1. Introduction
The progress in numerical weather prediction (NWP) since 1970 has greatly benefited from the assimilation of satellite data (Eyre 2008). As a growing number of satellite observations are assimilated in forecasting systems and contribute to the improvement of initial conditions for the forecasts, an accurate simulation of satellite radiances is becoming more important in satellite data assimilation. The accuracy of simulated radiances is also critical for other applications such as optimal estimation retrieval of atmospheric parameters including trace gases, for which model equivalents are utilized as a priori for the retrieval. An accurate simulation of satellite data is also important for the intercalibration of satellite instruments, as the radiances simulated with high spatial resolution NWP models are found to be useful for the evaluation of measurements from multiple satellites (Saunders et al. 2013; Lee and Ahn 2021).
Sources of errors in radiative transfer (RT) simulations may include errors in input data such as NWP model background fields and surface emissivity, also errors in cloud screening, and many others. Particularly, errors arising from neglecting the satellite viewing geometry in RT simulations may far exceed the instrument noise for strongly absorbing channels (Bormann and Healy 2006). Since satellites observe a target area on Earth from different viewing angles, the slant path from the satellite to the target on Earth varies with varying viewing angles. Therefore, it is important to take the slant satellite viewing geometry into account to achieve the simulation accuracy in the observation operator calculation, thereby producing better model equivalents with RT simulation during the assimilation. Traditionally, however, the RT simulation has exploited the vertical profiles of atmospheric variables on a model grid interpolated to a geolocation and has neglected the slanted viewing geometry.
Studies have investigated the effect of taking the horizontal structure of the atmosphere into account on the RT simulation of microwave/infrared radiances from low-Earth-orbiting (LEO) sensors such as the Advanced Microwave Sounding Unit-A and the Atmospheric Infrared Sounder (Joiner and Poli 2005; Poli et al. 2005). Bormann and Healy (2006) also have demonstrated the importance of taking the viewing geometry of limb sounders such as the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) in the radiance simulation and Healy et al. (2007) introduced two-dimensional observation operators to show the benefits of taking the horizontal structure of the atmosphere into account in the assimilation of GPS radio occultation measurements.
In an operational setting, the European Centre for Medium-Range Weather Forecasts (ECMWF) is the first NWP center that has fully taken the slanted viewing geometry into account for clear-sky sounder radiances, with the upgrade of the Integrated Forecasting System on 22 November 2016 (Bormann 2017), and this change in the ECMWF system has led to positive forecast impacts most noticeable in the stratosphere and also to significant reductions in the analysis increments in the upper stratosphere, particularly at high latitudes. Burrows (2018) extended this work for the assimilation of geostationary radiance data in the ECMWF system to reduce the forward-model error and to make use of geostationary data with larger zenith angles. The impact of slant satellite-viewing geometry on the simulation and assimilation of satellite radiances was also investigated in the Environment Canada’s weather forecast system (Shahabadi et al. 2018, 2020) and it was found that the radiance simulation and assimilation for both short-term and long-term periods are greatly improved with the slant-path modeling.
The Korea Meteorological Administration (KMA) launched a new domestic atmospheric model system, named Korean Integrated Model (KIM) in April 2020 (Hong et al. 2018) to replace the previous model adopted from the Met Office Unified Model (UM) in 2010. Since 2020, KMA has been operating both systems, that is, UM and KIM, in parallel, but neither system has been taking the slanted viewing geometry into account. As a preliminary step to take the slanted satellite viewing geometry into account in the NWP models operated in KMA, this study investigates the impact of slant path on the simulated radiance, Jacobians, and the retrievals. Since the impact of slant-path modeling on the assimilation of LEO observations has been widely studied, this study mainly focuses on the exploitation of slant-path modeling for the geostationary (GEO) satellites, particularly the second geostationary satellite of Korea, Geo-KOMPSAT-2A (GK2A) launched on 4 December 2018 and stationed at 128.2°E longitude (Kim et al. 2021). The contribution of observations from GEO satellites to forecast error reduction is known to be smaller than those from microwave sounders or hyperspectral infrared sounders (McNally 2014; Joo et al. 2013). However, GEO satellite data can play an important role to improve the forecast accuracy when assimilated over the domains with spatially and temporally sparse observations (Burrows 2019). Specifically, radiances from GEO satellites with short time interval (e.g., 2 min for rapid scan) can be used for high-resolution convective scale models, and using slant paths in these model fields is also important, especially because more grid boxes will be involved. In addition, the three water vapor absorption channels of the advanced imagers on board GEO satellites, such as the Advanced Himawari Imager (AHI) on board Himawari-8, Advanced Baseline Imager (ABI) on board the Geostationary Operational Environmental Satellite (GOES), and Advanced Meteorological Imager (AMI) on board GK2A, have weighting functions peaking at three different altitudes between 300 and 600 hPa and studies have found that the assimilation of water vapor observations from such GEO satellites with high spatiotemporal resolutions can benefit regional NWP models. For example, Wang et al. (2018) showed that the information from those water vapor channels contribute to improve heavy precipitation forecasts in a regional NWP model. Lu et al. (2019) also showed that the assimilation of moisture information from AHI reduces the average track error and improves the prediction of tropical cyclone intensity.
The slant-path effect is supposed to be greater for strongly absorbing spectral regions (e.g., water vapor channels in the advanced imagers) and taking the viewing geometry of GEO satellites into RT simulation contributes to reduce the forward-model error at larger zenith angles (Burrows 2018), thereby enabling more GEO data to be included in the assimilation. Additionally, the slant-path modeling can be particularly important where the horizontal variations of temperature and moisture significantly increase such as the frontal zones, which accompany interesting weather phenomena in the midlatitudes. Therefore, an accurate representation of the slant-path impact on the simulated radiances and ultimately the forecasts can be extremely important for such occasions.
This paper is organized as follows. Section 2 briefly describes the channel characteristics of GK2A AMI, KIM forecasts, and the RT simulation. Section 3 describes the method to get profiles along the slant path of a GEO satellite. Section 4 compares the simulated radiances and Jacobians with the traditional and the new method. The section also includes comparison results from a retrieval experiment based on an optimal estimation method for the retrieval of vertical profiles of temperature and moisture from AMI with the aid of KIM forecasts. The experiment was conducted to estimate the potential impact of taking the viewing geometry of GK2A into account in the assimilation of the KIM system. Last, section 5 summarizes and discusses the results.
2. Observations and RT simulations
a. Observations
The slant-path geometry is taken into account in the RT simulation of GK2A AMI thermal radiances in this study. AMI scans the full Earth disk every 10 min with a spatial resolution of 2 km (finer for the visible channels) and contains 10 infrared channels: one shortwave infrared channel (Ch07), three water vapor channels (Ch08, Ch09, and Ch10), four window channels (Ch11, Ch13, Ch14, and Ch15), one ozone absorption channel (Ch12), and one CO2 channel (Ch16), as described in Table 1. Currently, only 7 of 10 infrared channels, that is, the three water vapor channels and the four window channels, are assimilated in the operational KIM (version 3.6a as of January 2022) system, although the one-dimensional variational (1DVar) retrieval experiment conducted in this study utilizes the 9 infrared channels (Table 1). Except for the three water vapor channels peaking around 300–600 hPa (Fig. 1), the slant-path effect is supposed to be small for the lower peaking channels in clear-sky situations. Ch12 has two weighting function peaks, one in the stratosphere and the other near the surface, and thus slant-path modeling can have a relatively larger impact on the simulations for this channel. However, this ozone absorption channel is not utilized in the KIM system, and there is no available ozone forecast profile for the simulation. Thus, the study mainly focuses on the comparisons of the three water vapor channels before and after taking the slanted viewing geometry into account.
Information about the infrared channels (7–16) of AMI that are used in the operational KIM data assimilation system and in the 1DVar retrieval experiment for this study. Here, R = rejected, A = assimilated, and M = monitoring.
Weighting functions for the AMI infrared channels (nadir view; ocean) calculated using the U.S. Standard Atmosphere, 1976 temperature and moisture profiles at 51 pressure levels (COESA 1976).
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
b. RT simulations
For the simulation of the GK2A AMI radiances measured at 0000, 0600, 1200, and 1800 UTC from 1 to 31 May 2020, the 6-h forecast products from the KIM 3.5a global model are utilized. The spatial resolution of the KIM is 12 km, and the model top is around 80 km, with 91 vertical levels (Hong et al. 2018). The KIM forecast products utilized for the simulation include temperature and humidity profiles, temperature and mixing ratio at 2 m, surface temperature and pressure, and u–υ winds at 10 m. In addition, monthly ozone climatology (McPeters and Labow 2012) covering altitudes from 0 to 65 km is used for the study.
Note that the KIM has been operational at the KMA since April 2020, and it was not until October 2020 that the AMI measurements were utilized in the KIM system. Therefore, the KIM forecast profiles are independent from AMI observations during the study period (May 2020). In addition, the KIM system has been continually updated since its operation, and thus errors in the model forecasts in May of 2020 might be larger than those of the current operational version of the KIM. Specifically, when compared with the brightness temperature (BT) simulated with the fifth major reanalysis produced by ECMWF (ERA5), the BT simulated with the KIM 6-hourly forecasts has larger departures from the observations: for example, 0.8-K larger departures and 0.9–1.0-K larger standard deviation of departures for the high peaking water vapor channel (Ch08). Since the forecast model errors influence the accuracy of model equivalents (Bormann 2017), they should be considered in the interpretation of the slant-path effect.
For the forward simulation of clear-sky AMI BT and the calculation of Jacobians, a fast forward model, Radiative Transfer for TOVs (RTTOV) version 12.3 (Saunders et al. 2018), is utilized. The retrieval experiment utilizes the iterative physical retrieval algorithm based on 1DVar developed for the clear-sky atmospheric temperature and humidity profile from AMI thermal radiances (Lee et al. 2017). For both simulation and retrieval studies, the clear-sky AMI measurements averaged over a 3 × 3 field-of-view (FOV) area (1 FOV = 2 km) are used. To detect pixels contaminated with clouds, the GK2A level-2 cloud-mask product (Lee et al. 2018) was utilized, and an additional uniformity test was performed on the 3 × 3 FOV to filter out broken clouds.
3. Method
The viewing geometry of a GEO satellite is illustrated in Fig. 2. When a satellite observes the target point O on Earth’s surface with a satellite zenith angle θzen and azimuth angle θazi, the upwelling radiance measured at the top of the atmosphere can be calculated based on the RT equation and the required input atmospheric profiles for RT simulation is provided by model backgrounds in an NWP system. To produce accurate model equivalents via RT simulation, one will need the model fields along the line of sight (i.e., slanted profiles at S1, S2, …, Sn in the figure). So far, however, RT simulation has been conducted using the atmospheric profiles on the geolocation, that is the vertical profiles on the model grid interpolated to the geolocation point O. This will not make a noticeable difference in the simulated radiance in the lower atmosphere and/or for small θzen, but the difference increases with altitudes. The displacement Δx at each model level increases with height, and it reaches around 300 km at the highest model level, 80 km, where θzen ≈ 75°.
Illustration of satellite viewing geometry for the target point O with satellite zenith angle θzen and satellite azimuth angle θazi.
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
Unlike LEO satellites, GEO satellites are stationed above the equator and continuously observe the full-Earth disk area every 10–15 min, which makes it relatively easier to take the slant viewing geometry into account. Since the geolocation (latitude–longitude) of GEO satellites are fixed and the model top height is given, one can calculate the longest displacement (Δxlongest) for each target-point O using θzen and the height of model-top (Δz ∼ 80 km for KIM) using the relationship (Δx = Δz tanθzen). Using the property that all the points lying on the line OPn have the same azimuth angle θazi, all the model grid points above the line OPn with displacement smaller than Δxlongest can be searched. The number of grid points searched is subject to change with the horizontal resolution of the NWP model used. For example, the number of grid points reaches up to 30 at large zenith angles with a horizontal spacing of 12 km, as is the case for KIM, and the number would be much smaller with a coarser NWP model grid spacing. Bormann (2017) found that 6 profiles with a horizontal spacing of 32 km is sufficient to show the benefits of slant-path modeling and demonstrated that using a finer sampling (e.g., spatial resolution of ∼9.5 km) did not provide a benefit. A 12-km sampling is used in the present work, but it would be interesting to see whether similar results can be obtained with coarser sampling as suggested in the previous study.
To get the slant profile, that is, the atmospheric information at S1, S2, …, Sn corresponding to the model levels h1, h2, …, hn, one will need the model profiles at P1, P2, …, Pn. Since the points P1, P2, …, Pn are the projections onto the ground of the intersection between the slant path and the model levels, they do not necessarily coincide with the model grid points searched, and thus the atmospheric information at Pi corresponding to the model level hi is obtained by interpolating two neighboring model profiles (i.e., profiles at the two grid points that are closest to the point Pi).
An example of the difference between the slant and the vertical profile is shown in Fig. 3. In the figure, the thick solid black line indicates the difference between the two profiles (i.e., slanted − vertical) for temperature (Fig. 3a) and humidity (Fig. 3b), where θzen = 68° at 0000 UTC 2 May 2020. The color lines indicate the differences between the profiles at the model grids lying along the line OPn and the vertical profile at point O. In this example, the number of profiles along the line of sight is 20, and the “profile 1” (violet) is the closest from the target point O, and the “profile 20” (red) is the farthest. In the lower atmosphere, therefore, the slanted profile has values similar to the profile 1 but the values become closer to the profile 20 as the displacement increases in the upper atmosphere. As can be seen in the figure, the differences and variations between the vertical and slant profile are noticeable in the atmospheric layer high above the stratosphere for the temperature profile, indicating that the slant-path modeling would influence on the simulation of satellite channels sensitive to the upper atmosphere. As shown in Fig. 1, however, the AMI does not have a high-peaking temperature channel, and thus the variations above 100 hPa shown in the temperature profile (Fig. 3a) would not affect the simulated AMI radiances even with the slant-path modeling. Meanwhile, the variations and differences in the midtroposphere shown in the moisture profile (Fig. 3b) suggest that the slant-path modeling can improve the simulation accuracy for the AMI water vapor channels, which have the weighting function peaks between 300 and 600 hPa. This also suggests that only a small number of model profiles are sufficient for the slant-path modeling of AMI water vapor channels, because the heights affected by the slant-path modeling is below the troposphere. Figure 4 shows the displacement Δx for the 10-km model level (∼300 hPa). The displacement increases with increasing satellite zenith angles and reaches up to 37 km at θzen = 75°, which indicates that the maximum number of grid points (or model profiles) required for the slant-path calculation of AMI water vapor channels is four if the grid spacing is 12 km, as is the case for the KIM global model.
Difference between the slanted and vertical profile (thick solid black line) for (a) temperature and (b) mixing ratio at θzen = 68°. The colored lines indicate the difference between the profiles at the model grids lying along the slant plane and the vertical profile at point O.
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
Displacement for the 10-km model level with varying satellite zenith angles from 0° to 75° within the AMI full-disk domain.
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
Analyzing the difference between the slant and vertical profiles below 100 hPa, binned into 75 satellite zenith angles from 0° to 75°, reveals that the differences are close to zero with maximum difference ranging up to 0.23 K at around 100 hPa for temperature and up to 0.06 g kg−1 at around 600 hPa for moisture over the month of May 2020. The standard deviation (STD) of the differences for each 1° bin was also examined and the time series of 6-hourly STDs at four levels of interest are displayed in Fig. 5 for the month of May 2020. The dots in different colors indicate the STD for different satellite zenith angles between 0° (bluish) and 75° (reddish) at the level. From the figures, it is obvious that the STDs in the profile difference increase with increasing satellite zenith angles at all levels for both temperature (Fig. 5a) and moisture (Fig. 5b). The STDs in temperature difference increase with increasing altitude, showing the maximum (0.59 K) at 100 hPa. Hence, as mentioned above, sensors with high-peaking temperature sounding channels may see a significant improvement in the forward-model calculation if slant path is considered during the RT simulation. Unlike temperature, moisture profile differences (Fig. 5b) exhibit the largest STD at around 700 hPa and the magnitude of STD rapidly decreases with altitude. This can be viewed as mixed results of the rapid upward decline in the water vapor abundance and the small displacement in the azimuthal plane in the low atmosphere.
STD of difference between the newly calculated slanted profile and the original vertical profile for (a) temperature and (b) mixing ratio for May 2020. The differences are binned into 75 satellite zenith angles (from 0° to 75°) for each level and for the UTC time.
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
4. Results
a. Impact on the RT simulations
Simulated clear-sky AMI BTs with the conventional approach, where the vertical profile at the target point is used as the input for the RT simulation, are compared with the simulated BTs with the new approach, where the slant profile is used as the RTM input. The mean BT differences between the two approaches for the period 1–31 May 2020 are close to zero, with relatively large differences in the three water vapor absorption channels (Ch08, Ch09, and Ch10). However, the differences can be sizeable in the regions with large satellite zenith angles or with large horizontal water vapor variabilities in the low atmosphere. Figure 6 displays locations with absolute differences of the simulated BT that are larger than 3 K for the three water vapor channels from 1 to 10 May 2020. All three channels show similar distributions of BT difference, and particularly Ch08 and Ch09 display similar number of occurrences of large BT difference. Over the ocean, the events are mainly found at mid- to high latitudes in the North and South Pacific Ocean, and on land, the most significant differences are observed over India along the Himalayan border on the north and also from the central to the eastern parts of India as shown in the figures. A closer look at the KIM moisture field at 300 and 500 hPa (the representative level where the peak of the weighting function of Ch08 and Ch09 locates, respectively) revealed that the water vapor variabilities along the line of sight are relatively large over those regions in India in May 2020. To be specific, the differences in the model moisture field at 300 hPa between the vertical profile and the slanted profile are particularly high in India relative to other land areas, and the spatial distribution of the moisture difference (not shown) looks very similar to that of the BT difference for Ch08 (Fig. 6a). Similar pattern was found between the difference in KIM moisture field at 500 hPa and the BT difference for Ch09 (Fig. 6b). Chen and Liu (2016) also reported the large variations of precipitable water vapor in the south foothill of Himalayas, attributing the reason to the monsoon climate in these regions. Another factor that may have contributed to the large BT differences shown over India is the air mass flow along the Himalayas. The general orientation of Himalayan ranges, which runs in a northwest–southeast direction for about 2500 km, coincides with the direction of the line of sight of the GK2A satellite in the area. This geographical factor may have led to more water vapor variabilities along the slant path over India than in other land areas. Here it should be noted that the results can be model specific. Using a different NWP model, e.g., ECMWF with slightly different water vapor fields over India and with different spatial resolution, may exhibit different slant-path modeling effect for the same region.
Spatial distribution of absolute difference of BT, simulated with the two different methods, larger than 3 K for the three water vapor channels—(a) Ch08, (b) Ch09, and (c) Ch10—covering the period 1–10 May 2020.
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
Figure 7 displays the normalized difference in STD of AMI observations minus backgrounds (O − B) between the two methods, i.e., STD of O − B without the slant-path interpolation (σnoSlant) minus STD of O − B with the interpolation (σslant) normalized by σnoSlant. The STD differences are calculated for the nine infrared channels of AMI over the land (solid lines) and sea (dash–dotted lines) separately, averaged in each 5° satellite zenith angle bin for May 2020. The number of observations in each zenith angle bin is also presented in the right side of the figure. The relative reductions in the STD increase with increasing satellite zenith angles in general and are most noticeable in the three water vapor channels (Ch08, Ch09, and Ch10), which are sensitive to the mid- to upper-tropospheric atmosphere. At higher satellite zenith angles over land, the reduction tends to decrease with increasing angles from zenith angles of around 55°. This is mainly attributed to the dry atmospheric conditions between 300 and 500 hPa combined with low spatial variations of water vapor along the azimuthal angle over regions like eastern Kazakhstan, western Mongolia, and western India. All those regions are located at zenith angles greater than 55°. The mid- to upper-tropospheric moisture field over those regions tends to be much drier in May 2020. Smaller reductions in the STD or even negative reductions are frequently observed in such dry air conditions especially when there are little spatial variations in the moisture field along the azimuthal direction. The monthly mean relative reductions in O − B STD of the three water vapor channels between 40° and 60° satellite zenith angles are 1.3%, 1.5%, and 1.3% over the ocean and 2.1%, 2.1%, and 1.9% on land, respectively. Previous studies with LEO satellites have also shown that the slant-path effect is mainly confined to the regions with large satellite zenith angles or to the edges of the LEO swath and the largest reductions are found in the high-peaking channels such as the temperature sounding channel 9 of ATMS. Bormann (2017) reported 7% of global reduction in O − B STD of ATMS channel 9, peaking around the tropopause, but the reduction is much smaller in the channels with lower peaks such as the ATMS humidity sounding channel 22 sensitive to the upper troposphere (lower than the peak of channel 9), having 1.5% reduction over the period 25 January–24 February 2015.
Normalized difference in STD of O − B (%) between the two approaches over the land (solid lines) and sea (dash–dotted lines) averaged at 5° satellite zenith angle bins for May 2020 for the 9 infrared channels of AMI, displayed in different colors. The number of observations in each 5° bin over the sea and land is displayed on the right, with “e+0X” indicating that the leading number should be multiplied by 10X.
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
b. Impact on Jacobians
Changes in the temperature and moisture Jacobians after taking the slanted viewing geometry into account are also investigated. The Jacobian is the change in BT at the top of the atmosphere with respect to the change in the atmospheric parameters (temperatures and the concentration of trace gases) at a specific level. Since the shape and magnitude of the Jacobian determine the shape and magnitude of the analysis increment in data assimilation, the evaluation of the Jacobian is essential (Garand et al. 2001) for the future application of this method in the data assimilation system of KIM. The one-to-one comparison of the two types of Jacobians before and after taking the slant-path geometry into account reveals that the differences are noticeable in the temperature Jacobians for the three water vapor absorption channels. Figure 8 presents two extreme cases of temperature Jacobian comparisons that exhibit noticeable differences from each other on land (Fig. 8a) and over the ocean (Fig. 8b), representing possible results of particularly strong spatial humidity gradients combined with a large zenith angle. Both cases show that not only the shape but also the magnitude and height of the peak of the Jacobians change if the slant-path interpolation is used (displayed in purple). Particularly, the upward/downward moving of the peak of the Jacobians in the three water vapor channels represents the varying humidity conditions along the line of sight. This is because a wetter profile shifts the Jacobian higher for water vapor channels whereas a drier profile will shift it lower, so the Jacobian changes in the ways shown in the figure if the slant-path interpolation results in a wetter or drier profile.
One-to-one comparison of temperature Jacobians calculated with the vertical (black) and slanted (purple) background profile for (left) Ch08, (center) Ch09, and (right) Ch10 at 0000 UTC 2 May 2020 (a) at 22.07°N, 86.34°E (Case 1: land, θzen = 53.3°; surface skin temperature of 293.48 K; surface pressure of 969.23 hPa) and (b) at 18.996°S, 173.38°W (Case 2: sea, θzen = 68.4°; sea surface temperature of 301.44 K; surface pressure of 1011.76 hPa).
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
In addition, the differences between the two Jacobians (i.e., calculated with the atmospheric profiles with/without the slant-path interpolation) are compared at fixed satellite zenith angles within the AMI full-disk area at 0000 UTC 2 May 2020. To analyze the satellite-zenith-angle contribution to the slant-path effect in AMI infrared channels, the STD of the Jacobian differences are calculated for 8 different satellite-zenith-angle ranges (plotted in 8 different colors) and the results are presented in Fig. 9. For both temperature (Figs. 9a–c) and moisture (Figs. 9d–f) Jacobians, the slant-path effects for the three water vapor channels are mostly observed in the troposphere between around 200–800 hPa, although the absolute magnitudes of STD are not large overall. Note that the unit of STD for moisture Jacobians is “K,” meaning that the moisture Jacobians are multiplied by the water vapor amount at each model level. In general, the slant-path effects increase with increasing satellite zenith angles for both temperature and moisture Jacobians, and the moisture variabilities along a slant path around 600–700 hPa influence the most on the simulated radiances for Ch08 and Ch09.
The STD of (a)–(c) temperature and (d)–(f) moisture Jacobian differences (slant − vertical) with varying satellite zenith angles for (left) Ch08, (center) Ch09, and (right) Ch10 calculated at 0000 UTC 2 May 2020. Each color line represents STD calculated for a specific satellite zenith angle range as displayed in the top-right corner of (a) and (d).
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
c. Impact on the retrieval
As a preliminary step for using the slant-path interpolation in the data assimilation system in KMA in the near future, this study examines the slant-path effect on the retrieval by utilizing a stand-alone 1DVar algorithm for the retrieval of vertical temperature and moisture profiles from the GK2A AMI 9 infrared channels (refer to Table 1). The algorithm is based on the operational Atmospheric AMI Profile (AAP) retrieval algorithm (Lee et al. 2017) but modified to use the forecasts from KIM instead of UM as the first-guess and to apply the slant-path geometry. The algorithm iteratively adjusts the first-guess profiles, i.e., KIM 6-h forecasts, until it finds the optimal atmospheric state that minimizes the cost function. The retrieval was performed twice, one with the slant-path calculation and one without it, over the period 1–31 May 2020 at 6-h interval for clear-sky conditions. The results show that the slant-path calculation leads to faster convergence by providing improved first-guess profiles. For example, the number of retrievals with the first-guess profile (i.e., the first-guess profile is good enough and meets the convergence criteria even before the iterative retrieval process starts) is 10% or 8% larger with the slant-path modeling than the one without it in the latitudinal band of 60°–75°N or 45°–60°N, respectively.
The retrieval accuracy has been evaluated by comparing the retrieved vertical temperature and moisture profiles with the radiosonde observations (raob) at 77 stations, mainly including the stations with the Vaisala RS92 radiosonde within the GK2A AMI full disk area. To reduce the influence of radiosonde drift on the validation, the retrieved profiles collocated with the radiosonde are spatially averaged within the 100 km × 100 km grid box along the drift direction. Figure 10 shows the mean difference between the retrieved profiles and raob (bias) and the STD of the difference covering the period 1 to 31 May 2020 for temperature (Fig. 10a) and mixing ratio (Fig. 10b). The number of matches used for the validation is 1061. The vertical distribution of bias and STD (figures on the left side) exhibit very little difference between the two approaches. For a better look into the STD difference, the normalized difference in the STD, that is, 100 × (STDnoslant − STDslant)/STDnoslant, is presented together on the right side of the figure. Both temperature and mixing ratio show that the STD is reduced with the slant-path modeling below the 300 hPa level, particularly between 300 and 600 hPa with mean relative reduction of 4.5% for temperature and 1.04% for mixing ratio. In the comparison of the first-guess profiles (KIM forecast) and raob (not shown), the statistics has revealed that the normalized difference in the STD is 4.5% for temperature and 0.8% for mixing ratio. The little difference between the two statistics (i.e., first-guess vs raob and retrieval vs raob) indicates that the first guess has already been improved when the slanted-viewing geometry is taken into account; that is, slanted profiles agree better with the raob.
(left) Mean bias (dash–dotted lines) and STD (solid lines) of the retrieved (a) temperature and (b) mixing ratio compared with raob with the slant-path modeling (blue) and without it (black) in May 2020, and (right) the associated normalized difference in the STD [(STDnoslant − STDslant)/STDnoslant] (%) between the two methods.
Citation: Monthly Weather Review 151, 4; 10.1175/MWR-D-22-0080.1
The validation, however, has been conducted for only short period of time (May 2020) and only over land, and thus a more extensive validation including the sea will be needed to assure the positive slant-path effect on the retrieval from geostationary satellites.
5. Conclusions
For the future application in the operational KIM system in KMA, this study investigated the effect of taking the satellite viewing geometry into account for the simulation of GK2A AMI infrared channels. Using the 6-hourly forecasts of KIM, the upwelling radiances (converted to BT) and the Jacobians are calculated with the RTTOV version 12.3 over the period 1–31 May 2020. In addition, a 1DVar retrieval system was utilized to see the slant-path effect on the retrieval of temperature and moisture profiles from the AMI infrared channels. The results show that taking the satellite viewing geometry into account in the RT simulation can have a significant influence on the simulated BT for the water vapor absorption channels of AMI peaking between 300 and 600 hPa. The influence can be particularly significant in regions like India, over which the moisture field of KIM exhibits large variabilities along the line of sight in the mid troposphere. Over the ocean, the largest impact is observed in both Ch08 and Ch09 at high latitudes, particularly over the North and South Pacific Ocean. The one-to-one comparison of the Jacobians calculated with the two different background profiles revealed that not only the magnitude but also the shape and the peak height of the Jacobians are influenced by the treatment of slanted viewing geometry. The 1DVar retrieval experiment also shows that taking the viewing geometry of satellites into account contributes to a faster convergence during the retrieval process particularly at the latitudinal band between 45° and 75°N. The evaluation of retrieval accuracy using radiosonde observations also reveals that the new method helps slightly reduce the STD for the retrieved moisture profiles between 300 and 600 hPa. Currently, the slant-path interpolation is not applied to any satellite utilized in the KIM system, and the method and analyses in this study can be used to improve the observation operator for GEO satellites utilized in KIM and also for LEO satellites, the latter is expected to result in more sizeable improvements in the forecast system with the slant-path modeling.
Acknowledgments.
This work is supported by project KMA2018-00721 funded by the Development of Numerical Weather Prediction and Data Application Techniques in the Korea Meteorological Administration (KMA). The authors acknowledge the use of the KIM forecast data provided by the Numerical Modeling Center of KMA.
Data availability statement.
The GK2A AMI Level1B data can be downloaded from the KMA/NMSC Data service system (http://datasvc.nmsc.kma.go.kr/datasvc/html/main/main.do?lang=en). RTTOV can be downloaded from the NWP SAF website (https://nwp-saf.eumetsat.int/site/software/rttov/download/#Software). Due to confidentiality agreements, KIM forecast data can only be made available to bona fide researchers subject to a non-disclosure agreement. Details of the data and how to request access are available via telephone from +82-42-481-7508 at the Numerical Modeling Center at KMA.
REFERENCES
Bormann, N., 2017: Slant path radiative transfer for the assimilation of sounder radiances. Tellus, 69A, 1272779, https://doi.org/10.1080/16000870.2016.1272779.
Bormann, N., and S. B. Healy, 2006: A fast radiative-transfer model for the assimilation of MIPAS limb radiances: Accounting for horizontal gradients. Quart. J. Roy. Meteor. Soc., 132, 2357–2376, https://doi.org/10.1256/qj.05.160.
Burrows, C., 2018: Assimilation of radiance observations from geostationary satellites: First year report. EUMETSAT/ECMWF Fellowship Programme Research Rep. 47, ECMWF, 51 pp., https://www.ecmwf.int/sites/default/files/elibrary/2018/18551-assimilation-radiance-observations-geostationary-satellites-first-year-report.pdf.
Burrows, C., 2019: Assimilation of radiance observations from geostationary satellites: Second year report. EUMETSAT/ECMWF Fellowship Programme Research Rep. 51, ECMWF, 50 pp., https://www.ecmwf.int/en/elibrary/81120-assimilation-radiance-observations-geostationary-satellites-second-year-report.
Chen, B., and Z. Liu, 2016: Global water vapor variability and trend from the latest 36 year (1979 to 2014) data of ECMWF and NCEP reanalyses, radiosonde, GPS, and microwave satellite. J. Geophys. Res. Atmos., 121, 11 442–11 462, https://doi.org/10.1002/2016JD024917.
COESA, 1976: U.S. Standard Atmosphere, 1976. NOAA, 227 pp.
Eyre, J. R., 2008: Progress achieved on assimilation of satellite data in numerical weather prediction over the last 30 years. Proc. ECMWF Seminar on Recent Developments in the Use of Satellite Observations in Numerical Weather Prediction, Shinfield Park, Reading, United Kingdom, ECMWF, 28 pp., https://www.ecmwf.int/en/elibrary/9341-progress-achieved-assimilation-satellite-data-numerical-weather-prediction-over.
Garand, L., and Coauthors, 2001: Radiance and Jacobian intercomparison of radiative transfer models applied to HIRS and AMSU channels. J. Geophys. Res., 106, 24 017–24 031, https://doi.org/10.1029/2000JD000184.
Healy, S. B., J. R. Eyre, M. Hamrud, and J.-N. Thépaut, 2007: Assimilating GPS radio occultation measurements with two-dimensional bending angle observation operators. Quart. J. Roy. Meteor. Soc., 133, 1213–1227, https://doi.org/10.1002/qj.63.
Hong, S.-Y., and Coauthors, 2018: The Korean Integrated Model (KIM) system for global weather forecasting. Asia-Pac. J. Atmos. Sci., 54, 267–292, https://doi.org/10.1007/s13143-018-0028-9.
Joiner, J., and P. Poli, 2005: Note on the effect of horizontal gradients for nadir-viewing microwave and infrared sounders. Quart. J. Roy. Meteor. Soc., 131, 1783–1792, https://doi.org/10.1256/qj.04.125.
Joo, S., J. Eyre, and R. Marriott, 2013: The impact of MetOp and other satellite data within the Met Office global NWP system using an adjoint-based sensitivity method. Mon. Wea. Rev., 141, 3331–3342, https://doi.org/10.1175/MWR-D-12-00232.1.
Kim, D., M. Gu, T.-H. Oh, E.-K. Kim, and H.-J. Yang, 2021: Introduction of the Advanced Meteorological Imager of Geo-Kompsat-2A: In-orbit tests and performance validation. Remote Sens., 13, 1303, https://doi.org/10.3390/rs13071303.
Lee, B.-I., S.-R. Chung, H.-A. Kim, and S. Baek, 2018: Development of Cloud Detection Algorithm for Chollian-2A Satellite: Technical notes (in Korean and English). National Meteorological Satellite Center Algorithm Theoretical Basis Document: Version 4.0, 72 pp., https://book.kma.go.kr/search/DetailView.ax?sid=1&cid=32805.
Lee, S. J., and M.-H. Ahn, 2021: Synergistic benefits of intercomparison between simulated and measured radiances of imagers onboard geostationary satellites. IEEE Trans. Geosci. Remote Sens., 59, 10 725–10 737, https://doi.org/10.1109/TGRS.2021.3054030.
Lee, S. J., M.-H. Ahn, and S.-R. Chung, 2017: Atmospheric profile retrieval algorithm for next generation geostationary satellite of Korea and its application to the advanced Himawari imager. Remote Sens., 9, 1294, https://doi.org/10.3390/rs9121294.
Lu, J., T. Feng, J. Li, Z. Cai, X. Xu, L. Li, and J. Li, 2019: Impact of assimilating Himawari‐8‐derived layered precipitable water with varying cumulus and microphysics parameterization schemes on the simulation of Typhoon Hato. J. Geophys. Res. Atmos., 124, 3050–3071, https://doi.org/10.1029/2018JD029364.
McNally, A., 2014: The impact of satellite data on NWP. Seminar on the Use of Satellite Observations in NWP, Shinfield Park, Reading, United Kingdom, ECMWF, 10 pp., https://www.ecmwf.int/sites/default/files/elibrary/2015/11061-impact-satellite-data-nwp.pdf.
McPeters, R. D., and G. J. Labow, 2012: Climatology 2011: An MLS and sonde derived ozone climatology for satellite retrieval algorithms. J. Geophys. Res., 117, D10303, https://doi.org/10.1029/2011JD017006.
Poli, P., J. Joiner, and D. Lacroix, 2005: Application of radiative transfer to slanted line-of-sight geometry and evaluation with AIRS data. Proc. 14th Int. TOVS Study Conf., Beijing, China, UW/CIMSS, 6.4, 5 pp., http://cimss.ssec.wisc.edu/itwg/itsc/itsc14/proceedings/6_4_Poli.pdf.
Saunders, R., T. A. Blackmore, B. Candy, P. N. Francis, and T. J. Hewison, 2013: Monitoring satellite radiance biases using NWP models. IEEE Trans. Geosci. Remote Sens., 51, 1124–1138, https://doi.org/10.1109/TGRS.2012.2229283.
Saunders, R., and Coauthors, 2018: An update on the RTTOV fast radiative transfer model (currently at version 12). Geosci. Model Dev., 11, 2717–2737, https://doi.org/10.5194/gmd-11-2717-2018.
Shahabadi, M. B., J. Aparicio, and L. Garand, 2018: Impact of slant-path radiative transfer in the simulation and assimilation of satellite radiances in Environment Canada’s weather forecast system. Mon. Wea. Rev., 146, 4357–4372, https://doi.org/10.1175/MWR-D-18-0126.1.
Shahabadi, M. B., M. Buehner, J. Aparicio, and L. Garand, 2020: Implementation of slant-path radiative transfer in Environment Canada’s global deterministic weather prediction system. Mon. Wea. Rev., 148, 4231–4245, https://doi.org/10.1175/MWR-D-20-0060.1.
Wang, P., J. Li, B. Lu, T. J. Schmit, J. Lu, Y.-K. Lee, J. Li, and Z. Liu, 2018: Impact of moisture information from advanced Himawari imager measurements on heavy precipitation forecasts in a regional NWP model. J. Geophys. Res. Atmos., 123, 6022–6038, https://doi.org/10.1029/2017JD028012.
