Ecological hydrological monitoring method and system based on sensor network construction

Abstract

The invention discloses an ecological hydrological monitoring method and system based on sensor network construction, and the method comprises the steps: analyzing a monitoring demand, setting a sensor layout scheme, and carrying out the sensor network construction in a target monitoring region through the sensor layout scheme; performing ecological monitoring on the target region by using the constructed sensor network to obtain regional ecological monitoring information, and constructing multi-modal ecological monitoring features according to the regional ecological monitoring information; and greenhouse gas release prediction and carbon flux inversion are performed on the target area through the multi-modal ecological monitoring features, whether an ecological event exists in the target area is judged, and ecological adjustment early warning is performed. The analysis bottleneck and response disadvantage of a traditional monitoring method are broken through, high-precision decision response is provided for ecological system service evaluation and climate change response, and meanwhile support is provided for guaranteeing regional ecological stability.

Description

Technical Field

[0001] The present invention relates to the technical field of ecological monitoring, and in particular, to an ecological hydrological monitoring method and system based on a sensor network. Background Art

[0002] As the core means to understand the interaction between the terrestrial water cycle and the ecosystem, the development of ecological hydrological monitoring technology has always been closely related to the innovation of sensor networks and data analysis methods. Traditional ecological hydrological monitoring mainly relies on ground observations at discrete points and low-resolution remote sensing data, which has limitations such as insufficient spatial coverage, poor temporal continuity, and weak multi-factor collaborative analysis ability, and it is difficult to accurately depict the spatio-temporal heterogeneity of ecological hydrological processes.

[0003] In the prior art, although wireless sensor networks have been applied to the collection of local ecological parameters, most of the node deployments adopt static topological structures, which cannot adapt to signal attenuation and energy consumption balance under complex terrains, and do not fully consider the spatial correlation of sensor measurement errors and the constraints of ecological process mechanisms, resulting in significant scale effects and uncertainty accumulation in carbon flux inversion results. In addition, in response to sudden hydrological events (such as floods and droughts), due to the lack of dynamic data assimilation capabilities and coordinated scheduling mechanisms for edge computing resources, it is difficult to achieve real-time response of the monitoring system and online optimization of model parameters, which severely restricts the accurate assessment of ecosystem service functions and the effectiveness of disaster early warning.

[0004] Therefore, how to improve the comprehensiveness and real-time nature of ecological hydrological monitoring and also take into account the perception ability of ecological changes is an urgent problem to be solved. Summary of the Invention

[0005] The present invention overcomes the defects of the prior art and provides an ecological hydrological monitoring method and system based on a sensor network.

[0006] To achieve the above object, the first aspect of the present invention provides an ecological hydrological monitoring method based on a sensor network, including:

[0007] Analyze the monitoring requirements and set a sensor layout plan, and construct a sensor network in the target monitoring area through the sensor layout plan;

[0008] Use the constructed sensor network to conduct ecological monitoring on the target area to obtain regional ecological monitoring information, and construct multi-modal ecological monitoring features based on the regional ecological monitoring information;

[0009] Predict greenhouse gas emissions and invert carbon fluxes in the target area through the multi-modal ecological monitoring features, and determine whether there are ecological events in the target area to issue ecological adjustment warnings.

[0010] In this solution, the analysis of the monitoring requirements specifically includes:

[0011] Obtain the monitoring expectation information, decompose the monitoring target based on the monitoring expectation information to obtain the expected monitoring data for ecological hydrological monitoring of the target area, and obtain the expected monitoring information;

[0012] Generate a retrieval tag according to the expected monitoring information, perform monitoring sensor matching in the preset monitoring device database, analyze the monitoring sensors that match the current monitoring requirements, and obtain the matching sensor information;

[0013] Use data retrieval means to obtain the regional geographical information of the target area, discretize the target area into a number of grids of the same size based on the regional geographical information, and obtain the regional grid map;

[0014] Extract the obstacle features in the target area according to the regional geographical information, map the obstacles into a network mask through the extracted obstacle features, calibrate the non-obstacle area, and update the regional grid map to obtain the updated regional grid map.

[0015] In this solution, the setting of the sensor layout scheme constructs a sensor network in the target monitoring area through the sensor layout scheme, specifically including:

[0016] Obtain the matching sensor information, extract the functional features, energy consumption features, and communication ability features of the monitoring sensors that match the current monitoring requirements through the matching sensor information, and obtain the matching sensor feature information;

[0017] Obtain the updated regional grid map, construct a sensor network in combination with the matching sensor information, perform regional sensor network layout analysis using the greedy algorithm, and set the objective function according to coverage maximization and energy consumption minimization through weighted fusion;

[0018] Set the constraint conditions based on the communication distance, the number of nodes, and the obstacle distribution, add penalty terms to the objective function through the set constraint conditions, define each non-obstacle grid of the updated regional grid map as a feasible node, and perform population initialization in combination with the matching sensor feature information;

[0019] Calculate the fitness value of each individual in the initial population, select the optimal candidate individual through the calculated fitness value, and if there are multiple individuals with the same fitness value, select the individual with the smallest distance from the center of the uncovered area as the optimal candidate individual;

[0020] Perform iterative analysis until the termination condition is met and output the final solution set, generate the sensor network layout scheme for the target area according to the final solution set, and construct a sensor network in the target area.

[0021] In this solution, ecological monitoring of the target area is carried out using the constructed sensor network to obtain regional ecological monitoring information, and multi-modal ecological monitoring features are constructed based on the regional ecological monitoring information, specifically including:

[0022] Ecological monitoring of the target area is carried out through the sensor network constructed in the target area to obtain regional ecological monitoring information, and the regional ecological monitoring information includes hydrological monitoring data, greenhouse gas monitoring data, and meteorological monitoring data;

[0023] Data preprocessing is performed on the regional ecological monitoring information. The data standard deviation of the regional ecological monitoring information is calculated through data standard deviation calculation, and data outside the preset standard deviation range is defined as abnormal data, and outliers are removed and compensated;

[0024] Based on the preprocessed regional ecological monitoring information, three types of data dimensions are defined, namely spatial dimension, time dimension, and variable dimension, and the preprocessed regional ecological monitoring information is spatially and temporally aligned through time dimension features and spatial dimension features;

[0025] According to the spatially and temporally aligned regional ecological monitoring information, multi-dimensional tensors are constructed according to the spatial dimension, time dimension, and variable dimension to obtain a number of original tensors, and the Tucker decomposition method is introduced to decompose each original tensor to obtain a spatial factor matrix, a time factor matrix, and a variable factor matrix;

[0026] The spatial factor matrix, the time factor matrix, and the variable factor matrix are projected onto the same feature space for multi-dimensional feature fusion to generate multi-modal ecological monitoring features of the target area.

[0027] In this solution, greenhouse gas release prediction and carbon flux inversion of the target area are carried out through multi-modal ecological monitoring features, specifically including:

[0028] Obtain the multi-modal ecological monitoring features of the target area. Using the sensor positions as nodes, a directed edge is constructed according to the water flow direction and the edge weight is set, and each node is connected through the constructed directed edge to construct a topological structure diagram;

[0029] Based on the multi-modal ecological monitoring features of the target area, node embedding vectors are generated. The attention mechanism is introduced to calculate the attention scores of each node embedding vector, and the attention scores are used to weight each node embedding vector, and the topological structure diagram is updated through the weighted node embedding vectors;

[0030] The adjacency matrix is obtained through the updated topological structure diagram and imported into the spatio-temporal graph convolutional network. Node association features and trend features are extracted through spatial convolution and time convolution, and greenhouse gas release prediction is performed based on the extracted features to obtain greenhouse gas release prediction information;

[0031] Construct a carbon flux inversion model based on a Bayesian inference network, input the multimodal ecological monitoring features of the target area to obtain the transient posterior distribution of the corresponding areas of each sensor node, and perform carbon flux inversion inference through the obtained transient posterior distribution to obtain the carbon flux inversion inference results of several sensor node areas;

[0032] Perform weighted fusion according to the carbon flux inversion inference results of several sensor node areas to generate the total carbon flux inversion inference result of the target area, and obtain the regional carbon flux inversion information.

[0033] In this solution, determining whether there is an ecological event in the target area and performing ecological adjustment early warning specifically includes:

[0034] Obtain greenhouse gas release prediction information and regional carbon flux inversion information, and construct an ecological hydrological monitoring feature portrait of the target area in combination with the multimodal ecological monitoring features of the target area;

[0035] Use data retrieval to obtain several historical ecological events, extract the event features of each historical ecological event, construct the feature portraits of each historical ecological event through the event features of each historical ecological event, and set up an ecological event database;

[0036] Import the ecological hydrological monitoring feature portrait of the target area into the ecological event database for matching analysis, and calculate the similarity values between the ecological hydrological monitoring feature portrait and the feature portraits of each historical ecological event in the ecological event database;

[0037] Judge whether there is an ecological event in the current monitoring area through the calculated similarity values. If the similarity value is greater than the preset similarity threshold, it means that there is an ecological event in the current monitoring area, and generate a warning message for prompt in combination with the corresponding similar ecological event.

[0038] The second aspect of the present invention provides an ecological hydrological monitoring system constructed based on a sensor network. The system includes: a memory and a processor. The memory contains an ecological hydrological monitoring method program constructed based on a sensor network. When the ecological hydrological monitoring method program constructed based on a sensor network is executed by the processor, the following steps are implemented:

[0039] Analyze the monitoring requirements and set the sensor layout plan, and construct a sensor network in the target monitoring area through the sensor layout plan;

[0040] Use the constructed sensor network to conduct ecological monitoring on the target area to obtain regional ecological monitoring information, and construct multimodal ecological monitoring features according to the regional ecological monitoring information;

[0041] Predict greenhouse gas emissions and invert carbon fluxes for a target area through multimodal ecological monitoring features, and determine whether there are ecological events in the target area to issue ecological adjustment warnings.

[0042] The present invention discloses an ecological hydrological monitoring method and system based on a sensor network, including: analyzing monitoring requirements and setting a sensor layout plan, constructing a sensor network in a target monitoring area through the sensor layout plan; using the constructed sensor network to conduct ecological monitoring on the target area to obtain regional ecological monitoring information, and constructing multimodal ecological monitoring features based on the regional ecological monitoring information; predicting greenhouse gas emissions and inverting carbon fluxes for the target area through the multimodal ecological monitoring features, and determining whether there are ecological events in the target area to issue ecological adjustment warnings. It breaks through the analysis bottleneck and response disadvantages of traditional monitoring methods, provides high-precision decision-making responses for ecosystem service assessment and climate change response, and at the same time provides support for ensuring regional ecological stability. Brief Description of the Drawings

[0043] In order to more clearly illustrate the technical solutions in the embodiments or exemplifications of the present invention, the following will briefly introduce the drawings required for use in the embodiments or exemplifications. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the drawings shown.

[0044] Figure 1 It is a flowchart of an ecological hydrological monitoring method based on a sensor network provided by an embodiment of the present invention;

[0045] Figure 2 It is a flowchart of a regional ecological analysis and warning method based on ecological hydrological monitoring provided by an embodiment of the present invention;

[0046] Figure 3 It is a block diagram of an ecological hydrological monitoring system based on a sensor network provided by an embodiment of the present invention;

[0047] The realization, functional features and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the drawings. Detailed Embodiments

[0048] In order to be able to more clearly understand the above objects, features and advantages of the present invention, the following will further describe the present invention in detail in conjunction with the drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0049] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited by the specific embodiments disclosed below.

[0050] Figure 1 It is a flowchart of an ecological hydrological monitoring method based on a sensor network provided by an embodiment of the present invention;

[0051] As Figure 1 shown, the present invention provides a flowchart of an ecological hydrological monitoring method based on a sensor network, including:

[0052] S102, analyze the monitoring requirements and set the sensor layout plan, and construct a sensor network in the target monitoring area through the sensor layout plan;

[0053] S104, use the constructed sensor network to conduct ecological monitoring on the target area to obtain regional ecological monitoring information, and construct multi-modal ecological monitoring features based on the regional ecological monitoring information;

[0054] S106, predict greenhouse gas emissions and invert carbon fluxes in the target area through multi-modal ecological monitoring features, and determine whether there are ecological events in the target area to issue ecological adjustment warnings.

[0055] Further, in a preferred embodiment of the present invention, the analysis of the monitoring requirements specifically includes:

[0056] Obtain the monitoring expectation information, decompose the monitoring target based on the monitoring expectation information to obtain the expected monitoring data for ecological hydrological monitoring of the target area, and obtain the expected monitoring information;

[0057] Generate a retrieval tag according to the expected monitoring information, perform monitoring sensor matching in a preset monitoring device database, analyze the monitoring sensors matching the current monitoring requirements, and obtain the matching sensor information;

[0058] Use data retrieval means to obtain the regional geographical information of the target area, discretize the target area into a number of grids of the same size based on the regional geographical information, and obtain the regional grid map;

[0059] Extract the obstacle features in the target area according to the regional geographical information, map the obstacles into a network mask through the extracted obstacle features, calibrate the non-obstacle areas, and update the regional grid map to obtain the updated regional grid map.

[0060] It should be noted that in the initial stage of the eco-hydrological monitoring task, it is necessary to first clarify the core requirements of the monitoring through demand analysis and target breakdown, and form an operable technical index framework. Based on the preset ecosystem management objectives, the hydrological cycle elements (such as surface runoff, soil moisture) and ecological parameters (such as vegetation transpiration, carbon flux) in the target area are systematically broken down to establish a multi-dimensional monitoring index system. This process usually combines an expert knowledge base and a historical data feature library to transform the macroscopic monitoring expectations into specific quantitative parameters. For example, "evaluating the wetland carbon sink capacity" is transformed into an executable index such as "daily methane flux monitoring", and finally generates the expected monitoring information metadata including attributes such as parameter type, spatio-temporal resolution, and accuracy level. Subsequently, the preset sensor device database is called to encode the expected monitoring information into structured retrieval tags (such as "measurement range 10-100% vol", "sampling frequency ≥ 1Hz"), and multi-dimensional matching is implemented in the distributed monitoring device database. Through semantic similarity calculation and parameter constraint satisfaction evaluation, a set of sensor models that meet both performance requirements and environmental adaptability is selected, and a matching sensor information list including device technical parameters, deployment costs, and energy consumption characteristics is generated, providing a basis for hardware selection for subsequent physical deployment. After determining the monitoring device list, high-precision topographic maps, land use classification maps, and 3D point cloud models of the target area are obtained through a geospatial data engine. The target area is cut into uniform grid cells using a spatial discretization algorithm, and the grid size is dynamically adjusted according to the spatial variation characteristics of the monitoring elements (for example, a 50m×50m grid is used for soil moisture monitoring, while a 10m×10m grid may be used for canopy temperature monitoring). Based on a deep learning model, the outlines of obstacles such as buildings, water areas, and dense forests are extracted from satellite images, and through rasterization processing, they are mapped into a binary network mask matrix, where the obstacle grids are marked as non-deployable areas, and the remaining grids form the potential layout positions of sensor nodes. The finally generated updated regional grid map not only retains the original geographical coordinate framework but also embeds obstacle topological constraints, forming a digital monitoring base that takes into account geographical authenticity and deployment feasibility, laying a spatial decision-making foundation for the subsequent optimization layout of the sensor network.

[0061] Furthermore, in a preferred embodiment of the present invention, the setting of the sensor layout plan is used to construct a sensor network in the target monitoring area through the sensor layout plan, specifically including:

[0062] Obtain the matching sensor information, and extract the functional characteristics, energy consumption characteristics, and communication ability characteristics of the monitoring sensors that match the current monitoring requirements from the matching sensor information to obtain the matching sensor characteristic information;

[0063] Obtain the updated regional grid map, construct a sensor network in combination with the matched sensor information, perform layout analysis of the regional sensor network using the greedy algorithm, and set the objective function according to coverage maximization and energy consumption minimization in a weighted fusion manner;

[0064] Set constraints based on communication distance, number of nodes, and obstacle distribution, add penalty terms to the objective function through the set constraints, define each non-obstacle grid of the updated regional grid map as a feasible node, and perform population initialization in combination with the matched sensor feature information;

[0065] Calculate the fitness value of each individual in the initial population, select the optimal candidate individual through the calculated fitness value. If there are multiple individuals with the same fitness value, select the individual with the minimum distance from the center of the uncovered area as the optimal candidate individual;

[0066] Output the final solution set through iterative analysis until the termination condition is met, generate a sensor network layout plan for the target area according to the final solution set, and construct a sensor network in the target area.

[0067] It should be noted that after determining the set of sensor devices that match the current monitoring requirements, it is necessary to deeply analyze their functional attributes and performance parameters to construct a sensor feature knowledge base. By analyzing the functional characteristics (such as measurement accuracy, range) of the matching sensors, energy consumption characteristics (such as standby power consumption, continuous working duration), and communication capabilities (such as maximum transmission distance, protocol compatibility), a multi-dimensional feature vector is formed to quantitatively represent the applicability differences of different sensors in specific monitoring scenarios. For example, a high-precision carbon dioxide sensor may have low power consumption characteristics but limited communication distance. On this basis, the updated regional grid map is coupled with the sensor feature information to initiate the sensor network topology optimization process. The greedy algorithm is used as the core engine for layout optimization, with the dual objectives of maximizing coverage and minimizing energy consumption, to construct a weighted objective function: the coverage weight reflects the spatial heterogeneity of the monitoring requirements (such as the weight of the core protected area is higher than that of the edge area), and the energy consumption weight reflects the balance between the node deployment density and the battery life. To meet the actual deployment constraints, a communication radius threshold, an upper limit on the number of nodes, and an obstacle exclusion rule are introduced, and the hard constraints are transformed into penalty terms of the objective function through the Lagrange multiplier method to ensure the physical feasibility of the candidate solution set. Specifically, each non-obstacle grid is regarded as a potential deployment node, and an adjacency matrix is generated based on the sensor communication capabilities. When initializing the population, seed nodes are randomly selected and gradually expanded to ensure network connectivity. During the iteration process, the fitness value of each layout scheme is calculated, comprehensively considering the coverage area, energy consumption cost, and degree of constraint violation, and the Pareto front solution set is preferentially retained; when the fitness values of multiple individuals are the same, a spatial compactness index is introduced to select the scheme with the smallest distance from the centroid of the uncovered area to improve the deployment balance. The optimization is terminated by setting a convergence threshold or the maximum number of iterations, and the optimal sensor layout scheme that meets the multi-objective trade-off is output to guide the precise positioning of nodes and the configuration of network parameters during field deployment, and finally an ecological hydrological monitoring sensor network is constructed.

[0068] Furthermore, in a preferred embodiment of the present invention, the ecological monitoring of the target area is carried out by using the constructed sensor network to obtain regional ecological monitoring information, and multi-modal ecological monitoring features are constructed according to the regional ecological monitoring information, specifically including:

[0069] The ecological monitoring of the target area is carried out through the sensor network constructed in the target area to obtain regional ecological monitoring information, and the regional ecological monitoring information includes hydrological monitoring data, greenhouse gas monitoring data, and meteorological monitoring data;

[0070] The regional ecological monitoring information is preprocessed, and the data standard deviation of the regional ecological monitoring information is calculated through the data standard deviation. The data outside the preset standard deviation range is defined as abnormal data, and the outliers are removed and compensated;

[0071] Define three data dimensions based on the preprocessed regional ecological monitoring information, namely the spatial dimension, the temporal dimension, and the variable dimension, and perform spatio-temporal alignment on the preprocessed regional ecological monitoring information through the temporal dimension features and the spatial dimension features;

[0072] Construct multi-dimensional tensors according to the spatio-temporally aligned regional ecological monitoring information in the spatial dimension, the temporal dimension, and the variable dimension to obtain a number of original tensors, and introduce the Tucker decomposition method to decompose each original tensor to obtain the spatial factor matrix, the temporal factor matrix, and the variable factor matrix;

[0073] Project the spatial factor matrix, the temporal factor matrix, and the variable factor matrix onto the same feature space for multi-dimensional feature fusion to generate the multi-modal ecological monitoring features of the target area.

[0074] It should be noted that multi-source ecological monitoring data of the target area are collected in real time through the deployed sensor network, covering heterogeneous information such as hydrological parameters (such as soil moisture, surface runoff), greenhouse gas concentrations (CO2, CH4, N2O), and meteorological elements (air temperature, precipitation, wind speed), etc., to form an original monitoring data set. Data cleaning and quality enhancement are implemented for this data set: The standard deviation threshold method based on statistics is used to identify abnormal data points. For example, when the CO2 concentration continuously recorded by a certain sensor suddenly becomes ten times the background value and exceeds the three-standard-deviation range, it is determined as an outlier and removed; for missing data, the spatio-temporal proximity interpolation method or the sequence prediction model based on LSTM is used for compensation and reconstruction to ensure the spatio-temporal continuity of the data. After the preprocessing is completed, the data is mapped to a three-dimensional analysis framework: The spatial dimension represents the geographical coordinates of the monitoring grid, the time dimension corresponds to the time stamp sequence of data collection, and the variable dimension distinguishes different ecological parameter types. The sampling frequency and spatial coverage differences between sensors are eliminated through spatio-temporal alignment algorithms (such as dynamic time warping DTW or spatial resampling) to achieve seamless integration of multi-source data under a unified spatio-temporal benchmark. To further explore the internal correlations of ecological monitoring data, the spatio-temporally aligned data set is constructed into a multi-dimensional tensor structure, where each tensor element represents the observed value under a specific grid point, time point, and variable combination. The Tucker decomposition method is used to perform high-order principal component analysis on this tensor: The spatial factor matrix, time factor matrix, and variable factor matrix are iteratively decomposed through the alternating least squares (ALS) optimization algorithm. Among them, the spatial factor reveals the geographical distribution pattern, the time factor depicts the periodic fluctuations of ecological parameters, and the variable factor reflects the coupling relationship between multiple parameters, such as the positive correlation between temperature and CH4 emissions, etc. To fuse multi-dimensional features, the three types of factor matrices are projected into a shared latent semantic space, and a fused multi-modal ecological monitoring feature vector is generated through matrix splicing and non-linear transformation (such as attention weighting or kernel function mapping). This feature not only retains the spatio-temporal evolution law of the original data but also extracts the cross-modal interaction effects, such as the synergistic effects of extreme rainfall events on soil respiration and runoff carbon transport.

[0075] Furthermore, in a preferred embodiment of the present invention, the greenhouse gas release prediction and carbon flux inversion of the target area through the multi-modal ecological monitoring features specifically include:

[0076] Obtain the multi-modal ecological monitoring features of the target area. Taking the sensor positions as nodes, construct directed edges according to the water flow direction and set edge weights, and connect the nodes through the constructed directed edges to construct a topological structure diagram;

[0077] Generate node embedding vectors based on the multi-modal ecological monitoring features of the target area, introduce an attention mechanism to calculate the attention scores of each node embedding vector, weight each node embedding vector using the attention scores, and update the topological structure diagram with the weighted node embedding vectors;

[0078] Obtain the adjacency matrix from the updated topological structure diagram and import it into the spatio-temporal graph convolutional network. Extract node correlation features and trend features through spatial convolution and temporal convolution, and perform greenhouse gas emission prediction based on the extracted features to obtain greenhouse gas emission prediction information;

[0079] Construct a carbon flux inversion model based on the Bayesian inference network. Input the multi-modal ecological monitoring features of the target area to obtain the transient posterior distribution of the corresponding areas of each sensor node. Perform carbon flux inversion inference through the obtained transient posterior distribution to obtain the carbon flux inversion inference results of several sensor node areas;

[0080] Perform weighted fusion according to the carbon flux inversion inference results of several sensor node areas to generate the total carbon flux inversion inference result of the target area and obtain the regional carbon flux inversion information.

[0081] It is important to note that a dynamic topological structure graph is constructed based on the spatial distribution of the sensor network to represent the material and energy transfer pathways of ecological processes. By analyzing the target area's topographic elevation data and historical hydrological observations, the direction of surface runoff and groundwater recharge pathways are determined. These are abstracted into directed edge connections between nodes, and edge weights are assigned based on water flow velocity or pollutant diffusion rate (e.g., faster flow rates are associated with higher weights). This creates a topological network that reflects the coupled water-carbon transport mechanism. Furthermore, multimodal ecological monitoring features (such as temperature, humidity, NDVI, and CO2 concentration) are encoded into high-dimensional node embedding vectors. A multi-head self-attention mechanism is introduced to calculate the strength of inter-node connections. Through a query-key matching mechanism, attention scores are calculated between node embedding vectors, capturing long-range dependencies across regions. Based on this, node features are adaptively weighted and fused to update the edge weights and node states of the topological structure graph, enhancing the network's ability to represent complex ecological interactions. Subsequently, the updated topological graph adjacency matrix and node features are input into a spatiotemporal graph convolutional network (ST-GCN). Spatial convolutional layers aggregate the multimodal features of adjacent nodes. For example, Chebyshev polynomials are used to approximate the graph convolution kernel to extract spatial correlation patterns. Coupled with a temporal convolutional layer (TCN), these layers mine temporal trend features, such as the periodic changes in respiration driven by diurnal temperature differences. By stacking multiple layers of spatiotemporal convolutional modules, high-order spatiotemporal correlation features are gradually extracted, ultimately outputting the predicted greenhouse gas concentration and its spatial gradient distribution for each node within the future time window. Simultaneously, a Bayesian inference network is constructed to perform probabilistic inversion of carbon flux. Using multimodal features as input, the posterior distribution of the latent variable is learned through variational inference. Ecohydrological mechanism models (such as light response curves) are combined as prior constraints to generate a carbon flux probability density function for each sensor node's grid. This quantifies the uncertainty of the inversion results and yields a local carbon flux prediction. Finally, the local carbon flux prediction values are weighted and integrated to generate the overall net carbon flux probability distribution of the target area, and the regional carbon flux inversion information is output to provide a decision-making basis for ecological management.

[0082] Furthermore, in a preferred embodiment of the present invention, the step of determining whether an ecological event occurs in the target area and performing ecological adjustment warning specifically includes:

[0083] Obtain greenhouse gas release prediction information and regional carbon flux inversion information, and build an eco-hydrological monitoring characteristic profile of the target area by combining the multimodal ecological monitoring characteristics of the target area;

[0084] Using data retrieval to obtain a number of historical ecological events, extracting the event characteristics of each historical ecological event, constructing a feature portrait of each historical ecological event through the event characteristics of each historical ecological event, and setting up an ecological event database;

[0085] Import the ecological hydrological monitoring feature portrait of the target area into the ecological event database for matching analysis, and calculate the similarity values between the ecological hydrological monitoring feature portrait and the historical ecological event feature portraits in the ecological event database;

[0086] Judge whether there is an ecological event in the current monitoring area based on the calculated similarity value. If the similarity value is greater than the preset similarity threshold, it means that there is an ecological event in the current monitoring area, and a warning message will be generated and prompted in combination with the corresponding similar ecological event.

[0087] It should be noted that based on the greenhouse gas release prediction information and the regional carbon flux inversion information, multi-modal ecological monitoring features are fused to construct an ecological hydrological feature portrait to achieve a comprehensive assessment of the ecological state. By performing multi-dimensional feature encoding on the hydrological dynamics (such as runoff, soil moisture), spatial distribution of carbon flux, meteorological factors (temperature, precipitation) and multi-modal features of the sensor network (such as water level, flow velocity) within the region, a feature portrait containing time series evolution, spatial heterogeneity and variable interaction effects is generated. At the same time, by retrieving the historical ecological event library through the data engine, analyzing the spatio-temporal evolution trajectory and key driving parameters of events such as floods, droughts, and wetland degradation, abstracting the event features into structured labels and embedding them into the feature portrait, an ecological event database covering event precursor signals and process characteristics is constructed to provide ecological change warning assistance in real-time ecological hydrological monitoring. Specifically, a similarity measurement algorithm is used to match the real-time feature portrait of the target area with the historical event library. The matching degree of the multi-variable combination mode is quantified by cosine similarity. For example, when the feature portrait of a certain area shows a combination mode of decreasing soil moisture, decreasing vegetation index and abnormal fluctuation of CO2 flux within two consecutive weeks, and the similarity with the feature portrait of "wetland degradation" in the historical library reaches 85% (exceeding the preset threshold of 75%), a wetland degradation event warning will be triggered and pushed to the management department in real time to provide assistance for ecological area control and adaptive management.

[0088] Figure 2 Flowchart of a regional ecological analysis and warning method based on ecological hydrological monitoring provided by an embodiment of the present invention;

[0089] As Figure 2 shown, the present invention provides a flowchart of a regional ecological analysis and warning method based on ecological hydrological monitoring, including:

[0090] S202, construct a sensor network in the target monitoring area, and use the constructed sensor network to conduct ecological monitoring on the target area to obtain regional ecological monitoring information;

[0091] S204. Construct multimodal ecological monitoring features based on regional ecological monitoring information, conduct greenhouse gas emission prediction and carbon flux inversion for the target area, and obtain greenhouse gas emission prediction information and regional carbon flux inversion information.

[0092] S206. Perform multi-dimensional feature encoding based on the greenhouse gas emission prediction information, regional carbon flux inversion information, and multimodal ecological monitoring features of the target area to generate an ecological and hydrological monitoring feature portrait of the target area.

[0093] S208. Import the ecological and hydrological monitoring feature portrait of the target area into the ecological event database for matching analysis, and calculate the similarity values between the ecological and hydrological monitoring feature portrait and the feature portraits of each historical ecological event in the ecological event database.

[0094] S210. Determine whether there is an ecological event in the current monitoring area based on the calculated similarity values. If the similarity value is greater than the preset similarity threshold, it means that there is an ecological event in the current monitoring area, and a warning message will be generated in combination with the corresponding similar ecological event for prompt.

[0095] It should be noted that first, a sensor network is deployed in the target monitoring area, and ecological information in the area is collected in real time through these sensors, so as to obtain regional ecological monitoring information including hydrological parameters, greenhouse gas concentrations, and other ecological indicators. Subsequently, these monitoring information are used to construct multimodal ecological monitoring features, forming a feature set that can comprehensively describe the ecological state of the target area. Then, based on these multimodal ecological monitoring features, the system further conducts greenhouse gas emission prediction and carbon flux inversion. Greenhouse gas emission prediction estimates the greenhouse gas emissions in the area by analyzing the collected meteorological data such as gas concentration, temperature, and wind speed; while carbon flux inversion reveals the carbon cycle status in the area by calculating the carbon exchange amount between water bodies and the atmosphere. After obtaining the above prediction information, the greenhouse gas emission prediction information, regional carbon flux inversion information, and multimodal ecological monitoring features are subjected to multi-dimensional feature encoding to generate an ecological and hydrological monitoring feature portrait of the target area, reflecting the state and change trend of the regional ecology and hydrology. Next, it is imported into the ecological event database, and the similarity between the current area and historical events is calculated through matching analysis with the feature portraits of various historical ecological events stored in the database. Finally, by comparing the calculated similarity value with the preset threshold, the system determines whether there is an ecological event in the target monitoring area. If the similarity exceeds the preset threshold, it means that the current area may be experiencing a situation similar to historical ecological events, and a warning message will be generated for prompt, so as to provide a basis for timely intervention and decision-making, and improve the comprehensiveness and efficiency of ecological and hydrological monitoring work.

[0096] Figure 3An ecological hydrological monitoring system 3 constructed based on a sensor network provided by an embodiment of the present invention, the system includes: a memory 31, a processor 32, the memory 31 contains an ecological hydrological monitoring method program constructed based on a sensor network, and when the ecological hydrological monitoring method program constructed based on a sensor network is executed by the processor 32, the following steps are implemented:

[0097] Analyze the monitoring requirements and set the sensor layout plan, and construct a sensor network in the target monitoring area through the sensor layout plan;

[0098] Use the constructed sensor network to conduct ecological monitoring on the target area to obtain regional ecological monitoring information, and construct multi-modal ecological monitoring features based on the regional ecological monitoring information;

[0099] Predict the greenhouse gas emissions and invert the carbon flux of the target area through the multi-modal ecological monitoring features, and determine whether there are ecological events in the target area to issue an ecological adjustment warning.

[0100] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed with each other can be through some interfaces. The indirect coupling or communication connection of the devices or units can be electrical, mechanical, or other forms.

[0101] The units described above as separate components may or may not be physically separated. The components shown as units may or may not be physical units; they can be located in one place or distributed to multiple network units; some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0102] In addition, each functional unit in the embodiments of the present invention can be all integrated in a processing unit, or each unit can be separately used as a unit, or two or more units can be integrated in a unit; the above integrated units can be implemented in the form of hardware, or in the form of a combination of hardware and software functional units.

[0103] Those of ordinary skill in the art can understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including those of the above method embodiments; and the aforementioned storage medium includes: removable storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs and other various media that can store program codes.

[0104] Alternatively, if the above integrated units are implemented in the form of software function modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. And the aforementioned storage medium includes: removable storage devices, ROM, RAM, magnetic disks, or optical discs and other various media that can store program codes.

[0105] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An ecological hydrological monitoring method based on a sensor network, characterized in that, Including: Analyze the monitoring requirements and set the sensor layout plan, and construct a sensor network in the target monitoring area through the sensor layout plan; Use the constructed sensor network to conduct ecological monitoring on the target area to obtain regional ecological monitoring information, and construct multi-modal ecological monitoring features based on the regional ecological monitoring information; Through the multi-modal ecological monitoring features, predict the greenhouse gas emissions and invert the carbon flux in the target area, and judge whether there are ecological events in the target area to issue ecological adjustment warnings.

2. The ecological hydrological monitoring method based on a sensor network according to claim 1, characterized in that The analysis of the monitoring requirements specifically includes: Obtain the monitoring expectation information, decompose the monitoring target based on the monitoring expectation information to obtain the expected monitoring data for ecological and hydrological monitoring of the target area, and obtain the expected monitoring information; Generate retrieval tags according to the expected monitoring information, perform monitoring sensor matching in the preset monitoring device database, analyze the monitoring sensors that match the current monitoring requirements, and obtain the matching sensor information; Use data retrieval means to obtain the regional geographical information of the target area, discretize the target area into several grids of the same size based on the regional geographical information, and obtain the regional grid map; Extract the obstacle features in the target area according to the regional geographical information, map the obstacles into a network mask through the extracted obstacle features, calibrate the non-obstacle area, and update the regional grid map to obtain the updated regional grid map.

3. The ecological hydrological monitoring method based on a sensor network according to claim 1, characterized in that The setting of the sensor layout plan, and the construction of the sensor network in the target monitoring area through the sensor layout plan specifically includes: Obtain the matching sensor information, extract the functional features, energy consumption features and communication ability features of the monitoring sensors that match the current monitoring requirements through the matching sensor information, and obtain the matching sensor feature information; Obtain the updated regional grid map, combine the matching sensor information to construct the sensor network, use the greedy algorithm to analyze the regional sensor network layout, and set the objective function according to the maximum coverage and minimum energy consumption in a weighted fusion manner; Set the constraint conditions based on the communication distance, the number of nodes and the obstacle distribution, add penalty terms to the objective function through the set constraint conditions, define each non-obstacle grid of the updated regional grid map as a feasible node, and initialize the population in combination with the matching sensor feature information; Calculate the fitness value of each individual in the initial population, select the optimal candidate individual through the calculated fitness value, if there are multiple individuals with the same fitness value, then select the individual with the smallest distance from the center of the uncovered area as the optimal candidate individual; Through iterative analysis until the termination condition is met, output the final solution set, generate the sensor network layout plan for the target area according to the final solution set, and construct the sensor network in the target area.

4. The ecological hydrological monitoring method based on a sensor network according to claim 1, characterized in that, The use of the constructed sensor network to conduct ecological monitoring on the target area to obtain regional ecological monitoring information, and construct multi-modal ecological monitoring features based on the regional ecological monitoring information specifically includes: Conduct ecological monitoring on the target area through the sensor network constructed in the target area to obtain regional ecological monitoring information, and the regional ecological monitoring information includes hydrological monitoring data, greenhouse gas monitoring data and meteorological monitoring data; Perform data preprocessing on the regional ecological monitoring information, calculate the data standard deviation of the regional ecological monitoring information through data standard deviation calculation, define the data outside the preset standard deviation range as abnormal data, and perform outlier removal and compensation; Define three types of data dimensions based on the preprocessed regional ecological monitoring information, namely spatial dimension, time dimension, and variable dimension, and perform spatio-temporal alignment on the preprocessed regional ecological monitoring information through time dimension features and spatial dimension features; Construct multi-dimensional tensors according to the spatial dimension, time dimension, and variable dimension of the spatio-temporally aligned regional ecological monitoring information to obtain a number of original tensors, and introduce the Tucker decomposition method to decompose each original tensor to obtain a spatial factor matrix, a time factor matrix, and a variable factor matrix; Project the spatial factor matrix, time factor matrix, and variable factor matrix onto the same feature space for multi-dimensional feature fusion to generate multi-modal ecological monitoring features of the target area.

5. A method for ecological hydrological monitoring based on a sensor network according to claim 1, characterized in that, The greenhouse gas release prediction and carbon flux inversion of the target area are performed through the multi-modal ecological monitoring features, specifically including: Obtain the multi-modal ecological monitoring features of the target area, use the sensor position as a node, construct a directed edge according to the water flow direction and set the edge weight, and connect each node through the constructed directed edge to construct a topological structure diagram; Generate node embedding vectors based on the multi-modal ecological monitoring features of the target area, introduce an attention mechanism to calculate the attention scores of each node embedding vector, weight each node embedding vector using the attention scores, and update the topological structure diagram through the weighted node embedding vectors; Obtain the adjacency matrix through the updated topological structure diagram, import it into the spatio-temporal graph convolutional network, extract node correlation features and trend features through spatial convolution and temporal convolution, and perform greenhouse gas release prediction based on the extracted features to obtain greenhouse gas release prediction information; Construct a carbon flux inversion model based on the Bayesian inference network, input the multi-modal ecological monitoring features of the target area to obtain the transient posterior distribution of the corresponding areas of each sensor node, and perform carbon flux inversion inference through the obtained transient posterior distribution to obtain the carbon flux inversion inference results of several sensor node areas; Perform weighted fusion according to the carbon flux inversion inference results of several sensor node areas to generate the total carbon flux inversion inference result of the target area and obtain the regional carbon flux inversion information.

6. The ecological hydrological monitoring method based on a sensor network according to claim 1, characterized in that, Judge whether there is an ecological event in the target area and perform ecological adjustment early warning, specifically including: Obtain the greenhouse gas release prediction information and regional carbon flux inversion information, and construct an ecological hydrological monitoring feature portrait of the target area in combination with the multi-modal ecological monitoring features of the target area; Use data retrieval to obtain a number of historical ecological events, extract the event features of each historical ecological event, construct the feature portraits of each historical ecological event through the event features of each historical ecological event, and set up an ecological event database; Import the ecological hydrological monitoring feature portrait of the target area into the ecological event database for matching analysis, and calculate the similarity values between the ecological hydrological monitoring feature portrait and the feature portraits of each historical ecological event in the ecological event database; Determine whether there is an ecological event in the current monitoring area based on the calculated similarity value. If the similarity value is greater than the preset similarity threshold, it indicates that there is an ecological event in the current monitoring area, and a warning message is generated in combination with the corresponding similar ecological event for prompt.

7. An ecological hydrological monitoring system constructed based on a sensor network, characterized in that, The system includes: a memory and a processor. The memory contains an ecological hydrological monitoring method program constructed based on a sensor network. When the ecological hydrological monitoring method program constructed based on the sensor network is executed by the processor, the following steps are implemented: Analyze the monitoring requirements and set the sensor layout scheme, and construct a sensor network in the target monitoring area through the sensor layout scheme; Use the constructed sensor network to conduct ecological monitoring on the target area to obtain regional ecological monitoring information, and construct multi-modal ecological monitoring features based on the regional ecological monitoring information; Predict greenhouse gas emissions and invert carbon fluxes in the target area through multi-modal ecological monitoring features, and determine whether there is an ecological event in the target area to issue an ecological adjustment warning.

Images