Journal of Geophysical Research: Atmospheres

Intraseasonal temperature variability in the upper troposphere and lower stratosphere from the GPS radio occultation measurements

Authors


Corresponding author: B. Tian, Jet Propulsion Laboratory, California Institute of Technology, M/S 233-304, 4800 Oak Grove Dr., Pasadena, CA 91109, USA. (baijun.tian@jpl.nasa.gov)

Abstract

[1] In this study, we examine the detailed spatiotemporal patterns and vertical structure of the intraseasonal temperature variability in the upper troposphere and lower stratosphere (UTLS) associated with the Madden-Julian Oscillation (MJO) using the temperature profiles from the recent Global Positioning System radio occultation (GPS RO) measurements including the Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) mission. The MJO-related temperature anomalies in the UTLS are smaller near the equator (<0.6 K) than over the subtropics and extratropics (>1.2 K). Near the equator, the temperature anomalies exhibit an eastward tilt with height from the upper troposphere (UT) to the lower stratosphere (LS) and their magnitudes and signs are determined by the strength of convective anomalies and vertical pressure level. The subtropical temperature anomalies have similar magnitudes and patterns at a given location between the UT (250 hPa to 150 hPa) and the LS (150 hPa to 50 hPa) except for opposite signs that change around 150 hPa. The subtropical warm (cold) anomalies in the UT and cold (warm) anomalies in the LS are typically collocated with the subtropical positive (negative) tropopause height anomalies/cyclones (anticyclones) and flank or lie to the west of equatorial enhanced (suppressed) convection. We also compare the intraseasonal temperature variability in the UTLS related to the MJO between the GPS RO and Atmospheric Infrared Sounder (AIRS) measurements to highlight the new features of the GPS RO temperature anomalies and to evaluate the quality of the AIRS temperature in the UTLS considering the GPS RO temperature in the UTLS as the benchmark. Both AIRS and GPS RO have a very consistent vertical structure in the subtropical UTLS with a high correlation coefficient 0.92 and similar magnitudes. Both AIRS and GPS RO also show a generally consistent vertical structure of the intraseasonal temperature anomalies in the equatorial UTLS. However, GPS RO reveals many detailed fine-scale vertical structures of the equatorial temperature anomalies between 150 and 50 hPa that are not well captured by AIRS. Furthermore, the equatorial temperature anomalies are about 40% underestimated in AIRS in comparison to GPS RO, over the equatorial Indian and western Pacific Oceans for 250 hPa and over all longitudes for 100 hPa. The low sampling within the optically thick clouds and low vertical resolution near the tropopause may both contribute to these deficiencies of AIRS.

1. Introduction

[2] The upper troposphere and lower stratosphere (UTLS) is a coupling or transition layer ±5 km around the tropopause that shares the properties of both troposphere and stratosphere and is a topic of current active research [e.g., Fueglistaler et al., 2009; Gettelman et al., 2011]. Documenting and understanding the thermodynamic structure of the UTLS is essential to understand mechanisms of stratosphere-troposphere exchange [e.g., Holton et al., 1995], stratospheric dehydration and water vapor trends [e.g., Fujiwara et al., 2010; Mote et al., 1996; Schwartz et al., 2008b; Solomon et al., 2010; Stratospheric Processes and Their Role in Climate, 2000], and tropical thin cirrus cloud formation [e.g., Virts and Wallace, 2010; Virts et al., 2010].

[3] Observational studies of the UTLS thermal structure have traditionally been based on the global radiosonde network with large data sparse regions [e.g., Gettelman and Forster, 2002; Kiladis et al., 2001; Seidel et al., 2001; Seidel and Randel, 2006] or model-dependent analysis or reanalysis products with relatively low vertical resolution and attendant biases [e.g., Highwood and Hoskins, 1998; Hoinka, 1999; Kiladis et al., 2001; Randel et al., 2000]. A significant advance in the understanding of the UTLS thermal structure has recently come from Global Positioning System (GPS) radio occultation (RO) observations. The GPS RO systems can obtain temperature profiles in the UTLS with a high accuracy (less than 1 K), a high vertical resolution (∼200 m), under all-weather and all-cloud conditions, and with global coverage [Anthes et al., 2008; He et al., 2009]. The first GPS RO system, GPS Meteorology (GPS/MET) mission [Ware et al., 1996], operated between 1995 and 1997. The tropical tropopause thermal structure based on the GPS/MET data were performed by Nishida et al. [2000] and Randel et al. [2003]. Observations of the global tropopause became possible with the launch of the Challenging Minisatellite Payload (CHAMP) [Wickert et al., 2001] and Satellite de Aplicaciones Cientificas-C (SAC-C) missions [Hajj et al., 2004] in 2000 that increased the spatial and temporal sampling of the GPS RO data. Many studies have characterized the global tropopause parameters using the CHAMP and SAC-C data [e.g., Kishore et al., 2006; Randel and Wu, 2005; Schmidt et al., 2005, 2006, 2008]. The addition of the Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) (6 GPS RO satellites) [Anthes et al., 2008] as well as routine RO measurements from one of the Gravity Recovery And Climate Experiment (GRACE) twin satellites [Beyerle et al., 2005] in 2006 further increased the spatial and temporal sampling (including the diurnal cycle) of the GPS RO data to study the global tropopause. Using the COSMIC data, Son et al. [2011] recently documented the global spatiotemporal structure of the lapse-rate tropopause. The above studies using the GPS RO data have revealed many new features of the thermal structure of the UTLS and have significantly advanced our knowledge on this topic.

[4] The above studies, however, have focused mainly on the time-mean thermal structure of the UTLS instead of its variability. In particular, very few studies have analyzed the intraseasonal (30–90 day) variability of the UTLS thermal structure related to the Madden-Julian Oscillation (MJO) [Madden and Julian, 1971, 1972] using the GPS RO data. The MJO is the dominant form of the intraseasonal variability in the tropical atmosphere. It is characterized by slow (∼5 m s−1) eastward-propagating, large-scale oscillations in the tropical deep convection especially over the equatorial Indian and western Pacific Oceans during boreal winter (November–April) [Lau and Waliser, 2011; Zhang, 2005]. Accompanying the enhanced convection are low-level convergent winds and a planetary-scale circulation in the zonal plane (i.e., Walker circulation), with strong upper tropospheric zonal wind anomalies. Since its discovery, the MJO has continued to be a topic of significant interest due to its complex nature and profound influence on global weather and climate, such as monsoon onsets and breaks, El Niño-Southern Oscillation, tropical hurricanes, ocean biology, and atmospheric chemical composition [Lau and Waliser, 2011; Tian and Waliser, 2011].

[5] Many studies have documented the MJO's impact on the UTLS thermal structure. In their pioneering papers, Madden and Julian [1971, 1972] noticed that the MJO can impact the upper tropospheric (UT) wind and temperature through modulation of the direct convective forcing from the MJO. In the 1980s, based on analyzed circulation fields, many studies have demonstrated that the MJO could generate subtropical cyclonic and anticyclonic gyres in the UT as a Rossby-wave response to the equatorial convective forcing [e.g., Knutson and Weickmann, 1987; Weickmann et al., 1985]. Using the Microwave Sounding Unit (MSU)-derived UT temperature, Bantzer and Wallace [1996] showed that the MJO can impact the subtropical UT temperature through the subtropical UT cyclonic and anticyclonic gyres. Kiladis et al. [2001] documented the large-scale variability in tropopause height, temperature, and pressure at interannual and intraseasonal time scales relying on the radiosonde and National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis data. Based on the global vertical temperature profiles from the Atmospheric Infrared Sounder (AIRS) as well as contemporary reanalyses, Tian et al. [2006, 2010] documented the vertical thermal structure of the MJO from the surface to the lower stratosphere (LS). They found an eastward tilt with height of the intraseasonal temperature anomalies in the equatorial UTLS, with a Gill [1980]-like Kelvin-Rossby wave pattern of temperature anomalies at the 100-hPa level that propagated eastward with the deep convection anomaly. Virts and Wallace [2010] investigated the relationship between cirrus within the UTLS and the MJO, the annual cycle, and El Niño–Southern Oscillation (ENSO) using the first three years of data from the Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) mission. They demonstrated that the thermodynamic structure and variability of the UTLS is essential to understand mechanisms of tropical thin cirrus cloud formation.

[6] However, there are still many obvious limitations with the above studies. First, the spatial sampling is too sparse for studies based on radiosonde data. Second, there are model-dependent errors and vertical resolution is too coarse to resolve the fine vertical temperature structure near the tropopause for studies based on analysis/reanalysis data. Third, the vertical resolution of the previous satellite data may also be too coarse. For example, the MSU temperature data represent the mean UT (500–100 hPa) temperature and the vertical resolution of AIRS temperature data is only 1–2 km in the UTLS. Fourth, there may exist scene-dependent sampling bias in previous satellite data. For example, AIRS cannot measure the atmosphere when effective cloud amount is >70% which is common in the convective/cloudy phases of the MJO. Fifth, there may exist diurnal sampling biases in previous satellite data. For example, the two local time samples per day for AIRS cannot fully resolve the diurnal cycle.

[7] The main purpose of this work is to characterize and quantify the spatiotemporal patterns and vertical structure of the intraseasonal temperature variability in the UTLS related to the MJO using the temperature profiles from the recent GPS RO instruments. In addition to the high accuracy (less than 1 K), the high vertical resolution (∼200 m), and the all-weather- and all-cloud-condition sampling, the addition of COSMIC constellation [Anthes et al., 2008] in 2006 provided much-needed spatial and temporal data sampling (including the diurnal cycle) to study the intraseasonal temperature variability in the UTLS. Son et al. [2011] and Kim and Son [2012] have discussed the intraseasonal variance of the tropopause using the COSMIC data, but they did not address the detailed spatial and temporal patterns and vertical structure of the intraseasonal temperature variability in the UTLS that is the main topic of this study. We will also compare the intraseasonal temperature variability in the UTLS related to the MJO between the GPS RO and AIRS measurements to highlight the new features of the GPS RO results and to evaluate the quality of the AIRS temperature profiles in the UTLS considering GPS RO temperature profiles in the UTLS as the benchmark.

[8] The rest of this paper is organized as follows. Section 2 describes the data sets and MJO analysis methodology used in this study. Section 3 discusses the intraseasonal thermal structure in the UTLS from GPS RO and a comparison to that from AIRS. Section 4 summarizes the major findings of this paper.

2. Data and Methodology

2.1. GPS RO Data

[9] For this study, we use the temperature profiles retrieved at the Jet Propulsion Laboratory (JPL) using GPS RO data from the CHAMP, GRACE, SAC-C, and COSMIC missions for the period of 1 January 2006 to 31 December 2010. During this period, there are about 2000 temperature profiles per day that are distributed fairly uniformly in longitude including both oceans and land but with a relatively poor sampling over the tropics due to the high inclination orbits of the receiving satellites. With the assumption of local spherical symmetry, GPS RO amplitude and phase delay measurements can be used to retrieve bending angle profiles, which can be inverted directly to yield a vertical profile of microwave refractivity. The refractivity can be expressed as a function of temperature, pressure, and water vapor partial pressure. Above ∼10 km altitude in the tropics (lower at higher latitudes), the contribution from water vapor partial pressure can be neglected, and the refractivity can be written simply as N = 77.6 P/T, where P is the pressure in hPa and T is the temperature in K. The neglect of water vapor at 10 (15) km introduces a cold bias of about 0.2 (0.02) K in the tropics. This equation can be combined with the hydrostatic equation to derive the temperature and pressure profiles. Since the fractional refractivity error can be large in the upper stratosphere due to small bending angles, it is desirable to start the downward integration of the hydrostatic equation at ∼40 km using temperature from the global weather analysis such as the European Centre for Medium-Range Weather Forecasts (ECMWF). It has been shown through theoretical analysis that the temperature derived in this way is accurate to better than 1 K in the UTLS region (approximately 10–20 km) [Kursinski et al., 1997]. Differences between GPS RO temperatures and radiosondes and other satellite measurements are consistent with this estimate [Ho et al., 2007; Kuo et al., 2005; Rocken et al., 1997; Schwartz et al., 2008a; Wang et al., 2004]. For a more detailed description of the JPL retrieval algorithm, please refer to Hajj et al. [2002] and Ho et al. [2009].

[10] We note that the retrieved vertical profiles were obtained using the canonical transform method with a diffraction-limited vertical resolution of approximately 60 m [Gorbunov et al., 2004]. To reduce the noise in the measurements, a 200 m (1 km) vertical smoothing has been applied to each retrieved profile below (above) 20 km altitude. Thus in the UTLS region, the effective vertical resolution is about 200 m.

[11] To produce a gridded data set, we first interpolate each profile to a set of pressure levels (10–300 hPa with 25 hPa spacing below 100 hPa and 10 hPa spacing above 100 hPa). The horizontal resolution of the retrieved GPS RO temperature profiles is ∼200 km along track and 1 km cross track [Kursinski et al., 1997]. The temperature profiles are grouped and averaged daily in 5° × 10° latitude-longitude bins. We restrict our analysis to 300 hPa to 50 hPa (14 levels) because we are only interested in the UTLS process and the errors of GPS RO temperature retrievals become significant in the lower troposphere due to the contribution of water vapor to atmospheric refraction.

[12] To indicate the vertical movements of the tropopause related to the MJO, we use the tropopause height, temperature, and pressure derived from the GPS RO temperature profiles. The tropopause was defined as the lapse-rate tropopause according to the World Meteorological Organization (WMO) definition [World Meteorological Organization, 1957], i.e., the lowest level at which the lapse rate decreases to 2°C/km or less, provided that the average lapse rate between this level and all higher levels within 2 km does not exceed 2°C/km.

2.2. AIRS Data

[13] The AIRS/Advanced Microwave Sounding Unit (AMSU) sounding system [Aumann et al., 2003; Chahine et al., 2006] on the NASA Aqua platform has been operational since 1 September 2002. The AIRS and AMSU instruments are each cross-track scanning nadir sounders that are co-aligned and have a swath roughly 1650 km wide. The AIRS instrument is a 2378-channel grating spectrometer measuring infrared radiance at wavelengths in the range 3.7–15.4 μm with a horizontal resolution of about 13.5 km at nadir [Aumann et al., 2003]. The AMSU instrument is a 15-channel microwave radiometer with a horizontal resolution of about 45 km at nadir [Lambrigtsen, 2003]. Nine AIRS fields of view are contained within each AMSU field of view. The AIRS/AMSU geophysical retrieval method uses an iterative, least squares physical inversion of cloud-cleared infrared radiances, obtained from a combination of infrared and microwave observations. The algorithm is referred to as the AIRS/AMSU combined retrieval and was described by Susskind et al. [2006]. Following common practice, any discussion of AIRS in this work implicitly refers to the AIRS/AMSU system. The AIRS sounding system produces about 324,000 temperature profiles every day, separately by ascending and descending orbits. The horizontal resolution is about 45 km and the vertical resolution is about 1 km for AIRS temperature profile retrievals, referred to as Level-2 (L2) products [Susskind et al., 2006]. The AIRS Level-3 (L3) temperature profile product is the gridded averages of the AIRS L2 temperature profiles on horizontal 1°-latitude × 1°-longitude grids and 24 pressure levels from 1000 hPa to 1 hPa. The AIRS coverage is limited by the presence of optically thick clouds and can only retrieve temperature profiles for infrared cloud fraction (the product of emissivity and coverage) up to about 70% [Susskind et al., 2006]. This limitation induces an observation bias toward clear-sky scenes and spatially inhomogeneous sampling, low in cloudy regions such as the Intertropical Convergence Zone (ITCZ) but high in clear regions such as subtropics.

[14] For this study, we use the daily (arithmetic mean of ascending and descending nodes) AIRS V5 L3 atmospheric temperature profiles. For a direct comparison to the GPS RO results, we average the AIRS temperature profiles from its original 1°-latitude × 1°-longitude grids into 5°-latitude × 10°-longitude grids used for GPS RO data. We also use the same period (1 January 2006 to 31 December 2010) for both GPS RO and AIRS data. We also down-select the GPS RO and AIRS data from their original pressure levels to their common pressure levels (300, 250, 200, 150, 100, 70, 50 hPa) to plot their differences.

2.3. TRMM Data

[15] To indicate the spatial patterns and propagation characteristics of the equatorial convective anomalies associated with the MJO, we use the Tropical Rainfall Measuring Mission (TRMM) 3B42 rainfall data from 1 January 1998 to 31 March 2011. The TRMM 3B42 rainfall data are estimated from multiple satellites as well as gauge analyses where feasible at fine scales (0.25° × 0.25° and 3 hourly) [Huffman et al., 2007]. For this work, the rainfall data are averaged to daily on 2.5°-longitude × 2.5°-latitude grids.

2.4. Analysis Methodology

[16] For the MJO analysis and composite procedure, we use the multivariate empirical orthogonal function (EOF) method introduced by Wheeler and Hendon [2004], adopted widely by the MJO community [e.g., Waliser et al., 2009], and used in our previous studies [Li et al., 2010, 2012; Tian et al., 2010, 2011]. Briefly, the intraseasonal anomalies of daily data were obtained by removing the climatological-mean seasonal cycle and filtering via a 30–90-day band pass filter. Then, a composite MJO cycle (8 phases) was calculated by averaging the daily anomalies that occurred within each phase of the MJO cycle. The MJO phase for each day is determined by the Real-time Multivariate MJO (RMM) index (a pair of principal component time series called RMM1 and RMM2; available from 1974 to present at http://www.bom.gov.au/climate/mjo/). Only days with strong MJO activity (RMM12 + RMM22 >= 1) in boreal winter (November–April) from 1 January 2006 to 31 December 2010 are considered. The statistical assessment as to whether a composite mean at each point is different from zero is assessed by the two-sided Student's t test similar to the procedure of Tian et al. [2011].

3. Results

3.1. Horizontal Spatial Patterns of Intraseasonal Temperature Variability From GPS RO

[17] Figure 1 shows the composite boreal winter MJO cycle of latitude-longitude maps of GPS RO lapse-rate tropopause height anomalies. The contours show all anomalies while the color shades show only anomalies above 90% confidence limit. Red lines indicate positive values, whereas blue lines indicate negative values. The overlaid green lines denote TRMM rainfall anomalies with solid lines for positive values and dashed lines for negative values. We first discuss the TRMM rainfall anomalies. Positive rainfall anomalies indicate enhanced convection, while negative rainfall anomalies indicate suppressed convection. Figure 1 shows that the MJO convective anomalies are mainly confined to the equatorial Indian and western Pacific Oceans between 20°S and 20°N and they propagate eastward at a phase speed of around 5 m s−1 from the western/central Indian Ocean to the central Pacific. Phase 1 marks the onset of an MJO episode, with convection developing over the western and central equatorial Indian Ocean. The remnants of the previous active phase are visible along the date line near 10°S. In phase 2, these remnants have dissipated, while convection in the Indian Ocean has intensified and shifted eastward. The region of enhanced convection continues its eastward propagation through phase 4, weakening as it crosses the Maritime Continent because of interaction with local topography [Hsu and Lee, 2005]. In phase 5, the convective center emerges into the western Pacific, re-intensifies, and begins to split in two, with one center focused on the ITCZ and the other on the South Pacific Convergence Zone (SPCZ). In phases 7 and 8, only weak remnants of the convective center are still evident in the Pacific. The overall rainfall anomaly pattern in Figure 1 is consistent with that in Waliser et al. [2009, Figure 12] and many other studies.

Figure 1.

Composite boreal winter (November–April) MJO cycle of longitude-latitude maps of GPS RO lapse-rate tropopause height anomalies (km). The contours show all anomalies while the color shades show only anomalies above 90% confidence limit at the intervals of 0.1 km. Red lines indicate positive values, whereas blue lines indicate negative values. The overlaid green lines denote TRMM rainfall anomalies with solid lines for positive values and dashed lines for negative values. The contour interval is 1 mm day−1 with zero contours omitted.

[18] Statistically significant lapse-rate tropopause height anomalies can be found near the equator (from 10°S to 10°N) and over the subtropics and middle latitudes (20°N–50°N and 50°S–20°S). The equatorial lapse-rate tropopause height anomalies are generally less than 200 m and smaller than those over the subtropics and middle latitudes. The equatorial lapse-rate tropopause height anomalies are mostly evident (around 100–150 m) over the equatorial belt from southern America to Indian Ocean but are typically less than 100 m or almost negligible over the equatorial Pacific. The negative (positive) equatorial tropopause height anomalies are typically located to the west of the positive (negative) equatorial rainfall anomalies. The lapse-rate tropopause height anomalies over the subtropics and middle latitudes are generally large (>300 m) and are especially evident in the subtropical Pacific Ocean and eastern hemisphere. The subtropical tropopause height anomalies have a systematic relationship to the equatorial convective anomalies: the subtropical positive tropopause height anomalies typically flank or lie to the west of equatorial enhanced convection while the subtropical negative tropopause height anomalies flank or lie to the west of equatorial suppressed convection. These subtropical tropopause height anomalies propagate eastward at a phase speed of ∼5 m s−1, similar to the equatorial convective anomalies. Many studies have demonstrated that the subtropical tropopause height anomalies are driven by the subtropical UT cyclones or anticyclones that are generated by the equatorial convective anomalies as a Rossby-wave response [e.g., Hendon and Salby, 1994; Kiladis et al., 2001; Knutson and Weickmann, 1987; Tian et al., 2007]. The intraseasonal tropopause height anomalies from the tropics extend further poleward to high latitudes as a Rossby wave train.

[19] Figure 2 shows the composite boreal winter MJO cycle of latitude-longitude maps of GPS RO temperature anomalies at 250 hPa that represents the UT within the tropics (30°S–30°N) and middle latitudes up to around 45°N/S but the LS in high latitudes (see Figure 3). Figure 2 indicates that the intraseasonal temperature anomalies at 250 hPa are small (<0.4 K) near the equator (from 10°S to 10°N) but large (>1.2 K) over the subtropics (20°N–40°N and 40°S–20°S) and middle and high latitudes. Note that we only present the composite temperature anomalies, which are usually much smaller than those for individual events because the compositing procedure tends to reduce the signal amplitude. To demonstrate this point and to show the latitudinal difference of the intraseasonal temperature anomalies, Figure 3 shows the zonally averaged latitude-pressure cross-section of the standard deviation of the intraseasonal temperature anomalies from GPS RO. Figure 3 indicates that the intraseasonal temperature anomalies are small near the equator but large over the subtropics, middle and high latitudes at almost all levels. Over the equatorial region, the intraseasonal temperature anomalies peak around 80 hPa just over the boreal winter mean lapse-rate tropopause (around 100 hPa). Over the subtropics and middle and high latitudes, the intraseasonal temperature anomalies have double peaks, the first one just above the lapse-rate tropopause that decreases with latitude and the second one between 100 and 50 hPa.

Figure 2.

As in Figure 1 but for the GPS RO temperature anomalies (Kelvin) over the 250 hPa.

Figure 3.

Boreal winter zonally averaged latitude-pressure cross-section of the standard deviation of the intraseasonal temperature anomalies from GPS RO. The solid black line denotes the zonally averaged boreal winter mean lapse-rate tropopause pressure.

[20] Referring to Figure 2 again, the equatorial UT temperature anomalies are mostly found in the equatorial Indian Ocean and western Pacific, where the MJO-related convective anomalies are active. The subtropical or midlatitude UT temperature anomalies are especially evident in the Pacific Ocean and eastern hemisphere and mostly pronounced over the northern (winter) hemisphere. These equatorial and subtropical UT temperature anomalies have a systematic relationship to the equatorial convective anomalies. The equatorial temperature anomalies in the UT are typically collocated with the equatorial convective anomalies, with warm anomalies over the regions of enhanced convection and cold anomalies over the regions of suppressed convection. Over the subtropics, UT warm (cold) anomalies typically flank or lie to the west of equatorial enhanced (suppressed) convection. These UT temperature anomalies propagate eastward at a phase speed of ∼5 m s−1 similar to the equatorial convective anomalies. This Gill-like temperature anomaly pattern in the UT has been well documented in earlier studies [e.g., Bantzer and Wallace, 1996; Hendon and Salby, 1994; Tian et al., 2006]. Figures 1 and 2 also show that the subtropical UT temperature and tropopause height anomalies have very similar spatial and temporal patterns. Subtropical UT warm anomalies are generally located over regions of positive subtropical tropopause height and cold anomalies over regions of negative subtropical tropopause height. UT temperature anomalies are positively correlated tropopause height anomalies (with a correlation coefficient of 0.62).

[21] Composite boreal winter MJO cycle of latitude-longitude maps of GPS RO temperature anomalies over 150 hPa are plot but not shown. Over the equatorial and subtropical regions, the intraseasonal temperature anomalies at this level are much smaller (around 0.4 K) than those (around 1.2 K) in 250 hPa (Figure 1) and 100 hPa (shown later in Figure 4). Over the polar region, the intraseasonal temperature anomalies at this level (150 hPa) are similar to those (around 1.2 K) in 250 hPa (Figure 2).

Figure 4.

As in Figure 2 but for 100 hPa.

[22] Figure 4 shows the composite boreal winter MJO cycle of maps of GPS RO temperature anomalies at 100 hPa, representing the LS over the subtropics, middle and high latitudes and just around the tropopause over the equatorial region. Comparing Figures 2 and 4 shows that the spatial pattern of temperature anomalies at 100 hPa is almost identical to that at 250 hPa. However, over the equatorial and subtropical regions, the temperature anomalies have opposite signs between 100 hPa and 250 hPa. For example, temperature anomalies are large over the subtropics but small over the equatorial regions at both 100 and 250 hPa. The subtropical 100-hPa cold (warm) anomalies, in contrast to the 250-hPa warm (cold) anomalies, typically flank or lie to the west of equatorial enhanced (suppressed) convection and are generally collocated with the positive (negative) subtropical tropopause height anomalies. The subtropical temperature anomalies at both 100 and 250 hPa are highly correlated with the subtropical tropopause height anomalies with a negative correlation coefficient of −0.72 at 100 hPa and in contrast to a high positive correlation at 250 hPa. Over the regions of enhanced convection, equatorial cold anomalies are generally found at 100 hPa (Figure 4) in contrast to the equatorial warm anomalies at 250 hPa (Figure 2). However, the temperature anomalies over the equatorial regions seem to be much larger (0.6 K) at 100 hPa than that at 250 hPa. This Gill-like temperature anomaly pattern at 100 hPa has also been well documented in earlier studies [e.g., Bantzer and Wallace, 1996; Hendon and Salby, 1994; Tian et al., 2006]. Over the polar region, the intraseasonal temperature anomalies at this level (100 hPa) are similar to those (around 1.2 K) at 250 hPa (Figure 2).

[23] The composite MJO cycle of latitude-longitude maps of GPS RO temperature anomalies at 70 hPa representing the LS for all latitudes is presented in Figure 5. Comparing Figures 4 and 5, it can be seen that the subtropical, middle and high latitude temperature anomalies at 100 hPa and 70 hPa have almost the same pattern and signs. In contrast, the equatorial temperature anomalies at 70 hPa, especially over equatorial Indian and western Pacific Oceans, have different signs from those at 100 hPa and the same signs as those at 250 hPa. Equatorial warm anomalies at 70 hPa are generally found over the regions of enhanced convection, whereas equatorial cold anomalies at 70 hPa are generally found over the regions of suppressed convection. It seems the equatorial temperature anomalies propagate eastward with height from UT to LS and will be further discussed in the next subsection.

Figure 5.

As in Figure 2 but for 70 hPa.

3.2. Vertical Structure of Intraseasonal Temperature Variability From GPS RO

[24] Figure 6 shows the composite boreal winter MJO cycle of pressure-longitude cross-sections of GPS RO temperature anomalies over the equatorial region (averaged from 10°S to 10°N). The contours show all anomalies while the color shades show only anomalies above 90% confidence limits. The contour interval is 0.1 K with zero contours omitted. Red lines indicate positive values, whereas blue lines indicate negative values. The overlaid solid black and green lines denote TRMM rainfall and GPS RO lapse-rate tropopause height anomalies also over the equatorial region (scale at right). Figure 6 shows that the intraseasonal temperature anomalies in the equatorial UTLS exhibit an eastward tilt with height (around 75 hPa per 100 degree longitude) from the UT to the LS, which is a well-documented feature of the MJO. The signs of the equatorial temperature anomalies are determined by both vertical pressure level and convective anomalies. Over the peaks of enhanced convection, the temperature anomalies change signs near 150 hPa and 90 hPa. As a result, warm anomalies are generally found in the UT below 175 hPa and in the LS above 90 hPa, whereas cold anomalies are generally located in the UT between 150 and 100 hPa, with opposite signs in temperature for the peaks of suppressed convection. In between, the equatorial temperature anomalies vary depending on the strength of convective anomalies and vertical pressure level. This vertical structure is consistent with the spatial and temporal patterns at different levels shown above in Figures 2, 4, and 5, and has been well documented by many previous studies based on various data sources [e.g., Kiladis et al., 2001, 2005; Schwartz et al., 2008b; Tian et al., 2006, 2010; Virts and Wallace, 2010; Virts et al., 2010]. However, GPS RO reveals many detailed fine-scale vertical structures of the equatorial temperature anomalies between 150 and 50 hPa that are not well captured by previous data sets. A detailed comparison between GPS RO and AIRS temperature structure will be discussed in subsection 3.3.

Figure 6.

Composite boreal winter MJO cycle of pressure-longitude cross-sections of GPS RO temperature anomalies (Kelvin) over the equatorial region (10°S–10°N). The contours show all anomalies while the color shades show only anomalies above 90% confidence limit. The contour interval is 0.1 K with zero contours omitted. Red lines indicate positive values, whereas blue lines indicate negative values. The overlaid solid black and green lines denote TRMM rainfall and GPS RO lapse-rate tropopause height anomalies also for over the equatorial region (scale at right).

[25] The composite boreal winter MJO cycle of pressure-longitude cross sections of GPS RO temperature anomalies over the northern subtropics (i.e., averaged from 20°N to 40°N) is shown in Figure 7. The overlaid solid green lines denote the GPS RO tropopause height anomalies for the same latitudes, but the overlaid solid black lines denote the TRMM rainfall anomalies for the equatorial region (averaged from 10°S to 10°N). Comparing Figures 6 and 7 indicates that the subtropical temperature anomalies (>1.2 K) are much larger than those near the equator (0.6 K), which is consistent with Figures 2, 4, and 5 shown above. Figure 7 also indicates that the subtropical temperature anomalies are especially strong in the Pacific Ocean and eastern hemisphere and have similar magnitudes and patterns between the UT (250 hPa to 150 hPa) and the LS (150 hPa to 50 hPa) except for opposite signs. The sign changes occur around 150 hPa, the boreal winter mean lapse-rate tropopause pressure in the subtropics (see Figure 3). The subtropical temperature anomalies have a systematic relationship to the equatorial convective anomalies and the subtropical tropopause height anomalies. The subtropical LS cold anomalies together with UT warm anomalies are generally collocated with the positive subtropical tropopause height and typically flank or lie to the west of equatorial enhanced convection. In contrast, the subtropical LS warm anomalies together with UT cold anomalies are generally collocated with the negative subtropical tropopause height anomalies and typically flank or lie to the west of equatorial suppressed convection. They propagate eastward at a phase speed of ∼5 m s−1 similar to the equatorial convective anomalies. As a result, the subtropical tropopause height anomalies are positively correlated with the subtropical UT temperature anomalies, but negatively correlated with the subtropical LS temperature anomalies. These results are consistent with the spatial and temporal patterns in Figures 2, 4, and 5 shown above and indicate that the subtropical UTLS temperature anomalies result mainly from the subtropical UTLS cyclones or anticyclones that are generated by the equatorial convective anomalies as a Rossby-wave response.

Figure 7.

As in Figure 6 but over the northern subtropics (20°N–40°N). The overlaid solid green lines denote GPS RO lapse-rate tropopause height anomalies for the northern subtropics (20°N–40°N) (scale at right). The overlaid solid black lines denote TRMM rainfall anomalies for the equatorial region (10°S–10°N) (scale at right).

[26] Similar results are also found for the pressure-longitude cross sections of temperature anomalies for the southern subtropics between 40°S and 20°S (not shown). The temperature anomalies in the southern subtropics, however, are much weaker than those in the northern subtropics, consistent with the spatial maps discussed above.

3.3. A Comparison Between GPS RO and AIRS Temperature Anomalies

[27] The spatiotemporal patterns and vertical structure of the MJO from the GPS RO data are generally consistent with previous studies based on radiosonde, reanalyses, and satellite data, but there are several new features of the GPS RO results that have not been captured by previous data sets. In this section, we compare the intraseasonal temperature variability in the UTLS related to the MJO between AIRS and GPS RO to highlight the new features of the GPS RO temperature anomalies and to evaluate the quality of AIRS temperature in the UTLS considering GPS RO temperature in the UTLS as the benchmark. The composite MJO cycle of latitude-longitude maps of AIRS temperature anomalies at 250 hPa, similar to Figure 2, is shown in Figure 8. Comparing Figures 2 and 8 indicates that the spatial and temporal patterns of the intraseasonal temperature anomalies at 250 hPa are very similar between AIRS and GPS RO. For example, both AIRS and GPS RO show that the temperature anomalies are larger over the subtropical, extratropical and polar regions but smaller near the equator. Both AIRS and GPS RO show that the equatorial warm anomalies are typically collocated with the equatorial enhanced convection, while the subtropical warm anomalies flank or lie to the west of the equatorial enhanced convection. Furthermore, the magnitudes of the temperature anomalies, especially over the subtropical and extratropical regions, seem to be similar between AIRS and GPS RO. The subtropical and extratropical temperature anomaly differences between AIRS and GPS RO are much smaller compared to the anomaly themselves (<20%) and without a consistent pattern (not shown). Nevertheless, relatively large temperature anomaly differences (∼±0.3 K) between AIRS and GPS RO (not shown), compared to temperature anomalies themselves (∼0.6 K), are found over the equatorial Indian and western Pacific Oceans.

Figure 8.

As in Figure 2 but for AIRS.

[28] To quantify the difference in the magnitude of the AIRS and GPS RO temperature anomalies, we calculated the ratio of the AIRS temperature anomaly standard deviation to the GPS RO temperature anomaly standard deviation and the results are shown in Figure 9. Figure 9 indicates that the AIRS and GPS RO temperature anomalies are generally comparable (differences < 10%) in magnitude over most regions of the globe, especially over the subtropics and middle latitudes. However, the AIRS temperature anomalies are significantly (∼40%) underestimated in comparison to GPS RO over the equatorial Indian and western Pacific Oceans, the southern high latitudes, and some areas of the northern middle and high latitudes.

Figure 9.

The ratio of the AIRS and GPS RO temperature anomaly standard deviation at 250 hPa. The boxes in the map denote the subtropical Pacific Ocean and eastern hemisphere (20°N–40°N, 30°E–240°E) and the equatorial Indian and western Pacific Oceans (10°S–10°N, 60°E–150°E) used for Figures 1012.

[29] To demonstrate the consistency of the AIRS and GPS RO temperature anomalies over the subtropical Pacific and eastern hemisphere (20°N–40°N, 30°E–240°E), Figure 10a shows the scatterplot of the AIRS and GPS RO temperature anomalies over this region. Figure 10a further indicates that the AIRS and GPS RO temperature anomalies are very close in terms of sign and magnitude with a high correlation efficient of 0.94. To demonstrate the difference of the AIRS and GPS RO temperature anomalies over the equatorial Indian and western Pacific Oceans (10°S–10°N, 60°E–150°E), Figure 10b shows the scatterplot of the AIRS and GPS RO temperature anomalies over this region. Figure 10b further demonstrates that the AIRS temperature anomalies are systematically smaller (∼40%) than the GPS RO temperature anomalies with a correlation efficient of 0.71.

Figure 10.

The scatterplot and linear regression of GPS RO and AIRS temperature anomalies at 250 hPa over (a) the subtropical Pacific Ocean and eastern hemisphere (20°N–40°N, 30°E–240°E) and (b) the equatorial Indian and western Pacific Oceans (10°S–10°N, 60°E–150°E).

[30] To further demonstrate the different characteristics of the 250-hPa AIRS temperature anomalies over northern subtropical Pacific/Asia and equatorial Indian/western Pacific Oceans in comparison to the GPS RO temperature anomalies, Figure 11 shows the probability distribution function (PDF) of AIRS and GPS RO temperature anomalies at 250 hPa over these two regions. Figure 11 indicates that over the subtropical Pacific/Asia, AIRS and GPS RO temperature anomalies have almost the same distribution, that is, a flat and wide distribution with longer tails at the large temperature anomalies (±2 K). This is consistent with Figure 10a that the AIRS and GPS RO temperature anomalies are very close with each other over this region. However, over the equatorial Indian and western Pacific Oceans, AIRS and GPS RO temperature anomalies have a sharp and narrow distribution with strong peaks at the zero temperature anomalies. Most importantly, the distribution of AIRS temperature anomalies is much sharper and narrower than that of the GPS RO temperature anomalies, which is consistent with Figure 10b that the AIRS temperature anomalies are systematically smaller (∼40%) than the GPS RO temperature anomalies over the equatorial Indian and western Pacific Oceans.

Figure 11.

The PDF of AIRS and GPS RO intraseasonal temperature anomalies at 250 hPa over two selected regions: (a) northern subtropical Pacific and Asia (20°N–40°N; 30°E–240°E) and (b) equatorial Indian and western Pacific Oceans (10°S–10°N; 60°E–150°E).

[31] To illustrate the reason for the underestimation of AIRS temperature anomalies in comparison to GPS RO over the equatorial Indian and western Pacific Oceans, Figure 12 shows the scatterplots of the temperature anomaly differences between AIRS and GPS RO (AIRS–GPS RO) and TRMM rainfall anomalies over this region. Figure 12 indicates that temperature anomaly differences and rainfall anomalies are highly negatively correlated with a correlation coefficient of −0.73. This demonstrates that the underestimation of AIRS temperature anomalies at 250 hPa over the equatorial Indian and western Pacific Oceans is very likely due to the low sampling of AIRS under the presence of optically thick clouds in the deep convective cloud systems, which is a well known weakness of the infrared sounding instruments like the AIRS [Susskind et al., 2006].

Figure 12.

The scatterplot and linear regression of AIRS/GPS RO temperature anomaly differences (AIRS–GPS RO) and TRMM rainfall anomalies over the equatorial Indian and western Pacific Oceans (10°S–10°N, 60°E–150°E).

[32] Similar diagrams to Figures 812 but for 100 hPa have also been produced but not shown here. In general, similar conclusions for 250 hPa can also applied to 100 hPa with the following caveats. First, both the temperature anomalies and their differences change signs from 250 to 100 hPa over the equatorial and subtropical regions. Second, the AIRS temperature anomalies are smaller by ∼40% in comparison to GPS RO over all the longitudes of the equatorial belts at 100 hPa instead of confined to the equatorial Indian and western Pacific Oceans at 250 hPa. Third, the correlation between the temperature anomaly differences and rainfall anomalies over the equatorial region are much weaker at 100 hPa (0.3) than at 250 hPa (−0.73). This indicates that this underestimation of AIRS temperature anomalies at 100 hPa is due likely to other factors, such as coarse vertical resolution, in addition to the limited sampling of AIRS in the equatorial deep convective clouds.

[33] To further highlight the vertical and latitudinal structure of the AIRS and GPS RO temperature anomaly differences, Figure 13 shows zonally averaged latitude-pressure cross-section of the standard deviation of the intraseasonal temperature anomalies from AIRS and can be compared directly to Figure 3 from GPS RO. Comparing these two figures indicates the overall vertical and latitudinal structure of the intraseasonal temperature anomalies from AIRS are generally consistent with that from GPS RO. However, the intraseasonal temperature anomalies from AIRS are consistently smaller than those from GPS RO, especially over the tropics and the southern middle and high latitudes. In particular, at 80 hPa just above the lapse-rate tropopause over the equatorial region, the intraseasonal temperature anomalies from AIRS (∼0.7 K) are significantly smaller than those from GPS RO (∼1.5 K). This indicates that the coarse vertical resolution of AIRS, which cannot resolve the fine structure of the equatorial tropopause, may contribute to the underestimation of AIRS temperature anomalies at this level.

Figure 13.

As in Figure 3 but for AIRS.

[34] To further highlight the similarity and difference between AIRS and GPS RO over the equatorial region, the composite MJO cycle of pressure-longitude cross sections of temperature anomalies over the equatorial region (10°S–10°N) from AIRS are shown in Figure 14, which can be compared directly to Figure 6 from GPS RO. Comparing Figures 6 and 14 indicates that both AIRS and GPS RO show a consistent vertical structure of the intraseasonal temperature anomalies in the equatorial UTLS. The linear correlation between AIRS and GPS RO is very high at 250 and 100 hPa (∼0.86) (calculated based on the corresponding points between Figures 6 and 14) but very low around 150 hPa due probably to the small temperature anomalies there. These results indicate that the AIRS data are useful to study the intraseasonal temperature variability in the equatorial UTLS. However, GPS RO reveals many detailed fine-scale vertical structures of the equatorial temperature anomalies between 150 and 50 hPa that are not well captured by AIRS. Furthermore, the equatorial temperature anomalies are about 40% underestimated in AIRS in comparison to GPS RO mainly over the Indian Ocean at 250 hPa and over all longitudes for 100 hPa. As discussed above, the low sampling within the optically thick clouds and low vertical resolution of AIRS may both contribute to these deficiencies. Thus, when AIRS data are used to evaluate climate models or reanalyses in the equatorial UTLS, the low sampling within the optically thick clouds and low vertical resolution near the tropopause of AIRS data should be considered.

Figure 14.

As in Figure 6 but for AIRS.

[35] To further highlight the similarity and difference between AIRS and GPS RO over the subtropical region, the composite MJO cycle of pressure-longitude cross sections of temperature anomalies over the northern subtropics (20°N–40°N) from AIRS are shown in Figure 15, which can be compared directly to Figure 7 from GPS RO. Comparing Figures 7 and 15 demonstrates that AIRS and GPS RO have a consistent vertical structure in the subtropical UTLS with a high correlation coefficient 0.92 (calculated based on the corresponding points between Figures 7 and 15) and similar magnitudes. Their differences are small (<20%) compared to the anomalies themselves. It seems that the low sampling within the optically thick clouds and low vertical resolution of AIRS do not affect the intraseasonal temperature anomalies over the subtropics due probably to the lack of optically thick clouds and strong vertical movement of the tropopause over the subtropics. Similar results can be found for the southern subtropics but not shown. These results indicate that AIRS is a reliable data set to study the intraseasonal temperature variability in the UTLS over the subtropics and middle latitudes.

Figure 15.

As in Figure 7 but for AIRS.

4. Summary and Conclusions

[36] We have examined the detailed spatiotemporal patterns and vertical structure of the intraseasonal temperature variability in the UTLS associated with the MJO, using the state-of-the-art temperature profiles from the recent GPS RO instruments. The GPS RO systems can measure the temperature profiles in the UTLS with a high accuracy (errors less than 1 K), a high vertical resolution (∼200 m), under all-weather and all-cloud conditions, and with global coverage [Anthes et al., 2008; He et al., 2009]. The addition of COSMIC and GRACE mission data provides much improved spatial and temporal sampling, including the diurnal cycle, to study the intraseasonal temperature variability. The findings of this study based on the GPS RO data are summarized below. The intraseasonal temperature anomalies in the UTLS are smaller near the equator (<0.6 K) but larger (>1.2 K) over the subtropical, extratropical and polar regions. Near the equator, the temperature anomalies exhibit an eastward tilt with height from the UT to the LS depending on strength of convective anomalies and vertical pressure level. Over the subtropics, temperature anomalies are especially evident in the Pacific Ocean and eastern hemisphere and most pronounced over the northern (winter) hemisphere. Over these regions, temperature anomalies have similar magnitudes and patterns but opposite signs between the UT (250 hPa to 150 hPa) and the LS (150 hPa to 50 hPa). The sign changes occur around 150 hPa, the boreal winter mean lapse-rate tropopause pressure in the subtropics. The subtropical temperature anomalies have a systematic relationship to the equatorial convective anomalies and the subtropical tropopause height anomalies. The subtropical LS cold anomalies together with UT warm anomalies are generally collocated with the positive subtropical tropopause height anomalies and typically flank or lie to the west of equatorial enhanced convection. In contrast, the subtropical LS warm anomalies together with UT cold anomalies are generally collocated with the negative subtropical tropopause height anomalies and typically flank or lie to the west of equatorial suppressed convection. They propagate eastward at a phase speed of ∼5 m s−1 similar to the equatorial convective anomalies. As a result, the subtropical tropopause height anomalies are positively correlated with the subtropical UT temperature anomalies, but negatively correlated with the subtropical LS temperature anomalies. These results indicate that the subtropical UTLS temperature anomalies result mainly from the subtropical UTLS cyclones or anticyclones that are generated by the equatorial convective anomalies as a Rossby-wave response.

[37] The current GPS RO results are generally consistent with previous studies based on radiosonde, reanalyses, and satellite data, but there are several new features of the GPS RO results that have not been captured by previous data sets. We also compare the intraseasonal temperature variability in the UTLS related to the MJO between AIRS and GPS RO to highlight the new features of the GPS RO temperature anomalies and to evaluate the quality of AIRS temperature in the UTLS considering GPS RO temperature in the UTLS as the benchmark. Both AIRS and GPS RO show a consistent spatiotemporal pattern and vertical structure of the intraseasonal temperature anomalies in the UTLS including larger anomalies over subtropics/extratropics and smaller anomalies near the equator, an eastward tilt of the equatorial temperature anomalies with height from the UT to the LS, contrasting subtropical temperature anomalies between UT and LS, and the same relationships to the equatorial convective anomalies and subtropical tropopause anomalies. In particular, AIRS and GPS RO have a very consistent vertical structure in the subtropical UTLS with a high correlation coefficient 0.92 and similar magnitudes. Their differences are small (<20%) compared to the anomalies themselves. These results indicate that AIRS is a reliable data set to study the intraseasonal temperature variability in the UTLS over the subtropics and middle latitudes. Both AIRS and GPS RO also show a generally consistent vertical structure of the intraseasonal temperature anomalies in the equatorial UTLS. However, GPS RO reveals many detailed fine-scale vertical structures of the equatorial temperature anomalies between 150 and 50 hPa that are not well captured by AIRS. Furthermore, the equatorial temperature anomalies are about 40% underestimated in AIRS in comparison to GPS RO, over the equatorial Indian and western Pacific Oceans for 250 hPa and over all longitudes for 100 hPa. The low sampling within the optically thick clouds and low vertical resolution of AIRS may both contribute to these deficiencies. These results indicate that AIRS is a useful data set to study the intraseasonal temperature variability in the UTLS near the equator, but with a caution of biases caused by the low sampling within the optically thick clouds and low vertical resolution near the tropopause.

[38] The intraseasonal temperature variability in the UTLS from GPS RO presented here can be used as a satellite-based benchmark diagnostic to help evaluate the Coupled Model Intercomparison Project phase 5 (CMIP5) models. The intraseasonal temperature variability in the UTLS from AIRS can also serve such purpose given their global dense coverage but with a caution of their biases near the equator caused by the low sampling within the optically thick clouds and low vertical resolution near the tropopause.

Acknowledgments

[39] This research was performed at Jet Propulsion Laboratory (JPL), California Institute of Technology (Caltech), under a contract with National Aeronautics and Space Administration (NASA). It was supported jointly by the Research and Technology Development program and Atmospheric Infrared Sounder (AIRS) project at JPL. We thank three anonymous reviewers for their constructive comments that helped improve the quality of this paper.