Observations of ozonesonde measurements of the NDACC/SHADOZ (Network for the Detection of Atmospheric Composition Change and the Southern Hemisphere ADditional OZonesondes)
program and humidity profiles from the daily Météo-France
radiosondes at Réunion island (21.1
The variability of ozone in the tropical upper troposphere (10–16 km in altitude) is important for the climate as it influences the radiative budget (Lacis et al., 1990; Thuburn and Craig, 2002; Riese et al., 2012) and modifies the oxidizing capacity of the atmosphere and the lifetime of other chemical species. The tropical ozone budget in the upper troposphere is influenced by stratospheric intrusions, convective transport from the surface, advection from mid-latitudes and chemical reactions.
Due to the complex interplay between dynamics and chemistry, tropical ozone concentrations observed by radiosondes at stations within the Southern Hemisphere ADditional OZonesondes (SHADOZ) network show large spatial and temporal variability (Thompson et al., 2003a; Fueglistaler et al., 2009). The average tropical ozone mixing ratio in the upper troposphere has a value of 40 ppbv, varying between 25 and 60 ppbv. Causes for the ozone variability in the tropics are particularly difficult to ascertain without a careful analysis of the processes involved in the observed variability (Fueglistaler et al., 2009). One example is that the S shape, found in the mean ozone profile in the SHADOZ stations over the Pacific, was interpreted by Folkins and Martin (2005) to be a consequence of the vertical profile of the cloud mass flux divergence.
In general, the impact of convection on the ozone budget in the tropical upper troposphere is not well established. Solomon et al. (2005) used a statistical method to characterize the impact of convection on the local ozone minimum in the upper troposphere above the SHADOZ sites within the maritime continent (Fiji, Samoa, Tahiti and Java). They identified a minimum of 20 ppbv of ozone in 40 % of the ozone profiles. The 20 ppbv corresponds also to the ozone mixing ratio in the local oceanic boundary layer. The sites are located in a convectively active region (Hartmann, 1994; Laing and Fritsch, 1997; Solomon et al., 2005; Tissier et al., 2016) and have a higher probability to be influenced by local deep convection than other SHADOZ sites.
Tropical convection can transport air masses from the marine boundary layer to the upper troposphere (Jorgensen and LeMone, 1989; Pfister et al., 2010) in less than a day. Because the ozone chemical lifetime is on the order of 50 d, air masses within the convective outflow will retain the chemical signature of the boundary layer (Folkins et al., 2002, 2006). Ozone can therefore be used as a convective tracer, and in doing such, Solomon et al. (2005) estimated the mean level of convective outflow to be between 300 and 100 hPa, or 8 and 14 km in altitude, for SHADOZ stations located in the western Pacific.
At present, little is known about the impact of convection on SHADOZ sites
that are away from actively convective regions. In the Southern Hemisphere,
the position of Réunion island (21.1
In this paper, we analyze ozonesonde measurements of the NDACC/SHADOZ program and humidity profiles from daily Météo-France radiosondes from Réunion island between November 2013 and April 2016 to identify the origin of wet upper tropospheric air masses with low ozone mixing ratio observed above the island, and we try to understand the role of transport, detrainment and mixing processes on the composition of the tropical upper troposphere over Réunion island. We use infrared brightness temperature data from the METEOSAT-7 geostationary satellite to identify deep convective clouds over the SWIO region. The geographic origin of air masses measured by the radiosondes is estimated using Lagrangian back trajectories calculated by the FLEXible PARTicle dispersion model (FLEXPART) (Stohl et al., 2005). Section 2 presents the radiosonde measurements, satellite products and FLEXPART model used in this study. Section 3 presents the seasonal variability in ozone and humidity as well as the convective influence on the radiosonde measurements. Results on the mean level of convective outflow and the convective origin of the air masses measured over Réunion island are also presented in Sect. 3. A summary and conclusions are given in Sect. 4.
The ozonesondes at Réunion island are launched under the framework of
the Network for the Detection of Atmospheric Composition Change (NDACC) and
the Southern Hemisphere ADditional OZonesondes (SHADOZ) programs. The SHADOZ
project gathers ozonesonde and radiosonde (pressure, temperature, wind) data
from tropical and subtropical stations (Sterling et al., 2018; Witte et al., 2017, 2018; Thompson et al., 2017). Between 2014 and 2016, 158 ozonesondes
were launched at Réunion island (almost 3 per month). The majority of
the ozonesonde launches occur around 10:00 UTC at the airport (Gillot:
21.06
In addition, we use data from operational daily meteorological Meteomodem
M10 radiosonde launches performed by Météo-France (MF) at 12:00 UTC at
the airport since 2013. The MF dataset provides relative humidity (RH)
measurements with respect to water at a higher frequency than the SHADOZ
data, and this is important to study the day-to-day variability of the impact
of convection on the upper troposphere. The Meteomodem M10 radiosondes
provide measurements of temperature, pressure and RH with respect to water and
zonal and meridional winds. We also calculated RH with respect to ice (RHi) when
the temperature is below 0
For both MF and SHADOZ sondes, the average ascent speed of the balloon is 5 m s
As noted previously, this study focuses on austral summer conditions (November to April) and in particular the austral summer seasons 2013–2014, 2014–2015 and 2015–2016 (hereafter referred to as summer 2014, 2015 and 2016, respectively). Figure 1 shows the NDACC/SHADOZ 2013–2016 seasonal average ozone mixing ratio profiles as well as the overall mean 4-year average. The 4-year average profile over 2013–2016 increases in the troposphere (from 25 ppbv at the surface to 200 ppbv at 17 km). In austral autumn (March, April and May) and winter (June, July and August) the ozone values are lower than the mean climatology in the troposphere above 3 km. Ozone values increase in the lower troposphere during the dry season (from May to September) when biomass-burning plumes from southern Africa and Madagascar can be transported eastward and result in ozone production in the lower troposphere over the Indian Ocean (Sinha et al., 2004). The maximum of tropospheric ozone occurs in austral spring (September, October and November) at the end of the biomass-burning season (which extends from July to October; Marenco et al., 1990). Using ozonesonde and lidar from Réunion island from 1998 to 2006, Clain et al. (2009) showed that the influence of stratosphere–troposphere exchange induced by the subtropical jet stream is maximum in austral winter (June to August) when the jet moves closer to the island. They established that the 4–10 and 10–16 km altitude ranges can be directly influenced by biomass burning and stratosphere–troposphere exchange. The influence of stratosphere–troposphere exchange is in agreement with high-ozone and low-water-vapor layers, which are ubiquitous over Réunion island in austral winter. Austral summer (DJF) exhibits low ozone values, and in particular, below 3 km and between 9 and 14 km summertime ozone is at an annual low. The source of these low values will be discussed later in this paper.
Climatological mean ozone profile (from November 2013 to April 2016) in black. The 2013–2016 seasonal mean ozone profile for summer (December, January, February) is in red, spring (March, April, May) is in cyan, winter (June, July, and August) is in blue and autumn (September, October, November) is in grey.
METEOSAT-7 is a geostationary satellite positioned at the longitude
58
We assume black-body radiation (Slingo et al., 2004; Tissier, 2016) to
estimate the brightness temperature. Young et al. (2013) have classified
clouds in the tropics from the CloudSat and MODIS database for 1 year of
observations (2007) over the 30
In order to fold FLEXPART weekly products with METEOSAT-7 infrared
brightness temperature data, the latter dataset is interpolated to a regular
latitude–longitude grid with a 1
Average map of the deep convective cloud occurrence (DCCO) for austral summer conditions (NDJFMA from November 2013 to April 2016). The yellow contour is for DCCO > 7 %, the green contour is for DCCO > 12 % and the dark green contour is for DCCO > 17 %.
A weakness of the methodology relates to our treatment of convective tower anvils, which may have brightness temperatures colder than 230 K. However, we assume that only convective centers correspond to cloud tops with a brightness temperature below 230 K. We are using this assumption to identify the deep convective clouds and compare their distribution with the vertical transport from the boundary layer to the upper troposphere calculated by the FLEXPART model.
To estimate the convective origin of mid- to upper-tropospheric air masses
observed above Réunion island, we use the FLEXPART Lagrangian particle
dispersion model (Stohl et al., 2005). We use input meteorological fields
from the ECMWF Integrated Forecast System (IFS, current ECMWF operational
data) that have 137 vertical levels up to 0.01 hPa. The vertical resolution
varies:
We use FLEXPART to calculate back trajectories of particles from three bins at 1 km intervals in the upper troposphere (i.e., 10–11, 11–12, 12–13 km)
above Réunion island. Vertical bins are defined between 10 and 13 km to
trace the lower observed ozone values in the upper troposphere during
austral summer (Fig. 1). The bins have a horizontal latitude–longitude
resolution of
Figure 3 shows the time series of vertical profiles of RH from 2014 to 2016.
High RH values (above 80 %) observed below 2 km are typical values of
the tropical humid marine boundary layer (Folkins and Martins, 2005). The
mean value of RH in the upper troposphere (10–13 km) ranges from
Time–height cross sections of day-to-day variations of relative
humidity (RH) for October 2013–July 2014
Higher values of RH of
It is known that the El Niño–Southern Oscillation (ENSO) can affect
convective activity over the SWIO (e.g., Ho et al., 2006; Bessafi and
Wheeler, 2006). The NOAA Climate Prediction Center Ocean Niño Index
(ONI,
The differences in humidification in the upper troposphere can also be
affected by the Madden–Julian Oscillation (MJO). To define the state of the
MJO, we used the Real-time Multivariate MJO (RMM) indices RMM1 and RMM2
from the Australian Bureau of Meteorology
(
RH distribution in the upper troposphere (10–13 km) above
Réunion island. The 904 profiles for the period November 2013–April 2016 were
used to compute the distribution. Bins are every 5 %. The mean RH of the
distribution (30.4 %) is shown with a blue line, the mean RH for a hydrated
profile (WV > 121 ppmv) with a dotted red line and the mean RH for strong
hydration (WV >
During the three austral summers studied, the MJO was active over the Indian
Ocean for a similar number of days (14 %, 18 % and 18 % of the time in
austral summers 2014, 2015 and 2016, respectively). The averaged upper
tropospheric RH for an active MJO over the Indian Ocean is 30 %, almost
the same as the climatological RH over the period November 2013 to April 2016 (cf. Fig. 5). During some of these MJO events there was an increase in
RH, e.g., 5–11 December 2013 (50 %), 3–5 November 2015 (46 %), 13–20 January 2016 (52.4 %) and 1–3 February 2016 (54.8 %). Garot et al. (2017) studied the evolution of the distribution of upper-tropospheric
humidity (UTH) over the Indian Ocean with regard to the phase of the MJO
(active or suppressed). They used RH (with respect to water) measurements
from the Sounder for Atmospheric Profiling of Humidity in the Intertropics
by Radiometry (SAPHIR)/Megha-Tropiques radiometer, RH measured by upper-air
soundings, dynamic and thermodynamic fields produced by the ERA-Interim
model, and the cloud classifications defined from a series of geostationary
imagers to assess changes in the distribution of UTH when the development of
MJO takes place in the Indian Ocean. There is a strong difference in the
distribution of UTH according to the phase of MJO (active or suppressed).
During active (suppressed) phases, the distribution of UTH measured by
SAPHIR was moister (drier). However, their study focused on the equatorial
(8
Austral summer 2014 (Figs. 3 and 4) is affected by three tropical cyclone
events. Overall, summer 2014 is the driest of the 3 years. Consistent
with a higher ONI at
Figure 5 shows the histogram of RH between 10 and 13 km for the three austral summer periods (2014, 2015 and 2016). We choose a RH value of 25 % (corresponding to the median of the distribution) to characterize the upper tropospheric background, which should be dry without the effect of convective hydration. In the rest of the study, a RH threshold of 50 % is used to isolate upper tropospheric air masses that have likely been affected by deep convection. A threshold for RH had to be chosen to isolate the RH and ozone profiles that were most likely impacted by convection. The average water vapor mixing ratio between 10 and 13 km in austral summer (182 ppmv) is larger than in austral winter (65 ppmv), probably due to the effect of deep convection and associated moisture transport and cloudiness. The average RH of air masses with water vapor mixing ratios greater than 182 ppmv is 48.8 %. Thus, a RH threshold of 50 % is used to isolate upper tropospheric air masses that may have been affected by deep convection in the rest of the study.
In this part of the study, we use the NDACC/SHADOZ dataset to analyze the convective influence on air masses observed above Réunion island. The 2013–2016 NDACC/SHADOZ ozone dataset has a mean background value of 81 ppbv in the upper troposphere (average ozone mixing ratio between 10 and 13 km). Figure 6 shows the ozone distributions for the lower troposphere (below 5 km, green bars in Fig. 6) and the upper troposphere (10–13 km, grey bars in Fig. 6); 76.8 % of the lower tropospheric ozone data have values ranging from 15 to 40 ppbv. These values agree with ozone mixing ratios typically observed for air masses in the marine boundary layer (20 ppbv); we note that the values larger than 20 ppbv can be explained by mixing with air masses of the tropical free troposphere with climatologically higher ozone content. In the upper troposphere, ozone mixing ratios range from 30 to 110 ppbv (Fig. 6). To estimate the average residence time in the upper troposphere, we analyzed the evolution of RTLT for different back-trajectory durations (not shown). RTLT from 46 h back trajectories is mostly located in the vicinity of Réunion island, as well as the northeast of Madagascar. The 96 h RTLT pattern is significantly different and spreads over the eastern and northern regions of Madagascar for 2015 and 2016 and also west of Madagascar in 2014. The pattern of 120 h and 168 h RTLT is roughly similar to the 96 h RTLT, except that RTLT is more spread over the northeast and west of Madagascar. It means that most of the humid air masses reaching the 10–13 km layer above Réunion island were embedded in convective clouds and were transported from the lower troposphere to the upper troposphere within 96 h. The spread in the RTLT product from 96 to 168 h backward in time is the result of horizontal atmospheric transport in the lower troposphere. Therefore, we can estimate an average time of transport between the main convective sources and the upper troposphere over Réunion island to be 96 h.
NDACC/SHADOZ ozone distribution in the lower troposphere (0–5 km) shown in green. The total distribution in the upper troposphere (10–13 km) is shown in grey and for moist data (10–13 km and RH > 50 %) in blue. The distributions are based on 55 ozone profiles for austral summers 2014, 2015 and 2016. The mean values for each distribution are 73.8, 57 and 33.5 ppbv for the 10–13 total ozone distribution, 10–13 km ozone distribution with RH > 50 % and 0–5 km total ozone distribution, respectively.
For the upper troposphere, we further consider the ozone distribution for humid air masses by using a RH threshold of 50 %. We performed sensitivity tests by using RH thresholds ranging from 40 % to 55 %, and found that the ozone distribution in the upper troposphere is very similar for these different RH thresholds (not shown). One main mode appears in the ozone distribution for air masses with RH > 50 % (blue bars in Fig. 6) that is centered around 45 ppbv (56.4 % of data are between 30 and 57.5 ppbv) As explained previously, the mode centered around 45 ppbv in the wet distribution may be associated with vertical transport of low-ozone air masses from the marine boundary layer to the upper troposphere and subsequent mixing with tropospheric air masses with higher ozone content along their pathway.
Ozone mixing ratios higher than 70 ppbv are observed less frequently in the moist upper troposphere (16 % of the observations) than in the total distribution (43 % of the observations). However, the average ozone mixing ratio in the humid upper troposphere is on average higher than the ozone mixing ratio observed in the lower troposphere (45 ppbv against 31.7 ppbv, respectively). This again agrees with a convective transport pathway from the marine boundary layer to the upper troposphere and mixing along the pathway.
As suggested in Fig. 2, and later discussed in Sect. 3.5, the deep
convection that may commonly influence the upper troposphere above
Réunion island is not directly in the vicinity of the island but further
north in the ITCZ region. The difference between the ozone signature in the
low troposphere (31 ppbv) and directly above Réunion island (45 ppbv)
suggests that mixing processes occurred during the long-range transport
through the upper troposphere enriched in ozone (
Solomon et al. (2005) have studied ozone profiles at several tropical sites in the Southern Hemisphere to characterize the impact of deep convection on the ozone distribution in the tropical troposphere. They studied 6 years of measurements (1998 to 2004) from different stations of the SHADOZ network. In the Solomon et al. (2005) study, 40 % of the ozone profiles over the western tropical Pacific (WTP) stations (Fiji, Samoa, Tahiti and Java) have ozone mixing ratios lower than 20 ppbv within the upper troposphere (10 to 13 km); 20 ppbv is the average ozone mixing ratio found in the clean marine boundary layer of the WTP. The WTP is the most active convective basin of the Southern Hemisphere due to warmer SSTs in this region (Hartmann, 1994; Laing and Fritsch, 1997; Solomon et al., 2005; Tissier, 2016). Hence, ozone profiles in the WTP have a higher probability of being influenced by recent and nearby convection than other SHADOZ stations. This explains the weaker probability of ozone mixing ratios lower than 20 ppbv in the upper troposphere for other stations which are located further from the ITCZ region.
Figure 7 shows fractions of the ozone distribution lower than different ozone mixing ratios (25, 40, 45, 50, 55 and 60 ppbv) for ozone profiles observed during the austral summer seasons of 2013 to 2016. A low probability of measuring ozone mixing ratios lower than 25 ppbv is found at the top of the boundary layer. Furthermore, none of the ozone profiles have a mixing ratio lower than 20 ppbv between 8 and 13 km, confirming the results by Solomon et al. (2005), and less than 12 % of the profiles have an ozone mixing ratio lower than 40 ppbv. However, the fraction of ozone profiles displays a maximum of occurrence for ozone thresholds at 45 (22 %), 50 (27 %) and 55 ppbv (35 %) between 10 and 13 km, corresponding to the altitude of the mean level of convective outflow found in Solomon et al. (2005).
Vertical profiles of frequency of occurrence of ozone mixing ratios below 25, 40, 45, 50, 55 and 60 ppbv at Réunion island for austral summers (NDJFMA) 2014, 2015 and 2016.
We will show in the subsequent sections that the ozone chemical signature of convective outflow diagnosed from Fig. 7 is mainly associated with air masses detrained from the ITCZ. In comparison to the WTP region, the ITCZ is primarily located north of Réunion island (Schneider et al., 2014), even in austral summer (Fig. 2). Considering that Réunion island is farther from the ITCZ than the stations in the WTP, a longer time for long-range transport to occur is needed from the convective region to Réunion island, and thus mixing between low-ozone air masses in the boundary layer with high-ozone air masses in the upper troposphere can explain the values observed in the upper troposphere over Réunion island. Moreover, photochemical production of ozone during long-range transport after convective entrainment can increase the ozone of an air mass (Wang et al., 1998).
Hellen was a tropical cyclone that formed in the Mozambique Channel and was named on 26 March 2014. It became a category 4 tropical cyclone on the Saffir–Simpson scale on 30 March 2014. After reaching its maximum stage on 30 March 2014, Hellen was classified as category 1 on the Saffir–Simpson scale and was 1200 km away from Réunion island at the time of the sounding on 31 March 2014 at 12:00 UTC. Although Tropical Cyclone Hellen is not the most influential cyclone on the upper troposphere above Réunion island, it is a relevant case study as it is representative of tropical cyclones that form in the Mozambique Channel for the SWIO region. In addition, this system had a clear signature in the RH profile in the upper troposphere (relative maximum of RH of 60 % at 11 km altitude in Fig. 8c, red curve). Since RHi is around 100 % at 11 km and the decrease in humidity below the layer is slower than above the layer, it probably indicates a hydration effect due to sedimented ice crystals.
Patterns of the RTLT (fraction of residence time in the lower troposphere; see Sect. 2.3 for definition) during the week before 31 March 2014 for air masses sampled in the upper troposphere above Réunion island are displayed in Fig. 8a. RTLT can be considered a map of density probability function of origin of the thousands of trajectory particle source locations in the lower troposphere. High values of RTLT (filled contours in Fig. 8a) are observed over the Mozambique Channel and are coincident with the best track of Tropical Cyclone Hellen (red curve in Fig. 8a and b). Thus, the FLEXPART backward trajectories indicate that the air mass sampled on 31 March 2014 above Réunion island spent a significant amount of time in the lower troposphere during the previous week while Tropical Cyclone Hellen was intensifying over the Mozambique Channel. Additionally, the high values of RTLT coincide with a high weekly mean convective cloud cover (DCCO; see Sect. 2.2 for definition) for the same week (Fig. 8b). The weekly DCCO was higher over the Mozambique Channel in agreement with the presence of the tropical cyclone in this region during the week preceding 31 March 2014.
The two maps of RTLT and DCCO roughly display the same pattern (maximum
above the Mozambique Channel). A detailed analysis of RTLT was performed for
Tropical Cyclone Hellen with different residence times in the lower
troposphere with 48, 96, 120 and 168 h FLEXPART
back trajectories (not shown). After 48 h, no contribution in RTLT is
found. After 96 h, the RTLT is located north of the storm track, within
the convective region of Tropical Cyclone Hellen. After 120 and 168 h,
an counterclockwise dispersion toward Africa, outside the convective cells, is
found. It represents the fraction of air masses in the lower troposphere
that was advected toward the convective clouds before reaching the 10–13 km
altitude range. Hence, the collocation of RTLT with DCCO depends on the
collocation of the convective regions in FLEXPART
In this section, we will identify which tropical cyclones have influenced the upper troposphere above Réunion island. We display in Fig. 9 the trajectories of 23 tropical cyclones (8 in 2014, 9 in 2015 and 6 in 2016) that were within a 2100 km radius around Réunion island, representing 74 % of tropical cyclones that developed within the SWIO basin between summers 2014 and 2016 (from November 2013 to April 2016). Outside the 2100 km radius, the influence of tropical cyclones (TCs) on Réunion island's upper troposphere is found to be limited (not shown).
Map of best tracks of tropical cyclones for a range ring of 2100 km around Réunion island. The best tracks of tropical cyclones during austral summer 2014 (November 2013 to April 2014) are shown in orange with different symbols for each tropical cyclone (TC); best tracks are in cyan for austral summer 2015 (November 2014–April 2015) and in blue for austral summer 2016 (November 2015–April 2016, in green). The yellow star indicates the location of Réunion island.
There is significant variability in the number of SWIO cyclones that traverse (or maybe form in) the Mozambique Channel in a given year. For 2014 there were three (out of eight for the SWIO), for 2015 there was two (out of nine), and for 2016 there was none (out of six). Near Réunion island, a similar activity is found during the three summer seasons (about two cyclones per year in the direct vicinity of the island). In 2014, Tropical Cyclone Bejisa was the only cyclone that directly impacted Réunion island. During the three austral summer seasons of 2014, 2015 and 2016, half of the tropical cyclones formed northeast of Réunion island (12 in total).
In order to determine the tropospheric origin of upper tropospheric air
masses observed over Réunion island during summers 2014, 2015 and 2016
(Fig. 3), we integrated the RTLT gridded over the domain of study
(1
sRTMT in Fig. 10 represents the origin in the middle troposphere. An
increase in sRTMT is associated with a vertical transport in the troposphere
weaker than events that increase sRTLT, such as deep convection. A study by
Schumacher et al. (2015) has shown that vertical transport within stratiform
clouds can reach 10 m s
Figure 11 shows the monthly averaged maps of the product of DCCO and
RTLT, which represents the probability of convective influence from a given
region on the upper troposphere above Réunion island. At the beginning
of the austral summer seasons (November 2013, 2014 and 2015), the main
convective regions that influence the upper troposphere above Réunion
island are located in central Africa (Congo Basin and Angola). Then from November
to January, the influential convective region moves to the east towards
Mozambique Channel. For seasons 2014 and 2015, most of the influential
convective regions are linked to cyclonic activity. TC Bejisa (B-2014, near
Réunion island) and TC Deliwe (D-2014, in the Mozambique Channel) were
the most influential convective events in January 2014. For January 2015,
two tropical cyclones were active in the SWIO, Bansi (B-2015, near
Réunion island) and Chedza (C-2015, Mozambique Channel). In February
2014, three cyclones formed in the SWIO basin. Despite the short distance
from the island, TC Edilson (E-2014) did not have a significant influence on
the upper troposphere above Réunion island, while a more remote tropical
cyclone such as Guito (G-2014, in Mozambique Channel) significantly hydrated
the upper troposphere above Réunion island. In March 2014, the little
patch in the Mozambique Channel was directly linked with the Tropical Cyclone Hellen
(H-2014). In February 2015 TC activity decreased and convection over
Madagascar hydrated the upper troposphere above Réunion island. TC
Haliba (H-2015) caused the maximum value of RTLT
Monthly-average product of DCCO and RTLT normalized by the total residence time in the whole atmospheric column. From top right to bottom left: November 2013, December 2013, January 2014, February 2014, March 2014, April 2014, November 2014, December 2014, January 2015, February 2015, March 2015 and April 2015. The values for each contour are indicated by the numbers in blue, orange and red at the bottom of each panel. The location of Réunion island is indicated by a white star in each plot. In each subplot, the letters indicate tropical cyclones that occurred during a given month.
We analyzed ozonesonde measurements from the NDACC/SHADOZ program and humidity profiles from the daily Météo-France radiosondes at Réunion island between November 2013 and April 2016 to identify the origin of wet upper-tropospheric air masses with low ozone mixing ratio observed above the island, located in the subtropics of the SWIO basin.
A seasonal variability in hydration events in the upper troposphere was found. The variability was linked to the seasonal variability of convective activity within the SWIO basin. An increase in the convective activity in austral summer 2016 (a strong El Niño year) compared to austral summers 2014 and 2015 was associated with higher upper-tropospheric hydration. In the upper troposphere, ozone mixing ratios were lower (mean of 57 ppbv) in humid air masses (RH > 50 %) compared to the background mean ozone mixing ratio (73.8 ppbv).
A convective signature was identified in the ozone profile dataset by studying the probability of occurrence of different ozone thresholds. It was found that ozone mixing ratios lower than 45 to 50 ppbv had a local maximum of occurrence near the surface and between 10 and 13 km in altitude, indicative of the mean level of convective outflow, in agreement with Solomon et al. (2005) and Avery et al. (2010).
Combining FLEXPART Lagrangian back trajectories with METEOSAT-7 infrared
brightness temperature products, we established the origin of convective
influence on the upper troposphere above Réunion island. We found that
the ozone chemical signature of convective outflow above Réunion island
is associated with air masses detrained from the ITCZ located northwest of
the island and tropical cyclones in the vicinity of the island (2100 km
around the island). A higher correlation between tropical cyclone activity
and high upper-tropospheric RH values was found in austral summers 2014 and
2015. It was found that isolated convection within the ITCZ was more
pronounced in 2016 (most likely due to the strong El Niño), and as a result
the vertical transport associated with these isolated convective clouds was
misrepresented in the
Hence, it has been found that the upper troposphere above Réunion island is impacted by convective outflows in austral summer. Most of the time, deep convection is not observed in the direct vicinity of the island, as opposed to the western Pacific sites in the study by Solomon et al. (2005), but more than 1000 km away from the island in the tropics either from tropical storms or the ITCZ. In November and December, the air masses above Réunion island originate, on average, from central Africa and the Mozambique Channel. During January and February the source region is the northeast region of Madagascar and the Mozambique Channel.
The average chemical ozone signature of convective outflow was found to be 45 ppbv between 10 and 13 km in altitude, which differs from the 20 ppbv threshold used in Solomon et al. (2005). The higher threshold can be explained by vertical transport of low-ozone air masses from the marine boundary layer to the upper troposphere and subsequent mixing with tropospheric air masses with higher ozone content along their pathway when advected over more than 1000 km.
METEOSAT-7 data used in this study are available at
All authors contributed to the paper. DH wrote the article with contributions from SE, JB, KR and JPC. JMM and FP performed the ozone radiosonde measurements. SE and JB performed the FLEXPART simulations. DH processed the radiosonde and FLEXPART data. All authors revised the article draft.
The authors declare that they have no conflict of interest.
OPAR (Observatoire de Physique de l'Atmosphère à La Réunion, including Maïdo Observatory) is part of OSU-R (Observatoire des Sciences de l'Univers à La Réunion), which is being funded by Université de La Réunion, CNRS-INSU, 720 Météo-France and the French research infrastructure ACTRIS-France (Aerosols, Clouds and Trace gases Research InfraStructure). This work was supported by the French LEFE CNRS-INSU Program (VAPEURDO).
This paper was edited by Farahnaz Khosrawi and reviewed by three anonymous referees.