A method for tracing water pollutants based on environmental DNA and machine learning

Abstract

The present invention provides a method for tracing water pollutants based on environmental DNA and machine learning, comprising the following steps: Step S1: Regional investigation of pollutant sources and characterization of water body characteristics; Step S2: Extraction of environmental DNA, amplification with specific primers, and DNA sequencing and quantification; Step S3: Construction and training of a machine learning model using DNA data and water quality data; Step S4: Collecting water samples in the actual polluted area, inputting the data into the model, and outputting the types and proportions of pollution sources. By combining environmental DNA technology and machine learning models, the present invention innovatively proposes a method for tracing water pollutants that is adaptable to multiple pollution sources, accurate and efficient, and has high application value and broad promotion prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of water pollution source analysis, and particularly to a method for tracing water pollutants by combining environmental DNA (eDNA) technology and machine learning models. Background Art

[0002] Tracing water pollutants is of great significance for water environmental protection. How to accurately judge the type of pollution sources and locate the positions of pollution sources plays an important role in the protection of China's water environment. Traditional pollution source tracing methods usually rely on mathematical models and physical and chemical monitoring technologies, and have limitations when facing complex and dynamic actual situations of pollution sources, and their effects are very limited.

[0003] Environmental DNA (eDNA) technology has developed rapidly in recent years and has been increasingly widely used in ecological environment monitoring. It can be used to identify the biodiversity of organisms in water bodies or other environments. By designing specific primers for specific species, specific species in the environment can be identified. This technology provides a new solution idea for the problem of pollution source tracing. However, due to problems such as the accuracy of DNA detection itself, time-consuming and laborious detection, and sequencing costs, there are few reports on using environmental DNA for pollution source tracing at present, and related technologies and ideas are not yet mature.

[0004] In related technologies, for example, Chinese Patent Document CN119151457A provides a method for tracing pollution sources in a basin based on spatio-temporal sequence eDNA data, which can provide accurate pollution source tracing information in a short time. However, it mainly focuses on pollutants in water bodies and requires multi-point and long-term sampling and analysis of water bodies, which is relatively laborious. Chinese Patent Document CN107058580A provides a method for quantitatively analyzing the sources of water microbial pollution. It mainly focuses on pollution that can be characterized by microbial indicators such as fecal pollution, and uses a standard curve for quantification. Although it has strong pertinence, there are also many limitations, so it is difficult to be used for pollution source tracing in the scenario of multiple pollution sources. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for tracing water pollutants based on environmental DNA and machine learning, which analyzes the characteristics of pollution sources from multiple angles by combining environmental DNA technology and traditional water quality data, and realizes the accurate tracing of pollutants through a machine learning model.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A method for tracing water pollutants based on environmental DNA and machine learning, which is used to analyze the characteristics of pollution sources from multiple angles and realize the accurate tracing of pollutants, includes the following steps:

[0008] Step S1: Regional pollutant source investigation and water body characteristics characterization;

[0009] Step S11: Identify the types of potential pollution sources and obtain their raw sewage samples; determine the possible types of pollution sources through on-site investigation and historical data analysis, including industrial wastewater, agricultural runoff, and domestic sewage;

[0010] Step S12: Obtain water quality characteristic data of the pollution source;

[0011] The water samples were initially screened using a rapid water quality tester and spectrophotometry to obtain key pollution indicators, including pH, chemical oxygen demand, total phosphorus, total nitrogen, suspended solids, and fluorescence data;

[0012] The raw sewage sample is analyzed by high performance liquid chromatography, gas chromatography-mass spectrometry and fluorescence spectroscopy to obtain its chemical composition characteristics;

[0013] The microbial communities in the collected samples were sequenced by amplification of 16S rRNA and 18S rRNA to obtain a background database of environmental DNA;

[0014] Step S13: Tracking specific biomarkers through environmental DNA technology; targeting specific pollution sources, including characteristic microorganisms of industrial pollutant emissions and bacterial communities of agricultural pollution sources, screening specific DNA markers, and amplifying and analyzing them using real-time fluorescence quantitative PCR and metagenomic sequencing;

[0015] Step S2: environmental DNA extraction, specific primer amplification and DNA sequencing quantification;

[0016] Step S21: providing a multidimensional data set for model training; collecting environmental DNA samples from different pollution sources, extracting DNA data based on qPCR, digital PCR and high-throughput sequencing technology, and measuring species abundance;

[0017] Combined with water quality parameters, a pollution source feature dataset was constructed as a machine learning training sample;

[0018] Step S22: Establish a machine learning model to achieve pollution source analysis; use a multi-model integrated learning framework to perform deep algorithm training, including random forest, XGBoost, and support vector machine;

[0019] K-fold cross validation and Bayesian optimization are used to tune model parameters to ensure high accuracy and generalization ability of pollution source analysis;

[0020] Step S3: construct and train a machine learning model using DNA data and water quality data;

[0021] Step S31: Integrate the characteristic indicators of environmental DNA sequence data and corresponding water quality data from different pollution sources; form a training data set with multi-dimensional characteristics for creating a comprehensive description of pollution source characteristics and their corresponding molecular biology information;

[0022] Step S32: Achieve efficient pollution analysis through model training and validation; combine supervised learning and unsupervised learning methods, use deep neural networks to extract complex pollution characteristics, and at the same time use the K-means clustering algorithm for adaptive learning of pollutant categories; analyze the contribution degree of model features through Shapley value analysis and evaluate the importance of pollution factors;

[0023] Step S4: Collect water samples in the actual polluted area, input the data into the model and output the pollution source type and proportion;

[0024] Step S41: Collect water samples in the actual polluted area;

[0025] Adopt the spatial grid sampling method, set multiple monitoring points in the polluted area, and arrange them at multiple points from upstream to midstream to downstream to ensure wide spatial coverage;

[0026] Use drones and automatic sampling equipment for regular or fixed-point sampling to avoid human interference;

[0027] Step S42: Measure water quality data and environmental DNA data;

[0028] Use fluorescence quantitative PCR to amplify specific DNA fragments for gene fingerprint identification of pollution sources;

[0029] Adopt ultra-high resolution mass spectrometry to analyze water pollutants and obtain DNA sequence information by combining nanopore sequencing;

[0030] Step S43: Input the above data into the machine learning model to analyze the pollution source type and its proportion; combine multi-modal data fusion and input it into the model, including environmental DNA, chemical pollutants and water quality parameters, and analyze the contribution rate of pollution sources through the model;

[0031] Use probability prediction to calculate the proportion of pollution sources and generate a pollution source tracing report to provide a scientific basis for pollution control.

[0032] Preferably, the specific steps of step S13 are as follows:

[0033] Step S131: Classification of pollution source categories:

[0034] Industrial pollution: Screen microorganisms that are tolerant to specific metal ions;

[0035] Agricultural pollution: Screen including fecal indicator bacteria and pesticide-degrading bacteria;

[0036] Domestic sewage: Screening includes human-derived indicator bacteria and antibiotic resistance genes;

[0037] Step S132: Database comparison:

[0038] Perform comparison by combining the NCBI GenBank and Silva databases to screen for environmental DNA biomarkers associated with known pollution sources;

[0039] Adopt metagenomics analysis to identify specific genes, including heavy metal tolerance genes or antibiotic resistance genes;

[0040] Step S133: Select specific biomarker primers for subsequent amplification primer design:

[0041] Use Primer-BLAST and Oligo 7.0 to design specific PCR primers for pollution source-related microorganisms;

[0042] Adopt a primer visualization tool to optimize the primer melting temperature and secondary structure to avoid non-specific amplification; ensure that the primers can distinguish different pollution sources;

[0043] Step S134: Extract DNA from the water sample using a DNA extraction kit and store the DNA at low temperature. DNA extraction: Adopt the magnetic bead method or column extraction method to efficiently recover trace environmental DNA; use a reagent to remove PCR inhibitors in the water sample;

[0044] Step S135: Use qPCR to amplify specific DNA fragments; perform environmental DNA sequencing and quantitative analysis on the amplified samples.

[0045] Preferably, in the said step S21, use water samples from different pollution sources to prepare mixed water samples with different concentrations; and measure the DNA and water quality data of each mixed water sample to generate a multi-dimensional data set;

[0046] In the said step S22, the specific steps are as follows:

[0047] Step S221: Input: Environmental DNA data and water quality indicators;

[0048] Step S222: Output: Pollution source types and their proportions; including pollution source category classification, pollution source proportion calculation, and prediction of pollution source spatio-temporal distribution;

[0049] Step S223: Adopt a multi-model integrated learning framework, combine various machine learning or deep learning methods to improve the stability and accuracy of pollution source tracing.

[0050] Preferably, in the said step S31, fuse the environmental DNA data and water quality data; eliminate data noise through data preprocessing to improve data quality;

[0051] Feature engineering is adopted, including data splicing or time series feature extraction, to construct a multi-modal dataset of pollutants-biological markers-water quality parameters for training the model;

[0052] In step S32, by using K-means / GMM clustering for adaptive learning of pollutant categories, the accuracy of pollution classification can be improved;

[0053] A random forest, XGBoost or deep neural network model is used to train the pollution source tracing system. Combining Bayesian optimization and K-fold cross-validation can further improve the generalization ability of the model to ensure prediction accuracy;

[0054] Precision, Recall, F1-score, MSE and R² metrics are used to evaluate the performance of the model; at the same time, Shapley value analysis is introduced to quantify the contribution degree of pollution factors, and the proportion of pollution sources is calculated in combination with Bayesian inference;

[0055] Finally, the visualization of the distribution and change trend of pollution sources is realized through GIS heat maps and trend prediction curves, thereby improving the scientificity and accuracy of pollution control decisions. This system integrates supervised learning methods, can accurately predict the types and proportions of pollution sources, verify the model performance, and ensure its efficient and reliable pollution source analysis.

[0056] Preferably, in step S43, the specific steps are as follows:

[0057] Step S431: Data preprocessing and fusion: Clean, standardize and normalize water quality data, environmental DNA data, and mass spectrometry data; integrate data from different data sources through data fusion technology;

[0058] Feature engineering and model training: Extract features from the fused data, including pollutant concentrations, abundances of biomarkers, and similarities of gene fingerprints, as model inputs; select appropriate machine learning models for training according to the data characteristics;

[0059] Pollution source type analysis: Through model learning, identify the characteristics of different pollution sources and classify the pollution sources, including agricultural pollution, industrial emissions, and domestic sewage; calculate their proportions in the total pollution according to the contribution degrees of each pollution source;

[0060] Probability prediction and calculation of pollution source proportion: Use probability prediction algorithms to estimate the contribution proportions of each pollution source in water pollution; combine Bayesian inference or Markov chain Monte Carlo methods to calculate the probability distribution of each pollution source and generate a pollution source proportion report;

[0061] Pollution source tracing report generation: Combine the contribution ratios of each pollution source and regional environmental data to generate a pollution source tracing report; the report should include the pollution source type, the proportion of the pollution source, and the spatial distribution information of pollutants, providing a basis for formulating pollution control strategies.

[0062] Preferably, the DNA extraction is completed by a DNA extraction kit.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] (1) The present invention combines environmental DNA and traditional water quality data to comprehensively identify pollution sources from the perspectives of DNA and water quality indicators, improving the accuracy of source tracing.

[0065] (2) The present invention uses a machine learning model to process complex data, and the results are accurate and reliable, adapting to complex scenarios with multiple pollutant sources.

[0066] (3) The present invention has a wide application range and can adapt to the source tracing needs of different regions and different pollution sources.

[0067] (4) After the model of the present invention is trained, it can be reused, saving time and effort, and the output results are fast.

[0068] (5) The present invention has high recognition and accuracy for specific sewage types (such as aquaculture wastewater).

[0069] In summary, by combining environmental DNA technology and a machine learning model, the present invention innovatively proposes a method for tracing water pollutants that can adapt to multiple pollution sources, is accurate and efficient, and has high application value and broad promotion prospects. Brief Description of the Drawings

[0070] Figure 1 It is a schematic diagram of potential pollution sources and sampling points of a method for tracing water pollutants based on environmental DNA and machine learning provided by an embodiment of the present invention;

[0071] Figure 2 It is a schematic diagram of a method for tracing water pollutants based on environmental DNA and machine learning provided by an embodiment of the present invention. Detailed Description of the Embodiments

[0072] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0073] Such as Figure 1 And Figure 2As shown in the figure, a method for tracing water pollutants based on environmental DNA and machine learning provided by this embodiment is used to analyze the characteristics of pollution sources from multiple perspectives and achieve precise tracing of pollutants, including the following steps:

[0074] Step S1: Regional research on pollution source and characterization of water body characteristics;

[0075] Step S11: Identify potential pollution source types and obtain their raw sewage samples; through on-site investigation and historical data analysis, determine possible pollution source categories, including industrial wastewater, agricultural runoff, and domestic sewage;

[0076] Step S12: Obtain water quality characteristic data of pollution sources;

[0077] Use a rapid water quality detector and spectrophotometry to preliminarily screen water samples to obtain key pollution indicators, including pH value, chemical oxygen demand, total phosphorus, total nitrogen, suspended solids, and fluorescence data;

[0078] Analyze the raw sewage samples, including high performance liquid chromatography, gas chromatography-mass spectrometry, and fluorescence spectroscopy, to obtain their chemical composition characteristics;

[0079] Collect the microbial communities in the samples and obtain the background database of environmental DNA through 16S rRNA and 18S rRNA amplification sequencing;

[0080] Step S13: Trace specific biomarker through environmental DNA technology; for specific pollution sources, including characteristic microorganisms discharged from industrial pollutants and bacterial communities of agricultural pollution sources, screen specific DNA markers and use real-time fluorescence quantitative PCR and metagenomic sequencing for amplification and analysis;

[0081] Step S131: Classification of pollution source categories:

[0082] Industrial pollution: Screen microorganisms that are tolerant to specific metal ions;

[0083] Agricultural pollution: Screen fecal indicator bacteria and pesticide-degrading bacteria;

[0084] Domestic sewage: Screen human-derived indicator bacteria and antibiotic resistance genes;

[0085] Step S132: Database comparison:

[0086] Combine with NCBI GenBank and Silva databases for comparison to screen environmental DNA biomarkers associated with known pollution sources;

[0087] Use metagenomics analysis to identify specific genes, including heavy metal tolerance genes or antibiotic resistance genes;

[0088] Step S133: Select specific biomarker primers for subsequent amplification primer design:

[0089] Design specific PCR primers for pollution source-related microorganisms using Primer-BLAST and Oligo 7.0;

[0090] Use a primer visualization tool to optimize the primer melting temperature and secondary structure to avoid non-specific amplification; ensure that the primers can distinguish different pollution sources;

[0091] Step S134: Extract DNA from the water sample using a DNA extraction kit and store the DNA at low temperature. For DNA extraction: Efficiently recover trace environmental DNA using the magnetic bead method or column extraction method; Use a reagent to remove PCR inhibitors in the water sample;

[0092] Step S135: Use qPCR to amplify specific DNA fragments; Perform environmental DNA sequencing and quantitative analysis on the amplified samples.

[0093] Step S2: Environmental DNA extraction, specific primer amplification, and DNA sequencing quantification;

[0094] Step S21: Provide a multi-dimensional dataset for model training; Collect environmental DNA samples from different pollution sources, extract DNA data based on qPCR, digital PCR, and high-throughput sequencing technologies, and measure species abundance; Use water samples from different pollution sources to prepare mixed water samples with different concentrations; Measure the DNA and water quality data of each mixed water sample to generate a multi-dimensional dataset;

[0095] Combine water quality parameters to construct a pollution source characteristic dataset as a machine learning training sample;

[0096] Step S22: Establish a machine learning model to achieve pollution source analysis; Use a multi-model integrated learning framework for deep algorithm training, including random forest, XGBoost, and support vector machine;

[0097] Use K-fold cross-validation and Bayesian optimization to tune the model parameters to ensure high accuracy and generalization ability for pollution source analysis;

[0098] Step S221: Input: Environmental DNA data and water quality indicators;

[0099] Step S222: Output: Pollution source types and their proportions; including pollution source category classification, pollution source proportion calculation, and prediction of pollution source spatio-temporal distribution;

[0100] Step S223: Adopt a multi-model integrated learning framework, combine various machine learning or deep learning methods to improve the stability and accuracy of pollution source tracing.

[0101] Step S3: Construct and train a machine learning model using DNA data and water quality data;

[0102] Step S31: Integrate the characteristic indicators of environmental DNA sequence data and corresponding water quality data from different pollution sources; form a training dataset with multi-dimensional characteristics for creating a comprehensive description of pollution source characteristics and their corresponding molecular biological information;

[0103] Fuse environmental DNA data and water quality data; eliminate data noise through data preprocessing to improve data quality;

[0104] Adopt feature engineering, including data splicing or time series feature extraction, to construct a multi-modal dataset of pollutants-biological markers-water quality parameters for training the model;

[0105] Step S32: Achieve efficient pollution analysis through model training and validation; combine supervised learning and unsupervised learning methods, use deep neural networks to extract complex pollution characteristics, and at the same time use the K-means clustering algorithm for adaptive learning of pollutant categories; analyze the contribution degree of model features through Shapley value analysis and evaluate the importance of pollution factors;

[0106] By using K-means / GMM clustering for adaptive learning of pollutant categories, the accuracy of pollution classification can be improved;

[0107] Train the pollution source tracing system using random forest, XGBoost or deep neural network models, combine Bayesian optimization and K-fold cross-validation to further improve the generalization ability of the model to ensure prediction accuracy;

[0108] Use Precision, Recall, F1-score, MSE and R² metrics to evaluate the performance of the model; at the same time, introduce Shapley value analysis to quantify the contribution degree of pollution factors and calculate the proportion of pollution sources combined with Bayesian inference;

[0109] Finally, visualize the distribution and change trend of pollution sources through GIS heat maps and trend prediction curves, thereby improving the scientificity and accuracy of pollution control decisions. This system integrates supervised learning methods, can accurately predict the types and proportions of pollution sources, verify the model performance, and ensure its efficient and reliable pollution source analysis.

[0110] Step S4: Collect water samples in the actual polluted area, input the data into the model and output the types and proportions of pollution sources;

[0111] Step S41: Collect water samples in the actual polluted area;

[0112] Adopt the spatial sampling method, set multiple monitoring points in the polluted area, and deploy multiple points from upstream to midstream to downstream to ensure wide spatial coverage;

[0113] Use drones and automatic sampling equipment to sample at regular times or fixed points to avoid human interference;

[0114] Step S42: Measure water quality data and environmental DNA data;

[0115] Use fluorescence quantitative PCR to amplify specific DNA fragments and perform gene fingerprint identification on the pollution sources;

[0116] Adopt ultra-high resolution mass spectrometry to analyze water pollutants and combine nanopore sequencing to obtain DNA sequence information;

[0117] Step S43: Input the above data into the machine learning model to analyze the types and proportions of pollution sources; Combine multi-modal data fusion input into the model, including environmental DNA, chemical pollutants, and water quality parameters, and analyze the contribution rate of pollution sources through the model;

[0118] Adopt probability prediction to calculate the proportion of pollution sources and generate a pollution source tracing report to provide a scientific basis for pollution control.

[0119] Step S431: Data preprocessing and fusion: Clean, standardize, and normalize water quality data, environmental DNA data, and mass spectrometry data; Through data fusion technology, integrate data from different data sources;

[0120] Feature engineering and model training: Extract features from the fused data, including pollutant concentration, abundance of biomarkers, and similarity of gene fingerprints, as model inputs; Select a suitable machine learning model for training according to the data characteristics;

[0121] Analysis of pollution source types: Through model learning, identify the characteristics of different pollution sources and classify the pollution sources, including agricultural pollution, industrial emissions, and domestic sewage; Calculate their proportions in the total pollution according to the contribution degrees of each pollution source;

[0122] Probability prediction and calculation of the proportion of pollution sources: Use probability prediction algorithms to estimate the contribution proportions of each pollution source in water pollution; Combine Bayesian inference or Markov chain Monte Carlo methods to calculate the probability distribution of each pollution source and generate a report on the proportion of pollution sources;

[0123] Generation of pollution source tracing report: Combine the contribution proportions of each pollution source and regional environmental data to generate a pollution source tracing report; The report should include pollution source types, proportions of pollution sources, and spatial distribution information of pollutants, providing a basis for formulating pollution control strategies.

[0124] Preferably, the DNA extraction is completed by a DNA extraction kit.

[0125] Furthermore, in this embodiment, the DNA extraction is completed by a DNA extraction kit.

[0126] The following is a detailed description in combination with specific embodiments:

[0127] Research on the source tracing of water body pollution taking a certain river in a certain city of Hainan Province as an example

[0128] 1. Background and problem description

[0129] In a certain city of Hainan Province, there are serious water pollution problems in a certain river. According to the feedback of local water quality monitoring personnel, there is illegal sewage discharge in this river, and the water quality often deteriorates sharply, showing black and odorous phenomena. Through research, the local suspected pollution sources are mainly concentrated in the aquaculture industry around the river. In particular, some farms directly discharge aquaculture wastewater into the river illegally in order to save production costs, resulting in aggravated water pollution. Specifically, the farms involved include one grouper farm, two tilapia farms and some other fish farms.

[0130] 2. Pollution source identification and sampling

[0131] According to the above research results, several typical farms were selected as the investigation objects, and water samples were collected. By analyzing the wastewater samples of these farms and combining water quality data and environmental DNA data, basic data were provided for subsequent pollution source tracing.

[0132] Farm selection:

[0133] Grouper farm: There has been a long-term situation of discharging aquaculture wastewater in this farm, which may be one of the main pollution sources.

[0134] Tilapia farms: The wastewater of the two farms is also suspected of being directly discharged into the river, and the pollutant concentration is relatively high.

[0135] Other fish farms: Due to cost savings, some small farms also have similar situations of illegally discharging wastewater.

[0136] 3. Sample collection and data analysis

[0137] In order to effectively identify the pollution sources and their contribution ratios, the following steps were implemented:

[0138] Water sample collection: By using the spatial distribution sampling method, water samples were collected at different positions of the river, especially near the farms, to ensure that the collected water samples could reflect the distribution of pollution sources. The sampling points included multiple monitoring points in the upper, middle and lower reaches to ensure full coverage of the pollution sources.

[0139] Water quality detection: Use portable water quality analyzers and laboratory equipment to comprehensively analyze the collected water samples, mainly detecting chemical parameters in the water (such as pH, dissolved oxygen, ammonia nitrogen, heavy metals, COD, etc.) and biological indicators (such as the concentration of microorganisms in the water body).

[0140] DNA extraction and analysis: Use a specialized DNA extraction kit to extract environmental DNA (hereinafter referred to as eDNA) from the water samples. Amplify specific DNA markers (such as 16S rRNA gene, COI gene, etc.) by fluorescence quantitative PCR to identify the microbial community and pollution source fingerprints in the water body, and analyze the biomarker of the pollution source. Combine high-resolution mass spectrometry technology and nanopore sequencing technology to further analyze the chemical pollutants and DNA sequences in the water samples.

[0141] 4. Data fusion and machine learning modeling

[0142] To improve the accuracy of pollution source identification, machine learning methods are used to combine multi-modal data for pollution source tracing analysis:

[0143] Data fusion: Combine eDNA data with water quality data, and use feature engineering to process various data, such as standardizing and normalizing data such as pollutant concentration, abundance of microbial markers, chemical composition, etc., to ensure the consistency of data input.

[0144] Model training: Based on the collected data, use machine learning models such as random forest, XGBoost or deep neural network for training. Further optimize the model through methods such as cross-validation and Bayesian optimization to improve the prediction accuracy. The goal of the model is to predict the type of pollution source and its proportion by inputting water quality and eDNA data.

[0145] Model performance verification: Test water samples with different concentration ratios, adjust the model parameters and verify to ensure that the accuracy of the machine learning model reaches more than 90%.

[0146] 5. Pollution source monitoring and tracing report

[0147] After the model is established, entrust local river water quality monitoring personnel to sample the river water quality regularly and long-term, and input the collected water quality and eDNA data into the trained machine learning model for pollution source identification and analysis.

[0148] Long-term sampling and monitoring: River water quality monitoring personnel sample the river water quality irregularly, with a total of 8 samplings. In several of these samplings, it is found that the proportion of pollution sources from certain farms exceeds 10%, and even in some cases, exceeds 15%.

[0149] Pollution source tracing:

[0150] Tilapia farm: In two samplings, the pollution proportion of a certain tilapia farm exceeded 10%, indicating that there may be illegal sewage discharge behavior.

[0151] Generate pollution source tracing report: Combine all sampling results and the output of the machine learning model to generate a pollution source tracing report, which details the types of pollution sources, pollution proportions, and their spatial distributions, providing a scientific basis to help decision-making departments control pollution sources.

[0152] 6. Innovation and application prospects

[0153] In this study, by combining environmental DNA technology with machine learning models, a multi-source pollution source tracing method is innovatively proposed. This method can not only adapt to various pollution sources but also efficiently and accurately identify pollution sources and quantify their contribution proportions, with high practical value. Through this method, real-time pollution source monitoring and tracing capabilities can be provided for water quality management departments, assisting in the formulation and implementation of pollution control strategies.

[0154] This embodiment has broad application prospects, especially suitable for fields such as water body pollution source tracing, environmental monitoring, and pollution prevention and control. Through cooperation with local governments, environmental protection agencies, etc., this technology can be promoted and applied nationwide, effectively improving the scientificity and accuracy of water quality management.

[0155] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by terms such as "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. are based on the orientation or positional relationships shown in the drawings. These are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention.

[0156] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0157] The above are only the preferred specific embodiments 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, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes, shall be covered by the protection scope of the present invention.

Claims

1. A water pollutant source tracing method based on environmental DNA and machine learning is used to analyze the characteristics of pollution sources from multiple angles and achieve accurate source tracing of pollutants, characterized in that: The steps include: Step S1: Regional pollutant source investigation and water body characteristics characterization; Step S11: Identify the types of potential pollution sources and obtain their raw sewage samples; determine the possible types of pollution sources through on-site investigation and historical data analysis, including industrial wastewater, agricultural runoff, and domestic sewage; Step S12: Obtain water quality characteristic data of the pollution source; The water samples were initially screened using a rapid water quality tester and spectrophotometry to obtain key pollution indicators, including pH, chemical oxygen demand, total phosphorus, total nitrogen, suspended solids, and fluorescence data; The raw sewage sample is analyzed by high performance liquid chromatography, gas chromatography-mass spectrometry and fluorescence spectroscopy to obtain its chemical composition characteristics; The microbial communities in the collected samples were sequenced by amplification of 16S rRNA and 18S rRNA to obtain a background database of environmental DNA; Step S13: Tracking specific biomarkers through environmental DNA technology; targeting specific pollution sources, including characteristic microorganisms of industrial pollutant emissions and bacterial communities of agricultural pollution sources, screening specific DNA markers, and amplifying and analyzing them using real-time fluorescence quantitative PCR and metagenomic sequencing; The step S13 specifically includes the following steps: Step S131: Classification of pollution sources: Industrial pollution: Screening includes microorganisms tolerant to specific metal ions; Agricultural pollution: Screening includes fecal indicator bacteria and pesticide degrading bacteria; Domestic sewage: screening includes human indicator bacteria and antibiotic resistance genes; Step S132: Database comparison: Compare with NCBI GenBank and Silva databases to screen environmental DNA biomarkers associated with known pollution sources; Using metagenomic analysis to identify specific genes, including heavy metal tolerance genes or antibiotic resistance genes; Step S133: Select specific biomarker primers for subsequent amplification primer design: Primer-BLAST and Oligo 7.0 were used to design specific PCR primers targeting microorganisms associated with pollution sources; Use primer visualization tools to optimize primer melting temperature and secondary structure to avoid nonspecific amplification; Ensure that primers can distinguish between different sources of contamination; Step S134: extracting DNA from the water sample using a DNA extraction kit and storing it at low temperature. DNA extraction: using a magnetic bead method or a column extraction method to efficiently recover trace amounts of environmental DNA; Use PCR inhibitor removal reagent to remove PCR inhibitors from water samples; Step S135: amplifying specific DNA fragments using qPCR; Perform environmental DNA sequencing and quantitative analysis on amplified samples; Step S2: environmental DNA extraction, specific primer amplification and DNA sequencing quantification; Step S21: providing a multidimensional data set for model training; collecting environmental DNA samples from different pollution sources, extracting DNA data based on qPCR, digital PCR and high-throughput sequencing technology, and measuring species abundance; Combined with water quality parameters, a pollution source feature dataset was constructed as a machine learning training sample; In step S21, water samples from different pollution sources are used to prepare mixed water samples of different concentrations; and DNA and water quality data of each mixed water sample are measured to generate a multidimensional data set; Step S22: Establish a machine learning model to achieve pollution source analysis; use a multi-model integrated learning framework to perform deep algorithm training, including random forest, XGBoost, and support vector machine; K-fold cross validation and Bayesian optimization are used to tune model parameters to ensure high accuracy and generalization ability of pollution source analysis; In step S22, the specific steps are as follows: Step S221: Input: environmental DNA data and water quality indicators; Step S222: Output: pollution source types and their proportions; including pollution source classification, pollution source proportion calculation and pollution source spatiotemporal distribution prediction; Step S223: adopting a multi-model integrated learning framework and combining multiple machine learning or deep learning methods to improve the stability and accuracy of pollution source tracing; Step S3: construct and train a machine learning model using DNA data and water quality data; Step S31: Integrate the environmental DNA sequence data of different pollution sources and the characteristic indicators of the corresponding water quality data to form a training data set with multi-dimensional characteristics, which is used to create a comprehensive description of the characteristics of the pollution source and its corresponding molecular biological information; In the step S31, the environmental DNA data is merged with the water quality data; data noise is eliminated through data preprocessing to improve data quality; Feature engineering, including data concatenation or time series feature extraction, is used to construct a multimodal dataset of pollutants, biomarkers, and water quality parameters for model training; Step S32: Use model training and verification to achieve efficient pollution analysis; combine supervised learning and unsupervised learning methods, deep neural network to extract complex pollution features, and use K-means clustering algorithm to perform adaptive learning of pollutant categories; The Shapley value analysis analyzes the contribution of model characteristics and evaluates the importance of pollution factors; In step S32, the accuracy of pollution classification can be improved by using K-means / GMM clustering to perform pollutant classification adaptive learning; Use random forest, XGBoost or deep neural network models to train the pollution source tracing system, combined with Bayesian optimization and K-fold cross validation to further improve the generalization ability of the model to ensure prediction accuracy; The performance of the model is evaluated using the indicators of Precision, Recall, F1-score, MSE and R². At the same time, the Shapley value analysis is introduced to quantify the contribution of pollution factors, and the proportion of pollution sources is calculated in combination with Bayesian inference. Finally, the distribution and changing trends of pollution sources can be visualized through GIS heat maps and trend prediction curves, thereby improving the scientificity and accuracy of pollution control decisions. The system integrates supervised learning methods to accurately predict the types of pollution sources and their proportions, verify model performance, and ensure that it can efficiently and reliably analyze pollution sources. Step S4: Collect water samples in the actual polluted area, input the data into the model and output the type and proportion of pollution sources; Step S41: collecting water samples in the actual polluted area; The spatial sampling method is used to set up multiple monitoring points in the polluted area, upstream-middle Multiple points are deployed upstream and downstream to ensure wide spatial coverage; Use drones or automatic sampling equipment to collect samples at fixed times or locations to avoid human interference; Step S42: measuring water quality data and environmental DNA data; Use fluorescent quantitative PCR to amplify specific DNA fragments and perform genetic fingerprint identification of pollution sources; Ultra-high resolution mass spectrometry is used to analyze water pollutants, combined with nanopore sequencing to obtain DNA sequence information; Step S43: Input the above data into the machine learning model to analyze the types of pollution sources and their proportions; combine the multimodal data fusion input model, including environmental DNA, chemical pollutants and water quality parameters, and analyze the contribution rate of pollution sources through the model; Probabilistic prediction is used to calculate the proportion of pollution sources, and a pollution source tracing report is generated to provide a scientific basis for pollution control.

2. According to claim 1, a method for tracing the source of water pollutants based on environmental DNA and machine learning is characterized in that: In step S43, the specific steps are as follows: Step S431: Data preprocessing and fusion: Clean, standardize and normalize water quality data, environmental DNA data and mass spectrometry data; integrate data from different data sources through data fusion technology; Feature engineering and model training: Extract features from the fused data, including pollutant concentrations, biomarker abundance, and similarity of genetic fingerprints, as model inputs; select appropriate machine learning models for training based on data characteristics; Pollution source type analysis: Through model learning, the characteristics of different pollution sources are identified and classified, including agricultural pollution, industrial emissions, and domestic sewage; according to the contribution of each pollution source, its proportion in the total pollution is calculated; Probability prediction and pollution source proportion calculation: Use probability prediction algorithms to estimate the contribution of each pollution source to water pollution; combine Bayesian inference or Markov chain Monte Carlo method to calculate the probability distribution of each pollution source and generate a pollution source proportion report; Generation of pollution source tracing report: Generate a pollution source tracing report based on the contribution ratio of each pollution source and regional environmental data; the report should include information on pollution source type, pollution source ratio, and spatial distribution of pollutants, providing a basis for the formulation of pollution control strategies.

Images