Precision evaluation method for soil moisture inversion iterative algorithm

Abstract

The invention provides a precision evaluation method for a soil moisture inversion iterative algorithm, which comprises the following steps of: firstly, analyzing the number and value range of required parameters according to application conditions of a forward model, and establishing a basic data grid covering all parameter combination conditions according to the number and value range; then calculating a remote sensing signal simulated by a model for each intersection point in the grid through a forward model, then calculating a soil moisture inversion value based on the remote sensing signal simulated by the model through an iterative algorithm for pre-performing precision evaluation, and finally performing difference calculation on an inversion result and basic soil moisture data to obtain a soil moisture inversion value. And analyzing the distribution rule of the difference value in the grid so as to evaluate the precision of the iterative algorithm. According to the method provided by the invention, the precision of the soil moisture inversion iterative algorithm under various environmental parameter combination states can be intuitively expressed, and the application range and main defects of the soil moisture inversion iterative algorithm can be more scientifically and reasonably evaluated through grading and regional statistics.

Description

Technical Field

[0001] The present invention belongs to the technical field of algorithm verification, and particularly relates to a method for evaluating the accuracy of an iterative algorithm for soil moisture inversion. Background Art

[0002] Remote sensing inversion of soil moisture is currently the main way to obtain soil moisture information, and the low accuracy of soil moisture inversion is the focus of many inversion studies. The remote sensing inversion of soil moisture consists of three parts: data, model, and algorithm. Data is the information source of soil moisture, the model is the relationship function between remote sensing information and soil moisture, and the algorithm is the solution method for constructing an equation with soil moisture as the unknown based on the former two. This indicates that the factors affecting the inversion accuracy include the measurement error of remote sensing data, the simulation error of the model, and the solution error of the algorithm. However, in fact, regarding the inversion accuracy problem, the research mainly focuses on the accuracy analysis of data and models. For the accuracy analysis of algorithms, there is a lack of research attention and a complete accuracy evaluation method has not been formed.

[0003] In the initial stage of soil moisture inversion research, the mathematical function structure of the inversion model was simple and the analytical solution method was relatively common. In the case of not considering parameter approximation in the calculation process, this algorithm would not generate a solution error, so the initial research did not pay attention to the error generated by the inversion algorithm. However, with the improvement of the inversion model, the complexity of the model function relationship has increased sharply, and most studies cannot perform the analytical solution of soil moisture. To solve this problem, an iterative algorithm has been developed. It first evaluates the deviation degree between the inversion parameters and the actual values by establishing a cost function, and then continuously adjusts the inversion parameters by taking the derivative of the cost function until the cost function reaches the minimum value. Although this algorithm avoids the analytical solution of soil moisture, it also generates the accuracy problem of the algorithm. The accuracy problem of the iterative algorithm is reflected in that the soil moisture inversion result calculated according to the algorithm is an approximate expression of the model analytical calculation result within a certain parameter range (i.e., the effective region of the iterative algorithm), while in the invalid region of the algorithm, that is, when the iteration does not converge or the iteration falls into a local optimum or saddle point, the gap between the two is large. This may be the main reason for the low accuracy of soil moisture inversion, and carrying out the accuracy analysis of the model and data cannot fundamentally solve this problem, which is the main reason for carrying out the accuracy evaluation research on the iterative algorithm. However, currently, the number of studies on the accuracy evaluation of the inversion iterative algorithm is small, and there is a lack of specific evaluation methods and evaluation systems. Therefore, there is a need for an intuitive and analyzable accuracy evaluation method for the inversion iterative algorithm. Summary of the Invention

[0004] Aiming at the problem of the lack of an accuracy evaluation method for the soil moisture inversion iterative algorithm, an intuitive and analyzable accuracy evaluation method for the iterative algorithm is proposed by combining numerical simulation and regional statistics.

[0005] The prerequisite for applying this method is to select a forward model, determine the types, numbers, and ranges of input and output elements according to the model form and application scope, establish a parameter grid covering all element combinations, then directly calculate the remote sensing signals on each grid through the forward model based on the input elements, and then combine the remote sensing signals and some input elements to perform the inversion calculation of soil moisture through the inversion iteration algorithm to be evaluated. Finally, calculate the difference between the soil moisture inversion value and the basic data, analyze the distribution law of this difference on the grid, and the accuracy and applicable scope of the iteration algorithm can be evaluated through error classification and regional statistics.

[0006] A method for evaluating the accuracy of the inversion iteration algorithm for soil moisture mainly includes the following steps:

[0007] (1) Construction of the basic parameter grid

[0008] (1.1) Analysis of input and output parameters required by the forward model

[0009] First, clarify the input and output parameters of the forward model. The input parameters of the forward model generally include radar incidence angle, surface roughness, vegetation water content, and soil moisture, etc. The output parameters are generally the dielectric parameters of the surface, such as brightness temperature, backscattering coefficient, and reflectivity, etc. Secondly, analyze the application scope of the forward model through literature research and find out the allowable parameter range of the forward model.

[0010] (1.2) Setting of parameter intervals

[0011] After determining the ranges of the input and output parameters, it is necessary to determine the intervals of the input and output parameters according to the parameter ranges. To ensure the unbiasedness of the evaluation results, uniform intervals are set for each parameter. The selection of the interval size is relatively flexible, generally set between 1 / 10 and 1 / 40 of the length of the parameter range interval.

[0012] (1.3) Structural arrangement of basic data

[0013] Arrange the arrangement of the basic data according to the results of steps (1.1) and (1.2) to ensure that the algorithm accuracy evaluation can cover all parameter combination situations. The specific arrangement method is as follows: 1) Sort the soil moisture as the inversion target from low to high to generate a rectangular grid, and each grid represents a soil moisture value; 2) Introduce the distribution of other input parameters in each soil moisture grid. If the number of other input parameters is less than or equal to 2, only horizontal and vertical rectangular distributions are required within each soil moisture grid. If the number of other input parameters is greater than 2, small grids generated by further subdivision after introducing horizontal and vertical rectangular distributions are introduced to ensure that the accuracy evaluation can cover all parameter combination situations.

[0014] (2) Calculate the remote sensing signal according to the selected forward model function

[0015] After obtaining the basic data covering all parameter combinations, the remote sensing signal can be calculated according to the functional relationship simulated by the forward model. The forward model can be summarized in the following functional form, and the finally calculated value is the remote sensing signal value simulated by the inversion model on each grid.

[0016] I = f(m s , θ, s, …) Equation (1)

[0017] Where: I represents the target simulated by the forward model, generally the reflectivity, brightness temperature or backscattering coefficient; m s is the soil volumetric water content; θ is the radar incidence angle; s is the root mean square height of the surface.

[0018] (3) Inverse the soil moisture through a certain iterative algorithm by combining remote sensing information and some parameters

[0019] After obtaining the remote sensing signal simulated by the model, the soil moisture inversion can be carried out through an iterative algorithm by combining relevant known parameters. The specific method is as follows: 1. Select the auxiliary parameters required for inversion, and most parameters can be obtained from the basic data; 2. Select the iterative algorithm whose accuracy you want to evaluate, and apply it to calculate the soil moisture inversion. Finally, the inversion value of the soil moisture on each grid is obtained, which is also the soil moisture result calculated according to the iterative algorithm.

[0020] (4) Calculate the difference between the soil moisture inversion value and the basic soil moisture data, and analyze the distribution law of the error.

[0021] Based on the calculation result of the iterative algorithm obtained in step (3), and the soil moisture in the basic parameter grid in step (1) is the soil moisture determined by the forward model. Taking the difference between the two can obtain the error of the iterative algorithm. The calculation formula is shown in Equation 2. This difference is the basic index for evaluating the accuracy of the iterative algorithm. Through the distribution trend of this difference on the grid, the accuracy of the iterative algorithm in various parameter ranges can be intuitively seen. Through magnitude division and regional statistics, the applicable range and main defects of the algorithm can also be scientifically analyzed to determine the future improvement direction of the algorithm.

[0022] MSE = m s algorithm -m s fudaental Equation (2)

[0023] Where: MSE represents the difference between the theoretical calculation value of the inversion model and the calculation value of the inversion algorithm; m s algorithm is the calculation result of the inversion algorithm in step (3); m s fudaental is the soil moisture set in the construction of the basic parameter grid.

[0024] The method provided by the present invention can scientifically evaluate the accuracy, application scope and main defects of a soil moisture inversion iterative algorithm when applied to a certain model. At the same time, the performance form of accuracy evaluation is more intuitive, which provides technical support for the improvement and verification of the soil moisture inversion iterative algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the basic process of the present invention.

[0026] Figure 2 It is a schematic diagram of the structural arrangement of the basic input and output data in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0027] The present invention will be further described in detail below with reference to the accompanying drawings:

[0028] As Figure 1 shown in the process, a method for evaluating the accuracy of a soil moisture inversion iterative algorithm includes the following steps:

[0029] (1) Construction of the basic parameter grid

[0030] (1.1) Analysis of the input and output parameters required by the forward model

[0031] First, clarify the input and output parameters of the forward model. The input parameters of the forward model generally include radar incident angle, surface roughness, vegetation water content, soil moisture, etc., and the output parameters are generally the dielectric parameters of the surface, such as brightness temperature, backscattering coefficient, reflectivity, etc. Secondly, analyze the applicable range of the forward model through literature research to find out the parameter range allowed by the forward model.

[0032] (1.2) Setting of parameter intervals

[0033] After determining the ranges of the input and output parameters, it is necessary to determine the intervals of the input and output parameters according to the parameter ranges. To ensure the unbiasedness of the evaluation results, uniform intervals are set for each parameter. The selection of the interval size is relatively flexible, generally set between 1 / 10 and 1 / 40 of the length of the parameter range interval.

[0034] (1.3) Structural arrangement of the basic data

[0035] Arrange the arrangement of the basic data according to the results of steps (1.1) and (1.2) to ensure that the algorithm accuracy evaluation can cover all parameter combination situations. The specific arrangement method is as follows: 1) Sort the soil moisture as the inversion target from low to high to generate a rectangular grid, and each grid represents a soil moisture value; 2) Introduce the distribution of other input parameters in each soil moisture grid. If the number of other input parameters is less than or equal to 2, only horizontal and vertical rectangular distributions need to be carried out within each soil moisture grid. If the number of other input parameters is greater than 2, the small grids generated after introducing the horizontal and vertical rectangular distributions are further subdivided to ensure that the accuracy evaluation can cover all parameter combination situations.

[0036] (2) Calculate the remote sensing signal according to the selected inversion model function

[0037] After obtaining the basic data covering all parameter combinations, the remote sensing signal can be calculated according to the functional relationship simulated by the forward model. The forward model can be summarized in the following functional form, and finally the remote sensing signal value simulated by the inversion model on each grid is obtained.

[0038] I = f(m s , θ, s, …) Equation (1)

[0039] In the formula: I represents the target simulated by the forward model, generally the reflectivity, brightness temperature or backscattering coefficient; m s is the soil volumetric water content; θ is the radar incident angle; s is the root mean square height of the surface.

[0040] (3) Invert the soil moisture through a certain iterative algorithm by combining the remote sensing information and some parameters

[0041] After obtaining the remote sensing signal simulated by the model, the soil moisture inversion can be carried out through an iterative algorithm by combining relevant known parameters. The specific method is as follows: 1. Select the auxiliary parameters required for inversion, and most parameters can be obtained from the basic data; 2. Select the iterative algorithm whose accuracy you want to evaluate and apply it to calculate the soil moisture inversion. Finally, the inversion value of the soil moisture in each grid is obtained, which is also the soil moisture result calculated according to the iterative algorithm.

[0042] (4) Calculate the difference between the soil moisture inversion value and the basic soil moisture data, and analyze the distribution law of the difference.

[0043] Based on step (3), the calculation result of the iterative algorithm is obtained. The soil moisture in the basic parameter grid in step (1) is the soil moisture determined by the forward model. The difference between the two can obtain the error of the iterative algorithm. The calculation formula is shown in Equation 2. This difference is the basic index for evaluating the accuracy of the iterative algorithm. Through the distribution trend of this difference on the grid, the accuracy of the iterative algorithm within various parameter ranges can be intuitively seen. Through magnitude division and regional statistics, the applicable range and main defects of the algorithm can also be scientifically analyzed to determine the future improvement direction of the algorithm.

[0044] MSE = m s algorithm -m s fudaental Equation (2)

[0045] In the formula: MSE represents the difference between the theoretical calculation value of the inversion model and the calculation value of the inversion algorithm; m s algorithm is the calculation result of the iterative algorithm in step (3); m s fudaental is the soil moisture set in the construction of the basic parameter grid.

Claims

1. A method for evaluating the accuracy of an iterative soil moisture inversion algorithm, characterized in that: The following steps are involved: (1) Analyze the number and range of input and output parameters according to the application conditions of the forward model, and establish a basic data grid covering all parameter combinations; (2) Calculate the remote sensing signal of each grid based on the functional relationship determined by the forward model based on the basic data; (3) Calculate the soil moisture inversion value through an iterative algorithm with preliminary accuracy assessment; (4) Calculate the difference between the inverted value and the basic soil moisture data, analyze the distribution pattern of the difference on the grid, and evaluate the accuracy of the soil moisture algorithm.

2. The method for evaluating the accuracy of the soil moisture inversion iterative algorithm according to claim 1, characterized in that: Step (1) specifically includes the following sub-steps: (1.1) Analysis of the input and output parameters required for the forward model; First, the input and output parameters of the forward model are clarified. The input parameters of the forward model include radar incident angle, surface roughness, vegetation moisture content and soil moisture, and the output parameters are the dielectric parameters, brightness temperature, backscattering coefficient and reflectivity of the surface. Secondly, the scope of application of the forward model is analyzed through literature survey to find out the parameter range allowed by the forward model. (1.2) Parameter interval setting; After determining the range of input and output parameters, the intervals of input and output parameters are determined according to the parameter ranges; to ensure the unbiasedness of the evaluation results, each parameter is set at a uniform interval; the interval size is set between 1 / 10 and 1 / 40 of the length of the parameter range; (1.3) Structural arrangement of basic data The basic data are arranged according to the results of steps (1.1) and (1.2) to ensure that the algorithm accuracy evaluation can cover all parameter combinations. The specific arrangement is as follows: 1) Sort the soil moisture as the inversion target from low to high to generate a rectangular grid, and each grid represents a soil moisture value; 2) Introduce the distribution of other input parameters in each soil moisture grid. If the number of other input parameters is less than or equal to 2, it is only necessary to perform horizontal and vertical rectangular distribution in each soil moisture grid. If the number of other input parameters is greater than 2, then introduce horizontal and vertical rectangular distribution and continue to subdivide the generated small grid to ensure that the accuracy evaluation can cover all parameter combinations.

3. The method for evaluating the accuracy of the soil moisture inversion iterative algorithm according to claim 1, characterized in that: Step (2) specifically includes the following sub-steps: After obtaining the basic data covering all parameter combinations, the remote sensing signal is calculated according to the functional relationship simulated by the forward model. The forward model is summarized in the following functional form. The final result obtained after calculation is the remote sensing signal value simulated by the inversion model on each grid; I=f(m s ,θ,s,…) Formula (1) Where: I represents the target simulated by the forward model, which is reflectivity, brightness temperature or backscattering coefficient; m s is the volumetric water content of the soil; θ is the radar incident angle; s is the root mean square height of the ground surface.

4. The method for evaluating the accuracy of the soil moisture inversion iterative algorithm according to claim 1, characterized in that: Step (3) specifically includes the following sub-steps: (3.1) Select the auxiliary parameters required for inversion and obtain them from the basic data; (3.2) Select the iterative algorithm whose accuracy you want to evaluate and apply it to the soil moisture inversion calculation. Finally, the inverted value of soil moisture for each grid is obtained, which is the soil moisture result calculated based on the iterative algorithm.

5. The accuracy evaluation method for soil moisture inversion iterative algorithm according to claim 1 is characterized in that: Step (4) specifically includes the following sub-steps: Based on step (3), the calculation result of the iterative algorithm is obtained, and the soil moisture in the basic parameter grid in step (1) is the soil moisture determined by the forward model. The difference between the two is the error of the iterative algorithm. The calculation formula is shown in Formula 2: MSE=m s algorithm -m s fudaental Formula (2) Where: MSE represents the difference between the theoretical calculation value of the inversion model and the calculation value of the iterative algorithm; m s algorithm is the result of the iterative algorithm calculation in step (2); m s fudaental It is the soil moisture set in the basic parameter grid construction.

Images