A Model Simulation-Based Method and System for Verifying the Authenticity of Remote Sensing Products

By combining ground observations and model simulations to generate spatiotemporally continuous reference data, a multi-index system is constructed, which solves the problem of insufficient spatial representativeness in the authenticity verification of remote sensing products and realizes a comprehensive evaluation of the accuracy and uncertainty of remote sensing products.

CN123116601APending Publication Date: 2026-11-17KUNMING UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611350862.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-02
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing methods for verifying the authenticity of remote sensing products rely excessively on limited ground observations, resulting in insufficient spatial representativeness and scale consistency. This makes it difficult to accurately assess the true accuracy of products under large-scale and complex underlying surface conditions, and also makes it difficult to explain the main sources of product uncertainty.

Method used

By combining ground observations and model simulations, spatiotemporally continuous model simulation reference data is generated, and through standardization processing, reference data for product evaluation is formed. A multi-index system is constructed for comprehensive evaluation, including assessments of bias, correlation, distribution similarity, and uncertainty.

Benefits of technology

It generates spatiotemporally continuous model simulation reference data, overcomes the problem of insufficient spatial representativeness of ground observations, constructs a multi-level evaluation system, can quantitatively diagnose the impact of environmental factors on product uncertainty, and realizes a comprehensive evaluation of product accuracy and applicability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116601A_ABST
    Figure CN123116601A_ABST
Patent Text Reader

Abstract

This invention discloses a model-based method and system for verifying the authenticity of remote sensing products, belonging to the technical fields of hydrology and remote sensing, land surface process simulation, and quantitative remote sensing product authenticity verification. This method targets the surface variables corresponding to the remote sensing product to be verified, collecting meteorological drivers, topography, surface parameters, observational data, and necessary human activity data for the study area. A mechanistic model or coupled model matching the target variables is selected and calibrated to generate spatiotemporally continuous model simulation reference data. After standardizing the spatial scale, temporal scale, and variable definitions between the remote sensing product to be verified and the model simulation reference data, deviation, correlation, distribution consistency, and relative uncertainty indices are calculated to form a product accuracy evaluation result. Furthermore, a generalized summation model is used to quantify the contribution of environmental factors to product uncertainty, diagnose error sources, and output product suitability levels and improvement suggestions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the technical field of hydrology and remote sensing, land surface process simulation and quantitative remote sensing product authenticity verification, specifically involving a method and system for verifying the authenticity of remote sensing products based on model simulation. Background Technology

[0002] Remote sensing products can provide information on surface variables at regional to global scales, which is of great significance for water resource management, ecological environment monitoring, agricultural irrigation, disaster early warning and climate change research.

[0003] In recent years, remote sensing products based on satellite observation, active / passive remote sensing inversion, multi-source fusion, reanalysis, and land surface model assimilation have emerged, providing rich data sources for global and regional scale applications.

[0004] However, differences in sensor types, inversion algorithms, input data, and scale definitions among various remote sensing products lead to varying accuracy performance under different underlying surface conditions, climate zones, and intensities of human activity. Particularly in areas with high vegetation cover, complex terrain, strong surface heterogeneity, or human disturbances such as irrigation, product uncertainty increases significantly.

[0005] Traditional verification methods often rely on limited ground station observations or manual quadrat observations, which suffer from insufficient spatial representativeness. Ground observations typically only provide point-scale or small-range true values, making it difficult to reflect the spatial heterogeneity at the remote sensing pixel scale, leading to discrepancies between the veracity verification results and the actual product error.

[0006] Existing studies have proposed product authenticity verification methods based on ground observation networks, but these methods usually rely on dense observation stations or long-term manual sampling, which are difficult to meet the needs of large-scale remote sensing product authenticity verification in terms of economic cost, deployment conditions and spatial coverage. Summary of the Invention

[0007] The technical problem this invention aims to solve is that existing methods for verifying the authenticity of remote sensing products rely excessively on limited ground observations, lacking sufficient spatial representativeness and scale consistency. This makes it difficult to accurately assess the true accuracy of products under large-scale, complex underlying surface conditions and to explain the main sources of product uncertainty. This invention provides a method for verifying the authenticity of remote sensing products that combines ground observation constraints with the spatiotemporal continuity of model simulations. This method replaces or supplements limited ground observations and achieves a comprehensive evaluation of product accuracy, sources of uncertainty, and applicable scenarios at a unified scale.

[0008] This invention is applicable to the accuracy assessment and uncertainty analysis of multi-source remote sensing products, reanalysis products, or assimilation products in arid, semi-arid, humid, and complex underlying surface conditions. The target surface variables corresponding to the remote sensing products are soil moisture, evapotranspiration, surface temperature, or vegetation index.

[0009] This invention utilizes a model calibrated with observational data to generate spatiotemporally continuous model simulation reference data R0, which, after being standardized, serves as reference data R1 for verifying the authenticity of remote sensing products to be validated. By constructing a comprehensive evaluation index system, the authenticity of remote sensing products from different sources, at different spatial scales, and using different algorithms can be verified. In the scenario of verifying the authenticity of soil moisture, ParFlow-CLM can serve as a model instance for generating reference data.

[0010] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: a method for verifying the authenticity of remote sensing products based on model simulation, comprising the following steps: Step 1: Collect model-driven data, model parameter data, ground observation data, and human activity data related to the target surface variables represented by the remote sensing product P0 to be validated within the study area; Step 2: Select a mechanistic model or a coupled model based on the target surface variables, construct a study area model using the driving data and parameter data of the mechanistic model or coupled model, and perform parameter calibration and independent verification of the study area model using ground observation data; Step 3: Run the independently validated study area model to generate spatiotemporal continuous model simulation reference data R0 for the target surface variables; Step 4: Perform spatial scale, time scale, projection coordinates, variable definition and effective sample consistency processing on the remote sensing product P0 to be verified and the model simulation reference data R0 to obtain the consistent remote sensing product P1 to be verified and the model simulation reference data R1. Step 5: Compare the standardized remote sensing product P1 to be verified with the model simulation reference data R1 in spatiotemporal samples, calculate the deviation or error index, correlation index, distribution consistency index and relative uncertainty index, and obtain the product accuracy evaluation result. Step 6: Establish a generalized summation model with the relative uncertainty index as the response variable and environmental factors related to the target surface variable as explanatory variables. Determine the dominant environmental factors of product uncertainty based on the significance and deviation explanation rate of each environmental factor in the generalized summation model. Step 7: Based on the product accuracy evaluation results and the dominant environmental factors, output the accuracy, applicability level, uncertainty sources, and algorithm improvement suggestions of the unified remote sensing product P1 to be verified. Record the original remote sensing product P0 version, the unified remote sensing product P1 version, the study area model version, and the model-driven data or ground observation data so that the authenticity verification results can be updated on a rolling basis when any of the above versions change.

[0011] Specifically, in Step1: The target surface variables are soil moisture, evapotranspiration, surface temperature, or vegetation index. The model-driven data includes precipitation, temperature, radiation, wind speed, air pressure, and humidity; The model parameter data includes topographic elevation, land cover, and surface parameters that control the change process of target surface variables; The ground observation data are station observation data, flux observation data, or survey data of the target surface variables; The human activity data refers to irrigation data, water intake data, land use change data, or engineering control data that affect the target surface variables.

[0012] Step 2 includes the following steps: When the target surface variable is soil moisture, the coupling model is the ParFlow-CLM coupling model. The ParFlow module is used to calculate the three-dimensional variable saturated groundwater flow and surface runoff, and the CLM module is used to calculate the land surface energy balance and evapotranspiration. The effects of irrigation, water intake, land use change, and engineering regulation on target surface variables are incorporated into the coupled model as boundary conditions, source-sink terms, or parameter correction terms, and the Nash efficiency coefficient, root mean square error, and Pearson correlation coefficient are used as parameter calibration and independent verification indicators.

[0013] Specifically, in Step 3: The model simulation reference data R0 is the hourly or daily spatial continuous data of the target surface variables output by the model in the study area under the model's native spatial grid and temporal resolution, which is used as reference data for subsequent consistency processing.

[0014] Step 4 specifically includes: Step 4.1: Perform time aggregation or time interpolation on the model simulation reference data R0 based on the observation time of the remote sensing product P0 to be verified, so that the two have the same time scale and corresponding observation time; Step 4.2: Based on the pixel boundaries of the remote sensing product P0 to be verified, the model simulation reference data R0 is aggregated or resampled using area weighting to make the model simulation reference data R0 and the remote sensing product P0 to be verified have the same spatial grid. Based on the spatial grid and cell boundaries of the remote sensing product P0 to be verified, area-weighted aggregation or resampling method is selected for the model simulation reference data R0 to make the remote sensing product P0 to be verified and the model simulation reference data R0 have the same spatial grid. Step 4.3: Unify the remote sensing product P0 to be verified and the model simulation reference data R0 to the same projection coordinates; Step 4.4: Standardize the vertical representation definition based on the vertical representation method of the target surface variables: For target surface variables that have the meaning of soil layer stratification, standardize the soil layer depth; for target surface variables that do not have the meaning of soil layer stratification, standardize the vertical representation range. Step 4.5: Following the same valid sample screening rules, perform valid sample screening on the remote sensing product P0 to be verified and the model simulation reference data R0 to form a consistent remote sensing product P1 to be verified and model simulation reference data R1 with a one-to-one correspondence.

[0015] Specifically, in Step 5: The deviation or error indicators include mean absolute deviation, mean deviation, root mean square error, residual standard deviation, 95% uncertainty and t-statistic; The correlation indices include Pearson correlation coefficient, Spearman correlation coefficient, best fit curvature coefficient, Nash efficiency coefficient, Leighgate efficiency coefficient, and Wilmot consistency index. The distribution consistency index includes the Kolmogorov-Smirnov test integral and the comprehensive performance index; The relative uncertainty index is the relative uncertainty calculated based on the triangular cap method; In the i-th valid spatiotemporal sample participating in the index calculation, N represents the total number of valid paired samples, and i represents the i-th valid spatiotemporal sample. This represents the value of the remote sensing product P1 to be verified after standardization in the i-th valid spatiotemporal sample. This represents the value of the model simulation reference data R1 after standardization in the i-th valid spatiotemporal sample. This represents the arithmetic mean of the P1 values ​​of all remote sensing products to be verified. This represents the arithmetic mean of all model simulation reference values ​​R1, with the residuals uniformly denoted as... Unless otherwise specified, all indicators use the same batch of valid paired samples; When there are at least three independent product sequences, the triangular hat method is used to solve for the noise variance of the k-th product based on the covariance of the pairwise differences of each sequence. and in accordance with Calculate the k-th product sequence relative uncertainty .

[0016] Specifically, in Step 6: The environmental factors are determined based on the physical processes that control the spatial changes of surface variables of the target area; When the target surface variable is soil moisture, the environmental factors include topographic elevation, sand content, clay content, porosity, saturated hydraulic conductivity, organic matter content, leaf area index, and root abundance. After standardizing each environmental factor, the nonlinear relationship between the relative uncertainty and each environmental factor is fitted using a spline smoothing function. The dominant environmental factor is then determined based on the significance test results and the contribution ratio of the deviation explanation rate of each environmental factor.

[0017] Specifically, in Step 7: The product accuracy evaluation results are divided into three applicability levels: applicable, limited applicable, and not applicable, according to preset thresholds. After completing step 7, monitor the P1 version of the remote sensing product to be verified after the standardization process, the model version of the study area, the R1 version of the model simulation reference data after the standardization process, and the ground observation data version. When the original P0 version of the remote sensing product to be verified, the P1 version of the remote sensing product to be verified after the standardization process, the model version of the study area, the model-driven data, or the ground observation data change, return to step 4 and retain the results of this and previous authenticity verifications.

[0018] A model-simulation-based remote sensing product authenticity verification system executes the aforementioned model-simulation-based remote sensing product authenticity verification method. It includes a data management module, a model construction and calibration module, a reference data generation module, a product consistency processing module, a multi-index verification module, an uncertainty diagnosis module, and a result output and update module, all connected in sequence. The data management module executes Step 1; the model construction and calibration module executes Step 2; the reference data generation module executes Step 3; the product consistency processing module executes Step 4; the multi-index verification module executes Step 5; the uncertainty diagnosis module executes Step 6; and the result output and update module executes Step 7, recording the authenticity verification results and related version information. When changes are detected in the original remote sensing product version P0, the consistent remote sensing product version P1, the study area model version, the model-driven data, or the ground observation data, the next round of authenticity verification is triggered.

[0019] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the model simulation-based remote sensing product authenticity verification method according to any one of claims 1 to 8.

[0020] The beneficial effects of this invention are: 1. It can generate spatiotemporally continuous model simulation reference data R0, and after standardization processing, form model simulation reference data R1 for product evaluation, overcoming the deficiency of insufficient spatial representativeness of ground observations. 2. A multi-level evaluation system with 15 indicators was constructed, which can comprehensively evaluate product accuracy from multiple perspectives such as deviation, correlation, distribution similarity and uncertainty.

[0021] 3. It can quantitatively diagnose the impact of environmental factors on product uncertainty, providing direction for product algorithm improvement.

[0022] 4. This authenticity verification method can be updated on a rolling basis to adapt to the iteration of model and product versions. Attached Figure Description

[0023] Figure 1 This is a flowchart of the model simulation-based remote sensing product authenticity verification method of the present invention; Figure 2 This is a hierarchical structure diagram of the remote sensing product authenticity verification and evaluation index system of the present invention; Figure 3 The figures show time series and annual process response diagrams. Figure (a) is a time series diagram of soil moisture between the model simulation reference data R1 and ten standardized remote sensing products P1 from 2011 to 2013. Figures (b), (c), and (d) are time series diagrams of soil moisture and the difference between infiltration and evapotranspiration in 2011, 2012, and 2013, respectively. Figure 4 This is a scatter plot showing the overall accuracy of the standardized remote sensing product P1 to be verified and the standardized model simulation reference data R1. Figure 5 Figure (f) shows the correlation between the product P1 to be verified and the difference between the simulated infiltration and evapotranspiration in the model, and Figure (g) shows the bar chart of the effective sample number of each product P1 to be verified participating in the test. Figure 6 The following are spatial distribution maps of soil moisture-related environmental factors: Figure (a) shows the topographic elevation distribution, Figure (b) shows the sand content distribution, Figure (c) shows the clay content distribution, Figure (d) shows the porosity distribution, Figure (e) shows the saturated hydraulic conductivity distribution, Figure (f) shows the organic matter content distribution, Figure (g) shows the leaf area index distribution, and Figure (h) shows the root abundance distribution. Detailed Implementation

[0024] The present invention will be further described below with reference to the accompanying drawings and embodiments. The following embodiments are used to illustrate the technical path of the present invention and do not limit specific watersheds, model versions, grid scales, parameter values, or optimization algorithms.

[0025] Example 1: As Figure 1 As shown, the present invention sequentially performs the following steps: data acquisition Step 1, model construction and calibration Step 2, reference data generation Step 3, product consistency processing Step 4, multi-index verification Step 5, uncertainty diagnosis Step 6, and result output and update Step 7. Step 7 feeds back the product version, model version, and verification results to Step 4 to update the authenticity verification results as the version changes.

[0026] During the data acquisition phase, the target surface variable refers to the surface state quantity or flux directly characterized by the original remote sensing product P0 and output by the model, specifically soil moisture, evapotranspiration, surface temperature, or vegetation index. The acquisition includes meteorological driving data, topographic elevation and land cover data, surface parameter data controlling the changes in the target surface variable, ground observation data of the target surface variable, and human activity data affecting the target surface variable. When the target surface variable is soil moisture, relevant surface parameters specifically include river network, soil texture, porosity, and saturated hydraulic conductivity; human activity data specifically include irrigation volume and water intake processes.

[0027] In the model construction phase, specific models were selected and simulation units for the study area were established based on the physical processes corresponding to the target surface variables: soil moisture adopted the ParFlow-CLM hydrological-land surface coupling model, evapotranspiration adopted the CLM land surface process model, surface temperature adopted the CLM land surface energy balance model, and vegetation index adopted the Biome-BGC ecological process model. In the soil moisture authenticity verification example, the ParFlow module was responsible for the calculation of three-dimensional variable saturated groundwater flow and surface runoff, and the CLM module was responsible for the calculation of land surface energy balance and evapotranspiration.

[0028] For areas with intense human activity, processes that affect the target variable, such as irrigation, water intake, land use change, or engineering regulation, need to be incorporated into the model as boundary conditions, source-sink terms, or parameter correction terms to improve simulation accuracy.

[0029] During the model calibration and validation phase, key parameters of the selected specific model are calibrated using station observation data, flux observation data, or survey data of the target surface variables. Independent validation is then performed using observation data that does not overlap with the calibration period. The Nash efficiency coefficient (NSE), root mean square error (RMSE), and Pearson correlation coefficient are used. As calibration and validation metrics; when NSE, RMSE and When the preset thresholds are met, the model is deemed to have passed verification and reference data is generated.

[0030] In the reference data generation stage, the independently validated study area model outputs daily or hourly spatially continuous model simulation reference data R0 for the target time period. This model simulation reference data R0 is reference data that has not yet been standardized with the original remote sensing product P0 to be validated. Before entering the index calculation, P0 and R0 undergo standardization processing in terms of time scale, spatial scale, projected coordinates, variable units, vertical representation definition, and valid samples, respectively forming the standardized remote sensing product P1 to be validated and the standardized model simulation reference data R1.

[0031] In the stage of constructing the evaluation index system, four types of indicators are set: deviation or error, correlation, distribution consistency, and relative uncertainty. Their hierarchical structure is as follows: Figure 2 As shown, the calculation formulas and specific usage methods for each indicator are as follows. For each remote sensing product to be verified, the standardized remote sensing product P1 is matched with the corresponding standardized model simulation reference data R1 according to the same spatial location and observation time. Invalid data is eliminated, and a one-to-one corresponding valid spatiotemporal sample pair is formed. Then, each indicator is calculated in sequence. Each indicator is used for comprehensive evaluation and ranking among different products to be verified. Among them, the deviation or error indicator is used to measure the degree of numerical deviation of the standardized remote sensing product P1 from the standardized model simulation reference data R1; the correlation indicator is used to measure the consistency of their spatiotemporal changes; and the distribution consistency indicator is used to measure the degree of difference in their probability distributions. When there are at least three mutually independent product sequences, the triangular hat method is used to calculate the relative uncertainty index, which is the ratio of the noise standard deviation of the k-th product to the arithmetic mean of the product sequence to measure its relative noise level.

[0032] Category 1: Deviation / Error Mean Absolute Deviation (MAD):

[0033] MAD represents the average error magnitude, with an ideal value of 0.

[0034] Mean Deviation (MBD):

[0035] For each sample, retain the sign of the difference, sum them, and divide by N. A MBD greater than 0 indicates that the overall reference value is higher than the product value, and a MBD less than 0 indicates that the overall reference value is lower than the product value. The ideal value is 0.

[0036] Root Mean Square Error (RMSE):

[0037] The closer RMSE is to 0, the smaller the overall error; the ideal value is 0.

[0038] Residual standard deviation (SD):

[0039] The closer SD is to 0, the smaller the residual dispersion; the ideal value is 0.

[0040] 95% uncertainty (U95):

[0041] First calculate SD separately 2 and RMSE 2 The sums are then squared, and the result is multiplied by 1.96. U95 is used to characterize the 95% uncertainty under the original table definition, with an ideal value of 0.

[0042] t-statistic (TS):

[0043] First calculate MBD 2 RMSE 2 and RMSE 2 -MBD 2 Then use MBD 2 Divide by the difference, multiply by N-1, and then take the square root. Use the non-negative square root form as in the original table; when calculating, N>1 and the denominator must be non-negative, with an ideal value of 0.

[0044] In this category, MAD, MBD, and RMSE have been widely used in remote sensing soil moisture verification, while U95 and TS are less common in the field of soil moisture verification than the former three.

[0045] Category 2: Relevance Pearson correlation coefficient (R-Pearson):

[0046] The ideal value is 1, and the index is undefined when any sequence remains unchanged.

[0047] Spearman correlation coefficient (R-Spearman):

[0048] First and Convert to rank respectively and For parallel values, the average rank is used; then the sum of squares of the corresponding rank differences is calculated, with an ideal value of 1.

[0049] Best-fit curvature coefficients (SBF):

[0050] The ideal value is 1, and it is undefined when the denominator is 0.

[0051] Nash efficiency coefficient (NSE):

[0052] The ideal value is 1.

[0053] Leigate efficiency coefficient (LCE):

[0054] The ideal value is 1.

[0055] Wilmot Consistency Index (WIA):

[0056] The ideal value is 1.

[0057] In this category, R-Pearson, R-Spearman, and SBF range from [-1, 1], where +1 indicates perfect positive correlation, -1 indicates perfect negative correlation, and 0 indicates no correlation. For LCE and NSE, the range is [-∞, 1], where +1 indicates perfect correlation and -∞ indicates no correlation. For WIA, the range is [0, 1], where 1 indicates perfect correlation and 0 indicates no correlation.

[0058] Category 3: Distribution Consistency The third category of indicators employs the Kolmogorov–Smirnov integral (KSI) based on the cumulative distribution function and the combined performance index (CPI) based on the KSI. In statistics, the Kolmogorov–Smirnov test is typically used to determine whether there is a significant difference between the latent probability distributions of the target product being evaluated and the reference data used for verification, or whether their latent probability distributions differ significantly from the assumed distribution (such as a normal distribution) (referring only to tests based on finite samples).

[0059] Wherein, KSI is defined as: ; In the formula, x represents the distribution integral variable in the KSI formula. and These represent the maximum and minimum values ​​after merging the two sets of samples, D. n This indicates that the cumulative distribution functions of the two empirical sets at point x are distributed over the interval [0, 1]. The absolute difference within, Indicates by The critical area, determined by the length of the integration interval, is a characteristic quantity of this distribution: ; Where the coefficient It is the statistical property index of the cumulative distribution function, representing the KS critical difference and the number of points. It is inversely proportional to the square root, defined as:

[0060] in for The function can be set to a constant, representing the KS critical function value corresponding to N as specified in the original index system. Here, it is set... =1.63, The KS critical function is set according to the original indicator system, and the ideal value of the KS test integral (KSI) is 0. Comprehensive Performance Index (CPI):

[0061] All values ​​in CPI and KSI are in the same unit as soil moisture, with an ideal value of 0. KSI and CPI range from [0, ∞], where 0 indicates that their distributions are completely identical, and ∞ indicates that their distributions are completely different.

[0062] Category 4: Relative uncertainty In the absence of prior knowledge, the Three-Cornered Hat (TCH) method can be used to estimate the relative uncertainty (RU) of various soil moisture products (remote sensing, reanalysis, and assimilation soil moisture products). After considering certain constraints, the TCH method framework can be used to jointly calculate several correlated datasets to obtain the relative uncertainty of each product.

[0063] Tricorn hat relative uncertainty (TCH-RU): ;

[0064] In the TCH-RU formula This represents the i-th product sequence in the unified remote sensing product P1 to be verified. This represents a common reference quantity reflected by multiple standardized remote sensing product sequences P1. This represents the i-th noise sequence in the unified remote sensing product P1 to be verified. This represents the diagonal element in the i-th row and i-th column of the noise covariance matrix. This represents the i-th relative uncertainty calculated based on the unified remote sensing product P1 to be verified. The i-th product observation sequence... Represented as common reference quantity With noise sequence The sum; the noise covariance matrix is ​​estimated by the pairwise differences of at least three products, and the i-th diagonal element is taken. Finally, calculate according to the formula in the original table. . The closer it is to 0, the smaller the relative uncertainty. When the ratio is 0, it is undefined; the ideal value is 0.

[0065] In the product-reference data comparison phase, multiple remote sensing products P1 to be verified, after being standardized, are matched with the model simulation reference data R1, which has also been standardized, according to the same spatial grid, observation time, variable definition, projected coordinates, and valid samples. Then, the aforementioned indicators are calculated for each product, each grid, and each time period, and can be processed according to... Figure 3 Output the time series comparison results between the unified remote sensing product P1 to be verified and the unified model simulation reference data R1.

[0066] When the target surface variables have definite process response quantities, radar charts are used to display the correlation between each product and the process response quantity simulated by the model. In soil moisture validation, the difference between the infiltration amount I and the evapotranspiration amount ET calculated by the model, I-ET, is used as the net vertical water input. Pearson correlation coefficients and Spearman correlation coefficients between the soil moisture series and the I-ET series of each product are calculated to assess the product's ability to capture vertical water migration processes.

[0067] In the environmental factor contribution analysis phase, multiple environmental factors related to the target variable are collected, and a heatmap of the spatial distribution of environmental factors and their contribution to product uncertainty is output, such as... Figure 6 As shown. In soil moisture verification, the environmental factors may include DEM, sand content, clay content, porosity, saturated hydraulic conductivity Ks, organic matter content, leaf area index (LAI), and root abundance (RA). A generalized additive model (GAM) is used to construct the nonlinear relationship between relative uncertainty (RU) and environmental factors. For each product, grid-scale relative uncertainty is used as the response variable, and standardized values ​​of topographic elevation, sand content, clay content, porosity, saturated hydraulic conductivity, organic matter content, leaf area index, and root abundance are used as explanatory variables, with the number of effective samples as a covariate. By restricting the basic dimension of splines and using generalized cross-validation to select smoothing parameters, the following equation is fitted:

[0068] This is the relative uncertainty, which is essentially the expectation of the dependent variable. and For unknown parameters, For the first One nonlinear continuous independent variable This is a nonparametric smoothing function used to construct the nonlinear response relationship between the relative uncertainty of soil moisture products and the independent variable. =8), This represents the error term corresponding to the GAM model.

[0069] In this invention, the deviation explained by the GAM model is... The key output for quantifying the contribution of explanatory variables to the total relative uncertainty of the target soil moisture product is as follows:

[0070] In the formula, and These represent the residual deviation (RU) and null deviation (ND) of the GAM model, respectively. By analyzing the deviation explained by the GAM model and the significance of environmental factors, the dominant factors influencing the uncertainty of each product are identified.

[0071] In the authenticity verification results output stage, a comprehensive evaluation radar chart, an environmental factor impact heat map, a key factor effect spline curve, and a graph showing the relationship between relative uncertainty (RU) and other statistical indicators are generated for each product. Finally, a remote sensing product authenticity verification report is generated, providing information on product accuracy, applicability level, sources of uncertainty, and algorithm improvement suggestions.

[0072] The present invention also provides a system implementation method. The system includes a data management module, a model building and calibration module, a reference data generation module, a product consistency processing module, a multi-index verification module, an uncertainty diagnosis module, and a result output and update module that are connected in sequence. Each module executes Step 1 to Step 7. The result output and update module feeds back the product version, model version, and authenticity verification results to the data management module for rolling updates.

[0073] The present invention can also be implemented as a computer program, which is stored in a computer-readable storage medium and executed by a processor.

[0074] The following section uses the irrigation area in the middle reaches of the Heihe River Basin as an example to illustrate the specific application of this invention.

[0075] 1. Study Area and Data The study area is located in the middle reaches of the Heihe River Basin (100°~101°E, 38.5°~39.5°N). The terrain of this area transitions from the northern foothills of the Qilian Mountains to the center of the basin. It is a typical agricultural oasis area with low precipitation (100-200 mm) and high evaporation (1200-1800 mm). There is large-scale surface water and groundwater irrigation.

[0076] Collect the following data according to step 1: Driving data: meteorological data (precipitation, temperature, radiation, etc.), DEM (ASTER-GDEM), river network (MERITHydro), land cover (30m classification data), soil parameters (GSDE dataset), and aquifer parameters (GLHYMPS).

[0077] Irrigation data: Dataset of monthly surface water and groundwater irrigation volume in the Heihe River Basin with a resolution of 30 arcseconds. (2016)

[0078] Products to be verified: 10 products were selected, including JAXA (AMSR), LPRM-C, LPRM-X, SMOS-IC, ESA CCI-A, ESA CCI-P, ESA CCI-C, GLEAM-A, GLEAM-B and GLDAS-Noah.

[0079] Environmental factors: DEM, sand, clay, porosity, saturated hydraulic conductivity Ks, organic matter content, LAI, root abundance RA.

[0080] Ground validation data: Soil moisture observation data from the Daman Super Station, including TDRP (2cm depth) and CRNP (360m radius).

[0081] 2. ParFlow-CLM Model Construction and Reference Data Generation A ParFlow-CLM model was constructed for the Heihe River Basin, with a horizontal resolution of 0.005° (approximately 500m) and a vertical division of the soil aquifer system into 7 layers (6 soil layers + 1 aquifer layer, with a total thickness of 102m). Irrigation data was incorporated into the model as an upper boundary condition.

[0082] The model was driven by data from 2010 to 2013. Model calibration results show that the R-Pearson ratio of simulated soil moisture to CRNP observations reaches 0.93, and the RMSE is <0.06 m. 3 / m 3 ( Figure 3The model accurately captures the rising trend of soil moisture during the irrigation period (June-September) and the falling trend during winter. Therefore, the daily surface soil moisture (resampled to 0.25°) simulated by ParFlow-CLM from 2011 to 2013 was used as a reference dataset to test other products.

[0083] 3. Verification and Result Analysis of Multi-Indicator Authenticity Based on the 15 indicators in Table 1, the accuracy of 10 products relative to the ParFlow-CLM reference data was calculated. Key results are shown in Table 2.

[0084] Table 1. Statistical Table of Accuracy Evaluation Indicators for 10 Soil Moisture Products in the Heihe River Basin (Partial)

[0085] The results show that GLDAS products have the lowest MAD, RMSE, and RU among the products listed in Table 2, indicating superior overall performance. GLEAM-B has a higher NSE, and CCI-C has a lower relative uncertainty than most single remote sensing source products. LPRM series products suffer from overestimation or insufficient correlation. SMOS-IC has a large dynamic range but also the highest uncertainty. Overall, fused, reanalyzed, or assimilated products outperform most single remote sensing source products in this embodiment.

[0086] The correlation analysis results between each product and the infiltration-evapotranspiration (I-ET) simulated by the model showed that the JAXA product had the highest correlation with I-ET (R=0.42), indicating that it was most sensitive to the vertical water migration process; while the LPRM product had a negative correlation with I-ET, indicating that it had a deficiency in capturing water infiltration after irrigation / rainfall.

[0087] 4. Analysis of the impact of environmental factors on product uncertainty The contribution of eight environmental factors to RU was quantified using the GAM model. The results (Table 3) show that: LAI is the leading factor for all products (except SMOS-IC and CCI-A), contributing between 10.1% and 17.8% to the explanation of RU.

[0088] For assimilation products (GLEAM, GLDAS), the impact of LAI shows a monotonically decreasing trend, indicating that they are less sensitive to vegetation cover and the algorithm is more robust.

[0089] For a single remote sensing product (JAXA, SMOS-IC), the influence of LAI fluctuates non-linearly, with an inflection point at LAI ≈ 0.6, indicating that the accuracy of existing inversion algorithms is unstable in areas with medium vegetation cover.

[0090] Table 2. Dominant environmental factors for soil moisture products identified by the GAM model.

[0091] Further analysis revealed that the spatial variability coefficients of saturated hydraulic conductivity Ks and LAI showed the strongest correlation with the spatial variability coefficient of soil moisture, such as... Figure 6 As shown, this indicates that soil physical properties (Ks) and surface vegetation properties (LAI) are the two most critical environmental factors controlling the spatial distribution of soil moisture and the accuracy of product authenticity verification.

[0092] 5. Conclusion and Authenticity Verification Report Output Based on the above analysis, an authenticity verification report will be output: GLDAS products have the highest overall accuracy in the middle reaches of the Heihe River Basin; GLEAM-B performs well in terms of NSE, and CCI-C has relatively low uncertainty. They can be selected for different applications such as water resource management, drought monitoring, or uncertainty control.

[0093] While JAXA products are systematically undervalued, they are sensitive to vertical water processes and can be used for irrigation event identification and soil moisture memory studies.

[0094] The LPRM product performs poorly in this region and its inversion algorithm needs to be improved under conditions of high vegetation cover and irrigation.

[0095] Environmental factor impact analysis indicates that future improvements to soil moisture products should focus on enhancing the ability to characterize the spatial heterogeneity of LAI and Ks.

[0096] This embodiment uses ParFlow-CLM simulated soil moisture as reference data to demonstrate the effectiveness and practicality of the model simulation-based remote sensing product authenticity verification method proposed in this invention. This method overcomes the spatial limitations of ground observation and can provide a scientific basis for the accuracy evaluation and algorithm improvement of multi-source remote sensing products.

[0097] The above embodiments illustrate the preferred technical path of the present invention. The present invention is not limited to a specific watershed, specific model version, specific computing platform, or fixed threshold.

[0098] All equivalent technical solutions that generate reference data based on model simulation with observational constraints and use a multi-level index system to verify the authenticity of remote sensing products and diagnose the sources of product uncertainty should fall within the protection scope of this invention; ParFlow-CLM simulation of soil moisture is only one specific embodiment of this invention.

Claims

1. A method for verifying the authenticity of remote sensing products based on model simulation, characterized in that, Includes the following steps: Step 1: Collect model-driven data, model parameter data, ground observation data, and human activity data related to the target surface variables represented by the remote sensing product P0 to be validated within the study area; Step 2: Select a mechanistic model or a coupled model based on the target surface variables, construct a study area model using the driving data and parameter data of the mechanistic model or coupled model, and perform parameter calibration and independent verification of the study area model using ground observation data; Step 3: Run the independently validated study area model to generate spatiotemporal continuous model simulation reference data R0 for the target surface variables; Step 4: Perform spatial scale, time scale, projection coordinates, variable definition and effective sample consistency processing on the remote sensing product P0 to be verified and the model simulation reference data R0 to obtain the consistent remote sensing product P1 to be verified and the model simulation reference data R1. Step 5: Compare the standardized remote sensing product P1 to be verified with the model simulation reference data R1 in spatiotemporal samples, calculate the deviation or error index, correlation index, distribution consistency index and relative uncertainty index, and obtain the product accuracy evaluation result. Step 6: Establish a generalized summation model with the relative uncertainty index as the response variable and environmental factors related to the target surface variable as explanatory variables. Determine the dominant environmental factors of product uncertainty based on the significance and deviation explanation rate of each environmental factor in the generalized summation model. Step 7: Based on the product accuracy evaluation results and the dominant environmental factors, output the accuracy, applicability level, uncertainty sources, and algorithm improvement suggestions of the unified remote sensing product P1 to be verified. Record the original remote sensing product P0 version, the unified remote sensing product P1 version, the study area model version, and the model-driven data or ground observation data so that the authenticity verification results can be updated on a rolling basis when any of the above versions change.

2. The method for verifying the authenticity of remote sensing products based on model simulation according to claim 1, characterized in that: In Step 1: The target surface variables are soil moisture, evapotranspiration, surface temperature, or vegetation index. The model-driven data includes precipitation, temperature, radiation, wind speed, air pressure, and humidity; The model parameter data includes topographic elevation, land cover, and surface parameters that control the change process of target surface variables; The ground observation data are station observation data, flux observation data, or survey data of the target surface variables; The human activity data refers to irrigation data, water intake data, land use change data, or engineering control data that affect the target surface variables.

3. The method for verifying the authenticity of remote sensing products based on model simulation according to claim 2, characterized in that: Step 2 includes the following steps: When the target surface variable is soil moisture, the coupling model is the ParFlow-CLM coupling model. The ParFlow module is used to calculate the three-dimensional variable saturated groundwater flow and surface runoff, and the CLM module is used to calculate the land surface energy balance and evapotranspiration. The effects of irrigation, water intake, land use change, and engineering regulation on target surface variables are incorporated into the coupled model as boundary conditions, source-sink terms, or parameter correction terms, and the Nash efficiency coefficient, root mean square error, and Pearson correlation coefficient are used as parameter calibration and independent verification indicators.

4. The method for verifying the authenticity of remote sensing products based on model simulation according to claim 1, characterized in that: In Step 3: The model simulation reference data R0 is the hourly or daily spatial continuous data of the target surface variables output by the model in the study area under the model's native spatial grid and temporal resolution, which is used as reference data for subsequent consistency processing.

5. The method for verifying the authenticity of remote sensing products based on model simulation according to claim 1, characterized in that: Step 4 specifically includes: Step 4.1: Perform time aggregation or time interpolation on the model simulation reference data R0 based on the observation time of the remote sensing product P0 to be verified, so that the two have the same time scale and corresponding observation time; Step 4.2: Based on the pixel boundaries of the remote sensing product P0 to be verified, the model simulation reference data R0 is aggregated or resampled using area weighting to make the model simulation reference data R0 and the remote sensing product P0 to be verified have the same spatial grid. Based on the spatial grid and cell boundaries of the remote sensing product P0 to be verified, area-weighted aggregation or resampling method is selected for the model simulation reference data R0 to make the remote sensing product P0 to be verified and the model simulation reference data R0 have the same spatial grid. Step 4.3: Unify the remote sensing product P0 to be verified and the model simulation reference data R0 to the same projection coordinates; Step 4.4: Standardize the vertical representation definition based on the vertical representation method of the target surface variables: For target surface variables that have the meaning of soil layer stratification, standardize the soil layer depth; for target surface variables that do not have the meaning of soil layer stratification, standardize the vertical representation range. Step 4.5: Following the same valid sample screening rules, perform valid sample screening on the remote sensing product P0 to be verified and the model simulation reference data R0 to form a consistent remote sensing product P1 to be verified and model simulation reference data R1 with a one-to-one correspondence.

6. The method for verifying the authenticity of remote sensing products based on model simulation according to claim 1, characterized in that: In Step 5: The deviation or error indicators include mean absolute deviation, mean deviation, root mean square error, residual standard deviation, 95% uncertainty and t-statistic; The correlation indices include Pearson correlation coefficient, Spearman correlation coefficient, best fit curvature coefficient, Nash efficiency coefficient, Leighgate efficiency coefficient, and Wilmot consistency index. The distribution consistency index includes the Kolmogorov-Smirnov test integral and the comprehensive performance index; The relative uncertainty index is the relative uncertainty calculated based on the triangular cap method; In the i-th valid spatiotemporal sample participating in the index calculation, N represents the total number of valid paired samples, and i represents the i-th valid spatiotemporal sample. This represents the value of the remote sensing product P1 to be verified after standardization in the i-th valid spatiotemporal sample. This represents the value of the model simulation reference data R1 after standardization in the i-th valid spatiotemporal sample. This represents the arithmetic mean of the P1 values ​​of all remote sensing products to be verified. This represents the arithmetic mean of all model simulation reference values ​​R1, with the residuals uniformly denoted as... Unless otherwise specified, all indicators use the same batch of valid paired samples; When there are at least three independent product sequences, the triangular hat method is used to solve for the noise variance of the k-th product based on the covariance of the pairwise differences of each sequence. and in accordance with Calculate the k-th product sequence relative uncertainty .

7. The method for verifying the authenticity of remote sensing products based on model simulation according to claim 2, characterized in that: In Step 6: The environmental factors are determined based on the physical processes that control the spatial changes of surface variables of the target area; When the target surface variable is soil moisture, the environmental factors include topographic elevation, sand content, clay content, porosity, saturated hydraulic conductivity, organic matter content, leaf area index, and root abundance. After standardizing each environmental factor, the nonlinear relationship between the relative uncertainty and each environmental factor is fitted using a spline smoothing function. The dominant environmental factor is then determined based on the significance test results and the contribution ratio of the deviation explanation rate of each environmental factor.

8. The method for verifying the authenticity of remote sensing products based on model simulation according to claim 1, characterized in that: In Step 7: The product accuracy evaluation results are divided into three applicability levels: applicable, limited applicable, and not applicable, according to preset thresholds. After completing step 7, monitor the P1 version of the remote sensing product to be verified after the standardization process, the model version of the study area, the R1 version of the model simulation reference data after the standardization process, and the ground observation data version. When the original P0 version of the remote sensing product to be verified, the P1 version of the remote sensing product to be verified after the standardization process, the model version of the study area, the model-driven data, or the ground observation data change, return to step 4 and retain the results of this and previous authenticity verifications.

9. A model-simulation-based system for verifying the authenticity of remote sensing products, characterized in that, The method for verifying the authenticity of remote sensing products based on model simulation as described in any one of claims 1-8 includes a data management module, a model building and calibration module, a reference data generation module, a product standardization processing module, a multi-index verification module, an uncertainty diagnosis module, and a result output and update module, which are connected in sequence via communication. The data management module is used to execute Step 1; the model building and calibration module is used to execute Step 2; the reference data generation module is used to execute Step 3; the product standardization processing module is used to execute Step 4; the multi-index verification module is used to execute Step 5; the uncertainty diagnosis module is used to execute Step 6; and the result output and update module is used to execute Step 7, recording the authenticity verification results and related version information. When changes are detected in the original remote sensing product version P0 to be verified, the standardized remote sensing product version P1 to be verified, the model version of the study area, the model-driven data, or the ground observation data, the next round of authenticity verification process is triggered.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the model simulation-based remote sensing product authenticity verification method according to any one of claims 1 to 8.