Optical fingerprint sediment source tracing method based on multi-particle-size combination optimization

Abstract

The invention discloses a multi-particle-size combined optimized optical fingerprint sediment source tracing method which comprises the following steps: dividing sediment source land types into S type and G type, collecting corresponding types of samples, performing wet screening on the samples to three particle size ranges, and measuring spectrums of the samples; performing spectrum-based PCA-LDA discriminant analysis on a sample obtained after wet screening; preparing the samples subjected to discriminant analysis into a series of mixed samples according to different combination modes and different gradient proportions, and measuring spectral data of the mixed samples for modeling; modeling by utilizing a partial least square regression method, fitting the original spectral data of a modeling set sample and the spectral data subjected to spectral preprocessing with the source sample proportion value of the corresponding sample respectively, establishing a spectral tracing model, and screening out an optimal tracing model from the spectral tracing model; and utilizing the optimal tracing model to predict and evaluate the sediment source. The accuracy of an optical fingerprint tracing result can be improved.

Description

Technical Field

[0001] The invention belongs to the field of soil erosion sediment source identification, and in particular relates to an optical fingerprint sediment source tracing method with multi-particle size combination optimization. Background Art

[0002] Identifying the source of eroded sediment is the key to understanding the occurrence and development of soil erosion and preventing and controlling soil and water loss. It plays an important reference role in the targeted management of erosion source areas and the formulation of soil and water conservation measures. Since optical characteristics can comprehensively reflect the physical and chemical properties of samples, different erosion sources will have differences in material composition, organic matter content, particle size and aggregate characteristics, and their corresponding optical characteristics will also show heterogeneity, such as the shape, position, intensity and number of characteristic peaks. Therefore, different types of soils have different spectral characteristics, and the spectral characteristics of soils of the same type with different particle size ranges are also different. Optical fingerprint tracing technology uses the optical characteristics of sediment as a fingerprint factor, and uses its significant differences in sediments from different sources, and maintains conservation during erosion and migration, to determine the source of sediment and its relative contribution. Optical fingerprint tracing technology has shown great potential in the identification of sediment sources due to its obvious advantages such as high recognition, fast speed, low cost and non-destructive samples. However, sediment is often accompanied by particle sorting during erosion and transportation, that is, coarse particles are deposited first, and fine particles are enriched. The content of organic matter and other components in fine particles is usually higher than that in coarse particles, and the corresponding components are also enriched. These effects will affect the optical characteristics, and it is impossible to ensure that the optical characteristics are conservative from source to sink, which in turn affects the accuracy of the optical fingerprint tracing results of sediment sources. The traditional sediment source optical fingerprint tracing method does not consider the impact of particle sorting, organic matter enrichment and other effects on the accuracy of the tracing results. It is usually assumed that the optical properties of sediment will not change during erosion and transportation. When modeling, only the source soil type combination is used, and the particle size combination is not considered, resulting in a low match between the optical characteristics of the source sample and the sediment sample, so it is difficult to ensure the accuracy of the tracing results. Summary of the invention

[0003] The present invention provides an optical fingerprint sediment source tracing method with multi-particle size combination optimization. On the basis of traditional sediment source sample classification, the particle size is further classified to create a more comprehensive modeling set sample in a multi-particle size combination manner to solve the problem of inaccurate optical fingerprint tracing results caused by particle size sorting and organic matter enrichment.

[0004] The technical solution adopted by the present invention is:

[0005] The present application discloses a multi-particle size combination optimized optical fingerprint sediment source tracing method, comprising:

[0006] Classification of source types: The sediment source where samples are to be collected is divided into S-type source and G-type source based on the ditch edge line; among them, the S-type source is the new loess source, and the G-type source is the old loess source;

[0007] Sample collection: Five sampling areas were selected in the S-type source area, and samples were collected in each sampling area to obtain five S-type samples, represented by S1, S2, S3, S4, and S5; five sampling areas were selected in the G-type source area, and samples were collected in each sampling area to obtain five G-type samples, represented by G1, G2, G3, G4, and G5; the five S-type samples (S1-S5) and five G-type samples (G1-G5) collected were wet-sieved to obtain a total of thirty samples in three particle size ranges: S1-31, S1-63, S1-125, S2-31, S2-63, S2-125...S5-31, S5-63, S5-125, G1-31, G1-63, G1-125, G2-31, G2-63, G2-125…G5-31, G5-63, G5-125, among which, S1-31, S1-63, S1-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of S1 type samples, S2-31, S2-63, S2-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of S2 type samples, and so on; G1-31, G1-63, G1-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of G1 type samples, and so on;

[0008] Thirty samples obtained after wet screening were used to perform spectral-based sediment source sample discriminant analysis, the spectral data of each sample was measured, and the obtained spectral data was subjected to PCA-LDA discriminant analysis to determine whether samples of different source types and different particle size ranges can be effectively distinguished. If the judgment is passed, the next step is entered; otherwise, the discriminant analysis is performed again after reclassification until the discriminant analysis is passed; wherein the spectral data includes mid-infrared spectral data, visible light spectral data or near-infrared spectral data;

[0009] After the judgment analysis is passed, the samples with the same particle size from the same source are mixed in the same mass ratio to obtain six source samples S31, S63, S125, G31, G63, and G125. These six source samples are prepared into a series of mixed samples in different combinations and different gradient proportions, and their spectral data are measured; 2 / 3 of the samples are randomly selected from the mixed samples as the modeling set samples, and the remaining 1 / 3 of the samples are used as the prediction set samples; among them, sample S31 is S1-31, S2-31, S3-31, S4-31, The S5-31 sample is mixed in a mass ratio of 1:1:1:1:1; the sample S63 is mixed in a mass ratio of 1:1:1:1:1:1 by S1-63, S2-63, S3-63, S4-63, and S5-63 samples that have passed the judgment analysis; and the same applies to the sample S125; the sample G31 is mixed in a mass ratio of 1:1:1:1:1:1 by G1-31, G2-31, G3-31, G4-31, and G5-31 samples that have passed the judgment analysis; and the same applies to the samples G63 and G125;

[0010] Preprocessing the spectral data of the modeling set samples, using partial least squares regression method to build a model, fitting the original spectral data of the modeling set samples and the spectral data after spectral preprocessing with the source sample ratio value of the corresponding sample, establishing a spectral tracing model, and screening out the best tracing model;

[0011] The optimal tracer model is used to predict and evaluate the source of sediment.

[0012] As an optional technical solution, the collection of sample S1 includes: in the first sampling area of ​​the five sampling areas divided by the S-type source, peel off the upper soil sample with a thickness of 0-20 cm on the surface as the sampling layer, and randomly collect c sub-samples within the grid area of ​​a meter × b meter in the sampling layer for uniform mixing, wherein a and b are both 4-6, and c is 8-12; remove impurities from the mixed sample, air-dry it naturally, and then sieve it through a 125μm aperture sieve to obtain an S1 type sample.

[0013] As an optional technical solution, wet screening treatment of sample S1 includes: adding ultrapure water to sample S1 and mixing evenly, then placing it into an ultrasonic disperser for dispersion, and wet screening to three particle size ranges: 0-31μm, 31-63μm, and 63-125μm. Finally, after wet screening of the S1 sample, three samples S1-31, S1-63, and S1-125 are obtained.

[0014] As an optional technical solution, before determining the spectral data of each sample, the method further includes placing the thirty samples obtained after wet screening in an oven at 40° C. for drying.

[0015] As an optional technical solution, the combination includes at least one sample from the six source samples of S31, S63, S125, G31, G63 and G125.

[0016] As an optional technical solution, the preprocessing includes at least one of convolution smoothing (SG-Smoothing), standard normalization (SNV) and baseline correction (Baseline).

[0017] As an optional technical solution, the optimal tracer model is determined by the coefficient of determination R 2 The two indicators of screening are root mean square error (RMSE).

[0018] As an optional technical solution, before using the optimal tracer model to predict and evaluate the source of sediment, it also includes: preparing target sediment samples to obtain mixed samples with different sorting degrees, and the mixed samples are used to evaluate the prediction accuracy of the optimal tracer model under different particle sorting conditions.

[0019] The beneficial effects of the present invention are as follows: in the present application, two different types of source samples, new loess and old loess, are selected, and they are further wet-screened in three particle size ranges (0-31μm, 31-63μm, 63-125μm), and the wet-screened samples are discriminated and analyzed; the samples that pass the discriminant analysis are mixed according to the same source, the same particle size range and the same mass ratio to obtain six source sample samples S31, S63, S125, G31, G63, and G125; these six source sample samples are prepared into a series of mixed samples according to different combinations and different gradient proportions, and their spectral data are measured; about 2 / 3 of the samples are randomly selected from the mixed samples as the modeling set, and the remaining about 1 / 3 of the samples are used as the prediction set , and perform different preprocessing on the spectral data of the modeling set samples; use the partial least squares regression method to build a model, fit the original spectral data of the modeling set samples and the spectral data after spectral preprocessing with the source sample ratio value of the corresponding sample, establish a spectral tracing model, and screen out the best tracing model; and use the prepared mixed samples with different particle size sorting degrees to verify the best spectral tracing model. The verification results show that the maximum value of the mean absolute error MAE between the actual value and the predicted value is 9.2%, and the average value is 4.4%, while the maximum value of the mean absolute error MAE predicted by the traditional spectral modeling method is 35.2%, and the average value is 23.2%. It can be seen that the prediction accuracy of the present application has been greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the process flow of the optical fingerprint sediment source tracing method optimized for multiple particle size combinations.

[0021] Figure 2 Mid-infrared spectra of six source samples.

[0022] Figure 3 is the linear discriminant result.

[0023] Figure 4 This is a schematic diagram of the results of the mean absolute error value of the mixed sample H1-H6 source type prediction using the method of this embodiment.

[0024] Figure 5 This is a schematic diagram of the results of the mean absolute error value of the mixed sample H1-H6 source type prediction using the traditional method.

[0025] Figure 6 This is a schematic diagram of the predicted mean absolute error (MAE) of different particle size ranges of mixed samples H1-H6 using the method of this embodiment.

[0026] Figure 7 Schematic diagram of the spectral modeling method. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0028] Before describing the present embodiment, the background and overall inventive concept of the present application are first described.

[0029] Since optical characteristics can comprehensively reflect the physical and chemical properties of samples, different erosion sources will have differences in material composition, organic matter content, particle size and aggregate characteristics, and their corresponding optical characteristics will also show heterogeneity, such as the shape, position, intensity and number of characteristic peaks. Therefore, different types of soils have different spectral characteristics, and soils of the same type with different particle size ranges also have different spectral characteristics, such as Figure 2 The mid-infrared spectral data of wavenumber-absorbance of six source samples of two sediment source types and three particle size ranges are shown in Figure 2. The six source samples have different spectral characteristics. The PCA-LDA discriminant analysis is used to distinguish their spectral data. Figure 3As shown, the six source samples are clustered separately, and there is no crossover or overlap between the different source samples, indicating that the six source samples can be effectively distinguished by using the spectrum.

[0030] Optical fingerprint tracing of sediment sources is based on optical characteristics of stable properties. It uses the significant differences in sediments from different sources and the stability during transportation to determine the sediment sources and their relative contributions. However, in reality, sediments are often accompanied by particle sorting during erosion and transportation, that is, coarse particles are deposited first during transportation, while fine particles are preferentially transported to a longer distance. The content of organic matter and other components in fine particles is usually higher than that of coarse particles, and the corresponding components are also enriched. There is often a significant difference in the particle size between the collected sediment samples and the source samples. The greater the difference, the greater the difference in optical characteristics. If only the optical characteristics of the source samples are used for modeling, without considering the effects of particle size changes, the optical characteristics of the source samples and sediment samples will have a low match, that is, the optical characteristics are unstable during erosion and transportation, thus affecting the accuracy of the optical fingerprint tracing results of sediment sources.

[0031] Based on the traditional classification of sediment source samples, the present invention further classifies them by particle size and prepares a series of mixed samples in the form of multi-particle size combinations to create a more comprehensive modeling sample set that can represent sediment samples that may be formed under different erosion conditions, thereby establishing a comprehensive spectral database for subsequent modeling and prediction. Compared with the traditional spectral tracing method, this multi-particle size combination modeling method avoids the problem of low matching of optical characteristics of source samples and sediment samples caused by particle size sorting, and provides the possibility of achieving model prediction accuracy.

[0032] Take the two sediment sources of new loess and old loess as examples. Based on the classification of the two sediment sources of new loess and old loess, they are further subdivided into three particle size ranges (0-31μm, 31-63μm, 63-125μm), a total of 6 categories. A series of mixed samples (74) are prepared according to different combinations and proportions of these six source samples, which include sample combination intervals that may be formed by particle size sorting. The partial least squares regression method is used to model the model. 2 / 3 of the samples are randomly selected from the prepared mixed samples (74) as the modeling set. The partial least squares regression method is used to model the model. The original spectral data of the modeling set samples and the spectral data after spectral pretreatment are respectively fitted with the source sample ratio values ​​of the corresponding samples to establish a spectral tracer model, and the best tracer model is selected from them. The remaining 1 / 3 samples are used as the prediction set for model verification. In order to explore whether different spectral pretreatments can improve the prediction ability of the model, it is possible to try to use spectral data with different pretreatments to establish the model. In order to simulate the sediment samples that may be formed under different erosion conditions, according to the original proportion of each particle size range in the source soil, 6 mixed sediment samples with different soil particle sorting degrees were configured, recorded as H1, H2, H3, H4, H5, and H6, and their spectral data were measured. The spectral data of the above 6 mixed sediment samples were input into the constructed model respectively to calculate the respective source proportions, which not only include the proportion of soil types, but also the proportion of different particle size ranges corresponding to different soil types. The above-mentioned relevant data statistical analysis was completed using Unscrambler X 10.4 software (CAMO Software AS., Oslo, Norway).

[0033] The traditional spectral tracing method only considers the combination of source soil types, not the combination of particle sizes. It only mixes new loess and old loess in different proportions and uses partial least squares regression to build a model. However, the particle sizes of each type of soil in the modeling set samples are all original proportions, and do not include the particle combination formed by particle size sorting. The greater the particle size sorting of the sediment sample, that is, the lower the optical feature match between the source sample and the sediment sample, the lower the accuracy of the tracing result.

[0034] Example

[0035] This embodiment discloses an optical fingerprint sediment source tracing method with multi-particle size combination optimization, including:

[0036] a. Classification of source types: The sediment source areas where samples are to be collected are divided into S-type source areas and G-type source areas based on the ditch edge line; among them, the S-type source areas are new loess source areas, and the G-type source areas are old loess source areas;

[0037] b. Sample collection: five sampling areas were selected in the S-type source area, and samples were collected in each sampling area to obtain five S-type samples, denoted by S1, S2, S3, S4, and S5; five sampling areas were selected in the G-type source area, and samples were collected in each sampling area to obtain five G-type samples, denoted by G1, G2, G3, G4, and G5; the five S-type samples (S1-S5) and five G-type samples (G1-G5) collected were wet-sieved to obtain three types of particle size ranges, a total of thirty samples S1-31, S1-63, S1-125, S2-31, S2-63, S2-125...S5-31, S5-63, S5-125, G1-31, G1-63, G1-125 , G2-31, G2-63, G2-125…G5-31, G5-63, G5-125, among which, S1-31, S1-63, S1-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of S1 type samples, S2-31, S2-63, S2-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of S2 type samples, and so on; G1-31, G1-63, G1-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of G1 type samples, and so on;

[0038] c. Perform spectral-based discriminant analysis of sediment source samples using the thirty samples obtained after wet screening, measure the spectral data of each sample, and perform PCA-LDA discriminant analysis on the obtained spectral data to determine whether samples of different source types and different particle size ranges can be effectively distinguished. If the judgment is passed, proceed to the next step, otherwise, reclassify and perform discriminant analysis again until the discriminant analysis passes; wherein the spectral data includes mid-infrared spectral data, visible light spectral data or near-infrared spectral data; reclassification refers to reclassifying S-type samples and G-type samples that cannot pass the discriminant analysis according to particle size until the PCA-LDA analysis can completely distinguish samples from different sources and with different particle sizes;

[0039] d. After the judgment analysis is passed, the samples with the same particle size from the same source are mixed in the same mass ratio to obtain six source samples S31, S63, S125, G31, G63, and G125. These six source samples are prepared into a series of mixed samples in different combinations and different gradient proportions, and their spectral data are measured; 2 / 3 of the samples are randomly selected from the mixed samples as the modeling set samples, and the remaining 1 / 3 of the samples are used as the prediction set samples; among them, sample S31 is S1-31, S2-31, S3-31, S4-31 after the judgment analysis is passed. , S5-31 samples are mixed in a mass ratio of 1:1:1:1:1; sample S63 is formed by mixing S1-63, S2-63, S3-63, S4-63, and S5-63 samples that have passed the judgment analysis in a mass ratio of 1:1:1:1:1; and the same applies to sample S125; sample G31 is formed by mixing G1-31, G2-31, G3-31, G4-31, and G5-31 samples that have passed the judgment analysis in a mass ratio of 1:1:1:1:1; and the same applies to samples G63 and G125;

[0040] e. Preprocessing the spectral data of the samples in the modeling set, using partial least squares regression method to perform modeling, fitting the original spectral data of the samples in the modeling set and the spectral data after spectral preprocessing with the source sample ratio values ​​of the corresponding samples, establishing a spectral tracing model, and screening out the best tracing model;

[0041] f. Use the optimal tracer model to predict and evaluate the source of sediment.

[0042] As an optional implementation, the collection of sample S1 includes: in the first sampling area of ​​the five sampling areas divided by the S-type source, peel off the upper soil sample with a thickness of 0-20 cm on the surface as the sampling layer, randomly collect c sub-samples in the grid area of ​​a meter × b meter in the sampling layer and mix them evenly, wherein a and b are both 4-6, c is 8-12, and in this embodiment, a and b are both 5, and c is 10; remove the impurities from the mixed sample and air dry it naturally, and then sieve it through a 125μm aperture sieve to obtain an S1 type sample. The collection method of samples S2-S5 and samples G1-G5 is consistent with that of sample S1, and they can be collected in the corresponding divided collection areas.

[0043] As an optional embodiment, the wet sieving treatment of sample S1 includes: adding ultrapure water to sample S1 and mixing evenly, then placing it in an ultrasonic disperser for dispersion, and wet sieving to three particle size ranges: 0-31 μm, 31-63 μm, and 63-125 μm. Finally, after wet sieving the S1 type sample, three samples S1-31, S1-63, and S1-125 are obtained. The wet sieving treatment process of samples S2-S5 and samples G1-G5 is the same as the wet sieving treatment process of sample S1.

[0044] As an optional implementation, before determining the spectral data of each sample, the method further includes drying the thirty samples obtained after wet screening in an oven at 40°C.

[0045] As an optional embodiment, the combination includes at least one sample from the six source samples of S31, S63, S125, G31, G63 and G125.

[0046] As an optional implementation, the preprocessing includes at least one of convolution smoothing method SG-Smoothing (SG is Savitzky-Golay), standard normalization SNV (SNV is Standard Normalized Variate) and baseline correction Baseline.

[0047] As an optional embodiment, the optimal tracer model is determined by the coefficient of determination R 2 The two indicators of screening are root mean square error (RMSE).

[0048] As an optional implementation, before using the optimal tracer model to predict and evaluate the sediment source, the method also includes: preparing target sediment samples to obtain mixed samples with different sorting degrees, and the mixed samples are used to evaluate the prediction accuracy of the optimal tracer model under different particle sorting conditions.

[0049] In order to better understand the present application, the present application is further described below.

[0050] The Shayangou watershed, a typical small watershed in the hilly and gully area of ​​the Loess Plateau with severe soil erosion, was chosen as the actual application area.

[0051] 1. Classify the source areas and collect samples. The source areas are divided into two types: new loess (S) and old loess (G) based on the ditch edge line, and source soil samples are collected. Five sampling areas are divided in the S-type source area, and one sample is collected in each sampling area to obtain samples S1-S5. Similarly, five sampling areas are divided in the G-type source area, and one sample is collected in each sampling area to obtain samples G1-G5. The method of collecting samples in each sampling area is the same. Taking sample S1 as an example, representative and easily eroded soil samples are selected in the corresponding sampling area during collection. A stainless steel shovel is used to peel off the 0-20 cm thick upper soil sample as the sampling layer. Each sample consists of 5×5m in the sampling layer. 2 The 10 sub-samples collected within the grid area were mixed evenly. The collected samples were naturally air-dried after removing animal and plant remains, gravel and other debris, and uniformly passed through a 125 μm pore size sieve to obtain sample S1.

[0052] 2. Sample wet sieving: Ultrapure water was added to the sieved samples (10 samples, S1-S5 and G1-G5) and mixed evenly, and then placed in an ultrasonic disperser for dispersion. According to the particle composition characteristics of new loess, old loess and eroded sedimentary sediment, these ten samples were wet-sieved to three particle size ranges: 0-31μm, 31-63μm, and 63-125μm, respectively, and thirty samples were obtained, represented as: S1-31, S1-63, S1-125, S2-31, S2-63, S2-125, S3-31, S3-63, S3-125, S4-31, S4-63, S4-125, S5-31, S5-63, S5-125; G1-31, G1-63, G1-125, G2-31, G2-63, G2-125, G3-31, G3-63, G3-1 25, G4-31, G4-63, G4-125, G5-31, G5-63, G5-125; among them, S1-31, S1-63, S1-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of S1 type samples, S2-31, S2-63, S2-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of S2 type samples, S3-31, S3-63, S3-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-1 25μm samples, S4-31, S4-63, S4-125 respectively represent the samples with particle sizes ranging from 0-31μm, 31-63μm, and 63-125μm obtained after wet sieving of S4 samples, S5-31, S5-63, S5-125 respectively represent the samples with particle sizes ranging from 0-31μm, 31-63μm, and 63-125μm obtained after wet sieving of S5 samples; G1-31, G1-63, G1-125 respectively represent the samples with particle sizes ranging from 0-31μm, 31-63μm, and 63-125μm obtained after wet sieving of G1 samples, G2-31, G2-63, G2-125 respectively represent the samples with particle sizes ranging from 0-31μm, 31-63μm, and 63-125μm obtained after wet sieving of G2 samples, After sieving, the samples with particle sizes ranging from 0-31μm, 31-63μm, and 63-125μm are obtained. G3-31, G3-63, and G3-125 respectively indicate that the G3 samples with particle sizes ranging from 0-31μm, 31-63μm, and 63-125μm are obtained after wet sieving. G4-31, G4-63, and G4-125 respectively indicate that the G4 samples with particle sizes ranging from 0-31μm, 31-63μm, and 63-125μm are obtained after wet sieving. G5-31, G5-63, and G5-125 respectively indicate that the G5 samples with particle sizes ranging from 0-31μm, 31-63μm, and 63-125μm are obtained after wet sieving.

[0053] It should be noted that the sample particles with a particle size range of 0-31μm in the present application do not include particles of 0μm, but include sample particles of 31μm, all sample particles of 31-63μm in the present application do not include sample particles of 31μm, but include sample particles of 63μm, and all sample particles of 63-125μm in the present application do not include sample particles of 63μm, but include sample particles of 125μm.

[0054] 3. Discriminant analysis. The thirty samples obtained after wet screening were placed in an oven at 40°C for drying, and then their spectral data were measured. The spectral data can be mid-infrared spectral data, visible light spectral data or near-infrared spectral data. In this embodiment, taking mid-infrared spectroscopy as an example, the infrared spectral data can be measured using a ThermoNicolet iS50 infrared spectrometer (ThermoNicolet, USA). The mid-infrared spectra of the six source samples are shown in the figure below. Figure 2 Thirty samples (S1-31, S1-63, S1-125, S2-31, S2-63, S2-125, S3-31, S3-63, S3-125, S4-31, S4-63, S4-125, S5-31, S5-63, S5-125, G1-31, G1-63, G1-125, G2-31, G2-63, G2-125, G3-31, G3-63, G3-125, G4-3 1. PCA-LDA analysis (PCA is principal component analysis, LDA is linear discriminant analysis) was performed on the spectral data of G4-63, G4-125, G5-31, G5-63, and G5-125. The results showed that the six source samples were clustered separately, and there was no crossover or overlap between the samples of different source samples, indicating that the spectrum can effectively distinguish the six source samples (samples of the same particle size from the same type of source are the same source samples, such as S1-31, S2-31, S3-31, S4-31, and S5-31 are the same source samples). The specific results are as follows Figure 3After the analysis was passed, samples with the same particle size from the same source were mixed in the same mass ratio to obtain six source samples S31, S63, S125, G31, G63, and G125. Among them, sample S31 is obtained by mixing S1-31, S2-31, S3-31, S4-31, and S5-31 samples that have passed the judgment analysis in a mass ratio of 1:1:1:1:1; sample S63 is obtained by mixing S1-63, S2-63, S3-63, S4-63, and S5-63 samples that have passed the judgment analysis in a mass ratio of 1:1:1:1:1; and the same applies to sample S125; sample G31 is obtained by mixing G1-31, G2-31, G3-31, G4-31, and G5-31 samples that have passed the judgment analysis in a mass ratio of 1:1:1:1:1; and the same applies to samples G63 and G125.

[0055] 4. Prepare the modeling set samples. A total of six source samples S31, S63, S125, G31, G63, and G125 with three particle size ranges from two sources were prepared into a series of mixed samples in different combinations, and the gradient ratio was set under each combination condition. The purpose of this step is to create a more comprehensive modeling sample set that can represent the sediment samples that may be formed under different erosion conditions, thereby establishing a comprehensive spectral database for subsequent modeling and prediction. This multi-particle size combination modeling method provides the basis for achieving model accuracy. The specific ratios are shown in Table 1 below.

[0056] Table 1. Preparation ratio of modeling samples

[0057]

[0058]

[0059]

[0060] 5. Model building: Use regression method to build spectral tracer model. The model building process refers to Figure 7 In the modeling process, the spectral data of the modeling sample are first preprocessed. In this application, the data after various preprocessing (SG-Smoothing, SNV and Baseline) and the original data (Raw) are used to establish the spectral data tracer model respectively in combination with the partial least squares regression method (PLSR). 2 ) and root mean square error (RMSE) were used to select the best tracing model for subsequent analysis. 2 The higher the value, the better the model fitting effect, the smaller the RMSE, and the higher the stability of the model. Raw data refers to the spectral data that has not been preprocessed.

[0061] Specifically, the root mean square error RMSE is calculated according to the following formula:

[0062]

[0063] In the formula, y i is the measured value, is the estimated value, and n is the number of samples.

[0064] Specifically, the coefficient of determination R 2 Calculated according to the following formula:

[0065]

[0066] In the formula, y i is the measured value, is an estimated value, is the mean value, and n is the number of samples.

[0067] Specifically, 50 samples (about 2 / 3 of the total samples) out of the 74 total samples in Table 1 were randomly selected as the modeling set for model establishment, and the model was validated using the leave-one-out cross-validation method. The remaining 24 samples (about 1 / 3 of the total samples) with known true proportions were used as the prediction set to verify the accuracy of the tracing results. The established tracing models were screened, and it was found that SNV combined with partial least squares regression (PLSR) was the best tracing model. The specific results are shown in Table 2.

[0068] Table 2. Prediction accuracy of different preprocessing methods combined with partial least squares regression model

[0069]

[0070] Among them, RMSE C represents the root mean square error calculated using the modeling set, R C 2 represents the coefficient of determination calculated using the modeling set; RMSE p represents the root mean square error calculated using the prediction set, R p 2 Represents the coefficient of determination calculated using the prediction set.

[0071] 6. Preparation of target sediment samples: In order to simulate the sediment samples that may be formed under different erosion conditions. According to the original proportion of each particle size range in the source soil, 6 mixed samples with different soil particle size sorting degrees were prepared, recorded as H1, H2, H3, H4, H5, and H6. The proportion of the 0-31μm part of each mixed sample was calculated according to its original proportion of each particle size range in the source soil (Table 3) and the enrichment ratio of the 31-63μm and 63-125μm parts (Table 4), and finally 6 mixed samples with different degrees of soil particle size sorting were obtained (Table 5). These 6 mixed samples are used to evaluate the prediction accuracy of the model under different particle sorting conditions. The setting of the mixed sample not only considers soil from a single source, but also considers the combination of different sources and particle sizes, thereby more comprehensively simulating the sediment mixing process under natural conditions. In this way, researchers can evaluate the robustness and accuracy of the model when dealing with complex and variable sediment samples. See the following formula for specific preparation.

[0072]

[0073] In the formula, ER PS Refers to the enrichment ratio of particles within a certain particle size range; PS sediment It refers to the mass percentage (%) of particles within a certain size range in the target sediment; PS source It refers to the mass percentage (%) of particles within a certain size range in the original soil.

[0074] Table 3. Mass proportion of new and old loess in each particle size range

[0075]

[0076] Table 4. Enrichment ratio of different particle size ranges in different mixed samples

[0077]

[0078] Table 5. Mixing ratio

[0079]

[0080]

[0081] 7. Sediment source prediction and assessment: Use the best tracer model (SNV source prediction and) to predict the proportion of source type (i.e., S type or G type) of mixed samples H1-H6, such as Figure 4As shown, the results show that the mean absolute error (MAE) of the prediction of the source type of mixed samples H1-H6 is 1.19%, 2.12%, 8.33%, 1.11%, 9.17%, and 4.40%, respectively, and the average MAE value is 4.4%. In addition, the traditional spectral modeling method was used to predict the source type: the traditional spectral tracing method only divides the source samples into two categories, S and G, and performs PCA-LDA discriminant analysis on them. Under the condition that the discrimination passes, S and G are mixed in different proportions to prepare the mixed samples, and the spectrum of the mixed sample is obtained as the modeling set. Subsequently, the spectrum is preprocessed in different ways and combined with the PLSR method to model. The particle size classification is not performed for each category before modeling. The prediction results of the traditional spectral tracing method are shown in the figure below. Figure 5 As shown, the mean absolute error (MAE) of the predictions of mixed samples H1-H6 are 13.20%, 12.44%, 24.72%, 30.54%, 22.94%, and 35.22%, respectively, and the average MAE value is 23.2%. As the degree of particle size sorting (simulated by the degree of soil particle enrichment) increases, the accuracy of identification gradually decreases, and the highest MAE value is 35.2%. The results show that the use of this embodiment to further classify the sediment source samples based on the particle size for modeling can accurately identify the source contribution of the two source types of new and old loess. As the degree of particle size sorting increases, the recognition accuracy remains at a high level and does not change significantly, indicating that sediment source tracing through the modeling method of multi-particle size combination can maintain a high accuracy at different particle size sorting degrees. In addition, the modeling method optimized by multi-particle size combination can also predict the source contribution of different particle size ranges in the sample (i.e., S31, S63, S125, G31, G63, G125, a total of 6 sources) as shown in the figure. Figure 6 As shown, the predicted mean absolute error (MAE) of the contribution of different particle size ranges of mixed samples H1-H6 are 3.17%, 5.76%, 6.74%, 3.48%, 4.21%, and 3.84%, respectively, and the average MAE value is 4.5%. This shows that the model with multi-particle size combination optimization can not only identify the different soil types of sediment sources, but also accurately predict the contribution of each particle size in different soil types. The mean absolute error (MAE) values ​​of the model tracing sediment sources are all less than 10%, indicating that the prediction results are accurate and reliable.

[0082] The above description is only a specific implementation mode of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. All technical solutions that fall within the scope defined by the claims of the present invention fall within the protection scope of the present invention.

Claims

1. The optical fingerprint sediment source tracing method based on multi-particle size combination optimization is characterized by: include: Classification of source types: The sediment source where samples are to be collected is divided into S-type source and G-type source based on the ditch edge line; among them, the S-type source is the new loess source, and the G-type source is the old loess source; Sample collection: Five sampling areas were selected in the S-type source area, and samples were collected in each sampling area to obtain five S-type samples, represented by S1, S2, S3, S4, and S5; five sampling areas were selected in the G-type source area, and samples were collected in each sampling area to obtain five G-type samples, represented by G1, G2, G3, G4, and G5; the five S-type samples (S1-S5) and five G-type samples (G1-G5) collected were wet-sieved to obtain a total of thirty samples in three particle size ranges: S1-31, S1-63, S1-125, S2-31, S2-63, S2-125...S5-31, S5-63, S5-125, G1-31, G1-63, G1-125, G2-31, G2-63, G2-125…G5-31, G5-63, G5-125, among which, S1-31, S1-63, S1-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of S1 type samples, S2-31, S2-63, S2-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of S2 type samples, and so on; G1-31, G1-63, G1-125 respectively represent the samples with particle sizes of 0-31μm, 31-63μm, 63-125μm obtained after wet sieving of G1 type samples, and so on; Thirty samples obtained after wet screening were used to perform spectral-based sediment source sample discriminant analysis, the spectral data of each sample was measured, and the obtained spectral data was subjected to PCA-LDA discriminant analysis to determine whether samples of different source types and different particle size ranges can be effectively distinguished. If the judgment is passed, the next step is entered; otherwise, the discriminant analysis is performed again after reclassification until the discriminant analysis is passed; wherein the spectral data includes mid-infrared spectral data, visible light spectral data or near-infrared spectral data; After the judgment analysis is passed, the samples with the same particle size from the same source are mixed in the same mass ratio to obtain six source samples S31, S63, S125, G31, G63, and G125. These six source samples are prepared into a series of mixed samples in different combinations and different gradient proportions, and their spectral data are measured; 2 / 3 of the samples are randomly selected from the mixed samples as the modeling set samples, and the remaining 1 / 3 of the samples are used as the prediction set samples; among them, sample S31 is S1-31, S2-31, S3-31, S4-31, The S5-31 sample is mixed in a mass ratio of 1:1:1:1:1; the sample S63 is mixed in a mass ratio of 1:1:1:1:1:1 by S1-63, S2-63, S3-63, S4-63, and S5-63 samples that have passed the judgment analysis; and the same applies to the sample S125; the sample G31 is mixed in a mass ratio of 1:1:1:1:1:1 by G1-31, G2-31, G3-31, G4-31, and G5-31 samples that have passed the judgment analysis; and the same applies to the samples G63 and G125; Preprocessing the spectral data of the modeling set samples, using partial least squares regression method to build a model, fitting the original spectral data of the modeling set samples and the spectral data after spectral preprocessing with the source sample ratio value of the corresponding sample, establishing a spectral tracing model, and screening out the best tracing model; The optimal tracer model is used to predict and evaluate the source of sediment.

2. The optical fingerprint sediment source tracing method with multi-particle size combination optimization according to claim 1 is characterized in that: The collection of sample S1 includes: In the first sampling area of ​​the five sampling areas divided by the S-shaped source area, the upper soil sample with a thickness of 0-20 cm was peeled off as the sampling layer, and c sub-samples were collected at random points in the grid area of ​​a meter × b meter in the sampling layer and mixed evenly, where a and b were both 4-6, and c was 8-12; The mixed sample was free of impurities and naturally air-dried, and then preliminarily sieved through a 125 μm pore size sieve to obtain an S1 type sample.

3. The optical fingerprint sediment source tracing method with multi-particle size combination optimization according to claim 1 is characterized in that: The wet screening treatment of sample S1 includes: adding ultrapure water to sample S1 and mixing evenly, then putting it into an ultrasonic disperser for dispersion, and wet screening to three particle size ranges: 0-31μm, 31-63μm, and 63-125μm. Finally, after wet screening of the S1 sample, three samples S1-31, S1-63, and S1-125 were obtained.

4. The optical fingerprint sediment source tracing method with multi-particle size combination optimization according to claim 1 is characterized in that: Before determining the spectral data of each sample, the method further includes placing the thirty samples obtained after wet screening in an oven at 40° C. for drying.

5. The optical fingerprint sediment source tracing method with multi-particle size combination optimization according to claim 1 is characterized in that: The combination method includes at least one sample among the six source samples S31, S63, S125, G31, G63 and G125.

6. The optical fingerprint sediment source tracing method with multi-particle size combination optimization according to claim 1 is characterized in that: The preprocessing includes at least one of convolution smoothing (SG-Smoothing), standard normalization (SNV) and baseline correction (Baseline).

7. The optical fingerprint sediment source tracing method with multi-particle size combination optimization according to claim 1 is characterized in that: The optimal tracer model is determined by the coefficient of determination R 2 The two indicators of screening are root mean square error (RMSE).

8. The optical fingerprint sediment source tracing method with multi-particle size combination optimization according to claim 1 is characterized in that: Before using the optimal tracer model to predict and evaluate the sediment source, the method also includes: preparing target sediment samples to obtain mixed samples with different sorting degrees, and the mixed samples are used to evaluate the prediction accuracy of the optimal tracer model under different particle sorting conditions.

Images