Retrieval of Cloud Ice Water Path from FY-3F MWTS and MWHS

Abstract

Microwave sounding observations obtained from the National Oceanic and Atmospheric Administration (NOAA) and the European Meteorological Operational Satellite Program (METOP) satellites have been used for retrieving the cloud ice water path (IWP). However, the IWP algorithms developed in the past cannot be applied to the Fengyun-3F (FY-3F) microwave radiometers due to the differences in frequency of the primary channels and the fields of view. In this study, the IWP algorithm was tailored for the FY-3F satellite, and the retrieved IWP was compared with the fifth generation of reanalysis data from the European Centre for Medium-Range Weather Forecasts (ERA5) and the Meteorological Operational Satellite-C (METOP-C) products. The results indicate that the IWP distribution retrieved from FY-3F observations demonstrates strong consistency with the cloud ice distributions in ERA5 data and METOP-C products in low-latitude regions. However, discrepancies are observed among the three datasets in mid- to high-latitude regions. ERA5 data underestimate the frequency of high IWP values and overestimate the frequency of low IWP values. The IWP retrieval results from satellite datasets demonstrate a high level of consistency. Furthermore, an analysis of the IWP time series reveals that the retrieval algorithm used in this study better captures variability and seasonal characteristics of IWP compared to ERA5 data. Additionally, a comparison of FY-3F retrieval results with METOP-C products shows a high correlation and generally consistent distribution characteristics across latitude bands. These findings confirm the high accuracy of IWP retrieval from FY-3F data, which holds significant value for advancing IWP research in China.

1. Introduction

The Cloud Ice Water Path (IWP) refers to the total mass of all ice-phase particles within a vertical atmospheric column per unit area. These ice particles include ice crystals, graupel, and precipitating snow within clouds, and it is typically measured in kilograms per square meter (kg/m²). As a critical parameter for characterizing the microphysical properties of ice clouds, IWP is essential for understanding their radiative effects and role in the atmospheric water cycle [1]. Ice clouds have a substantial impact on the Earth’s radiation balance, influencing planetary albedo and solar energy production based on their coverage, distribution, and properties [2,3]. Global observations of ice clouds are essential for understanding the Earth’s climate system. Satellite remote sensing utilizes various observational wavelengths to measure distinct cloud microphysical properties. For clouds with relatively low optical thickness, the IWP can be measured using visible light sensors [4]. For thicker ice clouds, microwave sensors provide more accurate IWP measurements [5]. Algorithms for retrieving IWP using dual-channel microwave measurements from airborne millimeter-wave imaging radiometers have been under development for some time [6]. Weng and Grody demonstrated that, for a given particle volume density, the relationship between IWP and the effective diameter of ice particles can be derived using a two-stream model, and both variables can be retrieved from microwave measurements at two distinct frequencies [7]. However, since most retrieval algorithms rely on in-situ measurements over tropical oceans using aircraft, the distinction between atmospheric ice clouds and surface scatterers has not been thoroughly analyzed.

To address this, Zhao and Weng proposed a new algorithm for retrieving IWP from Advanced Microwave Sounding Unit-A (AMSU-A) measurements at 89 GHz and 150 GHz, improving its suitability for satellite-based observations [8]. Both IWP and the effective diameter of ice particles are related to the scattering parameters of ice particles, which are measured at two frequency channels (89 GHz and 150 GHz) of AMSU-A. The ratio of the scattering parameters at these frequencies directly yields the effective diameter of ice particles. With the particle volume density known, the IWP can then be derived from the scattering parameter at 150 GHz. This improved algorithm enables more accurate global retrieval of the effective diameter of ice particles and IWP from satellite data [8]. Buehler et al. noted that millimeter-wave frequency radiometers are sensitive to heavier precipitation, whereas sub-millimeter-wave frequencies are more responsive to smaller ice particles [9]. Sun and Weng expanded the two-stream radiation model and utilized the constant scanning angle along the scan line of the satellite-borne MISR/VSI to retrieve ice cloud parameters [10]. Wang et al. utilized observational data from the Microwave Humidity Sounder on the FY-3B satellite to retrieve global IWP using a deep neural network [11].

However, the application of IWP retrieval still faces several challenges. Due to limitations in the accuracy of IWP retrieval algorithms and the quality and quantity of available data, the precision of IWP obtained using deep learning methods requires further enhancement. Additionally, the lack of long-term continuous observational data hinders existing IWP datasets from providing reliable support for climate change research and long-term weather forecasting. Therefore, improving IWP retrieval algorithms and advancing the research and application of multi-source data fusion technologies are essential. This study will utilize observational data from the Microwave Temperature Sounder-III (MWTS-III) and the Microwave Humidity Sounder-II (MWHS-II) aboard the Fengyun-3F (FY-3F) satellite to explore IWP retrieval algorithms and analyze the resulting data, aiming to achieve high-precision IWP products through physical retrieval techniques.

The article is structured as follows: Section 2 introduces the MWTS-III and MWHS-II instruments on FY-3F, the Microwave Humidity Sounder (MHS) and AMSU-A instruments on the Meteorological Operational Satellite-C (METOP-C), and the fifth generation of reanalysis data from the European Centre for Medium-Range Weather Forecasts (ERA5), providing detailed information on each data source. Section 3 discusses the IWP retrieval algorithms. Section 4 compares and analyzes the IWP results retrieved from FY-3F data with the ERA5 dataset and evaluates the retrieval accuracy. Section 5 summarizes the research and discusses its limitations.

2. Satellite Data and ERA5 Data

2.1. Satellite Data

FY-3F satellite was successfully launched in August 2023 and is carrying onboard MWTS-III and MWHS-II. FY-3F is the seventh satellite in the Fengyun-3 series and the third satellite in the sun-synchronous dawn–dusk orbit [12]. As a dawn–orbit satellite, FY-3F has taken over the in-orbit operational mission of FY-3C, enhancing Earth system observation capabilities while continuing to conduct global imaging and vertical atmospheric sounding. The MWTS-III used in this study, compared to its predecessor, the Microwave Temperature Sounder -II (MWTS-II), includes two additional window channels with center frequencies of 23.8 GHz and 31.4 GHz, providing horizontal resolutions of 75 km and 65 km, respectively. These channels can detect precipitation, surface emissivity, liquid water, and water vapor [13]. MWTS-III is a total power radiometer with a scanning range of ±53.35° from the zenith direction, covering a width of 2700 km, and achieving a total field of view (FOV) of 98% within each scan line [14]. The MWHS-II on board FY-3F is the second instrument of the 03 batch, capable of retrieving vertical information on atmospheric humidity and temperature through four detection frequency bands and fifteen detection channels, with a horizontal resolution of 15 km. MWHS-II on FY-3F has adjusted the frequency of the window channel from 150 GHz to 166 GHz. Using the 166 GHz and 89 GHz window channels, IWP over the ocean can be retrieved. Channels 11 to 15, with center frequencies on the 183.31 GHz water vapor absorption line, can be used to retrieve water vapor distribution in different vertical layers of the atmosphere [15].

The European Meteorological Operational Satellite Program (METOP) is a crucial component of the European Polar System (EPS) space segment operated by the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT), providing real-time data to multiple European meteorological services, the National Oceanic and Atmospheric Administration (NOAA), and other international agencies. The third METOP satellite (METOP-C) was launched on 7 November 2018, carrying MHS and AMSU-A [16,17]. AMSU-A is a total power microwave radiometer with 15 channels operating in the frequency range of 23.8 GHz to 89 GHz. It uses a step-scan mode with a swath width of 2343 km for each channel, consisting of 30 FOVs per scan line and a scan period of 8 s. Each FOV in AMSU-A does not overlap with its adjacent FOV [18,19,20]. MHS has five frequency channels ranging from 89 GHz to 190 GHz. It includes two window channels at 89 GHz and 157 GHz, as well as three water vapor channels near 183.3 GHz, primarily used for observing tropospheric water vapor and cloud-rain information. MHS is a cross-track scanning instrument, with each scan containing 90 consecutive FOVs, covering a range of approximately 49.5° on both sides of the satellite orbit, and an antenna beamwidth of 1.11° at the half-power point, resulting in a swath width of 2348 km. Its spatial resolution at the nadir is approximately 17 km, enabling it to obtain detailed atmospheric information [21].

Table 1. Microwave sounding instrument channels used in the cloud ice water algorithms.

Instrument	Center Frequency (GHz)	Bandwidth (MHz)	Polarization	3-db Beamwidth (°)
MWTS-III	23.8	270	QH	2.20
	31.4	180	QH	2.20
AMSU-A	23.8	270	QV	3.30
	31.4	180	QV	3.30
MWHS-II	89	180	QH	2.00
	166	1500	QH	1.10
	183.31 ± 7	2000	QV	1.10
MHS	89	180	QV	1.10
	157	2800	QV	1.10
	190.311	2000	QV	1.10

presents the parameter specifications for each microwave channel utilized in this study.

2.2. ERA5 Data

ERA5 is the fifth-generation reanalysis dataset produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) [22]. Building upon ERA-Interim, it provides higher temporal and spatial resolutions. ERA5 contains a wealth of climatic data, including cloud microphysical variables such as cloud water and ice. However, it does not directly assimilate cloud-specific variables like IWP or cloud fraction. Instead, these properties are derived computationally through a numerical weather prediction model, constrained by assimilated data such as temperature, humidity, and wind speed, which determine the spatial and temporal distribution of clouds. Notably, the IWP data in ERA5 exclusively represents cloud ice and does not account for snow, defined as aggregated ice crystals. Comparisons with satellite observations show that the spatial distribution of IWP in ERA5 closely aligns with satellite-retrieved data [23,24]. The ERA5 dataset has a spatial resolution of 0.25° × 0.25° and a temporal resolution improved from 6 h to 1 h, offering comprehensive spatial and temporal coverage. Furthermore, ERA5 utilizes the Integrated Forecasting System (IFS) “Cy41r2” developed by ECMWF, which features a prognostic microphysical scheme with distinct variables for cloud liquid, cloud ice, rain, and snow. This advancement enables ERA5 to represent atmospheric ice more accurately compared to its predecessor, ERA-Interim [23]. This study uses FY-3F observational data to examine IWP retrieval algorithms and analyzes the results in conjunction with ERA5 data.

2.3. Cross-Calibration

The IWP retrieval algorithm used in this study was originally developed based on observational data from MHS and AMSU-A onboard satellites operated by NOAA and METOP polar-orbiting satellites of Europe [7,8,13]. However, the microwave observations in this study are derived from the MWTS-III and MWHS-II instruments onboard the FY-3F satellite. Variations in instrument characteristics and observational environments across different satellite systems can significantly affect IWP retrieval results. To address these discrepancies and improve the reliability and accuracy of retrievals, this study applies China’s Advanced Radiative Transfer Modeling System (ARMS) and employs the double-difference method to cross-calibrate the FY-3F observations with those from METOP satellites. This approach effectively mitigates systematic errors caused by platform differences, providing a robust data foundation for subsequent IWP retrievals.

The ARMS was officially released in late 2020. This system introduced an accurate and efficient scheme for calculating atmospheric transmittance and developed a comprehensive atmospheric spectral dataset. Its effectiveness in assimilation and retrieval applications for meteorological satellites has been extensively validated. Moreover, its performance is comparable to internationally recognized radiative transfer models, such as Radiative Transfer for TOVS (RTTOV) and Community Radiative Transfer Model (CRTM) [25,26]. The double-difference method, a widely used technique for satellite data cross-calibration, effectively reduces systematic biases between instruments, enhancing the consistency and reliability of observational data. Variations in design principles and orbital characteristics among sensors on different satellites can introduce systematic errors, complicating direct comparisons of satellite observations. To address this issue, the double-difference method utilizes third-party data to establish a consistent reference framework. This reference is then used to calculate and correct observation errors across satellites, ensuring accurate calibration. In this study, the ARMS was employed to generate simulated values as third-party data for calibrating FY-3F satellite observations to align with METOP-C satellite data. Accurate atmospheric and surface parameters are crucial for obtaining reliable satellite observation simulations from ARMS. Accordingly, ERA5 data under clear-sky conditions were spatially and temporally matched to provide input parameters for the ARMS system, producing simulated satellite observation values consistent with actual observational conditions. The first step in applying the double-difference method to correct FY-3F observation data involves calculating the differences between satellite observation values and their corresponding simulated values, as expressed in Equations (1) and (2):

$$D_F=O_F-B_F$$

(1)

$$D_M=O_M-B_M$$

(2)

In Equations (1) and (2), $O_F$ and $O_M$ represent the observation values from the FY-3F and METOP-C satellites, respectively. $B_F$ and $B_M$ denote the corresponding simulated values for these satellites, derived using the ARMS model. $D_F$ and $D_M$ signify the differences between the observation values and their respective simulated values. Since $B_F$ and $B_M$ are simulated values obtained from the ARMS system under identical atmospheric and surface conditions, they can be assumed equal ($B_F=B_M$). Using Equations (1) and (2), the double difference between the two satellites is calculated, as shown in Equation (3):

$$D_D=O_F-O_M$$

(3)

The double-difference value derived from Equation (3) is used to correct the FY-3F satellite observation data, ensuring alignment with the METOP-C satellite data. This method is applicable to similar instruments on different satellites and effectively addresses inconsistencies in observation data between satellite platforms.

Figure 1 displays scatter plots of observed brightness temperatures from corresponding channels of MWHS-II and MHS before and after correction using the double-difference method. Figure 1a,b shows the scatter plots of MWHS-II brightness temperatures before correction against MHS in corresponding channels, while Figure 1c,d illustrates the same after correction. It can be observed that, as FY-3F and METOP-C are satellites in similar orbits, the observed brightness temperatures from corresponding channels exhibit high consistency. Following correction using the double-difference method, the correlation between the brightness temperatures from the two satellites improves significantly, with reductions in both RMSE and MAE, thereby enhancing the accuracy of IWP retrieval.

Similarly, Figure 2 presents scatter plots of observed brightness temperatures from corresponding channels of MWTS-III and AMSU-A before and after correction using a radiative transfer simulation system. Figure 2a,b depicts the brightness temperatures of MWTS-III before correction against AMSU-A, while Figure 2c,d illustrates the same after correction. As in Figure 1, the observed brightness temperatures from corresponding channels of MWTS-III and AMSU-A demonstrate high consistency. Post-correction, the correlation improves, with noticeable reductions in RMSE and MAE. These results indicate that the double-difference method effectively reduces discrepancies in observational data between the FY-3F satellite and its counterparts, enhancing data reliability.

3. Description of Cloud Ice Water Algorithm

Algorithms for retrieving IWP from satellite observation data have been extensively developed over the years [7,8,13], as illustrated in Equation (4):

$$IWP=\mu D_e{\rho }_i(\Omega /{\Omega }_N)$$

(4)

where $\mu$ is the cosine of the zenith angle, $D_e$ is the ice particle effective diameters, ${\rho }_i$ is the ice particle bulk volume density, $\Omega$ is the ice cloud scattering parameter, and ${\Omega }_N$ is the normalized scattering parameter.

Weng and Grody derived the relationship between the brightness temperature emanating from ice clouds $T_B(z_t,\mu )$ and that at the cloud base $T_B(z_b,\mu )$ using the two-stream approximation at microwave frequencies [7], as expressed in Equation (5):

$$T_B(z_t,\mu )={\displaystyle \frac{T_B(z_b,\mu )}{1+\Omega (\mu )}}$$

(5)

where $z_t$ and $z_b$ are the heights of the top and base layers of clouds.

As shown in Equation (4), the IWP is proportional to $\Omega$, while the relationship between ${\Omega }_N$ and $D_e$ is nonlinear and may depend on the specific ice particle size distribution and bulk density. Consequently, measurements at two different frequencies are required to determine both IWP and $D_e$ for a given particle bulk density [7,27]. In this study, a constant bulk density for ice particles of 920.0 kg/m³ is assumed, making IWP uniquely related to $\Omega$ and $D_e$ [8].

The ratio of scattering parameters at two different frequencies, denoted as $r$, is determined using Equation (4):

$$r(D_e)={\displaystyle \frac{{\Omega }_{89}}{{\Omega }_{166}}}={\displaystyle \frac{{\Omega }_{N89}(x,m)}{{\Omega }_{N166}(x,m)}}$$

(6)

where $x$ is the size parameter ($x=\pi D/\lambda$) and $m$ is the complex refractive index. In 1992, Weng established an empirical relationship between $r$ and $D_e$ using simulation data from a radiative transfer model and experimentally derived the relationship between ${\Omega }_N$ and $D_e$ [28]. This relationship is expressed in the Equations (7) and (8):

$$D_e=a_0+a_{1}r+a_2{r}^2+a_3{r}^3$$

(7)

$${\Omega }_N=\exp \{b_0+b_1\ln (D_e)+b_2{[\ln (D_e)]}^2\}$$

(8)

Among these, $a_i (i=0,1,2,3)$ and $b_i (i=0,1,2)$ are parameters dependent on the bulk density and size distribution of ice particles. In this case $a_0=-0.24843$, $a_1=3.86726$, $a_2=-4.70782$, and $a_3=4.67150$.

Table 2. The parameter used in the calculation of ${\Omega }_N$.

Parameters	Thresholds	${\mathit{b}}_0$	${\mathit{b}}_1$	${\mathit{b}}_2$
${\Omega }_N$	$D_e<1.2 \mathrm{mm}$	−1.74663	1.90711	−0.73029
	$D_e\ge 1.2 \mathrm{mm}$	−1.58571	1.52230	1.52230

lists the parameters used in this study for calculating ${\Omega }_N$. In summary, once $T_B(z_b,\mu )$ is determined, the IWP can be retrieved [8].

Over the ocean, $T_B(z_b,\mu )$ can be derived using the radiative transfer equation. For a non-scattering atmosphere, this equation is expressed as [29]:

$$T_b=T_s[1-(1-\epsilon ){\gamma }^2]-\Delta T(1-\gamma )[1+(1-\epsilon )\gamma ]$$

(9)

Here, $\gamma$ is atmospheric transmittance, $\epsilon$ is the surface emissivity, $T_s$ is the surface temperature, and $\Delta T$ is the temperature difference between the surface and the atmospheric mean temperature. For a scatter-free, isothermal atmosphere, where $\Delta T=0$, the brightness temperature at microwave frequencies can be approximately represented using Equation (10) [29] as follows:

$$T_b=T_s[1-(1-\epsilon ){\gamma }^2]$$

(10)

In this context, $\gamma$ can be further expressed as:

$$\gamma =\exp [-({\tau }_0+{\tau }_v+{\tau }_l)/\mu )]$$

(11)

where ${\tau }_0$ is parameterized as a function of surface temperature, ${\tau }_l$ is derived as a function of cloud liquid water and cloud layer temperature, and ${\tau }_v$ is parameterized as a function of water vapor path [7].

Due to the complexity and variability of land surface types and the non-uniformity of background radiation, the application of the radiative transfer equation to estimate the brightness temperature at the base of ice clouds over land is limited. Zhao and Weng derived the brightness temperature at the base of ice clouds by utilizing observed brightness temperature data from the Advanced Microwave Sounding Unit (AMSU) under non-scattering conditions, combined with data from the Advanced Very High Resolution Radiometer (AVHRR), and applying an empirical relationship between low- and high-frequency channels [8]. After validation, the results showed errors comparable to those obtained over the ocean. This study uses brightness temperature data from the 23.8 GHz and 31.4 GHz channels on MWTS-III, along with data from the 89 GHz and 166 GHz channels on MWHS-II, to establish an empirical relationship between the brightness temperatures of low- and high-frequency channels. The empirical relationships used in this study are presented in Equations (12) and (13).

$$T_{b\_89}=122.85+1.08T_{b23}-0.54T_{b31}$$

(12)

$$T_{b\_166}=313.20+1.32T_{b23}-1.70T_{b31}$$

(13)

Figure 3 presents scatter plots illustrating the statistical relationship between the calculated brightness temperatures at the 89 GHz and 166 GHz channels, derived using Equations (12) and (13), in conjunction with brightness temperature data from the 23.8 GHz and 31.4 GHz channels of MWTS-III, and the corresponding observed brightness temperatures. As shown in Figure 3a, there is a strong correlation between the 89 GHz brightness temperatures obtained from the 23.8 GHz and 31.4 GHz channel data and the observed brightness temperatures, with $R^2=0.942$ and $RMSE=4.561$. As illustrated in Figure 3b, although the accuracy of the 166 GHz brightness temperatures obtained using the 23.8 GHz and 31.4 GHz channel data is slightly lower than that of the 89 GHz brightness temperatures, they still exhibit a high correlation with the observed brightness temperatures. This indicates that, despite the complexity of land surface types, high-precision brightness temperature data at the 89 GHz and 166 GHz channels can still be obtained, thereby facilitating the acquisition of more accurate IWP retrieval data.

4. Results

4.1. Spatio-Temporal Evolution Characteristics of IWP Retrieval Products

Figure 4 illustrates the daily mean values and differences in IWP retrieved from two satellite datasets (FY-3F and METOP-C) over global, oceanic, and terrestrial regions from June to July 2024.

The Figure 4 shows that the IWP retrieved from FY-3F and METOP-C data exhibit a consistent global variation, with a small difference between the two datasets. The average difference in IWP is approximately 0.021 kg/m², with a maximum difference of 0.092 kg/m². In oceanic and terrestrial regions, the IWP values also follow a consistent variation, with differences of 0.035 kg/m² and −0.027 kg/m², respectively. The maximum differences between the two datasets in these regions are 0.063 kg/m² in oceanic regions and −0.062 kg/m² in terrestrial regions. Despite the use of different retrieval algorithms for IWP in oceanic and terrestrial regions, the results show a high degree of accuracy. However, the IWP retrieved from FY-3F data is slightly lower than that from METOP-C data in terrestrial regions, while the reverse is observed in oceanic regions. This discrepancy may be related to differences in satellite observation times, instrument characteristics, and algorithmic variations, warranting further investigation. In summary, despite the use of distinct algorithms for oceanic and terrestrial regions, the accuracy of the IWP retrieved from both datasets is relatively consistent, demonstrating that FY-3F data can effectively retrieve IWP globally.

To analyze the seasonal variations in IWP, this study used FY-3F data from June and July 2024. The retrieved IWP values were averaged daily to investigate long-term variations. Figure 5 depicts the daily average IWP variations over oceanic regions. To validate these findings, a comparative analysis was performed using concurrent ERA5 data and METOP-C products. Figure 5 presents the variations of IWP across different latitude ranges over oceanic regions. The IWP values retrieved from satellite data show a high degree of consistency with ERA5 variations in low-latitude regions (30°S–30°N). However, in mid-to-high latitude regions (30–60°N and 30–60°S), the IWP values from ERA5 are generally higher than those retrieved from satellite data. This aligns with Dou et al., who observed that ERA5 struggles to accurately simulate the spatial distribution of atmospheric ice clouds at mid-to-high latitudes and often overestimates cloud ice in these regions [23]. Additionally, discrepancies between satellite-retrieved IWP and ERA5 data differ between the Northern and Southern Hemispheres. In the Northern Hemisphere, the average IWP value from ERA5 is 0.18 kg/m², significantly higher than the satellite-retrieved value of 0.12 kg/m². Conversely, in the Southern Hemisphere, the average IWP value from ERA5 is 0.17 kg/m², with a smaller difference from the retrieved value of 0.14 kg/m². Hong and Liu noted that ice cloud content is higher in winter than in summer, a pattern consistent with our retrieval results [30]. In comparison, ERA5 exhibits less pronounced seasonal variations in IWP, suggesting that our IWP retrieval more accurately captures expected seasonal variations. Moreover, the IWP values retrieved from FY-3F and METOP-C data exhibit similar variation patterns with minimal numerical differences, demonstrating that FY-3F data can be effectively utilized for global IWP retrieval.

4.2. Comparison Between FY-3F Retrieval Results and ERA5 Data

To assess the discrepancies between IWP retrievals from FY-3F satellite data and ERA5 reanalysis data, and to examine their specific characteristics within the cyclone region, this study focuses on retrieval results on 1 October 2024, when Super Typhoon Krathon attained its peak intensity over the Luzon Strait. Bilinear interpolation was used for spatiotemporal matching between the 24-hourly ERA5 reanalysis data and the FY-3F data for the day, resulting in the global IWP distribution shown in Figure 4. To minimize errors caused by excessively high latitudes during the IWP retrieval process, the analysis was limited to regions between 60°N and 60°S.

Figure 6 illustrates that the IWP retrieved using FY-3F data exhibits a high degree of consistency with ERA5 in spatial distribution. However, ERA5 data contains a larger proportion of low IWP values. This discrepancy can be attributed to the resolution differences: ERA5 has a spatial resolution of 25 km, whereas the retrieval algorithm employed in this study offers higher resolution. Additionally, it may result from the physical processes representing clouds and precipitation in numerical models, which include grid-scale cloud microphysics parameterization and sub-grid-scale convective parameterization schemes. Notably, the convective parameterization schemes for cumulus convection do not explicitly represent ice clouds in terms of water content, as these processes manifest primarily as precipitation. Consequently, ERA5’s ice cloud content excludes ice clouds formed through sub-grid-scale convective processes. In contrast, the retrieval algorithm used in this study captures IWP information more accurately on a global scale. Dou et al. noted that ERA5 struggles to effectively simulate atmospheric ice conditions in mid-to-high latitudes, underestimating high IWP values and overestimating low IWP values compared to satellite observations [23]. These findings align with the observations from this study’s retrieval products, supporting the validity of using FY-3F data for IWP retrieval. Moreover, the higher spatial resolution of FY-3F data allows for more detailed information, making it a valuable tool for global IWP studies.

Figure 7 presents the IWP retrieved from FY-3F satellite data on 1 October 2024, in both tropical and extratropical cyclones, alongside the corresponding IWP data from ERA5. The results demonstrate that the spatial distribution of IWP retrieved from FY-3F data closely aligns with ERA5 data in both regions. Moreover, the FY-3F retrievals provide higher spatial resolution, particularly in capturing fine-scale typhoon structures, enabling more detailed and accurate observations of IWP. These findings indicate that the retrieval method used in this study not only effectively retrieves high-resolution global IWP information but also captures finer details in localized regions.

To compare retrieval results across different latitudes, this study analyzed marine region data from 1 October 2024 and compared the results with ERA5 cloud ice data to examine the IWP distribution across various latitudes. Probability density distribution maps and cumulative distribution function curves were generated. As illustrated in Figure 8, the blue line represents IWP retrieved using FY-3F data, while the green line represents IWP data from ERA5. The results indicate that both datasets exhibit similar numerical distribution characteristics, with values concentrated in the low-value range and relatively sparse in the high-value range. However, ERA5 IWP data are more tightly concentrated in the low-value range compared to FY-3F retrievals. This discrepancy may result from the lower resolution of ERA5 data, which are derived from numerical model calculations and may lack the ability to capture detailed ice cloud information. Conversely, FY-3F IWP retrievals, based on a higher-resolution physical retrieval algorithm, provide more detailed and accurate representations of ice clouds, leading to more realistic outcomes.

Figure 9 displays the histogram distributions of the differences between IWP retrievals using FY-3F data and ERA5 reanalysis data. Panel (a) shows the difference distribution over ocean regions, while panel (b) illustrates the distribution over land regions. Panel (c) and (d) illustrate the distributions of differences between the two datasets for high and low IWP values in oceanic regions, respectively. As observed in Figure 7, despite employing different methods to estimate the brightness temperature at the base of ice clouds over oceans and land, the IWP retrievals from FY-3F data closely align with ERA5 data in both regions. The differences are predominantly centered around zero, with most values falling within the range of −0.25 to 0.25 kg/m², and only a few outliers displaying significant discrepancies. Additionally, the histogram distributions of both datasets follow a normal distribution. Over oceans, the mean IWP retrieved from FY-3F data is 0.21 kg/m², compared to 0.15 kg/m² from ERA5 data. Over land, the mean IWP from FY-3F data is 0.06 kg/m², compared to 0.08 kg/m² from ERA5 data. These findings indicate a high level of consistency between IWP retrievals from FY-3F satellite observations and ERA5 reanalysis data across both oceanic and terrestrial regions.

4.3. Comparison Between FY-3F Retrieval Results and METOP-C Products

To validate the accuracy of IWP retrievals from FY-3F satellite data, this study compared them with IWP products derived from METOP-C satellite data. Figure 8 illustrates the global IWP results retrieved from FY-3F and METOP-C satellite data on 24 July 2024. The comparison reveals a high degree of spatial consistency between the two datasets. Additionally, the IWP retrievals from FY-3F data exhibit higher spatial resolution compared to those from METOP-C data, underscoring the significance of FY-3F satellite data for advancing ice cloud research.

As illustrated in Figure 10, the IWP retrieval results from FY-3F and METOP-C data demonstrate high spatial consistency. However, significant differences are observed in the Southern Hemisphere, whereas the results in the Northern Hemisphere are nearly identical. Between 30°N and 60°N, the mean IWP retrieved using FY-3F data is 0.105 kg/m², compared to 0.125 kg/m² using METOP-C data. From 30°N to 30°S, the mean IWP retrieved from FY-3F data is 0.245 kg/m², while that from METOP-C data is 0.234 kg/m². Between 30°S and 60°S, the mean IWP retrieved using FY-3F data is 0.123 kg/m², compared to 0.111 kg/m² using METOP-C data. These findings indicate that the IWP retrieval results from both datasets are relatively consistent across different regions. Furthermore, MWTS-III’s wider swath compared to MHS enables the FY-3F data to provide better global coverage. This allows for more seamless integration of adjacent scan tracks in mid- to low-latitude regions, significantly enhancing the ability to observe the global distribution of ice clouds.

Figure 11 displays a histogram illustrating the differences in IWP values retrieved from FY-3F and METOP-C data. The histogram exhibits a normal distribution, characterized by a prominent peak near zero and symmetric tails on both sides, indicating a high degree of consistency between the IWP values derived from the two datasets. Most differences fall within the range of ±0.25 kg/m², which may result from variations in satellite observation angles, differences in data acquisition times, or other factors. To further examine the correlation between IWP values from the two datasets, points with identical locations and scan times were selected to create a correlation scatter plot, shown in Figure 12. The results reveal that IWP values retrieved from FY-3F data are highly correlated with those from METOP-C data, underscoring the accuracy and reliability of the IWP retrieval method using FY-3F data.

Figure 13 displays IWP retrieval results for typhoon and extratropical regions based on satellite observational data from 24 July 2024. The IWP retrieved from FY-3F data in the typhoon region shows high consistency with that from METOP-C data and provides higher resolution, enabling a detailed observation of the typhoon’s internal structure and path. However, due to its narrower scanning width, the METOP-C data exhibits more missing values, which can affect the analysis of IWP variations. In contrast, FY-3F data mitigates this issue and, with its higher resolution, captures atmospheric ice cloud information more effectively.

To further analyze the numerical distribution characteristics of IWP retrieved from both datasets in different regions, probability density distribution diagrams and cumulative distribution function curves were plotted for the typhoon and extratropical regions (Figure 14). These visualizations demonstrate that the central tendency of IWP values retrieved from the two datasets is relatively consistent across both regions. In the extratropical region, the mean IWP values retrieved from FY-3F and METOP-C data are 0.15 kg/m² and 0.18 kg/m², respectively, indicating close agreement. The FY-3F data shows a higher concentration of values in the lower range, possibly due to uncertainties at the edges of the scanning band, where the extratropical region lies. In the typhoon region, the mean IWP values retrieved from FY-3F and METOP-C data are 0.22 kg/m² and 0.23 kg/m², respectively, further underscoring the high consistency between the two datasets. These results confirm that the IWP retrieved from FY-3F data aligns closely with the METOP-C product.

This study analyzed the distribution characteristics of IWP retrieval results across different latitude bands using FY-3F and METOP-C observational data from June and July 2024. Statistical analyses of the IWP retrieved from the two datasets were performed across latitude bands, as shown in Figure 15. The results indicate that the IWP distribution characteristics retrieved from both datasets are relatively consistent across latitude bands. Between 60° and 35° latitude in both hemispheres, the IWP from both datasets exhibits a relatively stable variation. However, the peak IWP retrieved from FY-3F data occurs near 15°S, whereas the peak from METOP-C data is located near 7°S. This discrepancy in peak positions may stem from differences in observation times between the two datasets. Between 35°N and 60°N, the IWP values retrieved from FY-3F data are slightly lower than those from METOP-C, while in most other latitude ranges, the FY-3F values are slightly higher. The largest difference in IWP between the two datasets occurs at 15°S, with a maximum difference of 0.081 kg/m², while the smallest difference is at 59°N, with a minimum difference of 0.015 kg/m². Notably, both datasets exhibit relatively high peaks near 5°S, with a difference of 0.027 kg/m² in this latitude band. Overall, the IWP retrieval results from the two datasets show high consistency in their latitude band distributions, with differences remaining relatively small. The IWP retrieved from FY-3F data tends to be slightly higher than that from METOP-C data in most regions. This may be attributed to the higher resolution of FY-3F satellite observations, which capture more detailed IWP information. In contrast, the lower resolution of METOP-C satellite observations may result in the loss of some high-value regions in the IWP product, leading to the observed differences between the datasets.

5. Conclusions and Discussion

To accurately retrieve IWP and obtain high-resolution retrieval results, this study utilized observational data from MWTS-III and MWHS-II onboard the FY-3F satellite. A physical retrieval algorithm was applied to derive IWP values, which were then compared with the ERA5 reanalysis product and METOP-C data. Analysis of probability density distribution plots and cumulative distribution function curves revealed that the IWP retrieved using FY-3F data exhibited numerical distribution characteristics consistent with ERA5, with a generally similar global distribution.

Comparisons between satellite products and reanalysis data on both global and regional scales showed that FY-3F and METOP-C retrievals achieved higher resolution than ERA5, enabling clearer observations of IWP magnitude and density distribution—an important advancement for IWP research. In cyclonic regions, FY-3F retrievals effectively captured the internal structure and movement paths of typhoons, significantly aiding continuous observation and predictive analysis. Long-term variation analysis showed that FY-3F IWP retrievals aligned closely with ERA5 data while better capturing seasonal variation characteristics, enhancing their utility for studying IWP relationships with other atmospheric variables. A comparison of difference histograms and scatter plots between FY-3F and METOP-C products revealed a high correlation, indicating that FY-3F retrievals not only showed strong consistency with reanalysis data but also high alignment with other satellite observations. Latitude-based comparisons further confirmed the spatial and numerical consistency of FY-3F retrievals with METOP-C data. Separate statistical analyses for oceanic and land regions demonstrated the high accuracy and global applicability of FY-3F IWP retrievals.

However, due to the cross-track scanning nature of microwave sounders, pixel resolution varies significantly between nadir and scan edges, introducing inconsistencies in IWP representation. Given the large spatial variability of IWP, it is critical to account for the effects of resolution variations when using these products. Additionally, since ERA5 data are not absolute truth values, further validation of IWP retrieval accuracy is necessary. Future studies should integrate data from multiple satellites to enhance validation and refine retrieval accuracy.