Molecular fingerprint-based method for quantifying and tracing sources of lake dissolved organic matter and uses thereof

Abstract

The application provides a lake dissolved organic matter quantitative tracing method based on molecular fingerprint and application thereof, and relates to the technical field of environmental analysis and water ecological management. The method comprises the following steps: collecting end member samples such as lake water and sediments, surface soil, plant litter, chemical fertilizer, livestock and poultry manure, tail water of a sewage treatment plant, tail water of aquaculture, etc., and enriching DOM through solid phase extraction; then obtaining molecular formula data through mass spectrometry; screening stubborn inert molecules that satisfy H / C < 1.5, conversion times of 0 and exist only in a single end member, and verifying the ecological conservation thereof through -2 < betaNTI < 2 and -0.95 < RCBray < 0.95; adopting SHAP algorithm to optimize a high-contribution molecular feature subset, and inputting the high-contribution molecular feature subset into a MixSIAR model to output the contribution proportion of each end member and the 95% confidence interval. The method guarantees the reliability of the tracer from the mechanism, breaks through the bottleneck of spectral overlap, and realizes accurate and reliable quantitative disassembly of endogenous and exogenous DOM.

Description

Technical Field

[0001] This invention relates to the field of environmental analysis and water ecological management technology, and in particular to a method for quantitative tracing of dissolved organic matter in lakes based on molecular fingerprinting and its applications. Background Technology

[0002] Dissolved organic matter (DOM) is a key component of the carbon cycle in lakes. Accurately identifying the sources of DOM (such as release from endogenous sediments and input from external sources) is a prerequisite for developing scientific management strategies in lake management. Currently, the mainstream technical solutions for DOM source apportionment mainly rely on three-dimensional fluorescence spectroscopy (EEMs-PARAFAC) and stable isotope tracing techniques.

[0003] Figure 1 This is a schematic diagram illustrating the detection methods of mainstream existing DOM source parsing technologies.

[0004] See Figure 1 Three-dimensional fluorescence spectroscopy measures the fluorescence excitation-emission matrix of a water sample and uses parallel factor analysis (PARAFAC) to resolve different fluorescent components, thereby inferring the origin of DOM. Stable isotope techniques measure the ratio of carbon and nitrogen stable isotopes (such as δ¹³C, ... 5 N), which utilizes the differences in isotopic characteristics of organic matter from different sources and combines them with a mixture model for quantitative calculation.

[0005] However, when the aforementioned DOM source analysis method is applied to lakes affected by intense biogeochemical processes, the following significant drawbacks exist: 1. Overlapping fingerprints lead to resolution failure: In lakes with long residence times or strong human interference, the spectral features of DOMs from different sources highly overlap in fluorescence spectra, making it difficult for traditional spectroscopic methods to effectively distinguish between endogenous (such as sediment release) and exogenous (such as terrestrial input) contributions. 2. Existing spectral indicators or isotopic signals are easily modified by biogeochemical processes such as microbial degradation and photobleaching during environmental migration, which violates the "conservative mixing" assumption on which the source apportionment model relies, thus leading to serious biases in the source tracing results. 3. Lack of source-specific identification capability at the molecular level: Existing technologies are mostly based on macroscopic optical properties or element isotope ratios, which cannot accurately identify "conservative components" that truly have anti-degradation capabilities and can accurately record source information at the molecular level. Therefore, it is difficult to support refined endogenous-exogenous load dismantling and targeted treatment decisions.

[0006] Therefore, there is an urgent need for a technical method that can identify, verify, and quantify truly conserved DOM tracer components at the molecular scale to overcome the limitations of existing technologies in the complex environment of lakes.

[0007] In view of this, the present invention is hereby proposed. Summary of the Invention

[0008] The purpose of this invention is to provide a quantitative source tracing method for dissolved organic matter (DOM) in lakes based on molecular fingerprinting. This method integrates ultra-high resolution mass spectrometry, molecular transformation theory, ecological community construction theory and machine learning algorithms, thereby constructing a complete technical path from microscopic molecular identification to macroscopic quantitative analysis.

[0009] In order to achieve the above-mentioned objectives of the present invention, the following technical solution is adopted: According to one aspect of the present invention, a method for quantitative tracing of dissolved organic matter in lakes based on molecular fingerprinting is provided, comprising the following steps: (A): Collect samples of the lake water body to be tested and potential pollution source end-units in the watershed. The end-units include at least: sediments, topsoil, plant litter, fertilizers, livestock and poultry manure, sewage treatment plant effluent, aquaculture effluent, phytoplankton and phosphate rock. The aqueous or solid extracts of each endmember sample were filtered, and DOM was enriched by solid phase extraction (SPE) to obtain DOM samples enriched by each endmember. (B): Using the DOM samples enriched by each endmember in step (A) of mass spectrometry, obtain a dataset containing molecular formula and relative peak intensity; (C): Based on the molecular formula obtained in step (B), screen for those that satisfy: Molecules with H / C < 1.5, zero transformations in a molecular transformation network constructed based on measured lake water, and existing only in a single endmember; The above molecules are denoted as stubborn-inert (RI) molecules, and a fingerprint set of RI molecules specific to each endmember is established; (D): Calculate the β nearest neighbor index (βNTI) and the Raup-Crick index (RCBray) based on Bray-Curtis distance among lake water samples of the marker molecular fingerprint set obtained in step (C). When -2 < βNTI < 2 and -0.95 < RCBray < 0.95, the set of labeled molecules is confirmed to meet the conservation verification conditions. The criteria of -2 < βNTI < 2 and -0.95 < RCBray < 0.95 are based on the ecological null model theory: when |βNTI| is less than 2, it indicates that the community assembly is significantly biased towards randomness, while RCBray is less than 0.95, it indicates that the differences in observed composition are mainly driven by random drift rather than environmental filtering. The two together constitute a statistically sufficient condition for the conservatism verification. (E): Using the relative peak intensity, mass-to-nucleus ratio, nitrogen-to-carbon ratio, oxygen-to-carbon ratio, hydrogen-to-carbon ratio, modified aromaticity index, carbon oxidation state, molecular Gibbs free energy, Kendrick mass defect, and double bond equivalent molecular index of the marked molecules that have been confirmed by conservation, a machine learning classification model is constructed. The contribution of each molecular feature to the endmember classification is quantified by an interpretable algorithm, and the molecular indicators with the top two contributions are selected as the optimal traceability feature parameters. (F): Input the optimal source traceability feature parameters into the Bayesian mixture model, and after running, output the quantitative contribution ratio and confidence interval of each endmember to the DOM of the lake water body.

[0010] This invention provides a method for quantitative tracing of dissolved organic matter (DOC) in lakes based on molecular fingerprinting. The method involves collecting lake water samples and end-member samples from various potential pollution sources. After solid-phase extraction to enrich DOM (domestic organic matter), molecular formula data is obtained using Fourier transform ion cyclotron resonance mass spectrometry (FT-IMS). Resilient-inert (RI) molecules that meet the following criteria are screened: H / C < 1.5, have zero transformations in the molecular transformation network, and exist only in a single end-member. Their ecological conservation is then verified using βNTI and RCBray indices. A high-contribution molecule feature set is selected using the SHAP algorithm. The results are then input into a MixSIAR Bayesian mixture model to output the contribution ratio of each end-member and the 95% confidence interval. Compared with existing technologies, this invention has the following advantages: (a) Mechanism-driven conservatism safeguards: By introducing an ecological null model (βNTI and RCBray index) to empirically verify the selected RI molecule set, it was confirmed that its distribution is mainly dominated by random drift process, thus eliminating the interference of environmental filtering effects such as microbial degradation at the mechanism level, and ensuring the authenticity and reliability of the input parameters of the source apportionment model. (b) Improved molecular-level resolution: Fourier transform ion cyclotron resonance mass spectrometry (FT-ICRMS) was used to identify labeled molecules (resilient-inert RIs) that exist only in a single end member at the molecular level, overcoming the problem of "different substances with the same spectrum" caused by overlapping fluorescent components in three-dimensional fluorescence spectroscopy, and achieving accurate differentiation between endogenous sediment release and exogenous input (such as fertilizers, sewage treatment plant effluent, etc.). (c) Quantifiable decomposition of endogenous and exogenous contributions: By using a Bayesian mixture model (MixSIAR) to output the posterior distribution mean and 95% confidence interval of the contribution ratio of each endmember, not only are quantitative results provided, but the uncertainty is also simultaneously characterized, providing technical support with both accuracy and reliability for attribution and targeted management of lake DOM load.

[0011] To better illustrate this application, the applicant provides the following supplementary explanation by comparing it with existing technologies: (1) The conservation of tracers is empirically demonstrated through the ecological null model to ensure the accuracy of the traceability results: Existing technologies (such as three-dimensional fluorescence spectroscopy and stable isotope methods) all rely on the theoretical assumption that the tracer index remains "unchanged" during migration. However, in real environments, DOM components are easily modified by processes such as microbial degradation and photochemical transformation, leading to distortion of model input parameters.

[0012] This invention introduces an ecological null model to calculate the βNTI and RCBray indices of the RI molecular set across lake water samples. Using -2 < βNTI < 2 and -0.95 < RCBray < 0.95 as criteria, it confirms that the distribution of this set is primarily dominated by random drift (physical mixing) processes, thus excluding environmental filtering (such as selective biodegradation) interference at the mechanistic level. This verification step directly ensures the validity of the "conservative mixing" premise of the source apportionment model, providing a reliable basis for the endmember contribution ratios output by MixSIAR.

[0013] (2) Relying on molecular fingerprinting to achieve specific differentiation of pollution sources and overcome the bottleneck of fingerprint overlap: Existing spectroscopic techniques are limited by resolution and can only extract a few broad fluorescent components. DOMs from different sources (such as sediment humus and plant litter humus) often exhibit highly similar fluorescence characteristics, resulting in "different substances with the same spectrum" that are difficult to distinguish.

[0014] This invention utilizes the ultra-high quality resolution (≤100 ppb) of FT-ICRMS to screen from a massive number of molecules for RI molecules that exist only in a single endmember, satisfy H / C < 1.5, and have 0 transformations, thus constructing an endmember-specific molecular fingerprint set. This set contains hundreds of structurally well-defined molecular markers (such as C...). 23 H 14 O3, C 17 H8O4N2), whose source specificity far exceeds that of macroscopic optical signals, can effectively identify and distinguish endogenous and exogenous contributions with similar physicochemical properties.

[0015] (3) Achieve quantitative decomposition of endogenous and exogenous contributions and refined identification of anthropogenic sources: Because the old humus released from sediments is highly similar to terrestrial humus in terms of fluorescence and isotopic characteristics, traditional methods often misjudge endogenous release as exogenous input, leading to deviations in remediation strategies.

[0016] This invention utilizes the dual constraints of end-member uniqueness and conservatism in RI molecular fingerprinting to independently quantify the contribution of endogenous release from sediments. Simultaneously, each exogenous end-member (fertilizer, wastewater treatment plant effluent, livestock manure, etc.) possesses its own unique subset of RI molecules. By combining SHAP algorithm contribution ranking with MixSIAR modeling, the contribution ratio and 95% confidence interval of each end-member can be output. This capability supports a shift in lake DOM load attribution from extensive "external pollution interception" to targeted "endogenous control" and "multi-source collaborative governance."

[0017] In a preferred embodiment of the present invention, the sediment in the end-member of step (A) is an endogenous release source in the lake.

[0018] In a preferred embodiment of the present invention, the mass spectrometry detected in step (B) is Fourier transform ion cyclotron resonance mass spectrometry (FT-ICRMS).

[0019] As a preferred embodiment, the present invention employs Fourier transform ion cyclotron resonance mass spectrometry (FT-ICRMS) for detection. Utilizing its ultra-high mass resolution (≤100 ppb) and precise molecular formula identification capability, it achieves accurate differentiation and elemental composition determination of thousands of compounds in complex DOM components, providing the necessary molecular-level data foundation for subsequent screening of stubborn-inert (RI) molecules that meet the requirements of H / C < 1.5 and end-member uniqueness.

[0020] In a preferred embodiment of the present invention, the molecular transformation network in step (C) is constructed based on the DOM molecular formula data obtained in step (B). Potential transformation relationships between molecules are identified by matching the mass difference between molecular formulas with the theoretical mass difference in a preset molecular transformation database.

[0021] In a preferred embodiment of the present invention, the interpretability algorithm in step (E) is the SHAP algorithm.

[0022] As a preferred implementation method, this application uses the SHAP algorithm to quantify the contribution of each RI molecular feature to endmember classification, and selects the two molecular features with the highest contribution as the optimal source tracing feature parameters. This ensures that the selected features not only have statistical discriminative power, but also reflect the true causal contribution of each molecule in the model decision-making, thereby improving the feature set's ability to distinguish similar exogenous sources (such as chemical fertilizers and livestock manure) and its biological interpretability.

[0023] In a preferred embodiment of the present invention, the Bayesian mixture model in step (F) is a MixSIAR model; As a preferred implementation, this application employs the MixSIAR Bayesian mixture model to output the posterior distribution of the contribution ratio of each endmember and the 95% confidence interval, thereby characterizing the inherent uncertainty of the environmental mixing process within a mathematical framework, so that the source tracing results not only reflect point estimates, but also provide a quantifiable range of confidence.

[0024] In a preferred embodiment of the present invention, the quantitative contribution ratio of each endmember to the DOM of the lake water body in step (F) is the mean of the posterior distribution, and the confidence interval is a 95% confidence interval.

[0025] As a preferred implementation, this application uses the mean of the posterior distribution of the MixSIAR model as the quantitative contribution ratio of endmembers and uses a 95% confidence interval to characterize its uncertainty, thereby transforming the source tracing results from a single deterministic estimate into a statistically significant probabilistic output, making the source apportionment conclusions of lakes DOM have practical value that is quantifiable, comparable, and can support risk decision-making.

[0026] According to one aspect of the present invention, the above-described molecular fingerprint-based quantitative tracing method for dissolved organic matter in lakes is used in verifying the conservation of dissolved organic matter (DOM) molecules.

[0027] The above-mentioned molecular fingerprint-based quantitative source tracing method for dissolved organic matter (DOM) in lakes provided by this invention is used to verify the conservation of DOM molecules. This method calculates the βNTI and RCBray indices of the molecular set to be tested among lake water samples, and uses -2 < βNTI < 2 and -0.95 < RCBray < 0.95 as the criteria to confirm that the distribution of the set is mainly dominated by random drift processes. This enables a verifiable and repeatable determination of whether DOM molecules have real environmental conservation at the ecological mechanism level.

[0028] In a preferred embodiment of the present invention, the use includes: calculating the β nearest neighbor index (βNTI) and the Raup-Crick index (RCBray) of the RI molecular set in lake DOM source resolution. When -2 < βNTI < 2 and -0.95 < RCBray < 0.95, it is determined that the distribution of the RI molecule set is mainly dominated by random drift process, thus confirming its effectiveness as a conservative tracer. The RI molecular system was obtained by screening molecules that meet the following criteria: an H / C atomic ratio of less than 1.5, a transformation count of 0, and existence only in a single pollution source endmember. The H / C atomic ratio was calculated based on the FT-ICRMS measured molecular formula after isotopic correction.

[0029] In a preferred embodiment of the invention, the application is used to distinguish between endogenous and exogenous contributions to the DOM of a lake.

[0030] In a preferred embodiment of the present invention, the endogenous contribution originates from sediment release; the exogenous contribution is selected from at least one of topsoil, plant litter, fertilizer, livestock and poultry manure, sewage treatment plant effluent, and aquaculture effluent.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method for quantitative source tracing of dissolved organic matter (DOM) in lakes based on molecular fingerprints. The method involves collecting lake water samples and end-member samples from various potential pollution sources. After solid-phase extraction to enrich DOM, molecular formula data is obtained using Fourier transform ion cyclotron resonance mass spectrometry (FT-ICRMS). Resilient-inert (RI) molecules that meet the following criteria are screened: H / C < 1.5, have zero transformations in the molecular transformation network, and exist only in a single end-member. Their ecological conservation is verified using βNTI and RCBray indices. A high-contribution molecular feature set is selected using the SHAP algorithm. The MixSIAR Bayesian mixture model is then used to output the contribution ratio of each end-member and the 95% confidence interval. Compared with existing technologies, this invention introduces ecological zero-model verification into the DOM molecular tracer screening process for the first time, ensuring the conservation of input parameters mechanistically. Leveraging the molecular-level resolution of FT-ICRMS, it overcomes the internal / external source confusion problem caused by fingerprint overlap in spectroscopic methods. Furthermore, by simultaneously outputting quantitative results and uncertainty intervals through a Bayesian framework, it achieves accurate, reliable, and decision-oriented quantitative decomposition of the sources of DOM in lakes. Attached Figure Description

[0032] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0033] Figure 1 This is a schematic diagram of the detection process of the mainstream existing DOM source parsing technology solutions provided by the present invention; Figure 2 This is a schematic diagram of the detection process of the molecular fingerprint-based quantitative tracing method for dissolved organic matter in lakes provided in Embodiment 1 of the present invention. Detailed Implementation

[0034] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] The technical solution of the present invention will be further described below with reference to the embodiments.

[0036] Example 1 Figure 2 This is a schematic diagram of the detection process of the molecular fingerprint-based quantitative tracing method for dissolved organic matter in lakes provided in Embodiment 1 of the present invention.

[0037] A molecular fingerprint-based quantitative source tracing method for dissolved organic matter in lakes was implemented in the Caohai and Waihai lake areas of Dianchi Lake according to the following steps (see...). Figure 2 ): (I) Multi-source end-member sample collection and DOM enrichment: In August 2024, surface water samples (0–0.5 m) from the Caohai and Waihai lake areas of Dianchi Lake, as well as end-member samples from nine potential pollution sources within the watershed, were collected. 1. Lake sediments (columnar samples, 0–5 cm, representing endogenous release sources); 2. Topsoil of the watershed (0–10 cm tillage layer); 3. Litter of native dominant plants (Ottelia acuminata, Ceratophyllum demersum); 4. Commonly used nitrogen fertilizer (urea) and phosphate fertilizer (superphosphate) mixed samples; 5. Fresh livestock and poultry manure from large-scale fish farms; 6. Secondary treated effluent from a wastewater treatment plant in Kunming; 7. Water discharge from cage aquaculture ponds surrounding Caohai and Waihai lake areas; 8. Phytoplankton in Dianchi Lake's Caohai area and its outer lakes; 9. A phosphate mine in Kunming.

[0038] Water samples were vacuum filtered through a 0.45 μm polycarbonate membrane. Solid samples were extracted with ultrapure water (18.2 MΩ•cm) at a ratio of 1:10 (w / v) for 24 h (protected from light, 4℃), and the extract was filtered using the same method. The filtrate was enriched using an Agilent PPL solid-phase extraction column (500 mg / 6 mL), eluted with methanol (1:1, v / v), purged with nitrogen to near dryness, reconstituted with 500 μL of methanol-water (1:1, v / v), and stored at -20℃ for analysis.

[0039] (II) Acquisition of DOM molecules from samples: Take 1 μL of DOM complex solution from each endmember and detect it using Fourier transform ion cyclotron resonance mass spectrometry (FT-ICRMS) in positive ion mode.

[0040] The mass spectrometry parameters were set as follows: spray voltage 4.5 kV, drying gas temperature 200℃, nebulizer gas flow rate 10 L / min, scan range m / z 150–1000, and 200 transient signals accumulated per sample.

[0041] The raw data was processed using Compass DataAnalysis 5.0 software, and molecular formula identification (elemental composition limited to C1–) was performed using APLpy and NMRShiftDB databases. 50 H0– 100 O0– 20 (N0–5S0–2, mass error ≤ 1 ppm), generating a data matrix containing the above molecular formula and its relative peak intensities.

[0042] (III) Screening of "Resilient-Inert" (RI) Molecular Markers: Based on the molecular formula obtained in step S2, RI molecules were screened according to the following four criteria: 1. H / C atomic ratio < 1.5; 2. The number of transformations in the molecular transformation network is 0; The molecular transformation network is constructed based on the obtained DOM molecular formula data. It identifies potential transformation relationships between molecules by matching the mass difference between molecular formulas with the theoretical mass difference in the preset molecular transformation database. 3. This molecule was detected only in samples with a single endmember (relative peak intensity > 1 × 10⁻⁶). 4 And the strength of the remaining 9 endmembers is <500.

[0043] After screening, a total of 9 end-member specific RI molecule sets were obtained, totaling 292 RI molecules; among them, sediment end-members contained 50 RI molecules and wastewater effluent end-members contained 76 RI molecules.

[0044] (iv) Conservatism verification based on the ecological null model: The relative peak intensity matrix of the 9 endmember RI molecular sets obtained in step (III) in 10 water samples from the Caohai Lake area of ​​Dianchi Lake is input into a self-written Python script for calculation: β Nearest Neighbor Index (βNTI): Based on the Bray-Curtis distance matrix, tested with 999 random permutations; Raup-Crick exponent (RCBray): Calculated based on the same distance matrix.

[0045] The results showed that the mean βNTI of the 10 endmember RI molecule sets was 1.15 (<2) and the mean RCBray was -0.22 (<0.95), indicating that their distribution was significantly dominated by random drift processes, which met the criteria for conservative tracer verification.

[0046] (V) Selection of source tracing parameters based on machine learning: SHAP-driven source tracing feature optimization: Using molecular indices such as relative peak intensity, mass-to-nucleus ratio, nitrogen-to-carbon ratio, oxygen-to-carbon ratio, hydrogen-to-carbon ratio, modified aromaticity index, and carbon oxidation state of 9 endmember RI molecular sets as features, a random forest classification model (scikit-learn 1.3.0, n_estimators=500, max_depth=12) is constructed.

[0047] KernelExplainer was used to calculate the Shapley values ​​for each molecule on the test set (30% of the samples), and the molecules were sorted in descending order of their absolute contribution. The relative peak intensities and nitrogen-carbon ratios of molecules constituted the optimal set of source-tracing characteristic parameters.

[0048] (vi) Quantitative analysis of endmember contributions in Bayesian mixture models: Quantitative analysis of the MixSIAR Bayesian mixture model: The two molecular feature parameters obtained in step (V) were input into the MixSIAR model, and the prior distribution was set as Dirichlet(1,1,…,1) (7 dimensions), with the MCMC chain length being 10. 5 Next, burn-in period 2×10 4 Each time, the sampling interval is 100 steps.

[0049] The model output shows that the contribution ratio of each endmember in the DOM of the Caohai Lake area of ​​Dianchi Lake is as follows: Endogenous release from sediments: 29.5% (95% confidence interval: 1.7–52.6%). Fertilizer: 12.1% (95% confidence interval: 0.6~43.6%) Plant litter: 11.1% (95% confidence interval: 0.5–38.9%). Aquaculture wastewater: 9.8% (95% confidence interval: 0.5~27.7%) Wastewater treatment plant effluent: 9.3% (95% confidence interval: 0.5%~26.7%). Livestock and poultry manure: 7.7% (95% confidence interval: 0.4~23.3%); Topsoil: 7.5% (95% confidence interval: 0.4–22.4%).

[0050] Phytoplankton: 7.3% (95% confidence interval: 0.4–22.4%).

[0051] Phosphate rock: 5.7% (95% confidence interval: 0.3~17.1%).

[0052] The above results indicate that endogenous release from sediments is the primary source of DOM in Dianchi Lake, and its contribution ratio is significantly higher than that of all other exogenous endogenous endogenous components.

[0053] Comparative Example 1 A molecular fingerprint-based quantitative source tracing method for dissolved organic matter in lakes differs from Example 1 only in that: The spectral parameters obtained from three-dimensional fluorescence spectroscopy analysis were used as source-tracing characteristic parameters. The selected spectral parameters were fluorescence index and biological index, which were directly input into a Bayesian mixture model for quantitative analysis. The remaining sample collection, endmember settings, and model running conditions were the same as in Example 1. In this method, the screening, conservation verification, and feature optimization steps of molecular markers based on Fourier transform ion cyclotron resonance mass spectrometry were not employed.

[0054] The model output shows that the contribution ratios of each endmember in dissolved organic matter in the outer sea of ​​Dianchi Lake and Caohai Lake are as follows: Endogenous release from sediments: 5.3% (95% confidence interval: 0.3–15.4%). Fertilizer: 9.3% (95% confidence interval: 0.5~28.8%) Plant litter: 37.3% (95% confidence interval: 6.7–66.8%) Aquaculture wastewater: 5.7% (95% confidence interval: 0.4%–16.3%) Wastewater treatment plant effluent: 6.4% (95% confidence interval: 0.3~19.4%). Livestock and poultry manure: 12.6% (95% confidence interval: 0.5~41.4%). Topsoil: 8.1% (95% confidence interval: 0.4–24.4%).

[0055] Phytoplankton: 5.4% (95% confidence interval: 0.3–16.2%).

[0056] Phosphate rock: 9.9% (95% confidence interval: 0.4~31.0%).

[0057] Compared to Example 1, Comparative Example 1 identified plant litter as the primary source and endogenous release from sediment as a secondary source; while in Example 1, endogenous release from sediment was the primary source, contributing 29.5%. This difference indicates that when using only spectral parameters for source tracing, it is easily affected by biogeochemical processes such as photobleaching, microbial degradation, and adsorption-desorption within the lake, leading to overlapping or shifting of spectral fingerprints, thereby reducing the accuracy of end-member identification. In particular, it is easy to underestimate the contribution of endogenous release from sediment and overestimate the contribution of terrestrial inputs such as plant litter.

[0058] The results demonstrate that, compared to source tracing methods based solely on spectral parameters, Example 1, by introducing stable molecular marker screening and conservation verification steps, can effectively reduce the interference of environmental processes on tracer signals and improve the accuracy and robustness of source apportionment of dissolved organic matter in complex eutrophic lake systems.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for quantitative tracing of dissolved organic matter in lakes based on molecular fingerprinting, characterized in that, Includes the following steps: (A): Collect samples of the lake water body to be tested and potential pollution source end-units in the watershed. The pollution source end-units include: sediment, topsoil, plant litter, fertilizer, livestock and poultry manure, sewage treatment plant effluent, aquaculture effluent, phytoplankton, and phosphate rock. The aqueous or solid extracts of each end-member sample were filtered, and the dissolved organic matter was enriched by solid-phase extraction to obtain the dissolved organic matter samples enriched by each end-member. (B): Using mass spectrometry to detect the dissolved organic samples enriched by each endmember in step (A), a dataset containing molecular formulas and relative peak intensities is obtained; (C): Based on the molecular formula obtained in step (B), screen for those that satisfy: Molecules with H / C < 1.5, zero transformations in the molecular transformation network, and existing only in a single endmember are denoted as labeled molecules, and a fingerprint set of labeled molecules specific to each endmember is established. (D): Calculate the β nearest neighbor index βNTI and the Raup-Crick index RCBray based on Bray-Curtis distance among lake water samples of the marker molecular fingerprint set obtained in step (C); When -2 < βNTI < 2 and -0.95 < RCBray < 0.95, the set of labeled molecules is confirmed to meet the conservation verification conditions. (E): Using the relative peak intensity, mass-to-nucleus ratio, nitrogen-to-carbon ratio, oxygen-to-carbon ratio, hydrogen-to-carbon ratio, modified aromaticity index, carbon oxidation state, molecular Gibbs free energy, Kendrick mass defect, and double bond equivalent molecular index of the marked molecules that have been confirmed by conservation, a machine learning classification model is constructed. The contribution of each molecular feature to the endmember classification is quantified by the interpretability algorithm, and the molecular indicators with the top two contributions are selected as the optimal traceability feature parameters. (F): Input the optimal source tracing feature parameters into the Bayesian mixture model, and after running, output the quantitative contribution ratio and confidence interval of each endmember to the dissolved organic matter in the lake water.

2. The method for quantitative tracing of dissolved organic matter in lakes based on molecular fingerprinting according to claim 1, characterized in that, The sediment in the endmember of step (A) is an endogenous release source in the lake.

3. The method for quantitative tracing of dissolved organic matter in lakes based on molecular fingerprinting according to claim 1, characterized in that, The mass spectrometry detected in step (B) is Fourier transform ion cyclotron resonance mass spectrometry.

4. The method for quantitative tracing of dissolved organic matter in lakes based on molecular fingerprinting according to claim 1, characterized in that, The molecular transformation network in step (C) is constructed based on the DOM molecular formula data obtained in step (B). By matching the mass difference between molecular formulas with the theoretical mass difference in the preset molecular transformation database, potential transformation relationships between molecules are identified.

5. The method for quantitative tracing of dissolved organic matter in lakes based on molecular fingerprinting according to claim 1, characterized in that, The interpretability algorithm in step (E) is the SHAP algorithm.

6. The method for quantitative tracing of dissolved organic matter in lakes based on molecular fingerprinting according to claim 1, characterized in that, In step (F), the Bayesian mixture model is the MixSIAR model; In step (F), the quantitative contribution ratio of each endmember to dissolved organic matter in lake water is the mean of the posterior distribution, and the confidence interval is a 95% confidence interval.

7. The use of the molecular fingerprint-based quantitative tracing method for dissolved organic matter in lakes according to any one of claims 1 to 6 in verifying the conservation of dissolved organic matter molecules.

8. The use according to claim 7, characterized in that, The applications include: calculating the β nearest neighbor index βNTI and the Raup-Crick index RCBray of a set of labeled molecules in source apportionment of dissolved organic matter in lakes; When -2 < βNTI < 2 and -0.95 < RCBray < 0.95, it is determined that the distribution of the set of labeled molecules is mainly dominated by random drift process, thus confirming its effectiveness as a conservative tracer. The labeled molecules are obtained by screening for molecules that meet the following conditions: H / C atomic ratio less than 1.5, number of transformations of 0, and existence only in a single pollution source endmember.

9. The use according to claim 7, characterized in that, The application described herein is used to distinguish between endogenous and exogenous contributions of dissolved organic matter in lakes.

10. The use according to claim 9, characterized in that, The endogenous contribution originates from sediment release; The exogenous contribution is selected from at least one of the following: topsoil, plant litter, fertilizer, livestock and poultry manure, sewage treatment plant effluent, aquaculture effluent, phytoplankton, and phosphate rock.

Images