Agricultural large water surface non-point source pollution multi-source dynamic monitoring method and system

Abstract

The present application relates to the technical field of water environment monitoring, and discloses a kind of agricultural large water surface non-point source pollution multi-source dynamic monitoring method and system, comprising: synchronous acquisition satellite remote sensing, bank base water quality sensing, meteorological hydrology and unmanned aerial vehicle image multi-source data, pre-processing output standardization space-time data set, water body partition grid and sensor calibration parameter;Respectively execute remote sensing inversion and in-situ time series analysis, output single-source water quality parameter and corresponding confidence;Water quality parameter distribution map and fusion residual are obtained using space-time adaptive weighted fusion;Pollution tracing, risk classification and diffusion prediction are completed by introducing non-point source pollution diffusion model;According to the evaluation results and residual iterative optimization monitoring point and model parameter, form a closed loop.The present application realizes the wide coverage, high precision dynamic monitoring of large water surface non-point source pollution, and improves the accuracy of pollution tracing and early warning.

Description

Technical Field

[0001] This invention relates to the field of water environment monitoring technology, and in particular to a method and system for dynamic monitoring of multi-source non-point source pollution on large agricultural water surfaces. Background Technology

[0002] Agricultural water bodies, including lakes, reservoirs, and contiguous aquaculture ponds, are core carriers for agricultural irrigation, aquaculture production, and regional ecological conservation. Their water quality directly affects the quality and safety of agricultural products and the ecological health of the watershed. Affected by multiple factors such as nitrogen and phosphorus loss from farmland runoff, feed residues from aquaculture, and scouring from livestock excrement, agricultural non-point source pollution is characterized by dispersed emissions, complex causes, strong spatial and temporal heterogeneity, and sudden fluctuations. It is difficult to achieve precise monitoring across the entire area using fixed monitoring points, making it a long-standing focus and challenge in water environment management. Efficient, accurate, and comprehensive dynamic monitoring technology is a core prerequisite for controlling agricultural non-point source pollution and ensuring water quality safety.

[0003] Traditional water quality monitoring of large bodies of water mainly relies on regular manual sampling combined with laboratory testing, and a small number of fixed online monitoring stations on shore. Manual sampling is time-consuming, labor-intensive, and limited in the number of sampling points, only reflecting the local water quality at those points. It cannot depict the spatial distribution of pollution over large water areas and struggles to capture sudden non-point source pollution processes caused by rainfall runoff. Fixed online monitoring stations can achieve continuous high-frequency monitoring, but they are still essentially point-based monitoring. To cover large areas of water, a large number of stations need to be deployed, resulting in high construction and maintenance costs. Furthermore, most shore-based stations can only be deployed in near-shore areas, leaving monitoring blind spots in the central and remote areas of the water body, failing to fully present the dynamic evolution of pollution across the entire area.

[0004] In recent years, satellite remote sensing and UAV remote sensing technologies have been gradually applied to water environment monitoring, enabling large-scale, non-contact water quality inversion and compensating for the coverage shortcomings of point-based monitoring. However, satellite remote sensing suffers from significant interference from cloud cover, atmospheric conditions, and solar flares, resulting in low image effectiveness and long revisit cycles, making it difficult to meet the needs of daily dynamic monitoring. UAV hyperspectral monitoring offers high flexibility and good inversion accuracy, but its limited endurance restricts the coverage area of ​​a single operation, making it difficult to independently undertake routine monitoring tasks of large water surfaces. Existing multi-source monitoring fusion schemes mostly adopt simple weighted fusion with fixed weights, failing to dynamically adjust the fusion strategy based on data quality, spatial distance, and hydrological processes, resulting in insufficient stability of fusion accuracy. Furthermore, most schemes only output water quality parameter distribution results, lacking the ability to trace non-point source pollution and predict diffusion trends. After detecting pollution anomalies, they cannot quickly locate the source of pollution or predict the scope of impact, making it difficult to directly support pollution control and emergency response. Therefore, there is an urgent need for a multi-source collaborative, spatiotemporally adaptive, dynamic monitoring method for agricultural large water surface non-point source pollution that combines monitoring, assessment, and early warning tracing capabilities. Summary of the Invention

[0005] To address the aforementioned technical shortcomings, the purpose of this invention is to propose a multi-source dynamic monitoring method for agricultural large-scale water surface non-point source pollution. This method aims to solve the technical problems in existing technologies, such as the difficulty in balancing single-source monitoring coverage and accuracy, the rigid weighting of multi-source fusion, and the lack of source tracing, prediction, and closed-loop optimization mechanisms, which result in low monitoring accuracy, poor timeliness, and insufficient management support capabilities for large-scale water surface non-point source pollution.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a method for dynamic monitoring of multi-source non-point source pollution on large agricultural water surfaces.

[0007] The method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces includes: Step S10: Simultaneously acquire multi-source data from satellite remote sensing images, shore-based water quality sensors, meteorological and hydrological data, and UAV inspection data. After radiometric correction, spatiotemporal alignment, and grid division preprocessing, output standardized spatiotemporal datasets, water body spatial partition grids, and sensor calibration parameters. Step S20: Based on the standardized spatiotemporal dataset obtained in step S10, perform remote sensing image water quality inversion and in-situ sensing time series analysis, and output single-source water quality parameter results, time series anomaly characteristics and corresponding confidence index. Step S30: Based on the water body spatial partitioning grid of step S10 and the single-source water quality parameter results and confidence index output in step S20, multi-source data fusion is performed using a spatiotemporal adaptive weighted fusion mechanism, and the fused water quality parameter distribution map and fusion residual sequence are output. Step S40: Based on the water quality parameter distribution map output in Step S30 and the meteorological and hydrological data in Step S10, a non-point source pollution diffusion model is introduced to perform pollution source tracing and dynamic risk assessment, and output the pollution risk level, pollution source tracing results and diffusion prediction range. Step S50: Based on the pollution assessment results output in step S40 and the fusion residual sequence output in step S30, perform dynamic optimization of monitoring points and iterative calibration of model parameters, output monitoring scheduling instructions, and provide feedback to update the fusion weights and diffusion model parameters.

[0008] Preferably, step S10, which involves simultaneously acquiring multi-source data from satellite remote sensing images, shore-based water quality sensors, meteorological and hydrological data, and UAV inspection data, and preprocessing them through radiometric correction, spatiotemporal alignment, and grid division to obtain a standardized spatiotemporal dataset, a water body spatial partition grid, and sensor calibration parameters, specifically includes: Step S101: Simultaneously acquire multispectral satellite remote sensing images, ammonia nitrogen, total phosphorus, and dissolved oxygen water quality time series data from shore-based fixed stations, and simultaneously collect meteorological and hydrological data on wind speed, wind direction, water level, and flow rate, as well as UAV hyperspectral inspection image data. Step S102: Perform radiometric calibration, atmospheric correction and geometric registration on remote sensing and UAV imagery; perform outlier removal and sliding filtering on in-situ sensing data; complete spatiotemporal alignment of multi-source data based on a unified geodetic coordinate system and timestamps to generate a standardized spatiotemporal dataset. Step S103: Divide the monitored water surface into several equal grid units according to the topography and hydrological characteristics of the water area, calibrate the spatial coordinates and sensing coefficients of each shore-based station, and output the spatial zoning grid of the water body and the sensor calibration parameters.

[0009] Preferably, step S20, based on the standardized spatiotemporal dataset obtained in step S10, involves performing remote sensing image water quality inversion and in-situ sensing time-series analysis, and outputting single-source water quality parameter results, time-series anomaly characteristics, and corresponding confidence indices. This step specifically includes: Step S201: Based on the preprocessed multispectral and hyperspectral images, the water quality concentration parameters of each grid are inverted using the empirical ratio method and the partial least squares regression model. At the same time, the confidence level of remote sensing inversion is calculated based on the cloud cover and signal-to-noise ratio of the images. Step S202: Perform sliding window statistics and time series anomaly detection on the shore-based water quality time series data, extract abnormal features such as concentration mutations and continuous exceedances, and calculate the confidence level of in-situ monitoring based on data completeness and equipment status; Step S203: Output the single-source water quality parameter results, time-series anomaly characteristics and corresponding confidence index of each grid cell. The confidence value ranges from 0 to 1.

[0010] Preferably, in step S201, the confidence level of remote sensing inversion is graded according to cloud cover: the confidence level is 0.9 when the cloud cover is less than 10%, the confidence level is 0.7 when the cloud cover is 10%-30%, and the confidence level is below 0.4 when the cloud cover is greater than 30%.

[0011] Preferably, step S30, based on the water body spatial partitioning grid of step S10 and the single-source water quality parameter results and confidence index output in step S20, employs a spatiotemporal adaptive weighted fusion mechanism to perform multi-source data fusion, and outputs the fused water quality parameter distribution map and fusion residual sequence, specifically includes: Step S301: Using the spatial partitioning grid of the water body as the basic unit, the single-source confidence as the basic weight, and combining the spatial distance attenuation coefficient and the time attenuation coefficient, a spatiotemporal adaptive weighted fusion model is constructed. Step S302: Weighted fusion of multi-source water quality parameters for each grid cell to obtain continuously distributed water quality parameter raster data for the entire water area, and generate a fused water quality parameter distribution map; Step S303: Calculate the deviation between the fused value and the measured value at the in-situ station location as the fusion residual, and output the fused water quality parameter distribution map and the fusion residual sequence.

[0012] Preferably, step S40, based on the water quality parameter distribution map output in step S30 and the meteorological and hydrological data in step S10, introduces a non-point source pollution diffusion model to perform pollution source tracing and dynamic risk assessment, and outputs the pollution risk level, pollution source tracing results, and diffusion prediction range, specifically includes: Step S401: Using the water quality parameter distribution map as the initial concentration field, and combining wind speed, wind direction, water level, flow rate and hydrological parameters, construct a two-dimensional hydrodynamic-water quality coupled diffusion model; Step S402: Use a reverse tracing algorithm to locate the source direction and confluence location of pollution in high-concentration areas, and output the pollution source tracing results; perform a forward simulation of the pollution diffusion trend in the next 24 hours to delineate the diffusion prediction range; Step S403: Classify the pollution risk level according to the multiple of water quality concentration exceeding the standard, and divide it into three levels: light risk, medium risk and heavy risk. Output the pollution risk level, pollution source tracing results and diffusion prediction range.

[0013] Preferably, step S50, based on the pollution assessment results output in step S40 and the fusion residual sequence output in step S30, includes the following steps: performing dynamic optimization of monitoring points and iterative calibration of model parameters, outputting monitoring scheduling instructions, and providing feedback to update the fusion weights and diffusion model parameters. Step S501: For high-risk and high-residual areas, generate drone-based encrypted inspection and mobile monitoring point scheduling instructions to supplement monitoring data; Step S502: Based on the newly added monitoring data and the statistical results of the fusion residual, iteratively optimize the weight coefficients of the multi-source fusion and the parameters of the diffusion model, and update the fusion weights of step S30 and the diffusion model parameters of step S40. Step S503: When the fusion residuals of three consecutive cycles are all below the set threshold, set the monitoring steady-state flag and output the monitoring scheduling command and steady-state flag.

[0014] This invention also provides a multi-source dynamic monitoring system for non-point source pollution on large agricultural water surfaces, comprising: Multi-source data acquisition and preprocessing module: used to simultaneously acquire multi-source data from satellite remote sensing images, shore-based water quality sensors, meteorological and hydrological data, and UAV inspection data. After radiometric correction, spatiotemporal alignment and grid division preprocessing, it outputs standardized spatiotemporal datasets, water body spatial partition grids and sensor calibration parameters. Single-source water quality inversion analysis module: Based on the standardized spatiotemporal dataset obtained by the multi-source data acquisition and preprocessing module, it performs remote sensing image water quality inversion and in-situ sensing time series analysis, and outputs single-source water quality parameter results, time series anomaly characteristics and corresponding confidence indexes. Spatiotemporal adaptive fusion module: Used to perform multi-source data fusion based on the water body spatial partitioning grid of the multi-source data acquisition and preprocessing module and the single-source water quality inversion analysis module output single-source water quality parameter results and confidence index. It adopts a spatiotemporal adaptive weighted fusion mechanism to perform multi-source data fusion and outputs the fused water quality parameter distribution map and fusion residual sequence. Pollution source tracing and risk assessment module: Based on the water quality parameter distribution map output by the spatiotemporal adaptive fusion module and the meteorological and hydrological data from the multi-source data acquisition and preprocessing module, it introduces the non-point source pollution diffusion model to perform pollution source tracing and dynamic risk assessment, and outputs the pollution risk level, pollution source tracing results and diffusion prediction range. Dynamic optimization closed-loop calibration module: Based on the pollution assessment results output by the pollution source tracing risk assessment module and the fusion residual sequence output by the spatiotemporal adaptive fusion module, it performs dynamic optimization of monitoring points and iterative calibration of model parameters, outputs monitoring scheduling instructions and feeds back updates to fusion weights and diffusion model parameters.

[0015] The present invention also provides a multi-source dynamic monitoring device for non-point source pollution on large agricultural water surfaces. The multi-source dynamic monitoring device for non-point source pollution on large agricultural water surfaces includes: a memory, a processor, and a multi-source dynamic monitoring program for non-point source pollution on large agricultural water surfaces stored in the memory and executable on the processor. When the multi-source dynamic monitoring program for non-point source pollution on large agricultural water surfaces is executed by the processor, it implements the above-mentioned method.

[0016] The present invention also provides a computer program product, which includes a multi-source dynamic monitoring program for non-point source pollution on large agricultural water surfaces. When the multi-source dynamic monitoring program for non-point source pollution on large agricultural water surfaces is executed by a processor, it implements the above-mentioned method.

[0017] The beneficial effects of this invention are as follows: By integrating multi-source data from satellite remote sensing, shore-based sensing, UAV inspection, and meteorological and hydrological data, it combines the advantages of wide-area coverage and high-precision in-situ observation, constructing a monitoring data base that integrates space and ground. This solves the problem of the difficulty in balancing the range and accuracy of traditional single-source monitoring, and provides multi-dimensional data support for dynamic monitoring of non-point source pollution across the entire area.

[0018] A spatiotemporal adaptive weighted fusion mechanism is adopted, which dynamically allocates fusion weights by combining data confidence, spatial distance and time decay, replacing the simple fusion with fixed weights. This makes the fusion results more consistent with the actual water quality distribution, effectively reduces the fusion error caused by factors such as cloud cover and sparse stations, and improves the overall accuracy of the water quality distribution map of the whole region.

[0019] The introduction of a hydrodynamic-water quality coupled diffusion model extends the monitoring from current status to pollution source tracing and diffusion prediction. It can not only present the current pollution distribution, but also locate the pollution inflow source and predict the diffusion trend, providing a basis for decision-making for non-point source pollution control and emergency response.

[0020] By dynamically optimizing monitoring points driven by pollution risk and fusion residuals, and combining this with iterative calibration of model parameters, a closed-loop management system of "monitoring-fusion-evaluation-optimization" has been formed. This system can automatically supplement monitoring for high-risk and high-error areas, continuously improving the system's monitoring accuracy and adaptability. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the first embodiment of a multi-source dynamic monitoring method for non-point source pollution on large agricultural water surfaces according to the present invention. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0023] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0024] In the first embodiment, the method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces includes: Step S10: Simultaneously acquire multi-source data from satellite remote sensing images, shore-based water quality sensors, meteorological and hydrological data, and UAV inspection data. After radiometric correction, spatiotemporal alignment, and grid division preprocessing, output standardized spatiotemporal datasets, water body spatial partition grids, and sensor calibration parameters.

[0025] It should be noted that in this step, satellite remote sensing provides large-scale, periodic, global observation data; shore-based fixed stations provide high-precision, high-frequency in-situ water quality time-series data; meteorological and hydrological data provide environmental driving parameters for pollution diffusion; and UAV hyperspectral imagery provides mesoscale, high-precision supplementary observations for key areas. Radiometric and atmospheric corrections eliminate radiation interference from illumination and atmosphere on remote sensing images, and geometric registration ensures spatial consistency of multi-source images; outlier removal and filtering remove noise jumps in in-situ sensing data. Spatiotemporal alignment unifies data with different resolutions and sampling frequencies onto the same spatial coordinate system and time reference. Water body spatial zoning grids divide large water surfaces into standardized analysis units, providing a spatial carrier for subsequent gridded fusion and evaluation. The output standardized dataset is used for single-source inversion analysis in step S20, while the zoning grids and calibration parameters provide a spatial reference for subsequent fusion calculations.

[0026] Understandably, through this preprocessing step, heterogeneous, multi-source monitoring data of different scales and frequencies are integrated into a standardized dataset with spatiotemporal uniformity. This not only preserves the observation advantages of each data source, but also eliminates the fusion errors caused by spatial misalignment and temporal deviation, laying a unified data foundation for subsequent multi-source collaborative analysis.

[0027] It should be understood that traditional large-scale water surface monitoring often involves independent operation of various data sources, with inconsistent data standards, making it impossible to effectively coordinate and function. This step, through standardized preprocessing and grid division, incorporates scattered multi-source data into a unified spatiotemporal analysis framework, achieving the organic integration of multi-source monitoring data from space, air, and ground.

[0028] For example, in a monitoring scenario of an agricultural reservoir, the system simultaneously acquires multispectral images from Sentinel-2 satellite, hourly data on ammonia nitrogen and total phosphorus from three shore-based monitoring stations, wind speed and direction data from a meteorological station, and hyperspectral images from a UAV. After preprocessing, all data are unified to the UTM coordinate system and the 12:00 time reference of the day. The reservoir is divided into 50m×50m grid units, with each grid corresponding to a set of multi-source observation data, preparing for subsequent inversion and fusion.

[0029] Step S20: Based on the standardized spatiotemporal dataset obtained in step S10, perform remote sensing image water quality inversion and in-situ sensing time series analysis, and output single-source water quality parameter results, time series anomaly characteristics and corresponding confidence indexes.

[0030] It should be noted that this step analyzes the remote sensing and in-situ data sources independently. Remote sensing inversion is based on the statistical relationship between water spectral characteristics and water quality concentrations, retrieving concentrations of ammonia nitrogen, total phosphorus, etc., for each grid. Confidence levels are assessed comprehensively based on cloud cover and image signal-to-noise ratio, reflecting the reliability of the inversion results. In-situ time-series analysis performs statistical analysis and anomaly detection on the station data, identifying abrupt concentration changes and continuous increases. Confidence levels are assessed based on data completeness and equipment operating status, reflecting the quality level of the in-situ data. Both types of single-source results are mapped to a unified grid and accompanied by confidence level labels, providing input for subsequent adaptive fusion. The output single-source results are then fused with the confidence level input in step S30 for multi-source fusion.

[0031] Understandably, this step involves independently interpreting and assessing the quality of each data source. This not only fully explores the information value of each type of data but also quantifies their reliability through confidence metrics, avoiding interference from low-quality data with the fusion results and providing a basis for subsequent differentiated weighted fusion.

[0032] It should be understood that existing multi-source monitoring schemes often directly fuse raw data without distinguishing between good and bad data, making them susceptible to the influence of bad data and low-quality imagery, leading to distorted fusion results. This step introduces confidence assessment, assigning a quality label to each data point, enabling subsequent fusion to select the best data and improve the robustness of the fusion process.

[0033] For example, under clear weather conditions, satellite imagery shows only 5% cloud cover, resulting in an inversion confidence level of 0.9 and high reliability of water quality inversion results across all grids. However, in cloudy weather, with localized cloud cover reaching 40%, the inversion confidence level for the corresponding area drops to 0.35, reducing the reference value of the inversion results. The shore-based base station data integrity rate is 98%, the equipment is operating normally, and the original location confidence level is 0.95, indicating high data reliability.

[0034] Step S30: Based on the water body spatial partitioning grid of step S10 and the single-source water quality parameter results and confidence index output in step S20, multi-source data fusion is performed using a spatiotemporal adaptive weighted fusion mechanism, and the fused water quality parameter distribution map and fusion residual sequence are output.

[0035] It should be noted that the spatiotemporal adaptive weighted fusion uses grid cells as the unit, and the basic weights are determined by the confidence level of each data source; the higher the confidence level, the greater the weight. Simultaneously, a spatial distance attenuation coefficient is introduced to reduce the influence weight of in-situ data in areas far from the original site, thus increasing the weight of remote sensing data. A temporal attenuation coefficient is also introduced to gradually reduce the weight of observation data from distant time intervals. The final water quality concentration value for each grid is obtained through weighted calculation, forming a continuous global water quality parameter distribution map. The fusion residual is calculated based on the measured values ​​from the in-situ site, determining the deviation between the fused value and the measured value for the corresponding grid, which is used to measure the fusion accuracy. The output water quality distribution map is input into step S40 for pollution assessment, and the fusion residual is input into step S50 for closed-loop optimization.

[0036] Understandably, this step achieves complementary advantages of multi-source data through adaptive weight adjustment in both spatiotemporal dimensions: high-precision in-situ data is mainly used near the site, while remote sensing coverage is mainly used in areas far from the site. High-quality data has a high weight, while low-quality data has a low weight, so that the overall accuracy of the fusion result across the entire domain reaches the optimal level.

[0037] It should be understood that traditional multi-source fusion often uses fixed-weight averaging, which cannot adapt to the differences in data quality across different regions and time periods, and is prone to overfitting near stations and insufficient accuracy at distant stations. The adaptive weighting mechanism in this step can dynamically adjust weights according to spatial location and data quality, significantly improving the balance accuracy of large-scale water surface fusion.

[0038] For example, within a 500-meter radius of a shore-based station, in-situ data accounts for 70% of the weight, while remote sensing data accounts for 30%, resulting in a fusion result that closely approximates the high-precision measured values. In waters 2 kilometers away from the station, the in-situ weight drops to 20%, while the remote sensing weight increases to 80%, relying on remote sensing data to reflect the overall distribution. In cloudy areas, the remote sensing confidence level decreases, and the corresponding weight automatically decreases, while the in-situ data weight increases accordingly, ensuring the stability of the fusion result.

[0039] Step S40: Based on the water quality parameter distribution map output in Step S30 and the meteorological and hydrological data in Step S10, a non-point source pollution diffusion model is introduced to perform pollution source tracing and dynamic risk assessment, and output the pollution risk level, pollution source tracing results and diffusion prediction range.

[0040] It should be noted that the two-dimensional hydrodynamic-water quality coupled diffusion model uses the fused water quality concentration distribution as the initial condition and hydrological and meteorological parameters such as wind speed, wind direction, water level, and flow rate as driving conditions to simulate the transport and diffusion process of pollutants in water bodies. The reverse tracing starts from areas of high pollution concentration and extrapolates the transport path of pollutants to locate the source location and approximate direction of pollution inflow. The forward simulation predicts the movement path and concentration changes of the pollution plume over a certain period, delineating the diffusion impact range. Risk levels are classified according to the multiple by which the water quality concentration exceeds the standard limit, corresponding to different control levels. The output source tracing results, risk levels, and predicted ranges can be directly used for pollution control and emergency decision-making.

[0041] Understandably, this step extends from static water quality distribution monitoring to dynamic pollution source tracing and prediction, answering not only "where is the pollution?" but also "where does the pollution come from and where will it go?" This upgrades the monitoring system from simple data collection to a management tool with decision support capabilities.

[0042] It should be understood that traditional monitoring systems only output current water quality data and lack analytical and predictive capabilities. Once pollution is detected, they cannot quickly pinpoint the source or predict the extent of impact, leading to delays in control and response. The diffusion model and source tracing algorithm in this step endow the system with dynamic assessment and early warning capabilities, enhancing the proactiveness of non-point source pollution control.

[0043] For example, monitoring revealed that the total phosphorus concentration in the northwest area of ​​the reservoir exceeded the standard, reaching a high-risk level. Through reverse tracing algorithms, combined with northwesterly wind meteorological conditions, the source of pollution was determined to be mainly from the runoff inlet of farmland on the northwest bank; forward simulation predicted that the pollution plume would spread southeastward in the next 12 hours, affecting downstream aquaculture areas, and the system automatically delineated the warning area and pushed it to the management department.

[0044] Step S50: Based on the pollution assessment results output in step S40 and the fusion residual sequence output in step S30, perform dynamic optimization of monitoring points and iterative calibration of model parameters, output monitoring scheduling instructions, and provide feedback to update the fusion weights and diffusion model parameters.

[0045] It should be noted that the dynamic optimization of monitoring points targets high-risk areas and areas with high fusion residuals by deploying drones for intensive inspections or mobile monitoring stations to supplement monitoring data and fill monitoring blind spots. Model parameter calibration utilizes newly added measured data and fusion residual statistical results to optimize the weight coefficients of the fusion model and the diffusion and attenuation coefficients of the diffusion model through least-squares iteration, making the model more closely reflect the actual characteristics of the local waters. The parameter optimization results are fed back to steps S30 and S40 to continuously improve the accuracy of subsequent fusion and prediction. The steady-state indicator represents whether the system has entered a stable convergence state, providing a status indication for routine monitoring.

[0046] Understandably, this step enables dynamic scheduling of monitoring resources and self-optimization of model parameters, forming a closed-loop mechanism of monitoring-driven optimization and optimization feeding back into monitoring. This allows the system to continuously improve itself, maintain high-precision operation over a long period, and improve the efficiency of monitoring resource utilization.

[0047] It should be understood that traditional monitoring systems have fixed monitoring points and model parameters, making them unable to adapt to changes in seasons and hydrological conditions. Their accuracy gradually declines after long-term operation, and monitoring resources are evenly distributed, failing to strengthen key areas. The closed-loop optimization mechanism in this step solves these problems, enabling the system to have adaptive and self-evolving capabilities.

[0048] For example, during the rainy season, farmland runoff increases, and the fusion residual in the reservoir inlet area remains high. The system automatically generates drone inspection instructions to intensify flight monitoring around the inlet. Using newly added hyperspectral data, the fusion weight and diffusion coefficient of the area are iteratively optimized. After two cycles of calibration, the fusion residual in the area decreases by 40%, the monitoring accuracy is significantly improved, and the consistency of pollution diffusion prediction is also significantly improved.

[0049] Example 2: Furthermore, the present invention provides a multi-source dynamic monitoring system for agricultural large-scale water source pollution, employing a multi-source dynamic monitoring method for agricultural large-scale water source pollution from the above embodiments. This system addresses the technical problems in existing technologies, such as difficulty in balancing single-source monitoring coverage and accuracy, fixed weighting of multi-source fusion, and lack of source tracing, prediction, and closed-loop optimization mechanisms, leading to low monitoring accuracy, poor timeliness, and insufficient management support capabilities for large-scale water source pollution. The beneficial effects of the multi-source dynamic monitoring system for agricultural large-scale water source pollution provided by the present invention are the same as those of the multi-source dynamic monitoring method for agricultural large-scale water source pollution provided in the above embodiments. Other technical features of the multi-source dynamic monitoring system for agricultural large-scale water source pollution are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0050] Example 3: This invention provides a multi-source dynamic monitoring device for non-point source pollution on large agricultural water surfaces. The device includes: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which are then executed to enable the at least one processor to perform the multi-source dynamic monitoring method for non-point source pollution on large agricultural water surfaces described in Example 1. The multi-source dynamic monitoring device for non-point source pollution on large agricultural water surfaces in this invention can include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and vehicle-mounted terminals (e.g., vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. This multi-source dynamic monitoring device for non-point source pollution on large agricultural water surfaces is merely an example and should not limit the functionality or scope of the invention. A dynamic monitoring device for multi-source non-point source pollution in large agricultural water bodies may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) that can perform various appropriate actions and processes according to a program stored in a read-only memory or a program loaded from a storage device into a random access memory. The random access memory also stores various programs and data required for the operation of the dynamic monitoring device. The processing unit, read-only memory, and random access memory are interconnected via a bus. An I / O interface is also connected to the bus. Typically, the following systems can be connected to the I / O interface: input devices including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices including, for example, magnetic tapes, hard disks, etc.; and communication devices. The communication device allows the dynamic monitoring device to communicate wirelessly or wiredly with other devices to exchange data. While a dynamic monitoring device for multi-source non-point source pollution in large agricultural water bodies with various systems has been described, it should be understood that implementation or possession of all described systems is not required. It can be implemented alternatively or have more or fewer systems.

[0051] Example 4: This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method for dynamic monitoring of multi-source non-point source pollution on large agricultural water surfaces. The computer program product provided by this invention can solve the technical problems in the prior art, such as the difficulty in balancing single-source monitoring coverage and accuracy, the fixed weighting of multi-source fusion, and the lack of source tracing, prediction, and closed-loop optimization mechanisms, leading to low accuracy, poor timeliness, and insufficient management support capabilities in monitoring non-point source pollution on large water surfaces. Compared with the prior art, the beneficial effects of the computer program product provided by this invention are the same as those of the above-described method for dynamic monitoring of multi-source non-point source pollution on large agricultural water surfaces, and will not be repeated here.

[0052] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a read-only memory. When the computer program is executed by a processing device, it performs the functions defined in the methods of the embodiments disclosed in this invention.

[0053] It should be understood that the various parts disclosed in this invention can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0054] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the present invention and its equivalents, the present invention also intends to include these modifications and variations.

Claims

1. A method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces, characterized in that, The methods include: Step S10: Simultaneously acquire multi-source data from satellite remote sensing images, shore-based water quality sensors, meteorological and hydrological data, and UAV inspection data. After radiometric correction, spatiotemporal alignment, and grid division preprocessing, output standardized spatiotemporal datasets, water body spatial partition grids, and sensor calibration parameters. Step S20: Based on the standardized spatiotemporal dataset obtained in step S10, perform remote sensing image water quality inversion and in-situ sensing time series analysis, and output single-source water quality parameter results, time series anomaly characteristics and corresponding confidence index. Step S30: Based on the water body spatial partitioning grid of step S10 and the single-source water quality parameter results and confidence index output in step S20, multi-source data fusion is performed using a spatiotemporal adaptive weighted fusion mechanism, and the fused water quality parameter distribution map and fusion residual sequence are output. Step S40: Based on the water quality parameter distribution map output in Step S30 and the meteorological and hydrological data in Step S10, a non-point source pollution diffusion model is introduced to perform pollution source tracing and dynamic risk assessment, and output the pollution risk level, pollution source tracing results and diffusion prediction range. Step S50: Based on the pollution assessment results output in step S40 and the fusion residual sequence output in step S30, perform dynamic optimization of monitoring points and iterative calibration of model parameters, output monitoring scheduling instructions, and provide feedback to update the fusion weights and diffusion model parameters.

2. The method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces according to claim 1, characterized in that, Step S10 involves the simultaneous acquisition of multi-source data from satellite remote sensing images, shore-based water quality sensors, meteorological and hydrological data, and UAV inspection data. After radiometric correction, spatiotemporal alignment, and grid division preprocessing, a standardized spatiotemporal dataset, a water body spatial partition grid, and sensor calibration parameters are obtained. Specifically, this includes: Step S101: Simultaneously acquire multispectral satellite remote sensing images, ammonia nitrogen, total phosphorus, and dissolved oxygen water quality time series data from shore-based fixed stations, and simultaneously collect meteorological and hydrological data on wind speed, wind direction, water level, and flow rate, as well as UAV hyperspectral inspection image data. Step S102: Perform radiometric calibration, atmospheric correction and geometric registration on remote sensing and UAV imagery, remove outliers and perform sliding filtering on in-situ sensing data, and complete spatiotemporal alignment of multi-source data based on a unified geodetic coordinate system and timestamp to generate a standardized spatiotemporal dataset. Step S103: Divide the monitored water surface into several equal grid units according to the topography and hydrological characteristics of the water area, calibrate the spatial coordinates and sensing coefficients of each shore-based station, and output the spatial zoning grid of the water body and the sensor calibration parameters.

3. The method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces according to claim 2, characterized in that, In step S20, based on the standardized spatiotemporal dataset obtained in step S10, the steps of performing remote sensing image water quality inversion and in-situ sensing time-series analysis, and outputting single-source water quality parameter results, time-series anomaly characteristics, and corresponding confidence indices, specifically include: Step S201: Based on the preprocessed multispectral and hyperspectral images, the water quality concentration parameters of each grid are inverted using the empirical ratio method and the partial least squares regression model. At the same time, the confidence level of remote sensing inversion is calculated based on the cloud cover and signal-to-noise ratio of the images. Step S202: Perform sliding window statistics and time series anomaly detection on the shore-based water quality time series data, extract abnormal features such as concentration mutations and continuous exceedances, and calculate the confidence level of in-situ monitoring based on data completeness and equipment status; Step S203: Output the single-source water quality parameter results, time-series anomaly characteristics and corresponding confidence index of each grid cell. The confidence value ranges from 0 to 1.

4. The method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces according to claim 3, characterized in that, In step S201, the confidence level of remote sensing inversion is classified according to cloud cover. When the cloud cover is less than 10%, the confidence level is 0.9; when the cloud cover is 10%-30%, the confidence level is 0.7; when the cloud cover is greater than 30%, the confidence level is 0.4 or less, and >0.

5. The method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces according to claim 1, characterized in that, In step S30, based on the water body spatial partitioning grid of step S10 and the single-source water quality parameter results and confidence index output in step S20, a spatiotemporal adaptive weighted fusion mechanism is used to perform multi-source data fusion, and the steps of outputting the fused water quality parameter distribution map and fusion residual sequence are specifically included: Step S301: Using the spatial partitioning grid of the water body as the basic unit, the single-source confidence as the basic weight, and combining the spatial distance attenuation coefficient and the time attenuation coefficient, a spatiotemporal adaptive weighted fusion model is constructed. Step S302: Weighted fusion of multi-source water quality parameters for each grid cell to obtain continuously distributed water quality parameter raster data for the entire water area, and generate a fused water quality parameter distribution map; Step S303: Calculate the deviation between the fused value and the measured value at the in-situ station location as the fusion residual, and output the fused water quality parameter distribution map and the fusion residual sequence.

6. The method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces according to claim 1, characterized in that, In step S40, based on the water quality parameter distribution map output in step S30 and the meteorological and hydrological data in step S10, a non-point source pollution diffusion model is introduced to perform pollution source tracing and dynamic risk assessment, and the steps of outputting pollution risk level, pollution source tracing results, and diffusion prediction range specifically include: Step S401: Using the water quality parameter distribution map as the initial concentration field, and combining wind speed, wind direction, water level, flow rate and hydrological parameters, construct a two-dimensional hydrodynamic-water quality coupled diffusion model; Step S402: Use the reverse tracing algorithm to locate the source direction and confluence location of pollution in the high-concentration area, output the pollution source tracing results, positively simulate the pollution diffusion trend in the next 24 hours, and delineate the diffusion prediction range; Step S403: Classify the pollution risk level according to the multiple of water quality concentration exceeding the standard, and divide it into three levels: light risk, medium risk and heavy risk. Output the pollution risk level, pollution source tracing results and diffusion prediction range.

7. The method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces according to claim 1, characterized in that, In step S50, based on the pollution assessment results output in step S40 and the fusion residual sequence output in step S30, the steps of performing dynamic optimization of monitoring points and iterative calibration of model parameters, outputting monitoring scheduling instructions, and feeding back updates to the fusion weights and diffusion model parameters specifically include: Step S501: For high-risk and high-residual areas, generate drone-based encrypted inspection and mobile monitoring point scheduling instructions to supplement monitoring data; Step S502: Based on the newly added monitoring data and the statistical results of the fusion residual, iteratively optimize the weight coefficients of the multi-source fusion and the parameters of the diffusion model, and update the fusion weights of step S30 and the diffusion model parameters of step S40. Step S503: When the fusion residuals of three consecutive cycles are all below the set threshold, set the monitoring steady-state flag and output the monitoring scheduling command and steady-state flag.

8. A multi-source dynamic monitoring system for agricultural large-scale water surface non-point source pollution, applied to the multi-source dynamic monitoring method for agricultural large-scale water surface non-point source pollution as described in any one of claims 1 to 7, characterized in that, The multi-source dynamic monitoring system for non-point source pollution on large agricultural water surfaces includes: Multi-source data acquisition and preprocessing module: used to simultaneously acquire multi-source data from satellite remote sensing images, shore-based water quality sensors, meteorological and hydrological data, and UAV inspection data. After radiometric correction, spatiotemporal alignment and grid division preprocessing, it outputs standardized spatiotemporal datasets, water body spatial partition grids and sensor calibration parameters. Single-source water quality inversion analysis module: Based on the standardized spatiotemporal dataset obtained by the multi-source data acquisition and preprocessing module, it performs remote sensing image water quality inversion and in-situ sensing time series analysis, and outputs single-source water quality parameter results, time series anomaly characteristics and corresponding confidence indexes. Spatiotemporal adaptive fusion module: Used to perform multi-source data fusion based on the water body spatial partitioning grid of the multi-source data acquisition and preprocessing module and the single-source water quality inversion analysis module output single-source water quality parameter results and confidence index. It adopts a spatiotemporal adaptive weighted fusion mechanism to perform multi-source data fusion and outputs the fused water quality parameter distribution map and fusion residual sequence. Pollution source tracing and risk assessment module: Based on the water quality parameter distribution map output by the spatiotemporal adaptive fusion module and the meteorological and hydrological data from the multi-source data acquisition and preprocessing module, it introduces the non-point source pollution diffusion model to perform pollution source tracing and dynamic risk assessment, and outputs the pollution risk level, pollution source tracing results and diffusion prediction range. Dynamic optimization closed-loop calibration module: Based on the pollution assessment results output by the pollution source tracing risk assessment module and the fusion residual sequence output by the spatiotemporal adaptive fusion module, it performs dynamic optimization of monitoring points and iterative calibration of model parameters, outputs monitoring scheduling instructions and feeds back updates to fusion weights and diffusion model parameters.

9. A multi-source dynamic monitoring device for non-point source pollution on large agricultural water surfaces, characterized in that, The agricultural large water surface non-point source pollution multi-source dynamic monitoring device includes: a memory, a processor, and an agricultural large water surface non-point source pollution multi-source dynamic monitoring program stored in the memory and executable on the processor. When the agricultural large water surface non-point source pollution multi-source dynamic monitoring program is executed by the processor, it implements the agricultural large water surface non-point source pollution multi-source dynamic monitoring method according to any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes a multi-source dynamic monitoring program for non-point source pollution on large agricultural water surfaces. When the multi-source dynamic monitoring program for non-point source pollution on large agricultural water surfaces is executed by a processor, it implements a method for multi-source dynamic monitoring of non-point source pollution on large agricultural water surfaces according to any one of claims 1 to 7.

Images