field. The finite difference method is used to divide the velocity distribution on the glacier surface into a grid. Based on the grid and boundary conditions, the stress distribution on the glacier surface is calculated, and the stress values ​​are obtained. Based on the stress value, determine the stress concentration area, obtain the stress concentration distribution, analyze the glacier movement state, and obtain the glacier movement state distribution.

8. The glacier change assessment method based on multi-source data fusion according to claim 1, characterized in that, A time series analysis was performed on the distribution of the glacier's motion state to obtain the changing trend of the motion state and assess the changes, including: The periodic and trend components of the glacier's motion state distribution are extracted using the time series decomposition method to obtain the changing trend, and the local motion velocity of the changing trend is calculated using the sliding window method to determine the accelerated flow region. Unstable regions are extracted based on the accelerated flow region, and the changes in unstable regions in adjacent time periods are compared through differential analysis to obtain the dynamic characteristics of the regions and determine the dynamic monitoring results. The dynamic features of the region are input into a pre-trained time series prediction model to predict the future trend of glacier movement and to identify potential risk areas based on the future trend. The dynamic monitoring results and potential risk areas are spatially mapped to output a distribution map of glacier movement status.

Images