Columnar Water Vapor Retrieval by Using Data from the Polarized Scanning Atmospheric Corrector (PSAC) Onboard HJ-2 A/B Satellites

: As the latest members of Chinese Environmental Protection and Disaster Monitoring Satellite Constellation, the ﬁrst two of HuanjingJianzai-2 (HJ-2) series satellites were launched on 27 September 2020 by China and are usually abbreviated as HJ-2 A/B satellites. The polarized scanning atmospheric corrector (PSAC) is one of main sensors onboard HJ-2 A/B satellites, which is mainly used to monitor atmospheric components such as water vapor and aerosols. In this study, a columnar water vapor (CWV) retrieval algorithm using two bands (865 and 910 nm) is developed for PSAC. The validation results of PSAC CWV data based on ground-based CWV data derived from Aerosol Robotic Network (AERONET) show that PSAC CWV data has a high accuracy, and all statistical parameters of PSAC CWV data are better than those of Moderate-resolution Imaging Spectroradiometer (MODIS) CWV data released by NASA. Overall, there is no obvious overestimation or underestimation in PSAC CWV data. The root mean square error (RMSE), mean absolute error (MAE), relative error (RE), and percentage of CWV data with error within ± ( 0.05 + 0.10 ∗ CWV AERONET ) (PER10) of PSAC CWV data are 0.17 cm, 0.13 cm, 0.08, and 78.19%, respectively. The RMSE, MAE, RE, and PER10 of MODIS CWV data are 0.59 cm, 0.48 cm, 0.28, and 16.55%, respectively. Compared with MODIS CWV data, PSAC CWV data shows a 71% decrease in RMSE, a 73% decrease in MAE, a 71% decrease in RE, and a 372% increase in PER10. In addition, the results of day-to-day comparisons between PSAC CWV data and AERONET data show that PSAC CWV data can effectively characterize the change trend of CWV.


Introduction
Water vapor is one of the components of the atmosphere that is mainly distributed in the troposphere, and its proportion in the atmosphere usually varies in the range of 0-4%. Its main sources are evaporation of water and transpiration of plants. The water vapor content can be characterized by the columnar water vapor (CWV) with the unit of cm, which is the total amount of water vapor contained in a vertical column per unit cross-sectional area from the ground to the top of the atmosphere (TOA). Although water MW algorithm has been applied to water retrieval for sensors such as the Microwave Radiation Imager (MWRI) and the calibration microwave radiometer (CMR) [13,23]. On the whole, among the four algorithms, the NIR algorithm has the highest accuracy and the MW algorithm has the strongest applicability. In this work, the NIR algorithm is selected for water vapor retrieval of PSAC according to the band setting of PSAC.
In this study, we use the NIR algorithm to retrieve CWV from PSAC onboard HJ-2 satellites and evaluate the accuracy of retrieval algorithm based on ground-based data. The retrieval results can be used for atmospheric correction of data from other sensors onboard HJ-2 satellites and provide data support for studies related to weather forecasting and climate change. The paper is structured as follows. The data used are introduced in Section 2 and the CWV retrieval algorithm is presented in Section 3. Then, the CWV retrieval results and analysis are shown in Section 4. After that, the limitations of this study are discussed in Section 5. Finally, the paper is summarized in Section 6.

PSAC Data
The main purpose of HJ-2 series satellites is for monitoring and early warning of environmental pollution and disasters, which is operated by China Centre for Resources Satellite Data and Application (CCRSDA. Available online: http://www.cresda.com (accessed on 18 January 2022)). HJ-2A and HJ-2B are the first two of the HJ-2 series, and both satellites are polar orbiting satellites with an orbital altitude of 645 km. PSAC onboard HJ-2 satellite is mainly used for atmospheric monitoring. It has a spatial resolution of about 6 km and an imaging width of more than 800 km. PSAC has nine spectral channels with polarization detection capability, including two NIR channels (865 and 910 nm) dedicated to CWV retrieval. The spectral response functions (SRF) of two bands of PSAC used for CWV retrieval are shown in Figure 1, and more characteristics of them can be found in Table 1. Since CCRSDA only provides us with PSAC data from March 2021 to August 2021 during the in-orbit testing period, the time period chosen for this study was these 6 months.
as the Moderate-resolution Imaging Spectroradiometer (MODIS) and the Medium Resolution Spectral Imager (MERSI)-2 [20,21], the TIR algorithm has been used for water vapor retrieval for sensors such as the Spinning Enhanced Visible and Infrared Imager (SEVIRI) and the Advanced Very High Resolution Radiometer (AVHRR) [12,22], and the MW algorithm has been applied to water retrieval for sensors such as the Microwave Radiation Imager (MWRI) and the calibration microwave radiometer (CMR) [13,23]. On the whole, among the four algorithms, the NIR algorithm has the highest accuracy and the MW algorithm has the strongest applicability. In this work, the NIR algorithm is selected for water vapor retrieval of PSAC according to the band setting of PSAC.
In this study, we use the NIR algorithm to retrieve CWV from PSAC onboard HJ-2 satellites and evaluate the accuracy of retrieval algorithm based on ground-based data. The retrieval results can be used for atmospheric correction of data from other sensors onboard HJ-2 satellites and provide data support for studies related to weather forecasting and climate change. The paper is structured as follows. The data used are introduced in Section 2 and the CWV retrieval algorithm is presented in Section 3. Then, the CWV retrieval results and analysis are shown in Section 4. After that, the limitations of this study are discussed in Section 5. Finally, the paper is summarized in Section 6.

PSAC Data
The main purpose of HJ-2 series satellites is for monitoring and early warning of environmental pollution and disasters, which is operated by China Centre for Resources Satellite Data and Application (CCRSDA. Available online: http://www.cresda.com (accessed on 18 January 2022)). HJ-2A and HJ-2B are the first two of the HJ-2 series, and both satellites are polar orbiting satellites with an orbital altitude of 645 km. PSAC onboard HJ-2 satellite is mainly used for atmospheric monitoring. It has a spatial resolution of about 6 km and an imaging width of more than 800 km. PSAC has nine spectral channels with polarization detection capability, including two NIR channels (865 and 910 nm) dedicated to CWV retrieval. The spectral response functions (SRF) of two bands of PSAC used for CWV retrieval are shown in Figure 1, and more characteristics of them can be found in Table 1. Since CCRSDA only provides us with PSAC data from March 2021 to August 2021 during the in-orbit testing period, the time period chosen for this study was these 6 months.  The SRFs of two bands of PSAC used for CWV retrieval. The first line in the legend (blue line) represents the water vapor absorption transmittance (WVAT) of the atmosphere when the view zenith angle (VZA) and solar zenith angle (SZA) are equal to 0 and CWV is equal to 1 cm. The second line (yellow line) represents the WVAT when VZA and SZA are equal to 0 and CWV is equal to 3 cm. The next two are SRFs of the two bands of PSAC used for CWV retrieval.

Ground-Based CWV Data
The aerosol robotic network (AERONET) is a global ground-based aerosol observation network, which has been in operation for nearly 25 years [24]. Although it is specially established for observing aerosols rather than water vapor, it can also provide high-quality CWV data. The uncertainty of AERONET CWV data is about 10% [25,26]. Due to the continuity and high accuracy of AERONET CWV data, it is widely used for quality assessment of remote sensing CWV data [27]. In this paper, the last version (Version 3.0) of Level 1.5 AERONET data after cloud mask and quality control were used for the validation of remote sensing CWV data [28]. A total of 350 sites were used in this work, and their spatial distribution is shown in Figure 2. The AERONET CWV data mentioned above are all available from the Internet (https://aeronet.gsfc.nasa.gov/, last accessed on 23 January 2022). second line (yellow line) represents the WVAT when VZA and SZA are equal to 0 and CWV is equal to 3 cm. The next two are SRFs of the two bands of PSAC used for CWV retrieval.

Ground-Based CWV Data
The aerosol robotic network (AERONET) is a global ground-based aerosol observation network, which has been in operation for nearly 25 years [24]. Although it is specially established for observing aerosols rather than water vapor, it can also provide high-quality CWV data. The uncertainty of AERONET CWV data is about 10% [25,26]. Due to the continuity and high accuracy of AERONET CWV data, it is widely used for quality assessment of remote sensing CWV data [27]. In this paper, the last version (Version 3.0) of Level 1.5 AERONET data after cloud mask and quality control were used for the validation of remote sensing CWV data [28]. A total of 350 sites were used in this work, and their spatial distribution is shown in Figure 2. The AERONET CWV data mentioned above are all available from the Internet (https://aeronet.gsfc.nasa.gov/, last assess: 23 January 2022).

Water Vapor Retrieval
As mentioned before, the NIR water vapor retrieval algorithm is only applicable to sensors with both the water vapor absorption band and band not affected by water vapor absorption in the spectral range of 850-1250 nm [11]. Therefore, bands 5 and 6 (865 and 910 nm) of PSAC are bands selected for water vapor retrieval. Among them, band 5 is located in the spectral range of the atmospheric window and is basically not affected by gas absorption, while band 6 is strongly absorbed by the atmosphere, almost all of which is caused by water vapor.

Water Vapor Retrieval
As mentioned before, the NIR water vapor retrieval algorithm is only applicable to sensors with both the water vapor absorption band and band not affected by water vapor absorption in the spectral range of 850-1250 nm [11]. Therefore, bands 5 and 6 (865 and 910 nm) of PSAC are bands selected for water vapor retrieval. Among them, band 5 is located in the spectral range of the atmospheric window and is basically not affected by gas absorption, while band 6 is strongly absorbed by the atmosphere, almost all of which is caused by water vapor.
For NIR bands, the apparent reflectance observed by the satellite (R TOA ) at the top of the atmosphere can be expressed as [29,30]: where T g is the total gaseous absorption transmittance, ρ 0 denotes the atmospheric path reflectance caused by the atmosphere, R sur f is surface reflectance, T ↑ and T ↓ represent the total transmittance (direct + diffuse) upward and downward, respectively. Due to their relatively long and close wavelengths, the difference between path radiation and atmospheric transmittance of bands 5 and 6 can be ignored. In addition, the difference in surface reflectance between the two bands is very small. For the above reasons, it can be considered that the apparent reflectance of band 5 is equal to that of band 6 without considering water vapor absorption [21]. Since the effect of gases other than water vapor on the absorption of radiation observed by PSAC in bands 5 and 6 is negligible, the water vapor absorption transmittance of band 6 can be calculated by dividing the apparent reflectance of band 6 by that of band 5.
where T 910 nm g is the water vapor absorption transmittance of band 6, R 865 nm TOA and R 910 nm TOA are the TOA reflectance of bands 5 and 6, respectively. Previous studies have shown that the water vapor absorption transmittance of the NIR channel is strongly correlated with CWV. This relationship can be written as [31,32]: where V CWV denotes CWV, T g represents the water vapor absorption transmittance of NIR band, θ s is the solar zenith angle, θ v is the satellite zenith angle, and A, B, and C are three constants related to the characteristics of the NIR band, respectively. In this paper, the selected PSAC data was divided into two parts. Among them, PSAC data in East Asia (18-53 • N, 73-130 • E) was used to determine above three parameters (A, B, and C), and the remaining PSAC data was used to test the effectiveness of the water vapor retrieval algorithm. For the convenience of description, the region used to determine parameters of algorithm model is referred to as region 1, and the region used to test the algorithm is referred to as region 2. There are 38 AERONET sites in region 1 and 312 AERONET sites in region 2. In order to construct water vapor retrieval model for PSAC, we first matched PSAC L1 data containing TOA reflectance data, solar zenith angle data, and satellite zenith angle data with AERONET CWV data within region 1. In this paper, we use the temporal mean of AERONET PWV data within half an hour of observation time of PSAC to match the cloudless pixels where the AERONET site is located. Note that the TOA reflectance of band 2 and the TOA reflectance difference between adjacent pixels of bands 2 and 7 are used to judge whether a pixel is a cloud pixel. The specific judgment criteria are described in Section 3.2. There are a total of 839 valid matches.
After that, we can calculate the natural logarithm of water vapor absorption transmittance of band 6 of PSAC (i.e., ln T 910 nm g ) and the slanted water vapor content in the path of the sun-ground-satellite (i.e., V CWV * L) of each valid match based on the corresponding AERONET CWV data, solar zenith angle data, satellite zenith angle data, and TOA reflectance of bands 5 and 6. Finally, we can substitute the above matching results into Equation (3) to establish 839 equations containing only three unknowns (A, B, and C), and determine the values of A, B, and C by using the least squares method. The scatter plot of the slanted water vapor content in the path of the sun-ground-satellite and natural logarithm of water vapor absorption transmittance of band 6 of PSAC of the above matching results is shown in Figure 3.

Cloud Mask
The NIR CWV retrieval algorithm is only applicable to the pixels not affected by clouds, so it is necessary to remove the cloud-influenced part of the PSAC CWV retrieval results to ensure the quality of PSAC CWV data. Two methods are used to identify cloud pixels. If a pixel is identified as a cloud pixel by any of the methods, the pixel is removed as a cloud pixel.
(1) The TOA reflectance of band 2 (443 nm) and band 7 (1380 nm) is used for the cloud identification [33,34]. For a pixel that needs to be detected, if the TOA reflectance of band 2 of the pixel is greater than 0.4, the pixel would be regarded as a cloud pixel. Next, we identify cloud pixels based on the TOA reflectance of band 7. For a pixel that needs to be detected, it is first assumed that the pixel is not affected by clouds. Then we can calculate the CWV of the pixel based on Equation (5). Next, the water vapor absorption transmittance of band 7 can be calculated using the Second Simulation of the Satellite Signal in the Solar Spectrum (6SV) model based on the calculated CWV [35]. Since the TOA reflectance of band 7 is less than 1 for cloud-free pixels when water vapor content is equal to 0, so the actual observed TOA reflectance of band 7 should be less than the value of 1 multiplied by the water vapor absorption transmittance calculated above. In other words, if a pixel is not affected by clouds, the actual observed TOA reflectance of band 7 of the pixel is less

Cloud Mask
The NIR CWV retrieval algorithm is only applicable to the pixels not affected by clouds, so it is necessary to remove the cloud-influenced part of the PSAC CWV retrieval results to ensure the quality of PSAC CWV data. Two methods are used to identify cloud pixels. If a pixel is identified as a cloud pixel by any of the methods, the pixel is removed as a cloud pixel.
(1) The TOA reflectance of band 2 (443 nm) and band 7 (1380 nm) is used for the cloud identification [33,34]. For a pixel that needs to be detected, if the TOA reflectance of band 2 of the pixel is greater than 0.4, the pixel would be regarded as a cloud pixel. Next, we identify cloud pixels based on the TOA reflectance of band 7. For a pixel that needs to be detected, it is first assumed that the pixel is not affected by clouds. Then we can calculate the CWV of the pixel based on Equation (5). Next, the water vapor absorption transmittance of band 7 can be calculated using the Second Simulation of the Satellite Signal in the Solar Spectrum (6SV) model based on the calculated CWV [35]. Since the TOA reflectance of band 7 is less than 1 for cloud-free pixels when water vapor content is equal to 0, so the actual observed TOA reflectance of band 7 should be less than the value of 1 multiplied by Remote Sens. 2022, 14, 1376 7 of 13 the water vapor absorption transmittance calculated above. In other words, if a pixel is not affected by clouds, the actual observed TOA reflectance of band 7 of the pixel is less than the water vapor absorption transmittance of band 7 of the pixel calculated based on the above method. Therefore, if the actual observed TOA reflectance of band 7 of a pixel is not smaller than the water vapor absorption transmittance of band 7 of the pixel calculated based on the above method, the pixel would be regarded as a cloud pixel.
(2) The TOA reflectance difference between adjacent pixels of band 2 and band 7 is used for the cloud identification [5,36]. For a pixel that needs to be detected, the standard deviation of apparent reflectance of it and its nearest 8 pixels is used to characterize the difference. The specific calculation method is given in Equation (6). If the calculated standard deviation of band 2 is greater than 0.01, or the standard deviation of band 7 is greater than 0.005, the pixel would be considered as a cloud pixel.
where R i TOA is the TOA reflectance with a different serial number, R mean TOA is the average value of all 9 TOA reflectance participating in the calculation, and STD represents the calculated standard deviation.
Note that the above thresholds (i.e., 0.4, 0.01, and 0.005) were set with reference to the threshold used by the cloud mask algorithms of the Advanced Himawari Imager (AHI) and MODIS that use similar bands to PSAC. Since the cloud mask is not the focus of this paper, the requirement for the cloud mask algorithm in this paper is that most cloud pixels can be detected. Examples of cloud mask results based on the above methods are shown in Figure 4. By analyzing Figure 4, it can be seen that the cloud mask methods used in this work are basically effective in identifying cloudy pixels.
Remote Sens. 2022, 14, 1376 7 of 14 than the water vapor absorption transmittance of band 7 of the pixel calculated based on the above method. Therefore, if the actual observed TOA reflectance of band 7 of a pixel is not smaller than the water vapor absorption transmittance of band 7 of the pixel calculated based on the above method, the pixel would be regarded as a cloud pixel.
(2) The TOA reflectance difference between adjacent pixels of band 2 and band 7 is used for the cloud identification [5,36]. For a pixel that needs to be detected, the standard deviation of apparent reflectance of it and its nearest 8 pixels is used to characterize the difference. The specific calculation method is given in Equation (6). If the calculated standard deviation of band 2 is greater than 0.01, or the standard deviation of band 7 is greater than 0.005, the pixel would be considered as a cloud pixel.
where is the TOA reflectance with a different serial number, is the average value of all 9 TOA reflectance participating in the calculation, and STD represents the calculated standard deviation.
Note that the above thresholds (i.e., 0.4, 0.01, and 0.005) were set with reference to the threshold used by the cloud mask algorithms of the Advanced Himawari Imager (AHI) and MODIS that use similar bands to PSAC. Since the cloud mask is not the focus of this paper, the requirement for the cloud mask algorithm in this paper is that most cloud pixels can be detected. Examples of cloud mask results based on the above methods are shown in Figure 4. By analyzing Figure 4, it can be seen that the cloud mask methods used in this work are basically effective in identifying cloudy pixels.

Results
In this section, the quality of PSAC CWV data is evaluated in two ways, including overall validation results and analysis of day-to-day variation in CWV.

Results
In this section, the quality of PSAC CWV data is evaluated in two ways, including overall validation results and analysis of day-to-day variation in CWV.

Overall Validation Results
To clearly understand the advantages and disadvantages of PSAC CWV data, we not only used AERONET data to assess the quality of PSAC CWV data but compared its accuracy with that of widely used and fully validated MODIS CWV data derived from MODIS onboard Terra (MOD05) [11,37,38]. Existing validation results show that MOD05 has a high correlation with ground-based CWV data, but MOD05 has a significant overestimation [39,40]. Since the accuracy of MOD05 derived from Terra/MODIS and MYD05 derived from MODIS/Aqua is close [40], MOYD05 is not used together with MOD05. The above MODIS CWV data can be downloaded free of charge from the Internet (https://ladsweb.modaps.eosdis.nasa.gov/, last accessed on 26 January 2022). A total of five parameters were used for quality assessment of remote sensing CWV data, including root mean square error (RMSE), mean absolute error (MAE), mean bias (MB), relative error (RE), and percentage of CWV data with error within ±(0.05 + 0.1 * CWV AERONET ) (PER10). The calculation equations of RMSE, MAE, MB, and RE are as follows: where N is the number of effective matching results between AERONET CWV data and remote sensing CWV data, V i cwv is the value of remote sensing CWV data of the ith matching result, and V i, cwv is the value of ground-based CWV data corresponding to V i cwv . In this work, we used the temporal mean of ground-based observations within half an hour of the satellite observation time to match the spatial average of remote sensing CWV data within the range of 9 × 9 km centered on the ground-based site [27,39]. The scatter plots of the matching results between PSAC CWV data, MODIS CWV data, and AERONET data are shown in Figure 5. From Figure 5, it can be found that there is no obvious overestimation or underestimation in PSAC CWV data (MB = −0.03 cm). However, there is a large overestimation in the MODIS CWV dataset (MB = 0.47 cm), which is consistent with the existing quality assessment results of MODIS CWV data [39,40]. The RMSE, MAE, RE, and PER10 of PSAC CWV data are 0.17 cm, 0.13 cm, 0.08, and 78.19%, respectively. The RMSE, MAE, RE, and PER10 of MODIS CWV data are 0.59 cm, 0.48 cm, 0.28, and 16.55%, respectively. Compared with MODIS CWV data, PSAC CWV data shows a 71% decrease in RMSE, a 73% decrease in MAE, a 71% decrease in RE. By analyzing the values of above statistical parameters, it can be found that, whether the MAE and RMSE used to characterize absolute error or the RE used to characterize relative error is selected for quality assessment of remote sensing CWV data, the accuracy of PSAC CWV data is significantly better than that of MODIS CWV data.
The obvious overestimation of MODIS CWV data is the main reason why the absolute error and relative error of MODIS CWV data are obviously worse than those of PSAC CWV data. In addition, due to the big systematic error of MODIS CWV data, the PER10 of PSAC CWV data (78.19%) is significantly larger than that of MODIS CWV data (16.55%). Since the PER10 of PSAC CWV data is greater than 68.27%, it can be considered that the expected error of PSAC CWV data is less than ±(0.05 + 0.1 * CWV AERONET ). In summary, it can be concluded that PSAC CWV data has no large systematic deviation and has a high accuracy, which is higher than that of MODIS CWV data.

Analysis of Day-to-Day Variation in CWV
As shown in Figure 6, the day-to-day comparisons between PSAC CWV data and AERONET CWV data were performed at four AERONET sites in different regions to evaluate the time consistency between them. The selected sites are NEON_LENO (88.16°W, 31.85°N), SEDE_BOKER (34.78°E, 30.86°N), Tomsk (85.05°E, 56.48°N), and Ussuriysk (132.16°E, 43.70°N), respectively. In general, AERONET CWV data and PSAC CWV data have very high consistency, and more than 90% of them having a difference of less than 0.2 cm. This indicates that PSAC CWV data can effectively characterize the daily variation in CWV.

Analysis of Day-to-Day Variation in CWV
As shown in Figure 5, the day-to-day comparisons between PSAC CWV data and AERONET CWV data were performed at four AERONET sites in different regions to evaluate the time consistency between them.  N), respectively. In general, AERONET CWV data and PSAC CWV data have very high consistency, and more than 90% of them having a difference of less than 0.2 cm. This indicates that PSAC CWV data can effectively characterize the daily variation in CWV.

Discussion
Compared with the widely used water vapor retrieval algorithm based on look-up table, the semi-empirical algorithm developed in this study has an extremely high retrieval efficiency because it does not need to use the radiative transfer model such as the MODerate resolution atmospheric TRANsmission (MODTRAN) computer code [41] to build the look-up table before retrieval and retrieve CWV by querying the look-up table

Discussion
Compared with the widely used water vapor retrieval algorithm based on look-up table, the semi-empirical algorithm developed in this study has an extremely high retrieval efficiency because it does not need to use the radiative transfer model such as the MODerate resolution atmospheric TRANsmission (MODTRAN) computer code [41] to build the lookup table before retrieval and retrieve CWV by querying the look-up table [20]. Additionally, our algorithm has very strong portability. Although it is only used for PSAC at present, it can be easily used for CWV retrieval of sensors such as MODIS and MERSI-2 that have both a water vapor absorption band and a band unaffected by water vapor absorption in the spectral range of 850-1250 nm [11].
Although PSAC CWV data developed in this work has a high accuracy and can effectively characterize the change trend of CWV, it still has some limitations. First, because the water vapor retrieval algorithm developed in this study needs to use historical PSAC data and AERONET data to determine the parameters of the algorithm model (A, B, and C in Equation (3)), it cannot be used for CWV retrieval immediately after the launch of HJ-2 satellites. However, HJ-2 satellites will undergo nearly one year of in-orbit testing before operational operation, a large amount of data will be accumulated during this period, which is sufficient for building the algorithm model. Therefore, it is feasible to use it as an operational algorithm.
In addition, as shown in Figure 6, the retrieval error of PSAC CWV data generally increases with the increase in CWV. This is unavoidable mainly because the sensitivity of band 6 to water vapor gradually decreases with the increasing water vapor content [42]. Although the sensitivity of all NIR bands to water vapor decreases with increasing water vapor content, the sensitivity of different water vapor absorption channels to water vapor is different. This means that the retrieval error of remote sensing CWV data can be reduced by using a band that is more sensitive to water vapor at high water vapor content compared to band 6. For example, MODIS and MERSI-2 are equipped with three NIR water vapor absorption bands to improve the accuracy of CWV retrieval results. Therefore, to improve the accuracy of CWV retrieval results at high water vapor content, two or more water vapor absorption bands are necessary for the subsequent PSACs.
increases with the increase in CWV. This is unavoidable mainly because the sensitivity of band 6 to water vapor gradually decreases with the increasing water vapor content [42]. Although the sensitivity of all NIR bands to water vapor decreases with increasing water vapor content, the sensitivity of different water vapor absorption channels to water vapor is different. This means that the retrieval error of remote sensing CWV data can be reduced by using a band that is more sensitive to water vapor at high water vapor content compared to band 6. For example, MODIS and MERSI-2 are equipped with three NIR water vapor absorption bands to improve the accuracy of CWV retrieval results. Therefore, to improve the accuracy of CWV retrieval results at high water vapor content, two or more water vapor absorption bands are necessary for the subsequent PSACs.

Conclusions
In this study, we developed a water vapor retrieval algorithm using two bands (865 and 910 nm) for PSAC onboard HJ-2 satellites, which is constructed based on the high correlation between the water vapor absorption transmittance of band 6 at 910 nm and water vapor content. The semi-empirical algorithm consists of three parts: (1) calculating the water vapor absorption transmittance of band 6 of PSAC; (2) constructing the relationship between the water vapor absorption transmittance of band 6 and CWV; (3) inverting CWV based on the constructed relationship. Since it can directly calculate CWV based on PSAC data and does not need to build the look-up table before retrieval and retrieve CWV by querying the look-up table, it has a very high retrieval efficiency.
To clearly assess the quality of PSAC CWV data, we not only use AERONET data to validate PSAC CWV data but compare its accuracy with that of widely used MODIS CWV data. The validation results based on ground-based CWV data from 312 AERONET sites

Conclusions
In this study, we developed a water vapor retrieval algorithm using two bands (865 and 910 nm) for PSAC onboard HJ-2 satellites, which is constructed based on the high correlation between the water vapor absorption transmittance of band 6 at 910 nm and water vapor content. The semi-empirical algorithm consists of three parts: (1) calculating the water vapor absorption transmittance of band 6 of PSAC; (2) constructing the relationship between the water vapor absorption transmittance of band 6 and CWV; (3) inverting CWV based on the constructed relationship. Since it can directly calculate CWV based on PSAC data and does not need to build the look-up table before retrieval and retrieve CWV by querying the look-up table, it has a very high retrieval efficiency.
To clearly assess the quality of PSAC CWV data, we not only use AERONET data to validate PSAC CWV data but compare its accuracy with that of widely used MODIS CWV data. The validation results based on ground-based CWV data from 312 AERONET sites show that PSAC CWV has a very high accuracy, and its accuracy is significantly better than that of MODIS CWV data released by NASA. The RMSE, MAE, RE, and PER10 of PSAC CWV data are 0.17 cm, 0.13 cm, 0.08, and 78.19%, respectively. The RMSE, MAE, RE, and PER10 of MODIS CWV data are 0.59 cm, 0.48 cm, 0.28, and 16.55%, respectively. Compared with MODIS CWV data, PSAC CWV data shows a 71% decrease in RMSE, a 73% decrease in MAE, a 71% decrease in RE, and a 372% increase in PER10. In addition, time-series comparisons between PSAC CWV data and AERONET data were performed to evaluate the time consistency between them. The results show that PSAC CWV data can effectively characterize the change trend of CWV.

Informed Consent Statement: Not applicable.
Data Availability Statement: PSAC data can be obtained from China Centre for Resources Satellite Data and Application (Available online: http://www.cresda.com (accessed on 18 January 2022)). AERONET data can be downloaded from the official website of AERONET (https://aeronet.gsfc. nasa.gov/, last accessed on 23 January 2022). MODIS data can be downloaded from the official website of NASA (https://ladsweb.modaps.eosdis.nasa.gov/, last accessed on 26 January 2022).