As mentioned earlier, in the tropical region, the amount of water vapour transported into the lower stratosphere from the troposphere is controlled by the cold tropical tropopause temperatures (Fueglistaler et al., 2009). Large convection around the eye of the cyclone and strong updrafts near the eye wall transports a large amount of water vapour into the lower stratosphere through the tropopause. In this way, cyclones will affect the tropopause structure (altitude/temperature). Thus, before quantification of STE, we show the tropopause characteristics observed during the TCs. We used post-processed products of level 2 dry temperature profiles with vertical resolution around 200 m provided by the COSMIC Data Analysis and Archival Center (CDAAC) for estimating the tropopause parameters during cyclones in the period from 2007 to 2013. COSMIC GPS-RO is a constellation of six microsatellites equipped with GPS receivers (Anthes et al., 2008). We also used CHAllenging Minisatellite Payload (CHAMP) GPS-RO data that are available between the years 2002 and 2006 and COSMIC data from 2007 to 2013 for obtaining background climatology of tropopause parameters over the north Indian Ocean.

The climatological mean of all the tropopause parameters is obtained by combining GPS-RO measurements obtained from CHAMP and COSMIC (2002–2013). The tropopause parameters include cold-point tropopause altitude (CPH) and temperature (CPT), lapse rate tropopause altitude (LRH) and temperature (LRT), and the thickness of the tropical tropopause layer (TTL), defined as the layer between convective outflow level (COH) and CPH, and these parameters are calculated for each profile of GPS-RO collected during the above-mentioned period. First, we separated the RO profiles available with respect to distance away from the cyclone centre, around 1000 km for an individual cyclone for each day of the respective cyclone. After separating, we calculated the tropopause parameters as mentioned above for each RO profile. The total number of occultations used in the present study is shown in Fig. 2a. Then we separated the tropopause parameters with respect to the different cyclone intensity. After estimating the tropopause parameters for all the 16 TCs with respect to different intensity, a cyclone-centre composite of all tropopause parameters is obtained. After careful analysis, it is found that there is not much variation in the tropopause parameters observed between D and DD, and between CS and SCS, and thus they are combined to DD and CS, respectively. To quantify the effect of the TCs on the tropopause characteristics, the climatological mean is removed from the individual tropopause parameters. The climatological mean tropopause parameters is estimated from the temperature profiles obtained by using GPS-RO data from 2002 to 2013. We also calculated the difference of tropopause parameters for different cyclone intensities (figures are not shown). Figure 3 shows the cyclone-centred composite of the mean difference in the tropopause parameters (CPH, LRH, CPT, LRT, COH, and TTL thickness) between climatological mean (2002–2013) and individual tropopause parameters observed during cyclones (irrespective of cyclone intensity), and more detailed results on the effect of TCs on the tropopause variations and mean temperature structure in UTLS region during TCs can be found in Ravindra Babu et al. (2015). We have reported that the CPH (LRH) is lowered by 0.6 km (0.4 km) in most of the areas within a 500 km radius of the cyclone centre, and the temperature (CPT/LRT) is more or less colder or equal to the climatological values from the area around 1000 km from the cyclone centre. Note that the effect of a cyclone can be felt up to 2000 km away but since the latitudinal variation also comes into the picture when we consider a 2000 km radius, we restrict our discussion related to variability to within 1000 km of the cyclone centre. COH (TTL thickness) has increased (reduced) up to 2 km within 500 km of the cyclones and in some areas up to 1000 km. Note that this decrease in TTL thickness is not only because of pushing up of the COH but also due to a decrease of CPH. From the above results, we concluded that the tropical tropopause is significantly affected by the cyclones and the effect is more prominent within 500 km of the cyclone centre. These changes in the tropopause parameters are expected to influence water vapour and ozone transport in the UTLS region during cyclones.

To see the variability and the transport of ozone during the passage of cyclones, we investigate the spatial and vertical variability of ozone in the UTLS region using MLS satellite observations. As mentioned in Sect. 2.1, we also separated the MLS profiles based on the distance from the TC centre for each day of the individual cyclone. From all the 16 cyclone cases, we separated the available MLS profiles with respect to distance from the cyclone centre, around 1000 km, and we also separated the MLS profiles with respect to different intensities of the cyclones. Figure 4 shows the normalized cyclone-centred composite of mean ozone mixing ratio (OMR) observed during cyclones (irrespective of cyclone intensity) at 82, 100, 121, and 146 hPa pressure levels during 2007–2013. Note that we have a reasonable number of MLS profiles (1517) from 16 cyclones to generate the meaningful cyclone-centre composite of ozone. Black circles are drawn to show distances 250, 500, 750, and 1000 km away from the cyclone centre. Since large variability in OMR is noticed from one pressure level to another, we normalized the values to the highest OMR value at a given pressure level. The highest OMR values at 82, 100, 121, and 146 hPa pressure levels are 0.38, 0.28, 0.19, and 0.13 ppmv, respectively. Large spatial variations in the OMR are observed with respect to the cyclone centre. At 82 hPa, higher OMR (∼ 0.4 ppmv) in the south-west (SW) side up to 1000 km and comparatively low OMR values (∼ 0.2 ppmv) are noticed in the north of the cyclone centre. At 100 hPa, an increase in the OMR (∼ 0.2 ppmv) near the cyclone centre within 500 km is clearly observed. This enhancement in OMR extends up to 146 hPa and is more prominent slightly to the western and eastern side of the cyclone. In general, the large subsidence located at the top of the cyclone centre is expected to bring lower stratospheric ozone to the upper troposphere. This might be the reason for the enhancement of ozone in the cyclone centre within 500 km. Several earlier studies have reported on the intrusion of the stratospheric air into the troposphere due to the subsidence in the eye region (Penn, 1965; Baray et al., 1999; Das et al., 2009, 2016a). The present results also support this aspect that the detrainment of ozone which reached 146 hPa might be due to strong subsidence. Interestingly, an enhancement in OMR in the south-east (SE) side at 121 hPa, but not at either 100 or at 146 hPa, can be noticed; this needs to be investigated further. Thus, in general, higher ozone concentrations are observed at the cyclone centre within 500 km and are slightly aligned to the western side of the cyclone centre.

In order to quantify the impact of cyclones on UTLS ozone more clearly, we have obtained anomalies by subtracting the mean cyclone-centred ozone observed during cyclones from the background climatology of UTLS ozone that is calculated by using the total available MLS profiles from 2007 to 2013. Figure 4e–h show the normalized mean difference of cyclone-centred ozone obtained after removing the background climatology values for different pressure levels shown in Fig. 4a–d. The maximum difference in OMR for corresponding normalized value at 82, 100, 121, and 146 hPa pressure levels is -0.089, -0.19, -0.09, and -0.06 ppmv, respectively. Enhancement in the OMR (∼ 0.1 ppmv) up to 1000 km from the cyclone centre is observed at 82 hPa. Interestingly, at 100 hPa, the OMR is more or less uniform throughout the radius 1000 km of the cyclone centre except ∼ 500 km radius from the centre where significant increase of OMR (∼ 0.2 ppmv) is observed. This increase in the OMR is within 500 km of the cyclone centre and it extends up to 121 hPa. However, enhancement in OMR at 146 hPa extends up to 1000 km but is distributed towards eastern and western sides of the cyclone centre. Thus, it is clear that the detrainment of lower stratospheric ozone will reach up to 146 hPa during the cyclone period due to the presence of strong subsidence in the cyclone centre. We also calculated the cyclone-centre composite of ozone based on different cyclone intensities such as DD, SCS, and VSCS. After carefully going through them, we have found that this detrainment of ozone reaching up to 146 hPa is more in the higher intensity period of the TCs. We do not know what happens below this pressure level due to limitation in the present data; however, studies (Das et al., 2016a; Jiang et al., 2015) have shown that LS ozone can reach as low as the boundary layer during cyclones. It will be interesting to see the variability in the water vapour, as a large amount of it is expected to cross the tropopause during the cyclone period and reach the lower stratosphere.

As mentioned earlier, an enormous amount of water vapour is expected to be pumped from the lower troposphere to the upper troposphere, and it can even penetrate into the lower stratosphere during cyclones. To see the linkage between tropopause variability and the transport of water vapour during cyclones, we investigated the horizontal and vertical variability of water vapour in the UTLS region using MLS satellite observations. Figure 5 shows the normalized cyclone-centred composite of mean water vapour mixing ratio observed during cyclones (irrespective of cyclone intensity) at 82, 100, 121, and 146 hPa pressure levels observed by MLS during 2007–2013. Black circles are drawn to show distances 250, 500, 750, and 1000 km away from the cyclone centre. The highest water vapour mixing ratio (WVMR) values for corresponding normalized values at 82, 100, 121, and 146 hPa pressure levels are 4.44, 4.49, 6.9, and 16.03 ppmv, respectively. Significantly higher WVMR values are noticed, extending from 500 up to 1000 km from the cyclone centre at 121 (∼ 6.5 ppmv) and 146 hPa (∼ 15 ppmv) levels, with more prominence on the eastern side of the cyclone centre. Comparatively low values are noticed in the centre of the cyclone, especially at 121 hPa. These results compare well with higher WVMR observed on the eastern side of cyclones over Atlantic and Pacific oceans (Ray and Rosenlof, 2007). These results also compare well with those reported by Ravindra Babu et al. (2015), who used GPS-RO measured relative humidity (RH) and found enhancement in RH on the eastern side of the centre in the upper troposphere (10–15 km) over the north Indian Ocean. Higher WVMR values observed on the eastern side of the cyclone centre might be due to the upper level anticyclonic circulation over the cyclones. It is interesting to note that high WVMR lies not at the centre but extends from 500 to 1000 km away from the centre of the cyclone. The WVMR is higher at 121 and 146 hPa than at 100 and 82 hPa. It seems less water vapour has been transported to 100 and 82 hPa from below. As we know, water vapour mostly originates from the lower troposphere and decreases with height. Therefore, vertical transport of water vapour from the lower troposphere to the UTLS may lead to water vapour enhanced at 121 and 146 hPa, and at some time, it reaches higher altitudes. The higher WVMR presented at 100 and 82 hPa levels shows the signature of the tropospheric air even entering into the lower stratosphere during cyclones.

In order to quantify the impact of cyclones on the UTLS water vapour more clearly, we have obtained anomalies by subtracting the mean cyclone-centred water vapour observed during cyclones from the background climatology mean of UTLS water vapour. Figure 5e–h show the normalized mean difference of the cyclone-centred WVMR obtained after removing the background climatology values for different pressure levels shown in Fig. 5a–d. The maximum difference in WVMR for corresponding normalized values at 82, 100, 121, and 146 hPa pressure levels is -0.44, -0.81, -2.55 and -9.09 ppmv, respectively. More than 7 ppmv differences are observed at 146 hPa within 1000 km of the centre, and at 121 hPa, a difference of ∼ 2 ppmv is noticed, extending up to 2000 km (figure not shown) on the eastern side of the centre. At 100 and 82 hPa levels, the increase in the WVMR is ∼ 0.8 and ∼ 0.6 ppmv, respectively, and the enhancement is more observed in the north-east (NE) side of the cyclone centre. Thus, a clear STE is evident during the cyclone over the north Indian Ocean where a clear enhancement in the water vapour (ozone) in the lower stratosphere (upper troposphere) is observed. For quantifying the amount of STE, we calculated the cross-tropopause mass flux for each cyclone by considering the spatial extent within 500 km of the cyclone centre, and results are presented in Sect. 3.3.

We adopted the method given by Wei (1987) to estimate the cross-tropopause mass flux, F. F is defined as F=1g-ω+Vh×∇Ptp+∂Ptp∂t=-ωg+1gVh×∇Ptp+1g∂Ptp∂t=FAM+FTM, where ω is the vertical pressure-velocity, Vh is the horizontal vector wind, Ptp is the pressure at the tropopause, g is the acceleration due to gravity, FAM is the air mass exchange due to horizontal and vertical air motions, and FTM is the air mass exchange due to tropopause motion.

The wind information is taken from ERA-Interim, and the tropopause temperature and pressure within 500 km of the cyclone centre is estimated from COSMIC GPS-RO measurements (Ravindra Babu et al., 2015). These values are considered for the maximum intensity day for each of the 16 cyclones and the respective cross-tropopause flux is estimated. Since the above-mentioned results showed that the higher OMR values are observed in the west and north-west (NW) side and more water vapour is located on the eastern side of the cyclone centre, we separated the area into four sectors with respect to the cyclone centre as C1 (NW side), C2 (NE side), C3 (SW side), and C4 (SE side), respectively, as shown in Fig. 4a. List of cyclones used in the present study with their names, cyclone intensity (CI), centre latitude, centre longitude, and minimum estimated central pressure on their peak intensity day are provided in Table 3. The total flux F (Eq. 1) depends on the air mass exchange due to horizontal and vertical air motion (FAM), and the air mass exchange due to tropopause motion itself (FTM). Since the number of COSMIC GPS-RO measurements is not sufficient to estimate the second term (FTM) for each event, we only calculated the first part of the equation (FAM) individually for each of cyclone with respect to the different sectors mentioned above, and the values are presented in Table 3. However, we roughly estimated the contribution of the second term by assuming change in the tropopause pressure by 0.5 hPa increase (decrease) within 6 h. We could see that the cross-tropopause flux for CS is 0.25 ± 0.07×10-3 kg m-2 s-1 (-0.36 ± 0.07×10-3 kg m-2 s-1), and for VSCS it is -0.24 ± 0.3×10-3 kg m-2 s-1 (-0.85 ± 0.3×10-3 kg m-2 s-1). If there is a change in the tropopause pressure by a 1 hPa increase (decrease), the flux for CS is 0.55 ± 0.07×10-3 kg m-2 s-1 (-0.66 ± 0.07×10-3 kg m-2 s-1), and for VSCS it is 0.06 ± 0.3×10-3 kg m-2 s-1 (-1.16 ± 0.3×10-3 kg m-2 s-1).

Figure 6 shows the cross-tropopause flux estimated in each sector from the centre of the cyclone for the different cyclone intensities (estimated based on the cyclone centre pressure). Red lines show the best fit. It clearly shows that the downward flux is always more in C1 and C3 sectors, whereas the C2 sector shows a more upward flux. The flux itself varies with cyclone intensity, and it is found that there is an increase in the downward flux as the cyclone centre pressure decreases, particularly for C1 and C3 sectors, whereas in the C4 sector, an increase in the upward flux is seen as the cyclone intensity increases, but the flux is always upward (positive) in the C2 sector, irrespective of the cyclone intensity. The second term (in Eq. 1) itself corresponds to the air mass exchange from the tropopause motion, and generally during the cyclone period there is an ∼ 400 m difference in tropopause altitude (LRH) within 500 km of the centre of the cyclone (Fig. 3). Thus, the spatial and temporal variation of the tropopause during the cyclones itself is very important to decide whether the flux is downward or upward. Interestingly, C1 and C3 sectors of the cyclone show dominant downward mean flux and C2 and C4 sectors show dominant upward mean flux with values of 0.4 ± 0.4×10-3, 1.2 ± 1.0×10-3, 0.2 ± 0.1×10-3, and 0.12 ± 0.3×10-3 kg m-2, respectively. These results strongly support our findings of higher ozone in the NW and SW sides and higher water vapour in the NE side of the cyclone centre. The mean flux is observed to vary with the intensity of the cyclone. Mean flux for the severe cyclonic storms (CS) is -0.05 ± 0.29×10-3 kg m-2, whereas for very severe cyclonic storms (VSCS), it is -0.5 ± 1.07×10-3 kg m-2. Reutter et al. (2015) reported that the upward and downward mass fluxes across the tropopause are more dominant in deeper cyclones compared to less intense cyclones over the North Atlantic. Our results are comparable with their results, with the averaged mass flux of the stratosphere to the troposphere as 0.3×10-3 kg m-2 s-1 (340 kg km-2 s-1) in the vicinity of cyclones over the North Atlantic Ocean. They also reported that more transport across the tropopause occurred on the west side of the cyclone centre during intensifying and mature stages of the cyclones over the North Atlantic region.