Aod ground-based measurement reference true value evaluation method and product authenticity inspection method

Abstract

The present application relates to the technical field of atmospheric environment remote sensing, and particularly relates to an AOD ground-based measurement reference true value evaluation method and a product authenticity inspection method, wherein the AOD ground-based measurement reference true value evaluation method comprises obtaining a ground reference true value and a first uncertainty index and a second uncertainty index of the ground reference true value; the first uncertainty index is used to indicate AOD inspection uncertainty introduced due to time difference; and the second uncertainty index is used to indicate AOD inspection uncertainty introduced due to spatial scale difference. The above scheme quantitatively separates and evaluates the inspection uncertainty introduced due to time-space scale mismatch by constructing a time variation model and a spatial variation model, and solves the problem that the intrinsic precision of a product cannot be scientifically evaluated due to the mixing of time-space representation deviation and satellite inversion error in a traditional method.

Description

Technical Field

[0001] This invention relates to the field of atmospheric environment remote sensing technology, and more specifically, to a method for evaluating the ground-based reference true value of AOD measurements and a method for verifying the authenticity of products. Background Technology

[0002] Aerosol optical depth (AOD) is a key parameter characterizing the content of aerosols in the atmosphere and their effect on the attenuation of solar radiation. It has significant scientific value in climate change research, atmospheric environmental monitoring, air quality assessment, and quantitative inversion of remote sensing data. In recent years, with the rapid development of aerospace remote sensing technology, many satellite platforms equipped with dedicated optical payloads (such as MODIS, VIIRS, Himawari, and MERSI) have the capability to produce aerosol optical depth products with global coverage and high-frequency observations. These satellite remote sensing products provide indispensable data support for macroscopic atmospheric environmental research, and their accuracy and reliability directly determine the credibility of related scientific conclusions and operational applications. To ensure the quality of satellite AOD products, ground-based solar photometer (such as CE318) observation networks are typically used as "reference values," and their authenticity is verified through satellite-to-ground comparison.

[0003] However, existing verification technologies have a fundamental flaw: a mismatch in spatiotemporal scales exists between ground-based single-point measurements and satellite pixel observations. Spatially, ground-based stations provide point measurements representing a range of tens to hundreds of meters, while satellite pixels typically cover an average area of ​​several to tens of kilometers. This difference in spatial representativeness introduces systematic biases under conditions of non-uniform aerosol distribution. Temporally, satellite transit times are fixed (e.g., MODISTerra satellite transits approximately at 10:30 AM local time daily), while ground-based observations are discrete sampling. This unavoidable temporal difference leads to spurious errors during periods of rapid aerosol change (e.g., pollution events, weather process evolution). The uncertainties introduced by this spatiotemporal scale difference cannot be quantitatively assessed or effectively eliminated using existing technologies, making it difficult for the verification results to scientifically reflect the inversion accuracy of the satellite products themselves, thus affecting the quality improvement and application effectiveness of remote sensing data products. Summary of the Invention

[0004] The purpose of this invention is to provide a solution to the above-mentioned technical problems.

[0005] In a first aspect, embodiments of the present invention provide a method for evaluating the true reference value of AOD foundation measurement, the method comprising:

[0006] Obtain a ground reference true value and a first uncertainty index and a second uncertainty index of the ground reference true value; wherein, the first uncertainty index is used to indicate the uncertainty of AOD test due to time difference; and the second uncertainty index is used to indicate the uncertainty of AOD test due to spatial scale difference.

[0007] The method for determining the first uncertainty index includes: collecting high-frequency time-series data from a ground-based solar photometer on the day the satellite passes overhead; constructing a time variation model based on the time-series data; quantifying the uncertainty introduced by the time difference based on the time variation model, and obtaining the first uncertainty index;

[0008] The method for determining the second uncertainty index includes: acquiring satellite aerosol optical thickness image products that are closest to the measurement time of the ground-based solar photometer; constructing a spatial variation model in the station pixel-scale region based on the image products; quantifying the uncertainty introduced by spatial scale differences based on the spatial variation model to obtain the second uncertainty index.

[0009] Furthermore, the construction of the time-varying model based on the time-series data includes:

[0010] A subset of observations within a continuous time window is extracted from the high-frequency time-series data, centered on the satellite transit time.

[0011] The observation subset is optimized and fitted to obtain the time variation trend function of aerosol optical thickness.

[0012] Furthermore, the optimization fitting aims to minimize the error between the calculated value of the trend function and the measured data for the corresponding time period.

[0013] Furthermore, the optimization fitting of the observed subset includes:

[0014] The observation subset is fitted with a time series regression method to obtain the trend function; wherein the time series regression method includes at least one of multinomial regression, spline regression, moving average filtering, or autoregressive model.

[0015] Furthermore, the step of quantifying the uncertainty introduced by time differences based on the time variation model to obtain the first uncertainty index includes:

[0016] The aerosol optical thickness value is calculated at each moment within the satellite transit time window based on the time variation model; wherein, the time window is centered on the satellite observation time and covers the time difference between the ground observation time and the satellite observation time.

[0017] The calculated values ​​of aerosol optical thickness at all times within the time window are statistically analyzed, and the standard deviation of this set of calculated values ​​is used as the first uncertainty index.

[0018] Furthermore, the construction of a spatial variation model based on the image product at the site pixel scale region includes:

[0019] Extract pixel datasets of pixel-scale regions centered on ground stations from the satellite aerosol optical thickness image products;

[0020] The spatial variation model is obtained by fitting the pixel dataset.

[0021] Further, fitting the pixel dataset includes:

[0022] The spatial trend surface modeling method is used to fit a function to the pixel dataset to obtain the spatial variation model; wherein, the spatial trend surface modeling method includes at least one of first-order or second-order polynomial trend surface fitting, Kriging interpolation or inverse distance weighted interpolation.

[0023] Furthermore, the step of quantifying the uncertainty introduced by spatial scale differences based on the spatial change model to obtain the second uncertainty index includes:

[0024] The aerosol optical thickness value at the ground station location is calculated based on the spatial variation model, and the regional average value of the aerosol optical thickness value of all pixels within the pixel-scale region is calculated.

[0025] The calculated value of the station location is compared with the average value of the region, and the deviation is used as the second uncertainty index.

[0026] Furthermore, the first uncertainty index and the second uncertainty index are used to correct the spatiotemporal representativeness bias in the authenticity verification of aerosol optical thickness satellite remote sensing products.

[0027] Secondly, embodiments of the present invention provide a method for verifying the authenticity of AOD products, the method comprising:

[0028] Obtain the true reference value of aerosol optical thickness measured by a ground-based solar photometer and the satellite aerosol optical thickness product data to be tested;

[0029] The first uncertainty index and the second uncertainty index of the aerosol optical thickness reference true value are determined by the method provided in the first aspect or any possible implementation of the first aspect;

[0030] Based on the first uncertainty index and the second uncertainty index, the spatiotemporal representativeness deviation correction is applied to the comparison test results between the satellite aerosol optical thickness product data and the reference true value to obtain the corrected authenticity test results.

[0031] Output the corrected authenticity test results.

[0032] The aforementioned scheme quantitatively separates and evaluates the verification uncertainty introduced by the mismatch between spatiotemporal scales by constructing temporal and spatial variation models. This solves the problem that the intrinsic accuracy of products cannot be scientifically evaluated due to the mixture of spatiotemporal representativeness bias and satellite inversion error in traditional methods. On the other hand, it uses time series regression to optimize and fit high-frequency time series data, and employs spatial trend surface modeling to fit functions to pixel-scale regions. This achieves continuous characterization of aerosol optical thickness variation trends at minute-level temporal resolution and pixel-level spatial resolution, reducing the idealized dependence on assumptions about aerosol dynamic evolution and spatial heterogeneity. Furthermore, the scheme is platform-independent and process-embeddable, making it applicable to the cross-validation of multi-source ground-based observation networks and satellite remote sensing products. By outputting quantitative spatiotemporal uncertainty indicators to support the generation of quality markers and uncertainty-based weight allocation, it promotes the transformation of remote sensing product verification from qualitative comparison to quantitative measurement.

[0033] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart illustrating the first uncertainty determination scheme in the AOD foundation measurement reference truth evaluation method provided in this embodiment of the invention.

[0036] Figure 2 This is a flowchart illustrating the second uncertainty determination scheme in the AOD foundation measurement reference truth evaluation method provided in this embodiment of the invention.

[0037] Figure 3A flowchart illustrating the AOD foundation measurement reference truth evaluation method in a certain application scenario provided in this application embodiment;

[0038] Figure 4 This is a schematic diagram of the time change modeling results in a certain application scenario provided by an embodiment of the present invention;

[0039] Figure 5 This is a schematic diagram of the spatial change modeling results in a certain application scenario provided by an embodiment of the present invention. Detailed Implementation

[0040] The embodiments of the technical solution of the present invention will be described in detail below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only examples, and should not be used to limit the scope of protection of the present invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and the foregoing description of the drawings are intended to cover non-exclusive inclusion. In the description of the embodiments of the present invention, technical terms such as "first," "second," etc., are only used to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly indicating the number, specific order, or primary or secondary relationship of the indicated technical features. In the description of the embodiments of the present invention, "a plurality of" means two or more, unless otherwise explicitly specified. The reference to "embodiment" herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0041] This invention provides a method for evaluating the true reference value of AOD foundation measurement, comprising:

[0042] Obtain the ground reference true value and the first uncertainty index and the second uncertainty index of the ground reference true value; wherein, the first uncertainty index is used to indicate the uncertainty of the AOD test due to time difference; and the second uncertainty index is used to indicate the uncertainty of the AOD test due to spatial scale difference.

[0043] The aforementioned aerosol optical depth (AOD) refers to a core physical parameter used in atmospheric science and remote sensing to quantitatively characterize the ability of aerosol particles in the entire atmospheric column to attenuate solar radiation. Mathematically, it is the integral of the aerosol extinction coefficient along a vertical path from the Earth's surface to the top of the atmosphere, usually expressed in dimensionless form. AOD directly reflects the comprehensive extinction effect of atmospheric aerosol concentration, particle size distribution, chemical composition, and vertical distribution characteristics on incident solar radiation, including both scattering and absorption processes. In satellite remote sensing quantitative inversion and atmospheric environmental monitoring applications, AOD, as a key atmospheric state variable and radiative forcing factor, is of significant scientific importance for accurate measurement in climate change research, air quality assessment, atmospheric correction, and the verification of remote sensing product accuracy.

[0044] The aforementioned ground-based AOD measurement reference true value refers to the aerosol optical thickness observation value in a specific band (usually the center wavelength of 550nm) obtained by directly observing solar radiation through a high-precision solar photometer (such as the CE318 model) deployed on the ground, based on the Beer-Lambert law inversion. This value is recognized by the international scientific community and remote sensing product verification system as the accuracy evaluation benchmark for satellite-inverted aerosol optical thickness products. Spatially, such ground-based observations characterize the aerosol extinction characteristics of the atmospheric column at a scale of tens to hundreds of meters above the site, and temporally reflect the instantaneous atmospheric state at the time of observation. It has the advantages of high observation accuracy, high spectral resolution, and complete calibration traceability chain. However, it is essentially a discrete point measurement and the time sampling is constrained by the observation conditions. This constitutes a strong scale mismatch with the spatial average and temporal differences represented by satellite remote sensing products. Therefore, in the authenticity verification, it is necessary to evaluate its spatiotemporal representativeness and use scientific characterization as the applicability boundary of the reference true value.

[0045] It is understandable that in verifying the authenticity of aerosol optical thickness satellite remote sensing products, the core premise for using ground-based solar photometer measurements as the reference true value is that their spatiotemporal representativeness must match the satellite observation scale. However, ground-based observations are essentially discrete point sampling, which can only characterize the local atmospheric column characteristics within a range of tens to hundreds of meters in space, while satellite pixel inversion values ​​are essentially a column extinction weighted average of a surface area of ​​several to tens of kilometers. There is an inherent spatial scale difference between the two. Simultaneously, the inherent temporal offset between the satellite's transit time and the ground observation time means that during rapid aerosol changes (such as pollution transport and boundary layer evolution), the observed atmospheric state is not the same. Traditional verification methods mix the pseudo-errors introduced by this spatiotemporal scale mismatch with errors in the satellite inversion algorithm, making it impossible to scientifically identify the root cause of product deviations, leading to distorted assessments of product accuracy and even misleading the direction of algorithm improvement. The introduction of the first uncertainty index aims to quantitatively characterize the dispersion of aerosol optical thickness variation caused by differences in observation time in the temporal dimension. The introduction of the second uncertainty index aims to quantify the representativeness deviation caused by the average difference between station location and pixel area in the spatial dimension. By incorporating the two as independent parameters into the verification system, the spatiotemporal pseudo-errors and satellite intrinsic errors can be separated and attributed, giving the reference true value a quantifiable confidence label, thereby improving the scientificity and objectivity of the verification results and supporting the standardization of remote sensing product accuracy evaluation and the establishment of an uncertainty transmission system.

[0046] Please see Figure 1 The following describes how the first uncertainty indicator mentioned above is determined:

[0047] The methods for determining the first uncertainty indicator include:

[0048] Step S110: Collect high-frequency time-series data from the ground-based solar photometer on the day the satellite passes overhead.

[0049] It is understandable that, to meet the technical requirements of data representativeness and sampling density for the aerosol optical thickness temporal variation model, high-frequency time-series data on the day of satellite transit is needed. Aerosol optical thickness exhibits certain nonlinear fluctuations on an hourly scale due to the influence of diurnal boundary layer variations, pollution transport processes, cloud-aerosol interactions, and meteorological dynamics. Traditional sparse sampling schemes cannot analyze such rapid evolutionary characteristics. High-frequency time-series data on the day of satellite transit can accurately capture the entire process of atmospheric state evolution that matches the satellite's spatiotemporal observations, ensuring that the constructed temporal variation model reflects the true dynamics of aerosol extinction characteristics on that specific date, rather than historical climatological averages, thus avoiding model bias introduced by differences in atmospheric processes on different days. High-frequency sampling (usually referring to a time resolution better than 15 minutes) provides sufficient degrees of freedom and information on inflection points for model fitting, supporting the ability to reconstruct the continuous trend of aerosol optical thickness at the minute level. This allows the aerosol optical thickness variation within any time difference between satellite observation time and ground observation time to be accurately interpolated and estimated. Ultimately, this ensures that the standard deviation calculation of the first uncertainty index is based on complete statistical samples rather than sparse extrapolation, improving the quantification accuracy and reliability of the uncertainty introduced by time differences.

[0050] Step S120: Construct a time variation model based on time series data.

[0051] Optionally, step S120 may include: extracting a subset of observations within a continuous time window from high-frequency time-series data, centered on the satellite transit time; optimizing and fitting the subset of observations to obtain a time-varying trend function of aerosol optical thickness.

[0052] Optionally, the above optimization fitting aims to minimize the error between the calculated value of the trend function and the measured data for the corresponding time period.

[0053] Optionally, the above-mentioned optimization fitting of the observation subset includes: using a time series regression method to fit a function to the observation subset to obtain a trend function; wherein, the time series regression method includes at least one of multinomial regression, spline regression, moving average filtering, or autoregressive model.

[0054] The aforementioned time-varying model refers to a function characterizing the continuous evolution of aerosol optical thickness over time, constructed by mathematically fitting a subset of observations within a continuous time window centered on the satellite's transit time using a time series regression method based on high-frequency time-series data collected by a ground-based solar photometer on the day of satellite transit.

[0055] The aforementioned observation subset uses the satellite transit time as the time reference center, extracting a local set of observation data from the complete high-frequency time-series data sequence that meets the requirements of continuity and representativeness. The extraction process defines a continuous time window centered on the satellite observation time, covering the time difference between the ground observation time and the satellite observation time, and extending symmetrically or asymmetrically to both sides. This ensures that the extracted observation subset contains all observation samples sufficient to characterize the dynamic evolution of aerosol optical thickness before and after the satellite transit, while guaranteeing the integrity and continuity of the time series. This avoids model fitting distortion caused by missing data or excessively large sampling intervals, thus providing a statistically significant and densely distributed input dataset for subsequent construction of time-varying models.

[0056] The core technical approach to optimizing and fitting the observation subset to obtain the time-varying trend function of aerosol optical thickness lies in using time series regression to perform continuous mathematical reconstruction of discrete observation data. The optimization process first requires selecting at least one algorithmic framework from polynomial regression, spline regression, moving average filtering, or autoregressive models that suits the time-varying characteristics and time sampling density of aerosol optical thickness. This framework approximates the measured aerosol optical thickness values ​​at each time step within the observation subset, constructing a continuously differentiable trend function expression for the time independent variable. During the fitting implementation phase, the global optimization objective is to minimize the error between the calculated value of the trend function at each time step within the observation subset's time domain and the corresponding measured data. Parameter estimation methods such as least squares estimation, maximum likelihood estimation, or Bayesian inference are used to solve for the optimal parameter set of the model. This ensures that the obtained time-varying trend function can both fully fit the dynamic evolution characteristics of the observation subset and smooth high-frequency noise and random fluctuations, ultimately forming a continuous function representation with minute-level time resolution interpolation capabilities, providing a complete numerical calculation interface for subsequent uncertainty quantification.

[0057] The aforementioned polynomial regression refers to performing global least-squares fitting on a subset of observations based on a family of polynomial functions. It constructs a single continuous function by estimating the polynomial coefficients to characterize the overall evolution trend of aerosol optical thickness over time. Spline regression, on the other hand, uses piecewise defined polynomial functions to fit data between adjacent time intervals, applying continuity and smoothness constraints at nodes to flexibly capture the local nonlinear variation characteristics of aerosol optical thickness. Moving average filtering is a time-domain low-pass filtering technique that smooths data through a weighted average of observations within a sliding time window, effectively suppressing high-frequency random noise and extracting the main trend components. The autoregressive model represents the current value of aerosol optical thickness as a linear combination of its historical lag terms, utilizing serial autocorrelation to construct a predictive model. It is understood that the aforementioned polynomial regression, spline regression, moving average filtering, or autoregressive models are all mature existing technologies. For their specific implementation methods and working principles, please refer to relevant technologies; these will not be elaborated further in this embodiment.

[0058] Step S130: Quantify the uncertainty introduced by time difference based on the time change model to obtain the first uncertainty index.

[0059] Optionally, step S130 may include: calculating the aerosol optical thickness value at each moment within the satellite transit time window based on a time variation model; wherein the time window is centered on the satellite observation time and covers the time difference range between the ground observation time and the satellite observation time; statistically analyzing the calculated aerosol optical thickness values ​​at all moments within the time window, and using the standard deviation of the set of calculated values ​​as the first uncertainty index.

[0060] The uncertainty introduced by the aforementioned time difference refers to the verification error component caused by the dynamic evolution of atmospheric aerosol optical thickness within the time shift window, due to the asynchronous time offset between the ground-based solar photometer observation time and the satellite payload's overflight observation time. The physical root of this uncertainty lies in the fact that aerosol concentration, particle size distribution, and vertical structure are regulated by diurnal boundary layer variations, pollution transport processes, and meteorological dynamics, exhibiting certain non-stationary fluctuations on a minute-to-hour scale. This results in atmospheric extinction states observed at two different times not being the same physical realization. Traditional verification methods ignore this temporal evolution effect, directly comparing observations from different times, leading to a mixture of time-shift spurious errors and satellite inversion errors, making it impossible to scientifically attribute the true source of product deviation. The above scheme reconstructs the continuous evolution of aerosol optical thickness within the time window by constructing a time-varying model, and uses the standard deviation of the set of model calculation values ​​at each moment within the window as a quantitative indicator of time difference uncertainty. This standard deviation characterizes the degree of dispersion of natural variation of aerosol optical thickness caused by asynchronous observation times, thereby achieving quantitative separation and elimination of this part of pseudo-errors, and ensuring that the authenticity test results only reflect the intrinsic accuracy of satellite products.

[0061] Step S130 reconstructs the continuous evolution of aerosol optical thickness within the satellite transit time window based on a time-varying trend function, and quantifies the degree of variability dispersion within this time domain using probabilistic statistical methods. Specifically, a continuous time window is defined, centered on the satellite observation time, covering the time difference between the ground observation time and the satellite observation time. The constructed time variation model is used to perform numerical calculations at each discrete moment within this window, obtaining a set of calculated aerosol optical thickness values. The standard deviation of this set is the first uncertainty index, mathematically representing the second-moment measure of the degree of variation in aerosol optical thickness within the time domain. A larger standard deviation indicates more severe natural fluctuations in aerosol optical thickness due to atmospheric dynamics within this time window, and a higher level of uncertainty introduced by the asynchronous observation times. This provides a quantifiable correction basis for the authenticity verification results based on time representativeness bias.

[0062] Please see Figure 2 The determination method for the second uncertainty indicator mentioned above is described below:

[0063] The methods for determining the second uncertainty indicator include:

[0064] Step S210: Collect satellite aerosol optical thickness image products that are closest to the measurement time of the ground-based solar photometer.

[0065] The aforementioned satellite aerosol optical thickness image products refer to quantitative remote sensing data products generated from multispectral / multi-angle atmospheric top radiance observation data acquired by optical remote sensors mounted on satellite platforms, after radiometric calibration, cloud detection, aerosol inversion algorithm processing, and atmospheric radiative transfer model calculations. The data format is a gridded image structure, with each pixel value representing the aerosol optical thickness inversion result of the entire atmospheric column within that spatial resolution unit. These products are typically distributed as Level 2 standard data products, possessing a clearly defined geographic coordinate reference system and quality assurance flags. Common examples include MODIS aerosol products (MOD04 / MYD04), VIIRS aerosol products (IP AOT), and Himawari aerosol products. Their spatial resolution is typically in the kilometer range (e.g., 3km to 10km), and their spectral bands are concentrated in the visible to near-infrared region (e.g., 550nm). They are used to provide large-scale, high-frequency global or regional aerosol distribution information and are key parameter products for satellite remote sensing monitoring of the atmospheric environment.

[0066] Step S210 above involves acquiring satellite aerosol optical thickness imagery products with the closest measurement time to that of a ground-based solar photometer. This aims to minimize asynchronous observation errors introduced by the temporal evolution of atmospheric conditions, ensuring the physical consistency of the satellite-to-ground comparison verification. Aerosol optical thickness exhibits dynamic characteristics, with its values ​​fluctuating non-stationarily on an hourly scale due to various factors such as diurnal variations in boundary layer thermal structure, local pollution source emission rhythms, regional pollution transport processes, and weather system evolution. If the time interval between satellite and ground observations is too long, the atmospheric extinction states represented by the two observations will not be the same physical realization, introducing temporal spurious errors that are difficult to quantitatively separate. Selecting imagery products with the closest time intervals allows the time difference between satellite and ground observations to be controlled within an acceptable range (usually less than 30 minutes), minimizing the impact of natural variations in aerosol optical thickness on the verification results. This satisfies the fundamental principle of "spatiotemporal matching" in authenticity verification, ensuring that the second uncertainty index only reflects spatial scale differences without being mixed with strong temporal variation components, thereby improving the scientific validity and credibility of the verification results.

[0067] Step S220: Construct a spatial variation model in the pixel-scale region of the site based on the image product.

[0068] Optionally, step S220 may include: extracting a pixel dataset of a pixel-scale region centered on a ground station from the satellite aerosol optical thickness image product; fitting the pixel dataset to obtain a spatial variation model.

[0069] Optionally, the above-mentioned fitting of the pixel dataset includes: using a spatial trend surface modeling method to perform function fitting on the pixel dataset to obtain a spatial variation model; wherein, the spatial trend surface modeling method includes at least one of first-order or second-order polynomial trend surface fitting, Kriging interpolation, or inverse distance weighted interpolation.

[0070] The aforementioned pixel-scale region refers to a specific spatial range extracted from satellite aerosol optical thickness imagery products, centered on the geographical location of the ground-based solar photometer station, and matching the spatial resolution of the satellite products. This region typically covers multiple consecutive pixels at the kilometer scale, used to construct a pixel dataset that reflects the spatial distribution characteristics of aerosol optical thickness. The spatial extent of this region needs to be adaptively determined based on the spatial resolution parameters of the satellite payload (such as 10km or 3km resolution for MODIS). Its technical purpose is to capture the spatial variation information of aerosol optical thickness in the area surrounding the station, providing geographically representative and statistically sufficient basic data support for the subsequent construction of spatial variation models.

[0071] The aforementioned spatial variation model refers to a continuous function representation constructed by modeling the spatial trend surface of aerosol optical thickness values ​​for each pixel within a pixel-scale region. This model, based on spatial interpolation algorithms such as first- or second-order polynomial trend surface fitting, Kriging interpolation, or inverse distance weighted interpolation, establishes a mathematical relationship between aerosol optical thickness values ​​and geographic coordinates. This is used to accurately characterize the spatial distribution pattern and gradient variation law of aerosol extinction characteristics in the region. The core function of the spatial variation model is its ability to calculate the theoretical aerosol optical thickness value at any precise location of a ground station and the area-weighted average value for the entire region. By quantifying the systematic deviation between the calculated values ​​at station locations and the regional average, it enables a quantitative assessment of the uncertainty introduced by the spatial scale mismatch between ground point measurements and satellite pixel surface observations.

[0072] The above-described implementation method for extracting a pixel dataset centered on a ground station from satellite aerosol optical thickness image products includes, for example: obtaining the precise geographic coordinates (longitude and latitude) of the ground solar photometer station, and using the geographic reference information embedded in the satellite image product (such as affine transformation parameters or latitude and longitude lookup tables) to convert the station's geographic coordinates into row and column indices in the image coordinate system; then, based on the spatial resolution parameters of the satellite payload and the preset regional coverage area (usually corresponding to a neighborhood window of multiple pixels), determining the range of all valid pixel row and column indices within a rectangular or circular spatial area centered on the pixel location of the station; finally, by reading the pixel values ​​of the corresponding data layer in the image product storage format (usually HDF, NetCDF, or GeoTIFF), and combining them with quality control flags to filter valid observed pixels, the extracted pixel geographic coordinates, aerosol optical thickness values, and quality control information are organized into a structured dataset, which constitutes the basic input for constructing a spatial variation model.

[0073] The above implementation method for fitting a pixel dataset to obtain a spatial variation model is as follows: Based on the extracted pixel dataset, where each data point contains pixel geographic coordinates (longitude and latitude) and its corresponding aerosol optical thickness value, an appropriate modeling algorithm is selected according to the spatial smoothness and gradient characteristics of the regional aerosol distribution: When the region exhibits global linear or quadratic gradient changes, a first-order or second-order polynomial trend surface is used for fitting, and the polynomial coefficients are estimated by the least squares method to construct a polynomial trend function; When the region has local heterogeneity and spatial autocorrelation, the Kriging interpolation method is selected, and the spatial autocorrelation is modeled based on the theory of variogram (such as the spherical model and the exponential model), and the Kriging estimator of the spatial stochastic process is obtained through unbiased optimal estimation; If the regional variation characteristics are unknown and fast calculation is required, inverse distance weighted interpolation is used, and a deterministic interpolation function is constructed by weighting the distance power function as the weight. After performing a function approximation on discrete pixel data using any of the above methods, a spatial variation model covering the entire pixel-scale region is generated. This model is a continuous computable function with respect to geographic coordinates, capable of outputting the theoretical value and regional average value of aerosol optical thickness at any location within the region, supporting subsequent quantitative calculations of spatial scale difference uncertainties.

[0074] The aforementioned first- or second-order polynomial trend surface fitting is a globally deterministic spatial interpolation method. It uses the least squares principle to fit a polynomial surface to the pixel dataset, establishing a low-order polynomial function relationship between aerosol optical thickness values ​​and geographic coordinates. The first-order trend surface employs a linear plane model (containing constant terms, linear longitude terms, and linear latitude terms), suitable for depicting aerosol optical thickness with a uniform spatial gradient distribution. The second-order trend surface adds quadratic longitude terms, quadratic latitude terms, and a longitude-latitude intersection term to form a quadratic surface model, reflecting the nonlinear curvature trend of aerosol optical thickness in space. Kriging interpolation, on the other hand, is a geostatistical spatial interpolation method based on regionalized variable theory and variogram analysis. It treats aerosol optical thickness as a spatially autocorrelated stochastic process, quantitatively describing the spatial dependence of aerosol optical thickness values ​​between pixels as the distance decreases by constructing a variogram (usually a spherical model, exponential model, or Gaussian model). This method utilizes known pixel data within the neighborhood of the point to be estimated to solve the Kriging equations and obtain the optimal unbiased linear estimator, minimizing the variance of the estimation error. It also outputs the estimation variance to quantify the interpolation uncertainty. The aforementioned inverse distance-weighted interpolation is a deterministic local interpolation method based on the first law of geography, namely the principle of spatial proximity. It uses the reciprocal of the power function of the Euclidean distance between the point to be estimated and known pixels as weights, and calculates the value of the point to be estimated by weighted averaging of the aerosol optical thickness values ​​of pixels in the neighborhood. The core of this method lies in the definition of the weight function, typically using a power exponent parameter to control the distance decay rate (the larger the exponent, the smaller the influence of distant pixels). The local neighborhood range participating in the interpolation can be determined by setting the search radius or the number of nearest neighbor pixels. It is understood that the specific implementation methods and working principles of the aforementioned first- or second-order polynomial trend surface fitting, Kriging interpolation, and inverse distance-weighted interpolation are detailed in related technologies; these will not be elaborated further in this embodiment.

[0075] Step S230: Quantify the uncertainty introduced by spatial scale differences based on the spatial change model to obtain the second uncertainty index.

[0076] Optionally, step S230 may include: calculating the aerosol optical thickness value at the ground station location based on the spatial variation model, and calculating the regional average value of the aerosol optical thickness values ​​of all pixels within the pixel-scale region; performing a deviation calculation between the calculated value at the station location and the regional average value, and using the deviation as a second uncertainty index.

[0077] The uncertainty introduced by the aforementioned spatial scale difference refers to the systematic verification error arising from the mismatch in spatial representativeness between single-point measurements from ground-based solar photometers and satellite remote sensing pixel area observations. Specifically, ground-based station observations essentially represent the aerosol extinction characteristics of a local atmospheric column at a scale of tens to hundreds of meters within the instrument's field of view, while satellite pixel inversion values ​​characterize the area-weighted average of aerosol optical thickness over a spatial region of several to tens of kilometers. The two differ by orders of magnitude in spatial scale. Due to the non-uniformity of aerosol concentration and composition in spatial distribution (e.g., urban pollution hotspots, biomass combustion plumes, dust transport belts), station measurements often fail to accurately represent the average state of the entire satellite pixel coverage area. This spatial representativeness bias constitutes the spatial scale difference uncertainty. Traditional verification methods directly equate station point measurements to the pixel area mean, ignoring the systematic errors caused by the spatial heterogeneity of aerosols. This results in verification results mixed with satellite inversion errors and spatial scale pseudo-errors, making it impossible to scientifically assess the intrinsic accuracy of the product. This invention constructs a spatial variation model to calculate the deviation between the theoretical value of the station location and the average value of the pixel area. This deviation is used as a second uncertainty index for quantitative characterization, thereby separating and eliminating spurious errors introduced by spatial scale mismatch. This supports the systematic correction of spatiotemporal representativeness deviations in authenticity verification and improves the scientificity and objectivity of the accuracy evaluation of satellite aerosol optical thickness products.

[0078] Step S230 above: The spatial representativeness difference between the station location point measurement and the pixel-scale regional average is solved using a spatial variation model. The theoretical value of aerosol optical thickness (AOP) at the location is calculated by substituting the precise geographic coordinates of the ground station into the constructed spatial variation trend function (representing the atmospheric extinction state corresponding to a single-point measurement by the ground-based instrument). Simultaneously, the average AOP of all locations within the pixel-scale region is calculated using integration or area-weighted averaging (representing the spatial mean observed from the satellite pixel surface). The difference between the calculated station location value and the regional average is calculated, and the resulting deviation quantitatively characterizes the systematic verification error caused by spatial scale mismatch. This deviation directly reflects the degree of insufficient station representativeness under aerosol spatial non-uniformity; the larger the absolute value, the stronger the uncertainty introduced by spatial scale differences. This allows for a quantifiable assessment of spatial representativeness deviation, providing a spatial uncertainty measure independent of time differences for verifying the authenticity of aerosol optical thickness satellite products.

[0079] Optionally, the aforementioned first and second uncertainty indices are used to correct spatiotemporal representativeness bias in the authenticity verification of aerosol optical thickness satellite remote sensing products. For example, in the authenticity verification of aerosol optical thickness satellite remote sensing products, the first and second uncertainty indices correct for spatiotemporal representativeness bias through quantitative separation and systematic subtraction mechanisms. Traditional satellite-to-ground comparison directly calculates the difference between ground-based single-point measurements and the average value of satellite pixels to determine the verification bias. This bias is a mixture of intrinsic errors of the satellite inversion algorithm, temporal asynchrony spurious errors, and spatial scale mismatch spurious errors. By introducing a first uncertainty index (standard deviation of temporal difference) and a second uncertainty index (spatial scale bias), the two are first orthogonally synthesized with the uncertainty of the satellite product's own algorithm to obtain the total verification uncertainty considering spatiotemporal representativeness bias, thus avoiding underestimation of verification risk. Secondly, when calculating the true inversion bias, the second uncertainty index is directly subtracted from the original satellite-to-ground difference to eliminate spatial systematic bias, and the first uncertainty index is used to reduce the weight or remove samples with drastic temporal variations, ensuring that the verification statistic only reflects the physical inversion performance of the satellite product. Furthermore, a quality marker system is constructed based on the spatiotemporal uncertainty index to mark the spatiotemporal representativeness level of each pair of satellite-to-ground matching samples, supporting subsequent calculation of the weighted root mean square error and confidence interval assessment based on uncertainty. Ultimately, this achieves a shift from qualitative comparison to quantitative error attribution, improving the scientific rigor and objectivity of the verification results for aerosol optical thickness products.

[0080] Please see Figure 3 To facilitate understanding of the working principle of the above-described AOD foundation measurement reference truth evaluation method, this embodiment of the invention also provides a specific application case of this method in a certain application scenario. In this application scenario, the above-described AOD foundation measurement reference truth evaluation method mainly includes:

[0081] S1: Obtain time-series measurement data of ground AOD measurement stations on a certain date, divide the time period and establish a model of the change of ground measurement AOD over time, and use the established model to evaluate the uncertainty of AOD product inspection caused by the difference between satellite and ground measurement time.

[0082] S1.1: Acquire ground-based AOD time-series measurement data. Obtain AOD time-series data for the entire day of a given date from a solar photometer (such as CE318), with a time resolution of not less than 15 minutes.

[0083] S1.2: Define key time periods. Based on the satellite transit time (e.g., 10:30 AM for MODIS), define time periods within ±1 hour or other appropriate ranges, and extract ground observation AOD data within the corresponding time periods.

[0084] S1.3: Establish a time variation model of ground AOD, and use multinomial fitting, moving average, spline curve or other regression methods to model the extracted ground AOD time series data to form a time variation trend function of AOD;

[0085] S1.4: Estimate the uncertainty caused by the time difference between satellite and ground observation. Using the established time variation model, calculate the minute-by-minute AOD value during the time period from the satellite's transit time to the ground measurement time, and statistically analyze the standard deviation of the AOD value during this time period as an indicator of the uncertainty of AOD introduced by the time difference.

[0086] S2: Acquire satellite AOD images of the AOD site and its surrounding area, extract AOD images of a 10km*10km area centered on the measurement site, and establish an AOD spatial variation model using a trend surface model. Use this model to evaluate the uncertainty of AOD product inspection caused by the spatial scale difference between ground point measurement and satellite pixel-scale measurement.

[0087] S2.1: Acquire satellite AOD imagery data and obtain satellite AOD product data (such as MODIS, VIIRS, Himawari, etc.) that are as close as possible to the ground measurement time to ensure that the product quality level meets the usage requirements;

[0088] S2.2: Extract imagery of the area surrounding the station. Centered on the ground station, extract the satellite AOD data area within a 10km×10km range. The size of the extracted area can be adjusted according to actual needs.

[0089] S2.3: Construct an AOD spatial variation trend surface model to perform spatial modeling of satellite AOD pixels in the extracted area. Common methods include first-order or second-order trend surfaces, Kriging interpolation, etc., to establish a spatial variation model.

[0090] S2.4: Assess the uncertainty caused by spatial scale differences. Use a spatial model to calculate the AOD value of the measurement point location and compare it with the average value of the region (e.g., the average AOD of a 10km×10km region). The deviation is the test uncertainty index caused by spatial scale differences.

[0091] Furthermore, to facilitate understanding of the calculation principles of the first and second uncertainty indices in the above-described AOD foundation measurement reference truth value evaluation method, this embodiment of the invention also provides a specific application example of this method in another application scenario. In this application scenario, the above-described AOD foundation measurement reference truth value evaluation method mainly includes:

[0092] Step 1: Assessment of Uncertainty Caused by Time Difference

[0093] (1) Obtain ground AOD time series measurement data. Taking the AERONET Baotou-AOE station as an example, obtain AOD time series data for a certain day in 2023 with a time resolution of 15 minutes and a band of 550nm.

[0094] (2) Divide the key time period. The MODIS Aqua transit time is approximately 13:30 (UTC+8) at the AERONET Baotou-AOE station, i.e. 05:30 UTC. Use this as the center to define a ±1 hour time period, i.e. 04:30–06:30 UTC.

[0095] (3) Establish a ground-based AOD time variation model, extract data within this time period, and use cubic spline curve fitting to obtain the following AOD time variation function:

[0096]

[0097] The fitted curve is as follows Figure 4 As shown.

[0098] (4) Estimate the uncertainty caused by the satellite-ground time difference. Assume the most recent ground observation is 05:15 UTC and the satellite transit is 05:30 UTC. Calculate the AOD value between 05:15 and 05:30 based on the fitted curve and calculate its standard deviation:

[0099]

[0100] The uncertainty in AOD product verification caused by time difference is 0.0013.

[0101] Step 2: Assessment of Uncertainty Caused by Spatial Scale Differences

[0102] (1) Acquire satellite AOD image data, select MODIS Level 2 AOD product (MOD04_L2), and keep the time as close as possible to the ground observation; extract a 10km×10km area centered on the station to obtain AOD data of 9×9 pixels in the area;

[0103] (2) A second-order trend surface model was used for fitting to construct the spatial change trend surface model of AOD, as follows:

[0104]

[0105] The fitted curve is as follows Figure 5 As shown.

[0106] (3) Calculate the AOD value of the trend surface model at the ground station location, and then compare it with the average value of the entire region to assess the uncertainty caused by spatial scale differences, as follows:

[0107]

[0108] That is, the uncertainty of AOD verification caused by spatial scale differences is 0.002.

[0109] It is understandable that the above-mentioned AOD foundation measurement reference true value evaluation method has the following beneficial effects:

[0110] (1) Systematically evaluate the spatiotemporal representativeness of ground-based AOD reference true values ​​to improve the scientificity and reliability of AOD product verification: For the first time, “time difference” and “spatial scale difference” are introduced into the uncertainty analysis of the authenticity verification of aerosol optical thickness products. The sources of deviation between ground point measurement and satellite inversion results are systematically quantified, which can provide more physically meaningful reference true values ​​for the evaluation of satellite AOD products and improve the objectivity and scientificity of product verification.

[0111] (2) Introducing time-varying modeling to innovatively solve the error assessment problem caused by the inconsistency between satellite and ground observation times: By constructing a trend model of ground AOD changing with time (such as polynomial, moving average, or spline curve), the AOD value at the satellite transit time is estimated at minute-level time resolution, and its time uncertainty is measured by standard deviation. This method effectively solves the shortcomings of traditional "most recent observation method" or "fixed time averaging method" in terms of insufficient time representativeness.

[0112] (3) Establish a spatial trend surface model by combining satellite products to accurately assess the uncertainty caused by differences in spatial scale: Use regional data extracted from satellite images to construct a spatial variation trend model of AOD and quantify the error caused by the spatial inconsistency between ground points and satellite pixels. Compared with the traditional coarse method that uses single pixels or average values ​​as substitutes, this method is more refined and can fully reveal the spatial variation characteristics of local aerosol distribution.

[0113] (4) It takes into account the universality of different satellite AOD products, the method is highly adaptable and has good promotion value: it does not depend on specific satellite or ground instrument platforms, and can be widely applied to various satellite AOD products such as MODIS, VIIRS, Himawari and other satellites as well as ground measurement data of various solar photometers such as CE318, and has good platform independence and universality.

[0114] (5) It can serve as an important component of the AOD product error quantification framework and promote the standardization of remote sensing data quality: It provides a unified and quantitative spatiotemporal uncertainty assessment framework for AOD authenticity verification, and can serve as an important means of assessing the representativeness of the reference true value in the remote sensing data verification process, which helps to promote the standardization process of remote sensing product error modeling and quality control.

[0115] Based on the same inventive concept, embodiments of the present invention also provide a method for verifying the authenticity of AOD products, comprising: acquiring a reference true value of aerosol optical thickness measured by a ground-based solar photometer and satellite aerosol optical thickness product data to be verified; determining a first uncertainty index and a second uncertainty index of the aerosol optical thickness reference true value using the AOD ground-based measurement reference true value evaluation method provided in this application embodiment; performing spatiotemporal representativeness deviation correction on the comparison verification result between the satellite aerosol optical thickness product data and the reference true value based on the first uncertainty index and the second uncertainty index, obtaining a corrected authenticity verification result; and outputting the corrected authenticity verification result.

[0116] The aforementioned technical implementation of correcting the spatiotemporal representativeness bias of satellite aerosol optical thickness product data compared with the reference true value based on the first and second uncertainty indices lies in separating and eliminating pseudo-errors through error decomposition and systematic subtraction mechanisms. Specifically: First, the original deviation between the satellite pixel aerosol optical thickness product data and the ground-based solar photometer reference true value is calculated. This deviation is the convolution result of time difference, spatial scale difference, and intrinsic error of the satellite inversion algorithm. Then, the second uncertainty index (i.e., the deviation between the calculated value at the station location and the average value of the pixel area) is used to directly subtract the systematic deviation component introduced by the spatial scale mismatch from the original deviation, so that the corrected deviation only includes... The test incorporates contributions from time variation and algorithm error. Simultaneously, a first uncertainty index (i.e., the standard deviation of calculated values ​​within the time window) is introduced as a measure of the intensity of time variation. Samples with a time difference standard deviation exceeding a preset threshold are either downweighted or marked as low-quality samples and removed to suppress the impact of time-asynchronous pseudo-errors on the test statistics. Furthermore, the corrected deviation is orthogonally synthesized with the first and second uncertainty indices to obtain the total verification uncertainty considering spatiotemporal representativeness deviations. Based on this, a corrected authenticity test result is constructed, including but not limited to the corrected root mean square error, the mean deviation, and confidence intervals, ensuring that the final output test result scientifically represents the intrinsic accuracy of the satellite product rather than mixed spatiotemporal pseudo-errors.

[0117] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for evaluating the true reference value of AOD foundation measurement, characterized in that, The method includes: Obtain a ground reference true value and a first uncertainty index and a second uncertainty index of the ground reference true value; wherein, the first uncertainty index is used to indicate the uncertainty of AOD test due to time difference; and the second uncertainty index is used to indicate the uncertainty of AOD test due to spatial scale difference. The method for determining the first uncertainty index includes: Collect high-frequency time-series data from the ground-based solar photometer on the day the satellite passes overhead; A time-varying model is constructed based on the aforementioned time-series data; The uncertainty introduced by time difference is quantified based on the time change model, and the first uncertainty index is obtained; The method for determining the second uncertainty index includes: Collect satellite aerosol optical thickness image products that are closest to the measurement time of the ground-based solar photometer; A spatial variation model is constructed in the pixel-scale region of the site based on the image product; The uncertainty introduced by spatial scale differences is quantified based on the spatial change model, and the second uncertainty index is obtained.

2. The method for evaluating the true reference value of AOD foundation measurement according to claim 1, characterized in that, The construction of the time-varying model based on the time-series data includes: A subset of observations within a continuous time window is extracted from the high-frequency time-series data, centered on the satellite transit time. The observation subset is optimized and fitted to obtain the time variation trend function of aerosol optical thickness.

3. The method for evaluating the true reference value of AOD foundation measurement according to claim 2, characterized in that, The optimization fitting aims to minimize the error between the calculated value of the trend function and the measured data for the corresponding time period.

4. The method for evaluating the true reference value of AOD foundation measurement according to claim 2, characterized in that, The optimization fitting of the observed subset includes: The observation subset is fitted with a time series regression method to obtain the trend function; wherein the time series regression method includes at least one of multinomial regression, spline regression, moving average filtering, or autoregressive model.

5. The method for evaluating the true reference value of AOD foundation measurement according to claim 1, characterized in that, The process of quantifying the uncertainty introduced by time differences based on the time change model and obtaining the first uncertainty index includes: The aerosol optical thickness value is calculated at each moment within the satellite transit time window based on the time variation model; wherein, the time window is centered on the satellite observation time and covers the time difference between the ground observation time and the satellite observation time. The calculated values ​​of aerosol optical thickness at all times within the time window are statistically analyzed, and the standard deviation of this set of calculated values ​​is used as the first uncertainty index.

6. The method for evaluating the true reference value of AOD foundation measurement according to claim 1, characterized in that, The construction of a spatial variation model based on the image product at the site pixel scale region includes: Extract pixel datasets of pixel-scale regions centered on ground stations from the satellite aerosol optical thickness image products; The spatial variation model is obtained by fitting the pixel dataset.

7. The method for evaluating the true reference value of AOD foundation measurement according to claim 6, characterized in that, The fitting of the pixel dataset includes: The spatial trend surface modeling method is used to fit a function to the pixel dataset to obtain the spatial variation model; wherein, the spatial trend surface modeling method includes at least one of first-order or second-order polynomial trend surface fitting, Kriging interpolation or inverse distance weighted interpolation.

8. The method for evaluating the true reference value of AOD foundation measurement according to claim 1, characterized in that, The process of quantifying the uncertainty introduced by spatial scale differences based on the spatial change model to obtain the second uncertainty index includes: The aerosol optical thickness value at the ground station location is calculated based on the spatial variation model, and the regional average value of the aerosol optical thickness value of all pixels within the pixel-scale region is calculated. The calculated value of the station location is compared with the average value of the region, and the deviation is used as the second uncertainty index.

9. The method for evaluating the true reference value of AOD foundation measurement according to any one of claims 1 to 8, characterized in that, The first uncertainty index and the second uncertainty index are used to correct the spatiotemporal representativeness deviation in the authenticity verification of aerosol optical thickness satellite remote sensing products.

10. A method for verifying the authenticity of AOD products, characterized in that, The method includes: Obtain the true reference value of aerosol optical thickness measured by a ground-based solar photometer and the satellite aerosol optical thickness product data to be tested; The first uncertainty index and the second uncertainty index of the aerosol optical thickness reference true value are determined by the method of any one of claims 1 to 9. Based on the first uncertainty index and the second uncertainty index, the spatiotemporal representativeness deviation correction is applied to the comparison test results between the satellite aerosol optical thickness product data and the reference true value to obtain the corrected authenticity test results. Output the corrected authenticity test results.

Images