Mesoscale convection system and occurrence of extreme low tropopause temperatures : observations over Asian summer monsoon region

The present study examines the process of how tropospheric air enters the stratosphere, particularly in association with tropical mesoscale convective systems (TMCS) which are considered to be one of the causative mechanisms for the observation of extremely low tropopause temperature over the tropics. The association between the phenomena of convection and the observation of extreme low tropopause temperature events is, therefore, examined over the Asian monsoon region using data from multiple platforms. Satellite observations show that the area of low outgoing long wave radiation (OLR), which is a proxy for the enhanced convection, is embedded with high altitude clouds top temperatures (≤193 K). A detailed analysis of OLR and 100 hPa temperature shows that both are modulated by westward propagating Rossby waves with a period of ∼15 days, indicating a close linkage between them. The process by which the tropospheric air enters the stratosphere may, in turn, be determined by how the areas of convection and low tropopause temperature (LTT) i.e. T ≤ 191 K are spatially located. In this context, the relative spatial distribution of low OLR and LTT areas is examined. Though, the locations of low OLR and LTT are noticed in the same broad area, the two do not always overlap, except for partial overlap in some cases. When there are multiple low OLR areas, the LTT area generally appears in between the low OLR areas. Implications of these observations are also discussed. The present analysis also shows that the horizontal mean winds have a role in the spatial distribution of low OLR and LTT. Correspondence to: A. R. Jain (atma.jain@gmail.com)


Introduction
To explain the observations of low water vapor mixing ratio over southern England, Brewer (1949) suggested that the observed stratospheric air must have passed through the tropical tropopause where the temperature is low enough to dry the air by freezing the water vapor.Extreme low tropopause temperatures are generally observed over western Pacific during November-March.Newell and Stewart (1981) have suggested that this region of low tropopause temperature expands in the north-west direction towards the Indian tropics during the spring and monsoon seasons (April-September).Jain et al. (2006) analyzed the observations of intense radiosonde/GPS-sonde campaigns, which were carried out over the Indian tropics, including the Bay of Bengal and adjacent area during monsoon and post-monsoon season, and noted that extreme low tropopause temperature (T CPT ≤ 191 K) do occur over these regions.Such extreme low tropopause temperatures are referred to, hereafter, as LTT.
Mesoscale tropical convective systems (TMCS) are considered to be one of the prime mechanisms for the generation of tropical LTT.Injection of cold air by convective overshoots followed by irreversible mixing was suggested by Danielsen (1982).Sherwood et al. (2003) have indicated that diabatic cooling by convective system just below the cold point tropopause (CPT) chills the air to a temperature lower than the CPT temperature, thus establishing a Published by Copernicus Publications on behalf of the European Geosciences Union.
A. R. Jain et al.: Mesoscale convection system and occurrence of extreme low tropopause temperatures new CPT altitude.According to them the most likely explanation of diabatic cooling is the transport of cold air by penetrative clouds.Kim and Dessler (2004) have also argued that the mixing of cold air from overshooting convection is a potentially reasonable mechanism for the cooling of tropical tropopause layer (TTL), and this mechanism may also be important for determining the humidity in this layer.More recently, Pommereaue and Held (2007) have examined the impact of convection on the thermal structure of TTL using a series of radiosonde and weather S-band radar observations carried out during the HIBISCUS campaign in south Brazil in February 2004.They have shown that the injection of cold air explains partly the cooling of TTL associated with convection to which radiation, adiabatic lofting and Rossby and Kelvin Waves may also contribute.They have also noted that the afternoon overshoot and turbulent mixing is the only process that allows irreversible mixing.
It is apparent from the above discussion that the convection activity has a significant role in cooling the tropical tropopause region.However, what is currently not well understood is that how the tropospheric air penetrates into the lower stratosphere following a convection event.The process could be determined by the relative locations of the areas of convection and LTT.In case, these two phenomena are collocated, the injection of cold air by irreversible mixing of convection overshoots, as proposed by Danielsen (1982), may be the relevant mechanism.In that case, the air would be dehydrated while passing through the cold tropopause region provided this air stays for sufficient time in the cool tropopause region for the ice crystals to be formed and fallen out.Pommereaue and Held (2007) have shown that the injection of cold air by convection events in the upper troposphere and lower stratosphere (i.e.stratospheric fountain) is not an isolated event, but common over the land convection area.However, if the areas of convection and LTT are spatially separated, the entry of air from troposphere to lower stratosphere may not be a simple one-stage process.In that case the convective air may be rising into TTL up to a certain height (say level of neutral buoyancy) and from that point it may be rising slowly into the stratosphere (Holton and Gettelman, 2001;Gettelman et al., 2001;and Sherwood and Dessler, 2001).It is possible that some of the air rising into the stratosphere may return into the troposphere through the region of low temperature.In these cases the air in the upper troposphere region, close to tropopause, would be dry as inferred by Jain et al. (2006).It is, therefore, of interest to know how the areas of convection and LTT are spatially distributed.An important question is that if the convective activity has a role in cooling the tropopause then how the areas covered by these two phenomena (i.e.convective system and LTT) could be spatially separated.The cooling of the upper troposphere and lower stratosphere (UTLS) by convection takes a finite time approximately of ∼4-10 h or even longer (Kim and Dessler, 2004).Prevailing horizontal winds, in this time interval, could spatially redistribute the regions of convec-tive activity and of colder UTLS.This spatial redistribution would basically depend on the strength of background wind.Other mechanisms such as the wave activity associated with convection could also spread the cooling over substantial distances (Kuang and Bretherton, 2004).
In the present study the focus is on the features of the Asian summer monsoon region.There are, however, very limited observations existing over these regions during monsoon season.Cornford and Spavins (1973) measured the heights of cumulonimbus clouds that form during the premonsoon season (April-June) in north-east India, using air born radar, cameras and horizontal gyroscope.These clouds are found to extend up to ∼65 000 ft (∼20 km).Cornford and Spavins (1973) have noted that the cumulonimbus clouds exceeding 60 000 ft (∼18 km) occur on most of the measurement days.Bhat (2003) analyzed temperature data of active and weak phase of convection from observations taken on board ocean research vessel (ORV) Sagar Kanya (SK) during the Bay of Bengal Monsoon Experiment (BOBMEX) campaign (20 July-29 August 1999) over the Bay of Bengal and noted that temperature near the tropopause height during the active phase of convection is cooler by ∼2 K than that observed during the weak phase.This provided a direct evidence of the influence of convection activity on the temperature of the tropopause region.Kumar (2006) made use of observations by the Indian Mesosphere-Stratosphere-Troposphere (MST) Radar located at Gadanki (13.45 • N, 79.18 • E), a tropical station in India (Rao et al., 1995) and reported (a) enhanced mass fluxes across the tropopause and (b) enhanced gravity wave activity in the lower stratosphere during the passage of deep convection.
The main objective of the present study is twofold-first to examine the association between the phenomena of convection and the observation of LTT events and secondly to examine the spatial distribution of these two phenomena, which has a bearing on understanding of the process of troposphere -to-stratosphere transport of minor constituents, such as water vapor.Observations carried out on multiple platforms are, therefore, used for the present study.Daily observations are used to capture the fine scale features of OLR and LTT.

Databases and methodology
In the present study the (i) radiosonde/GPS-sonde observations of atmospheric temperature taken during BOBMEX period and (ii) data of cloud top temperature (CTT) and OLR obtained from the satellite INSAT-1D launched by the Indian Space Research Organization (www.isro.org)are used.Radiosonde measurements of atmospheric temperature, during BOBMEX, were carried out at seven stations (shown in Fig. 1).The radiosondes used in this campaign are IMD -MK-III which has a height resolution of ∼300 m and the expected error in temperature measurements is ∼1.4K at the level of 100 hPa (Jain et al., 2006).The GPS-sonde flights over the Bay of Bengal were carried out during BOBMEX onboard the ORV Sagar Kanya (Bhat et al., 2001;Bhat et al., 2002).The GPS-sonde, Vaisala model RS80-15G, used in this campaign was a standard instrument.The expected height resolution and error in temperature measurement for these instruments, near the level of tropopause, are better than 50 m and less than 1 K, respectively.For this study the altitude of lowest temperature in the vertical profile (i.e.cold point tropopause (CPT)) has been taken as the level of tropopause.Radiosonde and GPS-sonde flights, for which CPT could be detected unambiguously, are used in the present study.
In addition to radiosonde observations, data from the geostationary satellite INSAT-1D, sub satellite point (0 • N, 74 • E) for BOBMEX campaign, are used to obtain spatial distribution of OLR and CTT.This satellite was operated from May 1990 to May 2002.It had a payload called the very high resolution radiometer (VHRR with two channels, viz.visible (0.55-0.75 µm) and infrared channel (10.5-12.5 µm).The resolution of visible image is 2.75 km and the same for infrared image is 11 km.In addition to INSAT-1D data, high resolution imagery and cloud top temperature (CTT) are also obtained using International satellite cloud climatology project (ISCCP)-DX data for the period of BOBMEX observation.The ISCCP-DX data set is produced by the analysis of infrared (IR) and visible (day time only) radiances from weather satellite images with pixels about 5 km across (on the average), sampled at about 30 km and at about 3 h intervals (Rossow and Pearl, 2007).In these data sets, biases in (detectable) cloud amounts have been reduced to ≤0.05, except the summertime polar-regions, where the bias may still be ∼0.10.Biases in CTT are ≤2 K for lower level clouds and ≤4 K for optically thin, upper level clouds, except when they occur over lower-level clouds (Rossow and Schiffer, 1999).
Observations by the satellite Kalpana-1 for three days during July-August 2008, have been used to distinguish between the convective cloud system and the cirrus clouds.This satellite has a meteorological payload VHRR which provides images in visible (0.55-0.75 µm), infrared (10.5-12.5 µm) and water vapor (5.7-7.1 µm) channels.The spatial resolution at the sub-satellite point is 2 km in visible channel and 8 km in infrared and water vapor channels.The infrared channel is, sometimes, referred as "clean channel" as it is comparatively less affected by the water vapor, whereas the water vapor channel is called "dirty channel", since this channel is sensitive to the presence of tropospheric water vapor.Observation of nearly the same temperature at the same pixels/areas by the clean and dirty channels indicates the presence of thick convective clouds (see http://www.nrlmry.navy.mil/).In addition to above satellite observations daily mean OLR data for the period 24 June-15 August 2002 are used from NOAA website.
Temperature and wind fields at 100 hPa from European centre for medium-range weather forecasts (ECMWF) reanalysis data (ERA-40 model) have been examined.This analysis is carried out for monthly variations in the year 1999 and day-to-day variations during BOBMEX campaign period.The height level of 100 hPa is selected as it denotes standard level near the tropopause.The spatial resolution of temperature reanalysis data set (ERA-40) used in this study is (1.125 • × 1.125 • ).Table 1 lists the stations and dates for which extreme low tropopause temperature (≤191 K) is observed at three or more stations including Sagar Kanya.It also gives the stations for which T > 191 K.For the events listed in Table 1, simultaneous radiosonde/GPS-sonde data, INSAT-1D OLR data, and ECMWF reanalysis data sets are available.In addition to ECMWF data, National centers for environmental prediction (NCEP) reanalysis (Kalnay et al., 1996), data of daily mean temperature field at 100 hPa level for the period 24 June-15 August 2002 are also used.The data sets, as mentioned above, provided a unique opportunity to understand better the influence of intense tropical convection on the temperature field at the levels close to the tropopause.

Seasonal movement of the intertropical convergence zone (ITCZ) and observations of monsoon associated mesoscale convection events
Seasonal movement of the ITCZ is well known.In the month of January, the ITCZ is normally located in the region of 12-15 • S.During summer monsoon, the ITCZ merges with the monsoon trough and in the month of July, the ITCZ is located in the region of ∼20-24 • N latitude range (Holton et al., 2003).-193 193.0-196.9 195.3After the meteorological centers have started receiving satellite pictures of cloud systems on routine, the ITCZ is defined in terms of maximum cloudiness.This is due to the fact that the ITCZ is accompanied by an organized band of clouds.Some of these are convective clouds.This makes it feasible to use OLR for the detection of the ITCZ.Okoola (1998) used OLR to study spatial evaluation of the active convection pattern across the equatorial eastern Africa region during spring season in the Northern Hemisphere.In present study seasonal movement of the ITCZ, during the year 1999, is examined using monthly average OLR data from INSAT-1D.Figure 2 shows location of the ITCZ (solid line) for the months of January 1999 and July 1999.In Fig. 2, spatial distribution of monthly mean temperature at 100 hPa level from ECMWF reanalysis is also plotted.A careful examination of Fig. 2 shows that in the month of January both the ITCZ and minima in 100 hPa temperature lie south of the equator, whereas in July both the ITCZ and minima in 100 hPa temperature lie north of the equator, indicating similarity in the seasonal movement and a close association between the two phenomena.The ITCZ is generally associated with intense convection activity.Therefore, with movement of the ITCZ to the northward and its merging with the monsoon trough indicates that monsoon-associated convection must occur over the Indian tropical region and this is, indeed, the case.2. Clouds with such cold top temperature are noted to be located in the low OLR areas as seen from OLR contour maps.
The low CTT, as observed in Fig. 3, are compared with 100 hPa ECMWF reanalysis temperature field over these areas.Table 2 gives a comparison of the observed CTT and the reanalysis temperature at 100 hPa level, which is taken to be representative of ambient temperature.It can be noted from Table 2 that, in general, CTT temperature is lower by ∼10 K than the ambient temperature.This indicates the presence of clouds with CTT cooler than the surroundings.
To examine (i) whether deep penetrating convective systems with cold tops really occur over the Indian region and (ii) to distinguish such convective clouds system from cirrus, 3 days observational data obtained from Kalpana-1 satellite during July-August 2008 are used (Fig. 4).Some locations (i.e.same pixels/area) where the temperature from clean and dirty channels is close enough (i.e. with in 10 K) are marked  with arrows in Fig. 4, indicating the presence of thick convective clouds at these locations (http://www.nrlmry.navy.mil/).Right hand panels in Fig. 4 show high resolution OLR observations (0.25 • × 0.25 • ) from Kalpana-1 satellite corresponding to imagery in the left hand panels.It is evident from Fig. 4 that locations of close temperatures from infrared (clean) and water vapor (dirty) channels lie within the area of low OLR, i.e.OLR ≤160 W m −2 which is consistent with the presence of convection representing CTT≤235 K corresponding to the altitude of ≤200 hPa (i.e.≥12 km).These results are consistent with those of Cornford and Spavins (1973) who have observed clouds extending up to 65 000 ft (∼20 km) and also noted that cumulonimbus clouds exceeding 60 000 ft (∼18 km) occurred on most of the measurements days.
Monsoon-associated mesoscale convection is considered as one of the source mechanisms for cooling the tropopause.It would, therefore, be interesting to examine the occurrence of mesoscale convection and its association with occurrence of LTT.

Occurrence of low temperature in tropopause region in association with convection activity
In the present study, to examine the spatial distribution of the areas of enhanced convective systems, OLR has been used as a proxy for cloud/convective system.However, caution should be taken when interpreting OLR as a convection/rainfall, since thick, high clouds can produce identical OLR as connective cloud, while substantially different production of rainfall occur between the two cloud types (Lyons, 1991).In the present study following four approaches have been adopted to examine the spatial distribution of enhanced convection activity (i.e.low OLR) and of low temperature in tropopause region.

Radiosonde/GPS-sonde observations and OLR from INSAT-1D satellite
Radiosonde/GPS-sonde observations taken during BOB-MEX campaign and simultaneous OLR observations by INSAT-1D satellite are examined.All the cases with observations of low CPT temperature (T CPT ≤ 191 K) at 3 or more stations including Sagar Kanya (listed in Table 1), are studied in detail.Figure 5 shows all the 9 cases listed in Table 1.In Fig. 5

ECMWF reanalysis temperature at 100 hPa level and OLR from INSAT-1D satellite
In this approach ECMWF reanalysis temperature data (M100 model) at 100 hPa level (T 100 ) with a spatial resolution of 1.125 • × 1.125 • and simultaneous OLR observations from INSAT-1D are examined.Figure 5 shows all the 9 cases listed in Table 1.In Fig. 5 areas covered by low T 100 are hatched and areas of low OLR, i.e.OLR≤160 W m −2 are shaded grey (as mentioned in Sect.3.2.1). Figure 5 also shows that areas of low T 100 and low OLR are not always collocated, though they are located in same broad area.On some days, some of the areas of low OLR and low T 100 are noted to overlap/partially overlap as seen more clearly from panels (f), (g), (h) and (i) of Fig. 5. On some days, multiple centers of low OLR are also observed.In such cases, areas of low T 100 tend to be located in between the centers of low OLR which is more evident from panels (d-i) of Fig. 5.  Danielsen (1982) is operating in the cases where the areas of low OLR and low T 100 are overlapping or partially overlapping.In some cases, when these two areas are not overlapping, such as the case when low T 100 appears in between the multiple centers of low OLR, the mechanism suggested by Sherwood (2000) is more likely to be operating.It is likely that on some days, both the mechanisms are operating, but at somewhat different geographical locations.

Analysis of time series of daily mean T 100 and daily mean OLR
The results shown here bring out a close association between enhanced convection activity and occurrence of low T 100 /T CPT .This association is further examined using time series of daily mean OLR data from NOAA website and daily mean T 100 data from NCEP reanalysis for the period of 24 June-15 August 2002.This particular period was selected as the necessary OLR and NCEP reanalysis data were readily available.Two specific regions, viz.Indian land mass (ILM: 10-20 • N, 72.5-82.5 • E) and Bay of Bengal (BOB: 10-20 • N, 85-90 • E) are selected and daily mean OLR and T 100 are averaged over these areas.Panels (a) and (b) of Fig. 6 show the time series of anomaly in OLR and T 100 over the BOB and ILM regions, respectively.Panels (c) and (d) of Fig. 6 show the corresponding spectrum obtained using a fast Fourier transform (FFT) algorithm which applies de-trending 41 Figure 6. and necessary Hanning window (Blackman and Tukey, 1959) before taking FFT.Running mean over 5 days was taken for each series before applying FFT to filter out oscillations with period of <5 days.Spectrum of temperature and OLR time series show a broad peak with a period of 16-21 days.The OLR spectrum over BOB region shows an additional peak at ∼7-9 day period.Jain et al. (2010), using Arabian Sea Monsoon Experiment (ARMEX) campaign data, have shown that westward propagating Rossby waves of period of ∼15 days and eastward propagating intraseasonal oscillation modulate temperature at 100 hPa level during summer monsoon season.Figure 6a, b shows that waves of period of ∼15 days modulate OLR as well as T 100 .Therefore, this particular oscillation is selected by passing the OLR and T 100 anomaly time series through a band pass filter of 13-18 days period.Spectra of filtered time series are obtained using FFT  technique and the same are shown for ILM and BOB regions in Fig. 7a, b.It is apparent from Fig. 7 that OLR and T 100 spectrum over ILM region is similar.The spectra of OLR and T 100 differ slightly over BOB, because short-period waves (13-16 days period) are dominant in modulating OLR over this region.Since the two sets of spectra are similar over ILM region, a cross-correlation analysis is carried out.Figure 7c shows the cross-correlation function between T 100 and OLR time series over this region with OLR leading in phase by ∼5 days.Observations of low OLR and its variability are supposed to be the active source region for the generation of atmospheric waves.In this case, Rossby waves are identified to be present in the ILM region (Fig. 7).Westward movement of OLR with leading in phase in comparison to temperature suggests that quasi-periodic behavior of the source (convection i.e.OLR) has generated this wave mode in the upper troposphere and the delay of ∼5 days represents the time taken for the energy transfer to the medium which oscillates in resonance with the same characteristics.

Role of horizontal winds and convection generated gravity waves in spatial distribution of areas of low T 100 and of low OLR
A close association between OLR and T 100 /T CPT is evident from the results presented in Sect.3.2.This suggests that spatial distribution of these two parameters must be related to the processes involved in cooling the tropopause region by convection.This would throw light on the mechanism of intrusion of tropospheric air into stratosphere.There are the following two types of processes through which convection could cool the tropopause region and then redistribute spatially the regions of low temperatures at the tropopause level.
2. The convection-generated gravity waves penetrating to UTLS are modulating the temperature of the region and, in turn, spreading the cooling to the other areas (Tsuda et al., 1994;Potter and Holton, 1995;Kuang and Bretherton, 2004;and Dhaka et al., 2006).
An attempt is now made to examine the role of each of the above processes in determining the relative spatial distribution of convection activity and low tropopause temperature.

Role of horizontal winds
Horizontal wind and associated advection could have a role in determining the spatial distribution of the areas of low OLR and of low T 100 .It takes some time for convection to cool the tropopause region, which could be ∼4-10 h or even longer (Kim and Desseler, 2004).In this time interval, winds could shift the location of convective systems and thus the areas of both low OLR and T 100 may appear spatially separated.The period 24 August 1999, 00:00 GMT-28 August 1999, 18:00 GMT is selected to examine the role of horizontal winds, based on the synoptic situation (Morwal and Kumar, 2002) prevailing over the Indian tropical region during this period which is briefly described below: On 24 August 1999, a low pressure area was formed over west central Bay off the north Andhra coast which persisted till 25 August.An associated cyclonic circulation extended up to mid-tropospheric levels.Another low pressure system was formed on 26 August over north-west Bay off the Orissa coast and the associated cyclonic circulation extended up to mid-tropospheric levels tilting south-westwards with height.On 27 August, the low pressure area over north-west Bay off     clearly indicate that horizontal winds control the movements of cloud systems as well as of low T 100 systems.Since the areas of both cloud and low T 100 systems are mainly confined to 10-20 • N and 75-120 • E, these features closely resemble the tropical easterly jet (TEJ), indicating that cloud and low T 100 systems are closely linked to TEJ.

Role of convection generated gravity waves
The role of convection-generated gravity waves (GW) in cooling the temperature in tropopause region is examined by critically analyzing the temperature profiles obtained from GPS-sondes launched from ORV Sagar Kanya.Wave structure is observed to modulate the vertical temperature profile.Figure 9 shows a typical example, where temperature profile, perturbations from mean and the spectrum of temperature perturbations are shown for 24 August 1999, 00:00 GMT.The wave amplitude is ∼1-4 K with a mean value of 2.2 K and the vertical wavelength is in the range of 1.2-9.6 km with a mean value of 3.0±2.6km (Table 3).The convectiongenerated GW with a dominant period of 23-49 min is reported by Kumar (2006) from the observations made by using Indian MST Radar at Gadanki (13.5 • N, 79.2 • E).
The present observations indicate modulation of temperature in upper tropospheric and lower stratospheric region by convection-generated GW.These waves may also spread the cooling eastward (i.e.upwind) through formation of ice clouds (Potter and Holton, 1995).However, in the present study, GW effects in spreading the areas of cooling could not be separated out from the wind effects.

Summary
The objective of the present study was to understand the association between the convection activity and occurrence of extremely low temperatures in tropopause region using the observations from multiple platforms.The observations are summarized as follows: 1.The seasonal movement of the ITCZ is evident.In northern summer monsoon season, intense convection activity with CTT of ∼173-193 K is observed.These CTT,s are noted to be cooler than the expected ambient temperature.It is also noted that clouds with lower CTT lie within the area of low OLR, indicating that these are convective clouds, though presence of cirrus clouds along with convective clouds cannot be completely ruled out.
2. Low temperature in tropopause region and intense convection activity (i.e.low OLR) occur in the same broad area.Areas of LTT and of low OLR do not always overlap.In cases where area of low T 100 /T CPT does not overlap the low OLR area, the value of OLR at such locations is greater than 160 W m −2 .Multiple centers of enhanced convection (low OLR) are observed on number of days and areas of low T 100 /T CPT appear in between such low OLR areas.These observations point out that the mechanisms, one suggested by Danielsen (1982) and the other one suggested by Sherwood (2000), have a role in the troposphere-to-stratosphere transport of air.
On the days when both the mechanisms are operating, geographical areas where the two mechanisms are operating would naturally be separated.
3. Westward propagating Rossby waves of the period of ∼15 days are observed to modulate both OLR and T 100 over the Indian land mass (ILM) region.This again indicates a close association between the two phenomena.A cross-correlation analysis between OLR and T 100 time series shows that OLR leads in phase by ∼5 days.
4. An examination of series of high resolution cloud imageries show that cloud systems as well as the areas of occurrence of low T 100 both move under the influence of horizontal winds.It is also noted that movement of the cloud systems and that of areas of low T 100 is closely linked with TEJ winds which appear over this region during Asian summer monsoon. 32 Figure 1.Fig.1. Map showing the locations of the seven radiosonde stations operated during BOBMEX campaign period.

Fig. 1 .
Figure 1.Fig.1. Map showing the locations of the seven radiosonde stations operated during BOBMEX campaign period.
Fig-ure 3 shows INSAT-1D imageries for three days along with corresponding OLR distribution maps.Following important points may be noted from Fig. 3. 1. Cloud top temperature (CTT) as low as 193-173 K is observed over these areas.This indicates that cloud top is at a height ≥17 km.

Fig. 2 .
Fig. 2. Plot of contour maps of monthly mean temperature field at 100 hPa level from ECMWF reanalysis.Solid line shows location of ITCZ as determined using monthly mean OLR distribution from satellite INSAT-1D.(a) January 1999, (b) July 1999.

Fig. 6 .
Fig. 6.Time series of OLR and Temperature at 100 hPa level (T 100 ) anomaly for the ARMEX campaign period over (a) Bay of Bengal (BOB) and (b) Indian Land Mass (ILM) regions.Spectrum corresponding these time series are shown in panels (c) and (d), respectively.

Fig. 7 .
Fig. 7. Spectrum of OLR and T 100 anomaly time series over ILM BOB regions, after passing through a band pass filter of 13-18 days period, are shown in panel (a) and (b), respectively and panel (c) shows cross correlation function between T 100 and OLR time series over ILM region.

Fig. 8 .
Fig. 8. Left hand side panels show satellite imageries drawn using 6 hourly ICSSP-DX data.Horizontal wind field at 100 hPa level, corresponding to each of the satellite imagery, from ECMWF reanalysis is superposed.Right hand side panels show contour maps of temperature field at 100 hPa level.Contours of low T 100 temperature are shaded blue.Resolution of ECMWF reanalysis data presented in this particular figure is 2.5 • × 2.5 • .(a) 24 August 1999, 00:00 GMT to 24 August 1999, 18:00 GMT.(b) 25 August 1999, 00:00 GMT to 25 August 1999, 18:00 GMT.

Fig. 9 .
Fig. 9. Figure shows signature of temperature modulation by convection generated wave activity.(a) Vertical temperature profile taken on 24 August 1999, 00:00 GMT using GPS Sonde launched from ORV Sagar Kanya, (b) temperature perturbation from mean trend and (c) spectrum of the temperature perturbation from mean trend in the height profile of temperature.

Table 1 .
List of date, time and stations for which CPT temperature (T CPT ≤ 191 K) is observed at three or more stations simultaneously.Stations for which T CPT > 191 K are also listed.

Table 2 .
Cloud top temperature (CTT) and ECMWF reanalysis temperature over the region of cloud with cold CTT.

T 100 from NCEP reanalysis data and daily mean OLR data
Observations in Sects.3.2.1 and 3.2.2 are based on simultaneous measurements of OLR and T 100 /T CPT for 9 days.To examine spatial distribution of OLR and T 100 for a longer period, 54 days data (24 June-15 August 2002) are used.Daily mean OLR is obtained from NOAA website and daily mean T 100 field is taken from NCEP reanalysis.Extreme low temperatures (i.e.T 100 ≤ 191 K) are observed for 29 days.Low T 100 and low OLR are observed to occur over the same broad area, though all OLR and low T 100 areas do not overlap (see Auxiliary Material File 1 in http://www.ann-geophys.net/28/927/2010/angeo-28-927-2010-supplement.zip).It may be mentioned here that in such cases, OLR is greater than 160 W m −2 at the location of low T 100.However, in some cases, some of the low OLR and low T 100 areas are found to be overlapped or partially overlapped.Multiple patches of OLR are observed on number of days.Out of 29 days of low T 100 observations, on 24 days, area of low T 100 is observed in between the areas of multiple low OLR.This indicates that in these cases, low T 100 is produced either by independent multiple centers of convection activity or by one single convection event fragmented by wind pattern.These observations suggest that the mechanism suggested by

Table 3 .
Vertical wavelength of the wave activity in UTLS.The areas of low T 100 generally appear between the cloud systems, though some of the low T 100 areas may overlap or partially overlap the location of some of the cloud systems.4.Speed of the westward movement of the cloud and lowT 100 systems, is ∼10 ms −1 , which is consistent with the expected phase velocity of Rossby waves.