Thermal structure of intense convective clouds derived from GPS radio occultations

Thermal structure associated with deep convective clouds is investigated using Global Positioning System (GPS) radio occultation measurements. GPS data are insensitive to the presence of clouds, and provide high vertical resolution and high accuracy measurements to identify associated temperature behavior. Deep convective systems are 5 identiﬁed using International Satellite Cloud Climatology Project (ISCCP) satellite data, and cloud tops are accurately measured using Cloud-Aerosol Lidar with Orthogonal Polarization (CALIPSO) lidar observations; we focus on 53 cases of near-coincident GPS occultations with CALIPSO proﬁles over deep convection. Results show a sharp spike in GPS bending angle highly correlated to the top of the clouds, corresponding 10 to anomalously cold temperatures within the clouds. Above the clouds the temperatures return to background conditions, and there is a strong inversion at cloud top. For cloud tops below 14 km, the temperature lapse rate within the cloud often approaches a moist adiabat, consistent with rapid undiluted ascent within the convective systems.


Introduction
Deep convective systems play a fundamental role in atmospheric circulation and climate. Thunderstorms and mesoscale convective systems produce fast vertical transport, redistributing water vapor and chemical constituents and influencing the thermal structure of the Upper Troposphere and Lower Stratosphere (UTLS). Measurements of atmospheric parameters such as temperature and water vapor with high vertical reso-20 lution and accuracy near tropopause altitudes is difficult during severe weather events, and the processes that determine the details of UTLS transport are the subject of ongoing research (e.g., Pan et al., 2010). Hydration and dehydration of the UTLS are also poorly understood; models and measurements suggest that overshooting convection can penetrate the tropopause and, with different mechanisms, either hydrate or Introduction  (Kley et al., 1982;Danielsen, 1993;Wang, 2003;Jensen et al., 2007;Corti et al., 2008). The mean thermal structure in the tropics is linked to the influence of deep convection: a maximum in the observed lapse rate occurs near 12 km, associated with the top of convective influence (Gettelman and Forster, 2002;Folkins and Martin, 2005). 5 However, transient variability associated with deep convection is more difficult to identify, as balloon or aircraft measurements are difficult in regions of intense convection. Sherwood et al. (2003) used regular radiosonde measurements stratified according to the intensity of nearby convection to identify convective influence, identifying middle tropospheric warming and cooling at altitudes near the tropopause (typically above the 10 top of convection). Such cooling can result from convective overshooting, and also from excitation of gravity or Kelvin waves by the convection (Garcia and Salby, 1987;Randel and Wu, 2005). Chae et al. (2011) used 3-4 km thick layer measurements from the Aura Microwave Limb Sounder (MLS) to identify the average thermal structure of cloudy regions in the deep tropics, finding average cold temperatures (∼2 K) over ∼10- 15 14 km, and warm anomalies (∼1 K) above (∼18 km). In this paper we study the thermal structure of deep convective systems using observations from Global Positioning System (GPS) radio occultation measurements.
The GPS radio occultation (RO) technique (Kursinski et al., 1997) enables measurement of atmospheric thermal structure in any meteorological condition, since the 20 radiowave signal is not affected by clouds. Several GPS RO missions are working at present, providing a high density of vertical profiles with a good time and space coverage. Here we use GPS data to quantify the thermal structure of intense deep convection, based on identifying numerous GPS occultations that are co-located with large convective systems. The convective systems are identified using the Interna-25 tional Satellite Cloud Climatology Project (ISCCP) Deep Convection Tracking Database (Rossow et al., 1996), and the detailed cloud structure (cloud top) is provided by Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) measurements (Hunt et al., 2009 (Anthes et al., 2008), the Gravity Recovery And Climate Experiment (GRACE) twin satellites 5 (Beyerle et al., 2005), and the CHAllenging Minisatellite Payload (CHAMP) satellite (Wickert, 2001 The radio occultation technique uses the phase and amplitude of two L-band signals (L1 at 1575.42 MHz and L2 at 1227.60 MHz) transmitted from GPS satellites and received by the GPS receivers on board of Low Earth Orbit (LEO) satellites (e.g., COSMIC). When the signal passes through the atmosphere it is refracted due to the different density and it can be characterized by three parameters: the 15 bending angle α, the impact parameter a and the tangent radius r p . Under local spherical symmetry conditions the refractive index n can be derived from α, a and r p using the Abel inversion (Fjeldbo et al., 1971): where a r p = n(r p ) × r p is the impact parameter of the ray with radius r p . The 20 refractivity profile is then: Atmospheric refractivity (N) is dependent on the temperature (T ), pressure (p) and water vapor pressure (e) profiles according to: The water vapor term (proportional to e T 2 ) is small in the upper troposphere and stratosphere, where the water vapor mixing ratios are <100 ppmv, so that the 5 refractivity can be used to directly calculate the so-called dry temperature and dry pressure. More generally Eq. (3) can be solved by assimilating the refractivity together with first-guess profiles from the ECMWF analysis, using a onedimensional variational (1DVar) technique (Biondi et al., 2011). This provides so-called wet profile retrievals for T , p and e. We have analyzed the following 10 GPS level 2 data products in netCDF format: 1. Atmospheric Profile (atmPrf), which is a product containing the bending angle, refractivity, impact parameter and the so-called dry pressure and dry temperature (derived assuming no water vapor). All the parameters are reported versus the geometric height above the mean sea level and the coordinates 15 of the perigee point with high vertical resolution (order of a few meters).
2. European Centre for Medium-Range Weather Forecasts (ECMWF) Profile (ecmPrf), which is a low resolution product generated from the ECMWF gridded analysis (ECMWF TOGA 2.5 degree Global Surface and Upper Air Analysis) co-located with occultation profile for comparison purposes. It contains 20 refractivity, pressure, temperature and moisture profile versus mean sea level altitude and coordinates. We use these data simply for comparison to the higher resolution GPS measurements.
3. Wet Profile (wetPrf), which is an interpolated product with a vertical resolution of 100 m obtained using a one dimensional variational (1DVar)  together with ECMWF low resolution analysis data. This profile contains latitude and longitude of the perigee point, pressure, temperature, water vapor pressure, refractivity and mean sea level altitude of the perigee point.

b. Cloud measurements
Detailed information on the space-time behavior of convective systems comes 5 from the ISCCP Deep Convection Tracking Database, which is a globally gridded product with a nominal time resolution of 3 h. This is a database created to identify and describe the properties of mesoscale deep convective cloud systems (Rossow et al., 1996), using brightness temperatures from geostationary satellite measurements. Convective Systems (CS) are detected by cloud top temperatures <245 K, and Convective Clusters (CC) are identified as adjacent cloudy pixels with cloud top temperatures <220 K. The CS represents all high-level cloud systems, including isolated cirrus, deep convective towers, and anvil clouds. The dataset provides information related to location, size, shape and the properties of the clouds, including CC in case they are present, their location and properties. 15 The detailed vertical structure of clouds (and in particular the cloud top altitude) is studied using the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) instrument on board the CALIPSO satellite. The CALIPSO lidar measurements provide high-resolution vertical profiles of aerosols and clouds along satellite orbit tracks, with horizontal resolution of 333 m and vertical resolution between 20 30 m and 60 m depending on the altitude working at two-wavelength (532 nm, 1064 nm). The CALIPSO data are available from June 2006 and we used the lidar product level 1 (CAL LID L1), release ValStage1 and version V03 to analyze the total attenuated backscatter.
ACPD 11,2011 Thermal structure of intense convective clouds

a. GPS measurements and background climatology
Using all the GPS ROs collected from the CDAAC website from 1995 to 2009, we have created a gridded monthly climatology (with one-degree latitude-longitude resolution) containing the averaged bending angle and temperature profiles. This 5 climatology is used to compare with the profiles nearby to deep convection to identify anomalous bending angle and temperature behavior.
One key parameter that we analyze and discuss in this paper is the bending angle percentage anomaly with respect to the climatology (Biondi et al., 2011): where α conv is the bending angle profile for the convective systems and α clim is the climatological bending angle from the gridded reference. All of the analyses are based on the raw (unoptimized) bending angle from the atmPrf product and the temperature profiles form the wetPrf product.
b. Identifying co-locating GPS, ISCCP and CALIPSO measurements 15 Our objective is to identify GPS occultations which are co-located with large-scale intense deep convective systems (from ISCCP data), and furthermore closely overlap with the CALIPSO orbital track data. We then further identify cases where CALIPSO observations occur within the spatial radius of the CC, and within a time window of 2 h. This results in 53 cases where the GPS RO is co-located with CALIPSO, within a Convective Cluster; these 53 events form the basis for our analysis below. These convective systems span a wide range of latitudes (∼50 • N-S) and longitudes; 20 cases are within the latitude band 15 • N-S, and 38 within 30 • N-S. The cloud top altitude from CALIOP is defined as the maximum height where the total attenuated backscatter at 532 nm is larger than the background noise.

Results
We first highlight results from an individual case study, with the spatial structure illus-10 trated in Fig. 1. In this case the Convective System was located on the eastern Pacific basin on 2 February 2008, covering a broad region between ∼120 • W and 130 • W, polewards of 20 • N. Figure 1 shows the GOES brightness temperature difference between the 6.5 micron and 11 micron channels (which highlights high altitude clouds), together with the GPS RO tangent point (bold black line, centered near 21 • N; 123 • W) 15 and the CALIPSO track (bold white line). The GPS RO was acquired at 9:53:00 UTC in a Convective Cluster with minimum temperature of 210 K. CALIPSO overpassed the storm just 15 min later, at 10:08:00 UTC, 32 km from the RO tangent point. Figure 2 (left panel) shows the total attenuated backscatter from CALIOP along the orbit track in Fig. 1 as a function of altitude, highlighting a relatively flat cloud top near 20 12.5 km. The right panels in Fig. 2 show the corresponding bending angle anomaly from GPS RO (green line), together with the temperature profile from GPS RO (red line), the temperature profile from the ECMWF analysis (cyan), and the climatological temperature profile (blue line). The bending angle anomaly profile has a large positive spike centered near 12.5 km, very close to the altitude of the cloud top identified by 25 CALIPSO. The GPS temperature profile shows that the upper cloud layer corresponds to relative cold anomalies (compared to the background climatology), with a relative ACPD 11,2011 Thermal structure of intense convective clouds cold point near the cloud top (coincident with the maximum bending angle anomaly). In this case the maximum cold anomaly is approximately 9 K compared to the background climatology. The CALIOP attenuated backscatter shows the maximum cloud top at 12.64 km and the altitude of the top (averaged over the satellite track) at 12.16 km. Below the cloud (not shown), the temperature follows the climatological profiles up 5 to approximately 9 km; above 9 km the lapse rate within the cloud increases, reaching about 10 K km −1 just below the cloud top. There is a strong inversion near the cloud top, and above the cloud the temperature becomes warmer than the climatology, although this latter behavior is not observed in every case (or in the average statistics shown below). The temperature profile from the ECMWF analysis agrees with the profile from 10 GPS RO, but smears out the detailed vertical structure near the cloud top due to the coarse vertical resolution; in particular the sharp warming above the cloud is not well resolved.
Two further examples of temperature profiles acquired during intense convective systems with cloud tops reaching different altitudes are shown in Fig. 3. The first case 15 (Fig. 3a) was in January 2008 in Micronesia, where the cloud top altitude reached an altitude (16.8 km) near the tropical tropopause. This is associated with a sharp maximum in bending angle anomaly, and cold temperatures (up to 8 K) compared to the background climatology. The second case (Fig. 3b) was in April 2008 in the western Pacific basin, with a cloud top altitude near 14 km. In this case the bending angle 20 anomaly was more than 30 %, with a temperature anomaly near the cloud top of ∼12 K, and a strong inversion above. The ECMWF analysis produces a low resolution version of this detailed structure, and completely misses the strong inversion.
In all the cases analyzed here, the bending angle anomaly shows a spike in the UTLS (corresponding to a relative cold point in the temperature profile) with altitude lower than 25 the standard tropopause, frequently revealing a double tropopause structure. This may be one mechanism for the (infrequent) occurrence of double tropopauses in the tropics observed in GPS data (Schmidt et al., 2006;Randel et al., 2007). 11,2011 Thermal structure of intense convective clouds  Figure 4 shows a scatter plot of the altitude of the bending angle anomaly spike versus the maximum cloud top altitude from CALIOP, for each of the 53 cases considered here. This shows remarkably good agreement, with a correlation of 0.93 and a root mean square difference of 0.85 km. Since the GPS RO and CALIPSO track are never exactly co-located we also compare the bending angle profile with the altitude of the 5 cloud averaged over the satellite track (Fig. 5), and this results in a slightly lower correlation (0.85) and an apparent bias (wherein the averaged CALIPSO cloud top is ∼1 km lower than the bending angle anomaly).

ACPD
Using the cloud top altitude from CALIOP as a reference, we plotted the bending angle anomalies (Fig. 6), temperature profiles (Fig. 7), and lapse rate (−d T/d z) (Fig. 8), 10 as a function of altitude below and above the cloud top. These statistics are separated into cases where the cloud top altitude is below (22 cases) or above 14 km (31 cases), in order to identify any systematic differences with the depth of convection. The bending angle anomalies (Fig. 6) show a strong maximum near cloud top. The average bending angle percentage anomaly is 10 % (reaching above 30 % for the event shown 15 in Fig. 3b). The overall anomaly signals are stronger for profiles where the cloud tops are below 14 km (Fig. 6b). Temperature anomalies (Fig. 7) show relative coldest differences near the cloud top (consistent with the largest bending angle anomalies). The mean temperature anomaly is somewhat larger (∼8 K) for cloud tops below 14 km, compared to cloud tops above 14 km (∼3 K). In both cases, the temperature anomalies 20 decrease with altitude above the cloud top, and on average there are no significant anomalies above the cloud. Figure 8 compares the lapse rate (−d T/d z) structure associated with the convective systems, and for comparison the climatological background lapse rate is also included (black line). The colder temperatures within the clouds result in a significantly 25 larger lapse rate below cloud top, especially for the cases with cloud tops below 14 km (Fig. 8b). In these cases the observed convective system lapse rates often approach a wet adiabat, consistent with rapid undiluted ascent within the convective system. The wet adiabat lapse rate (e.g., 9 K km −1 at 300 hPa and −40 • C) is less than the 29102 ACPD 11,2011 Thermal structure of intense convective clouds dry adiabatic lapse rate of 9.8 K km −1 . Above the cloud top the lapse rate typically becomes negative, corresponding to the systematic inversion seen in the temperature profiles above cloud top (e.g., Figs. 2-3).

Conclusions
This study has utilized GPS RO temperature profiles to investigate the thermal struc-5 ture associated with deep convective systems, as identified in co-located ISCCP and CALIPSO data. The GPS data are well suited to this problem, as the radiowave signals are unaffected by clouds. The results show systematic anomalies in RO bending angle closely aligned with the top of convective systems (Fig. 4); it is important to appreciate that these results are derived from completely independent data sets (namely satellite radio occultation vs. space-based lidar measurements). These bending angle anomalies are associated with cold temperature anomalies over the cloud layer, peaking near cloud top. A schematic of the thermal structure of the convective systems derived from the results here is shown in Fig. 9. Temperatures decrease from the surface with a standard lapse rate until a few kilometers from the cloud top. At this altitude 15 the lapse rate increases (often approaching a moist adiabatic lapse rate, as quantified in Fig. 8) and revealing a relative temperature minimum (cold point) near the top of the convective system. Above the cloud top a strong inversion occurs, re-establishing a climatological temperature profile at higher altitudes. These upper level inversions are qualitatively similar to the lower level boundary layer inversions observed by GPS mea-20 surements in (Sokolovskiy et al., 2007). For cloud-top altitudes well below the tropical tropopause (∼17 km), the cloud-top inversion layers often result in a double tropopause structure, with the lower one at the altitude of the cloud top and the higher one at the climatological tropopause. As a note, there is reasonable agreement between the GPS RO temperature profiles and nearby ECMWF analysis (Figs. 2-3 There are several aspects of the results presented here which argue that the GPS bending angle and temperature results are robust aspects of intense convective systems. As noted above, completely independent measurements (lidar vs. GPS RO) identify cloud top altitudes with excellent agreement (Fig. 4). Further, very similar thermal structure is observed in many of the individual cases, especially for the 21 cases 5 with cloud tops below 14 km . In these cases, the lapse rate structure identifies a physically plausible behavior of approaching a moist adiabat at upper levels within the cloud system, consistent with rapid, unmixed ascent. As a note, the convective systems are also associated with enhanced water vapor over the altitudes of the cloud (up to several hundred ppmv according to collocated AIRS (Randel and Park, 2006) measurements). While these enhanced water vapor amounts will contribute to refractivity anomalies (Eq. 3), the corresponding changes in temperature are expected to be small (<0.5 K; S. Sokolovskiy, personal communication, 2011), and hence this is not an important source of bias in our results. The assumption of spherical symmetry inherent in the GPS refractive index retrieval may also be violated for the measure-15 ments near strong convective systems, although this probably does not introduce large biases.
It is interesting to compare the results of this study with the recent analysis of Chae et al. (2011), based on MLS temperatures combined with CALIOP cloud data. Note that there are significant differences between the temperature data used in these studies, 20 namely the use of MLS temperatures with vertical resolution of ∼3-4 km (versus <1 km for GPS RO), and much more extensive sampling for MLS (daily measurements spanning the tropics over 9 months vs. 53 individual cases studied here). The Chae et al. (2011) results reveal systematic cool temperatures below cloud top for cloudy regions, with average differences of ∼1-2 K. This vertical structure is similar to the GPS results; 25 the larger average temperature anomalies identified in our study may reflect a choice of isolating the most intense convective systems, or the higher vertical resolution of the GPS data. Chae et al. (2011) also identify systematic warm anomalies above cloud top, which are absent in our average results.  Fig. 9. Atmospheric thermal structure associated with intense convective systems, as observed by GPS RO. The blue line shows a background climatological temperature profile and the red line is the temperature profile during the convective systems.