Open AccessArticle

Synchronous Atmospheric Correction of Wide-Swath and Wide-Field Remote Sensing Image from HJ-2A/B Satellite

Honglian Huang

¹,

Yuxuan Wang

^1,2,

Xiao Liu

^1,*,

Rufang Ti

Xiaobing Sun

¹,

Zhenhai Liu

Xuefeng Lei

Jun Lin

³ and

Lanlan Fan

Key Laboratory of Optical Calibration and Characterization, Hefei Institutes of Physical Science, Chinese Academy of Sciences, Hefei 230031, China

Science Island Branch of Graduate School, University of Science and Technology of China, Hefei 230026, China

China Centre for Resources Satellite Data and Application, Beijing 100094, China

Author to whom correspondence should be addressed.

Remote Sens. 2025, 17(5), 884; https://doi.org/10.3390/rs17050884

Submission received: 26 January 2025 / Revised: 25 February 2025 / Accepted: 26 February 2025 / Published: 1 March 2025

Download

Browse Figures

Versions Notes

Abstract

The Chinese HuanjingJianzai-2 (HJ-2) A/B satellites are equipped with advanced sensors, including a Multispectral Camera (MSC) and a Polarized Scanning Atmospheric Corrector (PSAC). To address the challenges of atmospheric correction (AC) for the MSC’s wide-swath, wide-field images, this study proposes a pixel-by-pixel method incorporating Bidirectional Reflectance Distribution Function (BRDF) effects. The approach uses synchronous atmospheric parameters from the PSAC, an atmospheric correction lookup table, and a semi-empirical BRDF model to produce surface reflectance (SR) products through radiative, adjacency effect, and BRDF corrections. The corrected images showed significant improvements in clarity and contrast compared to pre-correction images, with minimum increases of 55.91% and 35.63%, respectively. Validation experiments in Dunhuang and Hefei, China, demonstrated high consistency between the corrected SR and ground-truth data, with maximum deviations below 0.03. For surface types not covered by ground measurements, comparisons with Sentinel-2 SR products yielded maximum deviations below 0.04. These results highlight the effectiveness of the proposed method in improving image quality and accuracy, providing reliable data support for applications such as disaster monitoring, water resource management, and crop monitoring.

Keywords:

atmospheric correction; HuanjingJianzai-2 (HJ-2) satellites; polarized scanning atmospheric corrector (PSAC); multispectral camera; wide-swath and wide-field image

1. Introduction

During remote sensing imaging, the radiative signals received by satellite sensors represent a coupled interaction between surface and atmospheric information. The primary objective of land observation sensors is to accurately capture surface information, while atmospheric contributions often act as interference [1,2]. Atmospheric scattering and absorption induce blur and distort the true surface reflectance, significantly affecting satellite image quality. Therefore, atmospheric correction is essential to retrieve accurate surface reflectance from satellite images, enhance image clarity, and improve overall image quality.

Common atmospheric correction methods include image feature-based relative correction and radiative transfer model-based correction. The relative correction method analyzes image information or compares it with historical data, allowing for correction in the absence of ground spectral data or synchronous atmospheric measurements. However, this approach typically has limited accuracy. Currently, most optical satellites employ radiative transfer model-based methods to generate surface reflectance products [3,4,5].

Radiative transfer model-based atmospheric correction methods require synchronously acquired atmospheric parameters as inputs [6,7,8]. The Earth’s atmospheric composition varies spatially and temporally, particularly for parameters such as water vapor and aerosols. To achieve high correction accuracy, fixed atmospheric models or distribution parameters are insufficient. Instead, synchronous atmospheric parameters for the target region must be acquired [9,10,11].

Radiative transfer model-based methods can be broadly categorized based on how atmospheric parameters are acquired. One such approach employs spectral bands specifically designed for atmospheric parameter retrieval, as demonstrated in the MODIS and Sentinel series. These bands typically have lower spatial resolutions and narrower bandwidths; they retrieve atmospheric parameters that are then used to correct higher-resolution bands [12,13,14]. The other approach involves equipping satellites with instruments for real-time atmospheric parameter acquisition. For instance, the Earth Observing-1 (EO-1) satellite features the Linear Etalon Imaging Spectrometer Array Atmospheric Corrector (LAC), while the WorldView-3 satellite includes the Clouds, Aerosols, Vapor, Ice, and Snow (CAVIS) instrument for cloud and aerosol detection. Similarly, China’s Gao Fen Duo Mo satellite is equipped with the Synchronous Monitoring Atmospheric Corrector (SMAC) [15]. The HuanjingJianzai-2 (HJ-2) A/B satellites incorporate a Polarized Scanning Atmospheric Corrector (PSAC) [16]. Among these instruments, the SMAC and PSAC incorporate polarization observations, which can effectively enhance aerosol retrieval accuracy, particularly over highly reflective surfaces, thereby improving atmospheric correction precision. Therefore, this approach has become a key trend in the development of high-spatial-resolution atmospheric correction in recent years.

The HuanjingJianzai satellite series focuses on disaster prevention, mitigation, and environmental protection while also addressing needs in land resources, agriculture, forestry, and seismology. Beyond acquiring clear imagery, these satellites aim to produce high-precision quantitative products such as surface albedo, vegetation indices, water body indices, snow cover indices, soil moisture, land cover change, and water quality indicators. These products are vital for disaster assessment, recovery, and resource monitoring, with their quality directly dependent on the accuracy of surface reflectance derived from atmospheric correction.

The HJ-2A/B satellites were launched on 27 September 2020, equipped with a Multispectral Camera (MSC), a hyperspectral imager (HSI), an infrared spectroradiometer (IRS), and a Polarized Scanning Atmospheric Corrector (PSAC) [16,17,18]. The PSAC provides real-time atmospheric parameters to support atmospheric correction for the MSC. The MSC has a swath width of 800 km and a total field of view of 62.6°. Across the entire swath, atmospheric parameters such as aerosol optical depth and water vapor content vary with geographic location. Additionally, the solar zenith angle, solar azimuth angle, camera viewing zenith angle, and viewing azimuth angle also change with geographic location. Furthermore, due to the directional nature of surface reflectance, the BRDF effect cannot be neglected for wide-field cameras, as surface reflectance varies with observation angles. Therefore, atmospheric correction for wide-swath, wide-field remote sensing images requires pixel-by-pixel adjustments that account for variations in the viewing zenith angle, viewing azimuth angle, solar zenith angle, solar azimuth angle, aerosol optical depth, water vapor content, and the influence of BRDF effects.

This paper presents a pixel-by-pixel synchronous atmospheric correction method for wide-swath, wide-field remote sensing images, incorporating BRDF effects. The method is applied to correct MSC images over various surface types. Validation experiments assess the improvements in image quality and reflectance retrieval accuracy achieved through this approach. The structure of this paper is as follows: Section 2 introduces the HJ-2A/B satellites, MSC, and PSAC instrument. Section 3 details the per-pixel synchronous atmospheric correction method and its implementation process. Section 4 presents the atmospheric correction results and validation experiments. Finally, Section 5 concludes the study.

2. HJ-2 A/B Satellites and Sensors

China’s HuanjingJianzai-2 (HJ-2) series satellites are sun-synchronous orbit satellites with an orbital altitude of 645 km. They are designed to replace the HJ-1A/B satellites, which have exceeded their operational lifespan, and are primarily used for disaster prevention and mitigation as well as environmental protection. The HJ-2A/B satellites are the first two in the HJ-2 series, equipped with the same four payloads, and have a descending node local time of approximately 10:30 a.m. The HJ-2A and HJ-2B satellites are arranged in a 180° phase configuration to achieve higher temporal resolution imagery. Their MSC and IRS can cover the entire Chinese region within two days, providing efficient data support for applications such as natural disaster monitoring, land use macro-monitoring, water resource management and protection, crop monitoring and yield estimation, and earthquake emergency response and disaster relief [19,20].

To support these diverse applications, the Chinese HuanjingJianzai-2 (HJ-2) A/B satellites are equipped with a Multispectral Camera (MSC), a hyperspectral imager (HSI), an infrared spectroradiometer (IRS), and a Polarized Scanning Atmospheric Corrector (PSAC). Table 1 provides an overview of the basic parameters of these sensors. For clarity in this paper, bands 4 (B4) and 5 (B5) of the MSC have been swapped to present the wavelengths in ascending order, facilitating easier analysis and comparison.

The PSAC is designed with nine spectral bands covering the visible to short-wave infrared range, each with four polarization detection channels. It is also equipped with a comprehensive on-board radiometric and polarization calibration system, achieving a polarization detection accuracy of 0.005 and a radiometric calibration accuracy of 7%. These capabilities are used to obtain synchronized results for cloud detection, atmospheric aerosol, and water vapor content [21,22]. The MSC consists of four CCDs, arranged in a stitched configuration, with a total swath width of 800 km and a total field of view angle of 62.6°. Each CCD has five bands, and compared to the HJ-1A/B satellites, an additional red-edge band has been added for monitoring the growth of green vegetation.

The HJ-2 A/B satellites are China’s second civilian remote sensing satellites equipped with a synchronous atmospheric correction instrument, following the Gao Fen Duo Mo satellite. The atmospheric correction instruments on both satellites are polarization sensors, but they operate in different modes. The SMAC instrument on the Gaofen Multi-Mode Satellite employs a split-aperture detection method, with each channel using a dual-pixel imaging mode to perform push-broom observations along the satellite’s flight direction.

The Polarized Scanning Atmospheric Corrector (PSAC) on the HJ-2 A/B satellites employs a combined split-aperture and split-amplitude simultaneous polarization measurement method to enable the simultaneous detection of all polarization channels. To match the MSC’s 800 km swath width, the PSAC performs cross-track scanning at a spatial resolution of 6 km, ensuring compatibility for synchronized observations. Figure 1 illustrates the coordinated detection capability of the PSAC and MSC. In the figure, the green arrow represents the satellite’s flight direction, while the blue arrow indicates the along-track push-broom direction of the MSC, which aligns with the satellite’s movement. The red arrow denotes the cross-track scanning direction of the PSAC, which is perpendicular to the satellite’s flight direction. This demonstrates that the PSAC and MSC can seamlessly perform synchronized observations.

3. Principle and Processing Flow of Atmospheric Correction Algorithm

Assuming that the target surface observed by the satellite is a uniform Lambertian surface, the observed Top-of-Atmosphere (TOA) radiance can be expressed using Equation (1):

L (θ_{s}, θ_{v}, φ) = T_{g} (θ_{s}, θ_{v}) \cdot [L_{p a t h} (θ_{s}, θ_{v}, φ) + \frac{ρ_{1} F_{s} μ_{s}}{π (1 - ρ_{1} S)} T (θ_{s}) T (θ_{v})]

(1)

where θ_s is the solar zenith angle (SZA), θ_v is the viewing zenith angle (VZA), φ is the relative azimuth angle (RAA), and μ_s is the cosine of the SZA. F_s is the extraterrestrial solar irradiance, representing the solar irradiance incident vertically outside the atmosphere, and it varies with wavelength. L_path is the atmospheric path radiance, T_g is the direct transmittance, and T(θ_v) and T(θ_s) represent the upward and downward diffuse transmittance, respectively. L_path, T_g, T(θ_v), and T(θ_s) are all atmospheric radiances, which can be calculated using an atmospheric radiative transfer model.

Based on this theoretical foundation, the atmospheric correction process integrates multiple computational steps to accurately retrieve surface reflectance. Figure 2 illustrates the atmospheric correction workflow proposed in this study. The subsequent sections will provide a detailed explanation of the processing steps outlined in the flowchart, including data matching, aerosol and water vapor retrieval, the construction of atmospheric correction lookup tables, atmospheric radiative correction, adjacency effect correction, and BRDF correction.

3.1. PSAC Data Preprocessing

The PSAC is equipped with nine polarization bands, each observed in four polarization directions: 0°, 45°, 90°, and 135°. The raw data collected in these orientations are transmitted to the ground as Level-0 strip data. A series of preprocessing operations, including data extraction and validation, polarization radiometric calibration, and geometric positioning, is performed to prepare the data for subsequent analysis.

The preprocessing yields normalized Stokes parameters (I, Q, and U) along with associated metadata, such as viewing geometry, solar geometry, latitude, longitude, altitude, and other auxiliary information. These processed data are stored in HDF format with a swath width of 800 km, constituting the Level-1 (L1) product of the PSAC. These products provide the foundational input for aerosol and water vapor retrieval.

The normalized Stokes parameters are derived using the following equation:

[\begin{matrix} I \\ Q \\ U \end{matrix}] = \frac{π d^{2}}{F_{s}} [\begin{matrix} I_{0} + I_{90} \\ I_{0} - I_{90} \\ I_{45} - I_{135} \end{matrix}]

(2)

where I₀, I₄₅, I₉₀, and I₁₃₅ represent the intensity values observed in the four polarization directions, while d is the astronomical distance between the Earth and the sun at the time of observation, which is used to correct the extraterrestrial irradiance.

3.2. Atmospheric Parameter Retrieval

The basic principle of atmospheric parameter retrieval is based on Equation (1), where various methods are employed to subtract surface reflectance information to derive the atmospheric aerosol optical depth (AOD) and column water vapor (CWV). Since the presence of clouds significantly affects sensor measurements, cloud detection is a critical preprocessing step to ensure accurate parameter retrieval.

Cloud detection is performed using the PSAC L1 product, applying a combination of techniques such as single-band reflectance thresholds, multi-band ratios, the normalized dust index, the normalized snow cover index, and the 1.38 µm cirrus band. These methods provide robust identification of cloud pixels under varying atmospheric conditions.

For non-cloud pixels, the normalized I, Q, and U values of four bands at 0.443 μm, 0.555 μm, 0.670 μm, and 0.865 μm are used to calculate the reflectance and polarized reflectance at the sensor’s aperture. Then, by constructing a surface reflectance ratio database and subtracting the surface information, AOD retrieval is achieved. The CWV is retrieved using the normalized radiance values I from the water vapor absorption band at 0.91 μm and the adjacent non-water vapor absorption band at 0.865 μm. This process enables the simultaneous retrieval of atmospheric parameters, resulting in the Level-2 (L2) product from the atmospheric correction instrument. For detailed algorithms, refer to [16].

The PSAC L2 product is stored in HDF files on a scene-by-scene basis. It includes not only atmospheric parameters, such as cloud detection results, aerosol optical depth, and column water vapor, but also auxiliary information, including latitude and longitude, land–ocean classification, and altitude. These data products provide valuable input for atmospheric correction.

3.3. Data Matching

To perform atmospheric correction on the images captured by the Multispectral Camera (MSC) using the AOD and CWV simultaneously retrieved by the PSAC, it is necessary to first match the L2 data from the PSAC with the L1 data from the MSC. The PSAC L2 product includes cloud detection, AOD, and CWV, all accompanied by geolocation (latitude and longitude) and altitude information. The MSC L1 product consists of radiometrically calibrated data, including scientific data (TIFF images) and corresponding calibration coefficient files, RPB parameter files, and auxiliary metadata in XML format. Both the MSC L1 product and the PSAC L2 product are stored scene by scene. The four CCDs of the MSC are not stitched together, with each scene having a swath width of 200 km. In contrast, the PSAC L2 product has a swath width of 800 km per scene.

Scene-by-scene matching is performed by extracting the imaging date and time from the XML metadata of the MSC L1 product and aligning it with the corresponding PSAC L2 product acquired at the same time. Since the MSC L1 data are not geometrically corrected, the geolocation of each pixel is derived by decoding the rational polynomial coefficients (RPCs). Using this pixel-level geolocation information, the atmospheric parameters retrieved by the PSAC are mapped onto the MSC image on a pixel-by-pixel basis.

Given the resolution disparity between the two datasets, where the PSAC has a spatial resolution of 6 km while the MSC operates at 16 m, each PSAC pixel corresponds to a 375 × 375 pixel block in the MSC image. This results in block-like structures in the mapped AOD and CWV, as shown in Figure 3a,c, where both axes represent pixel indices of the MSC image.

While this mapping process effectively integrates PSAC atmospheric parameters into MSC images, it introduces block artifacts in the atmospheric correction results, especially in regions with abrupt changes in the AOD and CWV. To address this issue, quality checks are performed on the PSAC retrievals to exclude invalid data. Spatial smoothing is then applied using a linear interpolation method, which ensures a gradual transition between adjacent blocks. This approach effectively resolves block artifacts and enhances the visual and quantitative quality of the corrected MSC images.

3.4. Atmospheric Correction LUT

For wide-swath, wide-field multispectral images captured by the MSC, it is not feasible to use a single set of aerosol optical depth (AOD), column water vapor (CWV), viewing geometry, and solar geometry data for atmospheric correction. Instead, atmospheric radiance must be calculated regionally. To improve computational efficiency and avoid repeated calls to the radiative transfer model, an atmospheric correction lookup table (LUT) is pre-constructed. The LUT provides precomputed atmospheric parameters, streamlining the correction process.

Based on the orbit and overhead pass times of the HJ-2A/B satellites, the ranges for the viewing solar zenith angle (SZA), zenith angle (VZA) and relative azimuth angle (RAA) are set. Additionally, the ranges and step sizes for CWV and the AOD are defined, along with the aerosol model provided by the 6SV atmospheric radiative transfer model. These parameter settings are detailed in Table 2, where the six aerosol models are as follows: continental model, maritime model, urban model, desertic model, biomass burning model, and stratospheric models.

For each CCD and spectral band of the MSC, the atmospheric correction LUT is constructed using the 6SV model. The spectral response functions of each band are incorporated into the simulations to compute the key atmospheric parameters: path radiance (L_path), direct transmittance (T_dir), diffuse transmittance (T_diff), and atmospheric hemispherical albedo (S). These parameters form the basis of the LUT.

Atmospheric correction is closely related to the aerosol model, with each model representing an independent LUT. To ensure efficient organization, each LUT is named based on its corresponding spectral band, CCD, and aerosol model. By leveraging precomputed LUTs, atmospheric correction calculations for the MSC images are significantly accelerated, enabling the timely processing of wide-swath satellite data.

3.5. Atmospheric Radiative Correction

To perform atmospheric correction on the MSC L1 product, the first step is to calculate the Top-of-Atmosphere radiance for each band of the multispectral image using the radiometric calibration coefficients. The calculation is expressed as follows:

L = K \cdot D N + B

(3)

where DN is the pixel gray value for each band of the multispectral image, K is the slope of the radiometric calibration coefficient, B the intercept of the radiometric calibration coefficient, and L represents the TOA radiance.

Based on the cloud detection results, cloudy pixels are treated differently from cloud-free pixels. For cloudy pixels, the TOA reflectance is directly calculated as the atmospheric correction result using the following formula:

ρ_{A O D} = \frac{π L}{{\cos θ_{s} F}_{s}}

(4)

For cloud-free pixels, the influence of atmospheric scattering and absorption is removed to retrieve the surface composite reflectance ρ₁, which incorporates the adjacency effect. Equation (1) can be rewritten as follows:

ρ_{1} = \frac{L (θ_{s}, θ_{v}, φ) / T_{g} (θ_{s}, θ_{v}) - L_{p a t h} (θ_{s}, θ_{v}, φ)}{T (θ_{s}) T (θ_{v}) F_{s} μ_{s} / π + ({L (θ_{s}, θ_{v}, φ) - L}_{p a t h} (θ_{s}, θ_{v}, φ)) S}

(5)

In the above equation, the parameters T_g, T(θ_v), T(θ_s), L_path, and S can all be obtained by referencing a pre-constructed atmospheric correction lookup table.

3.6. Adjacency Effect Correction

Equation (5) assumes a homogeneous surface, but real-world surfaces are often complex and heterogeneous. This necessitates the consideration of the influence of adjacent pixels, which can be modeled as the convolution of the target surface’s radiative field with the atmospheric point spread function (PSF). Based on the spatial resolution of MSC, the weighting function of the atmospheric point spread function is calculated to determine the size of the adjacency effect window.

ρ_{M} = \sum_{x, y = 1}^{N} ρ_{1} (x, y) P (x, y)

(6)

Here,

ρ_{M}

represents the background reflectance around the target pixel, N is the size of the adjacency effect window, and P(x,y) is the weighting function describing the contribution of a point at a distance from the target pixel [8]. For the HJ-2A/B satellite’s MSC, considering its spatial resolution and computational efficiency, N = 10 is selected.

Based on the calculated surface composite reflectance (

ρ_{1}

) and background reflectance (

ρ_{M}

), the true surface reflectance (

ρ_{t}

) can be computed as:

ρ_{t} = ρ_{1} + (ρ_{1} - ρ_{M})

(7)

where q is the adjacency effect correction coefficient, derived from the ratio of diffuse to direct atmospheric radiation. This correction ensures that the influence of scattered light from adjacent areas is accounted for, improving the accuracy of surface reflectance retrieval in heterogeneous regions.

3.7. BRDF Correction

For the MSC, which has a field of view angle of 62.6°, the effects of the Bidirectional Reflectance Distribution Function (BRDF) cannot be ignored. Wide-field satellite sensors provide limited observational angular information, making it challenging to fit BRDF model parameters directly from multi-angular datasets. To address this, MODIS BRDF products are used as prior knowledge to fit the parameters of a semi-empirical kernel-driven BRDF model.

The kernel-driven model is widely used to describe BRDF characteristics of surface features, offering simplicity, computational efficiency, and physical interpretability. It approximates the surface BRDF as a linear combination of three kernels: the isotropic kernel for Lambertian scattering, the volumetric scattering kernel for turbid medium scattering, and the geometric optics kernel for surface scattering. The model is expressed as:

R (θ_{s}, θ_{v}, φ, λ) = f_{i s o} (λ) + f_{v o l} (λ) K_{v o l} (θ_{s}, θ_{v}, φ) + f_{g e o} (λ) K_{g e o} (θ_{s}, θ_{v}, φ)

(8)

where f_iso represents the Lambertian reflectance coefficient; K_vol and f_vol denote the volumetric scattering kernel function and its corresponding kernel coefficient, respectively; and K_geo and f_geo represent the geometric optics kernel function and its associated kernel coefficient, respectively.

This study employs the RossThick–LiSparseR (RTSLR) model, which consists of the RossThick volumetric scattering kernel combined with the LiSparseR geometric optical kernel. This model is widely used and is also the basis for the operational MODIS BRDF and albedo products [23].

To account for the seasonal variation in the reflectance characteristics of vegetation and other surface features, the MODIS BRDF data product is first statistically analyzed for each season and then classified by land cover type. Using Equation (8), BRDF kernel coefficients are separately fitted for each land cover type based on the corresponding BRDF data, yielding BRDF model coefficients specific to different seasons and land cover types. These coefficients are then applied to MSC images, where land cover classification enables BRDF correction. By mitigating anisotropic reflectance effects, BRDF correction ensures more accurate surface reflectance retrieval, ultimately improving the quality of MSC images [24].

4. Atmospheric Correction Results and Analysis

4.1. Comparison of Images Before and After Atmospheric Correction

The atmospheric correction algorithm described in Section 3 effectively addresses atmospheric scattering, adjacency effects, and BRDF-related distortions. Using atmospheric parameters retrieved synchronously by the PSAC, atmospheric correction was applied to wide-swath, wide-field-of-view multispectral images. The corrected images encompass various land cover types, including urban areas, vegetation, water bodies, and soils. Figure 4, Figure 5 and Figure 6 display some of the atmospheric correction results. Figure 4a, Figure 5a and Figure 6a show the images before atmospheric correction, while Figure 4b, Figure 5b and Figure 6b show the images after atmospheric correction, with both presented as color composites using the first three bands of the MSC.

Figure 4 shows the atmospheric correction results for Beijing Airport, captured by the HJ-2A satellite’s MSC (CCD1). Before atmospheric correction, the image exhibits low contrast in land features, with blurred boundaries of the airport runway and buildings. After atmospheric correction, the boundaries of the airport runway and buildings become clear. Figure 5 presents a comparison of the atmospheric correction results for the Indian plains, captured by the HJ-2B satellite’s MSC (CCD3). The corrected image is notably clearer, with the contours of the rivers being more distinctly visible. Figure 6 displays the comparison of atmospheric correction results for a mountainous urban area, captured by the HJ-2A satellite’s MSC (CCD3). Compared to the pre-correction image, the corrected image shows more natural and vibrant colors, with clearer texture details. These indicate that atmospheric correction effectively removes the blurring effects caused by atmospheric interference, thereby improving the recognition of land features.

To quantitatively evaluate the improvement in image quality after atmospheric correction, two metrics—contrast and clarity—were calculated for the images before and after correction.

(a) Contrast

Contrast represents the ratio between bright and dark regions in an image, reflecting the gradient transition from light to dark. A higher contrast value indicates more gradient levels and richer image details [25,26]. Contrast is calculated as follows:

C = \sqrt{\frac{1}{\sum_{i = 1}^{N} w_{i}} \sum_{i = 1}^{N} w_{i} \frac{{({D N}_{i} - D N)}^{2}}{{D N}^{2}}}

(9)

where N represents the number of pixels in the local region of the image to be evaluated. w is the weighting function,

w_{i} = 0.5 (\cos (\frac{π}{\sqrt{x^{2} + y^{2}}} \sqrt{({x_{i} - x_{c})}^{2} + ({y - y_{c})}^{2}}) + 1)

, x and y represent the width and height of the local image region, (x_c, y_c) denotes the center point of the local image region, and (x_i, y_i) refers to the position of the i-th pixel in the local image region. DN_i is the grayscale value of the i-th pixel in the local image region, while DN is the weighted mean of the grayscale values in the local image region,

D N = \sqrt{\frac{1}{\sum_{i = 1}^{N} w_{i}} \sum_{i = 1}^{N} w_{i} {D N}_{i}}

The calculation of contrast primarily relies on texture and edge information. Therefore, when comparing two images with the same content, an image with less texture, overall blurriness, and weakened edges typically has a lower contrast, while a clearer image with well-defined features generally exhibits higher contrast.

(b) Clarity

Clarity, often referred to as image sharpness, is directly represented by the grayscale gradient, and different gradient operators can be used to calculate the clarity for image quality assessment [27,28]. In this study, the Sobel operator is employed to compute clarity, as defined by the following equations:

S_{x} = [\begin{matrix} - 1 & 0 & 1 \\ - 2 & 0 & 2 \\ - 1 & 0 & 1 \end{matrix}] S_{x} = [\begin{matrix} 1 & 2 & 1 \\ 0 & 0 & 0 \\ - 1 & - 2 & - 1 \end{matrix}]

(10)

The Sobel operator is applied sequentially through convolution with the image grayscale values to obtain the Sobel clarity:

S = \sqrt{{|D N (i, j) * S_{x}|}^{2} + {|D N (i, j) * S_{y}|}^{2}}

(11)

Here, S denotes the Sobel clarity, where a higher value indicates better clarity. When comparing two images with identical content, the image with higher quality will have richer details and appear clearer, resulting in a higher clarity value. Conversely, the clarity value will be lower for an image with less detail and lower quality.

Figure 7, Figure 8 and Figure 9 display the calculated contrast and clarity values before and after atmospheric correction for three locations: Daxing Airport in China, the Indian Plains, and Xianning in Hubei, China. “AC-B” represents images before atmospheric correction, “AC-A” represents images after atmospheric correction, and “Improvement” quantifies the percentage enhancement. A 100% improvement means the corrected image has double the contrast or clarity of the uncorrected image.

As shown, the contrast and clarity improvements vary by location, with the highest values observed at Xianning, followed by the Indian Plains and Daxing Airport. This variation corresponds to differences in aerosol optical depth across locations, as higher AOD leads to greater atmospheric effects and more pronounced improvements after correction.

Additionally, the improvements in contrast and clarity decrease with an increasing wavelength, indicating that shorter wavelengths in the visible spectrum are more sensitive to atmospheric scattering. This finding aligns with the Rayleigh scattering theory, which predicts stronger scattering effects at shorter wavelengths.

Overall, the atmospheric correction algorithm significantly enhances image quality, improving the visibility of surface features and supporting a more accurate analysis of remote sensing data.

4.2. Validation of Reflectance Through In Situ Measurement Comparison

To validate the atmospheric correction algorithm described in Section 3, reflectance synchronization experiments were conducted in Dunhuang, Gansu Province, and Hefei, Anhui Province, China. The surface reflectance values obtained after atmospheric correction were compared with in situ measured reflectance to evaluate the accuracy of the correction.

A surface reflectance validation experiment was conducted in Dunhuang, Gansu Province, China, from 23 January to 3 February 2021. The primary surface type at the Dunhuang site is the Gobi Desert, with coordinates at (40.0926°N, 94.3934°E). A secondary site featuring a high-reflectance surface covered with bright pebbles is located at (40.4750°N, 94.3806°E). Figure 10 and Figure 11 show the images before and after atmospheric correction, with red-marked areas indicating the regions where ground experiments were conducted. After atmospheric correction, the atmospheric effects were effectively removed, enhancing image details and surface textures.

A similar experiment was conducted in Hefei, Anhui Province, China, from 22 March to 26 March 2021. The experimental site included a wheat field at (31.5234°N, 117.3376°E) and a river water site at (31.5281°N, 117.3413°E). Figure 12 compares the pre- and post-correction images, with red marked and blue marked areas representing the ground measurement regions for the wheat field and river water, respectively. The corrected images show significant improvements in visual quality, with clearer surface features and reduced atmospheric interference.

During the ground experiment, surface reflectance was measured using a FieldSpec hand-held spectroradiometer during HJ-2A/B satellite overpasses. The FieldSpec, a product of Analytical Spectral Devices (ASD), USA, features high spectral resolution and portability, enabling precise in situ measurements across various surface types. The measurement method involved setting up multiple measurement points across the test area and employing a walking approach to ensure uniform spatial coverage.

At each measurement point, five sets of data were collected to calculate the average error. During the data processing phase, the measurements from individual points were first averaged, followed by averaging the results from all points within the experimental area. Invalid data were excluded using a variance-based threshold to minimize the impact of outliers. The processed surface reflectance results are presented in Figure 13, which shows the averaged reflectance values across the experimental area.

By applying the spectral response function (

a (λ)

) of the MSC, the surface reflectance spectrum (

ρ (λ)

) measured in the field was converted into band-equivalent reflectance (

ρ_{b a n d}

), as shown in the following equation:

ρ_{b a n d} = \frac{\int a (λ) ρ (λ) d λ}{\int a (λ) d λ}

(12)

Here,

ρ_{b a n d}

represents the band-equivalent reflectance, ensuring compatibility between in situ measurements and satellite-derived reflectance.

The band-equivalent reflectance was compared with the atmospheric-corrected surface reflectance (ACSR), and the results are presented in Figure 14. The figure highlights significant differences between the Top-of-Atmosphere (TOA) reflectance and the in situ measured band-equivalent reflectance for various surface types, including the Dunhuang site, high-reflectance site, vegetation, and water bodies. Notably, the TOA reflectance in the blue band is substantially higher than the in situ measured reflectance, and the overall curve shape shows inconsistency. However, after atmospheric correction, the surface reflectance aligns closely with the in situ measured band-equivalent reflectance, demonstrating the effectiveness of the correction process.

The atmospheric-corrected surface reflectance derived from the HJ-2A/B satellite’s MSC was compared with the in situ measured reflectance for the Dunhuang site, high-reflectance site, wheat field, and river water, which were treated as ground truth values. The calculated deviations are presented in Table 3. For these four surface types, larger reflectance deviations between the ACSR and the measured reflectance were observed in band 1 and band 2, which correspond to the blue and green bands with relatively shorter wavelengths, making them more susceptible to atmospheric influence. The largest reflectance deviation occurred in band 5 for the wheat field, which may be related to the higher aerosol optical depth on that day. However, overall, the deviations between the ACSR and the measured reflectance were mostly below 0.02, with the maximum deviation not exceeding 0.03. These results confirm the effectiveness of the correction in retrieving accurate surface reflectance values for diverse land cover types.

4.3. Comparison and Validation with Sentinel-2 Surface Reflectance Products

Sentinel-2 is part of the “Global Environment and Security Monitoring” program, carrying a multispectral imager with 13 bands at spatial resolutions of 10 m, 20 m, and 60 m and a swath width of 290 m. Among these, bands 2, 3, 4, 5, and 8 have central wavelengths and spatial resolutions similar to those of the HJ-2A/B satellite’s MSC. The specifics are detailed in Table 4.

The Sentinel-2 L2A product provides atmospheric-corrected surface reflectance images, which are the surface reflectance products. The accuracy of Sentinel-2′s surface reflectance products has undergone extensive comparative validation, and a quality assessment report has been released, making it a reliable product for surface reflectance accuracy verification [29,30,31]. To overcome the limitation of ground-based in situ measurements, which cannot encompass a wide variety of land cover types, this study utilizes Sentinel-2 surface reflectance products to compare and validate the atmospheric-corrected surface reflectance of the HJ-2A/B. This approach enables a comprehensive assessment of surface reflectance accuracy across diverse land cover types.

The surface reflectance from the three images in Section 4.1 were compared with the corresponding reflectance products from Sentinel-2 satellite for the same locations, including land cover types such as airport aprons, bare soil, water bodies, buildings, and vegetation. Areas with similar times and locations were selected, and the mean values were calculated for each, along with their absolute errors. The results are presented in Table 5.

As shown in Table 5, the surface reflectance from the HJ-2A/B MSC closely matches Sentinel-2 SR across different land cover types, including bare soil, water bodies, vegetation, aprons, and buildings, with most absolute errors being smaller than 0.04. Higher errors (>0.04) were observed for aprons and buildings, attributed to their higher reflectance values and differences in spectral response between HJ-2A/B band 4 and Sentinel-2 band 5. These discrepancies are further influenced by differences in the central wavelengths and bandwidths between the two satellites.

Overall, the results demonstrate that the atmospheric correction applied to HJ-2A/B MSC data achieves high consistency with Sentinel-2 SR products, validating the correction’s effectiveness across diverse surface types. Sentinel-2 validation complements the in situ measurements by providing broader geographic coverage and additional land cover variability.

5. Discussion

This study introduces a pixel-by-pixel synchronous atmospheric correction method for wide-swath and wide-field imaging, specifically tailored to the HJ-2A/B MSC, while incorporating BRDF effects. Atmospheric parameters were first retrieved using the PSAC, followed by spatiotemporal matching to synchronize these parameters with the target remote sensing images. An atmospheric correction lookup table was then constructed using a radiative transfer model, taking into account the spectral response functions and band configurations of the MSC. Using this lookup table, pixel-level atmospheric parameters were derived based on the viewing and solar geometries to perform synchronous atmospheric correction for the wide-field remote sensing images. Finally, BRDF correction was implemented by leveraging MODIS surface reflectance products and land cover-based pure-pixel matching, ensuring accurate and reliable surface reflectance retrieval.

Compared to traditional atmospheric correction methods, the proposed approach benefits from the real-time acquisition of atmospheric parameters, avoiding the inaccuracies introduced by fixed models or interpolated atmospheric data. The validation experiments confirmed that the corrected images showed notable improvements in visual quality, with contrast increasing by at least 55.91% and clarity by 146.46% in the visible bands. The infrared bands also exhibited significant enhancements, with contrast and clarity improvements of at least 35.63% and 90.51%, respectively. These findings highlight the effectiveness of the method in reducing atmospheric interference and improving surface reflectance retrieval.

Synchronous satellite-ground validation experiments were carried out at locations including the Dunhuang site and Hefei in China, focusing on typical land cover types such as the Gobi Desert, vegetation, and water bodies. Ground-measured reflectance was compared with the surface reflectance values obtained after atmospheric correction, showing retrieval deviations of less than 0.03. For surface types not covered by ground experiments, Sentinel-2 surface reflectance products were used as validation references, with maximum deviations below 0.043. The accuracy of the reflectance products was consistent with Sentinel-2 and MODIS products [13,31], confirming the reliability of the proposed method.

Despite these promising results, several challenges remain. The retrieval of atmospheric parameters may be affected by extreme atmospheric conditions, such as high aerosol loads or cloud contamination. Additionally, the BRDF correction approach, while effective for most land cover types, requires further refinement for complex terrains such as mountainous regions, where topographic effects introduce additional uncertainties.

The practical applications of this method span various domains, including environmental monitoring, disaster response, and agricultural assessments. The improved accuracy of surface reflectance retrieval enhances land cover classification, vegetation monitoring, and water quality assessments. As remote sensing technology advances, real-time atmospheric correction methods like the one proposed in this study will become increasingly essential for ensuring high-quality satellite imagery and reliable quantitative remote sensing products.

6. Conclusions

This study proposed a pixel-by-pixel synchronous atmospheric correction method for wide-swath, wide-field imaging, specifically designed for the HJ-2A/B MSC, while incorporating BRDF effects. The method integrates real-time atmospheric parameter retrieval using the PSAC, a radiative transfer model-based correction approach, and a semi-empirical BRDF correction.

Validation experiments demonstrated significant improvements in image contrast, clarity, and surface reflectance accuracy. Compared to pre-atmospheric correction images, the corrected images exhibited at least a 35.63% increase in contrast and a minimum 90.51% improvement in clarity. Comparisons with in situ measurements showed maximum deviations below 0.03, while validation against Sentinel-2 surface reflectance products confirmed that most deviations were below 0.04, highlighting the reliability of the correction method.

By leveraging synchronous atmospheric parameters, this method significantly improves the quality of wide-field remote sensing images and generates high-precision reflectance products. These products provide reliable data support for applications such as disaster monitoring, water resource management, and crop monitoring. Future work will focus on refining BRDF correction for mountainous regions to enhance the accuracy of surface reflectance retrieval in these complex terrains.

Author Contributions

This work was carried out in collaboration with all of the authors. Conceptualization, H.H., X.L. (Xiao Liu) and X.S.; methodology, H.H.; software, Y.W.; validation, H.H., X.L. (Xiao Liu) and R.T.; data acquisition, H.H., Z.L., X.L. (Xuefeng Lei), J.L. and L.F.; writing and revision, H.H. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported, in part, by the Aerospace Science and Technology Innovation and Application Research Project under Grant E23Y0H555S1 and Grant 6250251020 and, in part, by the Key Laboratory Fund Programme of the Chinese Academy of Sciences under Grant E33Y0HB42P1.

Data Availability Statement

The HJ-2A/B satellite images and PSAC data can be obtained from the China Centre for Resources Satellite Data and Application (available online: https://data.cresda.cn/#/home (accessed on 13 January 2025)). MODIS data can be downloaded from the official website of NASA (https://ladsweb.modaps.eosdis.nasa.gov/, last accessed on 17 January 2025). Sentinel-2 data can be downloaded from the official website of ESA (https://dataspace.copernicus.eu/, last accessed on 23 January 2025).

Acknowledgments

The MODIS data used in this study were provided by the National Aeronautics and Space Administration (NASA), and the Sentinel-2 data were provided by the European Space Agency (ESA). The authors sincerely thank their support in providing the data for this research.

Conflicts of Interest

The authors declare no conflicts of interest.

References

Kaufman, Y.J. Atmospheric effect on spatial resolution of surface imagery. Appl. Optics 1997, 23, 3399–3408. [Google Scholar]
Lyapustin, A.I.; Kaufman, Y.J. Role of adjacency effect in the remote sensing of aerosol. J. Geophys. Res. Atmos. 2001, 106, 11909–11916. [Google Scholar] [CrossRef]
Sayler, K.; Zanter, K. Landsat 8-9 Collection 2 (C2) Level 2 Science Product (L2SP) Guide Version 5.0.; Department of the Interior U.S. Geological Survey: Reston, VA, USA, 2023; pp. 5–7. [Google Scholar]
Louis, J.; Devignot, O.; Pessiot, L. S2 MPC: Level-2A Algorithm Theoretical Basis Document; EESA: Paris, France, 2021; pp. 44–67. [Google Scholar]
Román, M.O.; Justice, C.; Paynter, I.; Boucher, P.B.; Devadiga, S.; Endsley, A.; Erb, A.; Friedl, M.; Gao, H.; Giglio, L.; et al. Continuity between NASA MODIS Collection 6.1 and VIIRS Collection 2 land products. Remote Sens. Environ. 2024, 302, 113963. [Google Scholar] [CrossRef]
Vermote, E.F.; Kotchenova, S. Atmospheric correction for the monitoring of land surfaces. J. Geophys. Res. Atmos. 2008, 113. [Google Scholar] [CrossRef]
Atmospheric Correction Module: QUAC and FLAASH User’s Guide; Version 4.7; ITT Visual Information Solutions Inc.: Boulder, CO, USA, 2009; pp. 17–39.
Richter, R.; Schlapfer, D. Atmospheric/Topographic Correction for Satellite Imagery; ATCOR-2/3 User Guide, Version 9.5; ReSe Applications Schläpfer: Wil, Switzerland, 2023; Available online: www.rese-apps.com/software/download (accessed on 2 January 2025).
Doxani, G.; Vermote, E.; Roger, J.-C.; Gascon, F.; Adriaensen, S.; Frantz, D.; Hagolle, O.; Hollstein, A.; Kirches, G.; Li, F.; et al. Atmospheric Correction Inter-Comparison Exercise. Remote Sens. 2018, 10, 352. [Google Scholar] [CrossRef]
Kotchenova, S.Y.; Vermote, E.F.; Levy, R.; Lyapustin, A. Radiative transfer codes for atmospheric correction and aerosol retrieval: Intercomparison study. Appl. Optics 2008, 47, 2215–2226. [Google Scholar] [CrossRef] [PubMed]
Chraibi, E.; De Boissieu, F.; Barbier, N.; Luque, S.; Féret, J.B. Stability in time and consistency between atmospheric corrections: Assessing the re-liability of Sentinel-2 products for biodiversity monitoring in tropical forests. Int. J. Appl. Earth Obs. Geoinf. 2022, 112, 102884. [Google Scholar] [CrossRef]
Marujo, R.F.; Fronza, J.G.; Soares, A.R.; Queiroz, G.R.; Ferreira, K.R. Evaluating the impact of LaSRC and Sen2cor atmospheric correction algorithms on Landsat-8/OLI and Sentinel-2/MSI data over AERONET stations in brazilian territory. ISPRS Ann. Photogramm. Remote Sens. Spat. Inf. Sci. 2021, 3, 271–277. [Google Scholar] [CrossRef]
Roger, J.C.; Ray, J.P.; Vermote, E.F. MODIS Surface Reflectance User’s Guide (Collection 6). Version 1.6; 2022. Available online: https://modis-land.gsfc.nasa.gov (accessed on 3 January 2025).
Vermote, E.; Justice, C.; Claverie, M.; Franch, B. Preliminary analysis of the performance of the Landsat 8/OLI land surface reflectance product. Remote Sens. Environ. 2016, 185, 46–56. [Google Scholar] [CrossRef]
Wang, T.; Du, L.; Yi, W.; Hong, J.; Zhang, L.; Zheng, J.; Li, C.; Ma, X.; Zhang, D.; Fang, W.; et al. An adaptive atmospheric correction algorithm for the effective adjacency effect correction of submeter-scale spatial resolution optical satellite images: Application to a WorldView-3 panchromatic image. Remote Sens. Environ. 2021, 259, 112412. [Google Scholar] [CrossRef]
Xu, L.; Xiong, W.; Yi, W.; Qiu, Z.; Liu, X.; Zhang, D.; Fang, W.; Li, Z.; Hou, W.; Lin, J.; et al. Synchronous Atmospheric Correction of High Spatial Resolution Images from Gao Fen Duo Mo Satellite. Remote Sens. 2022, 14, 4427. [Google Scholar] [CrossRef]
Li, Z.; Xie, Y.; Hou, W.; Liu, Z.; Bai, Z.; Hong, J.; Ma, Y.; Huang, H.; Lei, X.; Sun, X.; et al. In-Orbit Test of the Polarized Scanning Atmospheric Corrector (PSAC) Onboard Chinese Environmental Protection and Disaster Monitoring Satellite Constellation HJ-2 A/B. IEEE Trans. Geosci. Remote Sens. 2022, 60, 4108217. [Google Scholar] [CrossRef]
Lei, X.; Liu, Z.; Tao, F.; Hou, W.; Huang, H.; Xie, Y.; Zhao, X.; Dong, H.; Zou, P.; Song, M.; et al. Geolocation Error Estimation Method for the Wide Swath Polarized Scanning Atmospheric Corrector Onboard HJ-2 A/B Satellites. IEEE Trans. Geosci. Remote Sens. 2022, 60, 5626609. [Google Scholar] [CrossRef]
Li, Y.; Wu, J.; Zhong, B.; Shi, X.; Xu, K.; Ao, K.; Sun, B.; Ding, X.; Wang, X.; Liu, Q.; et al. Methods of Sandy Land Detection in a Sparse-Vegetation Scene Based on the Fusion of HJ-2A Hyperspectral and GF-3 SAR Data. Remote Sens. 2022, 14, 1203. [Google Scholar] [CrossRef]
Zhao, E.; Gao, C.; Han, Q.; Yao, Y.; Wang, Y.; Yu, C.; Yu, H. An Operational Land Surface Temperature Retrieval Methodology for Chinese Sec-ond-Generation Huanjing Disaster Monitoring Satellite Data. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2022, 15, 1283–1292. [Google Scholar] [CrossRef]
Xie, Y.; Hou, W.; Li, Z.; Zhu, S.; Liu, Z.; Hong, J.; Ma, Y.; Fan, C.; Guang, J.; Yang, B.; et al. Columnar Water Vapor Retrieval by Using Data from the Polarized Scanning Atmospheric Corrector (PSAC) Onboard HJ-2 A/B Satellites. Remote Sens. 2022, 14, 1376. [Google Scholar] [CrossRef]
Shi, Z.; Li, Z.; Hou, W.; Mei, L.; Sun, L.; Jia, C.; Zhang, Y.; Li, K.; Xu, H.; Liu, Z.; et al. Aerosol Optical Depth Retrieval Based on Neural Network Model Using Polarized Scanning At-mospheric Corrector (PSAC) Data. IEEE Trans. Geosci. Remote Sens. 2022, 60, 4109018. [Google Scholar] [CrossRef]
Wang, Z.; Schaaf, C.B.; Sun, Q.; Shuai, Y.; Román, M.O. Capturing rapid land surface dynamics with Collection V006 MODIS BRDF/NBAR/Albedo (MCD43) products. Remote Sens. Environ. 2018, 207, 50–64. [Google Scholar] [CrossRef]
Wu, H.Z.; Huang, H.L.; Sun, X.B.; Liu, X.; Ti, R.; Wang, Y. BRDF Correction Based on Wide-Field Images of Environment-2 Satellite. Acta Opt. Sin. 2024, 44, 1228006. [Google Scholar]
Beghdadi, A. On Measures of Visual Contrast and Their Use in Image Processing. In Proceedings of the Colour and Visual Computing Symposium, Gjøvik, Norway, 16–17 September 2020. [Google Scholar]
Frazor, R.A.; Geisler, W.S. Local luminance and contrast in natural images. Vis. Res. 2006, 46, 1585–1598. [Google Scholar] [CrossRef]
Ryan, R.; Baldridge, B.; Schowengerdt, R.A.; Choi, T.; Helder, D.L.; Blonski, S. IKONOS spatial resolution and image interpretability characterization. Remote Sens. Environ. 2003, 88, 37–52. [Google Scholar] [CrossRef]
Shi, W.Z.; Zhu, C.Q.; Zhu, C.Y.; Yang, X.M. Multi-Band Wavelet for Fusing SPOT Panchromatic and Multispectral Images. Photogramm. Eng. Remote Sens. 2003, 69, 513–520. [Google Scholar] [CrossRef]
Doxani, G.; Vermote, E.F.; Roger, J.C.; Skakun, S.; Gascon, F.; Collison, A.; De Keukelaere, L.; Desjardins, C.; Frantz, D.; Hagolle, O.; et al. Atmospheric Correction Inter-comparison eXercise, ACIX-II Land: An assessment of atmospheric correction processors for Landsat 8 and Sentinel-2 over land. Remote Sens. Environ. 2023, 285, 113412. [Google Scholar] [CrossRef]
Pancorbo, J.; Lamb, B.; Quemada, M.; Hively, W.; Gonzalez-Fernandez, I.; Molina, I. Sentinel-2 and WorldView-3 atmospheric correction and signal normalization based on ground-truth spectroradiometric measurements. ISPRS J. Photogramm. Remote Sens. 2021, 173, 166–180. [Google Scholar] [CrossRef]
Clerc, S.; Devignot, O.; Pessiot, L. S2 MPC: Level 2A Data Quality Report; EESA and Opernicus: Paris, France, 2022; pp. 5–7. [Google Scholar]

Figure 1. Schematic of synchronized detection between PSAC and MSC.

Figure 2. Atmospheric correction flowchart for wide-swath and wide-field multispectral images.

Figure 3. Matching results of AOD and CWV for MSC image. (a) Matched AOD distribution. (b) AOD distribution after linear interpolation. (c) Matched CWV distribution. (d) CWV distribution after linear interpolation.

Figure 4. Comparison of pre- and post-atmospheric correction for HJ-2A satellite multispectral image of Beijing Daxing Airport, China. The red-marked and green-marked areas represent the selected regions for comparison and validation with Sentinel-2 data, as described in Section 4.3. (CCD1, 14 November 2022; AOD = 0.446; CWV = 0.51 g/cm²). (a) Before atmospheric correction. (b) After atmospheric correction.

Figure 5. Comparison of pre- and post-atmospheric correction for an HJ-2B satellite multispectral image of the Indian Plains region. The red-marked and green-marked areas represent the selected regions for comparison and validation with Sentinel-2 data, as described in Section 4.3. (CCD3, 25 November 2022; AOD = 0.208; CWV = 0.96 g/cm²). (a) Before atmospheric correction. (b) After atmospheric correction.

Figure 6. Comparison of pre- and post-atmospheric correction for HJ-2A satellite multispectral image of Xianning City, Hubei Province, China. The red-marked and green-marked areas represent the selected regions for comparison and validation with Sentinel-2 data, as described in Section 4.3. (CCD3, 23 December 2022; AOD = 0.564; CWV = 0.45 g/cm²). (a) Before atmospheric correction. (b) After atmospheric correction.

Figure 7. The contrast, clarity, and their improvements of the multispectral images of Daxing Airport, Beijing, China, before and after atmospheric correction from the HJ-2A satellite. (a) Contrast. (b) Clarity.

Figure 8. The contrast, clarity, and their improvements of the multispectral images of the Indian Plains region, before and after atmospheric correction from the HJ-2B satellite. (a) Contrast. (b) Clarity.

Figure 9. The contrast, clarity, and their improvements of the multispectral images of Xian Ning, Hubei Province, China, before and after atmospheric correction from the HJ-2A satellite. (a) Contrast. (b) Clarity.

Figure 10. Comparison of pre- and post-atmospheric correction for an HJ-2B satellite multispectral image at the Dunhuang site in China. The red-marked area represents the ground measurement region at the Dunhuang site. (a) Before atmospheric correction. (b) After atmospheric correction.

Figure 11. Comparison of pre- and post-atmospheric correction for HJ-2B satellite multispectral image at Northern high-reflectance site in Dunhuang, China. The red-marked area represents the ground measurement region at the high-reflectance site. (a) Before atmospheric correction. (b) After atmospheric correction.

Figure 12. Comparison of pre- and post-atmospheric correction for an HJ-2A satellite multispectral image at the suburban area of Hefei, Anhui Province, China. The red-marked and blue-marked areas represent the ground measurement regions for the wheat field and river water, respectively. (a) Before atmospheric correction. (b) After atmospheric correction.

Figure 13. The reflectance curve from the ground-based synchronized measurements. (a) Dunhuang, Gansu, China. (b) Hefei, Anhui, China.

Figure 14. Comparison chart of ground-measured reflectance and atmospheric-corrected SR. (a) Dunhuang site (25 January 2021, HJ-2B). (b) High-reflectance site (25 January 2021, HJ-2B). (c) Wheat field (25 March 2021, HJ-2A). (d) River water (25 March 2021, HJ-2A).

Table 1. Key instrument parameters of the four sensors on HJ-2A/B satellites.

Parameters	MSC	HSI	IRS	PSAC
Wavelength bands	B1: 0.45–0.52 μm	0.45–0.92 μm (Δλ = 5 nm) 0.90–2.50 μm (Δλ = 14 nm)	B1: 0.63–0.69 μm	B1: 0.400–0.420 μm
	B2: 0.52–0.59 μm		B2: 0.73–0.77 μm	B2: 0.433–0.453 μm
	B3: 0.63–0.69 μm		B3: 0.78–0.90 μm	B3: 0.545–0.565 μm
	B4: 0.69–0.73 μm		B4: 1.19–1.29 μm	B4: 0.660–0.680 μm
	B5: 0.77–0.89 μm		B5: 1.55–1.68 μm	B5: 0.845–0.885 μm
			B6: 2.08–2.35 μm	B6: 0.900–0.920 μm
			B7: 3.5–4.8 μm	B7: 1.35–1.40 μm
			B8: 10.5–11.4 μm	B8: 1.58–1.64 μm
			B9: 11.5–12.5 μm	B9: 2.21–2.29 μm
Nadir pixel resolution	16 m	48 m (0.45–0.92 μm) 96 m (0.90–2.50 μm)	48 m (B1–B5) 96 m (B6–B9)	6 km
Swath width	800 km	96 km	720 km	800 km
Total field of view (FOV)	62.6°	8.52°	60°	65°
Calibration accuracy	Absolute: ≤7% Relative: ≤3%	Absolute: ≤7% Relative: ≤3%	B1–B6: absolute ≤7%; relative ≤3%	Radiance: ≤7% DOLP: ≤0.005

Table 2. Atmospheric correction lookup table.

Parameters	Range	Step Size	Number
Altitude	0−4 km	1 km	5
SZA	0−80°	5°	17
VZA	0−65°	5°	14
RAA	0−180°	10°	19
CWV	0−10	1	11
AOD	0.0−0.2	0.05	5
	0.2−1.0	0.1	8
	1.0−2.0	0.2	5
	2.0−2.9	0.3	3
Aerosol model	6 types		6

Table 3. Comparison table of measured SR and atmospheric corrected SR.

Imaging Location	Imaging Time	AOD CWV (g/cm²)	Band	Measured Reflectance	ACSR	Absolute Error
Dunhuang site	25 January 2021 (HJ-2B)	AOD = 0.247 CWV = 0.33	B1	0.2095	0.1952	0.0143
			B2	0.2498	0.2301	0.0197
			B3	0.2708	0.2556	0.0152
			B4	0.2707	0.2536	0.0171
			B5	0.2796	0.2679	0.0117
Dunhuang site	27 January 2021 (HJ-2A)	AOD = 0.273 CWV = 0.28	B1	0.2143	0.1902	0.0241
			B2	0.2428	0.2446	0.0018
			B3	0.2687	0.2722	0.0035
			B4	0.2677	0.2605	0.0072
			B5	0.2746	0.2788	0.0042
High-reflectance site	25 January 2021 (HJ-2B)	AOD = 0.225 CWV = 0.29	B1	0.2615	0.2634	0.0019
			B2	0.3261	0.3176	0.0085
			B3	0.3747	0.3667	0.008
			B4	0.3815	0.3664	0.0151
			B5	0.4005	0.3911	0.0094
High-reflectance site	27 January 2021 (HJ-2A)	AOD = 0.228 CWV = 0.25	B1	0.2683	0.2631	0.0052
			B2	0.3261	0.3354	0.0199
			B3	0.375	0.3954	0.0204
			B4	0.383	0.3826	0.0004
			B5	0.3982	0.415	0.0168
Wheat field	23 March 2021 (HJ-2B)	AOD = 0.628 CWV = 2.35	B1	0.0423	0.0539	0.0116
			B2	0.0612	0.0746	0.0134
			B3	0.0339	0.0499	0.0161
			B4	0.1667	0.1597	0.0070
			B5	0.4349	0.4053	0.0296
Wheat field	25 March 2021 (HJ-2A)	AOD = 0.453 CWV = 1.53	B1	0.0424	0.0533	0.0109
			B2	0.0582	0.0694	0.0112
			B3	0.0331	0.0429	0.0098
			B4	0.1627	0.1521	0.0106
			B5	0.4301	0.4129	0.0172
River water	23 March 2021 (HJ-2B)	AOD = 0.599 CWV = 2.14	B1	0.0612	0.0793	0.0181
			B2	0.0909	0.1085	0.0176
			B3	0.0927	0.1015	0.0088
			B4	0.0820	0.0943	0.0123
			B5	0.0414	0.0541	0.0127
River water	25 March 2021 (HJ-2A)	AOD = 0.439 CWV = 1.40	B1	0.0623	0.0792	0.0169
			B2	0.0890	0.1038	0.0148
			B3	0.0910	0.0927	0.0017
			B4	0.0841	0.0756	0.0085
			B5	0.0422	0.0485	0.0063

Table 4. Spectral bandwidths and spatial resolutions of the Sentinel-2 and HJ-2A/B satellites.

Sentinel-2				HJ-2A/B
Name	Center (nm)	Spectral Width (nm)	SR (m)	Name	Center (nm)	Band Width (nm)	SR (m)
B2	490	65	10	B1	485	70	16
B3	560	35	10	B2	555	70	16
B4	665	30	10	B3	660	60	16
B5	705	15	20	B4	710	40	16
B8	842	115	10	B5	830	120	16

Table 5. Comparison of surface reflectance products between HJ-2A/B and Sentinel-2 satellites.

Imaging Location	Imaging Time	Band	SR of Sentinel-2	SR of HJ-2A/B	Absolute Error
Indian Plains bare soil (21.0601°N, 80.0198°E)	25 November 2022 (HJ-2B) and 24 November 2022 (S2A)	B1	0.0981	0.0941	0.0040
		B2	0.1346	0.1219	0.0127
		B3	0.1966	0.1689	0.0277
		B4	0.2288	0.2144	0.0144
		B5	0.2995	0.2733	0.0262
Indian Plains water body (21.0527°N, 80.0262°E)	25 November 2022 (HJ-2B) and 24 November 2022 (S2A)	B1	0.0569	0.0560	0.0009
		B2	0.0744	0.0635	0.0111
		B3	0.0607	0.0502	0.0105
		B4	0.0598	0.0488	0.0110
		B5	0.0363	0.0493	0.0130
Airport apron (39.4986°N, 116.4341°E)	14 November 2022 (HJ-2A) and 15 November 2022 (S2A)	B1	0.2950	0.2734	0.0216
		B2	0.3521	0.3200	0.0321
		B3	0.3822	0.3541	0.0279
		B4	0.3917	0.3493	0.0424
		B5	0.3886	0.3563	0.0323
Bare soil below airport (39.4802°N, 116.4012°E)	14 November 2022 (HJ-2A) and 15 November 2022 (S2A)	B1	0.0659	0.0844	0.0185
		B2	0.0994	0.0982	0.0012
		B3	0.1546	0.1347	0.0199
		B4	0.1791	0.1753	0.0038
		B5	0.2315	0.2252	0.0063
Buildings in Xianning (29.8952°N, 114.3308°E)	23 December 2022 (HJ-2A) and 25 December 2022 (S2A)	B1	0.1914	0.1779	0.0135
		B2	0.2147	0.1972	0.0175
		B3	0.2305	0.1948	0.0357
		B4	0.2405	0.1994	0.0411
		B5	0.2391	0.2192	0.0199
Vegetation in Xianning (29.9119°N, 114.3410°E)	23 December 2022 (HJ-2A) and 25 December 2022 (S2A)	B1	0.0328	0.0361	0.0033
		B2	0.0535	0.0435	0.0010
		B3	0.0378	0.0321	0.0057
		B4	0.0857	0.1097	0.0240
		B5	0.2723	0.2326	0.0397

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

Share and Cite

MDPI and ACS Style

Huang, H.; Wang, Y.; Liu, X.; Ti, R.; Sun, X.; Liu, Z.; Lei, X.; Lin, J.; Fan, L. Synchronous Atmospheric Correction of Wide-Swath and Wide-Field Remote Sensing Image from HJ-2A/B Satellite. Remote Sens. 2025, 17, 884. https://doi.org/10.3390/rs17050884

AMA Style

Huang H, Wang Y, Liu X, Ti R, Sun X, Liu Z, Lei X, Lin J, Fan L. Synchronous Atmospheric Correction of Wide-Swath and Wide-Field Remote Sensing Image from HJ-2A/B Satellite. Remote Sensing. 2025; 17(5):884. https://doi.org/10.3390/rs17050884

Chicago/Turabian Style

Huang, Honglian, Yuxuan Wang, Xiao Liu, Rufang Ti, Xiaobing Sun, Zhenhai Liu, Xuefeng Lei, Jun Lin, and Lanlan Fan. 2025. "Synchronous Atmospheric Correction of Wide-Swath and Wide-Field Remote Sensing Image from HJ-2A/B Satellite" Remote Sensing 17, no. 5: 884. https://doi.org/10.3390/rs17050884

APA Style

Huang, H., Wang, Y., Liu, X., Ti, R., Sun, X., Liu, Z., Lei, X., Lin, J., & Fan, L. (2025). Synchronous Atmospheric Correction of Wide-Swath and Wide-Field Remote Sensing Image from HJ-2A/B Satellite. Remote Sensing, 17(5), 884. https://doi.org/10.3390/rs17050884

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu