Method and system for detecting heavy metal pollution of agricultural land soil

Abstract

The invention discloses a detection method and system for agricultural land soil heavy metal pollution, and belongs to the technical field of soil heavy metal detection.The detection method comprises the steps that a dynamic grid division method is adopted for delimiting a detection area, and the detection area is divided into a plurality of space units; carrying out adaptive filtering and noise reduction processing on the in-situ rapid spectral signal to obtain denoised spectral data, and calculating the initial content of heavy metal elements in each space unit; constructing a multi-feature data set, training the multi-feature data set, and establishing a heavy metal pollution degree comprehensive evaluation model; and inputting the spatial unit feature data of the farmland to be detected into the comprehensive evaluation model, and outputting the comprehensive pollution index and risk level of each unit. According to the method and system for detecting heavy metal pollution of agricultural land soil provided by the invention, multi-source information can be fused, rapid and accurate detection can be realized, and the method has a dynamic risk assessment capability for detecting heavy metal pollution of agricultural land soil.

Description

Technical Field

[0001] This invention relates to the field of soil heavy metal detection technology, and in particular to a method and system for detecting heavy metal pollution in agricultural land soil. Background Technology

[0002] Soil is the foundation of agricultural production, and its quality and safety are directly related to the safety of agricultural products and human health. With the progress of industrialization and urbanization, the problem of heavy metal pollution in agricultural land soil is becoming increasingly serious. Heavy metals such as lead (Pb), cadmium (Cd), chromium (Cr), arsenic (As), and mercury (Hg) are highly toxic, difficult to degrade, and easily accumulate in soil, and can harm human health through the food chain.

[0003] Currently, methods for detecting heavy metals in soil mainly include laboratory chemical analysis methods (such as atomic absorption spectrometry and inductively coupled plasma mass spectrometry) and in-situ rapid detection methods (such as portable X-ray fluorescence spectrometry). Laboratory methods offer high precision but are cumbersome, time-consuming, and costly, making them unsuitable for large-scale, real-time monitoring. In-situ rapid detection methods are efficient, but their results are easily affected by factors such as soil matrix, moisture, and instrument noise, and their accuracy is insufficient, especially at low concentrations.

[0004] Furthermore, existing technologies are mostly limited to single-dimensional detection, focusing only on soil chemical content, microbial indicators, or static spatial interpolation and diffusion simulation. These methods fail to effectively integrate multi-source information from physical, chemical, and biological sources, and do not fully consider the dynamic and spatial heterogeneity of farmland environments, resulting in incomplete pollution assessments and a lack of forward-looking risk warnings. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting heavy metal pollution in agricultural land soil, so as to solve the problems existing in the background art.

[0006] To achieve the above objectives, the present invention provides a method for detecting heavy metal pollution in agricultural land soil, comprising the following steps: S1. Spatial grid construction and multi-source information acquisition: Based on the shape, area and surrounding pollution source distribution of the target farmland, the detection area is delineated using a dynamic grid division method, and the detection area is divided into multiple spatial units; soil physicochemical parameters and in-situ rapid spectral signals are collected in each spatial unit, and rhizosphere soil and microbial samples are collected simultaneously. S2. Preliminary pollution identification and microbial feature extraction: Adaptive filtering and noise reduction processing is performed on the in-situ rapid spectral signal to obtain denoised spectral data. Based on the denoised spectral data, the preliminary content of heavy metal elements in each spatial unit is calculated using a quantitative analysis model. Microbial DNA is extracted and high-throughput sequencing is performed on rhizosphere soil and microbial samples to obtain information on microbial community composition. A microbial symbiotic network is constructed to identify key microbial biomarkers and their abundance related to heavy metal pollution response. S3. Multi-source information fusion modeling: The preliminary content of heavy metal elements, abundance of key microbial biomarkers, soil physicochemical parameters and historical pollution source influencing factors of each spatial unit are fused to construct a multi-feature dataset; machine learning algorithms are used to train the multi-feature dataset to establish a comprehensive assessment model of heavy metal pollution level. S4. Pollution assessment and risk warning: Input the spatial unit characteristic data of the farmland to be tested into the comprehensive assessment model, and output the comprehensive pollution index and risk level of each unit; Based on the output results of all units, combine the geographic information system to generate a spatial distribution map of heavy metal pollution, a risk level map and a potential diffusion trend prediction map, and issue warnings for high-risk areas.

[0007] Preferably, the dynamic mesh generation method in S1 includes: Determine the center point for dividing the farmland based on the longitudinal and transverse length ratios; use the center point as a reference to initially divide the land into a nine-square grid. Combining historical or real-time data on the distribution of surrounding pollution sources, including main traffic arteries, industrial emission sources, and sewage irrigation areas, the nine-square grid is refined, and the grid is densified in the boundary areas where the pollution sources have a strong impact, to generate the final spatial unit division.

[0008] Preferably, the in-situ rapid spectral signal acquisition in S1 is performed using a portable X-ray fluorescence spectrometer; the adaptive filtering and noise reduction processing in S2 includes: Obtain a set of spectral signals measured multiple times within the same spatial cell; For each spectral signal, spectral values ​​are clustered, and the clustering results are arranged by wavelength to form multiple local windows; Analyze the anomaly degree of the difference between the spectral value and its neighboring values ​​within each window, and calculate the first degree of disturbance. Identify feature windows with a length greater than a threshold in the spectral signal, and calculate the second degree of perturbation by analyzing the similarity of the fitted curves between feature windows and the goodness of fit within them. Analyze the similarity between different spectral signals and select a set of similar spectral signals; Based on the first degree of disturbance, the second degree of disturbance, and the degree of dispersion of similar spectral signals at the same wavelength, the comprehensive degree of disturbance of each spectral value is calculated, and the length of the filtering window is dynamically adjusted accordingly for noise reduction.

[0009] Preferably, the construction of the microbial symbiotic network and the identification of key microbial biomarkers in S2 include: Based on species composition and abundance data obtained from high-throughput sequencing, a microbial symbiotic network is constructed with microbial species as nodes and inter-species correlations as edges. We perform network topology feature analysis on symbiotic networks and use community detection algorithms to divide functional modules. The centrality index of nodes in the network is calculated to identify microbial species that occupy key positions in the network and are significantly associated with known heavy metal pollution data, serving as key microbial biomarkers.

[0010] Preferably, the historical pollution source impact factors in S3 include: traffic impact factor, industrial coal combustion / incineration impact factor, and sewage irrigation impact factor; wherein, the traffic impact factor is calculated based on the traffic flow and distance of surrounding roads per unit time; the industrial coal combustion / incineration impact factor is calculated based on the number and distance of pollution sources within a preset radius; and the sewage irrigation impact factor is calculated based on the number of polluted water source input points and the historical irrigation range.

[0011] Preferably, the machine learning algorithm in S3 is one or more combinations of random forest, support vector machine or neural network; the output of the comprehensive evaluation model is a continuous comprehensive pollution index, which is divided into four risk levels: clean, lightly polluted, moderately polluted and heavily polluted according to a preset threshold.

[0012] Preferably, the generation of the potential diffusion trend prediction map in S4 includes: Based on soil pollution characteristics, regional hydrogeological parameters, and meteorological data, initial boundary conditions for pollution diffusion are set. Using identified high-pollution spatial units as potential pollution sources, and combining environmental driving factors, numerical simulation methods are used to predict the diffusion paths and ranges of heavy metal pollutants in time and space.

[0013] Preferred options also include: S5. Establish a pollution time-series database and a dynamic update mechanism; The comprehensive pollution index, risk level, spatial distribution data, and corresponding environmental parameters of each test are stored in a time-series database. Regularly retest the same farmland and use data from the time series database to train a time series prediction model to predict pollution trends in a specific future time period, thereby achieving dynamic early warning of risks.

[0014] A detection system for heavy metal pollution in agricultural land soil, comprising: The gridded acquisition module is used to execute step S1, completing spatial grid construction and multi-source information acquisition; The rapid detection and bioanalysis module is used to execute step S2, which completes the rapid detection of spectral signals, noise reduction processing, and microbial feature extraction. The intelligent analysis and modeling module is used to execute step S3 and complete the construction of a multi-source information fusion and comprehensive evaluation model. The assessment, early warning, and visualization module is used to execute step S4, completing pollution assessment, risk classification, spatial mapping, and early warning information dissemination.

[0015] Therefore, the present invention employs the above-mentioned method and system for detecting heavy metal pollution in agricultural land soil, which has the following beneficial effects: (1) By deeply integrating physical and chemical rapid detection, microbial ecological indicators and environmental pollution source information, the limitations of single technical means are overcome, and the soil pollution status and ecological health risks are comprehensively evaluated from multiple dimensions; (2) The dynamic grid division method is adopted to intelligently adjust the sampling density according to the terrain and pollution sources. While ensuring the overall detection efficiency, it focuses on capturing pollution hotspots, which significantly improves the accuracy of spatial assessment. (3) The introduction of advanced adaptive spectral noise reduction technology has improved the quality of rapid on-site detection data; the use of microbial symbiotic networks to identify key biomarkers has provided stable and sensitive biological evidence, enhancing the reliability of detection results; (4) Based on machine learning, a comprehensive evaluation model is constructed to achieve quantitative classification of pollution levels; combined with geographic information system (GIS) and diffusion model, spatial visualization of pollution and prediction of future trends are realized, providing strong intelligent decision support for precise management, safe utilization and early warning of risks of farmland soil.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the structure of a method and system for detecting heavy metal pollution in agricultural land soil according to the present invention; Figure 2 This is a schematic diagram of dynamic grid division and multi-source information acquisition in an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of multi-source information fusion modeling and risk assessment and early warning in an embodiment of the present invention. Detailed Implementation

[0019] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] Please see Figures 1-3 A method for detecting heavy metal pollution in agricultural land soil, comprising: S1. Spatial grid construction and multi-source information acquisition: Based on the shape, area and surrounding pollution source distribution of the target farmland, the detection area is delineated using a dynamic grid division method, and the detection area is divided into multiple spatial units; soil physicochemical parameters and in-situ rapid spectral signals are collected in each spatial unit, and rhizosphere soil and microbial samples are collected simultaneously.

[0021] For a rectangular farmland located in the suburbs, suspected of being affected by surrounding roads and historical upstream factories, the following steps were taken: First, the lengths of its longitudinal (long side) and transverse (short side) sides were measured, and the intersection of the two diagonals was determined as the center point. Using this center point, an initial nine-square grid was created, resulting in nine large grids.

[0022] Subsequently, the surrounding environment was investigated: a national highway (a major source of traffic pollution) lies to the north, and a closed chemical plant (a historical source of industrial pollution) is located approximately 2 kilometers to the east. Based on this pollution source information, the initial grid was dynamically refined. Specifically, the three large grids on the north side, closest to the national highway, were further subdivided into smaller units along the pollution gradient (from north to south); the two large grids downwind (southeast) near the historical plant were also refined. Ultimately, the entire farmland was divided into 49 spatial units (e.g., ...). Figure 2 (Illustrative image). In each spatial unit, at five points (center and four corners), in-situ soil spectral signals were collected using a portable XRF spectrometer (such as the Innov-X series). Measurements were repeated three times at each point, and the spectral signals were aggregated. Simultaneously, mixed soil samples were collected within the unit to determine physicochemical parameters such as pH, organic matter, and moisture content. Approximately 100g of rhizosphere soil near crop roots was also collected, placed in a sterile bag, and stored at low temperature for microbial analysis.

[0023] S2. Preliminary pollution identification and microbial feature extraction: Adaptive filtering and noise reduction processing is performed on the in-situ rapid spectral signal to obtain denoised spectral data. Based on the denoised spectral data, the preliminary content of heavy metal elements in each spatial unit is calculated using a quantitative analysis model. Microbial DNA is extracted and high-throughput sequencing is performed on rhizosphere soil and microbial samples to obtain information on microbial community composition. A microbial symbiotic network is constructed to identify key microbial biomarkers and their abundance related to heavy metal pollution response.

[0024] Adaptive denoising is performed on multiple XRF spectral signals acquired for each spatial cell. The process is briefly described as follows: DBSCAN clustering is used to form spectral value windows for each signal; the anomaly score of the spectral value differences within each window (e.g., using the LOF algorithm) is calculated to obtain the first level of perturbation; feature windows exceeding the maximum inter-class variance threshold are identified, and the similarity between windows and the goodness of fit within windows are calculated using curve fitting to obtain the second level of perturbation; the similarity between all signals in the cell is calculated, and similar signal groups are selected; the combined perturbation of the first and second levels of perturbation, along with the dispersion (e.g., variance) with similar signal groups, is calculated to obtain the comprehensive perturbation of each spectral value point. The size of the Gaussian filter window for each point is dynamically determined based on the comprehensive perturbation (larger window for larger perturbation, and vice versa), ultimately obtaining high-quality denoised spectral data. Based on this data, the preliminary content of several target heavy metals such as Cd, Pb, and As in the cell is calculated using a built-in or pre-calibrated quantitative model (e.g., partial least squares regression PLSR model).

[0025] Preserved rhizosphere soil samples were sent to the laboratory. Total microbial DNA was extracted using a commercial DNA extraction kit. The V3-V4 region of the 16S rRNA gene was amplified by PCR, followed by high-throughput paired-end sequencing on the Illumina MiSeq platform. The sequencing data underwent quality control, noise reduction, OTU clustering, and species annotation to obtain a microbial community composition and abundance table.

[0026] Using packages such as "igraph" in R, a microbial symbiotic network was constructed based on Spearman correlations between species. Modular analysis of the network was performed to identify different functional communities. Topological characteristics such as degree centrality and betweenness centrality of each node were calculated. Correlation analysis was conducted between the network topological characteristics and the preliminary heavy metal content data obtained in the previous steps to screen out several bacterial species (such as certain heavy metal-resistant Proteobacteria and Actinobacteria species) that showed significant abundance variations in contaminated units and occupied key pivotal positions in the symbiotic network. These were identified as key microbial biomarkers for this detection, and their relative abundance in each spatial unit was recorded.

[0027] S3. Multi-source information fusion modeling: The preliminary content of heavy metal elements, abundance of key microbial biomarkers, soil physicochemical parameters and historical pollution source influencing factors of each spatial unit are fused to construct a multi-feature dataset; machine learning algorithms are used to train the multi-feature dataset to establish a comprehensive assessment model of heavy metal pollution level.

[0028] A feature vector is constructed for each spatial unit, including: 1) the preliminary contents of Cd, Pb and As measured by XRF; 2) the relative abundance of the three key microbial biomarkers identified; 3) the soil pH and organic matter content of the unit; and 4) the historical pollution source impact factors of the unit: traffic factor (calculated based on the distance to the national highway and traffic flow model) and industrial factor (calculated based on the distance to the historical factory and the prevailing wind direction).

[0029] A training dataset was created by combining the feature vectors of all 49 spatial units and their corresponding accurate heavy metal contents measured by standard laboratory methods (such as ICP-MS) (as labels). The random forest regression algorithm was used to train this dataset and optimize the model parameters. The resulting model is the "Comprehensive Assessment Model for Heavy Metal Pollution Levels" for this farmland. This model can predict a comprehensive pollution index (CPI) that is closer to the true value based on multi-source feature vectors.

[0030] S4. Pollution assessment and risk warning: Input the spatial unit characteristic data of the farmland to be tested into the comprehensive assessment model, and output the comprehensive pollution index and risk level of each unit; Based on the output results of all units, combine the geographic information system to generate a spatial distribution map of heavy metal pollution, a risk level map and a potential diffusion trend prediction map, and issue warnings for high-risk areas.

[0031] The feature vectors of 49 spatial units (without lab labels) are input into a trained comprehensive evaluation model to obtain the CPI value for each unit. Thresholds are set: CPI < 1.0 indicates cleanliness, 1.0 ≤ CPI < 2.0 indicates slight contamination, 2.0 ≤ CPI < 3.0 indicates moderate contamination, and CPI ≥ 3.0 indicates severe contamination. Based on this, all units are classified into risk levels.

[0032] Using GIS software (such as ArcGIS), spatial interpolation (e.g., Kriging interpolation) is performed on the center coordinates, CPI value, and risk level of each cell to create a spatial distribution map of heavy metal pollution and a risk level zoning map for the farmland. Figure 3 (Illustration)

[0033] Several units with the highest CPI values ​​were selected as potential pollution sources. Combining this with local soil quality maps, groundwater flow data, and average annual rainfall, a simplified solute transport model was used to simulate the possible diffusion direction and range of heavy metals over the next 1-3 years, generating a potential diffusion trend prediction map. High-risk warning areas requiring close monitoring were marked on the map. Finally, the "three..." Figure 1 The "early warning" results are compiled into a test report and submitted to farmland managers or environmental regulatory departments.

[0034] A detection system for heavy metal pollution in agricultural land soil includes a grid-based data acquisition module: a mobile terminal (PAD) and an APP, which are used to assist field personnel in completing dynamic grid planning, point navigation, inputting XRF measurement data, photographing sampling points, and associating sample numbers.

[0035] Rapid detection and bioanalysis module: including a portable XRF spectrometer, a vehicle-mounted small microbial sample pretreatment device (for rapid DNA extraction and amplification), and spectral noise reduction algorithms and microbial sequencing data analysis workflows deployed on cloud or local servers.

[0036] Intelligent analysis and modeling module: This is the core computing engine, deployed on a server, and integrates functions such as multi-source data fusion, feature engineering, machine learning model training and invocation (random forest, neural network, etc.).

[0037] Assessment, early warning and visualization module: Developed based on Web GIS, it provides an interactive map interface for displaying pollution distribution maps, risk level maps, and diffusion prediction maps, and has functions such as threshold setting and automatic push of early warning information (SMS, email).

[0038] The modules communicate and exchange commands via wireless networks or data interfaces, forming a complete intelligent detection closed loop from the field to the cloud and then to the decision-making end.

[0039] Therefore, the present invention adopts the above-mentioned detection method and system for heavy metal pollution in agricultural land soil. By integrating in-situ rapid detection, microbial ecological analysis and multi-source information fusion modeling technology, it achieves efficient, accurate and multi-dimensional detection and risk assessment of heavy metal pollution in farmland soil.

[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting heavy metal contamination of agricultural land soil, characterized by, Includes the following steps: S1. Spatial grid construction and multi-source information acquisition: Based on the shape, area and surrounding pollution source distribution of the target farmland, the detection area is delineated using a dynamic grid division method, and the detection area is divided into multiple spatial units; soil physicochemical parameters and in-situ rapid spectral signals are collected in each spatial unit, and rhizosphere soil and microbial samples are collected simultaneously. S2. Preliminary pollution identification and microbial feature extraction: Adaptive filtering and noise reduction processing is performed on the in-situ rapid spectral signal to obtain denoised spectral data. Based on the denoised spectral data, the preliminary content of heavy metal elements in each spatial unit is calculated using a quantitative analysis model. Microbial DNA is extracted and high-throughput sequencing is performed on rhizosphere soil and microbial samples to obtain information on microbial community composition. A microbial symbiotic network is constructed to identify key microbial biomarkers and their abundance related to heavy metal pollution response. S3. Multi-source information fusion modeling: The preliminary content of heavy metal elements, abundance of key microbial biomarkers, soil physicochemical parameters and historical pollution source influencing factors of each spatial unit are fused to construct a multi-feature dataset. A comprehensive assessment model for the degree of heavy metal pollution was established by training a multi-feature dataset using machine learning algorithms. S4. Pollution assessment and risk warning: Input the spatial unit characteristic data of the farmland to be tested into the comprehensive assessment model, and output the comprehensive pollution index and risk level of each unit. Based on the output results of all units, a spatial distribution map of heavy metal pollution, a risk level map, and a potential diffusion trend prediction map are generated by combining the geographic information system, and early warnings are issued for high-risk areas.

2. The method for detecting heavy metal pollution in agricultural land soil according to claim 1, characterized in that, The dynamic mesh generation method in S1 includes: Determine the center point for dividing the farmland based on the longitudinal and transverse length ratios; use the center point as the reference to initially divide the land into a nine-square grid. Combining historical or real-time data on the distribution of surrounding pollution sources, including main traffic arteries, industrial emission sources, and sewage irrigation areas, the nine-square grid is refined, and the grid is densified in the boundary areas where the pollution sources have a strong impact, to generate the final spatial unit division.

3. The method for detecting heavy metal pollution in agricultural land soil according to claim 2, characterized in that: In S1, in-situ rapid spectral signal acquisition was performed using a portable X-ray fluorescence spectrometer; in S2, adaptive filtering and noise reduction processing included: Obtain a set of spectral signals measured multiple times within the same spatial cell; For each spectral signal, spectral values ​​are clustered, and the clustering results are arranged by wavelength to form multiple local windows; Analyze the anomaly degree of the difference between the spectral value and its neighboring values ​​within each window, and calculate the first degree of disturbance. Identify feature windows with a length greater than a threshold in the spectral signal, and calculate the second degree of perturbation by analyzing the similarity of the fitted curves between feature windows and the goodness of fit within them. Analyze the similarity between different spectral signals and select a set of similar spectral signals; Based on the first degree of disturbance, the second degree of disturbance, and the degree of dispersion of similar spectral signals at the same wavelength, the comprehensive degree of disturbance of each spectral value is calculated, and the length of the filtering window is dynamically adjusted accordingly for noise reduction.

4. The method for detecting heavy metal pollution in agricultural land soil according to claim 3, characterized in that, The construction of microbial symbiotic networks and the identification of key microbial biomarkers in S2 include: Based on the species composition and abundance data obtained by high-throughput sequencing, a microbial symbiotic network is constructed with microbial species as nodes and correlation between species as edges. The symbiotic network is analyzed for network topology characteristics, and community detection algorithms are used to divide functional modules. The centrality index of nodes in the network is calculated to identify microbial species that are in key positions in the network and are significantly related to known heavy metal pollution data, as key microbial biomarkers.

5. The method for detecting heavy metal pollution in agricultural land soil according to claim 4, characterized in that, The historical pollution source influencing factors in S3 include traffic influencing factors, industrial coal / burning influencing factors, and sewage irrigation influencing factors. The traffic influencing factors are calculated based on the traffic volume and distance of surrounding roads per unit time. The industrial coal / burning influencing factors are calculated based on the number and distance of pollution sources within a preset radius. The sewage irrigation influencing factors are calculated based on the number of pollution water input points and historical irrigation range.

6. The method for detecting heavy metal pollution in agricultural land soil according to claim 5, characterized in that: The machine learning algorithm in S3 is one or a combination of random forest, support vector machine, or neural network. The output of the comprehensive evaluation model is a continuous comprehensive pollution index, which is divided into four risk levels of clean, light pollution, moderate pollution, and heavy pollution according to a preset threshold.

7. The method for detecting heavy metal pollution in agricultural land soil according to claim 6, characterized in that, The generation of the potential diffusion trend prediction map in S4 includes: Based on soil pollution characteristics, regional hydrogeological parameters, and meteorological data, set the initial boundary conditions for pollution diffusion. Using the identified high-pollution spatial units as potential pollution sources, combined with environmental driving factors, the diffusion path and range of heavy metal pollutants in time and space are predicted by numerical simulation method.

8. The method for detecting heavy metal pollution in agricultural land soil according to claim 1, characterized in that, Further including: S5, establish a pollution time series database and dynamic updating mechanism; store the comprehensive pollution index, risk level, spatial distribution data and corresponding environmental parameters of each detection into the time series database; Regularly retest the same farmland, use the data in the time series database to train the time series prediction model, which is used to predict the pollution trend in a specific period of time in the future, and realize dynamic warning of risk.

9. A detection system for heavy metal pollution of agricultural land soil, applying a detection method for heavy metal pollution of agricultural land soil according to any one of claims 1-8, characterized in that, Including: Grid acquisition module, used to execute step S1, complete spatial grid construction and multi-source information acquisition; Fast detection and biological analysis module, used to execute step S2, complete fast detection, noise reduction processing and microbial feature extraction of spectral signal; Intelligent analysis and modeling module, used to execute step S3, complete multi-source information fusion and comprehensive evaluation model construction; Evaluation and warning and visualization module, used to execute step S4, complete pollution evaluation, risk classification, spatial mapping and warning information release.

Images