Soil organic matter and iron oxide joint inversion method based on radiation transfer model

Abstract

The invention relates to the technical field of soil organic matter detection, in particular to a soil organic matter and iron oxide joint inversion method based on a radiation transfer model. Comprising the following steps: S1, collecting a soil sample, and obtaining reflectivity spectrum data of the soil sample; s2, constructing a soil two-factor radiation transmission model; comprising the following sub-steps: establishing a relationship among the conversion reflectivity r, the soil organic matter content theta SOM and the soil iron oxide content theta SIO; establishing a relationship between the reflectivity R and the conversion reflectivity r by using a KM model to obtain a relationship among the reflectivity R, the soil organic matter content theta SOM and the soil iron oxide content theta SIO, and constructing a soil two-factor radiation transmission model; and S3, based on the soil two-factor radiation transmission model, performing joint inversion on the soil organic matter and iron oxide content by adopting a Bayesian method. The method has the advantages that the soil two-factor radiation transmission model has extremely high accuracy in the aspect of simulating spectral reflectivity; the problem of ill-conditioned inversion is solved by a combined inversion method.

Description

Technical Field

[0001] The present invention relates to the technical field of soil organic matter detection, and in particular to a joint inversion method of soil organic matter and iron oxides based on a radiation transmission model. Background Art

[0002] How to accurately and quickly analyze the physical and chemical properties of soil has always been a key issue that researchers have been working hard to explore. In recent years, the emergence of spectral remote sensing technology has provided a very potential solution to this problem. This technology can capture the reflectance information of the soil, reveal the intrinsic characteristics of the soil without loss, and has the characteristics of large-scale, strong timeliness and high efficiency.

[0003] However, there are still some problems with spectral remote sensing technology, which can be roughly summarized into the following categories:

[0004] Spectral technology mainly relies on statistical modeling methods when inverting soil composition. A significant problem with this type of method is that it is highly dependent on data, which means that the statistical model established in one area is often difficult to be effectively applied in other areas and has obvious limitations.

[0005] In addition, the spectral reflectance of soil is affected by many factors and involves the interaction of multiple substances and physical processes such as reflection, scattering, and absorption. The existing soil radiation transmission model only considers a single factor and ignores the complex composition of the soil, making it impossible to improve the accuracy of the model. The present invention intends to consider the comprehensive influence of soil organic matter and iron oxides on the spectral reflectance, construct a soil dual-factor radiation transmission model, and provide a guarantee for the subsequent joint inversion of soil components.

[0006] Currently, there is a lack of research on soil multi-component joint inversion, and there are pathological problems in soil multi-component inversion. How to solve the pathological inversion problem is a hot topic in the field of soil remote sensing inversion. Therefore, it is necessary to develop a soil multi-component joint inversion method based on the soil multi-factor influence radiation transfer model. Summary of the invention

[0007] In order to solve the above problems, the present invention provides a joint inversion method of soil organic matter and iron oxides based on a radiation transfer model.

[0008] The present invention aims to provide a joint inversion method of soil organic matter and iron oxides based on a radiation transfer model, which specifically comprises the following steps:

[0009] S1. Collect soil samples and obtain reflectance spectrum data of the soil samples;

[0010] S2. Constructing a soil dual-factor radiation transfer model based on the reflectance spectrum data; comprising the following sub-steps:

[0011] S21. Establish conversion reflectance r, soil organic matter content θ SOM and soil iron oxide contentθ SIO relationship;

[0012] S22. Use the KM model to establish the relationship between the reflectivity R and the converted reflectivity r, and obtain the reflectivity R, the soil organic matter content θ SOM and the soil iron oxide content θ SIO The relationship between soil radiation and soil radiation is constructed.

[0013] S3. Based on the soil two-factor radiation transfer model, the Bayesian method is used to jointly invert the soil organic matter and iron oxide contents.

[0014] Preferably, the conversion reflectance r, soil organic matter content θ SOM and soil iron oxide contentθ SIO The expression of the relationship is as follows:

[0015]

[0016] in:

[0017]

[0018] Where s1, k1 and R1 are the organic matter and iron oxide contents, respectively. SOM1 and θ SIO1 The scattering coefficient, absorption coefficient and reflectivity of the soil at the time of R i represents the Fresnel reflectivity of incident light from air to the soil surface; a1 represents the organic matter absorption coefficient k SOM The ratio of the soil scattering coefficient s1 with known organic matter and iron oxide content; a2 represents the iron oxide absorption coefficient k SIO The ratio of the scattering coefficient s1 of soil with known organic matter and iron oxide content; a3 represents the scattering coefficient s SOM The ratio of the soil scattering coefficient s1 with known organic matter and iron oxide content; a4 represents the iron oxide scattering coefficient s SIO Ratio to the scattering coefficient s1 of soils with known organic matter and iron oxide content.

[0019] Preferably, step S3 includes the following sub-steps:

[0020] S31. Determine the characteristic wavelength and divide the training set and the test set according to the ratio;

[0021] S32. Based on the training set data of characteristic wavelengths, a genetic optimization algorithm is used to obtain the prior distribution information of the model parameters, and the prior distribution information is input into the Bayesian model; the prior distribution information is a prior distribution function including a mean and a covariance matrix;

[0022] S33. Based on the unknown parameter φ in the prior distribution function, construct a joint posterior probability density function;

[0023] S34. Use the joint posterior probability density function to estimate the soil organic matter and iron oxide contents of the verification samples, and obtain the point estimates of the soil organic matter and iron oxide contents of the verification samples.

[0024] Preferably, the number of characteristic wavelengths is 20.

[0025] Preferably, the characteristic wavelength is determined using a cubic monotone Hermite interpolation method.

[0026] Preferably, the expression of the prior distribution function P(φ) is as follows:

[0027] P(φ)=P(θ)P(a);

[0028] In the formula, θ represents the variable of soil properties, including soil organic matter content θ SOM and soil iron oxide content; a represents the optical coefficient variable, including the organic matter absorption coefficient k SOM The ratio of the soil scattering coefficient s1 with known organic matter and iron oxide content, a1, and the iron oxide absorption coefficient k SIO The ratio a2 of the soil scattering coefficient s1 with known organic matter and iron oxide content, the organic matter scattering coefficient s SOM The ratio a3 of the soil scattering coefficient s1 with known organic matter and iron oxide content and the iron oxide scattering coefficient s SIO The ratio a4 to the scattering coefficient s1 of soil with known organic matter and iron oxide content.

[0029] Preferably, the joint posterior probability density function expression is as follows:

[0030]

[0031] Where P(R) is the normalization constant, P(ε) represents the prior distribution function of the model error, and P(R|φ,ε) is the likelihood function. The expression of P(R|φ,ε) is:

[0032]

[0033] Among them, R i represents the measured reflectivity in the ith band, R φ,represents the soil reflectivity simulated in the ith band as variable φ input into the soil two-factor radiative transfer model, σ 2 is the variance.

[0034] Preferably, step S2 also includes model performance evaluation, and the specific method is as follows: using the least squares algorithm to obtain the unknown optical parameters a1, a2, a3 and a4, and using the absolute coefficient and root mean square error as two indicators to evaluate the estimation performance of the soil dual-factor radiation transfer model.

[0035] Preferably, the step S1 specifically includes the following sub-steps:

[0036] S11. Preheat the ground object spectrometer and perform whiteboard calibration in a dark room;

[0037] S12. Use a ground object spectrometer to measure the reflectance spectrum of the soil sample, collect multiple spectra for each soil sample, and take the arithmetic mean of the multiple spectral curves of the soil sample as the reflectance spectrum data.

[0038] Preferably, the training set and the test set are divided in a ratio of 7:3 using the KS algorithm.

[0039] Compared with the prior art, the present invention can achieve the following beneficial effects:

[0040] The present invention intends to combine the Bayesian method and the soil two-factor radiation transfer model to realize the joint inversion of soil organic matter and iron oxides and solve the problem of ill-posed inversion: (1) Construct a soil two-factor radiation transfer model that comprehensively considers the effects of soil organic matter and iron oxides on spectral reflectance. The model has extremely high accuracy in simulating spectral reflectance, providing a guarantee for subsequent soil component inversion applications; (2) A soil organic matter and iron oxide joint inversion method combining the Bayesian method and the soil two-factor radiation transfer model is proposed to solve the problem of ill-posed inversion and provide a new solution for future soil component inversion work. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 1 is a scatter plot of the spectral reflectance simulated by the soil dual-factor radiation transfer model provided by an embodiment of the present invention on a training set and a test set and the measured reflectance; (a) is the training set; (b) is the test set.

[0042] Figure 2 It is a scatter plot of the actual soil organic matter SOM and the estimated soil organic matter SOM obtained by using Bayesian estimation according to an embodiment of the present invention.

[0043] Figure 3 3 is a scatter plot of the actual soil iron oxide SIO and the estimated soil iron oxide SIO obtained by using Bayesian estimation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the following description, the same modules are represented by the same reference numerals. In the case of the same reference numerals, their names and functions are also the same. Therefore, the detailed description thereof will not be repeated.

[0045] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not constitute a limitation of the present invention.

[0046] The present invention provides a joint inversion method of soil organic matter and iron oxides based on a radiation transfer model, which specifically comprises the following steps:

[0047] S1. Collect soil samples and obtain reflectance spectrum data of the soil samples; specifically, the steps include:

[0048] S11. Preheat the ground object spectrometer and perform whiteboard calibration in a dark room;

[0049] S12. Use a ground object spectrometer to measure the reflectance spectrum of the soil sample, collect multiple spectra for each soil sample, and take the arithmetic mean of the multiple spectral curves of the soil sample as the reflectance spectrum data.

[0050] S2. Constructing a soil dual-factor radiation transfer model based on the reflectance spectrum data; specifically comprising the following sub-steps:

[0051] S21. Establish conversion reflectance r, soil organic matter content θ SOM and soil iron oxide contentθ SIO The relationship between; the expression is as follows:

[0052]

[0053] in:

[0054]

[0055] Where s1, k1 and R1 are the organic matter and iron oxide contents, respectively. SOM1 and θ SIO1 The scattering coefficient, absorption coefficient and reflectivity of the soil at the time of R i represents the Fresnel reflectivity of incident light from air to the soil surface; a1 represents the organic matter absorption coefficient k SOM The ratio of the soil scattering coefficient s1 with known organic matter and iron oxide content; a2 represents the iron oxide absorption coefficient k SIOThe ratio of the scattering coefficient s1 of soil with known organic matter and iron oxide content; a3 represents the scattering coefficient s SOM The ratio of the soil scattering coefficient s1 with known organic matter and iron oxide content; a4 represents the iron oxide scattering coefficient s SIO Ratio of the scattering coefficient s1 of soil with known organic matter and iron oxide content.

[0056] The specific process is as follows: The absorption coefficient k and scattering coefficient s of the mixed material can be regarded as the linear weighted sum of the absorption coefficient and scattering coefficient of its components. Based on this assumption, k and s on the soil surface can be described as:

[0057] k=k SOM θ SOM +k SIO θ SIO +k Others (1-θ SOM -θ SIO )

[0058] s=s SOM θ SOM +s SIO θ SIO +s Others (1-θ SOM -θ SIO );

[0059] In the formula, SOM represents soil organic matter, SIO represents soil iron oxides, θ SOM , k SOM and SOM represent soil organic matter content, soil organic matter absorption coefficient and soil organic matter scattering coefficient, respectively, SIO , k SIO and SiO They represent soil iron oxide content, soil iron oxide absorption coefficient and soil iron oxide scattering coefficient, k Others and Others They represent the sum of the absorption coefficients and the sum of the scattering coefficients of the remaining components except SOM and SIO, respectively.

[0060] When the organic matter and iron oxide contents of a soil sample are known to be θ SOM1 and θ SIO1 When , the absorption coefficient k1 and scattering coefficient s1 of this soil are described as follows:

[0061] k1=k SOM θ SOM1 +k SIO θ SIO1 +k Others (1-θ SOM1 -θ SIO1 )

[0062] s1=s SOM θ SOM1 +s SIO θ SIO1 +s Others (1-θ SOM1 -θ SIO1 );

[0063] The ratio r of k and s can be expressed as:

[0064]

[0065] in:

[0066]

[0067] S22. The relationship between reflectance R and converted reflectance r was established using the Kubleka-Munk (KM) model to obtain reflectance R and soil organic matter content θ SOM and soil iron oxide contentθ SIO The relationship between soil radiation and soil radiation is constructed.

[0068]

[0069]

[0070] Among them, R i Represents the Fresnel reflectivity of incident light from air to the soil surface, and the expression is as follows:

[0071]

[0072] Among them, n soil and n air represent the refractive index of dry soil and the refractive index of air respectively.

[0073] Step S2 also includes model performance evaluation, the specific method is as follows:

[0074] S23. Obtain unknown optical parameters (a1, a2, a3, a4) using the least squares algorithm and use the absolute coefficient (R 2 ) and root mean square error (RMSE) were used to evaluate the estimation performance of the soil dual-factor radiative transfer model.

[0075] S3. Based on the soil dual-factor radiation transfer model, the Bayesian method is used to jointly invert the soil organic matter and iron oxide content; specifically, it includes the following sub-steps:

[0076] S31. Determine the characteristic wavelength and divide the training set and the test set according to the ratio;

[0077] Specifically, the cubic monotone Hermite interpolation method is used to determine 20 characteristic wavelengths; the KS algorithm is used to divide the training set and the test set into a ratio of 7:3.

[0078] S32. Based on the training set data of the characteristic wavelength, the mean and covariance matrix of the model parameters are obtained by using a genetic optimization algorithm; the prior distribution function P(φ) of the model parameters is constructed based on the mean and covariance matrix of each model parameter;

[0079] Specifically, the genetic optimization algorithm was used to obtain a1, a2, a3 and a4 (organic matter absorption coefficient k SOM The ratio of the soil scattering coefficient s1 with known organic matter and iron oxide content, a1, and the iron oxide absorption coefficient k SIO The ratio a2 of the soil scattering coefficient s1 with known organic matter and iron oxide content, the organic matter scattering coefficient s SOM The ratio a3 of the soil scattering coefficient s1 with known organic matter and iron oxide content and the iron oxide scattering coefficient s SIO The objective function of the genetic optimization algorithm is set to minimize the root mean square error RMSE between the measured reflectivity and the simulated reflectivity. SOM ,θ SIO , the mean and covariance matrix of a1, a2, a3 and a4, and bring them into the joint Gaussian distribution function to obtain the prior distribution function P(φ) of the model parameters, which is expressed as follows:

[0080] P(φ)=P(θ)P(a);

[0081] In the formula, θ represents the variable of soil properties (including soil organic matter content θ SOM and soil iron oxide contentθ SIO , θ=[θ SOM ,θ SIO ]), a represents the optical coefficient variable (a=[a1,a2,a3,a4]);

[0082] In this step, the six unknown parameters φ in the model are divided into two categories: one is the variable θ = [θ SOM ,θ SIO ]; the other is the optical coefficient variable a = [a1, a2, a3, a4]; assuming that the soil attribute variable θ and the optical coefficient variable a are independent of each other, they can be expressed as:

[0083] P(φ)=P(θ)P(a);

[0084] The optical coefficient variables a1, a2, a3, and a4 are high-dimensional parameters related to the wavelength. The high dimensionality of the variables will significantly increase the difficulty of solving ill-conditioned problems. In the prior art, there are usually more than 2000 wavelengths. In order to alleviate the ill-conditioned problem, the cubic monotone Hermite interpolation method is used to determine 20 characteristic wavelengths.

[0085] S33. Based on the unknown parameter φ in the prior distribution function, a joint posterior probability density function is constructed; considering the cumulative error ε introduced by the model and actual observations, the joint posterior probability density function expression of the unknown parameter φ for a given soil reflectivity R is as follows:

[0086]

[0087] Where P(R) is the normalization constant, P(ε) represents the prior distribution function of the model error (which follows the principle of taking 0 as the mean and σ 2 is the normal distribution of variance, variance σ 2 Following the uninformative inverse gamma prior distribution, which is conjugate with the normal distribution of the error), P(R|φ,ε) is the likelihood function, which is expressed as:

[0088]

[0089] Among them, R i represents the measured reflectivity in the ith band, R φ, represents the simulated soil reflectivity in the ith band when input into the soil two-factor radiative transfer model as variable φ.

[0090] S34. estimating the soil organic matter and iron oxide contents of the verification sample using the joint posterior probability density function to obtain point estimates of the soil organic matter and iron oxide contents of the verification sample;

[0091] Specifically, the Markov chain Monte Carlo method is used to calculate the maximum posterior probability as the point estimate of the soil attribute variable inversion; that is, under given conditions, find the parameter value that maximizes the marginal posterior probability, which can be expressed as:

[0092]

[0093] In addition, 95% confidence intervals of the posterior distributions were calculated to reflect the uncertainty in the variable estimates.

[0094] Figure 1 The scatter plots of the simulated spectral reflectance and measured reflectance of the soil dual-factor radiation transfer model on the training set and the test set are shown. It can be clearly seen from the figure that most of the points are closely around the 1:1 line, the RMSE of the training set is 0.0286, and R 2 =0.967, RMSE of the test set =0.0324, R2 =0.959. This result shows that the soil dual-factor radiation transfer model can accurately simulate the spectral reflectance of the soil.

[0095] Figure 2 and Figure 3 The scatter plots of the predicted and measured values ​​of soil organic matter (SOM) and soil iron oxide (SIO) are shown. The results show that the root mean square error (RMSE) values ​​of SOM and SIO obtained by Bayesian estimation are 3.196 and 2.556, respectively, and the predicted R p 2 The values ​​are 0.808 and 0.759 respectively. All SOM measured values ​​fall within the 95% confidence interval (HDI), and 97.1% of SIO measured values ​​fall within the 95% HDI interval.

[0096] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the disclosure of the present invention can be performed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and this document does not limit this.

[0097] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A joint inversion method for soil organic matter and iron oxides based on a radiation transfer model, characterized in that: The specific steps include: S1. Collect soil samples and obtain reflectance spectrum data of the soil samples; S2. Constructing a soil dual-factor radiation transfer model based on the reflectance spectrum data; comprising the following sub-steps: S21. Establish conversion reflectance r, soil organic matter content θ SOM and soil iron oxide content θ SIO relationship; S22. Use the KM model to establish the relationship between the reflectivity R and the converted reflectivity r, and obtain the reflectivity R, the soil organic matter content θ SOM and the soil iron oxide content θ SIO The relationship between soil radiation and soil radiation is constructed. S3. Based on the soil two-factor radiation transfer model, the Bayesian method is used to jointly invert the soil organic matter and iron oxide contents.

2. The method for joint inversion of soil organic matter and iron oxides based on a radiation transfer model according to claim 1, characterized in that: The conversion reflectivity r, soil organic matter content θ SOM and soil iron oxide content θ SIO The expression of the relationship is as follows: in: Where s1, k1 and R1 are the organic matter and iron oxide contents, respectively. SOM1 and θ SIO1 The scattering coefficient, absorption coefficient and reflectivity of the soil at the time of R i represents the Fresnel reflectivity of incident light from air to the soil surface; a1 represents the organic matter absorption coefficient k SOM The ratio of the soil scattering coefficient s1 with known organic matter and iron oxide content; a2 represents the iron oxide absorption coefficient k SIO The ratio of the scattering coefficient s1 of soil with known organic matter and iron oxide content; a3 represents the scattering coefficient s SOM The ratio of the soil scattering coefficient s1 with known organic matter and iron oxide content; a4 represents the iron oxide scattering coefficient s SIO Ratio of the scattering coefficient s1 of soil with known organic matter and iron oxide content.

3. The method for joint inversion of soil organic matter and iron oxides based on a radiation transfer model according to claim 2, characterized in that: The step S3 includes the following sub-steps: S31. Determine the characteristic wavelength and divide the training set and the test set according to the ratio; S32. Based on the training set data of characteristic wavelengths, a genetic optimization algorithm is used to obtain the prior distribution information of the model parameters, and the prior distribution information is input into the Bayesian model; the prior distribution information is a prior distribution function including a mean and a covariance matrix; S33. Based on the unknown parameter φ in the prior distribution function, construct a joint posterior probability density function; S34. Use the joint posterior probability density function to estimate the soil organic matter and iron oxide contents of the verification samples, and obtain the point estimates of the soil organic matter and iron oxide contents of the verification samples.

4. The method for joint inversion of soil organic matter and iron oxides based on a radiation transfer model according to claim 3, characterized in that: The number of the characteristic wavelengths is 20.

5. The method for joint inversion of soil organic matter and iron oxides based on a radiation transfer model according to claim 4, characterized in that: The characteristic wavelength is determined by using a cubic monotone Hermite interpolation method.

6. The method for joint inversion of soil organic matter and iron oxides based on a radiation transfer model according to claim 5, characterized in that: The expression of the prior distribution function P(φ) is as follows: P(φ)=P(θ)P(a); In the formula, θ represents the variable of soil properties, including soil organic matter content θ SOM and soil iron oxide content θ SIO ; a represents the optical coefficient variable, including the organic matter absorption coefficient k SOM The ratio of the soil scattering coefficient s1 with known organic matter and iron oxide content, a1, and the iron oxide absorption coefficient k SIO The ratio a2 of the soil scattering coefficient s1 with known organic matter and iron oxide content, the organic matter scattering coefficient s SOM The ratio a3 of the soil scattering coefficient s1 with known organic matter and iron oxide content and the iron oxide scattering coefficient s SIO The ratio a4 to the scattering coefficient s1 of soil with known organic matter and iron oxide content.

7. The method for joint inversion of soil organic matter and iron oxides based on a radiation transfer model according to claim 6, characterized in that: The joint posterior probability density function expression is as follows: Where P(R) is the normalization constant, P(ε) represents the prior distribution function of the model error, and P(R|φ,ε) is the likelihood function. The expression of P(R|φ,ε) is: Among them, R i represents the measured reflectivity in the ith band, R φ,i represents the soil reflectivity simulated in the ith band as variable φ input into the soil two-factor radiative transfer model, σ 2 is the variance.

8. The method for joint inversion of soil organic matter and iron oxides based on a radiation transfer model according to claim 7, characterized in that: The step S2 also includes model performance evaluation, and the specific method is as follows: using the least squares algorithm to obtain the unknown optical parameters a1, a2, a3 and a4, and using the absolute coefficient and root mean square error as two indicators to evaluate the estimation performance of the soil dual-factor radiation transfer model.

9. The method for joint inversion of soil organic matter and iron oxides based on a radiation transfer model according to claim 8, characterized in that: The step S1 specifically includes the following sub-steps: S11. Preheat the ground object spectrometer and perform whiteboard calibration in a dark room; S12. Use a ground object spectrometer to measure the reflectance spectrum of the soil sample, collect multiple spectra for each soil sample, and take the arithmetic mean of the multiple spectral curves of the soil sample as the reflectance spectrum data.

10. The method for joint inversion of soil organic matter and iron oxides based on a radiation transfer model according to claim 9, characterized in that: The training set and test set are divided into 7:3 ratios using the KS algorithm.

Images