InSAR permafrost deformation monitoring method considering soil environmental factors

Abstract

This invention provides an InSAR method for monitoring permafrost deformation that takes into account soil environmental factors. Based on a random forest model in machine learning, it downscales soil temperature and moisture data using multiple MODIS datasets and DEM products to obtain high-resolution downscaled soil temperature and moisture data. SAR image data of the target permafrost region is processed to obtain interferograms. Based on this, a permafrost deformation model that takes into account soil environmental factors is constructed using the downscaled soil temperature and moisture data. The deformation model is then used to process the interferograms, and an observation equation is established to obtain the deformation results of the target permafrost region. This method can establish the physical relationship between permafrost deformation and soil hydrothermal conditions, effectively improving the accuracy of permafrost deformation monitoring. Ultimately, it increases the accuracy of time-series deformation results, effectively enhancing the precision of permafrost deformation monitoring.

Description

Technical Field

[0001] This invention relates to the field of permafrost monitoring technology, and in particular to an InSAR permafrost deformation monitoring method that takes into account soil environmental factors. Background Technology

[0002] The complex surface deformation processes in permafrost regions are a major cause of engineering geological hazards such as thaw landslides and frost heave. Therefore, surface deformation monitoring in permafrost regions is of great significance. Traditional permafrost deformation monitoring methods include burying settlement meters, leveling observations, and GPS observations, which offer high observation accuracy and good temporal continuity. However, these methods mostly target single points or small areas, failing to acquire information on large-area surface deformation. Furthermore, the harsh natural environment and complex terrain of the Qinghai-Tibet Plateau permafrost region hinder large-scale field monitoring activities. Compared to traditional field monitoring methods, Synthetic Aperture Radar Interferometry (InSAR) technology offers all-weather, all-day monitoring capabilities, is independent of the monitoring environment and tools, and is more suitable for surface deformation monitoring in permafrost regions with harsh natural environments. In particular, time-series InSAR technology, represented by SBAS-InSAR, effectively mitigates the decoherence effects of D-InSAR technology. It can acquire large-scale, long-term, and high-precision surface deformation information at a lower cost and is widely used in permafrost regions with harsh natural environments, such as the Qinghai-Tibet Plateau.

[0003] For SBAS-InSAR technology, a reasonable phase deformation model is crucial for improving the accuracy of surface deformation monitoring. Existing permafrost deformation models mainly include purely mathematical models and deformation models that take into account external environmental factors. Among them, purely mathematical models rely on mathematical functions to fit phase changes, which is not conducive to revealing the physical laws and influencing factors of permafrost deformation. Existing deformation models that take into account external environmental factors only consider external environmental factors such as temperature and precipitation, ignoring the impact of changes in soil internal water and heat conditions on permafrost deformation. Furthermore, the sparse distribution of meteorological stations in permafrost areas makes it difficult to obtain high-precision, high-resolution meteorological data, which is not conducive to the construction of deformation models. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of the aforementioned background technology by providing an InSAR permafrost deformation monitoring scheme. This scheme constructs a permafrost deformation model that takes into account soil environmental factors, establishes the physical relationship between permafrost deformation and soil hydrothermal conditions, and better describes the permafrost deformation process. Simultaneously, it uses a random forest model to obtain high spatial resolution downscaled soil temperature and humidity data to construct the permafrost deformation model, thereby reducing the influence of the deformation model parameter resolution on the deformation calculation results.

[0005] To achieve the above objectives, this invention provides an InSAR method for monitoring frozen soil deformation that takes into account soil environmental factors, comprising the following steps:

[0006] S1, based on the random forest model, uses multiple MODIS and DEM products to downscale soil temperature and soil moisture data to obtain high-resolution downscaled soil temperature and soil moisture data.

[0007] S2, process N SAR image data covering the target permafrost area to obtain M interferograms after phase unwrapping;

[0008] S3, based on high-resolution soil temperature and soil moisture downscaling data, modeling is performed for each high coherence point in the interferogram that meets the preset coherence conditions, to obtain a frozen soil deformation model that takes into account soil environmental factors.

[0009] S4. Each interferogram is processed using the permafrost deformation model to obtain the phase values ​​of high coherence points in all interferograms. The deformation time series results of all high coherence points are integrated to obtain the deformation results of the target permafrost region.

[0010] Furthermore, S1 comprises two phases:

[0011] Training phase: First, the soil temperature and soil moisture proxy variables from MODIS and DEM products are aggregated and averaged to the original resolution. Then, the regression analysis module of the random forest model is used to establish a nonlinear function mapping relationship between the soil temperature and soil moisture proxy variables and the soil temperature and soil moisture data to obtain the downscaling model.

[0012] Prediction phase: Based on the assumption of scale invariance, high-resolution soil temperature and soil moisture proxy variables of the study area are substituted into the downscaling model to obtain high-resolution downscaling data of soil temperature and soil moisture.

[0013] Furthermore, before constructing the downscaling model during the training phase, soil temperature and soil moisture proxy variables are determined separately. The proxy variables are then aggregated and averaged to the original resolution of the soil temperature and soil moisture products. The soil temperature and soil moisture datasets are divided into training and testing datasets according to a preset ratio. The aggregated and averaged coarse-resolution soil temperature and soil moisture proxy variable datasets are input into the random forest model, and regression analysis is performed using the training data of the corresponding low-resolution dataset to be downscaled to obtain the downscaling model.

[0014] Furthermore, the core idea of ​​the downscaling model is expressed as follows:

[0015] ST c =f((R) c ), (LSTc ), (T c ), (LC c ))

[0016] SM c =f((R) c ), (LST c ), (T c ), (LC c ))

[0017] Wherein, the subscript c represents coarse resolution, ST represents soil temperature, SM represents soil moisture, R represents the index calculated from the MODIS surface reflectance product; LST represents the MODIS surface temperature product; T represents the topographic factor; LC represents the MODIS land cover product; and the functional relationship f(·) represents the nonlinear relationship between soil temperature or soil moisture and the corresponding proxy variable.

[0018] Furthermore, in the prediction stage, the original resolution proxy variables are input into the downscaling model to calculate high-resolution soil temperature and soil moisture estimates. The corresponding soil temperature and soil moisture residuals are then added to the obtained estimates to obtain the high-resolution soil temperature and soil moisture downscaling data, as shown in the following formula:

[0019] ST f =f((R) f ), (LST f ), (T f ), (LC f ))+ΔST c

[0020] SM f =f((R) f ), (LST f ), (T f ), (LC f ))+ΔSM c

[0021] In the formula, the subscript f represents high resolution, corresponding to the coarse resolution c in the downscaling model construction process; (ST c and ΔSM c These represent the residuals of soil temperature and soil moisture, respectively, and are used to indicate the deviation between the actual soil temperature and soil moisture and the soil temperature and soil moisture predicted by the downscaling model.

[0022] Furthermore, time-domain interpolation was performed on the downscaled soil temperature and soil moisture data to obtain downscaled soil temperature and soil moisture data consistent with the time of SAR image acquisition.

[0023] Furthermore, in the frozen soil deformation model in S3 that takes into account soil environmental factors, the surface deformation in the frozen soil region consists of two parts: linear subsidence and seasonal periodic deformation.

[0024]

[0025] Wherein, subscript d LP denoted as low-frequency deformation, t represents the time accumulation between the SAR image acquisition time and the reference image acquisition time, (x,r) are the image coordinates of the high coherence point, α1 represents the linear rate of permafrost deformation, β1 and β2 represent the parameters to be determined for soil temperature and soil moisture factors, respectively, and ST and SM represent the soil temperature and soil moisture data of the high coherence point at different times, respectively.

[0026] Furthermore, in S4, a phase equation is established based on the interferogram data:

[0027]

[0028] in, Let represent the unwrapped interferogram data of the i-th interferogram, i = 1, 2, ..., M, where M represents the total number of interferograms, λ represents the radar wavelength, and B... ⊥i Let θ represent the vertical baseline length of the i-th interferogram, Δh(x,r) represent the elevation residual of the high coherence point at coordinates (x,r), and θ i R is the radar incident angle. i The slant distance of the line of sight. Let t be the residual phase in the i-th interferogram, composed of atmospheric delay, high-frequency deformation, and noise. A and t B Indicates time;

[0029] Transforming the above equation to represent the phases of all interferograms in matrix form, we obtain the observation equation:

[0030]

[0031] Where P=[α1(x,r),β1(x,r),β2(x,r),Δh] T , represents the parameter vector matrix to be determined; This represents a vector matrix consisting of the phases in all interferograms. B is the coefficient matrix, expressed as follows:

[0032]

[0033] After obtaining the parameters to be determined, they are substituted into the phase equation to obtain the residual phase components. Composed of high-frequency deformation phase, atmospheric delay phase, and noise phase, it is expressed as follows:

[0034]

[0035] in and This represents the atmospheric delay phase portion of two SAR images in an interferometric pair. and They represent t respectively A and t B The cumulative amount of high-frequency components of the deformation phase in the LOS direction relative to the reference time t0, Δn i (x,r) represents the noise phase;

[0036] The residual phase component is filtered to remove atmospheric delay phase and noise phase, and the high-frequency deformation phase component is obtained. The low-frequency deformation phase component in the permafrost deformation is obtained. The low-frequency deformation phase component is added to the high-frequency deformation phase component and multiplied by the phase-deformation conversion coefficient λ / 4π. The result is converted into radar line-of-sight deformation to obtain the deformation time series of the study area and the time-series deformation result of the target permafrost area.

[0037] Furthermore, in S4, the least squares method is used to solve for the parameter vector matrix to be determined:

[0038]

[0039] The parameters to be determined are obtained.

[0040] The above-described solution of the present invention has the following beneficial effects:

[0041] The InSAR permafrost deformation monitoring method provided by this invention, which takes into account soil environmental factors, can establish the physical relationship between permafrost deformation and soil hydrothermal conditions, effectively improving the accuracy of permafrost deformation monitoring. Based on soil temperature and humidity data, a permafrost deformation model is constructed, establishing the physical relationship between permafrost deformation and soil hydrothermal conditions, thus better describing the permafrost deformation process. Considering that existing soil parameter products have poor resolution and cannot adequately reflect the spatial characteristics of surface deformation, a high spatial resolution downscaled soil temperature and humidity data is obtained based on a random forest model to construct the permafrost deformation model. This reduces the influence of the deformation model parameter resolution on the deformation solution results. Solving the observation equations based on this reduces the accuracy of time-series deformation results, effectively improving the accuracy of permafrost deformation monitoring.

[0042] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0043] Figure 1 This is a flowchart of the steps of the present invention;

[0044] Figure 2 A diagram verifying the accuracy of downscaling data for soil temperature and soil moisture provided in an embodiment of this application;

[0045] Figure 3 The following diagrams are provided for the solution coefficients and elevation residuals of the frozen soil deformation model considering soil environmental factors in an embodiment of this application, wherein (a) is the estimated value of the linear term coefficient, (b) is the estimated value of the soil temperature coefficient, (c) is the estimated value of the soil moisture coefficient, and (d) is the estimated value of the DEM residual.

[0046] Figure 4 A graph showing the measured data from two leveling monitoring stations provided in an embodiment of this application and the deformation results of the InSAR frozen soil deformation monitoring method that takes into account soil environmental factors. Detailed Implementation

[0047] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0048] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0049] It should also be noted that the illustrations provided in the following embodiments are merely schematic representations of the basic concept of this disclosure. The illustrations only show components relevant to this disclosure and are not drawn according to the actual number, shape, and size of components in implementation. In actual implementation, the type, quantity, and proportion of each component can be arbitrarily changed, and the component layout may be more complex. Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0050] The following is an exemplary description of the InSAR frozen soil deformation monitoring method that takes into account soil environmental factors provided in this application.

[0051] like Figure 1 As shown, an embodiment of the present invention provides an InSAR method for monitoring frozen soil deformation that takes into account soil environmental factors, comprising the following steps:

[0052] S1, based on the random forest model, uses multiple MODIS and DEM products to downscale ERA5_Land soil temperature data and SMAP_L4 soil moisture data to obtain high-resolution downscaled soil temperature and soil moisture data.

[0053] In this embodiment, the downscaling of soil temperature and soil moisture data based on the random forest model is mainly divided into two stages:

[0054] Training phase: First, the soil temperature and soil moisture proxy variables from MODIS and DEM products are aggregated and averaged to the original resolution (9km), which is consistent with the original resolution of ERA5_Land soil temperature data and SMAP_L4 soil moisture data. Then, the regression analysis module of the random forest model is used to establish a nonlinear function mapping relationship between the soil temperature and soil moisture proxy variables and the soil temperature and soil moisture data to obtain the downscaling model.

[0055] For example, the coverage of the above products includes the entire Qinghai-Tibet Engineering Corridor region. This region covers the subsequent soil temperature and soil moisture downscaling data verification area (Nagqu soil temperature and soil moisture data monitoring network) and the InSAR permafrost deformation monitoring area (Wudaoliang area of ​​Qinghai-Tibet Plateau, 34.86°~35.49°N, 92.58°~93.55°E).

[0056] It should be noted that the core idea of ​​downscaling based on the random forest model in this embodiment is to establish a nonlinear regression model between the dataset to be downscaled and the corresponding proxy variables. Therefore, before constructing the soil temperature and soil moisture downscaling models, it is necessary to determine the proxy variables for soil temperature and soil moisture respectively. As a nonlinear statistical ensemble regression algorithm, the training result of the random forest model is the voting result of all decision trees. The importance evaluation module, as the module that implements the feature selection function in the random forest model, plays a crucial role. It can rank the importance of the input proxy variables, thereby determining the best proxy variable for the dataset to be downscaled. There are two criteria for selecting proxy variables: a high correlation between a single proxy variable and the dataset to be downscaled; and the cross-correlation between proxy variables should be as low as possible. Since soil temperature and soil moisture in the permafrost region of the Qinghai-Tibet Plateau are affected by many complex factors, it is crucial to select appropriate proxy variables for soil temperature and soil moisture downscaling in different regions. The selection of features for soil temperature and soil moisture downscaling theoretically involves many factors, such as topographic factors such as elevation, slope, and aspect; environmental physical factors such as surface temperature and vegetation cover; and factors such as water distribution and land cover type. After determining the proxy variables for soil temperature and soil moisture, the proxy variables are first aggregated and averaged to the original resolution (9km) of the soil temperature and soil moisture products, respectively. Next, the soil temperature and soil moisture datasets are divided into training and testing datasets according to a certain ratio. Finally, the aggregated and averaged coarse-resolution datasets of soil temperature and soil moisture proxy variables are input into a random forest model, and regression analysis is performed using the training data from the corresponding low-resolution dataset to be downscaled. This yields the downscaling model, the core idea of ​​which can be expressed as follows:

[0057] ST c =f((R) c ), (LST c ), (T c ), (LC c ))

[0058] SM c =f((R) c ), (LST c ), (T c ), (LC c ))

[0059] Wherein, the subscript c represents coarse resolution, ST represents soil temperature, SM represents soil moisture, R represents indices such as NDVI (Normalized Difference Vegetation Index) and MNDWI (Normalized Difference Water Index) calculated from MODIS surface reflectance products (MOD09A1 and MYD09A1); LST represents MODIS surface temperature products (MOD11A2 and MYD11A2); T represents topographic factors such as DEM, slope, and aspect; LC represents MODIS land cover product (MCD12Q1); and the functional relationship f(·) represents the nonlinear relationship between soil temperature or soil moisture and the corresponding proxy variable, i.e., the downscaling model.

[0060] In this embodiment, various auxiliary data were used in the downscaling process, including datasets such as MOD09A1, MOD11A2, and MCD12Q1 from MODIS products, as well as elevation and slope data from DEM products.

[0061] Prediction phase: Based on the scale invariance assumption, the high-resolution (1km) soil temperature and soil moisture proxy variables of the study area were substituted into the downscaling model to obtain high-resolution downscaled data of soil temperature and soil moisture.

[0062] It should be noted that the main task of the prediction stage is to obtain the soil temperature and soil moisture residuals after obtaining a downscaling model with good generalization. Based on this, and according to the principle of scale invariance, the low-resolution soil temperature and soil moisture data can be downscaled to high-resolution soil temperature and soil moisture data. Specifically, firstly, the original resolution proxy variables are input into the downscaling model to calculate the high-resolution soil temperature and soil moisture estimates; then, the corresponding soil temperature and soil moisture residuals are added to the obtained estimates to obtain the high-resolution downscaled soil temperature and soil moisture data, as shown in the following formula:

[0063] ST f =f((R) f ), (LST f ), (T f ), (LC f ))+ΔST c

[0064] SM f =f((R) f ), (LST f ), (T f ), (LC f ))+ΔSM c

[0065] In the formula, the subscript f represents high resolution, corresponding to the coarse resolution c in the downscaling model construction process; ΔST cand ΔSM c These represent the residuals of soil temperature and soil moisture, respectively, and are used to indicate the deviation between the actual soil temperature and soil moisture and the soil temperature and soil moisture predicted by the downscaling model.

[0066] To verify the accuracy of the downscaling results for soil temperature and soil moisture, this embodiment uses data from 56 monitoring stations of the Nagqu Soil Temperature and Moisture Monitoring Network downloaded from the National Tibetan Plateau Scientific Data Center as verification data to verify the accuracy of the downscaling results. The verification results are as follows: Figure 2 As shown in the analysis, (1) the correlation coefficient between ERA5_Land soil surface temperature and the measured value is 0.9749, the root mean square error (RMSE) is 2.3581, and the average deviation is only 0.51141℃. This indicates that in the Nagqu monitoring network area, the downscaled data of ERA5_Land soil surface temperature has good consistency with the measured value, which also illustrates its applicability in the permafrost region of the Qinghai-Tibet Plateau to a certain extent; (2) the consistency between SMAP_L4 soil surface moisture product and the measured value is also high (R≈0.7071, RMSE≈0.0633), and the deviation between the two is small. Therefore, it can be seen that the downscaled data of ERA5_Land soil temperature and the downscaled data of SMAP_L4 soil moisture have high accuracy in the permafrost region of the Qinghai-Tibet Plateau.

[0067] Before constructing a permafrost deformation model using downscaled data, it is necessary to consider the inconsistency in the acquisition time resolution between soil temperature, soil moisture data, and SAR data. The time resolutions of soil temperature and soil moisture products are 1 hour and 3 hours, respectively; however, after downscaling, due to the influence of the time resolution of auxiliary products such as MODIS, the time resolution of the downscaled soil temperature and soil moisture products is 8 days; the time resolution of Sentinel-1 SAR imagery is 12 days. Therefore, in this embodiment, time-domain interpolation is performed on the downscaled soil temperature and soil moisture data to obtain downscaled soil temperature and soil moisture data consistent with the acquisition time of the SAR imagery.

[0068] S2 processes N SAR image data covering the target permafrost region to obtain M interferograms after phase unwrapping.

[0069] For example, the SAR image data mentioned above is from the Sentinel-1 satellite, which contains 112 interferometric wide swath (IW) up-orbit images with a range resolution of approximately 5 meters and an azimuth resolution of approximately 20 meters, spanning from March 21, 2017 to December 30, 2020.

[0070] In the SBAS processing of Sentinel-1 SAR data, the SAR image from February 3, 2019, was first used as the master image, and all acquired SAR images were registered to the same coordinate system. Then, considering the coherence of the interferometric pairs, the time baseline threshold was set to 36 days and the spatial baseline threshold was set to 150m, resulting in 259 interferograms, with 10 range and 2 azimuth multilooks. After that, the interferometric pairs were processed using pre-acquired precision orbit data to remove orbital errors and flat terrain effects, and the terrain phase was removed using 30m resolution SRTM DEM data. Finally, the phase unwrapping method was used, and the interferometric pairs were further screened based on the phase unwrapping quality, resulting in 171 interferograms with high coherence and unwrapping quality.

[0071] S3, based on high-resolution downscaling data of soil temperature and moisture, a model is built for each high-coherence point in the interferogram that meets the preset coherence conditions, resulting in a permafrost deformation model that takes into account soil environmental factors. It should be noted that the permafrost deformation model considering soil environmental factors in this embodiment assumes that surface deformation in the permafrost region consists of two parts: linear subsidence and seasonal periodic deformation.

[0072]

[0073] Wherein, subscript d LP denoted as low-frequency deformation, t represents the time accumulation between the SAR image acquisition time and the reference image acquisition time, (x,r) are the image coordinates of the high coherence point, α1 represents the linear rate of permafrost deformation, β1 and β2 represent the parameters to be determined for soil temperature and soil moisture factors, respectively, and ST and SM represent the soil temperature and soil moisture data of the high coherence point at different times, respectively.

[0074] S4. Each interferogram is processed using the permafrost deformation model to obtain the phase values ​​of high coherence points in all interferograms. The deformation time series results of all high coherence points are integrated to obtain the deformation results of the target permafrost region.

[0075] Specifically, the phase equation is established based on the interferogram data:

[0076]

[0077] in, Let represent the unwrapped interferogram data of the i-th interferogram, i = 1, 2, ..., M, where M represents the total number of interferograms, λ represents the radar wavelength, and B... ⊥i Let αh(x,r) represent the vertical baseline length of the i-th interferogram, αh(x,r) represent the elevation residual of the high coherence point at coordinates (x,r), and θ represent the vertical baseline length of the i-th interferogram. i R is the radar incident angle. i The slant distance of the line of sight. Let t be the residual phase in the i-th interferogram, composed of atmospheric delay, high-frequency deformation, and noise. A and t B Indicates time;

[0078] Transforming the above equation to represent the phases of all interferograms in matrix form, we obtain the observation equation:

[0079]

[0080] Where P=[α1(x,r),β1(x,r),β2(x,r),Δh] T , represents the parameter vector matrix to be determined; This represents a vector matrix consisting of the phases in all interferograms. B is the coefficient matrix, which can be expressed as follows:

[0081]

[0082] Solving for the parameter vector matrix using the least squares method:

[0083]

[0084] After obtaining the parameters to be determined, substituting them into the phase equation will yield the residual phase components. It consists of high-frequency deformation phase, atmospheric delay phase, and noise phase, and can be expressed as follows:

[0085]

[0086] in and This represents the atmospheric delay phase portion of two SAR images in an interferometric pair. and They represent t respectively A and t B The cumulative amount of high-frequency components of the deformation phase in the LOS direction relative to the reference time t0, Δn i (x,r) represents the noise phase.

[0087] Due to residual phase Each component in the model exhibits different spatiotemporal characteristics. Filtering effectively removes atmospheric delay and noise phases, allowing the acquisition of high-frequency deformation phase components, which in turn yields low-frequency deformation phase components in permafrost deformation. Adding the low-frequency and high-frequency deformation phase components and multiplying by the phase-deformation conversion coefficient λ / 4π converts the result to radar line-of-sight deformation, thus obtaining the deformation time series of the study area, i.e., the temporal deformation results of the target permafrost region. As described above, this scheme primarily focuses on improving the process of establishing observation equations based on high coherence points. It replaces the linear model used in the traditional SBAS-InSAR method with a permafrost deformation model that considers soil environmental factors. Furthermore, considering the spatial differences in soil hydrothermal conditions and the local characteristics of surface deformation in permafrost regions, high-resolution downscaling data of soil temperature and moisture are used to replace the original low-resolution data as input parameters for the deformation model, further reducing model errors and improving the accuracy of surface deformation inversion results in the study area.

[0088] The above steps will be illustrated with a specific example below.

[0089] Points with an average coherence greater than 0.5 and a minimum coherence greater than 0.4 across all coherence maps were selected as high coherence points. At each high coherence point, a set of equations for a frozen soil deformation model, taking into account soil environmental factors, was established. The least squares method was used to solve for the frozen soil deformation model coefficients and residual phase. The obtained frozen soil deformation model coefficients and elevation residuals, taking into account soil environmental factors, are shown below. Figure 3 As shown.

[0090] Figure 3 (a) represents the long-term deformation rate of the study area. As can be seen from the figure, between March 2017 and December 2020, the surface deformation of the Wudaoliang permafrost area was mainly subsidence. In particular, the surface subsidence was obvious in the Wudaoliang area, the Beiluhe Basin and the area around the salt lake in the northeast of the study area, with the maximum linear subsidence rate exceeding -25 mm / a. Figure 3(b) and (c) show the estimated soil temperature coefficient and soil moisture coefficient obtained from the deformation model, respectively, characterizing the degree of influence of changes in soil hydrothermal conditions on surface deformation in permafrost regions. As can be seen from the figures, the impact of changes in soil hydrothermal conditions on permafrost deformation exhibits significant spatial differences. This is because surface deformation in permafrost regions is not only affected by changes in soil hydrothermal conditions but also by factors such as soil type, land cover type, and topography. These factors, along with soil hydrothermal conditions, jointly influence surface deformation in permafrost regions. The spatial differences in these factors lead to varying sensitivities of permafrost deformation to changes in soil hydrothermal conditions across different regions. Within the study area, the rate of surface deformation in permafrost regions caused by changes in soil temperature ranges from -0.8 to 0.4 mm / ℃, while the rate of surface deformation caused by changes in soil moisture ranges from -150 to 150 mm / (m³ / m³). Surface deformation in the permafrost region surrounding the salt lake in the northeastern part of the study area is particularly sensitive to changes in soil hydrothermal conditions. Furthermore, permafrost deformation in the Beiluhe Basin in the southern part of the study area is also significantly affected by changes in soil temperature and moisture. Figure 3 (d) shows the estimated DEM residual values ​​obtained from the solution. As can be seen from the figure, most of the DEM residuals in the study area are between -10 and 10m, which is consistent with the relative accuracy of the 30-meter resolution SRTM DEM product. The small phase residual also indicates that the permafrost deformation model proposed in this scheme can better describe the surface deformation process in the permafrost region.

[0091] To verify the reliability of this method, the deformation monitoring results of two leveling monitoring stations (OP7 and OP8) located in the study area were compared with the deformation results of the InSAR frozen soil deformation monitoring method considering soil environmental factors provided in this application. The comparison curves are shown in the figure below. Figure 4 As shown in the figure, OP7 and OP8 represent the deformation results of two leveling monitoring stations, respectively. These results are used as measured data to verify and compare the accuracy of the most commonly used periodic model in the InSAR permafrost deformation monitoring field and the method provided in this application. Figure 4 As can be seen, at site OP7, the root mean square error (RMSE) of the periodic model curve is 22.13 mm, while the RMSE of this model is only 6.18 mm; at site OP8, the RMSE of the periodic model curve is 36.19 mm, while the RMSE of this model is 13.63 mm. The horizontal axis in the figure represents the date, and the vertical axis represents the deformation result, with units in millimeters (mm).

[0092] It can be seen that at site OP7, the InSAR frozen soil deformation model considering soil environmental factors proposed in this scheme is far superior to the traditional periodic model, with a root mean square error reduction of approximately 72.07%. At site OP8, the root mean square error of the deformation results obtained by the proposed model is approximately 62.34% lower than that of the periodic model. These conclusions demonstrate that the InSAR frozen soil deformation monitoring method considering soil environmental factors provided by this invention can effectively improve the accuracy of frozen soil deformation monitoring. This is because the method proposed in this invention constructs a frozen soil deformation model based on soil temperature and soil moisture data, which can establish the physical relationship between frozen soil deformation and soil hydrothermal conditions, thereby better describing the frozen soil deformation process. Furthermore, it fully considers the problem that existing soil parameter products, due to their poor resolution, cannot adequately account for the spatial characteristics of surface deformation. It uses a random forest model to obtain high spatial resolution downscaled soil temperature and moisture data to construct the frozen soil deformation model, reducing the influence of the deformation model parameter resolution on the deformation solution results, and based on this, constructs and solves the observation equations.

[0093] It can be seen that the method provided by the present invention can significantly increase the accuracy of time-series deformation results and effectively improve the accuracy of deformation monitoring in permafrost regions.

[0094] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An InSAR method for monitoring frozen soil deformation that takes into account soil environmental factors, characterized in that, Includes the following steps: S1, based on the random forest model, uses multiple MODIS and DEM products to downscale soil temperature and soil moisture data to obtain high-resolution downscaled soil temperature and soil moisture data. S1 consists of two phases: Training phase: First, the soil temperature and soil moisture proxy variables from MODIS and DEM products are aggregated and averaged to the original resolution. Then, the regression analysis module of the random forest model is used to establish a nonlinear function mapping relationship between the soil temperature and soil moisture proxy variables and the soil temperature and soil moisture data to obtain the downscaling model. Prediction phase: Based on the scale invariance assumption, the high-resolution soil temperature and soil moisture proxy variables of the study area are substituted into the downscaling model to obtain high-resolution downscaled soil temperature and soil moisture data. S2, for covering the target permafrost area N Each SAR image data is processed to obtain phase-unwrapped data. M An interference diagram; S3, based on high-resolution soil temperature and soil moisture downscaling data, modeling is performed for each high coherence point in the interferogram that meets the preset coherence conditions, to obtain a frozen soil deformation model that takes into account soil environmental factors. In the S3 permafrost deformation model that takes into account soil environmental factors, the surface deformation in the permafrost region consists of two parts: linear subsidence and seasonal periodic deformation. Among them, subscript Indicates low-frequency deformation. This represents the cumulative time between the time the SAR image was acquired and the time the reference image was acquired. The image coordinates of the high coherence points. The linear rate of permafrost deformation is represented by... and Let represent the parameters to be determined for soil temperature and soil moisture factors, respectively. and These represent the soil temperature and soil moisture data at different times at the high coherence point; S4. Each interferogram is processed using the permafrost deformation model to obtain the phase values ​​of high coherence points in all interferograms. The deformation time series results of all high coherence points are integrated to obtain the deformation results of the target permafrost region.

2. The InSAR frozen soil deformation monitoring method considering soil environmental factors according to claim 1, characterized in that, Before building the downscaling model during the training phase, proxy variables for soil temperature and soil moisture are determined separately. The proxy variables are then aggregated and averaged to the original resolution of the soil temperature and soil moisture products. The soil temperature and soil moisture datasets are then divided into training and testing datasets according to a preset ratio. The aggregated and averaged coarse-resolution proxy variable datasets of soil temperature and soil moisture are input into the random forest model, and regression analysis is performed using the training data of the corresponding low-resolution dataset to be downscaled to obtain the downscaling model.

3. The InSAR frozen soil deformation monitoring method considering soil environmental factors according to claim 2, characterized in that, The core idea of ​​the downscaling model is expressed as follows: Among them, subscript Represents coarse resolution. Indicates soil temperature, Indicates soil moisture. This represents the index calculated from MODIS surface reflectance products; Represents MODIS surface temperature products; Represents terrain factors; Indicates MODIS land cover products; functional relationship This indicates the nonlinear relationship between soil temperature or soil moisture and the corresponding proxy variable.

4. The InSAR frozen soil deformation monitoring method considering soil environmental factors according to claim 3, characterized in that, In the prediction phase, the original resolution proxy variables are input into the downscaling model to calculate high-resolution soil temperature and soil moisture estimates. The corresponding soil temperature and soil moisture residuals are then added to the estimated data to obtain the high-resolution downscaled soil temperature and soil moisture data, as shown in the following formula: In the formula, the subscript Represents high resolution, compared to coarse resolution in the downscaling model building process. correspond; and These represent the residuals of soil temperature and soil moisture, respectively, and are used to indicate the deviation between the actual soil temperature and soil moisture and the soil temperature and soil moisture predicted by the downscaling model.

5. The InSAR frozen soil deformation monitoring method considering soil environmental factors according to claim 4, characterized in that, The downscaled soil temperature and soil moisture data were interpolated in the time domain to obtain downscaled soil temperature and soil moisture data consistent with the time of SAR image acquisition.

6. The InSAR frozen soil deformation monitoring method considering soil environmental factors according to claim 5, characterized in that, In S4, the phase equation is established based on the interferogram data: in, Indicates the first Unwrapped interferogram data of an interferogram. =1,2,..., M , M This indicates the total number of interferograms. Indicates the radar wavelength. Indicates the first The vertical baseline length of an interferogram Indicates coordinates as High coherence point elevation residuals For radar incident angle, The slant distance of the line of sight. For the first The residual phase in the interferogram consists of atmospheric delay, high-frequency deformation, and noise. and Indicates time; Transforming the above equation to represent the phases of all interferograms in matrix form, we obtain the observation equation: in , represents the parameter vector matrix to be determined; This represents a vector matrix consisting of the phases in all interferograms. ; The coefficient matrix is ​​expressed as follows: After obtaining the parameters to be determined, they are substituted into the phase equation to obtain the residual phase components. , Composed of high-frequency deformation phase, atmospheric delay phase, and noise phase, it is expressed as follows: in and This represents the atmospheric delay phase portion of two SAR images in an interferometric pair. and They represent and Time relative to reference time Accumulation of high-frequency components of deformation phase in the LOS direction Indicates the noise phase; The residual phase component is filtered to remove atmospheric delay phase and noise phase, obtaining the high-frequency deformation phase component. This yields the low-frequency deformation phase component in the permafrost deformation. The low-frequency deformation phase component is then added to the high-frequency deformation phase component and multiplied by the phase-deformation conversion coefficient. The results are converted into radar line-of-sight deformation to obtain the deformation time series of the study area, thus obtaining the time-series deformation results of the target permafrost region.

7. The InSAR frozen soil deformation monitoring method considering soil environmental factors according to claim 6, characterized in that, In S4, the least squares method is used to solve for the parameter vector matrix: The parameters to be determined are obtained.

Images