Desert vegetation-underground water intelligent monitoring and early warning method based on artificial intelligence

Abstract

The invention relates to the technical field of intelligent early warning, in particular to a desert vegetation-underground water intelligent monitoring and early warning method based on artificial intelligence. The method comprises the following steps: obtaining a vegetation-groundwater multi-source monitoring interface corresponding to a target desert area, and carrying out vegetation growth and groundwater multi-source data acquisition and preprocessing to generate desert vegetation-groundwater standardized monitoring data; vegetation-groundwater ecological correlation characterization analysis processing is carried out through the desert vegetation-groundwater standardized monitoring data, desert ecological coupling analysis is carried out at the same time, and desert vegetation-groundwater ecological coupling relation data is generated; based on the desert vegetation-groundwater ecological coupling relation data, desert ecological vulnerable area identification analysis is carried out, an artificial intelligence coupling model is constructed to carry out ecological evolution trend deduction and early warning analysis, and desert vegetation-groundwater ecological early warning data are generated. According to the invention, cooperative monitoring and early warning of the vegetation growth state and underground water resources in the desert area can be realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent early warning technology, and in particular to an intelligent monitoring and early warning method for desert vegetation and groundwater based on artificial intelligence. Background Technology

[0002] With the intensification of global climate change, desertification is becoming increasingly severe, especially in arid and semi-arid regions. The relationship between desert vegetation and groundwater resources is close and has a significant impact on the sustainability of the ecological environment. The growth of desert vegetation is closely related to groundwater levels and quality; changes in groundwater resources directly affect vegetation growth and ecosystem stability. Existing desert vegetation monitoring methods mostly rely on remote sensing, manual observation, and ground sampling. While these methods can provide some data support, they suffer from low spatial resolution, long monitoring cycles, and the inability to achieve real-time monitoring of groundwater changes. This makes it difficult to obtain comprehensive information on groundwater changes, and the lack of integrated analysis of the relationship between vegetation and groundwater leads to poor early warning effectiveness. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide an intelligent monitoring and early warning method for desert vegetation and groundwater based on artificial intelligence, in order to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, an artificial intelligence-based intelligent monitoring and early warning method for desert vegetation and groundwater includes the following steps: Step S1: Obtain the vegetation-groundwater multi-source monitoring interface corresponding to the target desert area; collect and preprocess vegetation growth and groundwater multi-source data based on the vegetation-groundwater multi-source monitoring interface to generate standardized monitoring data of desert vegetation-groundwater. Step S2: Perform vegetation-groundwater ecological association characterization analysis and processing on the standardized monitoring data of desert vegetation-groundwater to generate desert vegetation-groundwater ecological association characterization data. Step S3: Based on the desert vegetation-groundwater ecological association characterization data, perform desert ecological coupling analysis to generate desert vegetation-groundwater ecological coupling relationship data; Step S4: Based on the desert vegetation-groundwater ecological coupling relationship data, conduct desert ecologically vulnerable area identification and analysis to generate desert vegetation-groundwater ecologically vulnerable area data; Step S5: Construct an artificial intelligence coupling model based on the desert vegetation-groundwater ecological coupling relationship data, and use the artificial intelligence coupling model to perform ecological evolution trend prediction and early warning analysis on the desert vegetation-groundwater ecologically fragile area data to generate desert vegetation-groundwater ecological early warning data.

[0005] The beneficial effects of this invention are:The intelligent monitoring and early warning method for desert vegetation-groundwater based on artificial intelligence proposed in this invention has the following advantages compared with existing technologies: By acquiring a multi-source monitoring interface for vegetation-groundwater, it breaks through the limitations of traditional single monitoring methods such as remote sensing, manual observation, and ground sampling. This enables comprehensive collection of vegetation growth data (such as vegetation coverage, plant height, and biomass) and groundwater hydrological data (such as groundwater level, water quality, and recharge), and allows for real-time data transmission through the interface, overcoming the shortcoming of traditional methods in not being able to monitor groundwater changes in real time. Through preprocessing steps such as cleaning, noise reduction, format standardization, and time-series alignment, standardized monitoring data for desert vegetation-groundwater is generated, eliminating format differences and quality risks from different data sources and ensuring data consistency and usability. Secondly, based on the standardized monitoring data, the intrinsic correlation between vegetation growth status and hydrological parameters such as groundwater level and water quality is deeply explored. Through characterization processing, abstract ecological correlations are transformed into quantifiable and analyzable characterization data, such as the vegetation coverage variation patterns corresponding to different groundwater level intervals and the correlation characteristics between water quality parameters and vegetation growth vitality. This correlation characterization analysis breaks through the limitations of traditional methods that isolate vegetation and groundwater monitoring data, clearly presenting the ecological logic of their mutual influence and constraints. Examples include the inhibitory effect of groundwater level decline on vegetation growth and the indirect impact of vegetation root distribution on groundwater recharge. Then, based on the ecological correlation characterization data, a coupling analysis is conducted from the perspective of the entire ecosystem. This not only clarifies the unidirectional influence between vegetation and groundwater but also reveals the coupling mechanism of their interaction and feedback. For instance, changes in groundwater level affect vegetation growth, while changes in vegetation cover alter surface runoff and infiltration processes, thus having a reciprocal effect on groundwater levels, forming a complete coupling cycle. The generated desert vegetation-groundwater ecological coupling relationship data quantifies key indicators such as coupling strength and coupling coordination, clearly defining the types of coupling relationships between the two in different regions (e.g., coordinated coupling and antagonistic coupling). Furthermore, by combining ecological coupling data and analyzing characteristics such as the degree of coordination and stability of vegetation-groundwater coupling, areas with imbalanced coupling and poor ecosystem stability can be identified. Examples include areas where groundwater levels are continuously declining and vegetation cover is simultaneously decreasing, and areas where water quality deterioration hinders vegetation growth. The generated desert vegetation-groundwater ecologically vulnerable zone data clarifies the spatial distribution, vulnerability level, and main vulnerability factors (such as excessive groundwater extraction and insufficient recharge due to drought). Finally, through an artificial intelligence coupling model built based on ecological coupling data, multi-dimensional dynamic data can be integrated to simulate the ecological evolution trends of vulnerable zones under different scenarios (such as climate change and human intervention). For example, it can predict the future trends of vegetation cover and groundwater level changes in vulnerable zones, and determine whether the ecosystem faces degradation risks and the rate of degradation.The ecological early warning data generated through early warning analysis clarifies the risk level, early warning period, and key risk points. It can also provide targeted intervention suggestions for ecological protection (such as rationally regulating the amount of groundwater resource extraction and carrying out vegetation restoration projects), thereby improving the efficiency of desert ecological early warning. Attached Figure Description

[0006] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the steps of the intelligent monitoring and early warning method for desert vegetation and groundwater based on artificial intelligence according to the present invention. Detailed Implementation

[0007] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0008] To achieve the above objectives, please refer to Figure 1 This invention provides an intelligent monitoring and early warning method for desert vegetation-groundwater based on artificial intelligence. In this example, the intelligent monitoring and early warning method for desert vegetation-groundwater based on artificial intelligence includes the following steps: Step S1: Obtain the vegetation-groundwater multi-source monitoring interface corresponding to the target desert area; collect and preprocess vegetation growth and groundwater multi-source data based on the vegetation-groundwater multi-source monitoring interface to generate standardized monitoring data of desert vegetation-groundwater. In this embodiment of the invention, a vegetation-groundwater multi-source monitoring interface is obtained through an ecological environment monitoring platform targeting a 150 km² target desert area on the southeastern edge of a desert. This interface adopts a standardized data transmission protocol and establishes a fixed association with three types of equipment: satellite remote sensing equipment, UAV aerial photography equipment, and groundwater monitoring wells, ensuring that the data transmission delay is ≤30 seconds. Based on this interface, synchronous data acquisition is initiated by the three types of equipment: the satellite remote sensing equipment acquires 10m resolution vegetation remote sensing image data at a frequency of 3 days / time, including raw calculation data of vegetation reflectance and vegetation index; the UAV aerial photography equipment acquires vegetation canopy coverage, plant height, community distribution images and coordinate information at a flight altitude of 150m at a frequency of 7 days / time; and 12 groundwater monitoring wells are arranged in a 2km×2km grid to acquire groundwater depth (accuracy 0.01m) and water temperature data at a frequency of 1 hour / time. The initial multi-source data collected underwent preprocessing: Duplicate data were removed using hash value comparison (redundant information with a duplicate data ratio ≥ 90%); non-isochronous data were unified into diurnal time-series data using linear interpolation; and all data coordinate systems were unified to Gaussian-Kruger projection using Gaussian projection transformation, with spatial matching completed using a 5m × 5m grid. Spatiotemporal co-functional feature parameters were extracted using feature extraction algorithms, and after min-max normalization mapping to the 0-1 interval to eliminate dimensional differences, standardized monitoring data of desert vegetation and groundwater containing 28 feature parameters were finally generated. The data format was organized according to "spatiotemporal coordinates - feature parameter category - standardized value - data reliability".

[0009] Step S2: Perform vegetation-groundwater ecological association characterization analysis and processing on the standardized monitoring data of desert vegetation-groundwater to generate desert vegetation-groundwater ecological association characterization data. In this embodiment of the invention, a method combining data grouping statistics and feature quantification is used to characterize and analyze the ecological association between vegetation and groundwater based on standardized monitoring data of desert vegetation-groundwater. First, the groundwater depth is divided into four intervals: 0-3m, 3-6m, 6-10m, and above 10m. Monitoring data from 30 grid units in each interval are selected, and growth parameters such as canopy cover, annual plant height growth, and diameter at breast height of the three main vegetation species (Haloxylon ammodendron, Artemisia arenaria, and Hippophae rhamnoides) are extracted to determine the vegetation growth parameter intervals corresponding to different groundwater conditions. Then, through field surveys and laboratory analysis, growth status data such as leaf water content and dead branch rate, reproductive capacity data such as seed germination rate and thousand-seed weight, and stress resistance data such as proline content and superoxide dismutase activity are obtained for each interval. All acquired data were subjected to feature quantification processing. Each indicator was normalized to the 0-1 interval according to min-max. The groundwater dynamic parameters in the standardized monitoring data were integrated with the quantified vegetation characteristic data to generate desert vegetation-groundwater ecological correlation characterization data containing 32 indicators including vegetation growth, reproduction, stress resistance and groundwater dynamics. The corresponding correlation between different groundwater conditions and vegetation growth status was clarified.

[0010] Step S3: Based on the desert vegetation-groundwater ecological association characterization data, perform desert ecological coupling analysis to generate desert vegetation-groundwater ecological coupling relationship data; In this embodiment of the invention, desert ecological coupling analysis is conducted using a combination of system analysis and cross-mapping based on desert vegetation-groundwater ecological correlation characterization data. First, a coupling analysis framework is constructed, comprising a data input layer, a parameter correlation layer, and a coupling analysis layer. The data input layer incorporates dynamic groundwater parameters such as groundwater infiltration velocity at different depths, precipitation infiltration recharge, and soil permeability coefficient. The parameter correlation layer establishes the spatial correspondence between dynamic groundwater parameters and vegetation growth areas. The coupling analysis layer sets two types of coupling analysis indicators: groundwater supply capacity and vegetation demand matching degree. Using a 5m×5m grid cell as the basic analysis unit, a cross-mapping algorithm is employed to achieve a one-to-one correspondence between vegetation growth adaptability characteristics and dynamic groundwater parameters. The linear correlation coefficient is calculated to obtain the correlation strength of a single parameter. After normalization, the arithmetic mean method is used to integrate the parameters to obtain a comprehensive coupling strength parameter (range 0-1, with values ​​closer to 1 indicating stronger coupling). The coupling intensity parameters were divided into five levels: strong coupling (0.8-1.0), relatively strong coupling (0.6-0.8), moderate coupling (0.4-0.6), relatively weak coupling (0.2-0.4), and weak coupling (0-0.2). The spatial distribution area and proportion of each level were statistically analyzed, and the influencing factors of coupling were analyzed. Finally, desert vegetation-groundwater ecological coupling relationship data were generated, which included spatial distribution data of coupling levels, details of correlation parameters of each level, and analysis of influencing factors.

[0011] Step S4: Based on the desert vegetation-groundwater ecological coupling relationship data, conduct desert ecologically vulnerable area identification and analysis to generate desert vegetation-groundwater ecologically vulnerable area data; In this embodiment of the invention, a multi-step screening and comprehensive judgment method is used to identify and analyze ecologically vulnerable areas in deserts based on desert vegetation-groundwater ecological coupling relationship data. First, historical ecological data from the past 20 years is collected. Through time-series correlation analysis and contribution analysis, four influencing factors, including annual precipitation and groundwater depth change rate, are screened out. A quantitative system of influencing factors is constructed, and the vital effects are calibrated. Spatial mapping rule matching and multi-level correlation analysis are used to match the coupling relationship data with the quantified influencing factors. A preliminary ecological risk assessment is completed using a spatial clustering algorithm, generating preliminary assessment results including risk level and spatial distribution. Ecosystem balance parameters are extracted from historical ecologically stable period data. Ecological risk benchmark interval parameters are calculated by combining anti-interference probability. Areas exceeding the interval are screened using a 25m×25m spatial neighborhood comparison. Boundary extraction (accuracy within 5m) and spatial topology analysis are performed on the screened areas to obtain boundary features and connectivity parameters. Judgment indicators such as boundary tortuosity ≥1.2 and maximum connectivity area ≥5000m² are set to determine potential areas of vegetation-groundwater supply imbalance. The weights of indicators such as groundwater supply and vegetation survival support are determined by analytic hierarchy process (AHP). The comprehensive ecological support capacity score is calculated, and the vulnerability level is divided according to the score threshold. Finally, data on desert vegetation-groundwater ecologically vulnerable areas containing vulnerability level, spatial distribution, and weakening factors are generated.

[0012] Step S5: Construct an artificial intelligence coupling model based on the desert vegetation-groundwater ecological coupling relationship data, and use the artificial intelligence coupling model to perform ecological evolution trend prediction and early warning analysis on the desert vegetation-groundwater ecologically fragile area data to generate desert vegetation-groundwater ecological early warning data.

[0013] In this embodiment of the invention, an artificial intelligence coupling model is constructed using a deep learning architecture based on desert vegetation-groundwater ecological coupling relationship data. This model enables the prediction and early warning analysis of the evolution trend of ecologically vulnerable areas, generating ecological early warning data. Model construction phase: A long short-term memory network is selected as the core architecture. The input layer has 32 neurons, corresponding to 32 core parameters in the ecological coupling relationship data (including coupling strength parameters, groundwater dynamic parameters, and vegetation growth parameters). Three hidden layers are set, with 64, 32, and 16 neurons in each layer, respectively. The ReLU activation function is used to enhance the model's nonlinear fitting ability, and a Dropout layer (dropout rate 0.2) is introduced to prevent overfitting. The output layer has 5 neurons, corresponding to the ecological evolution trend levels (stable, mild degradation, moderate degradation, severe degradation, and collapse). The model training used ecological coupling relationship data from the past 20 years (2004-2023) as the training set (70%), validation set (15%), and test set (15%). An adaptive momentum estimation optimizer was used during training, with an initial learning rate of 0.001, decreasing by 10% every 100 iterations. The cross-entropy loss function was used. Training stopped when the loss function value stabilized below 0.05 and the test set accuracy was ≥92%. After successful model validation, data from the desert vegetation-groundwater ecologically vulnerable area were input, and three simulation scenarios (natural evolution, extreme drought, and short-duration heavy rainfall) were set to predict ecological evolution trends: the natural evolution scenario used parameters based on the average meteorological and hydrological data of the past 10 years to predict the ecological state over the next 5 and 10 years; the extreme drought scenario simulated 120 consecutive days without precipitation, setting an annual evaporation increase of 20%; and the short-duration heavy rainfall scenario simulated a single day of 52mm precipitation, setting a short-term groundwater recharge increase of 30%. During the simulation, the rate of change of ecological status and the deviation of key parameters in each vulnerable area are calculated in real time. Ecosystem resilience parameters are introduced as early warning constraints, and early warning thresholds are set: a mild early warning corresponds to an ecological status change rate ≥ 0.05 / year, a moderate early warning corresponds to a key parameter deviation of 10%-20% from the baseline range, and a severe early warning corresponds to a key parameter deviation of > 20% from the baseline range or an evolution trend of severe degradation to collapse. The model outputs the evolution trend curves, key node times, and early warning levels of each vulnerable area under different scenarios, and generates desert vegetation-groundwater ecological early warning data by combining spatial distribution data. Specific simulation results are as follows: In the natural evolution scenario, the extremely vulnerable area will evolve from severe degradation to collapse in the next 5 years, with the key node being year 3 (groundwater depth exceeds the critical range), triggering a severe early warning; in the extreme drought scenario, the evolution rate accelerates, the key node is advanced to year 1.5, and the early warning level is upgraded to an emergency warning; in the short-term heavy rainfall scenario, the ecological status recovers slightly in the short term (1-2 years), but then continues to degrade, triggering a mild early warning.The ecological early warning data is ultimately organized in the format of "vulnerable area number - spatial coordinates - scenario inference - evolution trend - early warning level - key nodes - response suggestions," including a visualized spatial early warning distribution map and a time-series early warning curve. It clarifies the early warning priority and core intervention directions for each vulnerable area, providing precise support for desert ecological protection decision-making. An online update mechanism is incorporated during model operation, with the latest monitored ecological coupling relationship data being incorporated quarterly to fine-tune the model parameters, ensuring that the inference and early warning accuracy remains stable at over 90% over the long term.

[0014] Furthermore, step S1 includes the following steps: Step S11: Obtain the vegetation-groundwater multi-source monitoring interface corresponding to the target desert area. The vegetation-groundwater multi-source monitoring interface is associated with satellite remote sensing equipment, UAV aerial photography equipment and groundwater monitoring wells. In this embodiment of the invention, a vegetation-groundwater multi-source monitoring interface is obtained through an ecological environment monitoring platform for a target desert area on the southeastern edge of a desert. This interface adopts a standardized data transmission protocol and establishes a fixed association with three types of equipment. High-resolution optical satellites are selected for satellite remote sensing, with a monitoring range set to cover the target area of ​​150km. 2 The monitoring bands include visible light, near-infrared, and short-wave infrared; the drone aerial photography equipment is deployed at the core monitoring station in the target area, with a flight altitude of 150m, a forward overlap of 80%, and a lateral overlap of 70%; the groundwater monitoring wells are laid out in a 2km×2km grid, with 12 monitoring wells in the target area, each well having a built-in water level monitoring unit. The interface uses protocol parsing to achieve unified reception and distribution of monitoring data from the three types of equipment, ensuring that the data transmission delay is ≤30 seconds.

[0015] Step S12: Collect preliminary vegetation-groundwater multi-source data through the vegetation-groundwater multi-source monitoring interface, including vegetation remote sensing image data, vegetation aerial photography data, and groundwater depth and time series change data. In this embodiment of the invention, three types of equipment are simultaneously collecting data through a vegetation-groundwater multi-source monitoring interface. Satellite remote sensing equipment collects vegetation remote sensing image data at a frequency of 3 times per day, with a spatial resolution of 10m and a single image coverage of 20km × 20km, including raw calculation data for vegetation reflectance and vegetation index. Unmanned aerial vehicle (UAV) aerial photography equipment performs low-altitude aerial photography at a frequency of 7 times per day, collecting vegetation aerial photography data, including images of vegetation canopy cover, plant height, community distribution, and corresponding coordinate information. Groundwater monitoring wells collect groundwater depth data at a frequency of 1 hour per well, while also recording temporal variation data for the past 3 years. The data includes monitoring well number, collection time, depth value (accuracy 0.01m), and water temperature data. After collection, the data is aggregated through the interface to form preliminary vegetation-groundwater multi-source data, with a single batch of data not less than 500MB.

[0016] Step S13: Construct a spatial structure model corresponding to the target desert area. The spatial structure model includes sub-modules for terrain elevation, soil type, and vegetation distribution. In this embodiment of the invention, a spatial structure model of the target desert area is constructed using a three-dimensional modeling method. The model resolution is set to 5m×5m and includes three core sub-modules. The terrain elevation sub-module is constructed using digital elevation model data, integrating satellite remote sensing terrain data and field measurement elevation point data to generate parameters such as terrain relief, slope, and aspect. The soil type sub-module is constructed based on soil sampling data, classifying and labeling soils as sandy soil, sandy loam, and gravelly soil, and clarifying the spatial distribution range and thickness parameters of different soil types. The vegetation distribution sub-module is constructed by combining historical vegetation monitoring data, pre-setting the distribution characteristic parameters of major desert vegetation such as Haloxylon ammodendron, Artemisia argyi, and Hippophae rhamnoides. The modules are linked through a spatial coordinate system to ensure accurate spatial correspondence of the data in each sub-module.

[0017] Step S14: Map the preliminary vegetation-groundwater multi-source data to the spatial structure model, and sequentially complete data deduplication, temporal alignment and spatial matching processing to generate spatial matching results; In this embodiment of the invention, a spatial mapping algorithm is used to map preliminary vegetation-groundwater multi-source data to a spatial structure model according to coordinate information. Data deduplication is performed using a hash value comparison method to filter overlapping areas in satellite remote sensing images and UAV aerial images, retaining higher-resolution image data and removing redundant information with a duplicate data ratio of ≥90%. Temporal alignment uses a linear interpolation method, with 1 day as the time unit, to uniformly interpolate non-isochronous data from satellite remote sensing every 3 days and UAV every 7 days into daily-scale data, achieving temporal matching with the daily average of groundwater data every 1 hour. Spatial matching is achieved through Gaussian projection transformation, converting the coordinate system of all data to Gauss-Kruger projection, and using 5m×5m grid units to achieve precise grid matching between the data and the spatial structure model, generating a spatial matching result containing "grid number-vegetation data-groundwater data-topography and soil data".

[0018] Step S15: Extract the spatiotemporal collaborative feature parameters of vegetation-groundwater based on the spatial matching results, and generate standardized monitoring data of desert vegetation-groundwater.

[0019] In this embodiment of the invention, spatiotemporal collaborative feature parameters of vegetation and groundwater are extracted using a feature extraction algorithm based on spatial matching results. In the temporal dimension, the rate of change of vegetation index and the amplitude of canopy cover change for different vegetation types at monthly and quarterly scales are calculated, while simultaneously calculating the rate of change and fluctuation amplitude of groundwater depth for the corresponding time periods. In the spatial dimension, the spatial correlation coefficient between vegetation canopy cover and groundwater depth within the same grid cell, the correspondence parameters between vegetation distribution density and groundwater depth in different soil types, and the coupling feature parameters between topographic slope and the spatial distribution of vegetation and groundwater are extracted. The extracted feature parameters are standardized by using a min-max normalization method to uniformly map parameter values ​​to the 0-1 range, eliminating dimensional differences. Finally, standardized monitoring data of desert vegetation and groundwater containing 28 feature parameters is generated. The data format is organized according to "spatiotemporal coordinates - feature parameter category - standardized value - data reliability," providing input data for subsequent artificial intelligence monitoring and early warning models.

[0020] Furthermore, step S3 includes the following steps: Step S31: Based on the desert vegetation-groundwater ecological correlation characterization data, extract the vegetation growth parameter ranges corresponding to different groundwater conditions; In this embodiment of the invention, the desert vegetation-groundwater ecological correlation characterization data is processed using a data grouping and statistical method. This data is generated by integrating the standardized monitoring data mentioned above with the measured vegetation growth data. The groundwater depth is divided into four intervals: 0-3m, 3-6m, 6-10m, and above 10m. Each interval corresponds to the selection of monitoring data from 30 grid units. Growth parameters of three main desert vegetation species (Haloxylon ammodendron, Artemisia argyi, and Hippophae rhamnoides) were extracted from each interval, including canopy cover, annual plant height growth, and diameter at breast height (DBH). Statistical analysis determined the corresponding vegetation growth parameter intervals for different groundwater conditions: 0-3m interval: Haloxylon ammodendron canopy cover 25%-35%, annual plant height growth 15-25cm, DBH 0.8-1.2cm; 3-6m interval: canopy cover 18%-28%, annual plant height growth 8-18cm, DBH 0.5-0.9cm; 6-10m interval: canopy cover 10%-20%, annual plant height growth 3-10cm, DBH 0.3-0.6cm; above 10m interval: canopy cover <10%, annual plant height growth <5cm, DBH <0.4cm. This clarified the constraints of different groundwater conditions on vegetation growth.

[0021] Step S32: Obtain relevant data on vegetation growth status, reproductive capacity and stress resistance based on the vegetation growth parameter range, and generate vegetation growth adaptability characteristic parameters; In this embodiment of the invention, vegetation growth status, reproductive capacity, and stress resistance data are obtained by combining field surveys and laboratory analysis based on vegetation growth parameter ranges corresponding to various groundwater conditions. Growth status data, including leaf moisture content and dead branch rate, is obtained through UAV aerial image interpretation and field observation. Reproductive capacity data is obtained by collecting vegetation seeds and measuring seed germination rate, thousand-seed weight, and seedling survival rate. Stress resistance data is obtained through laboratory simulations of drought and high-temperature stress, including vegetation proline content and superoxide dismutase activity. The acquired data undergoes feature quantification processing, integrating indicators such as leaf moisture content (20%-30%), dead branch rate (<15%), seed germination rate (40%-60%), and proline content (150-250 μg / g) to generate vegetation growth adaptability characteristic parameters containing 12 quantitative indicators. Parameter values ​​are normalized to the 0-1 range using min-max normalization to ensure the data can be used for subsequent coupled analysis.

[0022] Step S33: Combine the flow distribution, recharge characteristics and groundwater dynamic parameters corresponding to soil permeability coefficient of groundwater at different depths in the target desert area to construct a coupling analysis framework; In this embodiment of the invention, a coupling analysis framework is constructed using a system analysis method. The framework comprises three layers: a data input layer, a parameter correlation layer, and a coupling analysis layer. The data input layer explicitly includes groundwater dynamic parameters corresponding to the flow distribution, recharge characteristics, and soil permeability coefficients of groundwater at different depths in the target desert area. The flow distribution parameters are groundwater infiltration velocities at different depths (0-5m, 5-10m, 10-20m), the recharge characteristic parameters are precipitation infiltration recharge and groundwater evaporation discharge, and the soil permeability coefficients are determined according to sandy soil, sandy loam, and gravelly soil classifications. The parameter correlation layer establishes the spatial correspondence between groundwater dynamic parameters and vegetation growth areas, clarifying the characterization range of groundwater parameters for different soil types. The coupling analysis layer sets a coupling analysis index system, including groundwater supply capacity indicators and vegetation demand matching indicators. The framework uses a spatial coordinate system to define the spatial location of each parameter, ensuring the accuracy and systematic nature of the coupling analysis.

[0023] Step S34: Based on the coupling analysis framework, cross-map the vegetation growth adaptation characteristic parameters and the groundwater dynamic parameters to generate vegetation-groundwater coupling intensity parameters; In this embodiment of the invention, a cross-mapping algorithm based on a coupling analysis framework is used to achieve the correlation and matching between vegetation growth adaptation characteristic parameters and groundwater dynamic parameters. Using a 5m×5m grid as the basic analysis unit, vegetation growth adaptation characteristic parameters (such as canopy cover, seed germination rate, and proline content) within the same grid are mapped one-to-one with groundwater dynamic parameters (such as groundwater infiltration velocity, precipitation infiltration recharge, and soil permeability coefficient). The correlation degree analysis method is used to calculate the vegetation-groundwater coupling strength parameter. By calculating the linear correlation coefficient between each vegetation parameter and groundwater parameter, and then normalizing the correlation coefficients to obtain the single-parameter correlation strength, the arithmetic mean method is used to integrate the single-parameter correlation strengths to obtain the comprehensive coupling strength parameter. For example, within a certain grid cell, the correlation coefficients between canopy cover and groundwater seepage velocity are 0.82, seed germination rate and precipitation infiltration recharge are 0.78, and proline content and soil permeability coefficient are 0.65. After normalization, the single-parameter correlation strengths are 0.85, 0.80, and 0.68, respectively. The arithmetic mean of these values ​​results in a comprehensive coupling strength parameter of 0.78. The final generated coupling strength parameter value ranges from 0 to 1, with values ​​closer to 1 indicating stronger coupling and values ​​closer to 0 indicating weaker coupling. The generated coupling strength data includes "grid number - coupling strength parameter - parameter composition details".

[0024] Step S35: Generate vegetation-groundwater ecological coupling relationship data through quantitative analysis of the vegetation-groundwater coupling intensity parameters.

[0025] In this embodiment of the invention, the vegetation-groundwater coupling intensity parameters are analyzed using a combination of quantitative grading and spatial analysis. The coupling intensity parameters are divided into five levels: strong coupling (0.8-1.0), relatively strong coupling (0.6-0.8), moderate coupling (0.4-0.6), relatively weak coupling (0.2-0.4), and weak coupling (0-0.2). The spatial distribution area and proportion of each level in the target desert area are statistically analyzed. For example, the strong coupling area is mainly distributed in sandy loam areas with groundwater depth of 0-3m, accounting for 12%; the relatively weak and weak coupling areas are mainly distributed in gravelly soil areas with groundwater depth of more than 10m, accounting for a total of 45%. Combining the vegetation type, soil characteristics, and groundwater conditions of different level areas, the influencing factors of the coupling intensity are analyzed, and the intrinsic correlation between vegetation growth and groundwater dynamics is identified. Finally, vegetation-groundwater ecological coupling relationship data is generated, including spatial distribution data of coupling levels, detailed correlation parameters for each level, and analysis of influencing factors of coupling, providing core correlation basis for subsequent intelligent monitoring and early warning.

[0026] Furthermore, step S4 includes the following steps: Step S41: Collect historical ecological data corresponding to the target desert area and screen out the influencing factors corresponding to the natural evolution of desert ecology; In this embodiment of the invention, historical ecological data for a target desert area on the southeastern edge of a desert over the past 20 years were collected through an ecological environment monitoring archive, a historical database of meteorological observation stations, and annual groundwater monitoring reports. The data spans from 2004 to 2023, with a uniform annual time resolution. The historical ecological data covers three core categories: vegetation community succession data, including temporal changes in the distribution area, dominance, and population density of Haloxylon ammodendron, Artemisia argyi, and Hippophae rhamnoides, obtained through the integration of historical remote sensing image interpretation and field survey records; groundwater dynamic evolution data, including annual average burial depth, recharge, and discharge data from 12 monitoring wells, with an accuracy controlled to 0.01m; and meteorological element change data, including annual precipitation, annual evaporation, and annual average temperature data, with precipitation accuracy of 0.1mm and temperature accuracy of 0.1℃. Temporal correlation analysis was used to explore the intrinsic relationship between vegetation community succession data and groundwater dynamic evolution data. Parameters such as temporal synchronicity coefficient and rate of change difference were calculated. Key driving parameters were screened in conjunction with meteorological element data. Redundant parameters with correlation coefficient absolute values ​​≥0.8 were eliminated through correlation analysis. Contribution analysis was used to determine the importance ranking. Finally, annual precipitation, groundwater depth change rate, annual evaporation, and Haloxylon ammodendron community response sensitivity were selected as influencing factors corresponding to the natural evolution of desert ecology, forming a structured list of influencing factors.

[0027] Step S42: Construct an impact factor quantification system, calibrate the influence intensity of each impact factor, and generate quantified impact factors; In this embodiment of the invention, a quantitative system for influencing factors is constructed by combining mechanism quantification mapping and condition adaptation correction, thereby calibrating the vitality effect of each influencing factor's intensity. First, based on the material cycle and energy flow patterns of desert ecosystems, data on the ecological mechanisms of each influencing factor are obtained, including quantitative indicators such as precipitation infiltration rate and the impact of groundwater depth changes on root water absorption range. An initial allocation model is constructed to calculate initial intensity parameters: annual precipitation (0.38), groundwater depth change rate (0.32), annual evaporation (0.25), and Haloxylon ammodendron community response sensitivity (0.20). Adaptive corrections are made by introducing typical vegetation community characteristics of the target area (Haloxylon ammodendron dominance 60%, Artemisia argyi dominance 25%) and groundwater hydrogeological conditions (sandy soil percentage 70%, aquifer thickness 8-15m). The corrected parameters are 0.35, 0.30, 0.27, and 0.23, respectively. The intensity coefficients are finally calibrated by combining the contribution difference parameters, generating quantitative standards (such as annual precipitation divided into 5 levels according to <100mm->250mm, corresponding to quantitative values ​​of 0.2-1.0) and final intensity coefficients (0.426, 0.313, 0.254, 0.184), forming a complete quantitative system for impact factors, and outputting the quantified impact factors.

[0028] Step S43: The desert vegetation-groundwater ecological coupling relationship data is correlated and matched with the quantified influencing factors to conduct a preliminary ecological risk assessment and generate preliminary assessment results; In this embodiment of the invention, spatial mapping rule matching and multi-level association analysis are employed to achieve association matching between desert vegetation-groundwater ecological coupling relationship data and quantified influencing factors, and to conduct a preliminary assessment of ecological risk. First, spatial distribution features (patch area of ​​coupling level, aggregation index) and temporal evolution features (area change rate, intensity mean change amplitude) are extracted from the coupling relationship data, generating a set of 12 coupling feature parameters. Based on association rule mining, spatial mapping rules are established to clarify the matching relationship between strong coupling features and annual precipitation (0.8-1.0, etc.), constructing a 4×5 dimension factor-coupling feature association matrix. A correlation degree superposition algorithm is used to calculate the basic correlation degree, and a spatiotemporal attenuation coefficient (spatial attenuation calculated based on distance from the core area, and temporal attenuation calculated based on the time span) and an ecological resilience correction coefficient are introduced for correction, resulting in ecological risk correlation parameters. A spatial clustering algorithm is used to divide the parameters into 5 risk levels (0.8-1.0 for extremely high risk to 0-0.2 for extremely low risk), and the area proportion and distribution area of ​​each level are statistically analyzed to generate a preliminary assessment result including risk level, spatial distribution, and core influencing factors.

[0029] Step S44: Based on the preliminary assessment results, conduct a potential assessment and analysis of supply imbalance to generate a list of potential areas for vegetation-groundwater supply imbalance; In this embodiment of the invention, a potential assessment analysis of supply imbalance is conducted based on preliminary assessment results using spatial neighborhood comparison and comprehensive judgment methods, generating a list of potential vegetation-groundwater supply imbalance areas. First, coupling relationship data from historical ecologically stable periods (2010-2015) are retrieved, and ecosystem balance parameters such as average coupling strength and average groundwater depth are extracted. Combined with an anti-interference probability of 79%, ecological risk baseline parameters (e.g., groundwater depth 3.95-5.05m, Haloxylon ammodendron canopy coverage 19.84%-26.16%) are calculated. A spatial neighborhood analysis method (neighborhood range 25m×25m) is used to compare the preliminary assessment results with the baseline parameters, screening out areas where the indicators exceed the range. Boundary extraction (accuracy within 5m) and spatial topology analysis are performed on the screened areas to obtain parameters such as boundary length, tortuosity, number of connected regions, and maximum connected area. Judgment indicators are set (boundary tortuosity ≥ 1.2, maximum connected area ≥ 5000m²). 2 If a region meets either of the following criteria (high risk or above), it is identified as a potential imbalance area. The region is then labeled with its area number, coordinate range, and other attributes, forming a list of eight potential areas.

[0030] Step S45: Perform ecological support analysis on the list of potential areas with vegetation-groundwater supply imbalance and output data on ecologically vulnerable desert vegetation-groundwater areas.

[0031] In this embodiment of the invention, an ecological support capacity quantitative assessment and spatial feature integration method is used to perform ecological support analysis on a list of potential areas with vegetation-groundwater supply imbalance, outputting data on ecologically vulnerable desert vegetation-groundwater areas. Three core support indicators are selected: groundwater supply capacity (recharge amount, burial depth stability), vegetation survival support capacity (community cover, species richness), and soil retention capacity (soil moisture content, permeability coefficient). The weights of these indicators are determined using the analytic hierarchy process (AHP) (groundwater supply capacity 0.4, vegetation survival support capacity 0.35, soil retention capacity 0.25). Indicator values ​​for each potential area are obtained through field sampling and monitoring data, and after standardization, a comprehensive ecological support capacity score (0-1 range, lower scores indicate weaker support capacity) is calculated. A vulnerability threshold is set (score < 0.3 for extremely high vulnerability, 0.3-0.5 for high vulnerability), and combined with the preliminary risk assessment, the potential areas are classified into vulnerability levels. By integrating data such as vulnerability levels, comprehensive support capacity scores, core weakness indicators, and spatial coordinate ranges of various regions, data on ecologically vulnerable desert vegetation-groundwater zones are output, clarifying the spatial distribution, vulnerability degree, and main weakening factors of each vulnerable zone.

[0032] Furthermore, step S41 includes the following steps: Step S411: Collect historical ecological data corresponding to the target desert area, including vegetation community succession data, groundwater dynamic evolution data, and meteorological element change data; In this embodiment of the invention, historical ecological data for a target desert area on the southeastern edge of a desert over the past 20 years were collected through an ecological environment monitoring archive, a historical database of meteorological observation stations, and annual groundwater monitoring reports. The data spans from 2004 to 2023, with a uniform annual time resolution. Vegetation community succession data includes changes in the community distribution area, dominance, and population density of *Hippophae rhamnoides*, *Artemisia arenaria*, and *Hippophae rhamnoides* in different years. This data was obtained by integrating historical remote sensing image interpretation with field survey records. Groundwater dynamic evolution data includes time-series changes in the annual average burial depth, groundwater recharge, and discharge of each monitoring well, with an accuracy controlled to 0.01m. Meteorological element change data includes annual precipitation, annual evaporation, annual average temperature, and number of days with strong winds per year, with precipitation accuracy of 0.1mm and temperature accuracy of 0.1℃. All data are organized in the format of "year-spatial partition-parameter type-value" to form a structured historical ecological dataset with a data integrity ≥95%.

[0033] Step S412: Extract vegetation-groundwater co-evolution correlation parameters based on the temporal correlation characteristics of vegetation community succession data and groundwater dynamic evolution data; In this embodiment of the invention, a temporal correlation analysis method is used to process vegetation community succession data and groundwater dynamic evolution data. Using the year as a time node, the correspondence between the two types of data at the same time node is established. By calculating the temporal synchronicity between the rate of change in vegetation community distribution area and the rate of change in groundwater depth, vegetation-groundwater co-evolution correlation parameters are extracted. Specifically, the difference and lag time between the rate of change in the distribution area of ​​*Haloxylon ammodendron* community and the corresponding rate of change in groundwater depth are calculated for each year. Simultaneously, the temporal correlation coefficients between the population density changes of *Artemisia arenaria* and *Hippophae rhamnoides* and the changes in groundwater recharge are calculated. Finally, eight vegetation-groundwater co-evolution correlation parameters are generated, including the temporal synchronicity coefficient, the difference in the rate of change, the lag time, and the population-hydrological correlation coefficient. The parameter values ​​are categorized and labeled by year and spatial region, clearly defining the co-evolution characteristics of different time periods and regions.

[0034] Step S413: Based on the vegetation-groundwater co-evolution correlation parameters, calculate the response sensitivity parameters of vegetation growth to changes in groundwater resources; In this embodiment of the invention, the response sensitivity parameters of vegetation growth to changes in groundwater resources are calculated using a responsiveness analysis method based on vegetation-groundwater co-evolution correlation parameters. Groundwater resource change indicators are defined as the annual changes in groundwater depth and groundwater recharge, while vegetation growth indicators are the corresponding annual changes in vegetation canopy cover and annual plant height growth. By calculating the ratio of the changes in vegetation growth indicators to the changes in groundwater resource indicators, the responsiveness values ​​for a single vegetation type and a single water resource indicator are obtained. These values ​​are then standardized to map to the 0-1 range to obtain the response sensitivity parameters. For example, the ratio of the change in canopy cover of *Haloxylon ammodendron* to the change in groundwater depth is 0.72, and the standardized response sensitivity parameter is 0.75; the ratio of the change in annual plant height growth of *Artemisia annua* to the change in groundwater recharge is 0.58, and the standardized response sensitivity parameter is 0.60, thus clarifying the sensitivity of different vegetation types to changes in groundwater resources.

[0035] Step S414: Combine response sensitivity parameters with meteorological element change data to identify key driving parameters affecting vegetation-groundwater ecological stability; In this embodiment of the invention, a coupling analysis method is employed to combine response sensitivity parameters with meteorological element change data to identify key driving parameters affecting vegetation-groundwater ecological stability. The response sensitivity parameters for each year are correlated and matched with corresponding annual meteorological element data (annual precipitation, annual evaporation, and annual average temperature). The time-series correlation coefficients between the response sensitivity parameters and each meteorological element are calculated, and meteorological elements with an absolute correlation coefficient ≥ 0.6 are selected as candidate driving parameters. Simultaneously, the indirect impact of meteorological element changes on groundwater resource changes is analyzed. The correlation coefficients between annual precipitation and groundwater recharge, and between annual evaporation and groundwater depth are calculated. Combined with the response sensitivity of vegetation growth to groundwater, annual precipitation, annual evaporation, groundwater depth change rate, and Haloxylon ammodendron community response sensitivity are ultimately identified as key driving parameters affecting vegetation-groundwater ecological stability, clarifying the direction of each parameter's effect on ecological stability.

[0036] Step S415: By deduplicating and ranking the importance of key driving parameters based on their correlation, the influencing factors corresponding to the natural evolution of desert ecology are screened out.

[0037] In this embodiment of the invention, a combination of correlation analysis and extreme value screening is used to deduplicate and rank the key driving parameters based on their correlations. By calculating the correlation coefficients between the key driving parameters, redundant parameters with an absolute correlation coefficient ≥ 0.8 are removed, such as a derived parameter highly correlated with annual evaporation, retaining four core key driving parameters. Contribution analysis is used to rank their importance, calculating the contribution of each key driving parameter to changes in vegetation-groundwater ecological stability. The contribution is determined based on the ratio of the parameter change magnitude to the ecological stability change magnitude. The final ranking is: annual precipitation (contribution 0.35) > groundwater depth change rate (contribution 0.28) > annual evaporation (contribution 0.22) > Haloxylon ammodendron community response sensitivity (contribution 0.15). Based on the ranking results, these four parameters are selected as influencing factors corresponding to the natural evolution of desert ecology, forming an influencing factor list, clarifying the quantitative range and mechanism of action of each factor, and providing core input for the subsequent construction of intelligent early warning models.

[0038] Furthermore, step S42 includes the following steps: Step S421: Based on the natural laws of material cycling and energy flow in desert ecosystems, obtain data on the ecological mechanisms of each influencing factor; In this embodiment of the invention, based on the natural laws of material cycling and energy flow in desert ecosystems, and through a combination of ecological mechanism research literature integration and field observation verification, data on the ecological mechanisms corresponding to four influencing factors—annual precipitation, groundwater depth change rate, annual evaporation, and Haloxylon ammodendron community response sensitivity—are obtained. The data on the ecological mechanisms of annual precipitation include the efficiency of precipitation infiltration in replenishing groundwater (35%-45% infiltration rate in sandy soil areas and 50%-60% in sandy loam areas), the effective utilization of precipitation directly nourishing vegetation (accounting for 20%-30% of precipitation), and the trigger threshold for plant seed germination (single precipitation ≥10mm); the data on the mechanism of groundwater depth change rate includes the impact of depth change on the root water absorption range of vegetation (the root water absorption range decreases by 15%-20% for every 1m increase in depth) and the magnitude of its regulation on soil moisture content; the data on the mechanism of annual evaporation includes the contribution of evaporation to groundwater discharge (25%-35%) and the intensity of its impact on plant leaf transpiration; the data on the mechanism of Haloxylon ammodendron community response sensitivity includes the feedback cycle of Haloxylon ammodendron growth status to water resource changes (15-30 days) and the regulatory effect of community cover changes on regional microclimate. All data are organized in the format of "influencing factor-action link-quantitative index-numerical range" to ensure the relevance and quantification of the mechanism data.

[0039] Step S422: Based on the ecological mechanism data, construct an initial allocation model for the intensity of influence factors and generate initial intensity parameters for each factor; In this embodiment of the invention, an initial allocation model for the intensity of influencing factors is constructed using a mechanism quantification mapping method based on ecological action mechanism data. The model takes quantitative indicators of ecological action mechanisms as input and quantifies the initial intensity parameters by measuring the scope and intensity of the action links. The model input indicators are set as the proportion of the influence range of the influencing factor on the vegetation-groundwater ecosystem, the duration of the effect, and the probability of triggering the action threshold. The output is the initial intensity parameters in the range of 0-1. Based on the data from the aforementioned mechanisms, the following initial parameters were calculated: annual precipitation initial intensity parameter 0.38 (effect range 65%, duration 90 days, trigger probability 70%); groundwater depth change rate initial intensity parameter 0.32 (effect range 58%, duration 120 days, trigger probability 65%); annual evaporation initial intensity parameter 0.25 (effect range 50%, duration 180 days, trigger probability 60%); and Haloxylon ammodendron community response sensitivity initial intensity parameter 0.20 (effect range 45%, duration 60 days, trigger probability 75%). This generated an initial parameter set containing the initial intensity of each factor and its calculation basis.

[0040] Step S423: Introduce typical vegetation community characteristics and groundwater geological conditions data corresponding to the target desert area, adaptively correct the initial action intensity parameters, and generate corrected action intensity parameters. In this embodiment of the invention, adaptive corrections are made by introducing typical vegetation community characteristics and groundwater geological conditions data corresponding to the target desert area (southern edge of a desert). Typical vegetation community characteristics data include a Haloxylon ammodendron community dominance of 60%, average cover of 18%, and root distribution depth of 2-6m; Artemisia argyi community dominance of 25% and root distribution depth of 1-3m. Groundwater geological conditions data include a regional groundwater aquifer thickness of 8-15m, sandy soil proportion of 70%, sandy loam proportion of 20%, gravelly soil proportion of 10%, and groundwater runoff rate of 0.5-1.2m / d. A condition-adaptive correction algorithm was used to correlate and calibrate the above data with the initial action intensity parameters. For example, considering that the high proportion of sandy soil in the region leads to a decrease in precipitation infiltration rate, the initial parameter of annual precipitation was corrected from 0.38 to 0.35; combined with the high dominance characteristics of the Haloxylon ammodendron community, the initial parameter of the Haloxylon ammodendron community response sensitivity was corrected from 0.20 to 0.23; based on the characteristics of groundwater runoff rate, the initial parameter of groundwater depth change rate was corrected from 0.32 to 0.30; and combined with the high evaporation characteristics of the region, the initial parameter of annual evaporation was corrected from 0.25 to 0.27, thus generating a set of corrected action intensity parameters.

[0041] Step S424: Based on the corrected effect intensity parameters, calculate the contribution difference parameters of each influencing factor to the vegetation-groundwater ecological balance; In this embodiment of the invention, the contribution difference parameter of each influencing factor to the vegetation-groundwater ecological balance is calculated using a difference calculation method based on the corrected effect intensity parameter. Based on the corrected effect intensity parameter, the difference between each factor parameter value and the average value of all factor parameters is calculated, and then normalized to obtain the contribution difference parameter. Specific calculations: The average value of the parameters after correction for each factor is (0.35+0.30+0.27+0.23) / 4=0.2875; the contribution difference parameter for annual precipitation is (0.35-0.2875) / 0.2875≈0.217; the contribution difference parameter for groundwater depth change rate is (0.30-0.2875) / 0.2875≈0.043; the contribution difference parameter for annual evaporation is (0.27-0.2875) / 0.2875≈-0.061; the contribution difference parameter for Haloxylon ammodendron community response sensitivity is (0.23-0.2875) / 0.2875≈-0.200. Positive values ​​for the contribution difference parameter indicate a contribution higher than the average level, while negative values ​​indicate a contribution lower than the average level, clearly indicating the relative differences in the contribution of each factor.

[0042] Step S425: Construct an impact factor quantification system by combining contribution difference parameters, and complete the final calibration of the quantification standards and effect intensity coefficients of each factor to generate quantified impact factors.

[0043] In this embodiment of the invention, an impact factor quantification system is constructed by combining contribution difference parameters. The system includes two core components: factor quantification standard setting and effect intensity coefficient calibration. The quantification standard setting adopts an interval segmented quantification method, which divides the actual monitoring values ​​of each impact factor into 5 levels and assigns corresponding quantification values. For example, annual precipitation is quantified as <100mm (0.2), 100-150mm (0.4), 150-200mm (0.6), 200-250mm (0.8), and >250mm (1.0); groundwater depth change rate is quantified as >0.5m / year (0.2), 0.3-0.5m / year (0.4), 0-0.3m / year (0.6), -0.3-0m / year (0.8), and <-0.3m / year (1.0). The intensity coefficient calibration employs a weighted correction based on contribution difference parameters. The final intensity coefficient is obtained by multiplying the contribution difference parameters of each factor by the initial intensity coefficient. For example, the final intensity coefficient for annual precipitation is 0.35 × (1 + 0.217) ≈ 0.426; the final intensity coefficient for groundwater depth change rate is 0.30 × (1 + 0.043) ≈ 0.313; the final intensity coefficient for annual evaporation is 0.27 × (1 - 0.061) ≈ 0.254; and the final intensity coefficient for Haloxylon ammodendron community response sensitivity is 0.23 × (1 - 0.200) ≈ 0.184. This final calibration of the quantitative standards and intensity coefficients generates quantified impact factors, providing precise input for the parameter configuration of subsequent artificial intelligence early warning models.

[0044] Furthermore, step S43 includes the following steps: Step S431: Obtain spatial distribution characteristics and temporal evolution characteristics data from the desert vegetation-groundwater ecological coupling relationship data, and generate a set of coupling characteristic parameters; In this embodiment of the invention, spatial distribution features and temporal evolution features are extracted from vegetation-groundwater ecological coupling relationship data using feature extraction methods. Spatial distribution features are extracted through spatial pattern analysis, extracting the spatial distribution range, number of patches, patch area, and spatial aggregation index for each coupling level (strong to weak coupling). This clarifies the distribution pattern of strong coupling areas concentrated in sandy loam soil areas with a groundwater depth of 0-3m, and weak coupling areas concentrated in gravelly soil areas above 10m. Temporal evolution features are extracted through temporal trend analysis, extracting the area change rate, mean change amplitude of coupling intensity, and slope of the change trend for each coupling level over the past 10 years. For example, the annual area change rate of strong coupling areas is -0.8%, and the annual area change rate of moderate coupling areas is +0.5%. The two types of feature data are standardized (spatial features are normalized according to the proportion of the total area of ​​the region, and temporal features are normalized according to the maximum value), integrating them to form a coupling feature parameter set containing 12 feature indicators. The parameter set is organized in the format of "spatial coordinates-time node-feature type-standardized value".

[0045] Step S432: Based on the set of coupled feature parameters and the quantification system of influence factors, establish spatial mapping rules and determine the matching relationship between each coupled feature and influence factor; In this embodiment of the invention, spatial mapping rules are established using an association rule mining method based on a set of coupled feature parameters and a quantitative system of influencing factors. First, the core dimensions of the quantitative system of influencing factors are defined, including the quantitative dimension of annual precipitation (0.2-1.0), the quantitative dimension of groundwater depth change rate (0.2-1.0), the quantitative dimension of annual evaporation (0.2-1.0), and the quantitative dimension of Haloxylon ammodendron community response sensitivity (0.2-1.0). By calculating the spatial correlation between the coupled feature parameters and the quantitative values ​​of each influencing factor, the mapping association basis is determined: the aggregation index in spatial distribution characteristics is correlated with the quantitative value of annual precipitation, and the patch area is correlated with the quantitative value of groundwater depth change rate; the area change rate in temporal evolution characteristics is correlated with the quantitative value of annual evaporation, and the magnitude of the mean intensity change is correlated with the quantitative value of Haloxylon ammodendron community response sensitivity. Based on the correlation results, spatial mapping rules are established to clarify the matching relationships between strong coupling features and annual precipitation (0.8-1.0) and groundwater depth change rate (0.8-1.0), and between weak coupling features and annual precipitation (0.2-0.4) and groundwater depth change rate (0.2-0.4), thus forming a structured mapping rule set.

[0046] Step S433: Match the coupled feature parameters to the corresponding quantization dimension of the influence factor according to the spatial mapping rules, and generate the factor-coupled feature correlation matrix; In this embodiment of the invention, a factor-coupled feature correlation matrix is ​​generated by matching the coupling feature parameters to the corresponding quantification dimensions of the influencing factors using a matrix construction method according to spatial mapping rules. The matrix has four row dimensions (annual precipitation, groundwater depth change rate, annual evaporation, and Haloxylon ammodendron community response sensitivity) and five column dimensions (strong coupling, relatively strong coupling, moderate coupling, relatively weak coupling, and weak coupling). The matrix elements are the matched correlation parameter values. During the matching process, the standardized parameter values ​​of each coupling feature are mapped to the quantification dimensions of the influencing factors according to the mapping rules. For example, the standardized value of the spatial aggregation index of the strong coupling feature (0.92) is matched to the 0.8-1.0 dimension of annual precipitation and assigned a value of 0.92 as a matrix element; the standardized value of the area change rate of the weak coupling feature (0.35) is matched to the 0.2-0.4 dimension of groundwater depth change and assigned a value of 0.35 as a matrix element. After completing the full-dimensional matching, the matrix elements are checked for consistency, outliers with matching errors > 5% are removed and corrected, ultimately generating a 4×5 dimension factor-coupled feature correlation matrix.

[0047] Step S434: Calculate the ecological risk correlation parameter corresponding to each region based on the factor-coupling feature correlation matrix. The ecological risk correlation parameter reflects the synergistic effect strength between the coupling feature parameter and the influencing factor. In this embodiment of the invention, the ecological risk correlation parameter corresponding to each region is calculated using a multi-level correlation analysis method based on the factor-coupling feature correlation matrix. First, the correlation element values ​​of each influencing factor and coupling feature in the matrix are extracted. Then, combined with the actual quantified values ​​of the influencing factors in each region, a correlation superposition algorithm is used to calculate the basic correlation. Next, a spatiotemporal attenuation coefficient (spatial attenuation calculated based on distance from the core area, and temporal attenuation calculated based on the data time series span) is introduced to correct the basic correlation. The correction formula is: Ecological risk correlation parameter = Basic correlation × Spatiotemporal attenuation coefficient × Ecological elasticity correction coefficient. For example, for a sandy loam region (ecological elasticity correction coefficient 0.95), the corresponding matrix element value is 0.82. The basic correlation is calculated as 0.82 × the average quantified value of the regional influencing factor 0.75 = 0.615. The spatiotemporal attenuation coefficient is 0.93. After correction, the ecological risk correlation parameter = 0.615 × 0.93 × 0.95 ≈ 0.542. The parameter value ranges from 0 to 1. The larger the value, the stronger the synergistic effect between the coupling characteristics and the influencing factors, and the higher the ecological risk. Ultimately, a set of ecological risk correlation parameters for each region is generated.

[0048] Step S435: Complete the preliminary ecological risk assessment by spatial clustering corresponding to the ecological risk correlation parameters, and generate the preliminary assessment results for each region.

[0049] In this embodiment of the invention, a spatial clustering algorithm is used to perform cluster analysis on the ecological risk correlation parameters to complete the preliminary ecological risk assessment. The number of clusters is set to 5, corresponding to 5 risk levels (extremely high risk, high risk, medium risk, low risk, and extremely low risk). The clustering thresholds are divided according to parameter values: 0.8-1.0 for extremely high risk, 0.6-0.8 for high risk, 0.4-0.6 for medium risk, 0.2-0.4 for low risk, and 0-0.2 for extremely low risk. During the clustering process, a 5m×5m grid unit is used as the clustering unit. The ecological risk correlation parameters of each unit are substituted into the algorithm to achieve spatial aggregation of similar units. For example, grid units with parameters of 0.85-0.98 are clustered into the extremely high risk class, and grid units with parameters of 0.1-0.18 are clustered into the extremely low risk class. After clustering, the area, proportion, and main distribution areas of each risk level are statistically analyzed. For example, the extremely high-risk area accounts for 8%, concentrated in gravelly soil areas with groundwater depth >10m and high annual evaporation; the extremely low-risk area accounts for 15%, concentrated in sandy loam soil areas with groundwater depth 0-3m. Based on the clustering results, preliminary assessment results for each region are generated, including risk level, spatial distribution, area proportion, and core influencing factors.

[0050] Furthermore, step S434 includes the following steps: Step S4341: Extract the element feature data from the factor-coupling feature correlation matrix to obtain the initial data of the effect strength of each correlation item; In this embodiment of the invention, a factor-coupling feature correlation matrix is ​​constructed based on the quantified influencing factors (annual precipitation, groundwater depth change rate, annual evaporation, and Haloxylon ammodendron community response sensitivity) and vegetation-groundwater ecological coupling relationship data generated above. The matrix has four rows representing four influencing factors and five columns representing five coupling level features (strong to weak coupling). The matrix elements are the correlation data between each influencing factor and its corresponding coupling feature. A matrix element extraction algorithm is used to extract the feature data of the elements in the matrix, clarifying the core attributes of each correlation item, including the spatial range (5m×5m grid cell), time period (annual), and basic action intensity value corresponding to the correlation item. Finally, the initial data of the action intensity of each correlation item are obtained, such as the initial intensity of the correlation item between annual precipitation and strong coupling feature 0.45, the initial intensity of the correlation item between groundwater depth change rate and moderate coupling feature 0.32, the initial intensity of the correlation item between annual evaporation and weak coupling feature 0.28, and the initial intensity of the correlation item between Haloxylon ammodendron community response sensitivity and strong coupling feature 0.21, forming a structured initial dataset of action intensity.

[0051] Step S4342: Based on the initial data of the interaction intensity, calculate the spatiotemporal decay coefficient of the correlation term, wherein the spatiotemporal decay coefficient reflects the variation characteristics of the correlation interaction with spatial distance and time span; In this embodiment of the invention, the spatiotemporal attenuation coefficient of the associated term is calculated using a spatiotemporal attenuation model based on the initial data of the effect intensity. The spatial attenuation coefficient is calculated using a distance attenuation function, with a set attenuation baseline distance of 5km. The attenuation coefficient decreases exponentially with increasing spatial distance, and the formula is: Spatial attenuation coefficient = e^(-0.2 × spatial distance / 5). For example, when the distance between the grid cell corresponding to a certain associated term and the core monitoring area is 2km, the spatial attenuation coefficient = e^(-0.2 × 2 / 5) = e^(-0.08) ≈ 0.923; when the distance is 6km, the spatial attenuation coefficient = e^(-0.2 × 6 / 5) = e^(-0.24) ≈ 0.787. The time decay coefficient is calculated using a time span decay function, with a base decay time of 5 years. The decay coefficient decreases linearly with increasing time span, and the formula is: Time decay coefficient = 1 - 0.05 × Time span / 5. For example, the time decay coefficient for a time span of 2 years = 1 - 0.05 × 2 / 5 = 0.98; and the time decay coefficient for a time span of 7 years = 1 - 0.05 × 7 / 5 = 0.93. The spatiotemporal decay coefficient is the product of the spatial decay coefficient and the time decay coefficient. For example, the spatiotemporal decay coefficient for a correlation term with a distance of 2 km and a time span of 2 years = 0.923 × 0.98 ≈ 0.905, accurately reflecting the decay characteristics of the correlation effect with spatial distance and time span.

[0052] Step S4343: Correct the initial data of the action intensity by combining the spatiotemporal attenuation coefficient, and generate the effective intensity parameters of the associated action; In this embodiment of the invention, the initial data of the action intensity is linearly corrected by combining the spatiotemporal attenuation coefficient. The correction formula is: associated effective action intensity parameter = initial data of action intensity × spatiotemporal attenuation coefficient × environmental adaptation coefficient, wherein the environmental adaptation coefficient is determined according to the soil type of the target area: 0.95 for sandy soil area, 1.0 for sandy loam area, and 0.9 for gravelly soil area. For each of the different correlation terms, the calculations were performed individually. For example, the correlation term between annual precipitation and strong coupling characteristics (initial strength 0.45, spatiotemporal attenuation coefficient 0.905, and regional environmental adaptation coefficient of sandy loam soil 1.0) was corrected to an effective strength parameter of 0.45 × 0.905 × 1.0 ≈ 0.407; the correlation term between groundwater depth change rate and moderate coupling characteristics (initial strength 0.32, spatiotemporal attenuation coefficient 0.85, and regional environmental adaptation coefficient of sandy soil 0.95) was corrected to an effective strength parameter of 0.32 × 0.85 × 0.95 ≈ 0.258; and the correlation term between annual evaporation and weak coupling characteristics (initial strength 0.28, spatiotemporal attenuation coefficient 0.787, and regional environmental adaptation coefficient of gravelly soil 0.9) was corrected to an effective strength parameter of 0.28 × 0.787 × 0.9 ≈ 0.199. After the correction is completed, a set of effective intensity parameters for the association effect is generated, which includes "association item number - initial intensity - spatiotemporal attenuation coefficient - environmental adaptation coefficient - effective intensity".

[0053] Step S4344: Based on the effective strength parameter of the association effect, construct a risk association degree calculation model and introduce ecosystem resilience related parameters as constraints; In this embodiment of the invention, a risk correlation calculation model is constructed using a multiple linear regression method based on the effective strength parameters of the correlation effects. The model input consists of four effective strength parameters of the correlation effects for each grid cell (annual precipitation-coupling characteristics, groundwater depth change rate-coupling characteristics, annual evaporation-coupling characteristics, and Haloxylon ammodendron community response sensitivity-coupling characteristics). The output is the ecological risk correlation parameter. Ecosystem resilience parameters are introduced as constraints, including vegetation recovery rate (Haloxylon ammodendron 0.08 cm / day, Artemisia argyi 0.05 cm / day) and soil water retention elasticity coefficient (sandy soil 0.6, sandy loam 0.8, gravelly soil 0.4). The constraint thresholds are set as vegetation recovery rate ≥ 0.03 cm / day and soil water retention elasticity coefficient ≥ 0.4. When the parameters are below the thresholds, the model output risk correlation parameter increases by 15%. The model structure is as follows: Risk correlation parameter = 0.3 × annual precipitation correlation effective strength + 0.28 × groundwater depth change rate correlation effective strength + 0.22 × annual evaporation correlation effective strength + 0.2 × Haloxylon ammodendron community response sensitivity correlation effective strength, ensuring that the model calculation logic matches the contribution of the influencing factors mentioned above.

[0054] Step S4345: Output the ecological risk correlation parameters for each region through the risk correlation calculation model.

[0055] In this invention, the target desert area is 150km. 2The area was divided into 600,000 grid cells of 5m × 5m. Effective strength parameters of the correlation effects and ecosystem resilience parameters were extracted for each grid cell and input into the risk correlation calculation model for batch calculation. For example, in a sandy loam grid cell (vegetation recovery rate 0.06 cm / day, soil water retention elasticity coefficient 0.8, both meeting the constraint thresholds), the effective strength parameters were 0.407, 0.258, 0.199, and 0.185. The calculated risk correlation parameter was: 0.3 × 0.407 + 0.28 × 0.258 + 0.22 × 0.199 + 0.2 × 0.185 ≈ 0.122 + 0.072 + 0.044 + 0.037 ≈ 0. 275; For a certain gravelly soil grid cell (vegetation recovery rate 0.02 cm / day, below the constraint threshold), the effective strength parameters are 0.35, 0.22, 0.18, and 0.16, respectively. The basic calculated value is 0.3×0.35+0.28×0.22+0.22×0.18+0.2×0.16≈0.105+0.062+0.040+0.032≈0.239. After a 15% adjustment, the risk correlation parameter is approximately 0.275. The final output is the ecological risk correlation parameter for each grid cell, with a value range of 0-1, forming spatial distribution data of ecological risk correlation covering the entire target area.

[0056] Furthermore, step S44 includes the following steps: Step S441: Obtain vegetation-groundwater ecological coupling relationship data for the target desert area during the corresponding historical ecologically stable period, and extract the corresponding ecosystem balance parameters during the historical ecologically stable period. In this embodiment of the invention, vegetation-groundwater ecological coupling relationship data for the target desert area (southern edge of a desert) during its historical ecologically stable period (2010-2015) is retrieved from an ecological environment monitoring archive. This data includes the coupling level, coupling strength parameters, spatial distribution coordinates, and temporal changes of each 5m×5m grid unit during the stable period. A parameter extraction algorithm is used to extract ecosystem balance parameters from the data. Extracted indicators include the average coupling strength parameter of each grid unit during the stable period, the average groundwater depth, the average vegetation canopy cover, and the average soil moisture content. The fluctuation range (mean ± standard deviation) of each parameter is also extracted. For example, the average coupling strength parameter of the sandy loam grid unit is 0.72, with a fluctuation range of 0.65-0.79; the average groundwater depth is 4.5m, with a fluctuation range of 3.8-5.2m; and the average canopy cover of *Haloxylon ammodendron* is 23%, with a fluctuation range of 19%-27%. The extracted parameters are organized in the format of "grid cell-parameter type-mean-fluctuation range" to form an ecosystem balance parameter set, with parameter integrity controlled at over 98%.

[0057] Step S442: Based on the ecosystem balance parameters, calculate the ecological risk benchmark interval parameters, which reflect the risk range for the stable operation of the ecosystem; In this embodiment of the invention, ecological risk baseline parameters are calculated using an interval quantification method based on ecosystem balance parameters. The interval width is determined by centering on the mean value of each balance parameter, combined with the parameter fluctuation range during the stable period and the ecosystem's resistance to disturbance threshold (79% probability of disturbance). The calculation logic is: baseline interval lower limit = mean - half-width of fluctuation range × resistance to disturbance; baseline interval upper limit = mean + half-width of fluctuation range × resistance to disturbance. For example, with a groundwater depth of 4.5m and a fluctuation range half-width of 0.7m, the baseline interval lower limit = 4.5 - 0.7 × 79% ≈ 3.95m, and the baseline interval upper limit = 4.5 + 0.7 × 79% ≈ 5.05m; with a Haloxylon ammodendron canopy cover of 23% and a fluctuation range half-width of 4%, the baseline interval lower limit = 23% - 4% × 79% ≈ 19.84%, and the baseline interval upper limit = 23% + 4% × 79% ≈ 26.16%. Simultaneously, the interval parameters are standardized to the 0-1 range, such as the standardization threshold of the groundwater depth benchmark interval being 0.32-0.68. Finally, an ecological risk benchmark interval parameter set is generated, which includes four types of indicators: groundwater depth, vegetation canopy coverage, soil moisture content, and coupling strength, thus clarifying the risk boundary range for the stable operation of the ecosystem.

[0058] Step S443: Compare the preliminary assessment results for each region with the ecological risk benchmark interval parameters in terms of spatial neighborhood to screen out regions whose spatial neighborhood range corresponding to the preliminary assessment results exceeds the ecological risk benchmark interval parameters. In this embodiment of the invention, a spatial neighborhood analysis method is used to compare the preliminary assessment results with the ecological risk benchmark interval parameters. The spatial neighborhood is defined as a 3×3 grid (i.e., a 25m×25m range) surrounding each grid unit. Using the grid unit as the basic comparison unit, the risk correlation parameters and coupling levels in the preliminary assessment results are extracted and compared one by one with the ecological risk benchmark interval parameters of the corresponding grid and neighboring grids. The comparison indicators include whether the risk correlation parameter exceeds the benchmark interval and whether the parameter corresponding to the coupling level deviates from the benchmark interval. Grid units and their neighboring areas with any indicator exceeding the benchmark interval are screened out. For example, a gravelly soil grid unit has a preliminary assessment risk correlation parameter of 0.65, corresponding to an ecological risk benchmark interval of 0.2-0.5. Simultaneously, the groundwater depth parameters of its three neighboring grid units exceed the benchmark interval of 3.95-5.05m. This area is determined to be outside the benchmark interval, forming a preliminary list of screened areas.

[0059] Step S444: Perform boundary extraction and spatial topology analysis on regions that exceed the ecological risk benchmark range parameters to generate regional boundary feature parameters and spatial connectivity parameters; In this embodiment of the invention, a boundary extraction algorithm is used to extract the boundary of areas exceeding the ecological risk benchmark range parameters. By identifying and selecting edge grid cells in the selected areas, a boundary line is fitted and generated, with the boundary line accuracy controlled within 5m. Extracted boundary feature parameters include boundary length, boundary tortuosity (torsuosity = actual boundary length / equivalent circular boundary length), and boundary orientation. Simultaneously, spatial connectivity parameters are calculated using spatial topology analysis. By determining the adjacency relationship of grid cells within the exceeded area, the number of connected regions, the maximum connected region area, and the connectivity density (connected region area / total region area) are calculated. For example, an exceeded region has a boundary length of 1200m, a boundary tortuosity of 1.35, and a boundary orientation of northwest-southeast; it contains 3 connected regions, with a maximum connected region area of ​​8000m². 2 With a connectivity density of 0.82, a comprehensive parameter set containing boundary features and connectivity features is generated.

[0060] Step S445: Combine regional boundary characteristic parameters, spatial connectivity parameters, and preliminary assessment results to generate a list of potential areas with vegetation-groundwater supply imbalance.

[0061] In this embodiment of the invention, a comprehensive assessment method is used to generate a list of potential vegetation-groundwater supply imbalance areas by combining regional boundary characteristic parameters, spatial connectivity parameters, and preliminary assessment results. The assessment indicators include boundary tortuosity ≥ 1.2 and maximum connected area ≥ 5000 m². 2 A preliminary risk assessment indicates a high-risk or higher level; meeting any two or more of the indicators qualifies a potential imbalance area. The identified potential areas are then labeled with attributes including area number, spatial coordinate range, boundary characteristic parameters, connectivity parameters, core exceedance indicators, and the preliminary risk assessment level. Simultaneously, the area of ​​each potential area and its proportion of the total target area are calculated. For example, potential area 1, designated Q1, has coordinates of 105°32′–105°35′ E, 37°21′–37°23′ N, a boundary inflection degree of 1.42, and a maximum connected area of ​​9500 m². 2 The core indicators exceeding the target were groundwater depth (6.2m) and risk correlation parameter (0.72), which ultimately formed a list of 8 potential areas, clarifying the core characteristics and imbalance risk basis of each area.

[0062] Furthermore, step S442 includes the following steps: Step S4421: Extract stability-related data corresponding to ecosystem structure and function from vegetation-groundwater ecological coupling relationship data during historical ecological stability periods; In this embodiment of the invention, ecologically stable period data was selected from nearly 20 years of historical ecological data, and the period from 2010 to 2015 was determined as the ecologically stable period for the target desert area. During this period, the fluctuation range of vegetation community coverage was <5%, and the annual change in groundwater depth was <0.2m. A feature screening method was used to extract stability-related data corresponding to ecosystem structure and function from the vegetation-groundwater ecological coupling relationship data during this period. Structural stability data included the dominance degree of Haloxylon ammodendron and Artemisia argyi communities, the coefficient of variation of population density, and the community evenness index; functional stability data included the groundwater recharge-discharge balance coefficient, vegetation net primary productivity, and soil moisture maintenance coefficient. The extracted data were standardized, and parameters such as the coefficient of variation and balance coefficient were uniformly mapped to the 0-1 range, forming a stability-related dataset containing 10 indicators. The data was organized in the format of "year-grid unit-indicator type-standardized value" to ensure that the data could be used for subsequent probability calculations.

[0063] Step S4422: Based on stability-related data, calculate the disturbance resistance probability and self-repair probability of the ecosystem; In this embodiment of the invention, the probability of disturbance resistance and the probability of self-repair of the ecosystem are calculated using probabilistic statistical methods based on stability-related data. The probability of disturbance resistance is calculated by analyzing the ability of stability indicators to maintain stability under slight disturbances. Disturbance thresholds are set as a 3% instantaneous decrease in vegetation cover and a 0.1m instantaneous increase in groundwater depth. The percentage of samples whose indicator values ​​remain within a stable range in the stability-related data is counted to obtain the probability of disturbance resistance. For example, the percentage of samples whose Haloxylon ammodendron community dominance remains stable under disturbance is 82%, and the percentage of samples whose groundwater recharge-discharge balance coefficient remains stable is 78%. The overall probability of disturbance resistance is calculated as (82% + 78% + average percentage of stable samples for the other 6 indicators) / 8 = 79%. The probability of self-repair is calculated by counting the percentage of samples whose stability indicators recover to a stable range after disturbance. The repair period is set to 90 days. The percentage of samples whose indicators recover to a stable range within 90 days after disturbance is counted to obtain a self-repair probability of 73%, thus clarifying the ecosystem's ability to resist disturbances and restore balance.

[0064] Step S4423: Combining the anti-disturbance probability and self-repair probability of the ecosystem, construct an ecological balance interval model and generate ecological balance critical interval parameters; In this embodiment of the invention, an ecological balance interval model is constructed by combining the anti-interference probability of 79% and the self-repair probability of 73% using an interval modeling method. The model uses the statistical characteristics of stability-related data as its core, taking the mean value of each stability indicator during its stable period as the interval center, and the mean ± 2 standard deviations as the initial interval range. Then, interval contraction adjustments are made by combining the anti-interference probability and the self-repair probability. The contraction magnitude is positively correlated with the probability value, and the formula is: Interval contraction coefficient = (anti-interference probability + self-repair probability) / 2 = 76%. The adjusted interval range = initial interval range × contraction coefficient. The model calculates and generates ecological balance critical interval parameters, such as the groundwater depth balance critical interval of 3.2-5.8m, the Haloxylon ammodendron community coverage balance critical interval of 18%-28%, and the soil moisture content balance critical interval of 12%-18%, forming a parameter set containing the critical intervals of eight core indicators. This parameter set clearly defines the core range for each indicator to maintain ecological balance.

[0065] Step S4424: Based on the ecological balance critical interval parameters, historical extreme climate event impact data are introduced for correction to generate the corrected balance interval parameters; In this embodiment of the invention, the impact data of historical extreme climate events in the target desert area over the past 20 years are incorporated and corrected based on ecological balance critical interval parameters. These extreme climate events include the severe drought of 2018 (120 consecutive days without precipitation) and the short-duration heavy rainfall of 2021 (52 mm of rainfall in a single day). Vegetation-groundwater response data for the periods of these events are extracted, and the magnitude of their impact on each critical balance interval is analyzed. For example, the severe drought caused a maximum deviation of 1.2 m in groundwater depth from the lower limit of the balance interval and a maximum deviation of 4% in Haloxylon ammodendron cover; the short-duration heavy rainfall caused a maximum deviation of 0.8 m in groundwater depth from the upper limit and a maximum deviation of 3% in soil moisture content. The impact magnitude weighted correction method is adopted, which assigns weights to the impact magnitude of extreme events according to their frequency of occurrence (severe drought once every 5 years and short-term heavy rainfall once every 3 years), calculates the correction amount, and finally generates the corrected equilibrium interval parameters: groundwater depth 2.9-6.1m, Haloxylon ammodendron community coverage 16%-30%, and soil moisture content 10%-20%, to ensure that the equilibrium interval parameters can adapt to extreme climate conditions.

[0066] Step S4425: Determine the corresponding upper and lower boundaries using the corrected equilibrium interval parameters to generate the ecological risk benchmark interval parameters.

[0067] In this embodiment of the invention, the corresponding upper and lower boundaries are determined by using the corrected equilibrium interval parameters, and an ecological risk benchmark interval parameter is generated using a boundary calibration method. First, the upper and lower boundaries of the corrected equilibrium intervals for each indicator are defined. 110% of the upper boundary is used as the upper risk warning boundary, and 90% of the lower boundary is used as the lower risk warning boundary. Exceeding these ranges indicates the existence of ecological risk. Ecological risk benchmark interval parameters are generated through calibration calculations, such as the groundwater depth risk benchmark interval of 2.6-6.7m (below 2.6m or above 6.7m is considered a risk interval), the Haloxylon ammodendron community coverage risk benchmark interval of 14.4%-33% (below 14.4% or above 33% is considered a risk interval), and the soil moisture content risk benchmark interval of 9%-22% (below 9% or above 22% is considered a risk interval). Simultaneously, the standardized thresholds corresponding to the risk benchmark intervals of each indicator (mapped to the 0-1 interval) are calculated. For example, the standardized threshold for the groundwater burial depth risk benchmark interval is 0.21-0.83. Finally, an ecological risk benchmark interval parameter set is formed, which includes the risk benchmark intervals of each indicator, standardized thresholds, and risk judgment rules, providing core judgment basis for subsequent accurate early warning of ecological risks.

[0068] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. An artificial intelligence-based desert vegetation-groundwater intelligent monitoring and early warning method, characterized in that, The method comprises the following steps: Step S1: obtaining a vegetation-underground water multi-source monitoring interface corresponding to a target desert area; based on the vegetation-underground water multi-source monitoring interface, vegetation growth and underground hydrology multi-source data collection and preprocessing are performed to generate desert vegetation-underground water standardized monitoring data; Step S2: through the desert vegetation-underground water standardized monitoring data, vegetation-underground water ecological correlation characterization analysis processing is performed to generate desert vegetation-underground water ecological correlation characterization data; Step S3: based on the desert vegetation-underground water ecological correlation characterization data, desert ecological coupling analysis is performed to generate desert vegetation-underground water ecological coupling relationship data; Step S4: based on the desert vegetation-underground water ecological coupling relationship data, desert ecological fragile area identification analysis is performed to generate desert vegetation-underground water ecological fragile area data; Step S5: based on the desert vegetation-underground water ecological coupling relationship data, an artificial intelligence coupling model is constructed, and the artificial intelligence coupling model is used for ecological evolution trend deduction and early warning analysis on the desert vegetation-underground water ecological fragile area data to generate desert vegetation-underground water ecological early warning data. 2.The artificial intelligence-based desert vegetation-groundwater intelligent monitoring and early warning method according to claim 1, characterized in that, Step S1 comprises the following steps: Step S11: obtaining a vegetation-underground water multi-source monitoring interface corresponding to a target desert area, the vegetation-underground water multi-source monitoring interface being associated with satellite remote sensing equipment, unmanned aerial vehicle aerial photography equipment and underground water monitoring well; Step S12: collecting preliminary vegetation-underground water multi-source data through the vegetation-underground water multi-source monitoring interface, including vegetation remote sensing image data, vegetation aerial photography data, underground water depth and time series change data; Step S13: constructing a spatial structure model corresponding to the target desert area, the spatial structure model comprising a terrain elevation, soil type and vegetation distribution sub-module; Step S14: mapping the preliminary vegetation-underground water multi-source data to the spatial structure model to sequentially complete data deduplication, time series alignment and spatial matching processing to generate a spatial matching result; Step S15: based on the spatial matching result, extracting vegetation-underground water spatio-temporal coordination feature parameters to generate desert vegetation-underground water standardized monitoring data. 3.The AI-based desert vegetation-groundwater intelligent monitoring and early warning method according to claim 1, characterized in that, Step S3 comprises the following steps: Step S31: based on the desert vegetation-underground water ecological correlation characterization data, extracting vegetation growth parameter intervals corresponding to different underground water conditions; Step S32: obtaining vegetation growth state, reproduction capacity and resistance related data according to the vegetation growth parameter intervals to generate vegetation growth adaptability feature parameters; Step S33: combining underground water dynamic parameters corresponding to flow distribution, supply characteristics and soil permeability coefficients of underground water of different depths in the target desert area, constructing a coupling effect analysis framework; Step S34: based on the coupling effect analysis framework, cross-mapping the vegetation growth adaptability feature parameters and the underground water dynamic parameters to generate vegetation-underground water coupling effect strength parameters; Step S35: through the vegetation-underground water coupling effect strength parameter quantitative analysis, generating vegetation-underground water ecological coupling relationship data.

4. The artificial intelligence-based desert vegetation-groundwater intelligent monitoring and early warning method according to claim 1, characterized in that, Step S4 comprises the following steps: Step S41: Collect historical ecological data corresponding to the target desert area, and screen out influence factors corresponding to natural evolution of the desert ecology; Step S42: Construct an influence factor quantification system, and calibrate the action strength of each influence factor to generate quantified influence factors; Step S43: Associate and match the desert vegetation-underground water ecological coupling relationship data with the quantified influence factors, and carry out preliminary ecological risk assessment to generate preliminary assessment results; Step S44: Based on the preliminary assessment results, carry out supply imbalance potential assessment analysis to generate a list of vegetation-underground water supply imbalance potential areas; Step S45: Perform ecological support analysis on the list of vegetation-underground water supply imbalance potential areas, and output desert vegetation-underground water ecological fragile area data. 5.The desert vegetation-groundwater intelligent monitoring and early warning method based on artificial intelligence according to claim 4, characterized in that, Step S41 includes the following steps: Step S411: Collect historical ecological data corresponding to the target desert area, including vegetation community succession data, hydrological dynamic evolution data, and meteorological element change data; Step S412: According to the time sequence correlation characteristics of the vegetation community succession data and the hydrological dynamic evolution data, extract the vegetation-underground water co-evolution correlation parameters; Step S413: Based on the vegetation-underground water co-evolution correlation parameters, calculate the response sensitivity parameters of vegetation growth to changes in underground water resources; Step S414: Combine the response sensitivity parameters with the meteorological element change data to identify key driving parameters affecting the stability of vegetation-underground water ecology; Step S415: Through the relevance deduplication and importance sorting of the key driving parameters, screen out influence factors corresponding to natural evolution of the desert ecology. 6.The AI-based desert vegetation-groundwater intelligent monitoring and early warning method according to claim 4, characterized in that, Step S42 includes the following steps: Step S421: Based on the natural laws of material circulation and energy flow in the desert ecosystem, obtain ecological action mechanism data corresponding to each influence factor; Step S422: According to the ecological action mechanism data, construct an initial allocation model of influence factor action strength to generate initial action strength parameters of each factor; Step S423: Introduce the typical vegetation community characteristics and hydrogeological conditions data corresponding to the target desert area to adaptively correct the initial action strength parameters and generate corrected action strength parameters; Step S424: Based on the corrected action strength parameters, calculate the contribution difference parameters of each influence factor to the vegetation-underground water ecological balance; Step S425: Combine the contribution difference parameters to construct an influence factor quantification system, and finally calibrate the quantification standards and action strength coefficients of each factor to generate quantified influence factors.

7. The artificial intelligence-based desert vegetation-groundwater intelligent monitoring and early warning method according to claim 4, characterized in that, Step S43 includes the following steps: Step S431: Obtain spatial distribution characteristics and time sequence evolution characteristic data in the desert vegetation-underground water ecological coupling relationship data to generate coupling characteristic parameter set; Step S432: Based on the coupling characteristic parameter set and the influence factor quantification system, establish a spatial mapping rule to determine the matching relationship between each coupling characteristic and influence factor; Step S433: According to the spatial mapping rule, match the coupling characteristic parameters to the corresponding influence factor quantification dimension to generate a factor-coupling characteristic association matrix; Step S434: calculating the ecological risk correlation degree parameters corresponding to each region based on the factor-coupling feature correlation matrix, the ecological risk correlation degree parameters reflecting the synergistic strength of the coupling feature parameters and the impact factors; Step S435: completing the preliminary ecological risk assessment through the spatial clustering of the ecological risk correlation degree parameters, and generating the preliminary assessment results corresponding to each region. 8.The AI-based desert vegetation-groundwater intelligent monitoring and early warning method according to claim 7, characterized in that, Step S434 includes the following steps: Step S4341: extracting the element feature data in the factor-coupling feature correlation matrix to obtain the initial data of the action strength of each correlation item; Step S4342: calculating the spatio-temporal decay coefficient of the correlation item based on the initial data of the action strength, the spatio-temporal decay coefficient reflecting the change characteristics of the correlation action with the spatial distance and the time span; Step S4343: modifying the initial data of the action strength in combination with the spatio-temporal decay coefficient to generate the effective strength parameter of the correlation action; Step S4344: constructing a risk correlation degree calculation model based on the effective strength parameter of the correlation action, and introducing the elasticity-related parameters of the ecological system as constraint conditions; Step S4345: outputting the ecological risk correlation degree parameters corresponding to each region through the risk correlation degree calculation model. 9.The desert vegetation-groundwater intelligent monitoring and early warning method based on artificial intelligence according to claim 4, characterized in that, Step S44 includes the following steps: Step S441: obtaining the vegetation-underground water ecological coupling relationship data of the target desert region corresponding to the historical ecological stable period, and extracting the corresponding ecological system balance parameters in the historical ecological stable period; Step S442: calculating the ecological risk benchmark interval parameters based on the ecological system balance parameters, the ecological risk benchmark interval parameters reflecting the risk range of the stable operation of the ecological system; Step S443: performing spatial neighborhood comparison on the preliminary assessment results corresponding to each region obtained through the preliminary assessment and the ecological risk benchmark interval parameters to screen out the regions whose spatial neighborhood range of the preliminary assessment results exceeds the ecological risk benchmark interval parameters; Step S444: performing boundary extraction and spatial topology analysis on the regions exceeding the ecological risk benchmark interval parameters to generate the region boundary feature parameters and the spatial connectivity parameters; Step S445: generating the list of potential vegetation-underground water supply imbalance regions in combination with the region boundary feature parameters, the spatial connectivity parameters and the preliminary assessment results. 10.The desert vegetation-groundwater intelligent monitoring and early warning method based on artificial intelligence according to claim 9, characterized in that, Step S442 includes the following steps: Step S4421: extracting the stability-related data corresponding to the structure and function of the ecological system from the vegetation-underground water ecological coupling relationship data of the historical ecological stable period; Step S4422: calculating the anti-interference probability and self-repair probability of the ecological system based on the stability-related data; Step S4423: constructing an ecological balance interval model in combination with the anti-interference probability and the self-repair probability of the ecological system to generate the ecological balance critical interval parameters; Step S4424: introducing the influence data of the historical extreme climate events to modify the ecological balance critical interval parameters to generate the modified balance interval parameters; Step S4425: determining the upper and lower boundaries corresponding to the modified balance interval parameters to generate the ecological risk benchmark interval parameters.

Images