COSMIC-2 RO Profile Ending at PBL Top with Strong Vertical Gradient of Refractivity

Abstract

The Formosa Satellite-7/Constellation Observing System for Meteorology, Ionosphere, and Climate-2 (Satellite-7/COSMIC-2), which was successfully launched on 25 June 2019, provides dense radio occultation (RO) observations over the tropics and subtropics. This study examines the RO-observed lowest altitude and its possible relationship to refractivity gradients and planetary boundary layer (PBL) heights. COSMIC-2 RO data over the Southeast Pacific region (SEP) and the South-Central Pacific (SCP) from August 2020 are employed to determine their RO-observed lowest altitudes, and the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 reanalysis data are used to obtain the gradients of refractivity. Results show that there are no ray perigees below the PBL top when the vertical gradient of $N-N(r)$ is strong (<−65 N-unit km⁻¹), where $N(r)$ represents the vertical profile of the spherically symmetric refractivity. Significantly strong local vertical gradients due to atmospheric ducting occur more frequently over the SEP than the SCP areas. For some cases, a strong local horizontal gradient of refractivity in the tangent direction of a ray near its perigee point can also limit the RO profile from going further below even when the vertical gradient of $N-N(r)$ is relatively weak. Fortunately, only about 0.6% COSMIC-2 RO profiles are unaffected by the above factors but cannot observe below 2-km altitude.

1. Introduction

The application of the Global Navigation Satellite System (GNSS) radio occultation (RO) sounding technique in meteorology began with the GPS/Meteorology (GPS/MET) experiment in 1995 [1]. Many GNSS RO missions have successfully been launched since then, including the Challenging Mini-Satellite Payload (CHAMP) [2], the Satelite de Aplicaciones Cientificas-C (SAC-C) [3], the Gravity Recovery and Climate Experiment (GRACE) [4], the Formosa Satellite-3/Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) [5], and the Formosa Satellite-7/COSMIC-2 [6]. The RO sounding technique measures atmospheric temperature and humidity information over the globe with high accuracy, precision, and vertical resolution in all weather conditions [7]. GNSS RO observations complement the satellite microwave and infrared measurements for numerical weather prediction (NWP). Many studies showed the values of GPS RO data for research on tropical cyclones [8,9,10], planetary boundary layer detection [11,12], and NWP [13,14,15,16].

The application of RO observations in the lower troposphere has always been challenging. Considering the large biases of RO observations in the lower troposphere, particularly in regions with large refractivity gradients, the European Met Office only operationally assimilated the RO data above 4-km altitude at the early stages of RO data assimilation to avoid biases. Since April 2008, all RO observations were assimilated by assuming large observation errors in the lower troposphere, and the corresponding experiments showed a small positive impact as a result of the inclusion of low-level RO data [17]. NCEP has also operationally assimilated COSMIC RO data since 1 May 2007, in which the RO observations below 4 km were assimilated if they passed quality control checks and obtained a significant improvement on the anomaly correlation scores and tropical wind forecasts [18,19].

The COSMIC was the first GNSS RO observing constellation system and delivered near-real-time RO data to operational weather centers all over the world since 2006 [5,20]. Different from its previous GNSS RO missions, COSMIC consists of six low-earth orbiting (LEO) satellites that move in six different orbits with the same inclination of 72°. COSMIC can provide about 3500 RO profiles globally. COSMIC-2 is a follow-on mission of COSMIC, and it was successfully launched on 25 June 2019. Like COSMIC, COSMIC-2 also contains six LEO satellites, whose orbital inclination is twice smaller than that of COSMIC (i.e., 24°). More than 5000 RO profiles are provided daily by COSMIC-2 within the latitudinal band (45°S, 45°N). Each COSMIC-2 LEO satellite carries an advanced RO receiver, which is named the Tri-GNSS Radio Occultation System (TGRS), with a high-gain beam-forming RO antenna that can produce a higher signal-to-ratio (SNR) than any RO receiver of previous RO missions. Therefore, COSMIC-2 delivers RO data of better precision and lower penetration into the troposphere [6]. About 50% of COSMIC-2 RO profiles can detect below 200 m above the surface. These features make COSMIC-2 data of unique value for the investigation of weather systems in the tropics and subtropics, especially in deep oceans.

RO observations can provide high-resolution vertical distributions of atmospheric refractivity, temperature, and moisture under almost all atmospheric weather conditions. The near-surface atmospheric profiling information is precious for studying tropical convection and hurricane genesis and development [21,22,23]. The sea-surface warm anomalies and the wet atmosphere in the lower troposphere are critical conditions for triggering strong convection [24,25,26]. However, in some specific cases, such as when the refractivity gradient is strong, the amplitude of RO signals is reduced because of its defocusing or undergoing strong fluctuation such that the RO signals cannot be distinguished from noise. This is most likely to happen in the atmospheric ducts which usually appear near the PBL and are accompanied by large negative gradients of atmospheric refractivity. Sokolovskiy et al. [27] utilized the RO observations from one of the COSMIC receivers which was configured to track down to the height of straight line (HSL) of −350 km in the tropics (30°S–30°N) on 5–6 October 2010 to investigate the RO signals very deep in the earth’s shadow, where HSL is defined as the height of a straight line that connects GPS and occulted LEO from the earth’s surface. Through the wave optics modeling of these RO signals, it was found that the atmospheric ducts confined the RO signals in the ducting areas, especially for the ducts of wide horizontal expansion. Since the refractivity gradients in the tropical lower troposphere are strong due to the impact of water vapor and its complicated small-scale structures, the fluctuation of the RO signals is quite strong, and tracking RO signals becomes more difficult.

Marquardt et al. [28] emphasized the importance of understanding how deep the GPS RO signals can be observed. When processing COSMIC-2 RO data, there are some COSMIC-2 RO profiles which cannot penetrate below 1 to 2 km above the surface. Since the planetary boundary layer (PBL) of the atmosphere is about 1 to 2 km and the PBL top is often characterized by a strong temperature inversion and a sharp moisture gradient, a strong vertical gradient of atmospheric refractivity can appear at the PBL top. A relationship of high lowest altitudes of RO profiles to a strong vertical gradient of atmospheric refractivity is thus sought in this study. Two oceanic areas, one over the Southeast Pacific region (SEP) and the other over the South-Central Pacific (SCP), are selected. It is well known that the SEP area is characterized by a persistent boundary layer cloud deck [29] and has a frequent occurrence of ducting [30,31] due to large negative refractivity gradient ($dN/dr$≤ −157 N-unit km⁻¹) [32], where $r$ is the distance from the perigee point of the RO ray path to the earth’s center. The persistent cloud deck over SEP is formed by both the large-scale subsidence in subtropical oceans and coastal upwelling-induced cold sea surface temperature [33,34]. Near the cloud deck, often accompanied by a sudden change in moisture and temperature, atmospheric ducts can easily occur. Some studies have used years of RO observations to investigate the atmospheric ducting events and found that RO refractivity over the SEP has a large bias and the ducts occur most frequently over the SEP [31,35]. These situations that are prone to occur in the SEP area are proven to easily affect the downward penetration of RO rays. Based on the above considerations, the SEP is selected as an area suitable for finding a statistical relationship between the lowest RO observing heights and the vertical levels of strong gradients of refractivity. The SCP area is a mid-ocean area located to the west of the SEP and will be used as a comparison to the SEP and to see if similar conclusions are valid in an area other than the SEP.

This paper is organized as follows. Section 2 and Section 3 describe the data and the methodology, respectively. Results of the relationships of RO lowest heights to the vertical and horizontal gradients of refractivity and clouds are given in Section 4. Section 5 presents the discussion. The conclusion and future work are provided in Section 6.

2. Data Description

The COSMIC-2 near-real-time (NRT) level 2 data (UCAR COSMIC Program) [36] are used in this study. These data were operationally processed by the COSMIC Data Analysis and Archive Center (CDAAC). The vertical profiles of bending angle and refractivity observations are provided at a vertical interval of 20 m, and the 1D-Var retrieved vertical profiles of temperature and specific humidity are given at an interval of 100 m. The observed bending angle profiles are used for obtaining the lowest height of RO observations. The COSMIC-2 RO data are available on the website www.cosmic.ucar.edu (accessed on 1 April 2022).

Figure 1 shows the monthly variations of data count of COSMIC-2 RO measurements in 2020 over the SEP area. The data count is the largest in August (about 8000 profiles), and the smallest in June (about 6000 profiles). The data count gradually decreases from the 5-km altitude to the surface (see Figure 1a) because some RO profiles end at certain heights. The vertical variations of RO data count are similar among different months. The data counts at 0.2, 3, and 8-km altitudes are shown in Figure 1b in different months of 2020. It can be seen that more than 50% of the COSMIC-2 RO observations can detect below 200 m, and there is a small monthly variation of RO data count at 0.2, 3, and 8 km. We employed the RO data in August 2020 for this study.

The ECMWF ERA5 reanalysis is used to obtain the atmospheric conditions near the lowest height of the COSMIC-2 RO profiles. The ERA5 reanalysis provides hourly data of global atmospheric temperature, humidity, cloud liquid/ice water content (LWC/IWC), cloud liquid/ice water path (LWP/IWP), and PBL top height at the horizontal resolution of 0.25 × 0.25°. Besides conventional observations and satellite radiance measurements, GPS RO observations were also assimilated in the ERA5 reanalysis [37]. For vertical profiles of the atmospheric variables, the ERA5 reanalysis provided to the public have a total of 37 levels from 1000 to 1 hPa. Figure 2 shows the monthly mean LWP distribution within the latitudinal band (60°S, 60°N) in August 2020 from the ERA5 reanalysis. The Intertropical Convergence Zone (ITCZ) is characterized by a zone of maximum LWP near the equator in the west and middle Pacific oceans in the Northern Hemisphere. It is seen that the SEP and SCP areas, which are defined as the regions (115°W–70°W, 40°S–0°) and (170°W–130°W, 40°S–0°), respectively, are located to the south of the ITCZ.

3. Methodology

3.1. Refractivity and Bending Angle Calculation

In order to use the ERA5 reanalysis to analyze the atmospheric refractivity distribution near the lowest height of the COSMIC-2 RO profiles that did not reach the surface, the refractivity should be calculated from ERA5 reanalysis. We use the following equation of refractivity provided by Smith and Weintraub [38]:

$$N=N_d+N_w=k_1\frac{P_d}{T}+k_2\frac{e}{T}+k_3\frac{e}{T^2}$$

(1)

where $k_1=77.6890, k_2=71.2952, k_3=375463$, and $e$ represents the water vapor pressure and can be calculated by:

$$e\approx \frac{qP}{0.622+0.378q}$$

(2)

Based on Equations (1) and (2), the ERA5 refractivity can be calculated from the ERA5 temperature ($T$) and specific humidity ($q$) at the 37 pressure levels. In the following, the first term on the right-hand-side of Equation (1) will be called the dry refractivity, and the second two terms on the right-hand-side of Equation (1) is called the wet refractivity. The vertical distributions of the dry and wet refractivity in the atmosphere will be shown and analyzed below.

Once the refractivity is calculated from the ERA5 reanalysis, the bending angle ($\alpha$) can be calculated by the Abel transformation [39]:

$$\alpha (a)=-2a{\displaystyle {\mathsf{\int }}_a^{\infty }\frac{1}{\sqrt{{\mu }^2-a^2}}d\ln n}$$

(3)

where $a$ represents the impact parameter, $n=1+{10}^{-6}N$ is the index of refractivity, $r$ is the geocentric radial distance, and $\mu =nr$. The refractivity and bending angle calculated from the ERA5 reanalysis will be used for a comparison with the COSMIC-2 RO profiles.

3.2. The Spherically Symmetric Refractivity Profile

Similar to the atmospheric density, the atmospheric refractivity decreases exponentially with altitude. A new variable $N(r)$, which is the spherically symmetric refractivity, is calculated to depict a reference change of refractivity with altitude. It is defined by the global mean refractivity at the same distance ($r$) from the perigee point of RO observations ($N(i(r))$) to the earth’s center:

$$\overline{N}(r)=\overline{N(i(r))}$$

(4)

where “$i(r)$” represents the ith observation point whose distance to the earth’s center is $r$. In the following, we will subtract this spherically symmetric refractivity from the total refractivity ($N$) to examine the vertical gradients of refractivity. In other words, the vertical gradient of $N-N(r)$, instead of $N$, can be compared among different altitudes.

4. Results

First, we will show two typical COSMIC-2 RO profiles, one profile (RO1) for which the GPS signals cannot propagate below the altitude where the large negative vertical gradient of refractivity exits, and the other profile (RO2) that occurred near the RO1 in both space and time but reached the surface. The ERA5 reanalysis is used to display the atmospheric conditions surrounding RO1 and RO2. Differences between RO1 and RO2 will be analyzed. Then, we will show the vertical profile of $N(r)$ that represents the reference variation of refractivity with altitude. Finally, a statistical relationship between the vertical gradient of $N-N(r)$ and the lowest RO observing height will be sought.

4.1. Two Typical Cases

The vertical variations of the bending angle, refractivity, temperature, and specific humidity at the locations of two chosen RO profiles are compared between the COSMIC-2 observations and the ERA5 reanalysis in Figure 3. The RO1 profile is a typical RO affected by a strong local vertical gradient of refractivity. Since the vertical resolution of the ERA5 reanalysis (shown by the ticks on the right y-axis) is much coarser than that of the COSMIC-2 RO observations, the ERA5 vertical profiles (dashed curves) seem smoother than the RO profiles (solid curve) for all four variables shown in the figure. The ERA5 reanalysis does not include the bending angle, so the dashed curves in Figure 3a are calculated by the Abel transformation based on the ERA5 refractivity profiles (see Section 3). The COSMIC-2 bending angle observations differ greatly from the ERA5 reanalysis at the heights where the observed bending angle changes greatly, such as at the impact height of 3 km and 5.5 km in Figure 3a. As seen in Figure 3b–d, the vertical variations of refractivity, temperature, and specific humidity between RO observations and the ERA5 reanalysis are much more consistent from 1 to 8 km than the lower troposphere (e.g., below 2 km). Furthermore, when a COSMIC-2 RO profile did not reach the surface, the ERA5 reanalysis illustrated the atmospheric conditions near and below the RO lowest height.

The RO1 is located near a shortwave trough of the geopotential height at 900 hPa, which is the pressure level closest to the lowest height of RO1 (Figure 4a). The RO observing time is 0637 UTC 11 August 2020. The refractivity at 900 hPa along the tangent direction of the RO1 ray path near the perigee point changes little. However, the vertical gradients of refractivity near the RO1 perigee point are strong. Figure 4b shows the vertical cross-section along the ray tangent direction at RO1 perigee point for total refractivity (color shaded), wet refractivity (solid black curve), and dry refractivity (dashed black curve). The distribution of the dry refractivity is nearly uniformly distributed in the vertical, while the wet refractivity varies greatly, and the contours are denser near the lowest height of RO1 than elsewhere, representing a large vertical gradient.

We may examine the horizontal and vertical distributions of the vertical gradient of the wet refractivity (Figure 5), which corresponds to Figure 4. It can be seen that there is a local area of strong vertical gradients of wet refractivity in the southwest part of a short-wave trough where the RO1 was located. If we examine the vertical variation of the vertical gradient of wet refractivity, a ducting band of strong negative vertical gradient of refractivity of more than −157 N-unit km⁻¹ appeared right below the lowest observing height of the RO1, which could limit the RO ray perigees below these altitudes. Therefore, RO1 is a typical RO profile for which the GPS radio signals cannot be distinguished at the GPS receiver due to the presence of a strong vertical gradient of refractivity in the low troposphere.

The second example, RO2, is a COSMIC profile which reached the surface. The RO2 was located to the west of the RO1 and observed at 0719 UTC 11 August 2020, which is only 40 min later than the occurrence of the RO1. The RO2 was located within a ridge area where the vertical gradients of refractivity are small (figure omitted). Thus, the RO2 profile can extend all the way from the top of the atmosphere to the earth’s surface.

4.2. Vertical Profiles of Spherically Symmetric Refractivity

In order to better compare the local variations of refractivity at different altitudes, we subtract the spherically symmetric refractivity profile ($N(r)$) from the total refractivity ($N$). Figure 6a shows the geolocation and the lowest altitudes from the earth’s surface of all COSMIC-2 RO profiles on 1 August 2020. As expected, there are some Ros whose observed lowest altitudes are high over high terrains, such as the Qinghai-Tibet Plateau, the Rocky Mountains, and the Andes Mountains, due probably to terrain occlusion. There are also some RO profiles with high lowest altitudes over oceans in the tropics. The reasons for the latter case are probably associated with the complicated small-scale structures of water vapor distributions that makes the vertical gradients of refractivity strong, thus truncating the RO signal propagation [40]. This will be explained below.

The spherically symmetric refractivity profile, which will be denoted as ${\overline{N}}^{obs}(r)$, can be calculated by averaging all RO observations with the same distance from the perigee points to the earth’s center. The vertical profile of spherically symmetric refractivity calculated by averaging the ERA5 reanalysis that is interpolated to RO locations is denoted as ${\overline{N}}^{ERA5}(r)$. Results for all COSMIC-2 RO profiles between 45°S and 45°N on 1 August 2020 are shown in Figure 6b. As expected, both ${\overline{N}}^{obs}(r)$ and ${\overline{N}}^{ERA5}(r)$ decrease with altitude. The spherically symmetric refractivity calculated from COSMIC-2 RO observations is slightly smaller than that calculated from the ERA5 reanalysis above and below 12 km where the two have the same values.

In the next subsection, we investigate the relationship of the vertical gradients of $N-{\overline{N}}^{obs}(r)$ to the lowest altitudes of COSMIC-2 observed RO profiles. Subtracting ${\overline{N}}^{obs}(r)$ from the total refractivity reduces the influence of a strong dependence of the atmospheric refractivity on altitude.

4.3. Statistical Results on the Relationship between the RO Lowest Altitudes and PBL Heights

The vertical gradients of $N-{\overline{N}}^{obs}(r)$ are calculated at the lowest altitudes of all the COSMIC-2 RO profiles in August 2020 over the SEP and SCP areas. The scatter plots in Figure 7 suggest a relationship between the lowest RO observing altitudes and the strong local vertical gradients of $N-{\overline{N}}^{obs}(r)$ in both SEP and SCP areas. The points on the left of the dashed line, which is line with the vertical gradient of $N-{\overline{N}}^{obs}(r)$ equal to −65 N-unit km⁻¹, represent the cases where the GPS RO signals in the low altitudes cannot be observed when encountering significant strong vertical gradients of $N-{\overline{N}}^{obs}(r)$. The results in the two areas are not exactly the same in that the strong negative vertical gradients of $N-{\overline{N}}^{obs}(r)$ occur more frequently in the SEP than in the SCP area, especially when the negative vertical gradients are less than −200 N-unit km⁻¹. This finding is consistent with previous studies that have pointed out that the atmospheric ducting occurs most frequently in the SEP area [30,31]. In Figure 7, the ERA5 PBL heights near the RO profiles are indicated in colors. It is seen that when the vertical gradients are smaller than −65 N-unit km⁻¹, the lowest altitudes of COSMIC-2 RO observations agree with the PBL heights. It is likely that the inversion at the PBL top hinders the propagation of RO signals. As shown by the black solid curve in Figure 7, however, this kind of RO only accounts for a very small fraction of all the COSMIC-2 RO data in these two areas.

We notice that there are still quite a few RO profiles on the right side of the dashed line whose lowest altitudes are high, suggesting that some RO profiles are unable to penetrate down into the lower troposphere even though the vertical gradient is weak. The impacts of the horizontal gradient of refractivity on how deep the GPS RO signals can be observed will be investigated in the following section.

Figure 8 gives a geographic look at the relationship between the lowest RO altitudes and the PBL heights for those RO cases whose negative vertical gradients of $N-{\overline{N}}^{obs}(r)$ at the RO lowest altitudes are smaller than −65 N-unit km⁻¹. We see correspondence between the lowest RO altitudes (top panels) and the PBL heights (bottom panels) in both the SEP (left panels) and SCP areas (right panels). More cases in the SEP area have higher PBL heights than the SCP area. This observation is consistent with the results in Figure 7 that showed that more significantly strong negative vertical gradients of $N-{\overline{N}}^{obs}(r)$ appeared in the SEP area than the SCP area.

4.4. Impact of Strong Horizontal Gradient of Refractivity

The horizontal gradient of refractivity is in general about two to three orders of magnitude smaller than the vertical gradient of refractivity. However, the relatively strong horizontal gradient along the tangent direction could also affect the propagation of the RO rays. In this section, we only use the RO data with weak vertical gradients of $N-{\overline{N}}^{obs}(r)$ to explore the impact of strong horizontal gradients so that the influence of strong vertical gradients is excluded.

Two COSMIC-2 RO profiles will be examined first. The RO3 profile is a typical RO profile located in an area with a large horizontal gradient of refractivity at a certain height, which may result in multipath propagation in the impact parameter domain and make the retrieval of refractivity impossible below this height [41,42,43]. The RO4 profile is located near the RO3 in both time and space but reached the surface and is chosen for comparison. Figure 9 shows the vertical profiles of the bending angle, refractivity, temperature, and specific humidity for the two profiles from the COSMIC-2 RO observations and the ERA5 reanalysis. The ERA5 profiles are smoother than the observed RO profiles for all four variables. The bending angle calculated by the Abel-transform from the ERA5 reanalysis for the RO4 profile (black dashed curve in Figure 9a) compares well with the observed bending angle (black solid curve in Figure 9a), but the ERA5 bending angle profile for the RO3 (red dashed curve) differs from the RO3 observation (red solid curve). Moreover, the refractivity and temperature profiles from the COSMIC-2 observations and the ERA5 reanalysis (Figure 9b,c) are very close, while the specific humidity profiles (Figure 9d) differ significantly, especially for the RO3 (red curves). To understand the reasons for the above differences, the atmospheric conditions near the two profiles are analyzed below.

Figure 10a shows the horizontal distributions of the geopotential height and refractivity at 775 hPa, which are the closest pressure levels to the RO3’s lowest altitude. It can be seen that the RO3 is located at a place where the refractivity changes dramatically along its tangent direction near the perigee (black arrow). In the vertical cross-section along the tangent direction of the RO3 perigee (Figure 10b), an abnormal distribution of upward protrusion of refractivity can be seen at the lowest altitude of the RO3. This distribution results in a strong horizontal gradient of refractivity at the lowest height of the RO3 profile, which is shown in Figure 11. The RO3 ray path passed a streak of strong horizontal gradient of refractivity. In the vertical cross-section, there is a zone of strong horizontal gradients of refractivity located at and below the lowest RO3 altitude. Therefore, the strong horizontal gradient of refractivity is considered as a possible reason for the RO3 not to go further down to the earth’s surface.

The RO4 profile is located in the northeast of a closed high of the geopotential height, characterizing a much more homogeneous distribution of the atmospheric conditions than those around the RO3 (figure omitted). The horizontal gradients surrounding RO4 are small.

Using the RO data with weak vertical gradients of $N-{\overline{N}}^{obs}(r)$ at the RO lowest heights, we provide scatter plots of the horizontal gradients of refractivity along the tangent directions of the ray paths at the perigee points with respect to the RO lowest heights in both the SEP and SCP areas (Figure 12). When the horizontal gradients of refractivity are greater than 0.15 N-unit km⁻¹, the lowest altitudes of most of these RO profiles are high. Although strong vertical gradients of refractivity occur more frequently in the SEP area (see Section 4.3), more cases of strong horizontal gradients of refractivity are found in the SCP area.

4.5. Influence of Clouds

In addition to the COSMIC-2 RO profiles that cannot reach near the surface due to a strong vertical gradient of $N-{\overline{N}}^{obs}(r)$ and/or strong horizontal gradient of refractivity, there are still some RO profiles whose lowest observing altitudes do not reach near the surface. These are the points left of the dashed line in the scatter plot Figure 12. Since the absorption and scattering effects of clouds contribute to the atmospheric refractivity [38,44], we may examine whether the presence of clouds might have affected the penetration depth of the RO profiles. The cloud contribution to the refractivity can be calculated as:

$$N_{cloud}=1.45w_{liquid}+0.69w_{ice}$$

(5)

where $w_{liquid}$ is the liquid water content in g m⁻³, and $w_{ice}$ is the ice water content in g m⁻³. Zou et al. [45] provided a detailed derivation of the two terms in Equation (5). For convenience, we will call $N_{cloud}$ the cloud refractivity. The cloud refractivity values at the lowest heights of the RO profiles are indicated in colors in Figure 12. It can be seen that the majority of the cloudy RO profiles have high lowest altitudes in both the SEP and SCP areas.

5. Discussion

More than 90% of COSMIC-2 RO profiles can detect the very lower troposphere. The percentages of COSMIC-2 RO profiles that cannot go below 200 m above the surface due to strong vertical gradients of refractivity, strong tangential horizontal gradients of refractivity, and clouds are about 7.3%, 1.4%, and 0.7%, respectively. There are still a few RO profiles whose lowest altitudes are relatively high but are not affected by the above factors. We examined the uncertainty (LSW) and confidence (CP) of the bending angle for these RO profiles in order to see if there are any obvious differences between the RO profiles whose lowest altitudes are low or high (figures omitted). Both the uncertainties and confidences of the RO profiles with low and high lowest altitudes increase with decreasing altitude. The variations and magnitudes of the uncertainties and confidences for RO profiles with the high lowest altitudes do not differ greatly from those with the low lowest altitudes. Therefore, the reason that these RO profiles that have weak refractivity gradients and are unaffected by clouds cannot penetrate down is not known and requires further investigation. Fortunately, they only account for about 0.6% of COSMIC-2 RO profiles.

6. Conclusions

The near-surface observations under different weather conditions are vital for the research on tropical convection and hurricane genesis and development. More than half of the COSMIC-2 RO profiles can provide high-vertical-resolution profiles with their lowest observing heights below 200 m, and only a few ROs cannot reach the lower troposphere. In order to understand what might have affected the downward penetrations of these RO profiles, the COSMIC-2 RO observations and the ERA5 reanalysis data in the SEP and SCP areas were examined and analyzed in this study. The strong vertical gradient of $N-{\overline{N}}^{obs}(r)$ is confirmed to prevent the RO signals from penetrating down, and the impacts of strong horizontal gradients of refractivity and clouds were also discussed.

First, when the strong negative vertical gradient of $N-{\overline{N}}^{obs}(r)$ is less than −65 N-unit km⁻¹, the lowest RO altitudes correspond well to the PBL heights. Secondly, the strong horizontal gradient of refractivity can also prevent the ROs from reaching the lower troposphere below it in most cases. For the two areas studied, the large negative vertical gradients of $N-{\overline{N}}^{obs}(r)$ appear more frequently in the SEP area where the atmospheric ducts occur than in the SCP area. The strong horizontal gradients of refractivity happen more frequently in the SCP area. Thirdly, the absorption and scattering effects of clouds can also affect the downward detection of RO observations to some extent. There are still a few RO profiles (~0.6%) with high lowest heights which cannot be attributed to the above three reasons, and their bending angle uncertainties and confidences distribute similarly to the profiles that reached the surface.

This study helps to understand the different penetration depths of COSMIC-2 RO profiles under different atmospheric conditions, and it will also benefit efforts to make better use of COSMIC-2 RO observations for studying weather in the tropical lower troposphere.