Three-dimensional partition DOM traceability model based on optical and isotope characteristics and application method

Abstract

The invention discloses a three-dimensional partition DOM traceability model based on optical and isotope characteristics. The three-dimensional partition DOM traceability model comprises the following steps: S1, establishing an end member library composed of four types of end members; s2, performing multi-index measurement on the basis of the collected samples of the sampling sites in the S1; s3, end member distinguishability test is carried out based on the data measured in the S2, and candidate indexes are selected; s4, inverting an end member contribution proportion of each sampling point; s5, screening an optimal combination; s6, carrying out partition modeling; and S7, pollution dynamic traceability and zoning treatment are carried out. Compared with the prior art, the three-dimensional partition DOM traceability model based on the optical and isotope characteristics and the application method thereof have the advantages that clear parameter combination screening, performance evaluation and partition implementation processes are provided, and the model and the application method support weather / rain period staging modeling and dynamic traceability.

Description

Technical Field

[0001] This invention relates to the field of urbanized watershed carbon cycle and pollution control technology, specifically to a three-dimensional partitioned DOM source tracing model and its application method based on optical and isotopic characteristics. Background Technology

[0002] Dissolved organic matter (DOM) in rivers is the most active organic carbon component in water bodies, participating in photochemical reactions, microbial metabolism, and metal complexation processes, and has a significant impact on water carbon cycling and water quality succession. The sources, structure, and chemical composition of DOM are complex, primarily controlled by land use patterns and hydrological disturbances. Therefore, accurately identifying DOM sources and their dynamic response mechanisms is a key scientific issue for understanding carbon cycling and pollution control in urbanized watersheds.

[0003] In multi-source coupled systems such as urbanized rivers, a single optical or isotopic index often cannot accurately separate signals from different endmembers, resulting in unstable source tracing and poor spatial adaptability.

[0004] To overcome this limitation, endmember hybrid analysis was introduced to achieve multi-source quantitative analysis through multi-parameter joint modeling.

[0005] This method, which constructs eigenvalue matrices of different endmembers and uses least squares or Bayesian algorithms to invert the endmember contribution ratio of mixed samples, has been widely used in hydrogeochemical research.

[0006] It also includes existing technologies such as the application of EMMA with three-dimensional optical-isotope combinations and the study of the contribution of different organic matter sources to urban river water during storm events; However, when the above-mentioned prior art is used, there are several issues, including: The model has several shortcomings, including: lack of systematic consideration of spatial zoning and hydrological variability; insufficient verification of model stability under both sunny and rainy conditions; failure to consider the combined inversion of stable carbon isotopes and end-member mixing models; failure to compare model performance and results at the land use zoning level; and reliance on two-dimensional optical parameter combinations.

[0007] This invention aims to address the problems of insufficient identification accuracy, poor spatial adaptability, and weak response to rainfall disturbances in existing DOM source tracing methods in complex urbanized watersheds. Summary of the Invention

[0008] The technical problem to be solved by this invention is to overcome the above-mentioned technical defects and provide a clear parameter combination screening, performance evaluation and zoning implementation process, supporting the phased modeling and dynamic tracing of sunny / rainy seasons, and a three-dimensional DOM tracing model and application method based on optical and isotopic characteristics.

[0009] To solve the above-mentioned technical problems, the technical solution provided by the present invention is: a three-dimensional partitioned DOM source tracing model based on optical and isotopic characteristics, comprising the following steps: S1: In the target watershed, sampling stations are set up according to functional zones. River samples are collected during sunny and rainy seasons. For each functional zone, an end-member library consisting of four types of end-members is established. S2: Based on the samples collected from the sampling sites in S1, multiple indicators are measured separately; S3: Based on the data measured in S2, perform endmember distinguishability test, remove indicators that are not significantly different between endmembers or have serious overlap in endmember ranges, and retain effective distinguishing indicators as candidate indicators. S4: Design multiple sets of two-dimensional and three-dimensional parameter combinations for candidate indicators to run endmember hybrid analysis models at the whole watershed scale, and invert the endmember contribution ratio of each sampling point; S5: Evaluate the performance of each parameter combination using multi-parameter combination metrics, and select the optimal combination; S6: Partition modeling. In each partition, a sub-model is independently built. Repeat S4-S5, compare the indicators, and determine the final traceability model for the corresponding functional area. S7: Based on the optimal parameter combination and zoning model, the dynamic impact of rainfall events on end-member contributions is quantified for dynamic pollution source tracing and zoning management.

[0010] Preferably, the functional areas in S1 include an upstream agricultural area, a midstream urban area, and a downstream built-up area; The river water samples were collected during the sunny period after a series of sunny days, and within 24 hours after rainfall during the rainy period. The terminator library contains four types of terminator samples: wastewater effluent, suspended algae, surface soil, and deep soil.

[0011] Preferably, the surface soil sampling depth is 0-1 cm, and the deep soil sampling depth is 5-10 cm.

[0012] Preferably, the multi-index determination in S2 includes organic carbon, optical index and stable carbon isotope determination; The optical parameters include PARAFAC components, biogenicity index, humification index, fluorescence index, specific UV absorption coefficient, spectral slope, and absorbance as determined by excitation-emission matrix fluorescence spectroscopy. The stable carbon isotopes include δ 13 C-DOC, δ 13 C-DIC; The absorbance is the absorption coefficient at 350 nm.

[0013] Preferably, the indicators in S5 include prediction goodness, mean relative error, and root mean square error. The S5 also includes a composite scoring system that combines indicators with spatial adaptability to determine the optimal model.

[0014] Preferably, the inversion calculation in S4 is performed using the Bayesian algorithm, the least squares method, or the Monte Carlo method.

[0015] Preferably, the index comparison in S6 includes using the sub-model as the final source tracing model for the corresponding functional area if the prediction goodness, average relative error, and root mean square error of the sub-model are better than those of the whole watershed model.

[0016] Preferably, the optimal parameter combination in S5 is a three-dimensional combination, including the biogenic index, the humification index, and stable carbon isotopes.

[0017] In another aspect, this invention discloses the application of the above-mentioned method in decision support for DOM source tracing and pollution control in urbanized watersheds.

[0018] The advantages of this invention compared to the prior art are: 1. Improved recognition accuracy: By incorporating optical indicators and stable carbon isotopes (δ¹³C-DOC) into a three-dimensional parameter combination, the ability to distinguish similar carbon signal sources such as wastewater and terrestrial sources is significantly improved (the BIX-HIX-δ¹³C-DOC combination in this invention improves the average G and significantly reduces RMSE compared to the traditional two-dimensional combination).

[0019] 2. Strong spatial adaptability: The zoning modeling strategy can adaptively set endmembers and parameter weights in different land use units, reduce the average deviation of the whole area, and improve the model fitting and source tracing credibility of different functional areas.

[0020] 3. Response to dynamic disturbances: It can quantify the changes in end-member contributions and input path transformations caused by rainfall, providing a quantitative basis for pollution emergency response and management under short-term hydrological disturbances. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the study area and sampling overview.

[0022] Figure 2 This is the result of optimizing the DOM source model construction process and parameter combination based on the EMMA method.

[0023] Figure 3 It compares the actual land use patterns in the study area and the global and regional analysis results of DOM sources under different model structures.

[0024] Figure 4 The results are based on the endmember performance metrics of different EMMA models across the entire watershed.

[0025] Figure 5It is based on the endmember selection results in a partially partitioned EMMA model.

[0026] Figure 6 This is a schematic diagram of the complete technical chain of the present invention.

[0027] As shown in the figure: Figure 2 The model structure in (a) is extended from two-dimensional optical parameters (such as BIX-HIX) to three-dimensional combinations of fused stable carbon isotopes (such as BIX-HIX-δ). 13 (c-D0C), and further optimize by introducing functional partitioning in space; (b) compare the model prediction goodness G value, mean relative error (MRE), and root mean square error (RMSE) corresponding to the combination of endmember parameters from two-dimensional to three-dimensional; (c) compare the optimal three-dimensional combination (BIX-HIX-δ) with the model prediction goodness G value, mean relative error (MRE), and root mean square error (RMSE). 13 Under C-D0C, the performance of the partition model after further modeling based on different partitions in terms of G, MRE and RMSE; Figure 3 (a) The actual land use pattern of the study area; (b) Based on the three-dimensional combination (FI-BIX-δ) 13 C-DOC and BIX-HIX-δ 13 (c) Results of DOM source analysis at the global scale (C-D0C). The bar chart shows the relative contribution ratios of four DOM sources—sewage, topsoil, deep soil, and algae—in the three functional zones (upstream farmland, midstream urban area, and downstream urban built-up area); 13 Based on C-DOC, sub-models were established for the upstream, midstream, and downstream areas respectively, and the effectiveness of regional modeling in identifying DOM sources during sunny periods and after rainfall was evaluated. Detailed Implementation

[0028] The present invention will now be described in further detail with reference to the accompanying drawings.

[0029] (I) Overall Process Framework 1. Data Collection and End-Member Database Establishment: Sampling stations were set up in the study section (Guangzhou section of the Pearl River) according to functional zones (upstream agricultural area, midstream urban area, downstream built-up area) (example: 21 stations). River water samples were collected during sunny periods (after consecutive sunny days) and rainy periods (within 24 hours after rainfall). Four types of end-member samples were collected at each zone or representative watershed location: wastewater treatment plant effluent, suspended algae, surface soil (0-1 cm, representing easily eroded input) and deep soil (5-10 cm, representing relatively inert release sources). DOC and optical parameters (EEMs-PARAFAC component, BIX, HIX, FI, β / α, SUVA) were measured for both river water and end-member samples. 254S 275-295 a 350 (etc.) and δ 13 C -DOC / δ 13 C-DIC.

[0030] 2. Data Preprocessing: Denoising, descattering, Raman normalization, and PARAFAC component analysis of EEMs data, with comparison and confirmation with OpenFluor. Endmember distinguishability tests (ANOVA / multiple comparisons) were performed on all candidate indicators, eliminating those with insignificant differences between endmembers or those with severely overlapping endmember ranges that made them indistinguishable.

[0031] 3. Parameter Combination Design and Model Running: Based on the aforementioned candidate indices, multiple sets of two-dimensional (arbitrary two-parameter combinations) and three-dimensional (multiple three-parameter combinations) parameter sets are designed (nearly 40 combinations in total in this invention example). EMMA (Bayesian / hybrid inverse solution) is run on each parameter combination at the whole watershed scale, outputting the endmember contribution estimates for each sample point.

[0032] 4. Model Performance Metrics and Selection: Figure 2 The study presents the overall modeling process. Based on the seven types of optical and stable isotope indexes of the watershed DOM identified in the preliminary stage, multiple parameter sets were formed by combining them to design various two-dimensional and three-dimensional parameter combination models, which were then run at the whole watershed scale. The performance was comprehensively evaluated using three indicators: G, MRE, and RMSE, and the optimal combination was selected. Furthermore, given that the aforementioned results have revealed significant spatial heterogeneity in the watershed DOM composition, the study introduces a land use-based spatial zoning strategy, constructing independent sub-models in different land use units to improve the model's structural adaptability and analytical accuracy under source term heterogeneity conditions. Each model runs independently in terms of endmember settings, parameter input, and fitting structure, better capturing the regional feature mixing mechanism and enhancing ecological interpretability. The results show that two-dimensional models (such as FI-BIX and BIX-HIX) are difficult to fully identify the multi-source mixing attributes of the DOM, with G values ​​generally below 0.60, RMSE above 0.80, and MRE between 1.8% and 6.9%, indicating limited analytical accuracy and stability.

[0033] In contrast, 3D models exhibit superior performance under the synergistic effect of multiple parameters, including various 3D combination schemes such as FI-BIX-δ. 13 C-DOC, FI-HIX-δ 13 C-DOC, BIX-HIX-δ 13 C-DOC, etc. A comprehensive evaluation revealed that BIX-HIX-δ 13The C-DOC combination, while maintaining strong structural differentiation among parameters, effectively captures the carbon source properties of the DOM, increases the average G value to 0.60, decreases RMSE to 0.40, and controls MRE within 4.1%, thereby improving the structural differentiation among endmembers and enhancing model stability, making it the optimal model structure of this invention.

[0034] By further introducing land use zoning on the basis of 3D modeling, the regional adaptability of the model is significantly improved. Figure 2 c). The G-value of the upstream agricultural area model increased to 0.62, and the RMSE decreased to 0.21, representing a reduction of approximately 50.0% in error compared to the global model. The MRE of the midstream urban area model decreased to 2.1%, while the RMSE remained at 0.21, demonstrating the most stable fitting accuracy. The G-value of the downstream built-up area model increased slightly, while the RMSE decreased by 43.9%. Overall, the regional models exhibited higher analytical accuracy and mathematical stability in each region.

[0035] More importantly, the evolution of the model structure significantly enhanced the ecological matching ability between the model structure and actual land use characteristics. Figure 3 ).

[0036] First, in the comparison of global modeling with multiple parameter combinations, two typical two-dimensional models, FI-BIX and BIX-HIX, were selected as representative models for further source tracing structure analysis and comparison, showing higher prediction quality (relatively large G value) and lower error (smaller MRE and RMSE). Specifically, the FI-BIX combination identified a higher contribution from topsoil in the source tracing results of the middle and lower reaches of the city (60.0% and 66.5% in sunny weather, and 64.5% and 67.2% in rainy weather), while the contribution from wastewater was lower (only 25-34%), deviating from the actual characteristics of the region's high urbanization level and significant point source emissions.

[0037] In comparison, the BIX-HIX combination demonstrated superior ability to identify biogenicity and humification levels, resulting in an improved contribution to urban wastewater (43.9% and 35.9% in the middle and lower reaches, respectively). However, it still failed to accurately capture the complex structure of pollution sources. A three-dimensional model incorporating stable isotope information (BIX-HIX-δ) was developed. 13 C-DOC significantly improved this bias. In the whole watershed model, wastewater became the main end-member in the middle and lower reaches, contributing 53.3% and 46.5% respectively, while the proportion of topsoil decreased to about 30%, reflecting the ability of carbon isotope indicators to supplement the identification of heavy isotope characteristics of wastewater.

[0038] However, even with a uniform structure, the global model still exhibits averaging bias when considering pollution characteristics across different regions. For instance, during the rainy season in the midstream urban area, the global model overestimated the contribution of topsoil (by 42.4%), while underestimating the contribution of wastewater (by 38.5%), which is inconsistent with the region's TN and COD levels during the rainy season. Mn Lift, SUVA 254 The low level and high FI (Firmware Intake) are inconsistent with the "sewage-dominated" signal, indicating that its spatial response capability is limited.

[0039] After adopting a land use-based three-dimensional zoning model, the DOM source structure identification results are more reasonable: the midstream urban area model clearly identifies sewage as the main source, contributing as much as 60.6% during the sunny season, and although it decreases during the rainy season, it still remains at a relatively high level.

[0040] This wastewater-dominated structure not only aligns with the region's high proportion of urban construction land and its frequent interference from sewage and pipeline overflows, but also corroborates the aforementioned water quality and optical characteristics, further demonstrating that the zoning modeling results can more accurately reflect the perception of the region's dominant pollution sources.

[0041] 5. Zonal Modeling: Based on the spatial heterogeneity of land use and end-member characteristics, sub-models are established for each functional zone (the end-member settings, end-member mean and variance, and parameter combinations for each zone can be set independently). Steps 3 and 4 are repeated to test whether the zonal model is significantly better than the global model (based on G, MRE, and RMSE as criteria).

[0042] 6. Comprehensive Scoring and Final Optimal Model Determination: A composite scoring system is established based on both mathematical performance (G, RMSE, MRE) and spatial fit (dominant endmember consistency rate, correlation test) to determine the optimal combination / zoning scheme. Only when the model significantly improves consistency with actual land use and water quality evidence while demonstrating mathematical superiority is it considered the optimal model best suited to actual land use (this invention ultimately identifies BIX-HIX-δ). 13 C-DOC is the optimal model for this type.

[0043] 7. Stage-based modeling and comparative analysis of sunny and rainy seasons: Based on the optimal parameter combination and regional model, the impact of rainfall on endmember contributions is quantified, the differences between sunny and rainy seasons are identified, and water quality indicators (TN, TP, COD) are considered. Mn The variation analysis of Chl-a and fluorescent components (PARAFAC) input paths and ecological responses can be used to construct risk warning thresholds and zonal governance recommendations based on DOM source structure.

[0044] In one embodiment: Besides BIX-HIX-δ 13In addition to C-DOC, other three-dimensional combinations can be used (such as FI-HIX-δ). 13 C-DOC, BIX-S 275-295 —δ 13 C-DOC achieves similar improvements under different watersheds or different end-member configurations; MCMC can be replaced by deterministic nonlinear least squares, Bayesian MCMC, or hybrid genetic / optimization algorithms, as long as the parameter combination screening and partition modeling process is retained; The zoning can be automatically generated based on land use layers or based on clustering algorithms (k-means, hierarchical clustering) using sample optical / chemical characteristics; δ can be added when conditions permit. 15 N, δ 2 H, radioactive isotopes, or molecular-level tracer indicators (FT-ICR-MS characteristic peaks) are used as additional dimensions to improve discriminative power.

[0045] like Figure 6 As shown, the complete technical chain from DOM indicator extraction and model optimization (2D to 3D, global to regional), to DOM source parsing results (taking the midstream as an example) and feature response is clearly presented, intuitively demonstrating the application of the 3D regional EMMA model in DOM source tracing of urban river networks; The left-hand input section contains two core inputs: first, potential source endmembers of DOM (sewage, topsoil, deep soil, algae); and second, river water samples from different functional zones across the study area during varying weather conditions, from which three key indicators of DOM are extracted: biogenic information (BIX), humic properties (HIX), and carbon source characteristics (δ¹⁸O). 13 C-DOC), as input parameters for the model; The middle section covers model optimization and computation: it first showcases a 2D EMMA model at the entire watershed scale (using the HIX-BIX two-dimensional index as an example), and then upgrades it to integrate δ... 13 The C-DOC 3D EMMA model is used; the entire watershed is partitioned, and model calculations are performed by functional zones to improve the accuracy of source resolution (it is more adaptable to the heterogeneity of different regions compared to the whole watershed model). The output module on the right presents the analysis results using the midstream urban core area (MU) as an example. Furthermore, these results correspond to the DOM feature changes actually detected under the combined influence of land use and rainfall.

[0046] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0047] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A three-dimensional zoned DOM provenance model based on optical and isotopic signatures, characterized in that: It comprises the following steps: S1: Dividing the target river basin into functional zones and deploying sampling sites, sampling river samples in sunny and rainy periods, and establishing an end-member library composed of four types of end-members for each functional zone; S2: Measuring multiple indexes based on the samples collected at the sampling sites in S1; S3: Conducting end-member distinguishability test based on the data measured in S2, eliminating indexes with insignificant differences between end-members or serious end-member range overlap, and retaining effective distinguishing indexes as candidate indexes; S4: Designing multiple sets of two-dimensional and three-dimensional parameter combinations for running the end-member mixing analysis model at the whole-basin scale, and inverting the end-member contribution proportion of each sampling point; S5: Evaluating the performance of each parameter combination by multiple parameter combinations, and selecting the optimal combination; S6: Dividing the model into zones, independently establishing a sub-model for each zone, repeating S4-S5, comparing the indexes, and determining the final tracing model for the corresponding functional zone; S7: Quantifying the dynamic influence of rainfall events on end-member contribution based on the optimal parameter combination and the zoned model, and applying it to pollution dynamic tracing and zoned governance.

2. The three-dimensional zoning DOM provenance model based on optical and isotopic characteristics according to claim 1, characterized in that: The functional zones in S1 include upstream farming areas, midstream urban areas, and downstream built-up areas; The river water samples are collected after consecutive sunny days in sunny periods and within 24 hours after rainfall in rainy periods; The four types of end-member samples in the end-member library include sewage treatment plant effluent, suspended algae, surface soil, and deep soil.

3. The three-dimensional zoning DOM provenance model based on optical and isotopic characteristics according to claim 2, characterized in that: The surface soil sampling depth is 0-1 cm, and the deep soil sampling depth is 5-10 cm.

4. The three-dimensional zoning DOM provenance model based on optical and isotopic characteristics according to claim 2, characterized in that: The multiple index measurements in S2 include organic carbon, optical indexes, and stable carbon isotope measurements; The optical indexes include PARAFAC components, biological source index, humification index, fluorescence index, specific ultraviolet absorption coefficient, spectral slope, and absorbance obtained by excitation-emission matrix fluorescence spectrum analysis; The stable carbon isotope includes δ 13 C-DOC, δ 13 C-DIC; The absorbance is the absorption coefficient at 350 nm.

5. The three-dimensional zoning DOM provenance model based on optical and isotopic characteristics according to claim 4, characterized in that: The indexes in S5 include prediction goodness, average relative error, and root mean square error, S5 also includes a composite scoring system to determine the optimal model in combination with the index and spatial adaptability.

6. The three-dimensional zoning DOM provenance model based on optical and isotopic characteristics according to claim 1, characterized in that: The inversion calculation in S4 is performed by Bayesian algorithm, least squares method, or Monte Carlo method.

7. The three-dimensional zoned DOM provenance model based on optical and isotopic signatures according to claim 5, characterized in that: The index comparison in S6 includes enabling the zoned sub-model as the final tracing model for the corresponding functional zone when the prediction goodness, average relative error, and root mean square error of the zoned sub-model are better than those of the whole-basin model.

8. The three-dimensional zoning DOM provenance model based on optical and isotopic characteristics according to claim 5, characterized in that: The optimal parameter combination in S5 is a three-dimensional combination including biological source index, humification index, and stable carbon isotope.

9. Application of the method of any one of claims 1-8 in the decision support of DOM tracing and pollution governance in urbanized river basins.

Images