Method for quantifying a fingerprint attribute difference threshold and method for tracing sediment sources

Abstract

The application provides a method for quantifying a fingerprint attribute difference threshold, which comprises the following steps: S1, obtaining a relative deviation data set; S2, configuring an initial data set corresponding to a plurality of sediment sources under the relative deviation condition based on the relative deviation data set; S3, mixing the initial data set according to a plurality of source proportions to obtain an initial virtual mixture corresponding to each source proportion under the relative deviation condition, and performing fingerprint attribute error fluctuation processing to obtain a virtual mixture; S4, evaluating a contribution rate average absolute error under the relative deviation condition based on the virtual mixture; S5, fitting a contribution rate average absolute error and a relative deviation relationship based on the contribution rate average absolute errors under different relative deviation conditions obtained in S4; and S6, determining the fingerprint attribute difference threshold based on the relationship and a contribution rate average absolute error threshold. The application can improve the accuracy of sediment source tracing results.

Description

Technical Field

[0001] This invention relates to the field of sediment source assessment, and particularly to the field of quantitative assessment of sediment sources, specifically to a method for quantifying fingerprint attribute difference thresholds and a method for tracing sediment sources. Background Technology

[0002] Against the backdrop of global climate and land-use change, soil erosion is occurring at a more severe and faster pace. This not only leads to land degradation and a rapid decline in land productivity, but also causes downstream water pollution, reservoir siltation, and marine environmental damage, posing a serious threat to food security and ecosystem stability, and ultimately affecting the sustainable development of human society (Amundson et al., 2015; Shukla et al., 2019; Owens, 2020; Borrellie et al., 2022). Therefore, the prevention and control of soil erosion has become a key focus worldwide, and accurate and efficient identification of the sources of erosion sediment is of great significance for addressing such environmental problems.

[0003] In closed drainage systems, rainfall and runoff drive soil erosion on slopes. According to the principle of mass conservation, the total amount of soil erosion from different sources equals the total amount of sediment at the downstream outlet. Therefore, obtaining the contribution rate of sediment from different sources is crucial for targeted soil erosion control. Sediment source fingerprinting is a quantitative tracing method for sediment sources applied in soil and water conservation (soil erosion). It uses stable physical, chemical, or biological information carried by soil and sediment as "fingerprint factors" to establish a mass conservation model from source to sediment. Solving the model yields the contribution rate of sediment from different sources. Currently, sediment source fingerprinting is widely used for quantitative sediment source tracing due to its speed, lack of spatiotemporal limitations, and low cost (Haddadchi et al., 2013; Koitere et al., 2013; Laceby et al., 2019). A fingerprint refers to the physical, chemical, or biological information adsorbed or encapsulated by soil or sediment, such as geochemical elements, soil magnetism, and color. When selecting these fingerprint factors, their temporal and spatial stability must be ensured. Furthermore, sufficient difference in fingerprint factor attribute values ​​between different sources is a prerequisite for accurate sediment source identification studies (Collinsetal, 2020; Evrard et al., 2022). When there are small differences in fingerprint factor attribute values ​​between sources, the predicted sediment source contribution is strongly affected by the non-conservatism (or variability) of fingerprint factors and sample measurement errors (Pulley et al., 2015). For example, if the differences in fingerprint factor attribute values ​​between different sources are small, even small changes in fingerprint factor attribute values ​​may affect the accuracy of the sediment source contribution rate results (Pulley et al., 2017). Conversely, if the differences in fingerprint factors between different sources are large enough, the system's robustness to the non-conservatism (i.e., variability) of fingerprint factors and measurement errors will be greatly increased, thereby ensuring the reliability of source classification.

[0004] However, in practical applications, although statistical methods can be used to screen for fingerprint factor combinations with the greatest differences in attributes among sediment sources, these differences are relative and cannot guarantee the rationality of source classification. Currently, there is a lack of research on the reliability of sediment source classification based on fingerprint factor attribute differences, increasing the uncertainty of sediment source prediction results. Therefore, quantifying the threshold of fingerprint attribute differences in sediment source areas is crucial to achieving reliable sediment source classification and thus improving the accuracy of sediment source tracing results. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for quantifying fingerprint attribute difference thresholds and a method for tracing sediment sources, in order to solve the problem that the accuracy of tracing results is affected by the unclear fingerprint factor difference thresholds of sediment source areas.

[0006] The objective of this invention is achieved through the following methods:

[0007] According to a first aspect of the present invention, a method for quantifying a fingerprint attribute difference threshold is provided, wherein the fingerprint attribute difference threshold is used to characterize the reliability of identifying sediment sources using fingerprint factors. The quantification method includes: S1, obtaining a relative deviation dataset composed of multiple relative deviations, wherein the relative deviations represent the differences of the same fingerprint factor in different sediment sources; S2, based on each relative deviation in the relative deviation dataset obtained in S1, configuring an initial dataset corresponding to multiple sediment sources under the relative deviation conditions, wherein the initial dataset includes a fingerprint factor attribute dataset corresponding to each sediment source, wherein the fingerprint factor attribute dataset consists of multiple fingerprint factor attribute values; S3, converting the initial dataset obtained in S2 under the relative deviation conditions... The initial dataset is mixed according to multiple different source ratios to obtain an initial virtual mixture corresponding to each source ratio under the relative deviation condition. The initial virtual mixture is then processed for fingerprint attribute error fluctuation to obtain a virtual mixture corresponding to each source ratio under the relative deviation condition. S4: The mean absolute error of contribution rate under the relative deviation condition is evaluated based on the virtual mixtures corresponding to all source ratios under the relative deviation condition obtained in S3. S5: The relationship between the mean absolute error of contribution rate and the relative deviation is fitted based on the mean absolute error of contribution rate under different relative deviation conditions obtained in S4. S6: The fingerprint attribute difference threshold is determined based on the relationship and the mean absolute error of contribution rate threshold.

[0008] In some embodiments of the present invention, in step S2, the initial dataset corresponding to multiple sediment sources under the relative deviation condition is configured as follows: S21, obtaining a preset sediment source type and a fingerprint factor combination adapted to the preset sediment source type; S22, randomly selecting one sediment source in step S21 and assigning a value to the fingerprint factor attribute dataset corresponding to that sediment source; S23, using the relative deviation and the fingerprint factor attribute dataset assigned in step S22 to calculate the fingerprint factor attribute dataset corresponding to the remaining sediment types among the preset sediment source types.

[0009] In some embodiments of the present invention, the preset sediment source types include two types. In S23, the fingerprint factor attribute value of each fingerprint factor attribute in the fingerprint factor attribute dataset corresponding to the remaining sediment types is calculated in the following manner:

[0010]

[0011]

[0012] Where R represents the relative deviation, c1 represents the fingerprint factor attribute value of fingerprint factor c in the first sediment source, and c2 represents the fingerprint factor attribute value of fingerprint factor c in the second sediment source.

[0013] In some embodiments of the present invention, in S3, the attribute value of each fingerprint factor in the initial virtual mixture corresponding to each source ratio is calculated in the following manner:

[0014]

[0015] in, This represents the value of the fingerprint factor after mixing in the initial virtual mixture. Indicates the types and quantities of sediment sources. Indicates the first The fingerprint factor attribute value of fingerprint factor c in the source of sediment. In the mixing ratio k, the first... The proportions corresponding to the sources of sediment, among which =1.

[0016] In some embodiments of the present invention, in S3, the virtual mixture corresponding to each source ratio under relative deviation conditions includes multiple fluctuated fingerprint factor attribute values, wherein each fluctuated fingerprint factor attribute value is calculated in the following manner:

[0017]

[0018] in, This represents the value of the fingerprint factor after mixing in the initial virtual mixture. Due to fingerprint attribute errors, This represents the fingerprint factor attribute value after fluctuation.

[0019] In some embodiments of the present invention, S4 includes: S41, evaluating the virtual contribution rate under relative deviation conditions using a mass balance model, wherein the virtual contribution rate uses virtual mixtures and initial datasets corresponding to each source ratio, and calculating the mean absolute error of the contribution rate corresponding to the source ratio based on the virtual contribution rate and the source ratio; S42, taking the average of the mean absolute errors of the contribution rates corresponding to all source ratios under each relative deviation condition in S41 as the mean absolute error of the contribution rate under that relative deviation condition.

[0020] In some embodiments of the present invention, in step S41, the virtual contribution rate is obtained by solving the virtual mixture and the initial dataset using the following mass balance model:

[0021]

[0022]

[0023]

[0024] Where f is the sum of squared residuals, C i The first in the virtual mixture Fingerprint factor attribute value, S ij The first one in the initial dataset The source of soil and sand Fingerprint factor attribute value, X j For the first The virtual contribution rate of each sediment source is given, where m represents the types of sediment sources and n represents the types of fingerprint factors, and m ≤ n.

[0025] In some embodiments of the present invention, in step S41, the average absolute error of the contribution rate corresponding to the source ratio is calculated based on the virtual contribution rate and the source ratio using the following rules:

[0026]

[0027] Where MAE represents the mean absolute error of the contribution rate corresponding to the source proportion, m is the amount of sediment source, and Y j Indicates the first The proportion of each type of sediment source in the mixing ratio, X j For the first Virtual contribution rate of sediment source.

[0028] According to a second aspect of the present invention, a method for tracing sediment sources is provided, the method comprising the following steps: T1, acquiring data to be evaluated from a target area for inputting a mass balance model, the data to be evaluated including calibration data under multiple source classifications, the calibration data including datasets under each source classification, wherein the datasets include initial datasets corresponding to multiple sediment sources under each source classification and real mixtures; T2, grouping the initial datasets corresponding to multiple sediment sources under each source classification in the data to be evaluated obtained in T1 according to a pairwise non-repeating grouping method, forming multiple grouped data under that source classification; T3, calculating each of the data in T2... T4. Using the fingerprint attribute difference threshold determined by the quantification method provided in the first aspect of the present invention, the source classification reliability verification is performed on all relative deviations obtained in step T3 for each source classification; wherein, when at least one of the relative deviations of fingerprint factors obtained in each group of data under a source classification is greater than the fingerprint attribute difference threshold, the source classification reliability verification is determined to be passed; T5. The data to be evaluated under the source classification that passed the source classification reliability verification in step T4 is used to evaluate the contribution rate of each sediment source under that source classification using a mass balance model.

[0029] Compared with existing technologies, this invention has the following advantages: By setting fingerprint factors to generate a series of deviations in different sediment sources, this invention simultaneously generates a series of mixtures with different sediment source proportions, and considers the impact of real-world fluctuation errors on the mixtures. It uses a mass balance model to output the source proportions of the series of mixtures based on these fluctuation errors, calculates the average absolute error between the known proportions and the model output proportions, obtains the correlation between the deviations and the average absolute error, and fits the relationship equation. It quantitatively transforms the absolute error threshold of the contribution rate (the maximum acceptable source contribution prediction error in sand source fingerprinting research) into a fingerprint attribute difference threshold, providing a quantitative reference for subsequent source reliability analysis, thereby improving the accuracy of sediment source contribution prediction. Attached Figure Description

[0030] The embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0031] Figure 1 This is a schematic diagram illustrating the relationship between the average absolute error of contribution rate and the relative deviation in a fingerprint attribute difference threshold quantification method provided in an embodiment of the present invention.

[0032] Figure 2 A schematic diagram illustrating sediment source classification under a source classification (e.g., land use type) within a study area, provided as an embodiment of the present invention;

[0033] Figure 3A relative deviation distribution diagram provided for an embodiment of the present invention that failed the source classification reliability verification;

[0034] Figure 4 This is a relative deviation distribution diagram for source classification reliability verification provided in an embodiment of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0036] As mentioned in the background section, "Currently, there is a lack of research on the reliability of sediment source classification based on fingerprint factor attribute differences, which increases the uncertainty of sediment source prediction results." To address this issue, this invention analyzes the overall process of tracing methods for different sediment sources and quantitatively converts the average absolute error threshold of the contribution rate into a fingerprint attribute difference threshold (the fingerprint attribute difference threshold is used to characterize the reliability of identifying sediment sources using fingerprint factors). This application improves the accuracy of sediment source contribution prediction. Specifically, using the fingerprint attribute difference threshold for evaluating the reliability of sediment source classification provides a quantitative reference and can significantly improve the accuracy of sediment source contribution rate prediction.

[0037] According to a first aspect of the present invention, a method for quantifying a fingerprint attribute difference threshold is provided, wherein the fingerprint attribute difference threshold is used to characterize the reliability of identifying sediment sources using fingerprint factors. The quantification method includes: S1, obtaining a relative deviation dataset composed of multiple relative deviations, wherein the relative deviations represent the differences of the same fingerprint factor in different sediment sources; S2, based on each relative deviation in the relative deviation dataset obtained in S1, configuring an initial dataset corresponding to multiple sediment sources under the relative deviation conditions, wherein the initial dataset includes a fingerprint factor attribute dataset corresponding to each sediment source, wherein the fingerprint factor attribute dataset consists of multiple fingerprint factor attribute values; S3, converting the initial dataset obtained in S2 under the relative deviation conditions... The initial dataset is mixed according to multiple different source ratios to obtain an initial virtual mixture corresponding to each source ratio under the relative deviation condition. The initial virtual mixture is then processed for fingerprint attribute error fluctuation to obtain a virtual mixture corresponding to each source ratio under the relative deviation condition. S4: The mean absolute error of contribution rate under the relative deviation condition is evaluated based on the virtual mixtures corresponding to all source ratios under the relative deviation condition obtained in S3. S5: The relationship between the mean absolute error of contribution rate and the relative deviation is fitted based on the mean absolute error of contribution rate under different relative deviation conditions obtained in S4. S6: The fingerprint attribute difference threshold is determined based on the relationship and the mean absolute error of contribution rate threshold.

[0038] It should be noted that the average absolute error threshold of contribution rate can be understood as the generally accepted average absolute error threshold of contribution rate in sediment source fingerprinting research. If the contribution rate is less than this threshold, the source contribution prediction result is considered correct and acceptable. This invention considers the fluctuation error in real-world conditions through a relative deviation dataset, generates multiple virtual mixtures, and configures initial datasets corresponding to multiple sediment sources under the given relative deviation conditions for each of the generated virtual mixtures. It also considers the impact of fluctuation error in the displayed situation on the mixtures and quantitatively converts the average absolute error threshold of contribution rate (the maximum acceptable source contribution prediction error in sediment source fingerprinting research) into a fingerprint attribute difference threshold, providing a quantitative reference for subsequent source reliability analysis and thus improving the accuracy of sediment source contribution prediction. Furthermore, configuring the initial dataset using relative deviation can eliminate interference from single factors affecting fingerprint factor attributes. After processing the initial virtual mixtures for fingerprint attribute error fluctuations, it ensures that the data in the virtual process more closely matches the actual collection situation, thereby better and more accurately quantifying the fingerprint attribute difference threshold.

[0039] To better illustrate the technical solution of the present invention, the following mainly uses mathematical calculation processes to provide a detailed illustrative explanation of the steps.

[0040] In the quantification method, a relative deviation dataset is obtained, consisting of multiple relative deviations. These relative deviations represent the differences of the same fingerprint factor in different sediment sources. The relative deviation dataset is set to simulate as many virtual mixtures as possible to obtain a better fingerprint attribute difference threshold. The relative deviations in the dataset can be set with reference to the mean absolute error of contribution rate (MAE) under the previous relative deviation condition, following the gradient increase trend of the relative deviation. Considering that during actual sediment sample collection, due to human interference, sampling errors, and other factors, the actual measured fingerprint factor attribute values ​​inevitably fluctuate compared to the ideal values, small differences in fingerprint factor attribute values ​​between sources may lead to inaccurate sediment source contribution prediction results. Preferably, when the MAE of the current evaluation is large, the increase in the relative deviation used in the next evaluation can be smaller to obtain a better evaluation fingerprint attribute difference threshold. Setting the relative deviation according to the gradient increase of the relative deviation can explore a better initial dataset, resulting in a more accurate fingerprint attribute difference threshold determined by the mean absolute error of contribution rate. Preferably, when the MAE obtained in the previous calculation is less than the gradient threshold (e.g., 0.16), the relative deviation used in the next calculation can be generated according to an arithmetic progression. For example, if the relative deviation set in the previous calculation is 2%, the evaluated MAE is 0.521; the relative deviation used in the next calculation can be 3%. When the MAE obtained in the previous calculation is greater than the gradient threshold (e.g., 0.16), the relative deviation used in the next calculation can be generated according to a proportional progression. For example, if the relative deviation set in the previous calculation is 6%, the evaluated MAE is 0.167; the relative deviation used in the next calculation can be 8%.

[0041] In addition, to simplify the calculation process, the initial dataset can be generated in advance through experiments on the rising trend of the relative deviation gradient.

[0042] In the quantification method, the process of configuring the initial dataset includes: based on each relative deviation in the obtained relative deviation dataset, configuring the initial dataset corresponding to multiple sediment sources under the relative deviation condition. The initial dataset includes the fingerprint factor attribute dataset corresponding to each sediment source, and the fingerprint factor attribute dataset consists of multiple fingerprint factor attribute values.

[0043] According to an embodiment of the present invention, in the process of initializing the dataset, the initial dataset corresponding to multiple sediment sources under the relative deviation condition is configured in the following manner: S21, obtaining a preset sediment source type and a fingerprint factor combination adapted to the preset sediment source type; S22, randomly selecting one sediment source in S21 and assigning a value to the fingerprint factor attribute dataset corresponding to the sediment source; S23, using the relative deviation and the fingerprint factor attribute dataset assigned in S22 to calculate the fingerprint factor attribute dataset corresponding to the remaining sediment types among the preset sediment source types.

[0044] It should be noted that by constraining the relative deviation, it can be ensured that the initial dataset is generated only under the single factor of relative deviation, thus improving the richness of the initial dataset corresponding to multiple sediment sources and ensuring the singularity of the influencing factor. For example, when the relative deviation is determined, the corresponding first fingerprint factor attribute values ​​and second fingerprint factor attribute values ​​can be many sets, which can simulate various situations in the real sampling process. This improves the accuracy of the quantification method for fingerprint attribute difference thresholds.

[0045] To better illustrate the assignment process, it is preferred that two types of sediment sources are preset in the quantification method. The following description uses the assignment process for two sediment sources as an example. In one embodiment of the present invention, in step S23, the value of each fingerprint factor attribute in the fingerprint factor attribute dataset corresponding to the remaining sediment types is calculated in the following manner:

[0046]

[0047]

[0048] Where R represents the relative deviation, c1 represents the fingerprint factor attribute value of fingerprint factor c in the first sediment source, and c2 represents the fingerprint factor attribute value of fingerprint factor c in the second sediment source. One of c1 and c2 can be randomly selected for assignment while solving for the other. The constraint of the relative deviation does not affect the output of the evaluation results. For example, if the relative deviation is set to 2%, the sediment source types are set to two (S1 and S2), and the fingerprint factors are set to A and B, then attribute values ​​(c1, c2, c2, c2, c2, c3, c4 ... S1A =100; c S1B =2000), and the fingerprint factor attribute value for another sediment source S2 is calculated to be c. S2A =98, c S2B =1960.

[0049] In the virtual mixture generation process, each initial dataset obtained above is subjected to the following operation: the initial datasets obtained under relative deviation conditions are mixed according to multiple source proportions to obtain an initial virtual mixture corresponding to each source proportion under the relative deviation conditions. Then, the initial virtual mixture is processed for fingerprint attribute error fluctuation to obtain a virtual mixture corresponding to each source proportion under the relative deviation conditions. In other words, the biggest advantage of virtual artificial mixtures is that the actual sediment source contribution is known, and the accuracy of the source contribution calculated by the model can be quantitatively evaluated based on the actual sediment source contribution.

[0050] It should be noted that mixing the initial dataset obtained under relative bias conditions according to multiple source proportions can increase the number of samples in the virtual mixture and improve the accuracy of fingerprint attribute difference threshold assessment. The preferred multiple source proportions are k1 (20%, 80%), k2 (30%, 70%), k3 (40%, 60%), k4 (50%, 50%), k5 (60%, 40%), k6 (70%, 30%), and k7 (80%, 20%).

[0051] According to one embodiment of the present invention, the attribute value of each fingerprint factor in the initial virtual mixture corresponding to each source ratio is calculated in the following manner:

[0052]

[0053] in, This represents the value of the fingerprint factor after mixing in the initial virtual mixture. Indicates the types and quantities of sediment sources. Indicates the first The fingerprint factor attribute value of fingerprint factor c in the source of sediment. In the mixing ratio k, the first... The proportions corresponding to the sources of sediment, among which =1.

[0054] Continuing with the example of sediment sources S1 and S2, when mixing at a ratio of k1, the attribute values ​​of the fingerprint factors in the initial virtual mixture after mixing are... M1A =100*20%+98*80%=98.4. Here, the percentages 20% and 80% are consistent with the standard percentages used in calculating the MAE value.

[0055] In the fluctuation processing, fingerprint attribute errors are introduced to simulate virtual mixtures under real-world conditions. According to one embodiment of the invention, the virtual mixture corresponding to each source ratio under relative deviation conditions includes multiple fluctuated fingerprint factor attribute values, wherein each fluctuated fingerprint factor attribute value is calculated as follows:

[0056]

[0057] in, This represents the value of the fingerprint factor after mixing in the initial virtual mixture. Due to fingerprint attribute errors, This represents the fingerprint factor attribute value after fluctuation. In this embodiment, The maximum setting is 2%. This reduces the overall computational complexity and improves the accuracy of the evaluation.

[0058] When evaluating the mean absolute error of contribution rate under relative deviation conditions, the mean absolute error of contribution rate under such relative deviation conditions is evaluated based on the virtual mixture corresponding to all source proportions under the relative deviation conditions. According to one embodiment of the present invention, the mean absolute error of contribution rate under relative deviation conditions is evaluated as follows: S41, a mass balance model is used to evaluate the virtual contribution rate under relative deviation conditions, the virtual contribution rate using the data corresponding to each source proportion and the initial dataset, and the mean absolute error of contribution rate corresponding to the source proportion is calculated based on the virtual contribution rate and the source proportion; S42, the average of the mean absolute errors of contribution rate corresponding to all source proportions under each relative deviation condition in S41 is taken as the mean absolute error of contribution rate under that relative deviation condition.

[0059] It should be noted that the virtual contribution rate of each sediment source was calculated using the OptQuest module in Oracle Crystal Ball. OptQuest overcomes the limitations of enumeration methods by intelligently searching for the optimal solution in the simulation model. In almost all cases, OptQuest can effectively find the best or near-best solution among a large number of possible alternatives (daSilvaetal, 2022). OptQuest uses heuristic algorithms to guide the search process to improve the solution, which avoids getting trapped in local solutions and maintains its effectiveness even when dealing with noisy model data (Oracle, 2018). In the calculation process, the maximum number of simulation runs was set to 2000, and the Latin hypercube sampling method was used. The above solution process is well known to those skilled in the art and will not be described in detail here.

[0060] According to one embodiment of the present invention, the virtual contribution rate is obtained by solving the virtual mixture and the initial dataset using the following mass balance model:

[0061]

[0062]

[0063]

[0064] Where f is the sum of squared residuals, C i The first in the virtual mixture Fingerprint factor attribute value, S ij The first one in the initial dataset The source of soil and sand Fingerprint factor attribute value, X j For the first The virtual contribution rate of each sediment source is given, where m represents the types of sediment sources and n represents the types of fingerprint factors, and m ≤ n.

[0065] According to one embodiment of the present invention, the mean absolute error of the contribution rate corresponding to the source ratio is calculated based on the virtual contribution rate and the source ratio using the following rules:

[0066]

[0067] Where MAE represents the mean absolute error of the contribution rate corresponding to the source proportion, m is the amount of sediment source, and Y j Indicates the first The proportion of each type of sediment source in the mixing ratio, X j For the first Virtual contribution rate of sediment source.

[0068] During the fitting process, the relationship between the mean absolute error of the contribution rate and the relative deviation is obtained based on the mean absolute error of the contribution rate under different relative deviation conditions. The fitting method used here is well known to those skilled in the art, such as the graphical method and the function approximation method, and will not be elaborated here.

[0069] To better illustrate the fitting process and the calculation process of the mean absolute error of the contribution rate, combined with Figure 1 Please provide a detailed explanation. For example... Figure 1As shown, the preferred mixing ratios are k1, k2, k3, and k4, representing four mixing ratios: 20% and 80%, 30% and 70%, 40% and 60%, and 50% and 50%, respectively. For ease of explanation, two types of sediment sources are assumed. A relative deviation (which can also be considered as the relative deviation of the same fingerprint factor attribute values ​​between different sources) of 1% is used for illustrative purposes. When the relative deviation is 1%, the initial dataset and virtual mixtures are constructed according to k1, k2, k3, and k4, respectively. Then, the mean absolute error of the contribution rate under each ratio is evaluated to obtain the mean absolute error of the contribution rate under each ratio. Preferably, the arithmetic mean of the mean absolute errors of the contribution rates under the four mixing ratios is used as the mean absolute error of the contribution rate under the condition of a 1% relative deviation. When calculating the mean absolute error of the contribution rate under each mixing ratio, each ratio value in the mixing ratio is taken as the true contribution rate of the sediment source. For example, when the mixing ratio is k1, 20% and 80% are the ratio values, corresponding to the true contribution rates of the two sediment sources in the initial dataset.

[0070] In determining the fingerprint attribute difference threshold, the threshold is determined based on the aforementioned interrelationship and the mean absolute error of contribution rate. The interrelationship (e.g., a function) between the mean absolute error of contribution rate and the relative deviation obtained during the fitting process is used to determine the relative deviation corresponding to the mean absolute error of contribution rate threshold, and this deviation value is used as the fingerprint attribute difference threshold. Figure 1 As shown, when the average absolute error threshold of the contribution rate is set to 13%, the relative deviation is 8%, which is the fingerprint attribute difference threshold.

[0071] In summary, the fingerprint attribute difference threshold quantification method provided by this invention can quantitatively convert the average absolute error threshold of contribution rate into a fingerprint attribute difference threshold, providing a quantitative and reliable reference for subsequent source reliability analysis, thereby improving the accuracy of contribution rates of different sediment sources.

[0072] According to a second aspect of the present invention, a method for tracing the source of sediment is provided, the tracing method comprising the following steps:

[0073] T1. Obtain the data to be evaluated in the target area for input into the mass balance model. The data to be evaluated includes calibration data under multiple source classifications. The calibration data includes datasets under each source classification, wherein the datasets include initial datasets and real mixtures corresponding to multiple sediment sources under each source classification. T2. Group the initial datasets corresponding to multiple sediment sources under each source classification in the data to be evaluated obtained in T1 in a pairwise non-repeating grouping manner to form multiple grouped data under that source classification. T3. Calculate the relative deviation of each fingerprint factor in the multiple grouped data under each source classification in T2. ​​T4. Use the fingerprint attribute difference threshold determined by the quantization method provided in the first aspect of the present invention to verify the source classification reliability of all relative deviations under each source classification obtained in step T3. Wherein, when at least one of the relative deviations of fingerprint factors obtained in each grouped data under a source classification is greater than the fingerprint attribute difference threshold, the source classification reliability verification is determined to be passed. T5. Use the mass balance model to evaluate the contribution rate of each sediment source under that source classification of the data to be evaluated that passed the source classification reliability verification in step T4.

[0074] The following is a detailed description in conjunction with the accompanying drawings.

[0075] like Figure 2 As shown, the source classification within the target area includes land use, geological attributes, small watersheds, soil depth, and other source classification methods (classification types). Once the land classification type is determined, the corresponding source quantity can be determined. For example... Figure 2 The classification is based on land use. It includes source 1 to source 3.

[0076] In the sediment source tracing method, the data to be evaluated in the target area for input into the mass balance model can be obtained in the following way:

[0077] First, a detailed survey (field investigation, remote sensing, etc.) is conducted in the target study area. Based on the survey results, source and sediment samples are collected in the target study area according to multiple sources. These samples are then naturally air-dried and sieved. Fingerprint factor attribute values ​​(including concentration, isotope ratios, etc., specific values ​​need to be determined based on the measured fingerprint factors) are measured in the laboratory. Statistical analysis is performed on the obtained source and sediment sample fingerprint factor attribute data, including conservation tests, Kruskal-Wallis H tests, and multivariate discriminant analysis. This yields the fingerprint factor combination with the greatest difference in attribute values ​​among sources (i.e., the optimal fingerprint factor combination, which is also the fingerprint factor used in the contribution rate quantification method), as well as the fingerprint factor attribute values ​​in the data to be evaluated.

[0078] The process of creating the largest fingerprint factor combination described above is a technique known to those skilled in the art and will not be elaborated upon here.

[0079] It should be noted that the initial datasets corresponding to multiple sediment sources under each source category in the data to be evaluated are grouped in a pairwise non-repeating manner to form the smallest unit (sediment source group) that distinguishes sediment sources. The smallest unit is preferably two sediment sources. For example, when there are 4 sediment sources under a land use category, they can be combined in a pairwise non-repeating manner according to mathematics, resulting in 6 smallest units.

[0080] Furthermore, the relative deviation of each fingerprint factor in multiple grouped data under each source category is calculated using the method for calculating relative deviation provided in the first aspect of this invention. Further details are omitted here.

[0081] Then, the fingerprint attribute difference threshold determined by the quantification method provided in the first aspect of the present invention is used to verify the reliability of the source classification for all relative deviations under each source classification obtained in step T3; wherein, when at least one of the relative deviations of the fingerprint factors obtained in each group of data under a source classification is greater than the fingerprint attribute difference threshold, the source classification reliability verification is determined to be successful. Here, the verification of source classification reliability is to ensure that the geochemical element combination (fingerprint factor) can distinguish any two of the multiple sediment sources (sediment sources in sediment source groups).

[0082] Finally, the data to be evaluated, which has passed the source classification reliability verification, is used to assess the contribution rate of each sediment source under that source classification using a mass balance model. The mass balance model and solution method used are the same as the quantification method for fingerprint attribute difference thresholds provided in this invention, and will not be repeated here.

[0083] The following section uses the Loess Plateau as a case study for practical application. Within this Loess Plateau, watershed 1 and watershed 2 are selected, with the source classification set as land classification type. Both watersheds are illustrated using land use types. In watershed 1, the sediment sources corresponding to the land use types include four types: farmland, grassland, forest land, and gullies. In watershed 2, the sediment sources corresponding to the land use types are grassland, farmland, and forest land.

[0084] Based on the erosion characteristics of various land uses in the watershed, source samples were collected from the top 0-5 cm layer. Before collection, surface stones, debris, and other impurities were removed. Then, at each sampling point, five soil samples were randomly collected from a 5×5m quadrat using a plastic shovel. These samples were uniformly mixed to form individual samples, placed in resealable bags, and a total of 14 farmland samples, 11 grassland samples, 9 ditch wall samples, and 5 sediment samples were collected. After all samples were collected, they were brought back indoors for natural drying, followed by sample sieving in preparation for fingerprint factor determination.

[0085] In this embodiment, the fingerprint factor selected is a geochemical element. First, sample pretreatment is performed: approximately 0.3 g of the sieved sample is weighed using a 0.01 g balance and placed in a digestion tube. 9 ml of nitric acid and 3 ml of hydrochloric acid are added for digestion. After digestion, the sample is transferred to a volumetric flask and diluted to 50 ml with pure water. The diluted liquid is then filtered through filter paper (glossy side up) into a beaker. Finally, the filtered liquid is transferred to a 50 ml centrifuge tube. Next, the geochemical elements are determined using inductively coupled plasma mass spectrometry (ICP-MS), mainly including 38 elements such as Cu, Pb, Zn, Fe, Al, Mg, and Bi.

[0086] In Basin 1, the optimal fingerprint factor combination was screened: First, the optimal fingerprint factor combination was screened from 38 geochemical elements using mathematical or statistical methods (conservatism test, Kruskal-Wallis H test and multivariate discriminant analysis), namely Co, Cr, Mg, P, Sc, Ti and Y. This combination represents the largest source fingerprint attribute difference among all elements.

[0087] In watershed 2, after conservation tests, Kruskal-Wallis H tests and multivariate discriminant analysis, the elements Al, Tb, Bi, Fe, P and Sr were selected as the optimal fingerprint factor combination.

[0088] like Figure 3 As shown, Source 1 comprises multiple sediment source groups according to source differentiation rules, including six sediment source groups: forest-farmland, grassland-farmland, ... gully-grassland, etc. The relative deviations of each sediment source group under the same fingerprint factor are shown below. Figure 3 As shown, only element Mg can distinguish between grassland-farmland, grassland-woodland, and gully-grassland, while element P can distinguish between gully-farmland. The relative deviations of other fingerprint factors in geochemical attribute values ​​between any two sources are all less than 8%. However, when performing sediment source analysis, it is still impossible to completely distinguish the four source areas, indicating that classifying sources by land use in this watershed leads to lower accuracy in predicting sediment source contributions. Therefore, land use classification was excluded in watershed 1, improving the accuracy of sediment source contribution prediction.

[0089] like Figure 4 As shown in the figure, in watershed 2, the multiple sediment source groups formed according to the source differentiation rules include three sediment source groups: forest-farmland, grassland-farmland, and grassland-forestland. The relative deviations of the corresponding sediment source groups under the same fingerprint factor in each sediment source group are as follows: Figure 4 As shown, elements Fe, Sr, and Tb can completely distinguish between grassland, farmland, and forest; element Al can completely distinguish between forest-farmland and grassland-farmland; element Bi can completely distinguish between grassland-forestland and grassland-farmland; and element P can completely distinguish between forest-farmland and grassland-forestland. This indicates that the selected geochemical element combination can distinguish between any two of the three sources.

[0090] The source contribution rate was then calculated using OptQuest. The results showed that grassland contributed 33.0% of the sediment, farmland contributed 60.6%, and gullies contributed 6.4%. This result indicates that in this watershed, farmland areas should be given special attention, and targeted management of farmland soil erosion can be carried out through soil and water conservation or other watershed management measures to reduce erosion and sediment production.

[0091] Therefore, the sediment source tracing method provided by this invention can effectively eliminate unreasonable source classifications, enabling reasonable contribution rate assessments based on source classification. In other words, it provides an efficient, feasible, scientific, and rapid quantitative method for determining the reliability of sediment source classification, thereby improving the accuracy of sediment source fingerprint identification. Furthermore, this invention is applicable to all source classification methods, such as land use, geological attributes, small watersheds, and soil depth; it has no limitations on the study area scale (area), exhibiting high applicability at the scale of catchment areas, watersheds, and regions; it can select the most suitable fingerprint factors for the study area from multiple fingerprint factors based on existing source classification methods; the entire process is easy to perform, computationally simple, widely applicable, and easy to promote in the future.

[0092] It should be noted that although the steps are described in a specific order above, it does not mean that the steps must be executed in the above specific order. In fact, some of these steps can be executed concurrently, or even in a different order, as long as the required function can be achieved.

[0093] This invention can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the invention.

[0094] Computer-readable storage media can be tangible devices that hold and store instructions for use by an instruction execution device. Computer-readable storage media can be, for example, including but not limited to, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof.

[0095] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for quantifying fingerprint attribute difference thresholds, characterized in that, The fingerprint attribute difference threshold is used to characterize the reliability of identifying the source of sediment using fingerprint factors, and the quantification method includes: S1. Obtain a relative deviation dataset consisting of multiple relative deviations, wherein the relative deviations are used to represent the differences of the same fingerprint factor in different sediment sources; S2. Based on each relative deviation in the relative deviation dataset obtained in S1, configure an initial dataset corresponding to multiple sediment sources under the relative deviation condition. The initial dataset includes a fingerprint factor attribute dataset corresponding to each sediment source. The fingerprint factor attribute dataset consists of multiple fingerprint factor attribute values. S3. The initial dataset obtained in S2 under the relative deviation condition is mixed according to multiple different source ratios to obtain an initial virtual mixture corresponding to each source ratio under the relative deviation condition. The initial virtual mixture is then subjected to fingerprint attribute error fluctuation processing to obtain a virtual mixture corresponding to each source ratio under the relative deviation condition. S4. Evaluate the mean absolute error of the contribution rate under the relative deviation condition based on the virtual mixture corresponding to all source proportions under the relative deviation condition in S3. S5. Based on the average absolute error of the contribution rate under different relative deviation conditions obtained in S4, the relationship between the average absolute error of the contribution rate and the relative deviation is fitted. S6. Determine the fingerprint attribute difference threshold based on the mutual relationship and the average absolute error threshold of the contribution rate.

2. The quantization method according to claim 1, characterized in that, In S2, the initial datasets corresponding to various sediment sources under the relative deviation condition are configured as follows: S21. Obtain a preset type of sediment source and a fingerprint factor combination that matches the preset type of sediment source. S22. Randomly select one of the sediment sources in S21 and assign values ​​to the fingerprint factor attribute dataset corresponding to that sediment source; S23. Calculate the fingerprint factor attribute dataset corresponding to the remaining sediment types in the preset sediment source types using the relative deviation and the fingerprint factor attribute dataset assigned in S22.

3. The quantization method according to claim 2, characterized in that, The preset sediment source types include two types. In S23, the fingerprint factor attribute value of each fingerprint factor attribute in the fingerprint factor attribute dataset corresponding to the remaining sediment types is calculated in the following way: Where R represents the relative deviation, c1 represents the fingerprint factor attribute value of fingerprint factor c in the first sediment source, and c2 represents the fingerprint factor attribute value of fingerprint factor c in the second sediment source.

4. The quantization method according to claim 1, characterized in that, In S3, the attribute value of each fingerprint factor in the initial virtual mixture corresponding to each source ratio is calculated as follows: in, This represents the value of the fingerprint factor after mixing in the initial virtual mixture. Indicates the types and quantities of sediment sources. Indicates the first The fingerprint factor attribute value of fingerprint factor c in the source of sediment. In the mixing ratio k, the first... The proportions corresponding to the sources of sediment, among which =1.

5. The quantization method according to claim 4, characterized in that, In S3, the virtual mixture corresponding to each source ratio under relative deviation conditions includes multiple fluctuated fingerprint factor attribute values, wherein each fluctuated fingerprint factor attribute value is calculated in the following manner: in, This represents the value of the fingerprint factor after mixing in the initial virtual mixture. Due to fingerprint attribute errors, This represents the fingerprint factor attribute value after fluctuation.

6. The quantization method according to claim 1, characterized in that, S4 includes: S41. The virtual contribution rate under relative deviation conditions is evaluated using a mass balance model. The data used for the virtual contribution rate are the virtual mixture and the initial dataset corresponding to each source ratio, and the mean absolute error of the contribution rate corresponding to each source ratio is calculated based on the virtual contribution rate and the source ratio. The virtual contribution rate is obtained by solving the virtual mixture and the initial dataset using the following mass balance model: Where f is the sum of squared residuals, C i The first in the virtual mixture Fingerprint factor attribute value, S ij The first one in the initial dataset The source of soil and sand Fingerprint factor attribute value, X j For the first The virtual contribution rate of each sediment source, where m is the type of sediment source and n is the type of fingerprint factor, and m≤n; In step S41, the mean absolute error of the contribution rate corresponding to the source ratio is calculated based on the virtual contribution rate and the source ratio, using the following rules: Where MAE represents the mean absolute error of the contribution rate corresponding to the source proportion, m is the amount of sediment source, and Y j Indicates the first The proportion of each type of sediment source in the mixing ratio, X j For the first The virtual contribution rate of sediment source; S42. The average absolute error of the contribution rate corresponding to all source proportions under each relative deviation condition in S41 is taken as the average absolute error of the contribution rate under that relative deviation condition.

7. A method for tracing the source of sediment, characterized in that, The tracing method includes the following steps: T1. Obtain the data to be evaluated in the target area for input into the mass balance model. The data to be evaluated includes calibration data under multiple source classifications. The calibration data includes datasets under each source classification, wherein the datasets include initial datasets and real mixtures corresponding to multiple sediment sources under the source classification. T2. The initial datasets corresponding to multiple sediment sources under each source category in the data to be evaluated obtained in T1 are grouped in a pairwise non-repeating grouping manner to form multiple grouped data under that source category; T3. Calculate the relative deviation of each fingerprint factor in multiple grouped data under each source category in T2; T4. Using the fingerprint attribute difference threshold determined by any one of the quantization methods described in claims 1-6, perform source classification reliability verification on all relative deviations under each source classification obtained in step T3; wherein, when at least one of the relative deviations of the fingerprint factors obtained in each group of data under a source classification is greater than the fingerprint attribute difference threshold, the source classification reliability verification is determined to be passed. T5. Using the mass balance model, evaluate the contribution rate of each sediment source under the source classification that has passed the source classification reliability verification in step T4.

8. A computer-readable storage medium, characterized in that, It stores a computer program that can be executed by a processor to implement the steps of the method according to any one of claims 1-7.

9. An electronic device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, cause the electronic device to perform the steps of the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, The computer program product includes computer program code that, when run on an electronic device, causes the electronic device to perform the method as described in any one of claims 1-7.

Images