Desert grassland hydrological threshold change identification and health grading system

Abstract

The invention relates to the technical field of ecological environment monitoring, and discloses a desert grassland hydrological threshold change identification and health grading system, which comprises a distributed sensing module for collecting water and soil environment data; the space function division module divides function subareas based on soil moisture and identifies a center node; the time phase state analysis module calculates a phase drift standard deviation of water planting response; a topology dynamic reconstruction module adjusts a threshold reconstruction network according to the phase drift and calculates a structure entropy; the self-adaptive scanning control module monitors the index change to judge the resonance condition, and adaptively adjusts the scanning strategy according to the electric quantity; the health grading module combines the phase drift and the structure entropy to calculate an early warning index so as to determine the health grade. According to the method, the critical precursor is captured through a spatial-temporal characteristic mutual feedback mechanism, and early-stage accurate identification of the hydrological threshold transformation process and quantitative grading of the desert grassland health state are realized.

Description

Technical Field

[0001] This invention relates to the field of ecological environment monitoring technology, specifically a desert grassland hydrological threshold change identification and health classification system. Background Technology

[0002] Desert grasslands lie in the ecological transition zone between arid and semi-arid regions, and their ecosystem stability is highly dependent on the maintenance of soil hydrological processes. Under continuous climate fluctuations and external disturbances, these ecosystems are highly susceptible to nonlinear abrupt changes, that is, crossing a specific hydrological threshold and irreversibly degrading from a vegetated state to a bare land state. Therefore, accurately identifying this hydrological threshold transition process and establishing a scientific health classification system is of great significance for desertification control and ecological restoration decisions.

[0003] Existing desert grassland monitoring technologies mainly rely on long-term vegetation index inversion from satellite remote sensing imagery or on obtaining single-point soil moisture and meteorological data from fixed ground observation stations. By analyzing the time-series trends of these single state variables, technicians typically use linear regression or statistical thresholding methods to infer the current state of the ecosystem, which is routinely applied in macro-scale monitoring or recording of environmentally stable periods.

[0004] However, critical transitions in desert ecosystems are often accompanied by complex spatiotemporal nonlinear feedback characteristics. Relying solely on single variables or simple linear statistical models makes it difficult to capture the weak precursory signals before system collapse, resulting in a significant lag in the identification of hydrological threshold changes. Furthermore, traditional monitoring networks often employ fixed-frequency sampling patterns, making it difficult to establish a dynamic balance between capturing instantaneous high-frequency hydrological fluctuations and ensuring long-term power supply for field equipment, easily leading to the omission of crucial abrupt changes. In addition, existing technologies lack an evaluation mechanism that effectively integrates the dynamic instability of the temporal dimension with the structural connectivity of the spatial dimension, making it difficult for the final health classification results to fully reflect the true resilience and potential degradation risk of desert grasslands.

[0005] Therefore, this invention proposes a desert grassland hydrological threshold change identification and health classification system to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a system for identifying hydrological threshold changes and classifying health in desert grasslands. This system solves the problems of lagging hydrological threshold identification due to the difficulty in effectively capturing the spatiotemporal feedback characteristics during critical ecosystem transitions, and the lack of a comprehensive spatiotemporal quantitative health classification evaluation mechanism.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a desert grassland hydrological threshold transition identification and health classification system, comprising: The distributed sensing module is used to collect data on soil volumetric water content, vegetation canopy temperature, and near-surface air temperature from sensor nodes within the monitoring area. The spatial functional zoning module is used to calculate the correlation coefficient of soil volumetric water content data between the sensor nodes to construct a similarity matrix, perform spectral clustering operation on the similarity matrix to divide the sensor nodes into non-overlapping functional sub-regions, and identify the topological center node in each functional sub-region. The temporal phase analysis module is used to aggregate the state data of sensor nodes in the functional sub-region to construct a local phase space, and to calculate the phase drift standard deviation between soil volumetric water content and vegetation canopy temperature response in the functional sub-region. The topology dynamic reconstruction module is used to dynamically adjust the network connectivity determination threshold according to the phase drift standard deviation, reconstruct the hydrological connectivity network of the functional sub-region based on the adjusted network connectivity determination threshold, and calculate the network structure entropy of the hydrological connectivity network. An adaptive scanning control module is used to monitor the rate of change of the phase drift standard deviation and the rate of change of the network structure entropy to determine the resonance triggering condition. When the resonance triggering condition is met, the average remaining power of the sensor nodes in the functional sub-region is obtained, and the full-domain high-frequency scanning strategy or the topology compensation scanning strategy is selected to be executed based on the average remaining power. The health grading module is used to calculate a comprehensive early warning index based on the phase drift standard deviation and the network structure entropy, and to determine the health level of the desert grassland based on the comprehensive early warning index.

[0008] Preferably, the distributed sensing module controls the sensor nodes to synchronously collect environmental parameters at each discrete time sampling point to generate raw observation data, and calculates the canopy atmospheric temperature difference for each sensor node; The canopy air temperature difference is obtained by subtracting the near-surface air temperature data from the vegetation canopy temperature data collected by the sensor nodes; The distributed sensing module is also used to perform outlier removal processing based on the Laida criterion on the time series of soil volumetric water content data and canopy atmospheric temperature difference.

[0009] Preferably, when performing spectral clustering, the spatial functional partitioning module calculates the degree matrix of the similarity matrix, constructs a non-normalized Laplace matrix, performs eigenvalue decomposition on the non-normalized Laplace matrix to obtain eigenvectors, and performs a clustering algorithm on the eigenvector matrix to divide the functional sub-regions. The process of the spatial functional zoning module identifying the topology center node includes: extracting a local similarity matrix from the similarity matrix that contains only nodes within a specific functional sub-region, calculating the principal feature vector of the local similarity matrix, and marking the sensor node within the functional sub-region corresponding to the component with the largest magnitude in the principal feature vector as the topology center node.

[0010] Preferably, the temporal phase analysis module uses the feature vector centrality output by the spatial functional zoning module to construct weighting coefficients, and performs weighted aggregation on the soil volumetric water content data and canopy atmospheric temperature difference data in the functional sub-region based on the weighting coefficients to generate regional soil volumetric water content sequence and regional vegetation response sequence; The time phase analysis module performs detrending and standardization processing on the regional soil volumetric water content sequence and the regional vegetation response sequence, constructs analytical signals using Hilbert transform to solve the instantaneous phase, calculates the instantaneous phase difference between the regional soil volumetric water content sequence and the regional vegetation response sequence, and statistically analyzes the fluctuation of the instantaneous phase difference within the sliding time window to obtain the phase drift standard deviation.

[0011] Preferably, the topology dynamic reconstruction module has a preset baseline threshold and a threshold upper limit, and establishes a monotonically non-decreasing mapping function to map the phase drift standard deviation to the real-time network connectivity determination threshold. The mapping function is configured such that the network connectivity determination threshold approaches the upper limit of the threshold as the phase drift standard deviation increases; When reconstructing the hydrological connectivity network, the topology dynamic reconstruction module traverses the elements in the local similarity matrix of the functional sub-regions, retains connection weights greater than or equal to the network connectivity determination threshold, and resets connection weights less than the network connectivity determination threshold to zero, thereby generating a weighted adjacency matrix of the hydrological connectivity network.

[0012] Preferably, the topology dynamic reconstruction module uses von Neumann entropy to calculate the network structure entropy; The process of calculating the network structure entropy includes: calculating the degree matrix based on the weighted adjacency matrix, constructing a combined Laplacian matrix, performing trace normalization on the combined Laplacian matrix to generate a density matrix, performing eigenvalue decomposition on the density matrix to obtain a sequence of non-negative eigenvalues, and calculating the entropy value of the non-negative eigenvalue sequence as the network structure entropy based on the Shannon entropy form.

[0013] Preferably, when determining the resonance triggering condition, the adaptive scanning control module calculates the time change rate of the phase drift standard deviation and the time change rate of the network structure entropy, respectively. The resonance triggering condition is set as follows: the time change rate of the phase drift standard deviation is greater than a preset positive threshold for the phase drift change rate, and the time change rate of the network structure entropy is less than a preset negative threshold for the network structure entropy change rate.

[0014] Preferably, the adaptive scanning control module has a preset energy safety threshold; When the average remaining power is greater than or equal to the energy safety threshold, the adaptive scanning control module executes the full-domain high-frequency scanning strategy to control all sensor nodes in the functional sub-region to increase the sampling frequency. When the average remaining power is less than the energy safety threshold, the adaptive scanning control module executes the topology compensation scanning strategy, controls the topology center node in the functional sub-region to perform high-frequency sampling, and controls other ordinary sensor nodes in the functional sub-region to enter a low-power sleep state.

[0015] Preferably, during the execution of the topology compensation scanning strategy, the adaptive scanning control module reconstructs the data of the ordinary sensor nodes that are in a low-power sleep state using the real-time observation values ​​of the topology center node; The reconstruction process uses a linear regression model, which uses the mean and standard deviation of the ordinary sensor nodes and the topology center node in the historical reference period, as well as the correlation coefficient between the two in the local similarity matrix, to infer the virtual data of the ordinary sensor nodes.

[0016] Preferably, the health grading module performs maximum and minimum standardization on the phase drift standard deviation to obtain the temporal risk component, and performs reverse normalization on the network structure entropy to obtain the spatial risk component. The health grading module performs a linear weighted summation of the time risk component and the spatial risk component to obtain the comprehensive early warning index. The health grading module compares the comprehensive early warning index with preset health thresholds, early warning thresholds, and danger thresholds to classify the health level of desert grassland into healthy state, sub-healthy state, early warning state, and danger state.

[0017] This invention provides a system for identifying hydrological threshold changes and classifying the health of desert grasslands. It offers the following advantages: 1. This invention constructs a local phase space encompassing soil moisture and vegetation temperature responses, utilizes Hilbert transform to extract the instantaneous phase of the signal, and calculates phase drift, effectively quantifying the temporal instability of the ecosystem. Simultaneously, based on the degree of phase drift, it dynamically adjusts the network connectivity criteria, reconstructs the hydrological connectivity network, and calculates the network structure entropy. This spatiotemporal feedback mechanism can keenly capture weak signals in the early stages of critical transitions in desert grasslands, achieving early and accurate identification of hydrological threshold transition processes and solving the problem that a single indicator is insufficient to characterize the degradation features of complex ecosystems.

[0018] 2. This invention determines resonance triggering conditions by monitoring the rate of change of phase drift and network structure entropy, and intelligently adjusts the scanning strategy when the system faces the risk of critical transition. Based on the remaining power of the sensor nodes, it adaptively selects either a full-domain high-frequency scan or a compensatory scan mode based on the topology center node, achieving high-precision capture of key ecological processes in energy-constrained desert environments. This mechanism ensures the long-term survival of the monitoring network while avoiding the omission of abrupt changes due to insufficient sampling frequency, significantly improving the system's robustness in responding to sudden hydrological events.

[0019] 3. This invention constructs a comprehensive early warning index by standardizing and fusing a phase drift index, which reflects temporal instability, with a network structure entropy index, which reflects spatial vulnerability. Based on this index, the system classifies the state of desert grasslands into four levels: healthy, sub-healthy, warning, and dangerous, establishing a quantitative ecological health grading evaluation system. This grading method overcomes the limitations of traditional assessments that rely solely on a single state variable, and can intuitively present the overall resilience and degradation risk of the ecosystem, providing a scientific basis for precise management and ecological restoration decisions in desert grasslands. Attached Figure Description

[0020] Figure 1 This is a system architecture diagram for identifying and classifying hydrological threshold changes in desert grasslands according to the present invention. Figure 2 This is a flowchart of the spatiotemporal feedback network reconstruction process of the present invention; Figure 3 This is a flowchart of the adaptive scanning control strategy of the present invention.

[0021] Among them, 10 is the distributed sensing module; 20 is the spatial functional zoning module; 30 is the temporal phase analysis module; 40 is the topology dynamic reconstruction module; 50 is the adaptive scanning control module; and 60 is the health grading module. Detailed Implementation

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see the appendix Figure 1 The system mainly includes: a distributed sensing module 10, a spatial functional zoning module 20, a temporal phase analysis module 30, a topology dynamic reconstruction module 40, an adaptive scanning control module 50, and a health grading module 60. These modules are connected and interact with each other via a data bus or wireless communication protocol, collaboratively completing the entire process from environmental parameter acquisition and spatiotemporal feature extraction to ecological health status determination.

[0024] The distributed sensing module 10 collects soil volumetric water content data, vegetation canopy temperature data, and near-surface air temperature data through several sensor nodes deployed in the monitoring area. The spatial functional zoning module 20 calculates the correlation coefficient of soil volumetric water content data between sensor nodes to construct a similarity matrix, and performs spectral clustering operation on the similarity matrix to divide the sensor nodes into several non-overlapping functional sub-regions. The temporal phase analysis module 30 aggregates the state data of all sensor nodes within the functional sub-region to construct a local phase space and calculates the phase drift standard deviation between soil volumetric water content and vegetation canopy temperature response within the functional sub-region. The topology dynamic reconstruction module 40 dynamically adjusts the network connectivity determination threshold according to the phase drift standard deviation, and reconstructs the hydrological connectivity network of the functional sub-region based on the adjusted network connectivity determination threshold, and then calculates the network structure entropy of the hydrological connectivity network. The adaptive scanning control module 50 monitors the rate of change of the phase drift standard deviation and the network structure entropy to determine the resonance triggering condition. When the resonance triggering condition is met, it obtains the average remaining power of the sensor nodes in the functional sub-region and selects to execute the full-domain high-frequency scanning strategy or the topology compensation scanning strategy based on the average remaining power. The health grading module 60 calculates the comprehensive early warning index based on the phase drift standard deviation and network structure entropy, and determines the health level of desert grassland based on the comprehensive early warning index.

[0025] See attached document Figure 1 The distributed sensing module 10, serving as the system's data acquisition front-end, is physically deployed within the desert grassland monitoring area to be identified and classified. This monitoring area refers to a selected, representative desert grassland plot. The distributed sensing module 10 includes several fixed-location sensor nodes distributed throughout the monitoring area, denoted as the sensor node set. ,in `Number` is a natural number representing the total number of sensor nodes. Each sensor node... It is a hardware unit that integrates data acquisition, processing, and transmission functions. These nodes work together to form a multimodal sensor network capable of acquiring groundwater hydrological information, surface vegetation physiological information, and environmental background information. The specific sensor configuration is as follows: Soil volumetric water content data collection: at each node location within the monitoring area. Soil moisture sensors with frequency domain reflectance (FDR) or time domain reflectance (TDR) probes are buried in the depth layer where the vegetation roots are mainly distributed to directly measure the soil volumetric water content.

[0026] Vegetation canopy temperature data collection: also at each node location A non-contact infrared temperature probe is installed on a bracket, and its field of view is adjusted to focus vertically downwards or tilted onto the surface of the vegetation canopy to measure the radiation temperature of the vegetation.

[0027] Near-surface air temperature data acquisition: Within the monitoring area (which can be co-located with the vegetation canopy temperature sensor, or a small number of representative reference points can be set up), a near-surface air temperature sensor with a thermistor or thermocouple probe is installed. This sensor is placed inside a radiation shield to avoid interference from direct solar radiation, and its installation height is consistent with the measurement height of the vegetation canopy infrared temperature sensor, thus ensuring the correspondence between the ambient background temperature and the canopy temperature on a spatial vertical scale.

[0028] The specific circuit connection methods, power supply module design, and signal analog-to-digital conversion circuits of the above-mentioned sensors are well-known technologies to those skilled in the art and will not be elaborated here.

[0029] The distributed sensing module 10 performs synchronous sampling operations at each discrete time sampling point. The system controls several sensor nodes to synchronously collect environmental parameters and generate raw observation data. To accurately characterize the physiological response of desert grassland under water-limited conditions, this invention not only directly utilizes the collected data but also preprocesses the data to construct an index reflecting the plant's water stress state, namely, the canopy atmospheric temperature difference.

[0030] The specific variable definitions and calculation process are as follows: Define time No. The raw observation data vector of each sensor node : ; in, This represents the soil volumetric water content data collected by the sensor nodes, in cubic meters per cubic meter. This represents the vegetation canopy temperature data collected by the sensor nodes, in degrees Celsius. This represents the near-surface air temperature data collected by the sensor node, in degrees Celsius.

[0031] Based on the above original observation data vector, calculate the node at time [time]. Canopy temperature difference : ; After completing data collection and index calculation, the distributed sensing module 10 pairs of data including soil volumetric water content. Temperature difference with the canopy The dataset undergoes outlier removal processing. Specifically, this invention processes the two time-series data generated by each sensor node (i.e., ... sequence sum The sequence was independently analyzed using the Laida criterion (3σ criterion). This method calculates the mean and standard deviation of the data within a sliding time window, and identifies outliers that deviate from the mean by more than three times the standard deviation as outliers for removal or interpolation correction. This step aims to eliminate noise caused by transient sensor hardware failures or external electromagnetic interference, ensuring the reliability of the data used in subsequent spatiotemporal analysis.

[0032] See attached document Figure 1 The spatial functional zoning module 20, as the core unit for the system to handle spatial heterogeneity, connects the distributed sensing module 10 with the subsequent temporal phase analysis module 30. The main function of the spatial functional zoning module 20 is to reorganize the physically discrete sensor nodes within the monitoring area into several functional sub-regions with consistent hydrological functions based on the temporal synchronization of soil moisture dynamics, and to select the most representative topological center node from each functional sub-region.

[0033] In the specific implementation process, the spatial functional zoning module 20 first establishes data associations between sensor nodes. This spatial functional zoning module 20 sets a time length of... A sliding time window is defined, and soil volumetric water content data sequences of all sensor nodes within this window are obtained from the distributed sensing module 10. For any two sensor nodes within the monitoring area... and Soil volumetric water content data sequences were extracted respectively. and The spatial functional zoning module 20 uses the Pearson correlation coefficient algorithm to quantify the degree of linear synchronization between the two sequences in the time dimension. For sequences containing... The monitoring network of sensor nodes, constructed by the spatial functional zoning module 20, forms a... 3D similarity matrix .

[0034] Pearson correlation coefficient and similarity matrix The calculation logic is as follows: ; ; in, Indicates the first time within the sliding time window Sensor nodes at each sampling time Soil volumetric water content data, This indicates the sensor nodes within the window. The arithmetic mean of soil volumetric water content data. To ensure that the clustering results reflect positive hydrological synergies, the spatial functional zoning module 20 performs a filtering operation, removing items with negative results (negative correlation) or autocorrelation (diagonal elements) from the similarity matrix. The middle value is zero.

[0035] Based on the constructed similarity matrix The spatial functional zoning module 20 uses a spectral clustering algorithm to group sensor nodes. This module maps the sensor network to a weighted undirected graph, where sensor nodes correspond to vertices of the graph, and the similarity matrix... The elements in the matrix correspond to the weights of the edges. The spatial functional partitioning module 20 calculates the degree matrix. degree matrix Given a diagonal matrix, the first element on its diagonal is... element Equal to the similarity matrix No. The sum of all elements in the row. Based on this, construct the nonnormalized Laplace matrix. The calculation formula is: .

[0036] Subsequently, the spatial functional zoning module 20 pairs of Laplace matrices Perform eigenvalue decomposition to solve the characteristic equation. This yields the eigenvalue sequence and its corresponding eigenvectors. The spatial functional partitioning module 20 determines the optimal number of clusters based on the principle of maximizing eigenvalue gaps. and select the first The eigenvectors corresponding to the smallest non-zero eigenvalues ​​form a feature matrix. K-means clustering is then applied to the row vectors of this feature matrix to divide all sensor nodes within the monitoring area into clusters. A number of non-overlapping functional sub-regions, denoted as Sensor nodes within each functional sub-region exhibit high statistical similarity in soil moisture dynamics. For the specific iterative convergence process of the K-means algorithm, those skilled in the art can refer to relevant textbooks or standard algorithm libraries; it is well-known technology in this field and will not be elaborated upon here.

[0037] After completing the partitioning, the spatial functional partitioning module 20 further partitions each independent functional sub-area. Micro-topological analysis is performed to identify topological central nodes. The spatial functional zoning module 20 uses a global similarity matrix... Extract only the sub-region containing this function The rows and columns of the internal nodes form a local similarity matrix. Based on this local similarity matrix Calculate the eigenvector centrality of each sensor node within the sub-region.

[0038] The specific process for identifying the topology center node is as follows: Calculate the local similarity matrix Maximum eigenvalue and its corresponding principal feature vector The principal eigenvector satisfies the following relation: ; Main eigenvectors The Each component The absolute value quantifies the connectivity and information influence of the corresponding sensor nodes within that functional sub-region. The spatial functional zoning module 20 traverses the functional sub-regions. For all sensor nodes within the functional sub-region, compare the magnitudes of their corresponding principal feature vector components, and mark the sensor node with the largest component magnitude as the topological center node of that functional sub-region. : ; The central node of this topology and its associated local similarity matrix Stored by this system.

[0039] See attached document Figure 1 The temporal phase analysis module 30 is connected to the spatial functional zoning module 20, and it explores the nonlinear dynamic characteristics of desert grassland in the time dimension. The temporal phase analysis module 30 uses each functional sub-region divided by the spatial functional zoning module 20 as the basic analysis unit, constructs a macroscopic local phase space through data aggregation, and quantifies the stability of desert grassland based on phase synchronization theory.

[0040] In practice, the time phase analysis module 30 first performs weighted aggregation on the multi-source data within the functional sub-region to generate state data representing the overall behavior of that sub-region. For any specific functional sub-region... The time phase analysis module 30 obtains the data of all sensor nodes within the sub-region at the same time. Soil volumetric water content data Data on atmospheric temperature difference in the canopy Here, the atmospheric temperature difference data for the canopy is... As a representation of the physiological response of vegetation canopy temperature data to the environment, it was incorporated into the aggregation process.

[0041] The temporal phase analysis module 30 uses the eigenvector centrality output by the spatial functional partitioning module 20 to construct weighted coefficients. Let the sensor node... The eigenvector centrality is Then its corresponding normalized weight for Based on this weight, the time phase analysis module 30 calculates the functional sub-region. Regional soil volumetric water content sequence and regional vegetation response sequences : ; ; The above and Together, they constitute the aggregated state data of this functional sub-region. Subsequently, the time phase analysis module 30 uses the regional soil volumetric water content sequence. The horizontal axis represents the regional vegetation response sequence. Using the vertical axis as the ordinate, a two-dimensional local phase space is constructed. In this local phase space, the trajectory of desert grassland state evolution over time intuitively demonstrates the dynamic coupling relationship between water-driven processes and vegetation responses.

[0042] Based on this, the temporal phase analysis module 30 calculates the standard deviation of the phase drift between the regional soil volumetric water content and the vegetation canopy temperature response (i.e., the regional vegetation response sequence). Since desert grassland ecological data typically exhibits non-stationary and non-linear characteristics, the temporal phase analysis module 30 employs Hilbert transform to extract the instantaneous phase of the signal.

[0043] First, set a sliding time window. Within this window, the sequence and Perform detrending and Z-score normalization on any processed sequence. Construct its analytical signal : ; in, Represents the Hilbert transform operator; The imaginary unit; Instantaneous amplitude; The instantaneous phase is calculated by the time phase analysis module 30 through signal analysis to determine the instantaneous phase of the regional soil volumetric water content sequence. Instantaneous phase of regional vegetation response sequence .

[0044] Next, calculate the time intervals of both. instantaneous phase difference : ; To eliminate phase winding, the calculation results were limited to... Within the interval. Finally, the time phase analysis module 30 statistically analyzes the instantaneous phase difference sequence within the sliding time window. Calculate the standard deviation of phase drift to account for fluctuations within the range. : ; in, This is the arithmetic mean of the instantaneous phase differences within this window. The standard deviation of the phase drift. It is a key indicator for measuring the degree of internal coupling in desert grasslands. Under steady-state conditions, vegetation changes synchronously with water fluctuations, and the phase difference remains stable. The response is relatively small; however, before the critical transition occurs, the vegetation's regulatory function is impaired, leading to a delayed or disordered response, manifested as drastic fluctuations in phase difference. Significantly increased.

[0045] See attached document Figure 1 and Figure 2 The topology dynamic reconstruction module 40, connected to the temporal phase analysis module 30, performs network structure analysis based on a spatiotemporal feedback mechanism. This module receives the phase drift standard deviation as a quantitative input to the temporal instability of desert grassland and dynamically adjusts the construction rules of the spatial network accordingly, thereby reconstructing the hydrological connectivity network of functional sub-regions and calculating the network structure entropy. This process aims to reveal the spatial topological vulnerability of desert grassland under ecological pressure.

[0046] In practice, the topology dynamic reconstruction module 40 first dynamically adjusts the network connectivity determination threshold based on the phase drift standard deviation. The network connectivity determination threshold refers to the minimum correlation coefficient standard used to determine whether there is a valid hydrological association between two sensor nodes when constructing the topology network. Only when the soil moisture correlation coefficient between the nodes is greater than or equal to this threshold does the topology dynamic reconstruction module 40 consider that there is a substantial connection between them.

[0047] The topology dynamic reconstruction module 40 has a preset benchmark threshold. and a threshold upper limit The baseline threshold refers to the lowest connectivity criterion used to determine the connectivity of desert grassland under ideal healthy conditions (i.e., minimal phase drift). Its value is typically set within the range of [0.3, 0.5], representing a statistically moderate level of correlation. The upper threshold refers to the highest connectivity criterion allowed for desert grassland under extremely unstable conditions (i.e., maximum phase drift). Its value is typically set within the range of [0.7, 0.9], representing a strong level of correlation. The purpose of setting the upper threshold is to prevent the network from completely breaking down due to excessively high thresholds, thus rendering the analysis meaningless.

[0048] The topology dynamic reconstruction module 40 establishes a monotonically non-decreasing mapping function to convert the input phase drift standard deviation. Mapped to real-time network connectivity determination threshold In this embodiment, the mapping function specifically adopts a sigmoid function form, and its mathematical expression is as follows: ; in, The preset critical reference value for phase drift represents the level of phase fluctuation when desert grassland begins to enter a state of alert. The sensitivity coefficient is used to control the response rate of the threshold as a function of phase drift. This function ensures that the decision threshold smoothly and monotonically decreases from the reference threshold as the standard deviation of the phase drift increases. To the upper limit of the threshold This approximation allows for more stringent adaptive logic where higher risks result in stricter screening. The mathematical properties of the Sigmoid function itself are well-known in applied mathematics and will not be elaborated upon here.

[0049] Subsequently, the topology dynamic reconstruction module 40 determines the threshold based on the adjusted network connectivity. Reconstruct the hydrological connectivity network of this functional sub-region. The hydrological connectivity network refers to a weighted topology graph, after threshold filtering, composed of all sensor nodes within the functional sub-region and their interconnections that meet high-strength correlation requirements. Physically, this hydrological connectivity network represents the set of backbone paths that can still maintain effective water transport and signal coupling under current ecological pressures.

[0050] The topology dynamic reconstruction module 40 obtains the local similarity matrix generated by the spatial functional partitioning module 20. For each element in the local similarity matrix, a threshold filter is applied. Specifically, each element in the local similarity matrix is ​​iterated through; if the element value is greater than or equal to the current network connectivity threshold, a threshold is applied. If the element value is less than the threshold, the connection weight is retained; if the element value is less than the threshold, the connection weight is reset to zero.

[0051] After the above filtering operations, a new sparse matrix is ​​formed, which is the weighted adjacency matrix of the hydrological connectivity network. : ; The weighted adjacency matrix The reconstructed hydrological connectivity network is mathematically defined, reflecting the hydrological transport paths that still maintain effective connections within functional sub-regions under current ecological pressures.

[0052] Next, the topology dynamic reconstruction module 40 calculates the network structure entropy of the hydrological connectivity network. In this embodiment, von Neumann entropy is used to quantify the network's structural complexity and information redundancy. The topology dynamic reconstruction module 40 first calculates the network structure entropy based on the weighted adjacency matrix. Calculate the degree matrix of the nodes The diagonal elements of the degree matrix are equal to the sum of the elements in the corresponding row of the weighted adjacency matrix. Then, a combined Laplace matrix is ​​constructed. .

[0053] To conform to the probabilistic definition of entropy, the topology dynamic reconstruction module 40 performs trace normalization on the combined Laplace matrix to generate the density matrix. The specific calculation involves combining the Laplace matrix. Divide by the trace of the matrix (i.e., the sum of the diagonal elements): ; The subsequent topology dynamic reconstruction module has 40 pairs of density matrices. Performing eigenvalue decomposition yields a sequence of nonnegative eigenvalues. And satisfy Based on this eigenvalue sequence, the network structure entropy is calculated in the form of Shannon entropy. : ; During the calculation process, if the eigenvalues If it is zero, then the corresponding term The network structure entropy is calculated based on the limit approaching zero. The lower the entropy value, the simpler, more fragmented, or isolated the hydrological connectivity network structure, and the worse the connectivity of desert grasslands. Conversely, the higher the entropy value, the more complex, homogeneous, and resilient the network structure. This network structure entropy serves as a key indicator for measuring the spatial health status.

[0054] See attached document Figure 1 and Figure 3The adaptive scan control module 50 constructs the closed-loop feedback control link of the system. The function of this adaptive scan control module 50 is to intelligently decide and execute the optimal scan strategy based on the energy status of the sensor network when a strong signal indicating an impending critical transition in the desert grassland is detected.

[0055] In practice, the adaptive scan control module 50 continuously monitors the phase drift standard deviation transmitted from the preceding module. With network structure entropy The adaptive scan control module 50 calculates the rate of change of these two indicators over time. In discrete time series analysis, the adaptive scan control module 50 approximates the rate of change using backward differencing: Rate of change of phase drift standard deviation ; Rate of change of network structure entropy ; Subsequently, the adaptive scanning control module 50 determines the resonance triggering condition based on the aforementioned rate of change. The resonance triggering condition refers to the phenomenon where the instability of desert grassland in the temporal dimension (manifested as increased phase drift) and the structural fragility in the spatial dimension (manifested as decreased network structure entropy) simultaneously deteriorate rapidly. The adaptive scanning control module 50 identifies this condition by setting dual thresholds: ; Among them, the positive threshold of the phase drift rate of change This is a critical rate parameter used to determine whether desert grassland is unstable in the time dimension. Its value is usually set to 2-3 times the standard deviation of long-term historical observations, aiming to filter out small fluctuations caused by environmental background noise; the network structure entropy change rate threshold. This is a critical rate parameter (positive value) used to determine whether the spatial structure of desert grasslands is rapidly collapsing. Its setting is intended to capture moments when the topological connectivity of hydrological connectivity networks declines sharply. When the rate of change of the phase drift standard deviation exceeds... Furthermore, the rate of change of network structure entropy is lower than that of negative networks. At that time, it is determined that the resonance triggering condition is met, which means that the desert grassland is facing an urgent risk of critical transformation.

[0056] When the resonance triggering condition is met, the adaptive scan control module 50 immediately obtains the average remaining power of all sensor nodes in the functional sub-region. The average remaining power is obtained by querying the power status registers of each node in the distributed sensing module 10 and calculating the arithmetic mean.

[0057] Next, the adaptive scanning control module 50 selects one of two mutually exclusive scanning strategies based on the average remaining power, and the determination criterion is whether the average remaining power is higher than the energy safety threshold. Energy security threshold It is a preset battery percentage warning line (e.g., set to 30% of the total capacity). Its physical meaning is to ensure that after the node performs high-energy-consuming operations, it still retains a minimum energy reserve sufficient to maintain basic communication and monitor network survival, preventing node failure and network paralysis due to oversampling.

[0058] The two mutually exclusive scanning strategies include the global high-frequency scanning strategy and the topology compensation scanning strategy.

[0059] Full-domain high-frequency scanning strategy: if the average remaining power If the adaptive scanning control module 50 sends control commands to all sensor nodes within the functional sub-region, uniformly increasing their sampling frequency to the preset maximum monitoring frequency. This strategy aims to capture all the details of the critical transition process between desert and grassland at the highest temporal resolution, providing the most comprehensive data support for accurate early warning and mechanistic analysis.

[0060] Topology compensation scanning strategy: if the average remaining power If this occurs, the strategy is activated to maintain core monitoring capabilities under energy constraints. This strategy only controls the topology center node within the functional sub-region. Perform high-frequency sampling (frequency is At the same time, it sends instructions to all other ordinary sensor nodes in the sub-region, causing them to enter a low-power sleep state.

[0061] During the execution of the topology compensation scan strategy, to ensure the integrity of spatial data, the adaptive scan control module 50 uses stored spatial correlation information to reconstruct the data of dormant nodes in real time. This module retrieves the local similarity matrix generated by the spatial functional partitioning module 20. And based on the real-time observations of the topology center node The simulation is then conducted. The real-time observation value of the topology center node refers to the physical measurement value of soil volumetric water content (or other relevant ecological indicators) directly collected and reported by the topology center node sensor, which is in a high-frequency sampling state, at the current sampling moment. It is the only real and accurate first-hand data source within the entire functional sub-region at this moment, and plays the role of the information cornerstone supporting the reconstruction of the spatial state of the entire sub-region.

[0062] The adaptive scan control module 50 uses the real-time observations through a linear regression model to calculate the value of any dormant node. virtual data Its reconstruction formula is as follows: ; in, and They are nodes The average soil volumetric water content of the topological center node during the historical reference period; and This represents the corresponding historical standard deviation; For the nodes in the local similarity matrix and topology center node The element value is the Pearson correlation coefficient between the two calculated in the spatial functional zoning module 20. .

[0063] Through the above mechanism, the adaptive scanning control module 50 realizes intelligent scheduling of monitoring resources (energy and bandwidth) at critical moments, ensuring that the desert grassland hydrological threshold change identification and health classification system can effectively respond to ecological critical changes under different energy conditions.

[0064] See attached document Figure 1 The health grading module 60 serves as the system's decision output. This module is responsible for standardizing and integrating the instability characteristics of desert grassland in the time dimension with the topological vulnerability characteristics in the spatial dimension, calculating a comprehensive early warning index, and classifying the health status of desert grassland accordingly.

[0065] In specific implementation, the health grading module 60 first receives the phase drift standard deviation of the functional sub-region. and network structure entropy Since these two indicators have different physical dimensions and directions of change (phase drift standard deviation is positively correlated with risk, while network structure entropy is negatively correlated with risk), the health grading module 60 first normalizes and unifies their directions. For the phase drift standard deviation, the health grading module 60 uses a max-min standardization method to map it into a time risk component. : ; in, This is the historical minimum value (or theoretical zero value) of the indicator. This is a preset risk ceiling value. After processing, the time risk component... The value range is [0,1], with a larger value indicating higher instability in the time dimension. For network structure entropy, considering that a lower entropy value represents a simpler network structure and weaker anti-interference ability (i.e., higher risk), the health grading module 60 performs reverse normalization to obtain the spatial risk component. : ; in, and These represent the theoretical minimum and maximum values ​​of the network structure entropy, respectively. This formula ensures the spatial risk component. The value range is also [0,1], and the larger the value, the higher the vulnerability of the spatial topology, thus keeping in line with the direction of change of the time risk component.

[0066] Subsequently, the health grading module 60 calculates a comprehensive early warning index based on these two risk components. This index is a dimensionless scalar value used to comprehensively quantify the overall risk level of a critical transition occurring in the current functional sub-region. The calculation formula employs a linear weighted summation model: ; in, and These are the time risk weight coefficient and the spatial risk weight coefficient, respectively, and they satisfy the normalization constraint. In this embodiment, the default settings are... This means that temporal instability and spatial fragmentation are considered equally important in contributing to the collapse of desert grassland ecological functions. In practical applications, technicians can also adjust the weighting ratios based on the specific ecological characteristics of different desert grassland types.

[0067] Finally, the health grading module 60 is based on the calculated comprehensive early warning index. Determine the health level of the desert grassland. This health grading module 60 presets a set of grading thresholds, including health thresholds. Warning threshold and danger threshold In this embodiment, take , , The health grading module 60 will integrate the early warning index. Compared to this set of thresholds, the value range is divided into four consecutive intervals, each corresponding to a specific health state: Health Level (Level I): When (Right now When the time frame is reached, the desert grassland is considered to be in a healthy state. At this time, the desert grassland has good temporal synchronization, strong spatial connectivity, and can effectively resist external interference.

[0068] Sub-health level (Level II): When (Right now When the desert grassland exhibits slight phase fluctuations or localized connection breaks, although no qualitative changes have occurred, it is determined to be in a sub-healthy state.

[0069] Warning Level (Level III): When (Right now When the indicators show significant deterioration, the desert grassland is considered to be in a warning state, approaching a critical point and facing a high risk of degradation.

[0070] Hazard level (Level IV): When (Right now When the comprehensive warning index is extremely high, it indicates that the desert grassland may have already experienced or is about to experience irreversible ecological functional collapse, requiring immediate intervention.

[0071] Through the above steps, the health grading module 60 transforms complex multidimensional monitoring data into intuitive grading results, achieving a precise quantitative assessment of the health status of desert grasslands.

[0072] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A system for identifying hydrological threshold transitions and classifying the health of desert grasslands, characterized in that, include: The distributed sensing module is used to collect data on soil volumetric water content, vegetation canopy temperature, and near-surface air temperature from sensor nodes within the monitoring area. The spatial functional zoning module is used to calculate the correlation coefficient of soil volumetric water content data between the sensor nodes to construct a similarity matrix, perform spectral clustering operation on the similarity matrix to divide the sensor nodes into non-overlapping functional sub-regions, and identify the topological center node in each functional sub-region. The temporal phase analysis module is used to aggregate the state data of sensor nodes in the functional sub-region to construct a local phase space, and to calculate the phase drift standard deviation between soil volumetric water content and vegetation canopy temperature response in the functional sub-region. The topology dynamic reconstruction module is used to dynamically adjust the network connectivity determination threshold according to the phase drift standard deviation, reconstruct the hydrological connectivity network of the functional sub-region based on the adjusted network connectivity determination threshold, and calculate the network structure entropy of the hydrological connectivity network. An adaptive scanning control module is used to monitor the rate of change of the phase drift standard deviation and the rate of change of the network structure entropy to determine the resonance triggering condition. When the resonance triggering condition is met, the average remaining power of the sensor nodes in the functional sub-region is obtained, and the full-domain high-frequency scanning strategy or the topology compensation scanning strategy is selected to be executed based on the average remaining power. The health grading module is used to calculate a comprehensive early warning index based on the phase drift standard deviation and the network structure entropy, and to determine the health level of the desert grassland based on the comprehensive early warning index.

2. The desert grassland hydrological threshold transition identification and health classification system according to claim 1, characterized in that, The distributed sensing module controls the sensor nodes to synchronously collect environmental parameters at each discrete time sampling point to generate raw observation data, and calculates the canopy atmospheric temperature difference for each sensor node. The canopy air temperature difference is obtained by subtracting the near-surface air temperature data from the vegetation canopy temperature data collected by the sensor nodes; The distributed sensing module is also used to perform outlier removal processing based on the Laida criterion on the time series of soil volumetric water content data and canopy atmospheric temperature difference.

3. The desert grassland hydrological threshold transition identification and health classification system according to claim 1, characterized in that, When performing spectral clustering, the spatial functional partitioning module calculates the degree matrix of the similarity matrix, constructs a non-normalized Laplace matrix, performs eigenvalue decomposition on the non-normalized Laplace matrix to obtain eigenvectors, and performs a clustering algorithm on the eigenvector matrix to divide the functional sub-regions. The process of the spatial functional zoning module identifying the topology center node includes: extracting a local similarity matrix from the similarity matrix that contains only nodes within a specific functional sub-region, calculating the principal feature vector of the local similarity matrix, and marking the sensor node within the functional sub-region corresponding to the component with the largest magnitude in the principal feature vector as the topology center node.

4. The desert grassland hydrological threshold transition identification and health classification system according to claim 3, characterized in that, The temporal phase analysis module uses the feature vector centrality output by the spatial functional zoning module to construct weighting coefficients. Based on the weighting coefficients, the soil volumetric water content data and canopy atmospheric temperature difference data in the functional sub-region are weighted and aggregated to generate regional soil volumetric water content sequence and regional vegetation response sequence. The time phase analysis module performs detrending and standardization processing on the regional soil volumetric water content sequence and the regional vegetation response sequence, constructs analytical signals using Hilbert transform to solve the instantaneous phase, calculates the instantaneous phase difference between the regional soil volumetric water content sequence and the regional vegetation response sequence, and statistically analyzes the fluctuation of the instantaneous phase difference within the sliding time window to obtain the phase drift standard deviation.

5. The desert grassland hydrological threshold transition identification and health classification system according to claim 1, characterized in that, The topology dynamic reconstruction module has a preset baseline threshold and threshold upper limit, and establishes a monotonically non-decreasing mapping function to map the phase drift standard deviation to the real-time network connectivity determination threshold. The mapping function is configured such that the network connectivity determination threshold approaches the upper limit of the threshold as the phase drift standard deviation increases; When reconstructing the hydrological connectivity network, the topology dynamic reconstruction module traverses the elements in the local similarity matrix of the functional sub-regions, retains connection weights greater than or equal to the network connectivity determination threshold, and resets connection weights less than the network connectivity determination threshold to zero, thereby generating a weighted adjacency matrix of the hydrological connectivity network.

6. The desert grassland hydrological threshold transition identification and health classification system according to claim 5, characterized in that, The topology dynamic reconstruction module uses von Neumann entropy to calculate the network structure entropy. The process of calculating the network structure entropy includes: calculating the degree matrix based on the weighted adjacency matrix, constructing a combined Laplacian matrix, performing trace normalization on the combined Laplacian matrix to generate a density matrix, performing eigenvalue decomposition on the density matrix to obtain a sequence of non-negative eigenvalues, and calculating the entropy value of the non-negative eigenvalue sequence as the network structure entropy based on the Shannon entropy form.

7. The desert grassland hydrological threshold transition identification and health classification system according to claim 1, characterized in that, When determining the resonance triggering condition, the adaptive scanning control module calculates the time change rate of the phase drift standard deviation and the time change rate of the network structure entropy, respectively. The resonance triggering condition is set as follows: the time change rate of the phase drift standard deviation is greater than a preset positive threshold for the phase drift change rate, and the time change rate of the network structure entropy is less than a preset negative threshold for the network structure entropy change rate.

8. The desert grassland hydrological threshold transition identification and health classification system according to claim 3, characterized in that, The adaptive scanning control module has a preset energy safety threshold. When the average remaining power is greater than or equal to the energy safety threshold, the adaptive scanning control module executes the full-domain high-frequency scanning strategy to control all sensor nodes in the functional sub-region to increase the sampling frequency. When the average remaining power is less than the energy safety threshold, the adaptive scanning control module executes the topology compensation scanning strategy, controls the topology center node in the functional sub-region to perform high-frequency sampling, and controls other ordinary sensor nodes in the functional sub-region to enter a low-power sleep state.

9. The desert grassland hydrological threshold transition identification and health classification system according to claim 8, characterized in that, During the execution of the topology compensation scanning strategy, the adaptive scanning control module reconstructs the data of the ordinary sensor nodes that are in a low-power sleep state using the real-time observation values ​​of the topology center node. The reconstruction process uses a linear regression model, which uses the mean and standard deviation of the ordinary sensor nodes and the topology center node in the historical reference period, as well as the correlation coefficient between the two in the local similarity matrix, to infer the virtual data of the ordinary sensor nodes.

10. The desert grassland hydrological threshold transition identification and health classification system according to claim 1, characterized in that, The health grading module performs maximum and minimum standardization on the phase drift standard deviation to obtain the temporal risk component, and performs reverse normalization on the network structure entropy to obtain the spatial risk component. The health grading module performs a linear weighted summation of the time risk component and the spatial risk component to obtain the comprehensive early warning index. The health grading module compares the comprehensive early warning index with preset health thresholds, early warning thresholds, and danger thresholds to classify the health level of desert grassland into healthy state, sub-healthy state, early warning state, and danger state.

Images