A method for soil moisture inversion by combining active and passive microwave

Abstract

The application discloses a kind of methods for soil moisture inversion of active and passive microwave combination, according to Mironov model, Zhao model, ω-τ model, Oh model, water cloud model forward model jointly simulating passive microwave brightness temperature and active microwave backscattering, by constructing active microwave backscattering and passive microwave brightness temperature observation and its forward model simulation value between active and passive joint cost function, utilize the meta-heuristic optimization algorithm based on Newton-Raphson method minimizes the cost function to give the best 1km scale soil moisture estimation value under the heterogeneous underlying surface;By introducing adaptive weight mechanism, according to the normalized minimum simulation error of active and passive channels, dynamic weight allocation is realized, the effective characterization of the spatiotemporal dynamic change of soil moisture is realized, and the precision and robustness of soil moisture inversion under complex surface, vegetation state and seasonal change environment are improved.

Description

Technical Field

[0001] This invention relates to a method for inverting soil moisture, and more particularly to a method for inverting soil moisture using a combination of active and passive microwave methods. Background Technology

[0002] Soil moisture is a key parameter in Earth system science, playing a crucial role in agricultural production, water resource management, and disaster early warning. Compared to visible-near-infrared and thermal infrared remote sensing, which are easily affected by clouds and rain, microwave remote sensing, with its all-weather, all-time capability and sensitivity to soil moisture, has become an important tool for soil moisture retrieval. Passive microwave remote sensing is sensitive to soil moisture, has relatively mature retrieval algorithms, and achieves high accuracy, but its spatial resolution is relatively low, typically at a scale of 10 km. Active microwave remote sensing has higher spatial resolution, reaching a scale of 10 m, but is easily affected by soil roughness and vegetation parameters, leading to lower accuracy in soil moisture retrieval. Therefore, combining the advantages of both active and passive microwave remote sensing to obtain high spatial resolution and high accuracy soil moisture data has become a research challenge.

[0003] Currently, the combined active and passive microwave inversion schemes for soil moisture retrieval are mainly divided into three categories: The first type involves using active microwave data to downscale the brightness temperature observed by passive microwave sensors with low spatial resolution, obtaining microwave brightness with the same spatial resolution as the active microwave data. Then, based on this brightness temperature data, high-resolution soil moisture is estimated using existing passive microwave soil moisture inversion algorithms.

[0004] The second type involves first using existing algorithms to retrieve low spatial resolution soil moisture using passive microwave brightness temperature, and then using the backscattering coefficient of active microwave observations, which has a relatively high spatial resolution, to downscale the passive microwave soil moisture retrieval results, thus obtaining soil moisture with a higher spatial resolution.

[0005] The third type utilizes both the backscattering coefficient from active microwave observations and the brightness temperature from passive microwave observations. Taking advantage of the high spatial resolution of active microwave and the sensitivity of passive microwave to soil moisture, it achieves joint inversion of soil moisture with high spatial resolution based on model-driven (radiative transfer mechanism model) or data-driven (regression, machine learning, deep learning, etc.) methods.

[0006] The above solution has the following technical defects: 1) Whether it is the first type of "downscaling first and then inversion" or the second type of "inversion first and then downscaling", these methods are essentially the "serialization" of observation data from active microwave and passive microwave sensors. The final soil moisture estimation accuracy is affected by the accuracy of the downscaling method, which can easily lead to the propagation of errors.

[0007] 2) Data-driven active-passive microwave soil moisture estimation methods rely on linear or nonlinear regression (machine learning, deep learning) approaches. They train a regression model using partial input data and then use the trained model parameters to estimate soil moisture. These methods lack a physical foundation, making the inversion accuracy highly susceptible to the accuracy of the training data, and their spatial and temporal transferability is weak.

[0008] 3) Model-driven active and passive microwave soil moisture joint inversion algorithms have a physical mechanism, thus exhibiting strong universality and transferability. However, such methods are still in the exploratory stage. In the soil moisture inversion process, the contribution (weight) of active and passive microwave data is usually set to 0.5 or a fixed constant, which has a strong subjectivity. Furthermore, the weights should be different for different soil and vegetation states, leading to reduced applicability of the fixed-weight method to different underlying surface types. Summary of the Invention

[0009] To address the shortcomings of the aforementioned technologies, this invention provides a method for jointly inverting soil moisture using active and passive microwave methods.

[0010] To solve the above technical problems, the technical solution adopted by the present invention is: a method for jointly inverting soil moisture using active and passive microwave methods, comprising the following steps: Step 1: Acquisition and Preprocessing of Relevant Products: Acquire Sentinel-1 active microwave backscattering coefficient data, SMAP passive microwave brightness and temperature data, and auxiliary data for the study area; perform projection transformation, resampling, cropping, and spatiotemporal scale matching on all data to ensure that all data products have the same spatiotemporal resolution and coverage. Step 2: Constructing a forward model: The Mironov model is used to calculate the soil dielectric constant, and the Zhao and Oh models are used to calculate the surface emissivity and backscattering of bare soil under different roughnesses, respectively; the τ-ω model and the water cloud model are used as forward models to jointly simulate the brightness temperature of passive microwaves and the backscattering of active microwaves under vegetation cover. Step 3: Calibration of vegetation and roughness parameters: Based on measured soil moisture data, the vegetation and roughness parameters in the forward model are calibrated in a targeted manner to ensure simulation accuracy and physical consistency. Step 4: Constructing the soil moisture inversion cost function: Construct a cost function that integrates Sentinel-1 active microwave backscattering and SMAP passive microwave brightness temperature. This function is defined as the sum of squares of the weighted normalized relative errors between the observed values ​​and the pseudo values ​​of the forward model. The inversion process uses a metaheuristic optimization algorithm based on the Newton-Raphson method to obtain the model soil moisture input value corresponding to the optimal solution of the cost function as the final inversion result; Step 5: Iteration stopping condition judgment: If the constructed inversion cost function is minimized, stop the current soil moisture input and go to step 6; otherwise, return to step 4. Step 6: Determining the soil moisture value of the study area.

[0011] Preferably, in step 1, the Sentinel-1 active microwave backscattering coefficient data is in C-band IW mode with a spatial resolution of 10m, and is resampled to 1km after radiometric calibration, thermal noise removal and terrain correction. The SMAP passive microwave brightness temperature data is the L-band brightness temperature with a spatial resolution of 9km and resampling to 1km. The auxiliary data included MODIS land cover, OpenLandMap soil texture, ERA5 soil temperature, and vegetation water content (VMC) estimated by MOD13A2 normalized vegetation index. Among them, MODIS land cover, OpenLandMap soil texture, and ERA5 soil temperature were resampled to 1 km, and the spatial resolution of vegetation water content (VMC) was 1 km.

[0012] Preferably, in step 2, the Mironov model uses the negative refractive index of the soil as a function of water content, and obtains the mixed dielectric constant of free and bound water through linear fitting, expressed by the formula: Formula 1 In the formula, the subscripts s, m, w, and a represent wet soil, mineral particles, water, and air, respectively. , , , These represent the mixed dielectric constants of wet soil, mineral particles, water, and air, respectively. Indicates volumetric water content. , These represent the volumetric water content of mineral particles and air, respectively. For wet soil and water volume content, the subscript can be omitted, and there is also and Then formula 1 simplifies to: Formula 2 Formula 3 in, It is the mixed dielectric constant of absolutely dry soil; Complex refractive index of wet soil It can be represented by the refractive index and the normalized decay exponent, as shown in Equation 4: Formula 4 Where n represents the refractive index of the wet soil, k represents the normalized attenuation index of the wet soil, and j represents the imaginary unit, satisfying j² = -1; The refractive index and normalized attenuation index of wet soil are estimated using Equations 5 and 6: Formula 5 Formula 6 in , , These represent the refractive indices of wet soil, dry soil, and water, respectively. , , These represent the normalized attenuation coefficients for wet soil, dry soil, and moisture content, respectively.

[0013] Dielectric constant With loss factor Calculated by the following formula: Formula 7 Formula 8

[0014] The dielectric constant was calculated using the Debye equation. With loss factor The refractive index and normalized decay index can be derived from Equation 4, and the complex dielectric constant of the soil can be estimated from these two parameters.

[0015] Preferably, the Oh model establishes a semi-empirical model of radar backscattering coefficient and surface parameters; The backscattering coefficient is expressed as: Formula 9 Formula 10 Formula 11 Formula 12 Formula 13 in, and These are the soil backscattering coefficients under two different polarizations. and Representing vertical and horizontal polarization modes, The Fresnel reflection coefficient represents the vertical polarization. The Fresnel reflection coefficient represents the horizontal polarization. It is the angle of incidence. It is the Fresnel reflectance of the nadir. , and These are empirical parameters. For free space wavenumber, It is the root mean square height of the Earth's surface.

[0016] Preferably, the Zhao model is expressed as: Formula 14 Formula 15 In the formula, For polarization mode, To polarize the microwave emissivity of surface soil, The Fresnel reflection coefficient of a smooth surface. For effective roughness parameters, Angle of incidence , These are the root mean square height and the relevant length, respectively. , , It is the fitting coefficient related to the incident angle; the model accurately simulates the soil emissivity in the L-band under conditions of soil moisture, surface roughness and incident angle over a large range.

[0017] Preferably, the water cloud model is represented as follows: Formula 16 Formula 17 Formula 18 In the formula, This represents the total radar backscattering coefficient. This indicates the contribution of vegetation layer to radar backscattering. This indicates the contribution of the exposed ground surface to radar backscattering. Indicates the double-layer transmittance of vegetation. This represents vegetation water content and can be calculated based on its normalized vegetation index (NDI). The radar incident angle, , These are parameters that depend on the vegetation type. This refers to the water content of the vegetation.

[0018] Preferably, the ω-τ model is expressed as: Formula 19 Formula 20 In the formula, Surface reflectance; and Let be the effective temperatures of the soil and vegetation, respectively. The model assumes that the physical temperature between the soil and vegetation is uniform. Approximately ; This refers to the water content of vegetation. , These are parameters that depend on the vegetation type. This represents the brightness temperature observed by a passive microwave radiometer. This indicates the optical thickness of the vegetation.

[0019] Preferably, in step 3, the cost function in the vegetation and roughness parameter retrieval process is given by the following formula: Formula 21 Formula 22 in, , , For roughness parameters; , , , Here, is a parameter dependent on vegetation type; n is the data logarithm; mm represents the polarization mode of the radar backscattering cross section. Polarization channels representing brightness temperature; , To simulate backscattering and brightness temperature values; , To observe backscattering and brightness temperature values, when the cost function , When minimized, the model outputs optimized vegetation and roughness parameters; , This represents the input parameters to be calibrated in the cost function. , represents the cost function for retrieving input parameters in the active microwave and passive microwave forward models, respectively; i represents the observation time series.

[0020] Preferably, in step 4, the cost function in the soil moisture retrieval process is as follows: Formula 23 in, It simulates vv, vh polarization backscattering and water cloud models. For the simulation of v, h polarization brightness temperature and ω-τ model; It is the observed polarized backscattering of vv and vh, These are the observed v, h polarization brightness temperature, and... This represents the input parameters to be calibrated in the cost function. The cost function for soil moisture retrieval is represented; In this cost function, To quantify the adaptive weighting coefficients of the contributions of the braking microwave and passive microwave observation channels to the inversion, a dynamic weighting strategy based on normalization minimum cost is adopted, calculated by the following formula: in, , Represent the cost functions in equations 21 and 22 respectively. , A function that normalizes the minimum value; , These represent the mean observed backscattering coefficients for vertical polarization and cross-polarization, respectively. , These represent the average observed brightness temperature values ​​for vertical polarization and horizontal polarization, respectively. To each , Normalization was performed, where, , Cost functions in equations 21 and 22 are respectively , Minimum value.

[0021] Preferably, in step 6, a curve is plotted showing the change between soil moisture and the corresponding inversion cost function value. The change of the inversion cost function value with the input soil moisture value is analyzed. The soil moisture value corresponding to the slow decrease rate of the inversion cost function value is selected, and the soil moisture input value determined by the metaheuristic optimization algorithm based on the Newton-Raphson method is the optimal soil moisture value.

[0022] This invention discloses a method for joint active and passive microwave remote sensing inversion of soil moisture, applicable to high-resolution soil moisture inversion of heterogeneous underlying surfaces. By constructing a forward model with physical mechanisms and adopting an adaptive weight cost function, the method can optimize inversion parameters and achieve high-precision soil moisture estimation at a resolution of 1 km. Moreover, the inversion exhibits strong spatiotemporal portability, realizing large-scale, all-weather, high spatial resolution, and high-precision microwave remote sensing estimation of soil moisture. This method is of great significance for applications in meteorological forecasting, hydrological modeling, ecological monitoring, agricultural management, and disaster early warning. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the technical process of the present invention. Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0025] Terminology Explanation:

[0026] Sentinel-1: The European Space Agency's C-band synthetic aperture radar satellite, used for active microwave observation.

[0027] SMAP: NASA's active-passive soil moisture satellite, used for passive microwave observation.

[0028] Soil moisture (SM): The water content in surface soil, usually expressed as a volume percentage (m³). 3 / m 3 )express.

[0029] Heterogeneous underlying surface: The land surface has spatial variations and heterogeneity, such as different surface features like vegetation type, soil texture, and topography.

[0030] Backscattering coefficient: usually denoted by the symbol σ 0 It is a dimensionless physical quantity that quantitatively describes the ability of a ground surface or volume to scatter electromagnetic wave energy toward the radar direction under actively emitted microwave illumination.

[0031] Brightness temperature: usually represented by the symbol T B or T b , is a physical quantity. It represents the thermodynamic temperature of an ideal blackbody when it emits the same radiance as the observed target at the same frequency.

[0032] To overcome the shortcomings of existing technologies, this invention proposes a method for joint active and passive microwave remote sensing inversion of soil moisture. This method is a high-resolution joint active and passive microwave remote sensing inversion method suitable for heterogeneous underlying surfaces, achieving high-precision estimation of soil moisture at a resolution of 1 km. The main technical problems addressed by this invention include the following three points: First, considering the complementarity of active and passive microwave observations, the problem of low spatial resolution in single passive microwave soil moisture retrieval or low accuracy in single active microwave soil moisture retrieval is solved by constructing a joint optimization cost function. Second, adaptive weights are applied based on the magnitude of the simulation error between active and passive microwaves, eliminating the influence of subjective factors in weight determination in existing joint active and passive microwave soil moisture retrieval algorithms. Third, by simulating all-day microwave signals through a mechanistic model, an iterative retrieval model is further constructed, solving the problem that existing regression or machine learning-based methods lack physical mechanisms, resulting in weak applicability of the algorithm to different regions in space and weak transferability of the algorithm to different time periods in time.

[0033] The method for jointly inverting soil moisture using active and passive microwave observations proposed in this invention leverages the complementarity of these observations. It scientifically and rationally determines the optimal inversion parameters and soil moisture estimates through cost function optimization, thereby obtaining soil moisture data at a 1km scale. The overall technical process of this invention is as follows: Figure 1 As shown, it mainly consists of the following steps:

[0034] Step 1: Acquisition and preprocessing of relevant products: Sentinel-1 active microwave backscattering coefficient, SMAP passive microwave brightness temperature product, and corresponding auxiliary data products such as sensor parameters, soil texture, vegetation water content (VWC), soil temperature, and land cover were acquired for the study area. Sentinel-1 data was obtained in C-band (5.405 GHz) IW mode with a spatial resolution of 10 m, and resampled to 1 km after radiometric calibration, thermal noise removal, and topographic correction. SMAP data was obtained in L-band (1.41 GHz) brightness temperature with a spatial resolution of 9 km, and resampled to 1 km. Auxiliary data included MODIS land cover (500 m, resampled to 1 km), OpenLandMap soil texture (250 m, resampled to 1 km), ERA5 soil temperature (approximately 9 km, resampled to 1 km), and VMC estimated from MOD13A2 normalized vegetation index (1 km). All data underwent projection transformation, resampling, cropping, and spatiotemporal scale matching to ensure that all data products had the same spatiotemporal resolution and coverage.

[0035] Step 2: Construct the forward model: This invention uses the Mironov model to calculate the soil dielectric constant, and then uses the Zhao and Oh models to calculate the surface emissivity and backscattering of bare soil under different roughnesses. Finally, the τ-ω model and the water cloud model are used as forward models to jointly simulate the brightness temperature of passive microwaves and the backscattering of active microwaves under vegetation cover.

[0036] The Mironov model considers free water and bound water as separate moisture components in the soil. Using the negative refractive index of the soil as a function of moisture, the mixed dielectric constant of free and bound water is obtained through linear fitting, and its formula can be expressed as: Formula 1 In the formula, s, m, w, and a represent wet soil, mineral particles, water, and air, respectively; , , , These represent the mixed dielectric constants of wet soil, mineral particles, water, and air, respectively. Indicates volumetric water content. , These represent the volumetric water content of mineral particles and air, respectively. For wet soil and water volume content, the subscript can be omitted, and there is also and Then formula 1 can be simplified to: Formula 2 Formula 3 in, It is the mixed dielectric constant of absolutely dry soil, determined by the dielectric constant of the solids in the soil and the soil bulk density. The complex refractive index of wet soil... It can be represented by the refractive index and the normalized decay exponent, as shown in Equation 4: Formula 4 Where n represents the refractive index of the wet soil, k represents the normalized attenuation index of the wet soil, and j represents the imaginary unit, satisfying j² = -1; The normalized attenuation exponent is understood here as the ratio of the standard attenuation coefficient to the free-space propagation constant. The refractive index and normalized attenuation exponent of wet soil can be estimated using Equations 5 and 6: Formula 5 Formula 6 in , , These represent the refractive indices of wet soil, dry soil, and water, respectively. , , These represent the normalized attenuation coefficients for wet soil, dry soil, and moisture content, respectively. If the refractive index and the normalized decay exponent are known, then the dielectric constant is... With loss factor It can be calculated using the following formula: Formula 7 Formula 8 The dielectric constant was calculated using the Debye equation. With loss factor The refractive index and normalized attenuation exponent are derived from Equation 4, and the complex permittivity of the soil can be estimated from these two parameters. This model is complex and requires a large number of parameters, some of which need to be obtained through field measurements. However, it is effective and is currently widely used.

[0037] The Oh model, developed by Oh et al., uses radar data with various polarization modes to establish the relationship between surface parameters and the same polarization ratio and cross polarization ratio, thus creating a semi-empirical model of radar backscattering coefficients and surface parameters. For bare soil, the backscattering observed by radar is only related to the soil's dielectric constant (soil moisture) and surface roughness, and its backscattering can be expressed as: Formula 9 Formula 10 Formula 11 Formula 12 Formula 13 in, and These are the soil backscattering coefficients under two different polarizations. and Representing vertical and horizontal polarization modes, The Fresnel reflection coefficient represents the vertical polarization. The Fresnel reflection coefficient represents the horizontal polarization. It is the angle of incidence. It is the Fresnel reflectance at the nadir. , and These are empirical parameters. For free space wavenumber, The model is based on the root mean square height of the ground surface. It can effectively simulate backscattering characteristics under different soil moisture conditions when the root mean square height of the ground surface is 0.1 to 6 cm and the correlation length is within the range of 2.6 to 19.7 cm of surface roughness.

[0038] The Zhao model is an exponentially correlated surface parameterized emission model constructed for L-band passive microwave soil moisture inversion, which precisely quantifies the impact of surface roughness on microwave radiation. Based on extensive simulation data from the Advanced Integral Equation Model (AIEM), this model employs an exponential autocorrelation function that better reflects land surface characteristics, comprehensively considering the synergistic effect of geometric roughness parameters and incident angle, and using effective roughness parameters (… The parameterization of microwave emissivity can be simplified and accurately simulated, and can be expressed as: Formula 14 Formula 15 In the formula, For polarization mode, To polarize the microwave emissivity of surface soil, The Fresnel reflection coefficient of a smooth surface. For effective roughness parameters, Angle of incidence , These are the root mean square height and the relevant length, respectively. , , It is the fitting coefficient related to the incident angle; the model accurately simulates the soil emissivity in the L-band under conditions of soil moisture, surface roughness and incident angle over a large range.

[0039] The water cloud model simplifies the scattering mechanism of vegetation cover, neglecting multiple backscatterings between vegetation and soil, and can be expressed as: Formula 16 Formula 17 Formula 18 In the formula, This represents the total radar backscattering coefficient. This indicates the contribution of vegetation layer to radar backscattering. This indicates the contribution of the exposed ground surface to radar backscattering. Indicates the double-layer transmittance of vegetation. This represents vegetation water content and can be calculated based on its normalized vegetation index (NDI). The radar incident angle, , These are parameters that depend on the vegetation type. This refers to the water content of the vegetation.

[0040] The ω-τ model uses surface brightness temperature to represent the brightness temperature of the vegetation canopy, without considering primary or secondary scattering within the vegetation layer, and can be expressed as: Formula 19 Formula 20 In the formula, Surface reflectance; and Let be the effective temperatures of the soil and vegetation, respectively. The model assumes that the physical temperature between the soil and vegetation is uniform. Approximately ; This refers to the water content of vegetation. , These are parameters that depend on the vegetation type. This represents the brightness temperature observed by a passive microwave radiometer. This indicates the optical thickness of the vegetation.

[0041] Step 3: Vegetation and roughness parameter calibration: To accommodate the heterogeneity of different underlying surface conditions, the roughness and vegetation parameters in the forward model need to be specifically calibrated based on measured soil moisture data to ensure the simulation accuracy and physical consistency of the model under different land surface types. The cost function in the parameter retrieval process is given by the following equation: Formula 21 Formula 22 in, , , For roughness parameters; , , , Here, is a parameter dependent on vegetation type; n is the data logarithm; mm represents the polarization mode of the radar backscattering cross section. Polarization channels representing brightness temperature; , To simulate backscattering and brightness temperature values; , To observe backscattering and brightness temperature values, when the cost function , When minimized, the model outputs optimized vegetation and roughness parameters. , This represents the input parameters to be calibrated in the cost function. , represents the cost function for retrieving input parameters in the active microwave and passive microwave forward models, respectively; i represents the observation time series.

[0042] Step 4: Construct the soil moisture inversion cost function: The core of active-passive co-inversion of soil moisture lies in constructing a cost function that comprehensively utilizes Sentinel-1 microwave backscattering and SMAP microwave brightness temperature. This function is defined as the sum of squares of the weighted normalized relative errors between the observed values ​​and the pseudo-values ​​of the forward model. The inversion process employs a metaheuristic optimization algorithm based on the Newton-Raphson method to obtain the model soil moisture input value corresponding to the optimal solution of the cost function, which is then used as the final inversion result. The cost function in the soil moisture retrieval process is as follows: Formula 23 in, It simulates vv, vh polarization backscattering and water cloud models. For the simulation of v, h polarization brightness temperature and ω-τ model; It is the observed polarized backscattering of vv and vh, These are the observed v, h polarization brightness temperature, and... This represents the input parameters to be calibrated in the cost function. This represents the cost function for soil moisture retrieval. In this cost function, To quantify the adaptive weighting coefficients of the contributions of the two observation channels, braking microwave (backscattering) and passive microwave (brightness temperature), to the inversion, a dynamic weighting strategy based on normalization minimum cost is adopted, calculated by the following formula: in, , Represent the cost functions in equations 21 and 22 respectively. , A function that normalizes the minimum value; , These represent the mean observed backscattering coefficients for vertical polarization and cross-polarization, respectively. , These represent the average observed brightness temperature values ​​for vertical polarization and horizontal polarization, respectively. To eliminate the differences in dimensions and magnitudes of different observed physical quantities, they were respectively... , Normalization was performed, where, , Cost functions in equations 21 and 22 are respectively , Minimum value; adaptive weights are essentially normalized weights of the inverses of the two normalized minimum costs, so that the observation channel with smaller simulation error in the forward model occupies a larger weight.

[0043] Step 5: Determine the stopping condition for iteration. Perform an iteration stopping condition check. If the constructed inversion cost function is minimized, stop the current soil moisture input and proceed to step 6; otherwise, return to step 4.

[0044] Step 6: Determination of soil moisture values ​​in the study area: A curve showing the change in soil moisture and the corresponding inversion cost function value was plotted. The variation of the inversion cost function value with the input soil moisture value was analyzed. The soil moisture value corresponding to the slow decrease rate of the inversion cost function value (e.g., a decrease rate of 0.01) was selected, and the soil moisture input value determined by the metaheuristic optimization algorithm based on the Newton-Raphson method was the optimal soil moisture value. Together, these two methods formed a joint active and passive microwave inversion scheme for heterogeneous underlying surfaces.

[0045] In summary, this invention uses forward modeling based on dielectric constant model (Mironov model), surface emissivity model (Zhao model), surface brightness temperature model (ω-τ model), bare soil backscattering calculation model (Oh model), and vegetation cover backscattering calculation model (water cloud model) to jointly simulate the brightness temperature of passive microwaves and the backscattering of active microwaves. By constructing a joint active-passive cost function between the backscattering of active microwaves and the brightness temperature observations and forward model simulations of passive microwaves, a metaheuristic optimization algorithm based on the Newton-Raphson method is used to minimize this cost function to give the optimal 1km scale soil moisture estimate for heterogeneous underlying surfaces.

[0046] By introducing an adaptive weighting mechanism, weights are dynamically allocated based on the normalized minimum simulation error of the active and passive channels, thus achieving an effective characterization of the spatiotemporal dynamic changes of soil moisture and improving the accuracy and robustness of soil moisture inversion under complex surface, vegetation conditions, and seasonal variations.

[0047] Compared with the prior art, the present invention has the following technical advantages:

[0048] 1) This invention takes into account the complementarity of active and passive microwave observations. By constructing a cost function for jointly optimizing soil moisture using active and passive microwave methods, it overcomes the problems of low spatial resolution in soil moisture inversion using single passive microwave or low accuracy in soil moisture inversion using single active microwave.

[0049] 2) The proposed cost function is built on the basis of a forward model driven by a mechanism model, which solves the problem of weak algorithm space and time transferability caused by the lack of physical basis in existing data-driven methods.

[0050] 3) By introducing an adaptive weighting mechanism based on normalized minimum simulation error, the problem of reduced applicability of soil moisture inversion accuracy to different underlying surface types caused by subjective factors in weight determination in existing joint inversion algorithms is overcome.

[0051] The above embodiments are not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the technical solution of the present invention are also within the protection scope of the present invention.

Claims

1. A method for jointly inverting soil moisture using active and passive microwave methods, characterized in that: Includes the following steps: Step 1: Acquisition and Preprocessing of Relevant Products: Acquire Sentinel-1 active microwave backscattering coefficient data, SMAP passive microwave brightness and temperature data, and auxiliary data for the study area; perform projection transformation, resampling, cropping, and spatiotemporal scale matching on all data to ensure that all data products have the same spatiotemporal resolution and coverage. Step 2: Constructing a forward model: The Mironov model is used to calculate the soil dielectric constant, and the Zhao and Oh models are used to calculate the surface emissivity and backscattering of bare soil under different roughnesses, respectively; the τ-ω model and the water cloud model are used as forward models to jointly simulate the brightness temperature of passive microwaves and the backscattering of active microwaves under vegetation cover. Step 3: Calibration of vegetation and roughness parameters: Based on measured soil moisture data, the vegetation and roughness parameters in the forward model are calibrated in a targeted manner to ensure simulation accuracy and physical consistency. Step 4: Constructing the soil moisture inversion cost function: Construct a cost function that integrates Sentinel-1 active microwave backscattering and SMAP passive microwave brightness temperature. This function is defined as the sum of squares of the weighted normalized relative errors between the observed values ​​and the pseudo values ​​of the forward model. The inversion process uses a metaheuristic optimization algorithm based on the Newton-Raphson method to obtain the model soil moisture input value corresponding to the optimal solution of the cost function as the final inversion result; Step 5: Iteration stopping condition judgment: If the constructed inversion cost function is minimized, stop the current soil moisture input and go to step 6; otherwise, return to step 4. Step 6: Determining the soil moisture value of the study area.

2. The method for jointly inverting soil moisture using active and passive microwave methods according to claim 1, characterized in that: In step 1, the Sentinel-1 active microwave backscattering coefficient data is in C-band IW mode with a spatial resolution of 10m. After radiometric calibration, thermal noise removal and terrain correction, it is resampled to 1km. The SMAP passive microwave brightness temperature data is the L-band brightness temperature with a spatial resolution of 9km and resampling to 1km. The auxiliary data included MODIS land cover, OpenLandMap soil texture, ERA5 soil temperature, and vegetation water content (VMC) estimated by MOD13A2 normalized vegetation index. Among them, MODIS land cover, OpenLandMap soil texture, and ERA5 soil temperature were resampled to 1 km, and the spatial resolution of vegetation water content (VMC) was 1 km.

3. The method for combined active and passive microwave inversion of soil moisture according to claim 2, characterized in that: In step 2, the Mironov model uses the negative refractive index of the soil as a function of water content, and obtains the mixed dielectric constant of free and bound water through linear fitting, expressed by the formula: Official 1 In the formula, the subscripts s, m, w, and a represent wet soil, mineral particles, water, and air, respectively. , , , These represent the mixed dielectric constants of wet soil, mineral particles, water, and air, respectively. Indicates volumetric water content. , These represent the volumetric water content of mineral particles and air, respectively. For wet soil and water volume content, the subscript can be omitted, and there is also and Then formula 1 simplifies to: Official 2 Official 3 in, It is the mixed dielectric constant of absolutely dry soil; Complex refractive index of wet soil It can be represented by the refractive index and the normalized decay exponent, as shown in Equation 4: Official 4 Where n represents the refractive index of the wet soil, k represents the normalized attenuation index of the wet soil, and j represents the imaginary unit, satisfying j² = -1. The refractive index and normalized attenuation index of wet soil are estimated using Equations 5 and 6: Official 5 Official 6 in , , These represent the refractive indices of wet soil, dry soil, and water, respectively. , , These represent the normalized attenuation coefficients for wet soil, dry soil, and moisture content, respectively. Dielectric constant With loss factor Calculated by the following formula: Official 7 Official 8 The dielectric constant was calculated using the Debye equation. With loss factor The refractive index and normalized decay index can be derived from Equation 4, and the complex dielectric constant of the soil can be estimated from these two parameters.

4. The method for combined active and passive microwave inversion of soil moisture according to claim 3, characterized in that: The Oh model establishes a semi-empirical model of radar backscattering coefficient and surface parameters. The backscattering coefficient is expressed as: Official 9 Official 10 Official 11 Official 12 Official 13 in, and These are the soil backscattering coefficients under two different polarizations. and Representing vertical and horizontal polarization modes, The Fresnel reflection coefficient represents the vertical polarization. The Fresnel reflection coefficient represents the horizontal polarization. It is the angle of incidence. It is the Fresnel reflectance of the nadir. , and These are empirical parameters. For free space wavenumber, It is the root mean square height of the Earth's surface.

5. The method for combined active and passive microwave inversion of soil moisture according to claim 4, characterized in that: The Zhao model is represented as follows: Official 14 Official 15 In the formula, For polarization mode, To polarize the microwave emissivity of surface soil, The Fresnel reflection coefficient of a smooth surface. For effective roughness parameters, Angle of incidence , These are the root mean square height and the relevant length, respectively. , , It is the fitting coefficient related to the incident angle; the model accurately simulates the soil emissivity in the L-band under conditions of soil moisture, surface roughness and incident angle over a large range.

6. The method for combined active and passive microwave inversion of soil moisture according to claim 5, characterized in that: The water cloud model is represented as follows: Official 16 Official 17 Official 18 In the formula, This represents the total radar backscattering coefficient. This indicates the contribution of vegetation layer to radar backscattering. This indicates the contribution of the exposed ground surface to radar backscattering. Indicates the double-layer transmittance of vegetation. This represents vegetation water content and can be calculated based on its normalized vegetation index (NDI). The radar incident angle, , These are parameters that depend on the vegetation type. This refers to the water content of the vegetation.

7. The method for combined active and passive microwave inversion of soil moisture according to claim 6, characterized in that: The ω-τ model is expressed as: Official 19 Official 20 In the formula, Surface reflectance; and Let be the effective temperatures of the soil and vegetation, respectively. The model assumes that the physical temperature between the soil and vegetation is uniform. Approximately ; This refers to the water content of vegetation. , These are parameters that depend on the vegetation type. This represents the brightness temperature observed by a passive microwave radiometer. This indicates the optical thickness of the vegetation.

8. The method for combined active and passive microwave inversion of soil moisture according to claim 7, characterized in that: In step 3, the cost function for the vegetation and roughness parameter retrieval process is given by the following formula: Official 21 Official 22 in, , , For roughness parameters; , , , Here, is a parameter dependent on vegetation type; n is the data logarithm; mm represents the polarization mode of the radar backscattering cross section. Polarization channels representing brightness temperature; , To simulate backscattering and brightness temperature values; , To observe backscattering and brightness temperature values, when the cost function , When minimized, the model outputs optimized vegetation and roughness parameters; , This represents the input parameters to be calibrated in the cost function. , represents the cost function for retrieving input parameters in the active microwave and passive microwave forward models, respectively; i represents the observation time series.

9. The method for jointly inverting soil moisture using active and passive microwave methods according to claim 8, characterized in that: In step 4, the cost function in the soil moisture retrieval process is as follows: Official 23 in, It simulates vv, vh polarization backscattering and water cloud models. For the simulation of v, h polarization brightness temperature and ω-τ model; It is the observed polarized backscattering of vv and vh, These are the observed v, h polarization brightness temperature, and... This represents the input parameters to be calibrated in the cost function. The cost function for soil moisture retrieval is represented; In this cost function, To quantify the adaptive weighting coefficients of the contributions of the braking microwave and passive microwave observation channels to the inversion, a dynamic weighting strategy based on normalization minimum cost is adopted, calculated by the following formula: in, , Represent the cost functions in equations 21 and 22 respectively. , A function that normalizes the minimum value; , These represent the mean observed backscattering coefficients for vertical polarization and cross-polarization, respectively. , These represent the average observed brightness temperature values ​​for vertical polarization and horizontal polarization, respectively. To each , Normalization was performed, where, , Cost functions in equations 21 and 22 are respectively , Minimum value.

10. The method for jointly inverting soil moisture using active and passive microwave methods according to claim 9, characterized in that: In step 6, a curve showing the change of soil moisture and the corresponding inversion cost function value is plotted. The change of the inversion cost function value with the input soil moisture value is analyzed. The soil moisture value corresponding to the slow decrease rate of the inversion cost function value is selected. The soil moisture input value determined by the metaheuristic optimization algorithm based on the Newton-Raphson method is the optimal soil moisture value.

Images