3.1 Background meteorological conditions over Chennai
Figure 1 shows the composite monthly mean variation of the temperature, relative humidity (RH), potential temperature gradient, zonal wind, meridional wind, mixing ratio of ozone and water vapour from the surface up to 22 km and OLR, Rainfall, CPT-H, and COT-H averaged for the period 2014–2019 over Chennai. Chennai is a coastal station located in the NE monsoon region; however, it experiences rainfalls from both SW and NE monsoon seasons. The time-height section of the temperature (Fig. 1a) shows strong seasonal variations within the atmospheric boundary layer (ABL) ~ 2.0 km and within the TTL. Within the ABL, the temperature becomes maximum (~ 304.6 K) during the premonsoon season and minimum (298.9 K) during the winter season. Within the ABL, temperature becomes maximum (~ 192 ± 2K) during the SW monsoon season and minimum (~ 190 ± 2 K) during the winter season. It is also observed that the TTL thickness is wider (~ 8 km) during October to May (NE monsoon to premonsoon) while narrower TTL thickness (~ 4 km) during the SW monsoon season. Both CPT and COT heights show a strong seasonality; however, they are roughly opposite in phase. The CPT-H becomes maximum (~ 17.5 ± 0.5 km) during the winter season, while it becomes lower (~ 16.7 ± 0.50.4 km) during the SW monsoon season. Whereas the COT-H becomes a minimum (11.2 ± 1.4 km) from December to April and a maximum (12.2 ± 1.6 km) from May to November. Over Chennai, the COT-H starts to increase with the beginning of thunderstorm activities during the end of the premonsoon season (May). These convective activities push the COT to a higher altitude and roughly maintain the same altitude during the SW monsoon season, with a slight decrease during October and then an increase in (November) due to the NE monsoon. The time height section of the relative humidity shows moist conditions with RH > 70% within the ABL throughout the year. The large variation of RH from winter to monsoon can be noticed in Fig. 1b. Near the surface, the RH is roughly opposite in face to the surface temperature. Above the ABL, dry conditions (RH < 20%) prevail from December to April. However, during the SW monsoon season, the moist condition (RH > 50%) prevails throughout the column from the surface to the tropopause due to the strong convective activities. The potential temperature gradient shows the minimum values that mark the COT-H. The CPT-H also separates the higher potential gradient in the lower stratosphere from the lower potential temperature gradient just below the tropopause. Thus, on the monthly mean basis, the change in the potential temperature gradient itself marks the CPT-H. Over Chennai, weak easterly prevails near the surface during the winter and premonsoon seasons, whereas strong westerly winds prevail during the SW monsoon. The low-level westerly jet (LLJ) dominates up to the altitude of 5 km with a core altitude of ~ 2 km. During December to April (winter and premonsoon seasons), westerly winds prevail above 5 km, while easterly wind prevails during the SW monsoon season. During the NE monsoon season, the winds above 5 km show a transitional state from easterly to westerly. During the SW monsoon season, the TTL lies in the vicinity of the tropical easterly jet (TEJ) streams. The time height section of the meridional wind shows that southerly wind from March to September and northerly wind from October to February near the surface (< 1 km). Above it to the mid-troposphere (8–9 km), weak northerly wind prevails throughout the years. In the upper troposphere (~ 10 km to CPT-H), southerly wind prevails from October to May, while northerly wind prevails during the SW monsoon season. It is interesting to note that the COT-H lies at the peak altitude of the southerly wind in the upper troposphere. The time height sections of the ozone mixing ratio show the higher concentrations (> 150 ppbv) in, the lower stratosphere above the CPT-H. Note that the colour bar is suppressed to less than 100 ppbv to enable the seasonal features of the ozone in the UTLS region. It can be seen that the ozone shows a strong seasonal variation in the UTLS (between 10 km to CPT-H) with a minimum ozone mixing ratio (~ 40 ppbv) from July to February and a maximum (~ 60 ppbv) mixing ratio from March to June. The CPT-H separates the stratospheric high ozone mixing ratio, while COT-H separates the tropospheric high ozone mixing ratio. Within the TTL, the ozone concentrations are found to be low.
The time height sections of the water vapor mixing ratio show the higher concentrations (> 10 ppmv) in the upper troposphere below the CPT-H. Note that colour bar is suppressed to less than 10 ppmv to enable the seasonal features of the water vapor in the UTLS region. It can be seen that the water vapor shows a strong seasonal variation in the depth of the higher water vapor mixing ratio, similar to the seasonal variation of the COT-H. The depth is up to ~ 14 km from November to April and ~ 15 km from May to October. During the SW monsoon season, the water vapor with higher concentrations ~ 6–7 ppmv reaches near the CPT-H, venerable to enter the lower stratosphere. Thus, we observed that the water vapor in the lower stratosphere up to about 21 km is ~ 3 ppmv during the winter season while it is ~ 5–6 ppmv during the SW monsoon season, indicating a tap reorder signal (Mote et al., 1996).
3.2 Typical identification of the tropical tropopause layer
The TTL is generally defined as the region between the main convective outflow level to cold point tropopause (Gettleman et al., 2002; Mehta et al., 2008). Both temperature and tracer profiles show the signatures of the TTL region and are hence used to define it (Pan et al., 2016). Broadly, the TTL is the region that extends from the top of the main convection (characterized by the level of minimum stability) to the level of maximum stability (Fueglistaler et al., 2009). Thus, to understand the complex physical and chemical processes in quantifying the TTL, we have utilized the temperature profiles from radiosonde and tracer (ozone and water vapor) profiles from MLS observations over Chennai in the northeast (NE) monsoon region during different seasons.
Figure 2 shows the typical profiles of the temperature, lapse rate, potential temperature, potential temperature gradient, ozone, and water vapor mixing ratio between 8 to 22 km observed on 24 February 2015, 11 April 2017, 31 August 2017, and 14 December 2017. The levels of the CPT, LRT, COT, LSM, and LMaxS are also marked. The typical temperature profiles indicate the CPT and LRT altitudes are higher during the winter season and lower during the SW monsoon season. On a typical day of winter (24 February 2015), it can be seen that CPT-H, LRT-H, and COT-H are ~ 17.6 km, ~ 16.6 km, and ~ 10.8 km, respectively. It is interesting to observe that the ozone mixing ratio gradually starts to increase just below the CPT while a steep increase from 18.6 km (just above the CPT). The level of steep increase in the ozone occurs near the LMaxS. Similarly, the typical ozone profiles observed during the premonsoon season, SW monsoon season, and NE monsoon season also indicate similar feature as the winter season. It can be seen that as the CPT-H lowers, the level of the minimum ozone mixing ratio and the level of the sharp transition from the lower ozone mixing ratio in the upper troposphere also lowers. The ozone mixing ratio is relatively higher during the winter season when compared to the SW monsoon season. Thus, with higher tropopause during winter higher ozone mixing ratio is observed. We can observe a secondary minimum that forms near the altitude of the LSM. Similarly, the typical water vapor profiles observed during the premonsoon season, SW monsoon season, and NE monsoon season also indicate that similar feature. The water vapor shows a strong seasonal variation with a higher value during the SW monsoon season and a minimum value during winter. The water vapor mixing ratio at the tropopause is found to be ~ 2 ppmv during winter and ~ 5 ppmv during the SW monsoons season.
3.3 Seasonal variations of the Tropical tropopause
Figures 3 (a)-(h) show the composite monthly mean and standard deviations of TTL parameters such as CPT-H, CPT-T, CPT-O, CPT-W, COT-H, COT-T, TTLt, and IRBT, respectively obtained from radiosonde observations during 2014–2019. Seasonal variations of the CPT are a well-known feature that becomes higher (CPT-H ~ 17.6 km) and colder (CPT-T ~ 190 K) in January and lower (CPT-H ~ 16.65 km) and warmer (CPT-T ~ 192 K) during August-September mainly due to the annual cycle of the BDC (Yuleva et al., 1994; Fueglistalar et al., 2014). Note that the CPT becomes the highest and coldest in January; however, it becomes the lowest in August and warmest in September. Recently, Annamalai and Mehta (2022), using the radiosonde observations at Gadanki (79.2oE, 13.45oN), located 120 km northwest of Chennai, reported that the warmest and lowest CPT could occur independently. The seasonal variations of the CPT over Chennai are similar to the CPT variation over Gadanki (Mehta et al., 2010; 2011) as well consistent with the earlier reports (e.g., Reid and Gage, 1996; Seidel et al., 2001) except some differences. Over Chennai, the CPT-H is almost steady from January to June and starts to lower from June to August and then rises up to December. The CPT-H steeply decreases from June to August and gradually increases from August to December. The CPT-T, however, remains steady from January to May and starts to warm from May to September and then cools up to December. It is observed that seasonal variations of the CPT-H and CPT-T do not show similar behavior. The LRT also shows strong seasonal variations similar to CPT; however, the amplitude of variation is smaller than CPT. It is to be noted that the range of seasonal variation of CPT-T over Chennai and Gadanki is smaller (~ 2 K) when compared with various other tropical stations such as Truk (7.44oN, 151.83oE), Rochambeau (4.83oN, 52.36oW), Singapore (1.03oN, 103.87oE), Seychelles (4.66oS, 55.53oE) and Darwin (12.41oS, 130.88oE) which have ~ 5 K range of variation (Mehta et al., 2011). Such difference is probably due to the strong influence of radiative processes from higher ozone heating over Chennai and Gadanki when compared to Truk. Using Southern Hemisphere Additional Ozonesondes (SHADOZ) data, Thompson et al. (2003) reported a higher ozone mixing ratio over the Indian ocean when compared to the western pacific.
The monthly mean ozone mixing ratio at CPT at Chennai varies from 50 ppbv during winter to 80 ppbv during the SW monsoon season, with a standard deviation of 10–20 ppbv. The maximum ozone concentration is found to be during July. The CPT is warmer when ozone concentration is higher at the CPT. However, the CPT-H and CPT-O are not completely linear, as stratospheric processes govern ozone. The CPT becomes warmest during September; however, the highest ozone is found during September.
We have obtained the monthly mean water vapor mixing ratio corresponding to CPT at Chennai, which is found to vary from 3 ppmv in winter to 5 ppmv in the SW monsoon season with a standard deviation of 0.5-1 ppmv. The CPT is relatively drier and colder during the winter and moister and warmer during the SW monsoon season. Over Chennai, the CPT is the most moister during September. Though the ozone and water vapor mixing ratio at the CPT becomes maximum during different months in the SW monsoon season, they at COT become maximum during June. At the COT level, the ozone is produced mainly from the troposphere, whereas at the CPT, the ozone mixing ratio is from the stratosphere.
The COT-H and COT-T also show a strong seasonal variation with higher (~ 13.2 km) COT-H during the SW monsoon season and lower (12.4 km) COT-H during the premonsoon season consistently. It is interesting to note that the COT-H becomes relatively higher also during the NE monsoon season. As Chennai is affected by the NE monsoon, during which strong convection occurs, however not strong as the SW monsoon season. Thus, in contrast to previous studies (Mehta et al., 2011; Hemant et al., 2014), COT-H is higher during the NE monsoon and winter seasons. The TTL thickness also shows strong seasonal variation, with a lower thickness (~ 3.5 km) during the SW monsoon season and a greater thickness (~ 5.0 km) during the premonsoon season. The seasonal pattern of the TTL is nearly similar to CPT-H. The seasonal variation of the IRBT data suggests the prevalence of convective events from June to November (covering SW and NE monsoons). Note that the low value of the IRBT indicates strong convection during SW and NE monsoon seasons.
3.4 Role of the upper tropospheric and lower stratospheric temperature, water vapor and ozone
To understand the different seasonal patterns of the CPT-T, CPT-O and CPT-W observed in the previous section, we have examined the temperature, ozone, and water vapor variation in the upper troposphere and lower stratosphere (from 10–20 km) as shown in Fig. 4. Interestingly, the upper tropospheric (UT) temperature anomalies show coherent seasonal variation between the altitude of 10–15 km with maxima during July. The pattern becomes weak and different from UT temperature between the altitudes 16–17 km leveled as the transition region from the UT to the lower stratosphere (LS). The LS temperature again shows coherent seasonal variation between the altitude of 17–20 km with maximum anomalies during August.
The similar phase of the UT and LS temperature anomalies over Chennai is consistent with the temperature anomalies over Gadanki (Mehta et al., 2011), located 120 km northwest of Chennai. However, it is to be noted that the UT and LS temperature anomalies are out of phase over the tropical western pacific (Reid et al., 1996). It can be seen that the CPT temperature anomalies have similar features as the LS temperature anomalies indicating that CPT is strongly influenced by wave driving mechanism (Yuleva et al., 1994), which controls the annual cycle of the LS temperature. However, the COT temperature anomalies are opposite in phase to both UT and LS temperature anomalies. The UT temperature and COT height are both governed by convective processes, and their annual cycles are the same as the annual cycle of the OLR. Thus, the UT temperature shows higher temperatures during the summer and lower temperatures during the winter; however, the COT is higher and colder in contrast to the UT temperature during the summer.
The water vapor anomalies also are in phase in the UT and LS, similar to temperature anomalies. The UT water vapor anomalies are higher during the summer due to the monsoon season; however, LS water vapor anomalies are higher during the post-monsoon season. It can be seen that the water vapor anomalies become higher when the LS temperature becomes colder. It is well known that the CPT temperature controls the water vapor entry to the LS. The seasonal variation of the ozone in the LS is well known due to variations in the incoming solar radiation. The higher the insolation, the higher the ozone production in the LS during the summer monsoon season. However, higher tropospheric ozone during summer is mainly due to strong convection
To understand the controls of the UT and LS processes in the variation of the CPT during different seasons, we have obtained the correlation coefficient between daily mean temperature, ozone, and water vapor at 18 km (typical LS temperature) with the temperature, ozone, and water vapor at different altitudes from 8 to 22 km as shown Fig. 5. Note that one can take the CPT temperature instead of the temperature at 18 km. Here we have taken an 18 km fixed height to avoid the complexity due to seasonal variation of the CPT height. In the earlier reports, the correlation is only obtained for the complete data, which does not account for the seasonal changes. However, Chennai experiences both NE and SW monsoons, extremely hot summers, and moderate winters resulting in different thermodynamic conditions during these seasons. Thus, the tropospheric effect on the CPT will be governed differently during the different seasons. From Fig. 5, it can be seen that correlations for the annual (whole data) significantly differ from different seasons. In general, the LS temperature is negatively correlated with tropospheric temperatures over the tropical pacific, which has been discussed in earlier studies (Gage and Reid, 1982; Reid, 1994) that is not true over the Indian monsoon region (Mehta et al., 2011). The annual correlation analysis for the temperature obtained over Chennai shows similar features as over Gadanki (Mehta et al., 2011); however, it has a contrasting seasonal feature.
Interestingly, the correlation profile during the SW monsoon season shows similar features as over the Pacific Ocean. At the same time, the correlation profile during the NE monsoon shows a similar feature as the annual correlation profile. The correlation in the UT is insignificant during the winter and premonsoon seasons. As mentioned earlier, the correlation changes from negative to positive in the transition layer (16–17 km). The poor correlation in the transition layer is due to the complex interaction between the UT and LS. In the LS, the correlation degrades with height due to the increasing effect of the QBO (Reid and Gage, 1996). The correlation profiles of the ozone and water vapor in the UT and LS are similar to the temperature correlation except for the transition layer. Thus, radiative heating due to ozone and water vapor could be an important factor for the observed profile of the temperature correlation.
As there is a complex interplay between the UT and LS, the annual cycles of the tropopause significantly deviate from the linearity associated with high tropopause potential temperature (Mehta et al., 2011, Reid and Gage, 1996). To further understand the roles of the annual cycle of the tropopause, we have also examined the annual cycles of the water vapor and ozone ate the tropopause. The monthly mean CPT-H, with CPT-T, CPT-O, and CPT-W, are shown in Fig. 6. Similar to the earlier reports, we also observed that the relationship between CPT-H and CPT-T is not linear due to complex radiative and dynamical effects that act differently on the CPT-H and CPT-T. It is observed that CPT-T increases with decreasing CPT-H from June to August, while CPT-T decreases with increasing CPT-H from September to November. Thus, CPT-H and CPT-T are linearly related from June to November. However, from December to May, CPT-H and CPT-T deviated from linearity. Such deviation from the liberality has been attributed due to additional heat input. To understand the radiative heating due to ozone and water vapor, we have also estimated the linear relation between the CPT-H, CPT-O, and CPT-W. It is interesting to observe that the CPT-H and CPT-O are linearly related during the same period from July to November. However, CPT-H and CPT-W are more or less linearly related throughout the years except during July. Over Chennai, it can be seen that the deviation during the linear relationship is due to an increase in the potential temperature at CPT (Figure not shown).
3.5 Role of convection on the UTLS temperature, water vapor, and ozone
Over Chennai, both NE and SW monsoons prevail from June to November, leading to frequent occurrences of convective days. We have segregated the convective and clear sky days during the different seasons following Meenu et al. (2010). Figure 7 shows the distributions of the clear sky days and cloudy days, such as VSC, SC, CC, DC, and VDC, during different seasons and for the entire year (annual). During the winter season, mostly (95% cases) clear sky prevails over Chennai, followed by ~ 3% cases of VSC and 1% cases of SC. Similarly, the premonsoon season mainly was (84% cases) dominated by clear sky conditions, followed by 4% cases of VSC, 7% cases of SC,1% cases of VDC. As mentioned earlier, Chennai experiences both SW and NE monsoons, and all types of convections occur during these seasons. The SW monsoon seasons are dominated by 49% clear-sky days and 8%, 17%, 3%, 12%, and 11% cases of VSC, SC, CC, DC, and VDC, respectively. Similarly, during NE monsoon season is dominated by 72% of clear sky days and different types of convections such as VSC (8%), SC (10%), CC (1%), DC (5%), and VDC (4%) cases, respectively.
It is well known that convection plays a significant role in the temperature and the tracers (H2O and O3) distributions, affecting the tropopause structure. To understand the different types of convection playing a role in the tropical tropopause, we have segregated the temperature, water vapor, and ozone profiles for NC, VSC, SC, CC, DC, and VDC cases for the UT and LS, as shown in Fig. 8. It is interesting to observe that the mean temperature, water vapor, and ozone profiles show a marked difference with increasing strength of the convection. In the temperature profile, the effect of the convection is most pronounced in the LS. The temperature profile is the coldest for the clear sky cases, which becomes warmer for VSC, SC, and CC and the warmest for the DC and VDC. The temperature anomalies of the convective cases obtained with respect to clear sky cases indicate a strong warning anomaly (0.8–1.6 K) in the mid-troposphere ~ 8–13 km with warming at ~ 10 km (Fig. 8a). The magnitude of the warming increases with the increasing convective strength. At the same time, an anomalous cooling (~ 0.7 K) just below the CPT in the upper troposphere (14–17 km) is observed. Such cooling is attributed due to the dry-adiabatic ascent caused by the mesoscale updraft and cloud-top radiative forcings. However, anomalously warmer LS temperature (above the CPT) can be attributed due to the subsidence of ozone-rich air from the lower stratosphere in association with the convective overshooting. The LS warming varies from 1.6–4.6 K, followed by increasing convective strength (Muhsin et al.,2018; Sherwood and Wahrlich, 1999).
The mean ozone profiles for different types of convections are shown in Fig. 8b. It can be seen that the ozone mixing ratio gradually decreases with the increasing strength of the convection. Note that CC cases show an exceptionally lower mixing ratio when compared to DC and VDC cases. However, higher ozone is also observed in the LS for the convective cases when compared to the clear sky cases. The lower ozone mixing ratio in the mid and upper troposphere can be attributed due to the loss of the tropospheric ozone by scavenging. While the higher ozone in the LS can be attributed to subsidence following the convection. Figure 8c shows the mean water vapor profiles during different convections. The water vapor is found to be higher during the convection when compared to the clear sky cases. In the LS, the water vapor mixing ratio is ~ 3 ppmv during the clear sky while ~ 4–5 ppmv during the convective cases. Thus, an increase in the LS temperature following the convection can be associated with reading heating due to an increase in the ozone and water vapor mixing ratio.
Figure 9 shows the mean CPT-H, CPT-T, CPT-O, CPT-W, COT-H, and TTL thickness along with their standard deviations for different convective cases. In general, CPT-H lowers with increasing strength of convection, whereas the CPT-T becomes warmer. We observed that CPT is the highest (~ 17.2 km) and the coldest (~ 190.2 K) during NC cases. It gradually lowers and becomes warmer with increasing strength of convections from VSC to CC. It lowers by ~ 0.5 km and becomes warmer by ~ 1.0 K for the CC cases. However, CPT-H becomes higher by ~ 0.1 km and colder by 0.8 K for VDC cases. As mentioned, the CPT lowers due to warming by the latent heat released from very shallow and convective clouds. However, as the convective clouds during VDC often reach higher altitudes close to the tropopause and even penetrate sometimes, the mean CPT relatively becomes higher and colder compared to CC cases. During the convective processes, enormous amounts of water vapor reach the upper troposphere, which is susceptible to crossing into the lower stratosphere. We observed that the water vapor mixing ratio at the CPT increases from 3.7 ppmv during NC cases to 4.5 ppmv with an increase in the strength of the convection. However, in the DC and VDC cases, the water vapor at the CPT remains roughly the same as in the CC cases.
Similarly, the ozone mixing ratio at the CPT also increases from 84 ppbv to 124 ppbv with the growing strength of the convection from VSC to CC. Such an increase in the ozone mixing ratio at the CPT during the convection appears related to the intrusion of the ozone-rich lower stratospheric air into the upper troposphere due to subsidence. However, for DC and VDC, the ozone mixing ratio at the CPT decreases from 124 ppmv to 100 ppmv. Such a decrease in the ozone mixing ratio could be due to the higher ozone destruction due to an increase in the water vapor mixing ratio.
We observed that the COT-H increases with increasing convective from VSC to CC and slightly decreases during DC and VDC. The increase in the COT-H is directly related to an increase in the top of the convection with increasing convective strength. However, during DC and VDCC, though convention penetrates the tropopause, the main outflow levels generally occur at the lower level. Thus, the COT-H is lower for penetrative convection than for non-penetrative convections (Sunilkumar et al., 2017; Hemanth et al., 2015). It is observed that the TTL thickness decreases with increasing strength of the convection from the VSC to CC; however, it slightly increases during the deep convective cases. The decrease in the TTL thickness is well-known during the convection due to an increase in the COT-H and a decrease in the CPT-H. During the clear sky condition, the TTL is relatively thicker, ~ 4–5 km; however, it decreases to 2–3 km during the convective days. The decrease in the TTL thickness may have important consequences on the decrease in the tropopause stability and hence an increase in the exchange processes between the upper troposphere and the lower stratosphere.
Figure 10a shows the mean temperature, potential temperature gradient, water vapor mixing ratio, and ozone mixing ratio for the clear sky (no convection), non-penetrative clouds (e.g., CC), and penetrative clouds (e.g., VDC). For the non-penetrative clouds, the TTL thickness decreases compared to clear sky cases. However, for penetrative clouds, the TTL thickness increase relative to non-penetrative clouds. The results mentioned above are illustrated by the schematic of shallow convection, non-penetrative clouds, and penetrative clouds (Kuang and Bretherton, 2004) are shown in Fig. 10b. The shallow convection is associated with weaker convergence and divergence at the lower troposphere ~ 5–6 km and subsequent release of less latent heat. Though the main divergence occurs much lower for shallow convection cases, secondary divergence will occur closer to the CPT (Mehta et al., 2008), taken as the COT. The COT, in this case, is relatively higher than the clear sky cases due to dry adiabatic ascent. However, CPT becomes warmer due to latent heat release and lowers relative to clear sky cases. For CC cases, convergence is stronger, leading to stronger divergence in the mid-troposphere ~ 10–11 km associated with stronger latent heat release. In this case, the secondary divergence close to the CPT is also taken as the COT. The latent heat release consequently lowers the CPT further than the shallow convection cases. However, for the penetrative clouds, the stronger divergence is associated with stronger convergence and an enormous amount of latent heat release. In this case, the main convective outflow level occurs at ~ 13 km, taken as the COT, as the secondary outflow even occurs above the CPT (Sunilkumar et al., 2010).
The penetrative deep convection is strongly coupled to the CPT (Kuang and Bretherton, 2004). In general, for the non-penetrative convection, the air parcel is adiabatically driven within a convective updraft and will no longer be positively buoyant above neutral buoyancy. However, the air parcel overshoots the level of neutral buoyancy and cools adiabatically for penetrative convection. Thus, CPT raises instead of lowering due to dry adiabatic ascent overcomes the lowering by latent heat release. Note that the subsidence associated with deep convection will also result in intrusion of the ozone-rich air and hence warming the CPT. The exact interaction among these processes is rather complex; our observation indicates higher and colder CPT associated with penetrative convection. The BDC removes the rising tropical air to the polar region (Yuleva et al., 1994; Reid and Gage, 1996)