## 1. Introduction

The middle and upper atmosphere are relevant to human survival and development, global climate, and environmental change (Kidston et al. 2015), and therefore research into these regions and determination of the parameters governing their behavior are of great practical value. For instance, Scaife et al. (2014) show the potential of predicting the quasi-biennial oscillation (QBO) and via downward coupling (not yet fully understood) the potential for seasonal prediction at midlatitude. Baldwin and Dunkerton (2001) show that a weakening or strengthening stratospheric vortex can alter circulation patterns down to the surface.

There are several methods to determine the parameters of the middle and upper atmosphere, such as weather balloons (Wang et al. 2005), sounding rockets (Fan et al. 2013), radar (Rechou et al. 2013), lidar (Sasi et al. 2003), and satellite remote sensing. The middle atmosphere influences the troposphere by coupling processes on a global scale. This calls for global observations over both land and sea, and hence for satellites. Therefore satellite remote sensing is of great importance in the study of temporal and spatial changes in parameters (e.g., Schmidt et al. 2010; W. Wang et al. 2013), the determination of structural characteristics, and the investigation of dynamic processes in the middle and upper atmosphere (e.g., Smith et al. 2007; Ern et al. 2011).

At present, China still lacks a meteorological satellite for determining the parameters of the middle and upper atmosphere. However, satellite remote sensing has been conducted on an international basis for the past 20 yr, mainly using the methods of occultation detection and limb sounding (Syndergaard 1998; Chen et al. 2011). Two examples that provide a multiyear continuous dataset are the Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC) and Sounding of the Atmosphere using Broadband Emission Radiometry (SABER). The COSMIC satellites use GPS radio occultation detection (Guo et al. 2011; Sheng 2013; Jiang et al. 2013). Among other advantages, this technology provides global coverage, all-weather observation, high accuracy, high vertical resolution, and long-term stability, since it is self-correcting without calibration. The main scientific goal of the Thermosphere, Ionosphere, Mesosphere Energetics, and Dynamics (TIMED) satellite is to explore dynamics and energy transfer processes in the mesosphere, lower thermosphere, and ionosphere region (Wu et al. 2006). The TIMED satellite carries the SABER detector, which is a 10-channel broadband radiometer that can obtain temperature data at 15–120-km height by observing radiation from CO_{2} in the 15-*μ*m band with limb sounding (Xu et al. 2007; Remsberg et al. 2008).

The accuracy and precision of such remote sensing instruments need to be validated. For instance, Ho et al. (2009) find that the results from COSMIC show a mean temperature deviation of 0.05 K, with a standard deviation of 1 K, in the upper troposphere and low stratosphere (8–30-km height). Below 8 km, the high water vapor content leads to complex propagation phenomena such as multipath propagation and superrefraction, leading to uncertainties in inversion results and increasing the temperature deviation (Rocken et al. 2000; Sokolovskiy et al. 2010; Ho et al. 2010). In addition, in comparison with global sounding and National Centers for Environmental Prediction global reanalysis data (Sun et al. 2010; Kishore et al. 2011; B. R. Wang et al. 2013), COSMIC occultation detection showed high precision, and it can be used to assess the accuracy of other detection techniques.

The accuracy of temperature data from TIMED/SABER remote sensing detection has been compared from various aspects (Sica et al. 2008; Ern et al. 2008). Remsberg et al. (2008) made a comprehensive assessment of data from the latest version (1.07) of SABER. The results show that the deviation of SABER temperature data is 1–3 K in the low stratosphere and ~1 K near the top of the stratosphere. The temperature deviation is ~2 K in the middle of the mesosphere, and the SABER-detected temperature is generally cold in the upper mesosphere. Fan et al. (2013) compared TIMED/SABER data with Chinese TK-1 rocket sounding data and showed that the atmospheric parameters obtained from the SABER data are close to those from the rocket sounding data. Especially at heights below 40 km, the temperatures obtained from the two detection methods are relatively consistent but deviate increasingly from one another as the height increases above 40 km.

There have been far fewer comparisons between COSMIC data and TIMED/SABER data, and only the results of Gong et al. (2013) have so far been published. The results show that the mean deviations between COSMIC and SABER approach 0 K at ~38 km, with positive values at higher altitudes and negative values at lower altitudes. The standard deviation of temperature reaches a minimum of ~1.8 K at ~20 km and a maximum of 6.5 K at ~55 km.

In a case study the current paper compares TIMED/SABER and COSMIC data with a focus on China. The accuracy is validated by in situ measurements from high-resolution radiosonde data. Special focus is given on the temporal and regional variation of the differences and according standard deviations.

The data and method are introduced in section 2. The comparison of high-resolution sounding data and satellite sounding data is presented in section 3. The comparison between the temperature obtained from COSMIC and that from TIMED/SABER is presented in section 4. Section 5 gives the conclusions of this comprehensive assessment.

## 2. Data and method

### a. Data source

#### 1) Radiosonde data

The high-resolution data came from balloon sounding experiments in Beijing (39.80°N, 116.47°E) and Xilinhot (43.95°N, 116.01°E) between 29 May and 28 June 2008. These experiments provided aerological sounding data from different radiosondes. The measuring data from Vaisala RS92 will be used in this paper. Vaisala RS92 was developed in Finland, and its good precision has been verified more than once (Steinbrecht et al. 2008; He et al. 2009). This paper will use the data from Vaisala RS92 to assess the accuracy of COSMIC and TIMED/SABER data at the corresponding heights.

#### 2) COSMIC data

COSMIC is a constellation composed of six microsatellites, distributed in six orbital planes. COSMIC was launched from Vandenberg Air Force Base in the United States on 15 April 2006. The GPS antenna and receiver mounted on each COSMIC satellite observe the refraction that occurs when a GPS signal passes through Earth’s atmosphere and the according time delay of the GPS signal (Anthes et al. 2008). The use of active radio signals enables measurements during the day and night. The use of L-band signals, with wavelengths of ~20 cm, ensures that the signals are negligibly influenced by aerosols and clouds (Sivakumar et al. 2011). A distinctive feature of the COSMIC mission, compared to previous missions, is the employed open-loop mode where the radio occultation mission is tracked during both set and rise neutral atmospheric occultation in the lower troposphere (Schreiner et al. 2007). The open-loop tracking technique will significantly reduce the GPS radio occultation inversion biases by eliminating tracking errors (Sokolovskiy et al. 2006). A model is used to retrieve data on electron density, temperature, pressure, and water vapor content along the path of the radio signal (Wickert et al. 2001).

The COSMIC data used in this paper were provided by the COSMIC Data Analysis and Archive Center (CDAAC) of the University Corporation for Atmospheric Research (UCAR) in the period from April 2006 to December 2009. The UCAR CDAAC radio occultation retrieval procedure begins with the phase and amplitude of the radio waves and precise positions and velocities of the satellites and ends with the retrieved refractivity profile at the estimated “occultation point.” Pressure and temperature are reconstructed from the retrieved refractivity by integration of the hydrostatic equation at altitudes where humidity may be neglected, that is, above the troposphere. The integration starts at 150 km, by setting pressure and temperature to zero. This “zero initialization” does not affect the retrieved temperature at ~80 km, which is close to the first guess (CIRA-86). At lower altitudes the retrieved temperatures are gradually affected by observations, according to an increase of their weight in the retrieved refractivity (Kuo et al. 2004) by means of an optimal estimation retrieval. Optimal estimation combines the information of an a priori profile with the information from the observation weighted with the respective noise/uncertainty (or atmospheric variability for a climatology used as a priori). Below 40 km the COSMIC retrievals are almost entirely determined by the GPS measurements. The vertical resolution of the retrieved temperature profiles is approximately 1 km at tropopause altitude. The data product atmPrf is dry-air postprocessing data from the second stage of processing. This is a dry temperature data product that does not include relative humidity in the inversion process and hence is reliable at and above the tropopause (>~15 km). The version numbers of the data from April 2006 to March 2009 and the data from April to December 2009 are 2007.3200 and 2009.2650, respectively.

#### 3) TIMED/SABER data

The TIMED satellite was launched on 7 December 2001. The orbit has an altitude of 625 km and an inclination of 74.1°. The satellite moves around Earth in about 1.6 h. It flies through the same latitude at two similar local times for a specific day, one for ascending and one for descending orbit nodes. Because of the precessional motion of ~12 min every day, complete local time coverage is achieved in 60 days. The SABER detector carried on TIMED began making observations in late January 2002. It measures CO_{2} infrared limb radiance from the tropopause to the lower thermosphere. Temperature profiles are retrieved below 70 km from 15-*μ*m CO_{2} emissions using the local thermodynamic equilibrium (LTE) algorithm (Mertens et al. 2001; Remsberg et al. 2008). SABER operational LTE retrieval is based on the multiple interleave approach (Remsberg et al. 2004). The original radiance profiles sampled at ~380 m sampling are split to five profiles, retrieved independently and merged to a common grid after the retrieval again. Thus, each individual retrieval is performed at a vertical step width close to the width of the vertical field of view of 2 km. The retrieved temperature profiles hence have a vertical resolution somewhat better than 2 km, but are oversampled on a vertical grid of ~380 m. At altitudes above 70 km, a nonlocal thermodynamic equilibrium (NLTE) retrieval is performed.

### b. Method

COSMIC atmPrf data are provided for the altitude range 0–60 km on a 100-m grid, SABER data are provided with ~380-m vertical sampling and radiosonde data have a much higher vertical resolution and are sampled every 5 m. To minimize interpolation errors, we decided to perform the comparison on the vertical grid of the atmPrf data. The radiosonde data were smoothed by a 100-m boxcar before interpolation. The data were assessed by comparison of individual profiles as well as statistic comparison using the methods described as following.

#### 1) Individual comparison method

*S*indicates satellite sounding data and superscript

*R*indicates radiosonde data, and the subscript

*i*indicates the serial number of the height (there are

*N*heights in all). The mean deviation and standard deviation in individual comparison are averaged over all altitudes for providing two simple data value to analyze the differences between a pair of profiles.

#### 2) Statistical comparative method

*j*is the serial number of the data (there are

*M*matching data in all).

### c. Mistime and misdistance

Two kinds of data differences may be caused by either the effects of the compared measurement techniques or atmospheric variation. In this paper we are interested in the effects of the measurement techniques and hence need to compare measurements at sufficiently close atmospheric conditions. Because atmospheric temperature changes with height, time, longitude, and latitude, we need to select satellite data that are measured sufficiently close to corresponding radiosonde data in terms of time, longitude, and latitude, and we compare these data at the same heights. The selection of sufficiently close data is defined by limits of mistime and misdistance (e.g., Wendt et al. 2013). While short misdistances and mistimes reduce the atmospheric variations, the selection of larger mistime and misdistance value has the advantage that a larger number of matching data are found, which enhances the statistical significance of the comparisons. Sofieva et al. (2008) showed that vertical wavenumber spectra of temperature fluctuations are similar, even for profiles separated significantly in space and time (a few hundreds of kilometers, a few hours). Limb sounding implies a weighting function along the line of sight (LOS) of approximately 200-km width affecting the sounding of atmospheric structures of comparable horizontal extent. For instance, atmospheric gravity waves (GWs) are only resolved if their horizontal wavelength projected onto the horizontal LOS is 200 km or longer (Preusse et al. 2002; Lange and Jacobi 2003), which may introduce differences in the vertical profiles measured by in situ and limb sounding instruments, even if they are measured at low misdistance and mistime (Preusse et al. 2003). Therefore comparisons might not further improve when misdistances are reduced below these shortest resolvable horizontal scales. For instance, the study of Zhang et al. (2011) shows insignificantly statistical differences of looser collocation (i.e., 100-, 200-, and 300-km radial misdistance with 1-, 2-, and 3-h mistime) between radiosonde and satellite data.

## 3. Comparison between satellite sounding data and radiosonde data

### a. Comparison of COSMIC occultation detection data and radiosonde data

The height ranges of COSMIC occultation data and radiosonde data are 0–60 and 0–32 km, respectively, and therefore we chose the latter common region for the purpose of comparison. As the product atmPrf of COSMIC is reliable at and above ~15 km, the reliable altitude ranges would be 15–32 km in this comparison. For the comparison we choose a maximum mistime of 3 h and allowed for 3° spatial separation both in latitude and in longitude, resulting in 16 pairs of COSMIC data matching radiosonde data.

In the following, two kinds of data are analyzed. Figure 1 provides an individual comparison of COSMIC detection data and radiosonde data. The geographic location of the COSMIC data is 39.32°N, 116.24°E, and the time is 0032 UTC 3 June 2008. The geographic location of the radiosonde data is 39.80°N, 116.47°E, and the time is 2300 UTC 2 June 2008. Data match each other closely in the vertical profile. As can be seen from Fig. 1a, the COSMIC data and the radiosonde data have almost identical spatial distributions of temperature. In Fig. 1b, the temperature difference between the two is mostly within ±3 K; only at about 7 and 10 km is the difference beyond this. The temperature difference reaches a maximum of 5.8 K at 10.5 km at the tropopause temperature minimum. Integrated over the reliable altitude ranges, the mean difference between the two profiles is −0.67 K, and the standard deviation is 1.26 K.

Figure 2 shows the statistical comparison between the COSMIC and radiosonde temperature data. The profiles of the mean temperature deviation and the standard deviation between the two sets of data are presented in Figs. 2a and 2b. The green box with dotted line in the graph indicates the reliable altitude ranges. The shadow area in Fig. 2a is the statistical error of the mean deviation. The mean temperature deviation in Fig. 2a is 0 K at about 9 km and is and is almost everywhere positive below and negative above this altitude. Within the reliable ranges, the mean deviations are between −2 and 0 K. There is a slight S around 10 km; this may be because the vertical resolution of radiosonde profiles is much higher than that of COSMIC profiles. Statistic uncertainties for mean deviations are ~1.5 K in the region 0–2 km and ~1 K in 2–32 km. In Fig. 2b, the standard deviation between the two is mostly less than 3 K in the reliable region 15–32 km. At the altitudes below, the deviation is greater at heights of 7–11 km, and the mean deviation reaches a maximum of ~4.9 K near 10.5 km.

Statistical comparison between COSMIC and radiosonde data. The green box with dotted line indicates the reliable altitude range. The shadow area is the statistical error of the mean deviation.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Statistical comparison between COSMIC and radiosonde data. The green box with dotted line indicates the reliable altitude range. The shadow area is the statistical error of the mean deviation.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Statistical comparison between COSMIC and radiosonde data. The green box with dotted line indicates the reliable altitude range. The shadow area is the statistical error of the mean deviation.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

### b. Comparison of SABER sounding data and radiosonde data

The height ranges of TIMED/SABER satellite remote sensing and of radiosonde detection are 15–135 and 0–32 km, respectively, and therefore we chose the common height range 15–32 km as the comparison region. In the height range 15–32 km, both the two sets of data are reliable. Eight SABER profiles matching with radiosonde data are found, with the matching condition (±3°, ±3 h). Figure 3 provides an individual comparison of SABER sounding data and radiosonde data. The geographic location of the SABER data is 37.88°N, 119.41°E, and the time is 1300 UTC 2 June 2008. The geographic location of radiosonde data is 39.80°N, 116.47°E, and the time is 1300 UTC 2 June 2008. As can be seen from Fig. 3a, there is no significant temperature difference between the SABER and radiosonde data. In Fig. 3b, the temperature difference is within ±4K throughout the region 15–32 km, but it approaches 4 K at heights of 24–28 km. The mean temperature difference between the two groups of data is 0.93 K, and the standard deviation is 1.70 K.

Individual comparison between SABER and radiosonde data.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Individual comparison between SABER and radiosonde data.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Individual comparison between SABER and radiosonde data.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Figure 4 shows the statistical comparison between the SABER and radiosonde temperature data. Figures 4a and 4b show the profiles of the mean temperature deviation and the standard deviation, respectively. The mean temperature deviation in Fig. 4a is positive at heights of 20–29 km and is almost negative at other heights. Statistic uncertainties for mean deviations are about 1.5 K at heights of 15–20 km and ~1 K at 20–32 km. In Fig. 4b, the standard deviation is within 1–4.8 K and mostly less than 4 K in the region 15–32 km. It is larger at heights of 15–19 and 22.5–25 km, reaching a maximum of 4.8 K at about 17 km.

Statistical comparison between SABER and radiosonde data. The shadow area is the statistical error of the mean deviation.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Statistical comparison between SABER and radiosonde data. The shadow area is the statistical error of the mean deviation.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Statistical comparison between SABER and radiosonde data. The shadow area is the statistical error of the mean deviation.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Because there is a lack of comparative data above 32 km, comparison between the two kinds of satellite sounding data and radiosonde data cannot illustrate the detection precision of satellite remote sensing over a larger range of heights. To analyze the accuracy of the two sets of data at heights of 15–60 km in the China region, the COSMIC occultation detection data and SABER satellite sounding data are compared in the next section.

## 4. Comparison of COSMIC occultation data and SABER sounding data

### a. Data matching condition

The COSMIC and SABER are both limb sounders. The common height range of COSMIC occultation detection and SABER satellite sounding is 15–60 km, that is, the upper troposphere, stratosphere, and lower mesosphere. Because of the optimal estimation retrieval the actual measurement contribution of GPS measurements to the retrieval product is negligible around 60 km. The COSMIC data is unreliable in the height range 40–60 km, so the reliable range in this comparison would be 15–40 km. The lower and upper atmosphere are coupled in this region, and so that the atmospheric conditions there are more complex. To reduce the effects of differences in time and space, we chose the COSMIC data as close to the SABER data in time and space as possible, with longitude and latitude differing by less than 1° and time by less than 1 h. There are 1272 pairs of matching data from COSMIC and SABER in the China region. Figure 5 shows the distribution of matching data, with the boundary line in blue and the data matching points shown as red crosses.

Distribution of matching data from COSMIC and SABER.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Distribution of matching data from COSMIC and SABER.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Distribution of matching data from COSMIC and SABER.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

### b. Individual and statistical comparisons of COSMIC and SABER data

Figure 6 provides an individual comparison of COSMIC occultation detection data and SABER sounding data. The geographic location of COSMIC data is 15.22°N, 134.26°E, and the time is 2200 UTC; 4 December 2007. The geographic location of the SABER data is 15.40°N, 133.39°E, and the time is 2100 UTC 4 December 2007. As can be seen from Fig. 6a, there is a difference between the COSMIC data and SABER data. In Fig. 6b, the temperature difference is larger near 15 km and in the region 50–60 km but is less than 5 K at other heights. The reason for increasing standard deviations at lowest altitudes maybe the saturation of CO_{2} emissions below 20 km (Remsberg et al. 2008). Integrated over the reliable altitude ranges, the mean temperature difference between the two groups of data is 0.98 K and the standard deviation 3.64 K.

Individual comparison between COSMIC and SABER data. The green box with dotted line indicates the reliable altitude range.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Individual comparison between COSMIC and SABER data. The green box with dotted line indicates the reliable altitude range.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Individual comparison between COSMIC and SABER data. The green box with dotted line indicates the reliable altitude range.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Figure 7 shows the statistical comparison between the COSMIC and SABER temperature data. Figures 7a and 7b give the profiles of the mean temperature deviation and the standard deviation, respectively, between the two sets of data. The green box with dotted line in the graph indicates the reliable altitude ranges. The mean temperature deviation in Fig. 7a is 0 K at about 44 km and is negative above and positive in the reliable ranges. In the reliable altitude range SABER temperatures are generally larger than COSMIC temperatures with maximum deviations of 3 K at 23 km. Statistic uncertainties for mean deviations are about 0.2 K at all attitudes. In Fig. 7b, the standard deviations are all 2.5 K or more in the green box. The deviation is large near 15 km and it decreases to reach a minimum of ~2.5 K at 20 km. There is a peak of ~3.7 K near 23 km, after which the standard deviation remains approximately the same at ~3.5 K from 25 to 35 km.

Statistical comparison between COSMIC and SABER data. The green box with dotted line indicates the reliable altitude range.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Statistical comparison between COSMIC and SABER data. The green box with dotted line indicates the reliable altitude range.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Statistical comparison between COSMIC and SABER data. The green box with dotted line indicates the reliable altitude range.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

### c. Seasonal change in the temperature deviation

To investigate the seasonal change in the temperature deviation, we divided the 1272 pairs of matching data according to season (spring, summer, autumn, and winter) and statistically analyzed the temperature deviation for different seasons. Table 1 gives the division of seasons and the distribution of matching data.

Matching data from COSMIC and SABER according to season.

Figure 8 shows the statistical results for the mean temperature deviation and standard deviation between COSMIC occultation detection and TIMED/SABER satellite sounding in the China region. As we can see from Fig. 8a, the mean temperature deviations of different seasons show a similar situation with that of the whole year. The seasonal difference in mean temperature deviation is small at heights of 15–35 km but larger at other heights. Figure 8b shows that the standard deviation is lowest in summer and greatest in winter, with the standard deviations for spring and autumn being intermediate. The seasonal difference in standard deviation is small at heights of 15–30 km, with that between summer and winter being less than 1 K in this range. However, at heights of 30–60 km, it is larger, reaching a maximum of ~3.7 K.

Mean temperature and standard deviation for different seasons and annually (colors).

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Mean temperature and standard deviation for different seasons and annually (colors).

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Mean temperature and standard deviation for different seasons and annually (colors).

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

### d. Changes of deviation with latitude

To investigate the effects of latitude variation, we divided the China region (3.86°–53.55°N, 73.66°–135°E) into four areas according to latitude and analyzed the temperature deviations in the different areas. Figure 9 gives the partition results, and it shows that the number of matching pairs increases with increasing latitude.

The temperature deviations in four areas partitioned by latitude with matching pairs indicated.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

The temperature deviations in four areas partitioned by latitude with matching pairs indicated.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

The temperature deviations in four areas partitioned by latitude with matching pairs indicated.

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Figure 10 gives the statistical results for the change in temperature deviation with latitude. As can be seen from Fig. 10a, the difference in the mean temperature deviation between the different areas is small at height of 17–32 km, but it is larger in lower and higher regions. Figure 10b shows that the standard deviation decreases with increasing latitude at heights of 15–20 km: the deviation is larger at lower latitude in this region, with a maximum of ~7.5 K at 15 km, and is smaller at higher latitudes. The effect of latitude is very small, and the standard deviations of the four different areas are close at heights of 20–24 km. In the range 25–55 km, the standard deviation is greatest in area 4, of higher latitude, and is smaller in areas 1 and 2. In the range 55–60 km, area 1, of lower latitude, has a larger deviation than the other three areas.

Mean temperature and standard deviation for the four different latitude areas and the whole region (colors).

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Mean temperature and standard deviation for the four different latitude areas and the whole region (colors).

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

Mean temperature and standard deviation for the four different latitude areas and the whole region (colors).

Citation: Journal of Applied Meteorology and Climatology 54, 9; 10.1175/JAMC-D-14-0151.1

In the south of China, cloud-top altitudes frequently are higher than 10 km. Infrared limb emissions are very sensitive to thin ice clouds, while COSMIC observations are insensitive to thin clouds. Given that SABER has a vertical field of view of 2 km, the increasing standard deviations below 20 km may indicate cloud contamination of SABER retrievals. In northern China in winter GWs were detected by a various instruments (Eckermann and Preusse 1999; Jiang et al. 2004). Since these are unlikely to be viewed exactly at the same location and in the same observation geometry (horizontal viewing direction) GW activity in winter at northern China may explain these increased standard deviations.

## 5. Conclusions

In this paper we used radiosonde data to validate the accuracy of COSMIC data at heights 0–32 km and TIMED/SABER data at 15–32 km. We then compared COSMIC data and TIMED/SABER data in the height range 15–40 km, during the period from April 2006 to December 2009. Our analysis leads to the following conclusions:

COSMIC has a high precision in the height range 0–32 km, with systematic bias with high-resolution data. This detection method retains its precision when the precision of radiosondes decreases, so it can be used to evaluate the performance of different sondes. The mean temperature deviation between the two methods is 0 K at ~9 km and is almost negative in higher region and positive in lower region. The standard deviation is mostly less than 3 K in the reliable region 15–32 km, and it is greater in the range 7–11 km, reaching a maximum of ~4.9 K at 10.5 km.

The precision of SABER sounding data is good in the height range 15–32 km and can accurately reflect the atmospheric condition. The mean temperature difference between SABER and radiosondes is positive at heights of 20–29 km and almost negative at other heights. The standard deviation between the two is within 1–4.8 K, being larger at 15–19 and 22.5–25 km, with its maximum at 17 km.

The temperature profiles obtained by COSMIC and SABER are consistent in the China region. The deviations between the two sets of data are mainly caused by systematic error and by differences in longitude, latitude, and time. In the altitude range 15–40 km SABER temperatures are generally larger than COSMIC temperatures with maximum difference of 3 K at 23 km. The standard deviations are all greater than 2.5 K in the region 15–40 km, being larger near 15 km and smallest, approaching 2.5 K, at 20 km. There is a peak in the standard deviation of ~3.7 K near 23 km, after which it remains approximately the same at ~3.5 K in the range 25–35 km.

The temperature deviation between COSMIC and SABER changes with the seasons, and there is a certain bias between the statistical results for different seasons. The seasonal difference in the mean deviation is small at heights of 15–35 km, but larger at other heights. The standard deviation is smallest in summer and largest in winter, and is intermediate in spring and autumn. The seasonal difference of the standard deviations is small at heights of 15–30 km, but is greater at 35–60 km and reaches a maximum of ~3.7 K.

The temperature deviation between COSMIC and SABER changes with latitude. The difference in the mean deviation between different areas is small at heights of 17–32 km, but it is larger in lower and higher regions. The standard deviation decreases with increasing latitude at heights of 15–20 km; it is larger at lower latitudes in this range, reaching a maximum of ~7.5 K at 15 km, and is smaller at higher latitudes. The effect of latitude variation is very small at heights of 20–24 km. Within the range 25–55 km, the area of highest latitude has the largest standard deviation and area of lowest latitude the smallest deviation.

The results presented in this paper indicate that the quality of COSMIC data is very high at altitudes lower than 40 km. The SABER data perform well at heights of 20–60 km but not well at lower altitudes. The temperature deviations between COSMIC and SABER change with latitudes and seasons in the China region. Through comparing data from COSMIC occultation detection and TIMED/SABER satellite sounding with high-resolution data, together with comparisons between the two techniques, the results described in this paper are consistent with those from previous research, and they can provide a basis for further study and applications of these techniques.

## Acknowledgments

The COSMIC data used in this paper were provided by the CDAAC of the University Corporation for Atmospheric Research, and the SABER data were provided by the TIMED/SABER team. We acknowledge the contributions to this work from the CDAAC and the TIMED/SABER team. Comments from four anonymous reviewers are also very much appreciated.

## REFERENCES

Anthes, R. A., P. A. Bernhardt, and Y. Chen, 2008: The COSMIC/FORMOSAT‐3 Mission: Early results.

,*Bull. Amer. Meteor. Soc.***89**, 313–333, doi:10.1175/BAMS-89-3-313.Baldwin, M. P., and T. J. Dunkerton, 2001: Stratospheric harbingers of anomalous weather regimes.

,*Science***294**, 581–584, doi:10.1126/science.1063315.Chen, S.-Y., C.-Y. Huang, Y.-H. Kuo, and S. Sokolovskiy, 2011: Observational error estimation of FORMOSAT-3/COSMIC GPS radio occultation data.

,*Mon. Wea. Rev.***139**, 853–865, doi:10.1175/2010MWR3260.1.Eckermann, S. D., and P. Preusse, 1999: Global measurements of stratospheric mountain waves from space.

,*Science***286**, 1534–1537, doi:10.1126/science.286.5444.1534.Ern, M., P. Preusse, M. Krebsbach, M. G. Mlynczak, and J. M. Russell III, 2008: Equatorial wave analysis from SABER and ECMWF temperatures.

,*Atmos. Chem. Phys.***8**, 845–869, doi:10.5194/acp-8-845-2008.Ern, M., P. Preusse, J. C. Gille, C. L. Hepplewhite, M. G. Mlynczak, J. M. Russell III, and M. Riese, 2011: Implications for atmospheric dynamics derived from global observations of gravity wave momentum flux in stratosphere and mesosphere.

,*J. Geophys. Res.***116**, D19107, doi:10.1029/2011JD015821.Fan, Z. Q., Z. Sheng, L. Wan, H. Q. Shi, and Y. Jiang, 2013: Comprehensive assessment of the accuracy of the data from near space meteorological rocket sounding.

,*Acta Phys. Sin.***62**, 199601, doi:10.7498/aps.62.199601.Gong, X. Y., X. Hu, X. C. Wu, and C. Y. Xiao, 2013: Comparison of temperature measurements between COSMIC atmospheric radio occultation and SABER/TIMED.

,*Chin. J. Geophys.***56**, 2152–2162.Guo, P., Y. H. Kuo, S. V. Sokolovskiy, and D. H. Lenschow, 2011: Estimating atmospheric boundary layer depth using COSMIC radio occultation data.

,*J. Atmos. Sci.***68**, 1703–1713, doi:10.1175/2011JAS3612.1.He, W. Y., S. P. Ho, H. B. Chen, X. J. Zhou, D. Hunt, and Y. H. Kuo, 2009: Assessment of radiosonde temperature measurements in the upper troposphere and lower stratosphere using COSMIC radio occultation data.

,*Geophys. Res. Lett.***36**, L17807, doi:10.1029/2009GL038712.Ho, S.-P., M. Goldberg, Y. H. Kuo, Z. Z. Cheng, and W. Schreiner, 2009: Calibration of temperature in the lower stratosphere from microwave measurements using COSMIC radio occultation data: Preliminary results.

,*Terr. Atmos. Oceanic Sci.***20**, 87–100, doi:10.3319/TAO.2007.12.06.01(F3C).Ho, S.-P., X. Zhou, Y.-H. Kuo, D. Hunt, and J. H. Wang, 2010: Global evaluation of radiosonde water vapor systematic biases using GPS radio occultation from COSMIC and ECMWF analysis.

,*Remote Sens.***2**, 1320–1330, doi:10.3390/rs2051320.Jiang, J. H., S. D. Eckermann, D. L. Wu, and J. Ma, 2004: A search for mountain waves in MLS stratospheric limb radiances from the winter Northern Hemisphere: Data analysis and global mountain wave modeling.

,*J. Geophys. Res.***109**, D03107, doi:10.1029/2003JD003974.Jiang, Y., Z. Sheng, and H. Q. Shi, 2013: Comparisons of tropopause derived from COSMIC measurements at Nanjing since August 2006.

,*J. Atmos. Sol.-Terr. Phys.***98**, 31–38, doi:10.1016/j.jastp.2013.03.013.Kidston J., A. A. Scaife, S. C. Hardiman, D. M. Mitchell, N. Butchart, M. P. Baldwin, and L. J. Gray, 2015: Stratospheric influence on tropospheric jet streams, storm tracks and surface weather.

,*Nat. Geosci.***8**, 433–440, doi:10.1038/ngeo2424.Kishore, P., and Coauthors, 2011: Global distribution of water vapour observed by COSMIC GPS RO: Comparison with GPS radiosonde, NCEP and JRA-25 reanalysis data sets.

,*J. Atmos. Sol.-Terr. Phys.***73**, 1849–1860, doi:10.1016/j.jastp.2011.04.017.Kuo, Y.-H., T.-K. Wee, S. Sokolovskiy, C. Rocken, W. Schreiner, D. Hunt, and R. A. Anthes, 2004: Inversion and error estimation of GPS radio occultation data.

,*J. Meteor. Soc. Japan***82**, 507–531, doi:10.2151/jmsj.2004.507.Lange M., and C. Jacobi, 2003: Analysis of gravity waves from radio occultation measurements.

*First CHAMP Mission Results for Gravity, Magnetic and Atmospheric Studies*, C. Reigber, H. Lühr, and P. Schwintzer, Eds., Springer, 479–484.Mertens, C. J., M. G. Mlynczak, M. Lopez-Puertas, P. P. Wintersteiner, R. H. Picard, J. R. Winick, L. L. Gordley, and J. M. Russell, 2001: Retrieval of mesospheric and lower thermospheric kinetic temperature from measurements of CO2 15 µm Earth limb emission under non-LTE conditions.

,*Geophys. Res. Lett.***28**, 1391–1394, doi:10.1029/2000GL012189.Preusse, P., A. D. Doernbrack, S. D. Eckermann, M. Riese, B. Schaeler, J. Bacmeister, D. Broutman, and K. U. Grossmann, 2002: Space based measurements of stratospheric mountain waves by CRISTA 1: Sensitivity, analysis method, and a case study.

,*J. Geophys. Res.***107**, 8178, doi:10.1029/2001JD000699.Preusse, P., S. D. Eckermann, M. Ern, F. J. Schmidlin, M. J. Alexander, and D. Offermann, 2003: Infrared limb sounding measurements of middle-atmosphere gravity waves by CRISTA.

*Remote Sensing of Clouds and the Atmosphere VII*, K. P. Schaefer et al., Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 4882), 134–148, doi:10.1117/12.463374.Rechou, A., J. Arnault, P. Dalin, and S. Kirkwood, 2013: Case study of stratospheric gravity waves of convective origin over Arctic Scandinavia—VHF radar observations and numerical modelling.

,*Ann. Geophys.***31**, 239–250, doi:10.5194/angeo-31-239-2013.Remsberg, E. E., and Coauthors, 2004: The Nimbus 7 LIMS version 6 radiance conditioning and temperature retrieval methods and results.

,*J. Quant. Spectrosc. Radiat. Transfer***86**, 395–424, doi:10.1016/j.jqsrt.2003.12.007.Remsberg, E. E., and Coauthors, 2008: Assessment of the quality of the Version 1.07 temperature-versus-pressure profiles of the middle atmosphere from TIMED/SABER.

,*J. Geophys. Res.***113**, D17101, doi:10.1029/2008JD010013.Rocken, C., Y.-H. Kuo, W. Schreiner, D. Hunt, and S. Sokolovskiy, 2000: COSMIC system description.

,*Terr. Atmos. Oceanic Sci.***11**, 21–52.Sasi, M. N., and Coauthors, 2003: A study of equatorial wave characteristics using rockets, balloons, lidar and radar.

,*Adv. Space Res.***32**, 813–818, doi:10.1016/S0273-1177(03)00412-5.Scaife, A. A., and Coauthors, 2014: Predictability of the quasi-biennial oscillation and its northern winter teleconnection on seasonal to decadal timescales.

,*Geophys. Res. Lett.***41**, 1752–1758, doi:10.1002/2013GL059160.Schmidt, T., J.-P. Cammas, H. G. J. Smit, S. Heise, J. Wickert, and A. Haser, 2010: Observational characteristics of the tropopause inversion layer derived from CHAMP/GRACE radio occultations and MOZAIC aircraft data.

,*J. Geophys. Res.***115**, D24304, doi:10.1029/2010JD014284.Schreiner, W., C. Rocken, S. Sokolovskiy, S. Syndergaard, and D. Hunt, 2007: Estimates of the precision of GPS radio occultations from the COSMIC/FORMOSAT-3 mission.

,*Geophys. Res. Lett.***34**, L04808, doi:10.1029/2006GL027557.Sheng, Z., 2013: The estimation of lower refractivity uncertainty from radar sea clutter using the Bayesian–MCMC method.

,*Chin. Phys.***22B**, 029302, doi:10.1088/1674-1056/22/2/029302.Sica, R. J., and Coauthors, 2008: Validation of the Atmospheric Chemistry Experiment (ACE) version 2.2 temperature using ground-based and space-borne measurements.

,*Atmos. Chem. Phys.***8**, 35–62, doi:10.5194/acp-8-35-2008.Sivakumar, V., P. Vishnu Prasanth, P. Kishore, H. Bencherif, and P. Keckhut, 2011: Rayleigh LIDAR and satellite (HALOE, SABER, CHAMP and COSMIC) measurements of stratosphere mesosphere temperature over a southern sub-tropical site, Reunion (20.8°S; 55.5°E): Climatology and comparison study.

,*Ann. Geophys.***29**, 649–662, doi:10.5194/angeo-29-649-2011.Smith, A. K., D. V. Pancheva, N. J. Mitchell, D. R. Marsh, J. M. Russell III, and M. G. Mlynczak, 2007: A link between variability of the semidiurnal tide and planetary waves in the opposite hemisphere.

*Geophys. Res. Lett*.,**34**, L07809, doi:10.1029/2006GL028929.Sofieva, V. F., F. Dalaudier, R. Kivi, and E. Kyroe, 2008: On the variability of temperature profiles in the stratosphere: Implications for validation.

,*Geophys. Res. Lett.***35**, L23808, doi:10.1029/2008GL035539.Sokolovskiy, S., C. Rocken, D. Hunt, W. Schreiner, J. Johnson, D. Masters, and S. Esterhuizen, 2006: GPS profiling of the lower troposphere from space: Inversion and demodulation of the open-loop radio occultation signals.

,*Geophys. Res. Lett.***33**, L14816, doi:10.1029/2006GL026112.Sokolovskiy, S., C. Rocken, W. Schreiner, and D. Hunt, 2010: On the uncertainty of radio occultation inversions in the lower troposphere.

,*J. Geophys. Res.***115**, D22111, doi:10.1029/2010JD014058.Steinbrecht, W., H. Claude, F. Schönenborn, U. Leiterer, H. Dier, and E. Lanzinger, 2008: Pressure and temperature differences between Vaisala RS80 and RS92 radiosonde systems.

,*J. Atmos. Oceanic Technol.***25**, 909–927, doi:10.1175/2007JTECHA999.1.Sun, B., A. Reale, D. J. Seidel, and D. C. Hunt, 2010: Comparison radiosonde and COSMIC atmospheric profile data to quantify differences among radiosonde types and the effects of imperfect collocation on comparison statistics.

,*J. Geophys. Res.***115**, D23104, doi:10.1029/2010JD014457.Syndergaard, S., 1998: Modeling the impact of the Earth’s oblateness on the retrieval of temperature and pressure profiles from limb sounding.

,*J. Atmos. Sol.-Terr. Phys.***60**, 171–180, doi:10.1016/S1364-6826(97)00056-4.Wang, B. R., X. Y. Liu, and J. K. Wang, 2013: Assessment of COSMIC radio occultation retrieval product using global radiosonde data.

,*Atmos. Meas. Tech.***6**, 1073–1083, doi:10.5194/amt-6-1073-2013.Wang, W., K. Matthes, T. Schmidt, and L. Neef, 2013: Recent variability of the tropical tropopause inversion layer.

,*Geophys. Res. Lett.***40**, 6308–6313, doi:10.1002/2013GL058350.Wang, Z., C.-P. Chang, B. Wang, and F.-F. Jin, 2005: Teleconnection from tropics to northern extratropics through a southerly conveyor.

,*J. Atmos. Sci.***62**, 4057–4070, doi:10.1175/JAS3600.1.Wendt, V., S. Wuest, M. G. Mlynczak, J. M. Russell III, J.-H. Yee, and M. Bittner, 2013: Impact of atmospheric variability on validation of satellite-based temperature measurements.

*J. Atmos. Sol.-Terr. Phys*.,**102**, 252–260, doi:10.1016/j.jastp.2013.05.022.Wickert, J., and Coauthors, 2001: Atmosphere sounding by GPS radio occultation: First results from CHAMP.

,*Geophys. Res. Lett.***28**, 3263–3266, doi:10.1029/2001GL013117.Wu, D. L., P. Preusse, S. D. Eckermann, J. H. Jiang, M. de la Torre Juarez, L. Coy, and D. Y. Wang, 2006: Remote sounding of atmospheric gravity waves with satellite limb and nadir techniques.

,*Adv. Space Res.***37**, 2269–2277, doi:10.1016/j.asr.2005.07.031.Xu, J., A. K. Smith, W. Yuan, H.-L. Liu, Q. Wu, M. G. Mlynczak, and J. M. Russell III, 2007: Global structure and long-term variations of zonal mean temperature observed by TIMED/SABER.

,*J. Geophys. Res.***112**, D24106, doi:10.1029/2007JD008546.Zhang, K., E. Fu, D. Silcock, Y. Wang, and Y. Kuleshov, 2011: An investigation of atmospheric temperature profiles in the Australian region using collocated GPS radio occultation and radiosonde data.

,*Atmos. Meas. Tech.***4**, 2087–2092, doi:10.5194/amt-4-2087-2011.