A digital analysis method for tracing water pollution based on off-site monitoring

Abstract

The present invention discloses a digital analysis method based on off-site monitoring of water pollution source tracing, comprising the following steps: S1, collecting water samples: collecting water samples at multiple predetermined sampling points in the target water area, S2, environmental DNA extraction: performing environmental DNA extraction operation on the collected water samples, S3, microbial gene detection: performing microbial gene detection on the extracted environmental DNA using high-throughput sequencing technology, S4, data analysis and model construction: performing data analysis based on multiple algorithms, S5, pollution source tracing determination: combining satellite remote sensing data, the water color, area changes, algae distribution, floating objects on the water surface and other information of the water area can be monitored.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality monitoring and pollution source tracing, and in particular to a digital analysis method based on off-site monitoring of water quality pollution source tracing. Background Art

[0002] With the acceleration of global industrialization and the continuous growth of population, water pollution and protection have become one of the major challenges facing the field of environmental science today. As a key link in water resource management and protection, the accuracy, timeliness and comprehensiveness of water quality monitoring technology are crucial for effectively evaluating water quality, tracing pollution sources and formulating reasonable pollution control strategies.

[0003] With the rapid development of molecular biology technology, environmental gene detection technology has begun to be applied to the field of water quality monitoring, providing a possibility to solve some problems of traditional methods. However, some existing analysis methods based on environmental gene detection data are not mature enough, lack systematicity and accuracy, and often cannot establish a stable and reliable quantitative relationship model between microbial genes and water quality parameters, resulting in difficulty in achieving high-precision calculation and real-time and continuous monitoring of water quality parameters such as DO and pH in practical applications. Some simple correlation analysis methods can only preliminarily determine that there is a certain correlation between microbial communities and water quality parameters, but cannot accurately quantify this relationship. When faced with complex and changeable water environments and mixed sources of pollution, it is difficult to meet the requirements of water quality monitoring for accuracy and reliability.

[0004] Based on this, the present application proposes a digital analysis method based on off-site monitoring of water pollution source tracing. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a digital analysis method based on off-site monitoring of water pollution source tracing.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A digital analysis method for tracing the source of water pollution based on off-site monitoring includes the following steps:

[0008] S1. Collecting water samples: Collect water samples at multiple predetermined sampling points in the target water area. The distribution of the predetermined sampling points is set according to the geographical characteristics of the water area, the direction of water flow and the distribution of surrounding potential pollution sources. The sampling frequency is determined according to the use function of the water area, historical pollution data and ecological sensitivity, ranging from once every hour to once every two weeks.

[0009] S2. Environmental DNA extraction: Perform environmental DNA extraction on the collected water samples using an optimized high-efficiency extraction kit or equipment that can effectively separate and enrich the environmental DNA of microorganisms in a complex water sample matrix and minimize interference from impurities.

[0010] S3. Microbial gene detection: High-throughput sequencing technology is used to perform microbial gene detection on the extracted environmental DNA. The detection range covers the gene sequences of various microbial groups such as bacteria, archaea, fungi, protozoa and algae. The sequencing depth is sufficient to accurately identify low-abundance microorganisms, and microbial gene sequences as low as the number of copies per milliliter of water sample can be detected.

[0011] S4. Data analysis and model building:

[0012] Based on the microbial community structure analysis algorithm: construct a model of the relationship between microbial community structure and water quality parameters, analyze the functional roles of different microbial groups in the ecosystem and their relationship with water quality parameters such as dissolved oxygen (DO) and pH value, and establish a mathematical model through learning from a large amount of sample data to infer water quality parameters based on microbial community structure data;

[0013] Based on the quantitative algorithm of specific gene markers: determine the specific gene markers closely related to DO and pH, such as genes encoding cytochrome oxidase and certain proton transporter genes, quantify the copy number of these gene markers in environmental DNA, and calculate them in combination with the pre-established relationship model between gene copy number and water quality parameters;

[0014] Based on machine learning multivariate regression algorithm: A variety of microbial gene information and some abiotic environmental factor data, such as water temperature and conductivity, are integrated, and a machine learning algorithm is used for multivariate regression analysis to train the model to learn the complex nonlinear relationship between microbial gene data, abiotic factors, DO and pH values.

[0015] S5. Pollution source determination: Combined with satellite remote sensing data, it is possible to monitor water color, area changes, algae distribution, floating objects on the water surface and other information; hydrological and meteorological data, including water level, flow rate, precipitation, water temperature, light, wind speed and other information, as well as historical water quality data. Through the established digital analysis model, the microbial genetic data, water quality parameter changes, satellite remote sensing data and hydrological and meteorological data are comprehensively analyzed to determine the source of pollution, pollution type, pollution diffusion path and scope.

[0016] Preferably, in step S3, the high-throughput sequencing technology is one or more combinations of Illumina sequencing platform, PacBio sequencing platform or Nanopore sequencing platform, and the sequencing read length meets the requirements for complete detection of key microbial gene sequences.

[0017] Preferably, in step S4, the construction of the relationship model between the microbial community structure and water quality parameters adopts multivariate statistical analysis methods such as principal component analysis, redundancy analysis or canonical correspondence analysis to improve the model's ability to analyze the relationship between complex microbial communities and water quality.

[0018] Preferably, in step S4, the quantitative method in the specific gene marker quantitative algorithm adopts real-time fluorescence quantitative PCR (qPCR) or digital PCR (dPCR) technology to ensure the accuracy and precision of gene copy number quantification.

[0019] Preferably, in step S5, the digital analytical model can distinguish different pollution types such as industrial pollution, agricultural non-point source pollution, domestic sewage pollution and natural geological pollution when determining the pollution source, and can analyze the contribution ratio of each pollution type to the mixed pollution.

[0020] Preferably, in step S5, the method also includes establishing a visualization platform to display information such as changes in water quality parameters, changes in microbial community structure, pollution tracing results, satellite remote sensing images, hydrological and meteorological data, etc. in the form of intuitive charts, maps or animations, so as to facilitate users to understand the water quality and pollution conditions of the water area in real time. The platform has functions such as data query, statistical analysis, and report generation.

[0021] The present invention has the following beneficial effects:

[0022] 1. Through data fusion and cross-validation, and by combining the advantages of different algorithms, the DO value error rate is controlled within a certain range, which is significantly lower than that of the traditional single algorithm or without fusion. It can more accurately reflect the actual dissolved oxygen and pH conditions of the water area, and provide a reliable data basis for water quality assessment.

[0023] 2. Through algorithm complementarity and dynamic switching mechanism, the algorithm can be flexibly selected according to the water quality characteristics and changes in each region. Whether it is the upstream stable area, the midstream complex area or the downstream sudden pollution situation, it can effectively respond to ensure accurate monitoring of water quality under various conditions, and ensure the ecological stability of the water area and the safety of the surrounding areas.

[0024] 3. Through the expansion of the microbial gene library and the coordination of algorithm optimization, the algorithm is continuously optimized during the long-term monitoring process. Not only the error rate of each algorithm is greatly reduced, but also the ability to warn of abnormal changes in water quality about a week in advance is also available, which is conducive to the formulation of response strategies in advance and reduces the negative impact of water quality deterioration, which is of great significance to the sustainable management of water resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 The pH values ​​calculated by different algorithms in Example 1 of the present invention and the fused line graph;

[0026] Figure 2 The DO values ​​calculated by different algorithms in Example 1 of the present invention and the fused line graph are shown. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0028] A digital analysis method for tracing the source of water pollution based on off-site monitoring includes the following steps:

[0029] S1. Collecting water samples: Collect water samples at multiple predetermined sampling points in the target water area. The distribution of the predetermined sampling points is set according to the geographical characteristics of the water area, the direction of water flow and the distribution of surrounding potential pollution sources. The sampling frequency is determined according to the use function of the water area, historical pollution data and ecological sensitivity, ranging from once every hour to once every two weeks.

[0030] S2. Environmental DNA extraction: Perform environmental DNA extraction on the collected water samples using an optimized high-efficiency extraction kit or equipment that can effectively separate and enrich the environmental DNA of microorganisms in a complex water sample matrix and minimize interference from impurities.

[0031] S3. Microbial gene detection: Use high-throughput sequencing technology to perform microbial gene detection on the extracted environmental DNA. The high-throughput sequencing technology is one or more combinations of Illumina sequencing platform, PacBio sequencing platform or Nanopore sequencing platform, and the sequencing read length meets the requirements for complete detection of key microbial gene sequences. The detection range covers the gene sequences of various microbial groups such as bacteria, archaea, fungi, protozoa and algae. The sequencing depth is sufficient to accurately identify low-abundance microorganisms, and microbial gene sequences as low as the number of copies per milliliter of water sample can be detected.

[0032] S4. Data analysis and model building:

[0033] Based on the microbial community structure analysis algorithm: construct a model of the relationship between microbial community structure and water quality parameters. The construction of the model of the relationship between microbial community structure and water quality parameters adopts multivariate statistical analysis methods such as principal component analysis, redundancy analysis or canonical correspondence analysis to improve the model's ability to analyze the relationship between complex microbial communities and water quality. Analyze the functional roles of different microbial groups in the ecosystem and their association with water quality parameters such as dissolved oxygen (DO) and pH value. Establish a mathematical model through learning from a large amount of sample data to infer water quality parameters based on microbial community structure data;

[0034] Based on the specific gene marker quantitative algorithm: determine the specific gene markers closely related to DO and pH, such as genes encoding cytochrome oxidase and certain proton transporter genes, quantify the copy number of these gene markers in environmental DNA, and calculate them in combination with the pre-established model of the relationship between gene copy number and water quality parameters. It should be noted that the quantitative method in the specific gene marker quantitative algorithm uses real-time fluorescence quantitative PCR (qPCR) or digital PCR (dPCR) technology, which can ensure the accuracy and precision of gene copy number quantification;

[0035] Based on machine learning multivariate regression algorithm: Comprehensive information of various microbial genes and some abiotic environmental factors data, such as water temperature, conductivity, etc., use machine learning algorithm for multivariate regression analysis, train the model to learn the complex nonlinear relationship between microbial gene data, abiotic factors and DO and pH values. It should be noted that the machine learning model used by the machine learning multivariate regression algorithm is random forest, support vector machine or neural network, etc. The model performance can be optimized through model evaluation indicators such as root mean square error, determination coefficient, etc.

[0036] S5. Pollution source determination: Combined with satellite remote sensing data, it is possible to monitor water color, area changes, algae distribution, floating objects on the water surface and other information of the water area; hydrological and meteorological data, including water level, flow rate, precipitation, water temperature, light, wind speed and other information, as well as historical water quality data. Through the established digital analysis model, the microbial gene data, water quality parameter changes, satellite remote sensing data and hydrological and meteorological data are comprehensively analyzed to determine the pollution source, pollution type, pollution diffusion path and scope. It should be noted that when determining the pollution source through the digital analysis model, different pollution types such as industrial pollution, agricultural non-point source pollution, domestic sewage pollution and natural geological pollution can be distinguished, and the contribution ratio of each pollution type to the mixed pollution can be analyzed. It also includes the establishment of a visualization platform to display water quality parameter changes, microbial community structure changes, pollution source tracing results, satellite remote sensing images, hydrological and meteorological data and other information in the form of intuitive charts, maps or animations, so that users can understand the water quality and pollution situation of the water area in real time. The platform has functions such as data query, statistical analysis and report generation.

[0037] Embodiment 1:

[0038] Step 1: Collect water samples:

[0039] In the estuarine wetland area of ​​about 15 square kilometers, six sampling points are evenly set according to the geographical characteristics of the water area (such as estuary shape, water depth distribution, etc.), water flow direction (changes in water flow direction and velocity affected by tides) and the distribution of surrounding potential pollution sources (such as nearby factories, farmlands, residential areas, etc.). The sampling frequency is set to once every three days, and a specially designed automatic sampling device is used to collect water samples at a predetermined time. The water samples are properly stored in a low temperature environment (such as 4°C) to prevent changes in microbial activity and DNA degradation, ensuring that the collected water samples can accurately reflect the water quality microbial status of the area.

[0040] Step 2: Environmental DNA extraction:

[0041] The collected water samples were processed using an optimized environmental DNA extraction kit. The kit uses a special cell lysis solution formula that can efficiently destroy the cell walls of various microorganisms and release the environmental DNA inside the cells. Impurities are removed through centrifugation, filtration and other steps, and then the environmental DNA is enriched and purified using a specific adsorption column. The extraction efficiency has been verified to be over 90%, that is, from a water sample that theoretically contains 100ng of environmental DNA, more than 90ng of high-quality environmental DNA can actually be extracted, and the content of impurities (such as proteins, polysaccharides, etc.) is less than 5%, to ensure the accuracy and reliability of subsequent testing.

[0042] Step 3: Microbial gene testing:

[0043] The extracted environmental DNA was sequenced using the Illumina high-throughput sequencing platform, and the sequencing read length was set to 300bp to ensure that the key gene sequences of most water-related microorganisms were fully covered. The data quality score Q30 reached more than 95%, that is, among the sequenced bases, bases with a quality value greater than 30 (indicating an error rate of less than 0.1%) accounted for more than 95% of the total sequenced bases, thereby ensuring the accuracy and reliability of the sequencing data. Through bioinformatics analysis of the sequencing data, gene sequences of various microbial groups such as bacteria, archaea, fungi, protozoa, and algae were detected, and low-abundance microorganisms could be identified, and microbial gene sequences as low as 10 copies per milliliter of water sample could be effectively detected.

[0044] Step 4: Data analysis and model building:

[0045] Based on microbial community structure analysis algorithm

[0046] Construct a relationship model between microbial community structure and water quality parameters: First, the microbial community structure data was sorted and standardized, and the principal component analysis (PCA) method was used to extract the main characteristic information of the microbial community structure. Then, a linear regression model between the main characteristic information of the microbial community and water quality parameters such as dissolved oxygen (DO) and pH value was established through a large amount of historical sample data (at least 50 samples). Assume that the main characteristic information of the microbial community is represented by the vector X = [x1, x2…, x n ], DO value is expressed by y DO The pH value is represented by y pH The established linear regression model is:

[0047] y DO =β DO0 +β DO1 x1+β DO2 x2+…+β DOn x n +∈ DO

[0048] y pH =β pH0 +β pH1 x1+β pH2 x2+…+β pHn x n +∈ pH

[0049] Among them, β DOi and β pHj (ij=0,1…,n) is the regression coefficient, which is determined by the least squares fitting method; ∈ DO and ∈ pH At the sampling point located in the center of the wetland, the DO value was 5.6 mg / L and the pH value was 7.3 based on the model.

[0050] Based on specific gene marker quantification algorithms:

[0051] Identify specific gene markers that are closely related to DO and pH, such as the gene encoding cytochrome oxidase (hypothesized to be Gene DO ) and some proton transporter genes (hypothesized to be Gene pH). Real-time fluorescence quantitative PCR (qPCR) technology is used to quantify the copy number of these gene markers in environmental DNA. First, a standard curve is made. A plasmid containing the target gene with a known concentration is used as a standard. After serial dilution, a qPCR reaction is performed. The standard curve is drawn with the cycle threshold (Ct value) as the vertical axis and the logarithm of the standard copy number as the horizontal axis. For the environmental DNA in the water sample, the qPCR reaction is also performed, and the copy number of the target gene is calculated from the standard curve based on the obtained Ct value. Then, the calculation is performed in combination with the pre-established model of the relationship between gene copy number and water quality parameters. Assuming that Gene DO The copy number is C DO , Gene pH The copy number is C pH , the established relationship model is:

[0052] y DO =a DO +b DO In(C DO )+δ DO

[0053] y pH =a pH +b pH In(C pH )+δ pH

[0054] Among them, a DO 、b DO 、a pH 、b pH is the model parameter, obtained by fitting a large amount of experimental data; δ DO and δ pH At the central sampling point of the wetland, the DO value was 5.4 mg / L and the pH value was 7.2 based on the algorithm.

[0055] Based on machine learning multivariate regression algorithm:

[0056] The genetic information of various microorganisms and some abiotic environmental factors data, such as water temperature (T) and electrical conductivity (EC), were integrated. The random forest (RandomForest) machine learning algorithm was used for multivariate regression analysis. First, the microbial gene information was feature encoded and digitized, for example, the abundance of different microbial genes was converted into numerical feature vectors. Then, these feature vectors were used together with abiotic environmental factors data such as water temperature and electrical conductivity as input variables, and DO value and pH value as output variables to construct a random forest regression model. By training a large amount of training data (at least 80% of historical data), the model parameters (such as the number of decision trees, tree depth, etc.) were optimized to minimize the root mean square error (RMSE) and maximize the coefficient of determination (R 2). At the central sampling point of the wetland, the DO value was calculated to be 5.8 mg / L and the pH value was 7.4 based on the model.

[0057] Based on historical data verification, weights are set for the three algorithms. A dynamic weighting method based on model performance evaluation is used to calculate the average error rate (ER) of each algorithm in the past 20 sampling data. alg1 , ER alg2 , ER alg3 ) and error stability indicators (such as the standard deviation SD of the error gl1 , SD gl2 , SD gl3 ). The weight calculation formula is:

[0058]

[0059] Among them, α is the weight adjustment coefficient, which is set to 0.5 according to the actual data. After calculation, the weight ω of the microbial community structure analysis algorithm alg1 =0.3, weight of specific gene marker quantitative algorithm ω alg2 =0.3, weight of machine learning multivariate regression algorithm ω alg3 =0.4. After weighted calculation, the final DO value is:

[0060] Y final =ω alg1 yDO1+ω alg1 yDO2+ω alg3 yDO3=5.6*0.3+5.4*0.3+5.8*0.4=5.62mg / L

[0061] y pHfinal =ω alg1 y pH1 +ω alg2 y pH2 +ω alg3 y pH3 =7.1*0.3+7.2*0.3+7.4*0.4=7.31

[0062] Compared with the actual measured value of high-precision water quality testing instruments, the DO value error rate calculation formula is:

[0063]

[0064] The pH error rate calculation formula is:

[0065]

[0066] Among them, yDO a ctual and ypH actual is the actual value measured by the high-precision water quality testing instrument. After calculation, the DO value error rate is ±2.8%, and the pH value error rate is ±2.0%. Before data fusion, the average error rates of the three algorithms were ±4.5%, ±5.0% and ±3.5% respectively.

[0067] Embodiment 2:

[0068] Step 1: Collect water samples:

[0069] For the 12-kilometer-long inland river that runs through the city, it is divided into three sections: upstream (0-4 kilometers), midstream (4-8 kilometers) and downstream (8-12 kilometers). Two sampling points are set up in each section based on the width of the river, changes in water depth and the distribution of surrounding pollution sources. The upstream sampling point is set near the natural river bank where the water flow is relatively stable, the midstream sampling point is located near urban drainage outlets and bridges and other areas that may affect water quality, and the downstream sampling point is close to the river estuary or the intersection with other water bodies. The sampling frequency is once every 5 days for the upstream, once every 3 days for the midstream, and once a day for the downstream. Water samples are collected using an automatic sampling device and stored at low temperatures.

[0070] Step 2: Environmental DNA extraction and microbial gene detection (same as Example 1)

[0071] The same optimized environmental DNA extraction kit and Illumina high-throughput sequencing platform were used to ensure extraction efficiency and sequencing quality, detect gene sequences of multiple microbial groups and detect low-abundance microbial genes.

[0072] Step 3: Data analysis and model building

[0073] Specific gene markers related to oxygen utilization and acid-base balance were determined, and the gene copy number was quantified using real-time fluorescence quantitative PCR (qPCR) technology and a relationship model was established (same as in Example 1). At the sampling point at 2 km, the DO value was quickly calculated to be 6.2 mg / L and the pH value was 7.5 by quantifying the relevant gene markers. The error rate compared with the actual measurement was about ±3.5%, and the error rate calculation formula was:

[0074]

[0075] The random forest machine learning algorithm was used to construct a multivariate regression model (same as in Example 1) by combining microbial gene information with abiotic environmental factors such as water temperature and conductivity. At the sampling point at 6 kilometers, the DO value was calculated to be 5.0 mg / L and the pH value was 7.1. The error rate was about ±4.0%, and the calculation method was the same as above.

[0076] A model of the relationship between microbial community structure and water quality parameters was constructed (same as in Example 1). When it was detected that the microbial community structure had changed dramatically due to a sudden sewage leak from a factory downstream, the algorithm was automatically switched to. At the sampling point at 10 kilometers, the impact of the new microbial community composition change on the water quality was identified in time, and the DO value was calculated to be 4.0 mg / L and the pH value to be 6.8. The error rate was ±5.0%, and the calculation method was the same as above.

[0077] Embodiment three:

[0078] Step 1: Collect water samples

[0079] In a lake with an area of ​​30 square kilometers, eight sampling points were set up based on the lake shape, depth distribution, location of rivers entering the lake, and surrounding land use. The sampling points cover different locations in the center of the lake, the lakeshore, and areas near the main inlets and outlets. The sampling frequency is set to once a week, and water samples are collected using a special sampling ship or buoy sampling device. After preliminary treatment on site, they are stored at low temperatures and transported to the laboratory for subsequent analysis.

[0080] Step 2: Environmental DNA extraction and microbial gene detection (same as Example 1)

[0081] The same optimized environmental DNA extraction kit and Illumina high-throughput sequencing platform were used to ensure extraction efficiency and sequencing quality, detect gene sequences of multiple microbial groups and detect low-abundance microbial genes.

[0082] Step 3: Data analysis and model building

[0083] Based on microbial community structure analysis algorithm (initial stage):

[0084] The DO and pH values ​​were estimated using a microbial community structure analysis algorithm, a specific gene marker quantification algorithm, and a machine learning multivariate regression algorithm, and the model was established in the same manner as in Example 1. The initial error rate was found to be between ±4.0% and ±5.5%, and the calculation method was the same as in Example 1.

[0085] The new gene information is fed back into three algorithms for optimization. In the algorithm based on microbial community structure analysis, the functional role of the microbial group to which the new gene belongs in the ecosystem is re-analyzed to improve the association model with water quality parameters; based on the specific gene marker quantitative algorithm, new quantitative markers are added, and the relationship model between gene copy number and water quality parameters is re-fitted; based on the machine learning multivariate regression algorithm, the new gene is input as a new variable for model training, and the model parameters are re-optimized to minimize the root mean square error (RMSE) and maximize the determination coefficient.

[0086] At the sampling point in the center of the lake, the error rate of DO value obtained based on the microbial community structure analysis algorithm was reduced to ±2.5%, and the error rate of pH value was ±1.8%; the error rate of DO value based on the specific gene marker quantitative algorithm was ±3.0%, and the error rate of pH value was ±2.2%; the error rate of DO value calculated based on the machine learning multivariate regression algorithm was ±2.2%, and the error rate of pH value was ±1.5%. The accuracy and stability of overall water quality monitoring have been significantly improved. The early warning capability assessment is determined by comparing the detection time difference of abnormal changes in water quality parameters before and after optimization. It can warn of abnormal changes in DO and pH values ​​about a week in advance, such as by monitoring the trend of changes in the abundance of microbial gene sequences, the deviation change between the predicted value and the actual value of the water quality parameter model, and other indicators. When these indicators exceed the set early warning threshold, a warning signal is issued. The early warning threshold is set according to historical data and actual needs, and has been verified and adjusted many times.

[0087] The microbial gene library used for testing is continuously expanded, with sources including the accumulation of long-term monitoring data of the lake and lakes with similar ecological environments, data mining of public microbial gene databases, and gene sequencing analysis of newly discovered microorganisms. For example, specific microbial genes found in areas with high temperatures and high nutrient concentrations are included in the gene library. The number of microbial species covered by the expanded gene library has increased by more than 30% compared to the initial number. Assuming that the gene library covered 100 microbial genes before the expansion and 100 microbial genes after the expansion, the proportion of gene species expansion is

[0088] Initially, three algorithms were used to estimate DO and pH values, and the error rate was found to be between ±4.0% and ±5.5%. As monitoring progressed, the microbial gene library used for testing was continuously expanded, and some microbial genes related to DO and pH values ​​under special water conditions were newly included, such as specific microbial genes found in high temperature and high nutrient concentration areas.

[0089] The new gene information was fed back into three algorithms. In the microbial community structure analysis algorithm, the association model between microbial functional groups and water quality parameters was improved; new quantitative markers were added based on the specific gene marker quantitative algorithm; and the new genes were used as new variables for model training based on the machine learning multivariate regression algorithm.

[0090] After half a year of collaborative work on gene library expansion and algorithm optimization, at the central sampling point of the lake, the DO value error rate based on the microbial community structure analysis algorithm was reduced to ±2.5%, and the pH value error rate was ±1.8%; the DO value error rate based on the specific gene marker quantitative algorithm was ±3.0%, and the pH value error rate was ±2.2%; the DO value error rate calculated based on the machine learning multivariate regression algorithm was ±2.2%, and the pH value error rate was ±1.5%. The accuracy and stability of overall water quality monitoring have been significantly improved. The early warning capability assessment is determined by comparing the detection time difference of abnormal changes in water quality parameters before and after optimization. It can warn of abnormal changes in DO and pH values ​​about a week in advance, such as by monitoring the trend of changes in the abundance of microbial gene sequences, the deviation change between the predicted value and the actual value of the water quality parameter model, and other indicators. When these indicators exceed the set early warning threshold, a warning signal is issued. The early warning threshold is set according to historical data and actual needs, and has been verified and adjusted many times.

[0091] It should be noted that, in the comparative examples, comparative example 1 shows that when the traditional single microorganism detection algorithm is used, the error of the estimation of DO and pH value is significantly higher than that of the improved algorithm. Comparative example 2 shows that when the conventional chemical detection estimation is used, its error rate is significantly higher than that of the example based on the microbial gene detection-related algorithm. The performance detection parameters of each group of embodiments and comparative examples are shown in Table 1:

[0092] Table 1: DO and pH values ​​and error rates obtained in various embodiments and comparative examples

[0093]

[0094]

[0095] In summary, the digital analysis method for tracing the source of water pollution based on off-site monitoring proposed in the present invention has shown excellent performance and significant advantages in multiple different embodiments. Through data fusion and cross-validation, the results of multiple environmental gene detection algorithms are integrated, and the target waters are sampled at intervals of three days within a one-month period. Figure 1 As shown in the figure, it can be seen that the pH value after data fusion (green solid line) fluctuates relatively less than the pH value of a single algorithm. For example, on the 18th day, the pH value based on the microbial community structure analysis algorithm reached a peak of about 7.7, while the pH value after data fusion fluctuated significantly less at this time, about 7.2. According to the data in Example 1, the error rate of the pH value after data fusion was ±2.0%, while before data fusion, the average error rates of the three algorithms were ±4.5%, ±5.0% and ±3.5%, respectively.

[0096] Specifically, Figure 2As shown, in the same one-month cycle, the target waters were sampled at intervals of three days. On the 18th day, the DO value based on the microbial community structure analysis algorithm reached a peak of about 6.2 mg / L, while the DO value after data fusion fluctuated significantly less at this moment, about 5.62 mg / L. According to the data in Example 1, the error rate of the DO value after data fusion was ±2.8%, while before data fusion, the average error rates of the three algorithms were ±4.5%, ±5.0% and ±3.5%, respectively. Both of the above experiments show that data fusion plays a significant role in reducing errors. Compared with the use of each algorithm alone before fusion, the error rate is significantly reduced, making the water quality parameter estimation more accurate, and providing a reliable basis for accurately grasping the dissolved oxygen and pH conditions of the waters.

[0097] Furthermore, in terms of algorithm complementarity and dynamic switching, appropriate algorithms are flexibly selected according to the water quality characteristics and changes of different river sections. In the stable upstream area, a specific gene marker quantitative algorithm is used to quickly obtain more accurate results. In the complex midstream area, a machine learning multivariate regression algorithm is used to deal with complex relationships. In the event of sudden pollution in the downstream, the algorithm is switched to the microbial community structure analysis algorithm for timely response. This dynamic switching mechanism can not only adapt to the needs of different scenarios, but also greatly reduce the overall average error rate. Compared with a single algorithm fixedly applied to the entire river section, it improves the detection efficiency and accuracy, and can respond quickly in the face of sudden pollution, buying precious time for timely response measures, and effectively ensuring the stability of the aquatic ecosystem and the safety of surrounding human activities.

[0098] The synergistic effect of the expansion of the microbial gene library and the optimization of the algorithm is fully reflected in the long-term monitoring example 3. As the monitoring process progresses, new microbial gene information is continuously incorporated and the algorithm is optimized. The error rate of each algorithm has dropped significantly within half a year. The DO value error rate can be reduced to ±2.2%, and the pH value error rate can reach ±1.5%. At the same time, it has the ability to warn of abnormal changes in water quality parameters about a week in advance. This not only improves the accuracy and stability of current water quality monitoring, but also provides valuable predictive information for long-term water quality management and ecological protection, which helps to plan and implement protection measures in advance, reduce the risk of water quality deterioration and the economic and ecological losses it may bring.

[0099] In addition, compared with traditional single microbial detection algorithms and conventional chemical detection estimation methods, the error rate of the present invention is greatly reduced. Due to its limitations, traditional methods are difficult to achieve the accuracy of the present invention in DO and pH value detection, and lack sensitivity and early warning capabilities for water quality changes. The present invention uses advanced environmental gene detection technology combined with innovative algorithm strategies to monitor water quality comprehensively, accurately and dynamically, providing strong technical support for water resource protection, ecosystem maintenance, and the harmonious coexistence of human activities and aquatic environments. It has significant innovation and superiority in the field of water quality monitoring, and has broad application prospects and great practical significance.

[0100] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A digital analysis method based on off-site monitoring of water pollution source tracing, characterized in that: The following steps are involved: S1. Collecting water samples: Collect water samples at multiple predetermined sampling points in the target water area. The distribution of the predetermined sampling points is set according to the geographical characteristics of the water area, the direction of water flow and the distribution of surrounding potential pollution sources. The sampling frequency is determined according to the use function of the water area, historical pollution data and ecological sensitivity, ranging from once every hour to once every two weeks. S2. Environmental DNA extraction: Perform environmental DNA extraction on the collected water samples using an optimized high-efficiency extraction kit or equipment that can effectively separate and enrich the environmental DNA of microorganisms in a complex water sample matrix and minimize interference from impurities. S3. Microbial gene detection: High-throughput sequencing technology is used to perform microbial gene detection on the extracted environmental DNA. The detection range covers the gene sequences of various microbial groups such as bacteria, archaea, fungi, protozoa and algae. The sequencing depth is sufficient to accurately identify low-abundance microorganisms, and microbial gene sequences as low as the number of copies per milliliter of water sample can be detected. S4. Data analysis and model building: Based on the microbial community structure analysis algorithm: construct a model of the relationship between microbial community structure and water quality parameters, analyze the functional roles of different microbial groups in the ecosystem and their relationship with water quality parameters such as dissolved oxygen (DO) and pH value, and establish a mathematical model through learning from a large amount of sample data to infer water quality parameters based on microbial community structure data; Based on the quantitative algorithm of specific gene markers: determine the specific gene markers closely related to DO and pH, such as genes encoding cytochrome oxidase and certain proton transporter genes, quantify the copy number of these gene markers in environmental DNA, and calculate them in combination with the pre-established relationship model between gene copy number and water quality parameters; Based on machine learning multivariate regression algorithm: A variety of microbial gene information and some abiotic environmental factor data, such as water temperature and conductivity, are integrated, and a machine learning algorithm is used for multivariate regression analysis to train the model to learn the complex nonlinear relationship between microbial gene data, abiotic factors, DO and pH values. S5. Pollution source determination: Combined with satellite remote sensing data, it is possible to monitor water color, area changes, algae distribution, floating objects on the water surface and other information; hydrological and meteorological data, including water level, flow rate, precipitation, water temperature, light, wind speed and other information, as well as historical water quality data. Through the established digital analysis model, the microbial genetic data, water quality parameter changes, satellite remote sensing data and hydrological and meteorological data are comprehensively analyzed to determine the source of pollution, pollution type, pollution diffusion path and scope.

2. According to claim 1, a digital analysis method based on off-site monitoring of water pollution source tracing is characterized in that: In step S3, the high-throughput sequencing technology is one or more combinations of Illumina sequencing platform, PacBio sequencing platform or Nanopore sequencing platform, and the sequencing read length meets the requirements for complete detection of key microbial gene sequences.

3. According to claim 1, a digital analysis method based on off-site monitoring of water pollution source tracing is characterized in that: In step S4, the construction of the relationship model between the microbial community structure and water quality parameters adopts multivariate statistical analysis methods such as principal component analysis, redundancy analysis or canonical correspondence analysis to improve the model's ability to analyze the relationship between complex microbial communities and water quality.

4. According to claim 1, a digital analysis method based on off-site monitoring of water pollution source tracing is characterized in that: In step S4, the quantitative method in the specific gene marker quantitative algorithm adopts real-time fluorescence quantitative PCR (qPCR) or digital PCR (dPCR) technology to ensure the accuracy and precision of gene copy number quantification.

5. According to claim 1, a digital analysis method based on off-site monitoring of water pollution source tracing is characterized in that: In step S4, the machine learning model adopted by the machine learning multivariate regression algorithm is a random forest, a support vector machine or a neural network, and the model performance is optimized through model evaluation indicators such as root mean square error, determination coefficient, etc.

6. According to claim 1, a digital analysis method based on off-site monitoring of water pollution source tracing is characterized in that: In step S5, the digital analytical model can distinguish different pollution types such as industrial pollution, agricultural non-point source pollution, domestic sewage pollution and natural geological pollution when determining the pollution source, and can analyze the contribution ratio of each pollution type to the mixed pollution.

7. According to claim 1, a digital analysis method based on off-site monitoring of water pollution source tracing is characterized in that: In step S5, the method also includes establishing a visualization platform to display information such as changes in water quality parameters, changes in microbial community structure, pollution tracing results, satellite remote sensing images, hydrological and meteorological data, etc. in the form of intuitive charts, maps or animations, so that users can understand the water quality and pollution conditions of the water area in real time. The platform has functions such as data query, statistical analysis, and report generation.

Images