Land degradation supervision method based on remote sensing monitoring

Abstract

The invention discloses a land degradation supervision method based on remote sensing monitoring, and relates to the technical field of land degradation monitoring. The method is used for solving the problems of hidden degradation risk identification and dynamic propagation simulation. According to a multi-temporal thermal infrared remote sensing image, daily variation characteristics of surface temperature are extracted through spectral analysis, a recessive salinization risk distribution map is constructed by combining measured data of soil salinity, and interference of vegetation coverage is overcome. And multi-source remote sensing and geographic data are fused, dominant driving factors significantly associated with the salting risk are screened, and the degeneration cause is clarified. And in combination with ecological restoration potential and treatment resource constraints, calculating an antagonism index of a driving factor and the restoration potential, and dividing degradation threat levels. And a natural hydrology and human activity dual-path propagation model is constructed, a risk space-time evolution process is simulated, and a future hotspot migration path and an optimal blocking window are predicted. According to the method, accurate early warning and active prevention and control of land degradation are realized through a technical chain of risk identification, driving analysis, threat grading and dynamic simulation.

Description

Technical Field

[0001] The present invention relates to the technical field of land degradation monitoring, and specifically to a land degradation supervision method based on remote sensing monitoring. Background Art

[0002] With the impact of global climate change and human activities, land degradation has become a severe problem faced globally. Especially in arid and semi-arid regions, land degradation has had a profound impact on the ecological environment, agricultural production, and social economy. Land degradation not only reduces the productivity of the land but also leads to a decline in biodiversity, increased soil erosion, and the loss of ecological service functions. To address this challenge, accurately monitoring the degree and change trend of land degradation, timely detecting the risk points of degradation, and taking effective control measures have become the urgent needs of governments and research institutions in various countries. As an efficient, non-destructive, and widely covered monitoring means, remote sensing technology has played an important role in land degradation monitoring, prediction, and assessment. In particular, data such as surface temperature and vegetation cover obtained through remote sensing image analysis can effectively reveal the spatial distribution characteristics and development dynamics of land degradation.

[0003] Although remote sensing monitoring has been widely used in the field of land degradation, the existing land degradation monitoring methods still have certain deficiencies. First, most of the existing methods focus on the single-index analysis of vegetation cover change or surface temperature, and cannot comprehensively utilize multi-dimensional and multi-source data, resulting in incomplete identification of land degradation driving factors and inability to fully reflect the diversity of degradation in a complex ecological environment. Second, many methods mainly rely on obvious degradation phenomena for monitoring, ignoring the long-term impact of hidden degradation phenomena on land degradation, which leads to the neglect of land degradation risks in some regions. The existing methods lack dynamic simulation of spatio-temporal evolution in degradation risk prediction and cannot accurately identify the migration path, outbreak threshold, and optimal control timing of future risks, thus affecting the timeliness and effectiveness of land degradation control measures. Therefore, there is an urgent need for a new method that can combine remote sensing technology and multi-source data analysis to comprehensively evaluate the risk of land degradation and conduct dynamic simulation in the spatio-temporal dimension to provide a scientific basis for land management. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a land degradation supervision method based on remote sensing monitoring, which solves the problems in the above background art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A land degradation supervision method based on remote sensing monitoring, comprising the following steps: S1. Collect multi-temporal thermal infrared remote sensing historical images of the target area, construct a pixel-by-pixel daily variation time series curve of surface temperature, apply spectral analysis method to the curve to extract thermal disturbance frequency and amplitude indicators, identify areas with high-frequency thermal anomalies within the year but without significant vegetation degradation, and perform regression fitting in combination with limited measured soil salinity data to form a hidden saline-alkali risk spatial distribution map; S2. According to the hidden saline-alkali risk spatial distribution map, fuse multi-source remote sensing data and auxiliary geographical data of the target area to construct a multi-dimensional driving factor raster atlas, and through spatial overlay analysis of the hidden saline-alkali risk spatial distribution map and each factor raster atlas, screen out the dominant driving factors significantly associated with the hidden saline-alkali risk spatial distribution as the land degradation driving factors of the target area; S3. According to the driving factors of land degradation, introduce regional ecological restoration potential indicators and governance resource constraint conditions, inversely associate the contribution intensity of the driving factors with the restoration potential to generate a restoration antagonism index, and based on the spatial distribution of the antagonism index, divide the potential threat levels of land degradation in the target area; S4. According to the potential threat levels of land degradation in the target area, fuse natural hydrological paths and human activity interference paths to construct a dual-path degradation propagation network model, use the threat level as the node weight, and simulate the slow release and diffusion of risks along the natural path and the accelerated migration along the human path through a spatio-temporal hybrid algorithm to generate a spatio-temporal evolution probability map of degradation risks, and mark the migration path, outbreak threshold and optimal blocking window period of future risk hotspots.

[0006] Further, the steps of collecting multi-temporal thermal infrared remote sensing historical images of the target area and constructing a pixel-by-pixel daily variation time series curve of surface temperature are as follows: Register and radiometrically calibrate the obtained remote sensing images in chronological order, and screen out representative daytime thermal infrared band data; Extract multi-temporal surface temperature values of the same day for each remote sensing pixel to form a single-day temperature curve, and superimpose it within the annual scale to form a pixel-by-pixel daily variation time series matrix; Through outlier removal, time resampling and multi-temporal normalization processing, enhance the curve stability and comparability, and construct a time series curve dataset that can reflect the dynamic characteristics of pixel thermal response.

[0007] Furthermore, the frequency spectrum analysis method is applied to the curve to extract the frequency and amplitude indicators of thermal disturbances, and the steps for identifying areas with high-frequency thermal anomalies but no significant vegetation degradation within the year are as follows: discrete Fourier transform or wavelet decomposition is performed on the daily surface temperature change time series curve of each pixel to extract the high-frequency components and dominant amplitude indicators in its frequency domain structure; a thermal disturbance spectrum feature map is constructed to identify areas with high-frequency periodic thermal anomalies on an annual scale; the vegetation index change image of the same period is integrated to eliminate the thermal disturbance areas caused by vegetation changes, and the target areas with independent thermal anomaly response characteristics but no significant vegetation degradation are screened out as suspected hidden salinization risk areas, and sporadic interference patches are eliminated through spatial clustering to generate hidden thermodynamic imbalance target areas.

[0008] Furthermore, regression fitting is performed in combination with limited measured soil salt data to form a spatial distribution map of implicit salinity risk. The steps are as follows: within the identified high-frequency thermal anomaly target area, multi-phase thermal disturbance spectral characteristics and auxiliary remote sensing indicators are extracted as regression factors, including surface dryness and humidity, vegetation coverage, and the high-frequency thermal anomaly target area is matched with the spatial position of the measured soil salinity point to construct a nonlinear regression model of thermal disturbance amplitude and salt concentration. The salinity risk of unsampled areas is extrapolated through a spatial interpolation algorithm, and the interpolation weights are optimized by integrating geographic constraints to generate a spatial distribution map of implicit salinity risk in the entire region, marking high-risk clusters and potential diffusion directions.

[0009] Furthermore, the steps of fusing the multi-source remote sensing data and auxiliary geographic data of the target area to construct a multidimensional driving factor raster map are as follows: integrating the multispectral vegetation stress index, radar-inverted soil moisture time series data, nighttime light intensity spatial distribution data, and auxiliary geographic data of groundwater level depth and cultivated land use intensity in the target area; and eliminating data scale differences by unifying the spatial resolution and aligning it with the coordinate system to generate a multidimensional driving factor raster map covering vegetation dynamics, hydrological conditions, and human activity intensity.

[0010] Furthermore, through the spatial overlay analysis of the spatial distribution map of implicit salinization risk and the raster maps of each factor, the dominant driving factors that are significantly correlated with the spatial distribution of implicit salinization risk are screened out as the driving factors of land degradation in the target area. The screening logic is as follows: through spatial heterogeneity association analysis, the local spatial correlation between the implicit salinization risk value and each driving factor is calculated pixel by pixel; the spatial differentiation characteristics of the factor contribution intensity are quantified through the geographically weighted regression model, and the driving factors that are significantly correlated both globally and locally and have high spatial consistency are screened out, and the redundant interference factors are eliminated to generate the spatial differentiation map of the dominant driving factors.

[0011] Furthermore, the steps of generating a restoration antagonism index by inversely correlating the contribution intensity of driving factors with the restoration potential are as follows: Normalize the contribution intensity of the selected dominant driving factors and perform non-linear weighted superposition with the regional ecological restoration potential index; Optimize the weight allocation by introducing geographical constraint conditions and generate the restoration antagonism index through weighted calculation.

[0012] Furthermore, based on the spatial distribution of the antagonism index, the steps of dividing the potential threat levels of land degradation in the target area are as follows: Use spatial clustering and natural breakpoint method to perform hierarchical cutting on the antagonism index raster layer, and combine with spatial continuity optimization to divide into three levels of potential threat levels, including high-risk areas, medium-risk areas and low-risk areas.

[0013] Furthermore, according to the potential threat levels of land degradation in the target area, the steps of constructing a dual-path degradation propagation network model by integrating natural hydrological paths and human activity interference paths are as follows: Use the threat level zoning results as the network node attribute weights, construct the natural hydrological diffusion path according to the groundwater flow direction and terrain slope, and construct the human activity interference path according to the cultivated land connectivity and irrigation canal distribution; Define the edge weight between nodes as the weighted superposition of the water-salt flux of the natural path and the pressure intensity of the human path to generate the topological structure of the dual-path degradation propagation network, which characterizes the composite diffusion mechanism of risk along natural and human paths.

[0014] Furthermore, the steps of simulating the slow release diffusion of risk along the natural path and the accelerated migration along the human path through a spatio-temporal hybrid algorithm, generating a spatio-temporal evolution probability map of degradation risk, and marking the migration path, outbreak threshold and optimal blocking window period of future risk hotspots are as follows: Based on the dual-path degradation propagation network model, use the threat level as the node weight, and simulate the dynamic risk propagation process through a spatio-temporal hybrid algorithm. The key steps include: Slow release diffusion of the natural path: Calculate the diffusion rate of salt according to the groundwater flow and terrain water collection ability, and simulate the migration process without intervention; Accelerated migration of the human path: Quantify the accelerated effect of cultivated land connectivity and irrigation canal distribution on risk propagation, and simulate the fast diffusion path caused by human activities; Generation of the probability map: Dynamically update the risk values of each node by iteratively simulating the risk accumulation process in stages through a time sliding window mechanism to generate a spatio-temporal evolution probability map of degradation risk; Marking of migration path and threshold: Extract the continuous areas in the probability map where the risk value exceeds the preset threshold, identify the main migration paths along the natural hydrological network and human paths, and mark the outbreak risk levels of high-risk hotspots; Identification of seasonal blocking windows: Combine the regional hydrological cycle and governance resource scheduling, analyze the temporal matching relationship between the risk propagation rate and the governance response efficiency, and determine the optimal blocking window period.

[0015] The present invention has the following beneficial effects:

[0016] (1) The land degradation supervision method based on remote sensing monitoring collects multi-temporal thermal infrared remote sensing historical images, constructs the daily change time series curve of surface temperature, combines the spectrum analysis method to extract the thermal disturbance frequency and amplitude indexes, effectively identifies potential salinization risk areas, and performs regression fitting with the measured soil data to generate a hidden salinization risk spatial distribution map. This method uses remote sensing data to efficiently and comprehensively monitor the potential risk areas of land degradation, avoids the time and space limitations of traditional ground surveys, and greatly improves the monitoring accuracy of land degradation.

[0017] (2) The land degradation supervision method based on remote sensing monitoring constructs a multi-dimensional driving factor raster map by fusing multi-source remote sensing data and auxiliary geographical data, and combines the spatial overlay analysis of land degradation driving factors to accurately identify the dominant factors significantly associated with land degradation risks, providing a scientific basis for the treatment of land degradation. At the same time, by constructing a dual-path degradation propagation network model and using a spatio-temporal hybrid algorithm to simulate the risk diffusion process, it can generate a spatio-temporal evolution probability map of degradation risks in real time, mark the migration path, outbreak threshold and optimal blocking window period of future risk hotspots, and provide more accurate decision-making support for land management and restoration.

[0018] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flowchart of the land degradation supervision method based on remote sensing monitoring of the present invention.

[0020] Figure 2 It is a flowchart of generating the restoration antagonism index of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] In the embodiments of the present application, the land degradation supervision method based on remote sensing monitoring is used to solve the problems existing in the traditional land degradation monitoring methods, such as limited monitoring scope, poor timeliness, and lack of accurate dynamic risk assessment and warning mechanisms. Through remote sensing technology, it is possible to obtain large-scale land degradation data in real time and comprehensively, and combine spatio-temporal analysis models to accurately predict the potential risks and hotspot areas of land degradation, providing a more scientific and refined basis for governance decision-making.

[0022] The general idea of the solution in the embodiments of the present application is as follows:

[0023] Collect multi-temporal thermal infrared remote sensing historical images of the target area, construct the daily change time series curve of surface temperature for each pixel, apply the spectrum analysis method to the curve to extract the thermal disturbance frequency and amplitude indexes, identify the areas with high-frequency thermal anomalies during the year but without significant vegetation degradation, and perform regression fitting in combination with limited measured soil salinity data to form a hidden salinization risk spatial distribution map.

[0024] According to the spatial distribution map of latent salinization risk, integrate the multi-source remote sensing data and auxiliary geographical data of the target area in the target area, construct a multi-dimensional driving factor raster atlas, and through the spatial overlay analysis of the spatial distribution map of latent salinization risk and each factor raster atlas, screen out the dominant driving factors significantly associated with the spatial distribution of latent salinization risk as the driving factors for land degradation in the target area.

[0025] According to the driving factors of land degradation, introduce the regional ecological restoration potential index and the constraints of governance resources, reverse-correlate the contribution intensity of the driving factors with the restoration potential, generate a restoration antagonism index, and based on the spatial distribution of the antagonism index, divide the potential threat levels of land degradation in the target area.

[0026] According to the potential threat levels of land degradation in the target area, integrate the natural hydrological path and the human activity interference path, construct a dual-path degradation propagation network model, use the threat level as the node weight, and through the spatio-temporal hybrid algorithm, simulate the slow release and diffusion of risks along the natural path and the accelerated migration along the human path, generate a spatio-temporal evolution probability atlas of degradation risks, and mark the migration path, outbreak threshold and optimal blocking window period of future risk hotspots.

[0027] Please refer to Figure 1 This embodiment of the present invention provides a technical solution: a land degradation supervision method based on remote sensing monitoring, including the following steps: S1. Collect multi-temporal thermal infrared remote sensing historical images of the target area, construct a pixel-by-pixel daily change time series curve of the surface temperature, apply the spectral analysis method to the curve to extract the thermal disturbance frequency and amplitude indicators, identify the areas with high-frequency thermal anomalies during the year but without significant vegetation degradation, and perform regression fitting in combination with limited measured soil salinity data to form a spatial distribution map of latent salinization risk; S2. According to the spatial distribution map of latent salinization risk, integrate the multi-source remote sensing data and auxiliary geographical data of the target area in the target area, construct a multi-dimensional driving factor raster atlas, and through the spatial overlay analysis of the spatial distribution map of latent salinization risk and each factor raster atlas, screen out the dominant driving factors significantly associated with the spatial distribution of latent salinization risk as the driving factors for land degradation in the target area; S3. According to the driving factors of land degradation, introduce the regional ecological restoration potential index and the constraints of governance resources, reverse-correlate the contribution intensity of the driving factors with the restoration potential, generate a restoration antagonism index, and based on the spatial distribution of the antagonism index, divide the potential threat levels of land degradation in the target area; S4. According to the potential threat levels of land degradation in the target area, integrate the natural hydrological path and the human activity interference path, construct a dual-path degradation propagation network model, use the threat level as the node weight, and through the spatio-temporal hybrid algorithm, simulate the slow release and diffusion of risks along the natural path and the accelerated migration along the human path, generate a spatio-temporal evolution probability atlas of degradation risks, and mark the migration path, outbreak threshold and optimal blocking window period of future risk hotspots.

[0028] In this implementation plan: S1. Multi-temporal thermal infrared remote sensing historical images: refer to thermal infrared image data obtained by remote sensing satellites or aircrafts that cover a certain period of time. These images can reflect the changes in surface temperature and help analyze the thermal characteristics of land degradation. Pixel-by-pixel daily surface temperature change time series curve: refers to, for each pixel point (i.e., the smallest unit in the image), constructing a time series of daily surface temperature changes based on its thermal infrared data. This can effectively reflect the fluctuation trend of the land surface temperature. Spectrum analysis method: This is a technique for identifying periodic changes by analyzing the frequency components in time series data. This method can extract the frequency and amplitude of thermal disturbances and reveal potential patterns of soil or vegetation degradation. Thermal disturbance frequency and amplitude index: An index extracted through spectrum analysis, representing the frequency and amplitude of surface temperature fluctuations. A relatively high frequency of thermal anomalies is usually caused by the accumulation of saline soils, evaporation, or abnormal water distribution, but not necessarily accompanied by significant vegetation degradation. Latent saline risk spatial distribution map: A map generated through regression fitting by combining the analysis results of remote sensing images and field soil salinity measurement data, used to represent the spatial distribution of latent saline risk within a region. Latent salinity refers to those that do not show signs of degradation on the surface but have potential salinization risks.

[0029] S2. Multi-source remote sensing data: Data from different remote sensing platforms or sensors, such as satellite images, ground measurement data, etc. These data can cover different bands, spatial resolutions, or time scales. Auxiliary geographic data: Include geographic information of the target area, such as land type, climate data, soil type, etc. These data provide more abundant background information for land degradation monitoring. Multi-dimensional driving factor raster atlas: By converting data of multiple driving factors (such as climate, soil, vegetation coverage, etc.) into a raster-form data atlas for spatial analysis. These factors will help identify the key factors affecting land degradation. Spatial overlay analysis: Refers to overlaying the latent salinization risk map and the raster atlas of each driving factor to find out which factors are significantly associated with the salinization risk. This process helps identify the dominant driving factors of land degradation. S3. Ecological restoration potential index: Used to measure the potential of a specific area in ecological restoration, such as soil improvement, vegetation restoration ability, etc. This index helps evaluate the restoration ability of land under different degradation conditions. Governance resource constraint conditions: Refer to the limitations of available resources (such as funds, technology, personnel, etc.) in the process of land governance. These constraint conditions need to be considered in the design of land governance strategies. Restoration antagonism index: An index calculated based on the inverse relationship between driving factors and restoration potential. It reflects the difficulty of land restoration in the face of the influence of certain driving factors. A higher antagonism index may mean that the land in this area is more difficult to restore. Threat level: According to the spatial distribution of the antagonism index, divide the potential threat levels of land degradation within the region. This step helps formulate regional priority governance strategies. S4. Natural hydrological path and human activity interference path: The natural hydrological path refers to the propagation paths of natural factors such as groundwater flow and surface water flow; the human activity interference path refers to the accelerating effect of human factors such as agricultural irrigation and urban construction on land degradation. Dual-path degradation propagation network model: Combine the impacts of natural and human paths to construct a network model containing nodes (representing different regional threat levels) and edges (representing propagation paths). Each node is assigned a different weight according to the threat level of land degradation. Spatiotemporal hybrid algorithm: An algorithm that combines time and space factors to simulate the propagation process of degradation risk on natural and human paths. This algorithm can predict the diffusion and change of risk and give different propagation rates according to different path types (natural or human). Degradation risk spatiotemporal evolution probability atlas: Based on the simulation results, generate a visualized atlas to show the spatial distribution and temporal evolution process of land degradation risk, helping to identify potential future risk hotspots. Migration path, outbreak threshold, and optimal blocking window period: According to the degradation risk atlas, extract the areas where the risk value exceeds a certain threshold to determine the main propagation path of the risk; through the analysis of historical degradation laws, set an outbreak threshold; combine seasonal climate conditions to determine the optimal governance window period (such as restoration measures before the rainy season or governance during the irrigation cycle).

[0030] Specifically, the steps for collecting multi-temporal thermal infrared remote sensing historical images of the target area and constructing the daily change time series curve of surface temperature for each pixel are as follows: Register and radiometrically calibrate the acquired remote sensing images in chronological order, and screen the representative daytime thermal infrared band data; Extract the multi-temporal surface temperature values of the same day for each remote sensing pixel to form a single-day temperature curve, and stack it within the annual scale to form a daily change time series matrix for each pixel; Through outlier removal, time resampling, and multi-temporal normalization processing, enhance the curve stability and comparability, and construct a time series curve dataset that can reflect the dynamic characteristics of pixel thermal response.

[0031] In this implementation plan, remote sensing image registration: Remote sensing images acquired at different times may have geometric offsets due to factors such as shooting angles, sensor positions, and ground position changes. Registration is to align these images through spatial coordinate alignment to ensure their correct alignment in the same geographical framework. This can ensure that subsequent analysis is not affected by geometric errors. The registration process usually involves algorithms such as geometric transformation and affine transformation of the image to ensure that the same ground objects in the image are consistent in space. Radiometric calibration: The temperature information obtained from remote sensing images is often affected by factors such as sensor characteristics, atmospheric conditions, and ground object reflectivity. Radiometric calibration is to convert the digital values in the remote sensing image into real surface temperature values. This process can eliminate the errors caused by the sensor and the external environment in the image, making the obtained surface temperature data have higher accuracy and credibility, and thus supporting subsequent analysis work. Screening thermal infrared bands: Thermal infrared remote sensing images usually contain multiple bands, and each band has a different response ability to surface temperature. The daytime thermal infrared band can more accurately reflect the changes in surface temperature. Therefore, it is necessary to screen out representative thermal infrared band data, which can effectively monitor the temperature changes of ground objects during the day. The purpose of screening is to ensure that the selected band can reflect the surface temperature characteristics of the target area to the greatest extent and avoid interference from irrelevant bands on the analysis results. Extracting multi-temporal surface temperature values of each remote sensing pixel within the same day to form a single-day temperature curve: Extracting multi-temporal surface temperature values: In each remote sensing image, each pixel corresponds to a position on the ground. By extracting the surface temperature of this pixel at different time points within the same day, the "intra-day temperature change" of this pixel can be obtained. This process can capture the temperature fluctuation characteristics at different times and help analyze the rising and falling trends and abnormal changes of daytime temperature. Forming a single-day temperature curve: For each pixel, by extracting the temperature values of this pixel within a day, we can obtain a curve of temperature changing with time. These temperature curves can intuitively reflect the change law of surface temperature within a day, such as typical characteristics such as daytime warming and nighttime cooling. The construction of the temperature curve provides the necessary time series data for subsequent analysis. Stacking within the annual scale to form a daily change time series matrix for each pixel: Annual scale stacking: Within a year, the temperature curves of each pixel will be different on different days. Therefore, it is necessary to stack the single-day temperature curves of different dates throughout the year to form complete annual change time series data. These data provide the complete annual surface temperature change situation for each pixel. Pixel-by-pixel time series matrix: The surface temperature change data of each pixel will form a time series matrix, where each row represents the temperature curve of a certain day, and each column represents the temperature value at a certain time point. In this way, the entire time series matrix provides a complete picture of the daily change of surface temperature for each pixel at the annual scale and can comprehensively reflect the annual fluctuations of land temperature.Outlier removal, time resampling and multi-temporal normalization: Outlier removal: Remote sensing data may contain outliers, which may be caused by sensor errors, atmospheric disturbances or other interference factors. Through outlier removal, these data points that do not conform to the normal change pattern can be eliminated to ensure the representativeness and accuracy of the temperature curve. Common methods include outlier detection based on statistical principles (such as detection based on standard deviation). Time resampling: The data acquisition time interval of remote sensing images is usually inconsistent, and there may be situations where the data acquisition frequency is low or high in different time periods. Time resampling unifies the data at different time points into a fixed time interval by homogenizing the data, which makes the time series data consistent and comparable during analysis. Through resampling, spatiotemporal change analysis can be performed more accurately. Multi-temporal normalization: Due to the different shooting times and environmental factors of remote sensing images, there may be radiation differences in images at different times. Through multi-temporal normalization, remote sensing data at different times can be mapped to a standard range to eliminate the impact caused by time differences. The normalized data is easier to compare and analyze, which enhances the comparability of surface temperature changes at different time points. Enhance the stability and comparability of the curve, and construct a time series curve dataset that can reflect the dynamic characteristics of pixel thermal response: Enhance the stability and comparability of the curve: Through the aforementioned outlier removal, time resampling and normalization processing, the final temperature curve is more stable and consistent. This means that the surface temperature changes at different time points and in different regions can be compared more accurately, reducing the impact of external interference factors. Time series curve dataset reflecting the dynamic characteristics of pixel thermal response: These processed temperature datasets can accurately reflect the thermal response characteristics of each pixel at different time scales, such as seasonal changes, diurnal changes, etc. This dataset provides the necessary basic data support for subsequent land degradation monitoring, salinization risk assessment, etc., and has important application value.

[0032] Specifically, the spectral analysis method is applied to the curve to extract the frequency and amplitude indicators of thermal disturbances, and the steps to identify areas with high-frequency thermal anomalies but no significant vegetation degradation within the year are as follows: discrete Fourier transform or wavelet decomposition is performed on the daily surface temperature change time series curve of each pixel to extract the high-frequency components and dominant amplitude indicators in its frequency domain structure; a thermal disturbance spectrum feature map is constructed to identify areas with high-frequency periodic thermal anomalies on an annual scale; the vegetation index change image of the same period is integrated to eliminate the thermal disturbance areas caused by vegetation changes, and the target areas with independent thermal anomaly response characteristics but no significant vegetation degradation are screened out as suspected hidden salinization risk areas, and sporadic interference patches are eliminated through spatial clustering to generate hidden thermodynamic imbalance target areas.

[0033] In this implementation scheme, the discrete Fourier transform: converts the surface temperature data in the time series into a frequency-domain signal to identify the high-frequency components in the temperature change, and then extracts the thermal disturbance characteristics within a year. Wavelet decomposition: decomposes the temperature curve through wavelet transform, which can more precisely identify the temperature changes on different time scales, especially local high-frequency disturbances. High-frequency component extraction: based on the results after Fourier transform or wavelet decomposition, the high-frequency components in the frequency domain can be extracted. High-frequency components are usually related to temperature fluctuations within a short period, and these fluctuations may be caused by sudden thermal anomalies, short-term climate changes, etc. In the time series of surface temperature, if the amplitude of the high-frequency component is large and does not change with the seasons, it usually means that there may be a thermal anomaly in this area. Dominant amplitude index: from the frequency-domain results, the amplitudes within certain frequency ranges can be selected as the dominant amplitude index, and these amplitudes reflect the intensity and period of the surface temperature fluctuations in this area. Usually, by analyzing the spectrogram of each pixel, the dominant period (i.e., the frequency with a large amplitude) and its corresponding amplitude can be determined. If the amplitude of the high-frequency fluctuation increases abnormally, it usually indicates that there is an abnormal thermal disturbance in this area. Thermal disturbance spectral characteristic map: based on the results of Fourier transform or wavelet transform, a thermal disturbance spectral characteristic map can be constructed for each pixel. This map shows the spectral characteristics of each pixel, especially the changes in the high-frequency fluctuation part. This process can identify the areas with obvious high-frequency periodic thermal anomalies on the annual scale, usually manifested as areas with large and irregular temperature fluctuation amplitudes. By extracting these areas from the overall image, areas with high-frequency thermal disturbances can be found, and these areas may represent potential hidden salinization risk areas, especially in the case of no obvious signs of vegetation degradation. Vegetation index change image: to exclude the temperature fluctuations caused by vegetation changes (such as drought, crop growth season changes, etc.), it is necessary to analyze in combination with the vegetation index image (such as NDVI) of the same period. The vegetation index can reflect the growth status and health level of the surface vegetation. By comparing with the changes in surface temperature, it can be judged whether the temperature fluctuations are caused by vegetation changes. Removing thermal disturbances caused by vegetation: through the correlation analysis with the vegetation index, the thermal disturbance areas caused by vegetation growth, decline or seasonal changes can be removed. This helps to identify the areas where the temperature changes are not caused by vegetation changes, and then ensures the effectiveness of the high-frequency thermal disturbance signal. Identification of suspected hidden salinization risk areas: after removing the interference caused by vegetation, the remaining areas may be thermal anomalies caused by soil salinization, groundwater level changes or other long-term thermal response changes. These areas are characterized by significant high-frequency temperature fluctuations but are not accompanied by obvious vegetation degradation. In the case of no obvious vegetation degradation, the existence of high-frequency thermal disturbances usually can be inferred as hidden salinization risk areas. Spatial clustering: using the spatial clustering algorithm to perform clustering analysis on the selected areas can identify the spatial distribution pattern of thermal disturbances.Cluster analysis can help eliminate those sporadic interference patches (such as temporary meteorological events or equipment errors) and retain those long-term and stable thermal anomaly regions. Generation of latent thermodynamic imbalance target areas: After eliminating interference through clustering, the finally obtained areas are target areas with independent thermal anomaly response characteristics and without accompanying vegetation degradation. These areas can be used as latent thermodynamic imbalance areas to provide targets for further monitoring of soil salinization, land degradation, etc.

[0034] Specifically, the steps for forming a latent salinization risk spatial distribution map by combining limited measured soil salinity data for regression fitting are as follows: In the identified high-frequency thermal anomaly target areas, extract multi-temporal thermal perturbation spectral characteristics and auxiliary remote sensing indicators as regression factors, including surface dryness / wetness and vegetation coverage, and match the spatial positions of the high-frequency thermal anomaly target areas with the measured soil salinity points to construct a non-linear regression model of thermal perturbation amplitude and salinity concentration. Use spatial interpolation algorithms to extrapolate salinization risks in un-sampled areas, and fuse geographical constraint conditions to optimize interpolation weights to generate a global latent salinization risk spatial distribution map, marking high-risk aggregation areas and potential diffusion directions.

[0035] In this implementation plan, multi-temporal thermal perturbation spectral characteristics: refer to the periodicity, amplitude, etc. characteristics of thermal perturbations extracted through spectral analysis. These characteristics reflect the characteristics of temperature fluctuations in the region and are key factors for identifying thermal anomaly regions. Auxiliary remote sensing indicators: include surface dryness / wetness (such as humidity indicators derived from thermal infrared or microwave remote sensing data) and vegetation coverage (such as the NDVI vegetation index). These indicators can reflect the moisture status and vegetation coverage of the surface and are crucial for the assessment of salinization risks. Combining these factors provides multi-dimensional input features for the regression model and can more accurately establish a prediction model for land degradation. In the identified high-frequency thermal anomaly target areas, factors highly correlated with soil salinization need to be extracted from remote sensing data, mainly including: multi-temporal thermal perturbation spectral characteristics, such as the main frequency amplitude of high-frequency thermal perturbations obtained by Fourier transform of the infrared band; surface dryness / wetness index, which can be retrieved through the outer shell temperature of the infrared; vegetation coverage, such as statistics (maximum value, mean value, etc.) of the annual normalized vegetation index (NDVI). Construct the regression factor vector of the pixel as follows: X j =[x j1 ,x j2 ,x j3 ; where: X j : the regression factor vector of the j-th pixel; x j1 : the thermal perturbation amplitude of this pixel; x j2 : the value of the dryness / wetness index; x j3 : the value of the vegetation coverage index. Register the spatial coordinates of the obtained measured soil salinity points with the regression factor image, and extract X j of the corresponding pixel, and pair the measured salinity value Y jForming a training dataset: Where: Y j : Measured soil salt concentration; n: Number of training samples. There is usually a non-linear response relationship between thermal disturbance and salt concentration. Therefore, a multiple non-linear regression model with quadratic terms is adopted: Where: Predicted salt concentration; β0: Regression constant term; β k : First-order regression coefficients (corresponding to thermal disturbance, dry-wetness, vegetation coverage); γ kl : Second-order interaction term coefficient, considering the coupling effect between factors; x i : The k-th regression factor (k = 1, 2, 3); This model can be fitted by the least squares method or extended to a support vector regression or random forest regression model. Using the above regression model to predict all unsampled pixels, the salt concentration estimation of the entire high-frequency thermal anomaly target area is obtained: Where: Predicted value of the salt concentration of the r-th unsampled pixel; X r : Regression factor vector extracted from this pixel; f: Trained regression function. This step completes the filling of the salinization risk value for the identified thermal anomaly area. For areas not covered by the model prediction, weighted spatial interpolation based on geographical factors is adopted, and geographical understandings such as slope and drainage area are introduced as weight adjustment factors. The interpolation formula is as follows: The weight is defined as: Where: w j (u): Interpolated salt value of the u-th unmeasured area; d ju : Spatial distance between the measured point j and the interpolated point u; q: Distance decay exponent (usually taken as 2); δ: Geographical constraint coefficient (such as whether it is a low-lying area with poor drainage, and the value is adjusted within the range of [0.5, 2]); This method gives priority to samples with consistent geographical conditions during interpolation to improve the prediction credibility. Combine the prediction result with the interpolation result to form a global salt risk layer, and the following methods are used for result expression: Classify the salt concentration (such as mild, moderate, and severe salinization); Extract high-risk aggregation areas: Extract high-salt sub-areas through regional connectivity analysis or isolines; Analyze the potential diffusion direction: Combine the slope aspect and the salt gradient field to draw salt expansion trend arrows to assist in the prevention and control planning.

[0036] Specifically, the steps for constructing a multi-dimensional driving factor raster map by integrating multi-source remote sensing data and auxiliary geographic data of the target area in the fusion target area are as follows: Integrate the multi-spectral vegetation stress index, radar-inverted soil moisture time-series data, spatial distribution data of night-time light intensity, and auxiliary geographic data such as groundwater level depth and cultivated land use intensity in the target area; Eliminate data scale differences through unified spatial resolution and coordinate system alignment to generate a multi-dimensional driving factor raster map covering vegetation dynamics, hydrological conditions, and human activity intensity.

[0037] In this implementation scheme, in this step, the vegetation stress index can be calculated from multispectral images; soil moisture time series data can be inverted based on radar backscatter coefficients; night light intensity data can be obtained from light sensors; groundwater level depth data can be derived from actual measurements at regional hydrological stations or official databases; cultivated land use intensity data can be obtained through land use status maps, remote sensing interpretation or statistical yearbooks. After the above data are obtained, they need to be uniformly preprocessed, including geometric correction, spatial registration and resampling, to ensure the comparability and integration of multi-source data in spatial and temporal scales. The selected driving factors should be typical and representative of the salinization driving mechanism, including natural factors and human disturbance factors. The data types are as follows: Multispectral vegetation stress index: such as normalized difference vegetation index, vegetation temperature difference index, photosynthetic efficiency index, which characterize the degree of vegetation stress and are important response characteristics of potential salinization risks; radar inversion soil moisture time series data: the surface soil moisture time series inverted using synthetic aperture radar data can reflect the rise of groundwater and abnormal surface humidity; night light intensity data: as a spatial proxy for the intensity of human activities; groundwater level depth data: an important hydrological parameter reflecting the possible path of regional groundwater salinity rise; cultivated land use intensity data: such as crop planting system (annual planting times), land multiple cropping index, cultivated land density, etc., reflecting the intensity of human transformation. Data preprocessing: resolution unification and spatial alignment. Due to different data sources, resolution (space, time), coordinate system, and data format are often inconsistent, so unified processing is required, including: spatial resolution resampling: use a consistent target grid resolution (such as 30m or 100m) for all factor data, and resample through bilinear interpolation or nearest neighbor method; coordinate system unification: unify each layer to the same geographic projection system; missing value and outlier processing: fill in the missing areas of remote sensing cloud occlusion, and remove abnormal pixels of night light saturation or radar interference; time series data integration: summarize the time series soil moisture, vegetation index and other data for multiple periods, and extract statistical features (such as mean, extreme value) as input factors. Construct a unified driving factor grid map. After preprocessing, each driving factor is organized into a multidimensional spatial map, corresponding to each pixel one by one, and the following structure is constructed: each pixel position corresponds to a driving factor vector; each vector contains multiple data layer information, covering natural factors and human factors; all vectors constitute a multidimensional driving factor grid map with a unified structure in space. This map can be used for subsequent salinization risk modeling, factor contribution analysis and spatiotemporal evolution monitoring.

[0038] Specifically, through the spatial overlay analysis of the spatial distribution map of implicit salinization risk and the raster maps of each factor, the dominant driving factors that are significantly correlated with the spatial distribution of implicit salinization risk are screened out as the driving factors of land degradation in the target area. The screening logic is as follows: through spatial heterogeneity association analysis, the local spatial correlation between the implicit salinization risk value and each driving factor is calculated pixel by pixel; the spatial differentiation characteristics of the factor contribution intensity are quantified through the geographically weighted regression model, and the driving factors that are significantly correlated both globally and locally and have high spatial consistency are screened out, and the redundant interference factors are eliminated to generate the spatial differentiation map of the dominant driving factors.

[0039] In this implementation plan, spatial heterogeneity association analysis: First, by spatially corresponding the spatial distribution map of implicit salinization risk with the grid map of driving factors in each dimension (such as surface humidity, vegetation stress index, groundwater level, etc.), for each remote sensing pixel, analyze the local correlation between its salinization risk value and each factor. The local spatial correlation coefficient is used to explore the influence degree and spatial aggregation characteristics of driving factors at the micro scale. For example: If the high salinization risk value in a certain area is concentrated, and the groundwater level in the corresponding area is generally shallow, the local spatial correlation coefficient is positive and significant, which means that the shallow groundwater level is an important driving factor of the salinization potential in the area. The geographically weighted regression (GWR) model is used to perform spatial regression modeling on each driving factor, and the following relationship is established: Where: S a : Implicit salinity risk value of the ath pixel; X ab : The bth driving factor value of the ath pixel (soil moisture, vegetation coverage); β b (u a ,v a ): Factor b at position (u a ,v a ), reflecting spatial heterogeneity; ε a: Error term. Model significance: The model can identify the regression intensity and direction of each factor in different spatial locations, clarify whether a certain factor has a significant impact both locally and globally, and thus assist in determining whether it is a dominant factor. Screening and redundancy elimination logic: Based on the output results of the geographically weighted regression model, the dominant driving factors are screened in combination with the following criteria: the regression coefficient is significant in most regions (for example, the p-value at a 95% confidence level is less than 0.05); the trend of spatial coefficient changes is highly consistent with the spatial distribution of salinity risk; the multicollinearity with other factors is low; if a driving factor is significant only in a local area, or is highly correlated with multiple other factors, it is determined to be a non-dominant factor or a redundant interference factor and is eliminated. Generation of spatial differentiation map of dominant driving factors: The dominant factors screened out by the above steps are spatially visualized with their regression weights or spatial significance indicators to generate a spatial differentiation map of the dominant factors. This map reflects the heterogeneity of the response mechanism of salinity risk in different regions, providing a scientific basis for differentiated prevention and control strategies.

[0040] See also Figure 2 Specifically, the contribution intensity of driving factors is inversely correlated with the restoration potential, and the steps to generate the restoration antagonism index are as follows: the contribution intensity of the screened dominant driving factors is normalized, and nonlinearly weighted superposition is performed with the regional ecological restoration potential index; the weight distribution is optimized by introducing geographical constraints, and the restoration antagonism index is generated through weighted calculation.

[0041] In this implementation, the contribution strength of the driving factors is normalized: the contribution strength of the dominant driving factors selected above (such as the local regression coefficient output by the GWR model) is normalized to ensure that different factors are comparable and prevent scale differences from affecting the weighted results. The formula is: Where: D k : the original contribution intensity value of the kth dominant driving factor at a certain pixel; D′ k : The normalized contribution intensity value, ranging from [0,1]. Introduction and spatial matching of ecological restoration potential indicators: Select ecological indicators that can reflect the natural recovery capacity of the region, such as NDVI seasonal fluctuation amplitude, soil water content, historical vegetation stability, etc., to construct an ecological restoration potential index map, and perform spatial pixel-level registration with the normalized maps of each driving factor. Construct the antagonistic relationship between the interference intensity of driving factors and the ecological restoration potential, that is, in areas with strong recovery capacity, even if the interference intensity is large, the salinization process can be alleviated. Therefore, the reverse coupling weighted model is used to highlight the balance effect between the two. The formula is expressed as: Among them: A q : The restoration antagonism index of the qth pixel. The larger the value, the stronger the interference and the weaker the restoration; D' qz: Normalized contribution intensity of the z-th driving factor in the q-th pixel; R z : Recovery potential index value of the z-th pixel; w z : Spatial weighting coefficient of the z-th driving factor; ε: A tiny constant to prevent the denominator from being zero. Geographic constraint optimization of weight assignment: By introducing geographic constraint conditions such as regional topographic and geomorphic zoning, land use types, and groundwater zoning, differential assignment of weights to each driving factor is carried out. The method can adopt the automatic weight assignment mechanism of the Analytic Hierarchy Process (AHP) to ensure that the weights fit the actual geographic background. Generating the recovery antagonism index map: After rasterizing the calculated recovery antagonism index, the recovery antagonism index map can be formed. This map reflects the tension relationship between the intensity of human activity interference and the ecological recovery ability in each region, providing a quantitative basis for saline-alkali risk governance and resource investment priority ranking.

[0042] Specifically, based on the spatial distribution of the antagonism index, the steps for dividing the potential threat levels of land degradation in the target area are as follows: Using spatial clustering and natural breakpoint method to perform hierarchical cutting on the antagonism index raster layer, and combining with spatial continuity optimization, three levels of potential threat levels are divided, including high-risk areas, medium-risk areas, and low-risk areas.

[0043] In this implementation scheme, spatial clustering: First, a spatial clustering method (such as K-means clustering, DBSCAN, etc.) is used to perform cluster analysis on the antagonism index raster data. The clustering method automatically divides the spatial units into several categories based on the numerical characteristics of the antagonism index. Different categories represent areas that exhibit similar degradation risk characteristics in space. Natural breakpoint method: This method determines the segmentation points that best reflect the essence of the data by statistically analyzing the distribution of the antagonism index values. Specifically, the natural breakpoint method will look for points with large changes in the median value of the data set, and use this as a basis to segment the raster layer, thereby dividing the continuous values of the antagonism index into categories with clear physical or ecological significance (such as low-risk, medium-risk, and high-risk areas). Combined with spatial continuity optimization: Spatial continuity optimization: The purpose of this optimization step is to ensure that the threat level after the division within the region is spatially consistent and reasonable. For example, in some areas, although the antagonism index may differ significantly in space, adjacent raster units have similar geographical and ecological characteristics and should be classified as the same threat level. Therefore, spatial continuity optimization will adjust the division results so that the threat levels of adjacent grids are more consistent, thereby avoiding unreasonable segmentation caused by data fluctuations. Three levels of potential threat levels: The target areas are divided into three threat levels through clustering and natural breakpoint methods combined with spatial optimization steps: High-risk areas: The antagonism index is low, indicating that these areas have a higher risk of land degradation, weak ecological restoration potential, and may be severely degraded. Medium-risk areas: The antagonism index is medium, showing a certain risk of degradation, but the restoration potential is acceptable, and moderate ecological restoration may be required. Low-risk areas: The antagonism index is high, indicating that these areas have a strong ecological restoration potential and a low risk of land degradation. Through this process, researchers can clearly identify the potential threats of land degradation in the target area and provide data support for subsequent risk management and restoration interventions.

[0044] Specifically, according to the potential threat level of land degradation in the target area, the natural hydrological path and the human activity interference path are integrated to construct a dual-path degradation propagation network model as follows: using the threat level zoning results as the network node attribute weights, constructing the natural hydrological diffusion path according to the groundwater flow direction and terrain slope, and constructing the human activity interference path according to the connectivity of cultivated land and the distribution of irrigation canals; defining the edge weights between nodes as the weighted superposition of the water-salt flux of the natural path and the pressure intensity of the artificial path, generating a dual-path degradation propagation network topology structure, and characterizing the composite diffusion mechanism of risks along the natural and artificial paths.

[0045] In this implementation plan, the threat level zoning is regarded as a network node attribute: in the previous step, a threat level layer containing high-risk, medium-risk, and low-risk areas has been obtained. Each spatial unit (such as a grid unit or an administrative unit) is regarded as a network node, and its threat level is used as the attribute weight of the node, reflecting the intensity of its degradation risk. Construct a natural hydrological diffusion path (Path 1): Based on the digital elevation model (DEM), analyze the terrain slope and groundwater flow direction, identify the potential paths of water and salt migration, and form a hydrological-dominated degradation diffusion path. For example, areas with large slope drops and concentrated water flows are likely to become channels for salt infiltration or surface accumulation. Construct a human activity interference path (Path 2): Based on the spatial connectivity of cultivated land (such as contiguous cultivated areas) and the distribution of irrigation canals and other agricultural infrastructure, construct an artificial interference path. This path reflects the risk that humans change the soil water and salt distribution and expand the degraded area through water conservancy activities. Define the edge weight (propagation intensity) between nodes:. Construct a dual-path degradation propagation network topology: When constructing a path to connect two nodes, the weight of the edge needs to be defined. The weight represents the intensity of degradation spreading from one node to another: for the natural path, the water and salt flux index is used to represent the diffusion ability, reflecting the intensity of salt-carrying movement of water bodies or groundwater; for the artificial path, the interference pressure intensity index, such as irrigation frequency, water consumption density, etc., is used; integrate the above node and edge information to form a graph model with a dual-path structure. This model can simulate the composite diffusion process of degradation risk along the natural path (water and salt migration) and the artificial path (farmland irrigation, etc.), and is used to reveal the potential expansion trend and key intervention paths.

[0046] Specifically, the steps to generate a spatio-temporal evolution probability map of degradation risk, mark the migration path, outbreak threshold, and optimal blocking window period of future risk hotspots by simulating the slow release diffusion of risk along the natural path and the accelerated migration along the artificial path through a spatio-temporal hybrid algorithm are as follows: Based on the dual-path degradation propagation network model, with the threat level as the node weight, simulate the dynamic propagation process of risk through a spatio-temporal hybrid algorithm. The key steps include: Slow release diffusion of the natural path: Calculate the diffusion rate of salt according to the groundwater flow and terrain water collection ability, and simulate the migration process without intervention; Accelerated migration of the artificial path: Quantify the accelerating effect of cultivated land connectivity and irrigation canal distribution on risk propagation, and simulate the rapid diffusion path caused by human activities; Generation of the probability map: Through the time sliding window mechanism, iteratively simulate the cumulative process of risk in stages, dynamically update the risk values of each node, and generate a spatio-temporal evolution probability map of degradation risk; Marking of the migration path and threshold: Extract the continuous areas in the probability map where the risk value exceeds the preset threshold, identify the main migration paths along the natural hydrological network and the artificial path, and mark the outbreak risk level of high-risk hotspots; Identification of the seasonal blocking window: Combine the regional hydrological cycle and governance resource scheduling, analyze the temporal matching relationship between the risk propagation rate and the governance response efficiency, and determine the optimal blocking window period.

[0047] In this implementation plan, natural path dilution and diffusion: simulation method: By calculating the direction of groundwater flow and the terrain's water collection capacity, the diffusion rate of salts on the natural path is deduced, and then the natural diffusion process without human intervention is simulated. Key elements: Groundwater flow: The direction and speed of groundwater flow directly affect the diffusion of salts. The groundwater flow model is used to quantify the spread of salts along the water flow path. Terrain water collection capacity: The digital elevation model (DEM) is used to calculate the terrain's water collection capacity, that is, the area where water converges, and then the diffusion pattern of salts is estimated. Formula explanation: The salt diffusion rate (r d ) can be calculated based on groundwater flow and terrain water collection capacity. The formula is: Where: K w represents the permeability coefficient of groundwater flow; A s is the area of the water convergence area; d w is the length of the water flow path; is the gradient of salt concentration. Accelerated migration along artificial paths: Simulation method: By quantifying the connectivity of cultivated land and the distribution of irrigation canal systems, the accelerating effect of human activities on the spread of degradation risks is simulated. Key elements: Connectivity of cultivated land: Through the spatial distribution and connectivity of cultivated land, quantify how human activities accelerate the spread of soil degradation risks through agricultural cultivation. Distribution of irrigation canal systems: The layout of irrigation canals directly affects the water flow path and thus the diffusion of salts. Formula explanation: The accelerated migration speed along artificial paths (r a ) can be calculated based on the connectivity of cultivated land and the irrigation system: r a =C g ·P c +C w ·P w ; where: C g is the cultivated land connectivity coefficient; P c is the measure of cultivated land connectivity; C w is the acceleration coefficient of the irrigation canal system; P wIt is the connectivity of the irrigation canal system. Probability map generation: Simulation process: Through the time sliding window mechanism, the risk propagation is iteratively simulated in stages, and the risk values of each node are gradually accumulated and updated. According to the changes in the risk values of each node within different time windows, a probability map of the spatio-temporal evolution of degradation risk is generated. Time sliding window mechanism: The total time range is divided into multiple short-term windows, and the risk values within each window are gradually updated, and the future risks are predicted based on historical data. Formula description: The risk value update formula can be expressed as: R(t) = R(t - 1) + ΔR(t); where: R(t) represents the risk value at time t; ΔR(t) is the change in risk within each time window, which depends on the propagation speed on natural and human paths. Migration path and threshold annotation: Risk hot spot extraction: According to the generated spatio-temporal evolution map of risk, continuous regions with risk values exceeding the set threshold are extracted, and the main migration paths along the natural hydrological network and human paths are identified. High-risk outbreak annotation: During the spatio-temporal evolution process, regions with risk values exceeding a certain threshold are marked, and the future outbreak risk levels of these regions are identified. Seasonal blocking window identification: Analysis method: Combining the regional hydrological cycle and governance resource scheduling, the temporal matching relationship between the risk propagation rate and the governance response efficiency is analyzed. Based on this analysis, the optimal "blocking window period" is determined, that is, intervention is carried out during the period when the risk propagation speed is slow and the governance response efficiency is high. Key elements: Hydrological cycle: It refers to the impact of the periodicity of hydrological changes (such as seasonal precipitation and water level changes) within the region on risk propagation. Governance response efficiency: It refers to the effect of taking governance measures to control risks within a specific time period.

[0048] In summary, the present application has at least the following effects:

[0049] The land degradation supervision method based on remote sensing monitoring can accurately identify the dominant driving factors significantly associated with the spatial distribution of hidden salinization risks through the fusion of multi-dimensional data and spatial analysis, providing a scientific basis for the monitoring and early warning of land degradation. Based on the spatio-temporal hybrid algorithm, it can simulate the propagation process of land degradation risks along natural and human paths and generate a probability map of the spatio-temporal evolution of risks, helping to identify future potential risk hot spots and their migration paths. By dividing the potential threat levels of land degradation and identifying the optimal blocking window period, it can provide support for the precise implementation of governance measures, ensuring intervention at the most effective time, thereby reducing the risk of land degradation. The dual-path degradation propagation network model and the dynamic analysis of the risk propagation process can provide global and local optimization solutions for land degradation governance strategies, effectively improving the governance efficiency and decision-making accuracy. Combining the analysis of the restoration antagonism index and the regional ecological restoration potential can more accurately evaluate the potential of land ecological restoration and provide scientific support for ecological protection and restoration.

[0050] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0051] The present invention is described with reference to the flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the processes or a plurality of processes and / or blocks Figure 1 one or more of the blocks or a plurality of blocks.

[0052] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the processes or a plurality of processes and / or blocks Figure 1 one or more of the blocks or a plurality of blocks.

[0053] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the processes or a plurality of processes and / or blocks Figure 1 one or more of the blocks or a plurality of blocks.

[0054] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0055] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A method for monitoring land degradation based on remote sensing, characterized in that The following steps are involved: S1. Collect multi-temporal thermal infrared remote sensing historical images of the target area, construct a pixel-by-pixel time series curve of daily surface temperature variation, apply spectrum analysis to the curve to extract thermal disturbance frequency and amplitude indicators, identify areas with high-frequency thermal anomalies but no significant vegetation degradation within the year, and perform regression fitting based on limited measured soil salinity data to form a spatial distribution map of hidden salinity risk; S2. Based on the spatial distribution map of hidden salinization risk, the multi-source remote sensing data and auxiliary geographic data of the target area are integrated to construct a multidimensional driving factor grid map. Through the spatial superposition analysis of the spatial distribution map of hidden salinization risk and the grid maps of each factor, the dominant driving factors significantly associated with the spatial distribution of hidden salinization risk are screened out as the driving factors of land degradation in the target area; S3. Based on the driving factors of land degradation, the regional ecological restoration potential index and governance resource constraints are introduced, the contribution intensity of the driving factors is inversely correlated with the restoration potential, and the restoration antagonism index is generated. Based on the spatial distribution of the antagonism index, the potential threat level of land degradation in the target area is divided; S4. According to the potential threat level of land degradation in the target area, the natural hydrological path and the human activity interference path are integrated to construct a dual-path degradation propagation network model. The threat level is used as the node weight. The spatiotemporal hybrid algorithm is used to simulate the slow-release diffusion of risks along the natural path and the accelerated migration of the artificial path, and a probability map of the spatiotemporal evolution of degradation risks is generated, marking the migration path, outbreak threshold and optimal blocking window period of future risk hotspots.

2. The method for monitoring and managing land degradation based on remote sensing according to claim 1, wherein: The steps to collect multi-temporal thermal infrared remote sensing historical images of the target area and construct a pixel-by-pixel time series curve of daily surface temperature variation are as follows: The acquired remote sensing images are registered and radiometrically calibrated in chronological order, and representative daytime thermal infrared band data are screened; For each remote sensing pixel, multi-temporal surface temperature values within the same day are extracted to form a single-day temperature curve, and then superimposed on an annual scale to form a pixel-by-pixel daily variation time series matrix; Through outlier removal, time resampling and multi-phase normalization processing, the stability and comparability of the curve are enhanced, and a time series curve dataset that can reflect the dynamic characteristics of pixel thermal response is constructed.

3. The land degradation supervision method based on remote sensing monitoring according to claim 2, characterized in that: The steps for applying spectrum analysis to the curve to extract the frequency and amplitude indicators of thermal disturbances and identifying areas with high-frequency thermal anomalies but no significant vegetation degradation during the year are as follows: Perform discrete Fourier transform or wavelet decomposition on the daily variation time series curve of the surface temperature of each pixel to extract the high-frequency components and dominant amplitude indicators in its frequency domain structure; Construct a characteristic spectrum of thermal disturbances to identify areas with high-frequency periodic thermal anomalies on an annual scale; By integrating the vegetation index change images of the same period, the thermal disturbance areas caused by vegetation changes are eliminated, and the target areas with independent thermal anomaly response characteristics but no significant vegetation degradation are screened out as suspected hidden salinization risk areas. Through spatial clustering, sporadic interference patches are eliminated to generate hidden thermodynamic imbalance target areas.

4. The land degradation supervision method based on remote sensing monitoring according to claim 3, characterized in that: The steps of performing regression fitting based on limited measured soil salinity data to form a spatial distribution map of implicit salinity risk are as follows: In the identified high-frequency thermal anomaly target area, extract multi-temporal thermal disturbance spectral features and auxiliary remote sensing indicators as regression factors, including surface dryness and vegetation coverage. Match the spatial positions of the high-frequency thermal anomaly target area with the measured points of soil salinity, construct a non-linear regression model of thermal disturbance amplitude and salinity concentration, extrapolate the saline-alkali risk of the un-sampled area through a spatial interpolation algorithm, optimize the interpolation weights by integrating geographical constraint conditions, generate a spatial distribution map of the whole-region latent saline-alkali risk, and mark the high-risk aggregation areas and potential diffusion directions.

5. The method for monitoring land degradation based on remote sensing according to claim 4, characterized in that: The steps for constructing a multi-dimensional driving factor raster map by integrating multi-source remote sensing data and auxiliary geographical data of the target area are as follows: Integrate the multi-spectral vegetation stress index, radar-inverted soil moisture time-series data, spatial distribution data of night light intensity, and auxiliary geographical data such as the depth of the groundwater table and the intensity of cultivated land use in the target area; eliminate the data scale differences through spatial resolution unification and coordinate system alignment, and generate a multi-dimensional driving factor raster map covering vegetation dynamics, hydrological conditions, and human activity intensity.

6. The method for monitoring land degradation based on remote sensing according to claim 5, characterized in that: The screening logic for the dominant driving factors significantly associated with the spatial distribution of the latent saline-alkali risk by spatial overlay analysis of the spatial distribution map of the latent saline-alkali risk and each factor raster map is as follows: Through spatial heterogeneity correlation analysis, calculate the local spatial correlation between the latent saline-alkali risk value and each driving factor pixel by pixel. Quantify the spatial differentiation characteristics of the factor contribution intensity through a geographically weighted regression model, screen out the driving factors that are significantly associated both globally and locally and have high spatial consistency, eliminate redundant interference factors, and generate a spatial differentiation map of the dominant driving factors.

7. The method for monitoring land degradation based on remote sensing according to claim 6, characterized in that: The steps for generating a restoration antagonism index by inversely correlating the factor contribution intensity and the restoration potential are as follows: Normalize the contribution intensity of the screened dominant driving factors, and perform non-linear weighted superposition with the regional ecological restoration potential index; optimize the weight allocation by introducing geographical constraint conditions, and generate a restoration antagonism index through weighted calculation.

8. The method for monitoring land degradation based on remote sensing according to claim 7, characterized in that: The steps for dividing the potential threat levels of land degradation in the target area based on the spatial distribution of the antagonism index are as follows: Use spatial clustering and natural breakpoints method to grade and cut the antagonism index raster layer, and optimize it in combination with spatial continuity to divide it into three levels of potential threat levels, including high-risk areas, medium-risk areas, and low-risk areas.

9. The method for monitoring and managing land degradation based on remote sensing according to claim 8, characterized in that: The steps for constructing a dual-path degradation propagation network model by integrating the natural hydrological path and the human activity interference path according to the potential threat levels of land degradation in the target area are as follows: Use the threat level zoning result as the network node attribute weight, construct a natural hydrological diffusion path according to the groundwater flow direction and terrain slope, and construct a human activity interference path according to the cultivated land connectivity and irrigation canal distribution; Define the edge weight between nodes as the weighted superposition of the water-salt flux of the natural path and the pressure intensity of the human path, generate the topological structure of the dual-path degradation propagation network, and characterize the composite diffusion mechanism of the risk along the natural and human paths.

10. The method for monitoring and managing land degradation based on remote sensing according to claim 9, wherein: The steps of simulating the slow release and diffusion of risks along natural paths and the accelerated migration along human paths through the spatio-temporal hybrid algorithm, generating the spatio-temporal evolution probability map of degradation risks, and marking the migration paths, outbreak thresholds, and optimal blocking windows of future risk hotspots are as follows: Based on the dual-path degradation propagation network model, with the threat level as the node weight, the spatio-temporal hybrid algorithm is used to simulate the dynamic risk propagation process. The key steps include: Slow release and diffusion along natural paths: Calculate the diffusion rate of salts according to groundwater flow and topographic water collection ability, and simulate the migration process without intervention; Accelerated migration along human paths: Quantify the accelerating effect of arable land connectivity and irrigation canal system distribution on risk propagation, and simulate the rapid diffusion paths caused by human activities; Generation of probability map: Through the time sliding window mechanism, iteratively simulate the risk accumulation process in stages, dynamically update the risk values of each node, and generate the spatio-temporal evolution probability map of degradation risks; Marking of migration paths and thresholds: Extract the continuous regions in the probability map where the risk values exceed the preset threshold, identify the main migration paths along the natural hydrological network and human paths, and mark the outbreak risk levels of high-risk hotspots; Identification of seasonal blocking windows: Combine the regional hydrological cycle and governance resource scheduling, analyze the temporal matching relationship between the risk propagation rate and the governance response efficiency, and determine the optimal blocking window period.

Images