1 Understanding convection features over Bay of Bengal using SST and 2 atmospheric variables 3 4 R. Uma, T.V. Lakshmi Kumar and M.S. Narayanan 5 Department of Physics, SRM University, Kattankulathur, Tamilnadu, India 6 7 Abstract 8 Tropical Oceanic region are frequently prone to deep convections. Hence it is very 9 essential to understand the features of convection with the help of oceanic and atmospheric 10 variables such as SST, OLR, Rainfall, relative humidity, Columnar Water Vapour (CWV) etc 11 and the linkage among them. In our present study, we have divided the Bay of Bengal (BoB) 12 into ten different sub regions (SR) and have attempted to study the connection between the above 13 stated variables during convective and non-convective events in the southwest monsoon (SWM) 14 season (June to September) for the period 1998 - 2010. The monthly behaviour of SST / OLR 15 decreased by 0.5 °C / 14 W/m2 from May to June and increased by 0.1 °C / 7 W/m2 from 16 September to October. Among ten SRs, SR 5 and SR 10 are observed to be coldest and warmest 17 respectively based on the SST variations. Intra-seasonal oscillations of the above mentioned 18 variables show the influences of quasi bi-weekly oscillations (QBWO) and Madden Julian 19 Oscillations (MJO). As the threshold values for SST, OLR and rainfall were already reported, we 20 have drawn our attention to deduce a threshold value for water vapour in lower level troposphere 21 (Water vapour density (WVD) at 850 mb) which highly influences the convection. In arriving at 22 a threshold of low level water vapour, we have analysed the convective and non-convective 23 events of each central 1β° x 1β° grid in all the SRs for the period from 1998 to 2010, along with 24 water vapour scale height. Our analysis inferred that the low level water vapour density at 1 25 850 mb varied above 12 g/m3during convective days and below 12 g/m3during non-convective 26 days. We noticed that the variability in water vapour density is more in non convective days than 27 in convective days over BoB. The results of the study may be useful to understand the water 28 vapour dynamics with SST, OLR and rainfall. 29 Keyword: Sea Surface Temperature, OLR, water vapor density, Convection, and Bay of Bengal 30 Author for correspondence: 31 32 33 34 35 36 37 38 Dr.T.V. Lakshmi Kumar Assistant Professor Department of Physics SRM University Kattankulathur, Tamilnadu, India – 603 203 E mail : lkumarap@gmail.com 2 39 1 Introduction 40 Convection is a physical process which makes the earth liveable by preventing the earth 41 from overheating due to solar radiation through vertical transport of heat and moisture in an 42 unstable atmosphere. This is the principle mechanism for the formation of clouds and 43 atmospheric circulation, which is responsible for the redistribution of heat from warm equatorial 44 regions to higher latitudes. Over the globe, oceans are dominant heat sources for the earth’s 45 troposphere as they are more capable of absorbing, retaining and transferring heat when 46 compared to land (Waliser and Graham 1993). Tropical Oceans are observed to be the regions 47 prone to frequent deep convections (Vinayachandran and Shetye et al. 1991). Hence, it is very 48 essential to understand the features of convection with the help of oceanic and atmospheric 49 variables such as SST, OLR, rainfall, relative humidity, CWV and linkage among them. 50 However, the basic element which drives the convection is the vertical transport of wind over a 51 specific region. The data provided by the satellite sensors of multiple channels covering the 52 whole ocean at finer saptio-temporal scale is of immense help to understand the convection 53 features. 54 Many aspects of relationship between SST and large scale convective systems using OLR 55 for deep convection has been revealed by several researchers (Waliser and Graham 1993, Zhang 56 1993). One of the major outcome of their studies is that intense deep convection occurs more 57 frequently when the SST values are in the range of 27 – 29 °C (Gadgil et al. 1984; Sud et al. 58 1999). They have used OLR (OLR<240 W/m2) as a tool to identify the location of deep 59 convection and reported that both intensity as well as the occurrence of deep convection 60 diminishes as SST increases beyond 29.5 °C. 3 61 Also, a few investigations had been carried out to infer the connection of SST with water 62 vapour at different levels of troposphere and deep convection, ( Holloway and Neelin et al. 2009; 63 Sherwood 2010) as water vapour involves in transfer of heat and moisture through updraft and 64 downdraft. The linkage of heavy rainfall events in relation to SST and water vapour as a 65 resultant of convection were also reported (Shankar et al. 2007; Zelinka and Hartman 2009). 66 Sherwood (1999a) reported that the humidity of lower troposphere is an important precursor for 67 initiation of deep convection in tropical western pacific. By conditionally averaging the 68 precipitation on CWV, Bretherton (2004) has noticed a sharp rise in rainfall at relatively high 69 CWV. Behaviour of moisture in terms of humidity is highly variable both vertically from lower 70 troposphere (850 mb) to upper (200 mb) troposphere and horizontally from warm pool to cold 71 pool when compared to temperature. Thus, positive feedback of water vapour plays a major role 72 in initiation and sustainability of deep convection. 73 As BoB is one of the most important heat source region of tropics and is prone to 74 frequent cyclonic activities, we have attempted to understand the features of convection over 75 BoB (0° to 20° N and 82° to 97° E) with the help of oceanic and atmospheric variables, such as 76 SST, OLR, Rainfall and CWV. In our present work, we first studied the variations in SST, OLR 77 during SWM season with an emphasis on convective and non- convective events. Further, we 78 have used wavelet analysis to study the intra-seasonal oscillations and focused on changes of 79 water vapour during convection with an aim to report threshold for water vapour of lower 80 troposphere on convective days over BoB. 81 4 82 2 Data and Methodology 83 The present study makes use of gridded SST, OLR, rainfall data sets for the analysis. The 84 authors 85 http://www.esrl.noaa.gov/psd/data/gridded/data ) and TMPA 3B42 V6 rainfall data sets (Data 86 source: http://disc.gsfc.nasa.gov ) on daily basis at a spatial resolution of 1° x 1° by re-gridding 87 the data sets which are available at a finer spatial resolution of 0.25° x 0.25°. CWV in the free 88 troposphere (850 – 500 mb) has been calculated by integrating the ERA interim (Data source: 89 http://apps.ecmwf.int/datasets/data/interim-full-daily ) specific humidity data from the 850 – 500 90 mb using Trapezoidal rule (maintaining the same spatial resolution of 1° x 1°).Daily OLR values 91 provided by CDC – NOAA (Data source: http://www.esrl.noaa.gov/psd/data/gridded/data ) at a 92 spatial resolution of 2.5° x 2.5°, were used as a proxy for identification of convective event. 93 These data sets are being widely used to study the atmospheric and oceanic processes (Shankar 94 2007, Holloway and Neelin et al. 2009). have used NOAA Optimally Interpolated SST (Data source: 95 The present study is confined to SWM season of India from 1998 – 2010. Ten 5° x 5° 96 SRs over Bay of Bengal (0° to 20° N and 82° to 97° E) are chosen such that each one of the SR 97 comprises of twenty five 1° x 1° grids (Fig.1). The analysis commenced by obtaining the daily 98 and monthly mean values of OLR, SST, rainfall and CWV for each of the ten 5° x 5° SRs and 99 study region as a whole. To study the Interasrasonal Variability (ISV) of the above mentioned 100 parameters wavelet analysis has been used. Wavelet transforms (WT) is a significant tool to 101 extract the multi-frequency non-stationary powers of a time series. Wavelets is a family of 102 functions constructed from translations and dilations of a single function called the "mother 103 wavelet"ψ(t). Morlet wavelet which is used as mother wavelet in our present study excels over 5 104 the other wavelets due to its ability to detect both time-dependent amplitude and phase for 105 different frequencies exhibited in the time series. It is defined by 106 ππ,π (π) = π √|π| π−π π( π ), π, πππΉ, π ≠ π (1) 107 where, a and b are scaling and translation parameter which measures the degree of 108 compression and translation respectively which determines the time location of the wavelet. 109 Wavelets have time-widths adapted to their frequencies. Wavelet transforms is significantly 110 applied in analysing non-stationary signal when compared to Fourier transform. Remarkable 111 difference between Fourier transform and Wavelet transform is that wavelets are well localized 112 in both time and frequency domain whereas the former is only localized in frequency domain. 113 Capability of Wavelet theory to reveal aspects of data such as trends, breakdown points, and 114 discontinuities in higher derivatives and self-similarity makes it as an excellent signal processing 115 technique which is missed by other signal analysis techniques (Sifuzzaman et al. 2009). 116 Theoretical details of wavelet analysis can be referred from Daubechies (1992)” 117 118 3 Region of interest 119 BoB is prone to frequent monsoon depressions and cyclonic disturbances. Hence it is 120 very important to study the ocean- atmosphere interaction over this region which helps to 121 understand the cyclogenisis, estimation of cyclone intensity etc. The study region (BoB) is 122 proven to be having a non-linearity between the precipitation and SST and there is a spatially 123 variable lag relationship between them (Roxy et al. 2013). Over the Arabian Sea, surface 124 convergence is strong enough to uplift the moisture which results in fast local convection where 125 as in BoB it is relatively weak hence the response in SST to the precipitation anomaly is slower. 6 126 Roxy (2014) have attempted to elucidate the relation between SST and rainfall in a large scale by 127 considering three different regions in Arabian Sea, BoB and South China Sea and concluded that 128 the lag between these two variables must be considered for the better understanding of the 129 convection phenomenon. It is also reported that the existence of upper threshold (29 °C) of SST 130 for convection as reported by Gadgil et al. (1984) is uncertain and convection takes place even at 131 SST greater than 30 °C (Roxy 2014). However, Roxy (2014) has not considered the threshold for 132 rainfall as an outcome of convection in the analysis. In our present analysis we have divided the 133 study region in to ten SRs where some regions are found to be more convective regions while 134 others are as low convective ones. These 10 SRs are also distinctly varied in terms of rainfall and 135 water vapor. 136 4 Definition of a convective event 137 We have considered a day as a convective day/event when all the three parameters discussed 138 above namely SST, OLR and rainfall attain their threshold values. Their threshold values are 139 mentioned below: 140 ο· SST: 27.5 °C < SST < 29 °C – Gadgil et al. (1984) 141 ο· OLR: OLR < 240 W/m2 – Zhang et al. (1993). 142 ο· Rainfall: Rainfall > 20 mm – Shankar et al. (2007). 143 It is to be noted that the authors of the present paper have considered the wind criteria prior to 144 the above mentioned criteria in carrying out the work. The ascending wind at 500 mb is taken as 145 convective event while the descending one is treated as non-convective. 146 5 Results and Discussions 7 147 5.1 Convective features of SST and OLR over BoB during SWM season 148 The long term monthly mean (1998 - 2010) and standard deviation (SD) values of SST 149 have been calculated and plotted for all individual SRs (for brevity sake, all the plots are not 150 shown). From the mean values of SST, we observed that among the ten SRs, SR 5/ SR 10 is 151 found to be the coldest/ warmest with a mean SST of 28.6 °C / 29.3 °C during SWM season 152 (Fig.2). SR 4 is found to be the highly convective SR from the frequency count of the maximum 153 number of convective events which has been elucidated in the forthcoming sections. The reason 154 for SR 5 to be the coldest is attributed to the intrusion of southwest monsoon current in this SR 155 during the period of May to September which cools the southern BoB (Shankar et al. 2007). 156 Vinayachandran and Shetye (1991) studied the structure of warm pools of equatorial Indian 157 Ocean (Arabian Sea and Bay of Bengal) and observed that the SST over the SR 10 of our present 158 study to be having the maximum SST during SWM season of the order of 29 - 30 °C. 159 Indian Ocean becomes the warmest area among the world oceans during April and May. 160 As it encounters the onset of SWM season, the SST values reduce due to strengthening of wind 161 circulation and remains in a range of 27 – 29 °C in the Bay during the months of June to August. 162 Then SST starts increasing from the mid September (Shenoi et al. 2002). The fall of SST is 163 found to be higher in SR 3 and 4 with 0.7 °C and 0.9 °C respectively as where SR 9 and 10 have 164 recorded lower values of 0.2 °C and 0.1 °C respectively. There is no significant increase of SST 165 in individual SRs from September to October. From the monthly mean and SD of all individual 166 SRs (figure not shown for all SRs), it has been observed that the SRs which are in the latitudinal 167 belt 5° – 20° N show a similar monthly behaviour as stated by Shenoi et al. (2002) and those 8 168 which are very near to the equator belt show a decrement in SST values throughout May to 169 October and seems to be warmer than other SRs with an SST range of 29 – 30 °C. 170 OLR being the measure of upwelling thermal radiation, its variability is highly dependent 171 on the SST variation. Lower/higher value of OLR indicates the presence of high cloud tops/low 172 clouds or clear conditions (Waliser and Graham 1993). Due to the above mentioned reasons, 173 OLR has been considered as a proxy to indentify deep convection in many studies to understand 174 the convection mechanism (Zhang 1993). On this basis, we carried out a similar daily and 175 monthly analysis with OLR data sets and observed that the monthly mean values of all ten SRs 176 are below 240 W/m2 which is reported as the threshold for the occurrence of convection (Zhang 177 et al. 1993). Fig.3 shows the monthly mean and SD of OLR for the cold SR (SR 5), warm SR 178 (SR 10) and highly convective SR (SR 4). Among the ten individual SRs (figure not shown for 179 all individual SRs), monthly mean OLR is observed to be lowest in SRs 3 and 4 and highest in 180 SR 5 during the SWM season. From Fig.3 it can be noted that OLR declined (14 W/m2) from 181 May to June, and increased (7 W/m2) from September to October . Alike SST, the monthly 182 behaviour of OLR is also inferred to be uniform in the SRs 1 to 7 (5° to 20° N). 183 5.2 Linkage between rainfall and convection 184 Rainfall being considered as the major outcome of a convective event, we used it as a 185 parameter to achieve our goal of understanding the features of convection. In the present study, 186 we made use of the rainfall in framing the criteria for a convective event/day over BoB which is 187 mentioned in the data and methodology section. 9 188 Based on the criteria for a convective event, we have identified the number of convective 189 days/events in all the individual SRs. Table 1 gives the information about the number of 190 convective events that occurred in each 5β° x 5β° SR during SWM season of the study period. 191 From Table 1, it is obvious that SR 4 is highly convective with a record of 266 192 convective events followed by SR 1(Number of events: 246) and SR 3(Number of events: 163). 193 These are the SRs located in the northern bay having the favourable conditions for deep 194 convective event with daily mean SST (27 – 29 °C) and OLR (185 – 205 W/m2).These results 195 show good agreement with the findings of Mooley and Shukla (1989) where, they reported that 196 the formation of number of low pressure system exceeds in northern Bay than in the other part of 197 the North Indian Ocean. Number of convective events are observed to be very low in the SRs 8, 198 9 and 10 (Table 1). This could be due to the impact of higher SSTs beyond the threshold i.e 199 above 29 °C in these SRs. The convective activity and intensity is reported to be decreasing with 200 increasing SST beyond the threshold due to evaporative cooling (Zhang et al.1995) and super 201 green house effect as reported by Imandar and Ramanathan 1994. 202 5.3 Intraseasonal variability (ISV) 203 It is reported that the ISV (10 – 60 day Oscillations) of SST generates favourable 204 conditions for convective activity (Roxy and Tanimoto 2007). On this basis, we attempted to 205 study the ISV of all the four variables over the study area. ISV of the SWM season is 206 characterized by two dominant modes, one between 10 and 20 days (Quasi Bi-Weekly 207 Oscillation (QBWO)) with a westward or north-westward propagation pattern over the monsoon 208 region and the other between 30 and 60 days which exhibits a northward or north-eastward 209 propagation called Madden Julian Oscillation (MJO), (Krishnamurti and Ardanuy 1980; 10 210 Goswami and Ajayamohan 2001). We studied these special characteristic features by using 211 wavelet power spectra. 212 Wavelet analysis has been widely used to reveal the characteristics and temporal 213 dynamics of various parameters like rainfall-river runoff (Lafreniere and Sharp 2002), large scale 214 features of rainfall over Indian landmass (Uma et al. 2013), SST and South Asian Summer 215 monsoon (Torrence and Compo 1998; Vianayachandran et al. 2012), and their relationship at 216 diurnal, intra-seasonal (10 – 20day (QBWO), 30-60 day (MJO)), intra-annual (2-7 years ENSO) 217 time scales. (Wang and Wang 1996) 218 Time series of SST, OLR, rainfall and CWV are subjected to “Morelet” wavelet analysis 219 for the period May to October to avoid the impact of edge effect. Fig.4 depicts the wavelet 220 power (absolute value squared) spectra of SST (a and e), OLR (b and f), rainfall (c and g) and 221 CWV (d and h) of the study region during a weak monsoon (2002) and an active monsoon 222 (2007) respectively. The area under the black line is the cone of influence. From Fig.4 (a-h), we 223 observe the presence of 30 – 60 days low frequency oscillations and 10 – 20 days high frequency 224 oscillations in all the four parameters. The spectra of all the four parameters exhibit more or less 225 similar pattern of maximum power in early June and the power slowly increases from 45 days 226 (spread of monsoon) and extends up to 120 days (July – August major monsoon months) and 227 starts decreasing as the season withdrawal takes place (Lawrance and Webster 2000). 228 Strength of the ISV in SST and OLR are observed to be high during good monsoon 229 (2007) than in a deficit monsoon (2002). Though the strength of the 10- 20 day oscillations of 230 SST and OLR are low, they are observed to be significant with 95% confidence level. Study of 231 Hareesh et al. (2001) on ISV in the central Bay during 1999 SWM season, have revealed the 232 existence of 10 – 32 days oscillation by applying Fast Fourier Transform (FFT) to SST time 11 233 series. The results of Han et al. (2006) on the impact of these sub monthly oscillations of wind 234 and OLR has shown that it can cause significant 10-30 days changes in SST over equatorial 235 basin and northern BoB. Spatially, this results with warming in the central western Indian Ocean 236 basin and northern BoB, cooling in the eastern equatorial warm pool as well as southern Bay 237 before the event peak. This higher amplitude of ISV of SST is attributed to the air-sea fluxes and 238 role of salinity since BoB receives more amount of fresh water from river runoff and excess of 239 rainfall (Hareesh et al. 2001). 240 SWM seasonal rainfall is reported to be very high over BoB and the number of low 241 pressure systems is also high when compared to Arabian Sea (Mooley and Shukla 1989). ISV of 242 the rainfall is observed to be high over the north-eastern BoB (Vinayachandran et al. 2012). 243 Coinciding with these findings, spectrum of rainfall over SRs 1, 3 and 4 (figure not shown) of 244 our present study also exhibited a strong ISV with higher power. 245 We also studied the wavelet power spectra of SST, OLR, rainfall and CWV of the highly 246 convective area (SR 4) located in the north-eastern Bay during a poor monsoon (2002) and good 247 (previously it was i) weak and active and ii) bad and good monsoon in few places, MSN sir 248 wanted to denote it as poor and good hence I have followed the same)monsoon (2007) (figure 249 not shown). Similar features are observed in the wavelet spectrum of CWV with distinct 250 variations in its power. The localized variations of 10 - 20 day seemed to be existing in both 251 good and poor monsoon years. The strength of the signal was found to be consistent in poor 252 monsoon year 2002 than in the good monsoon year 2007. From the above analysis, we conclude 253 that SST and other atmospheric variables have by and large similar variations over the study 254 area. Though the wavelet spectra are capable in depicting the intra seasonal variations of SST 255 and other atmospheric variables, the lead and lag of these parameters cannot be explored. It is 12 256 also important to see that the SSTs over a region influence the convection or the advective winds 257 modify the SSTs to cause deep convection. Bielli and Hartmann (2004) reported that the zonal 258 winds at the coastal areas of BoB travel to deep ocean thus modifying the SSTs over that region 259 to initiate convection. Similarly, Lanzante (1996) found that the SSTs maintained the lag 260 correlation of 10 - 12 days with the wind and has the 1-12 days of lead with the rainfall over a 261 region. Satheesan and Krishnamurthy (2005) studied the frequency spectrum coherence analysis 262 to identify the lead-lag between the parameters of zonal and meridional wind and temperature 263 and reported that they have 6 day and 12 day lag relation over Gadanki (13.5β° N, 79.2β° E). In the 264 present study, we tried to examine the lead and lag of SST and CWV over a few 1β° x 1β° grid 265 locations of the study area. We chose three 1β° x 1β° grids each from hot SR, cold SR and 266 convective SR of the study area and subjected to FFT. 267 Table 2 illustrates the lead and lag of SST/CWV during the study period over the above 268 mentioned SRs. The lead variable is mentioned in the table and the hyphen represents the zero 269 lead / lag. The lead lag showed that the CWV and SSTs maintained 6 to 10 days either lead or 270 lag during the study period. In the cold SR, most of the years have witnessed CWV leading the 271 SSTs. The SW monsoon winds from Arabian Sea converge at these regions causing the lower 272 SSTs which in turn the CWV predominantly leads the SSTs. In the hot SR as well as in the most 273 convective SR, the situation is mixed. In some of the years, CWV is leading and in the other 274 years, SST showed the lead. There is no much lead-lag relation between CWV and SSTs during 275 the years 2001 and 2002 of the study period. This might be due to the convection caused by local 276 SSTs or the convection caused by advective winds. 277 13 278 5.4 Relation between water vapor, rainfall and convection 279 It is a crucial task to understand the spatio-temporal variation and relation of water 280 vapour with convection. Free tropospheric (850 – 500 mb) moisture is reported to be more 281 variable when compared to other layers of the atmosphere (Holloway and Neelin 2009). They 282 found that moisture profile, conditionally averaged on precipitation over Nauru Island showed a 283 strong association between rainfall and moisture variability in the free troposphere and boundary 284 layer variability. This is also been confirmed from the observational studies of Brotherton et al. 285 2004 where they have observed a sharp rise in the rainfall at higher CWV. 286 We examined to see the percentage departure of CWV from the day of convection to the 287 very next day of convection by taking each 1β° x 1β° grid of the 5β° x 5β° SR 1. We resulted with 288 totally 2417 convective events in which 996 showed a positive anomaly and 1421 number of 289 events showed negative anomaly which shows that generally there is a decrease in CWV just 290 after the convective event. It is reported that intense precipitation events in the tropics are 291 preceded by an increase in low level humidity. Convection removes water vapour from a very 292 moist boundary layer through two processes i) precipitation and ii) transportation to the very dry 293 upper troposphere, resulting in dramatic changes to the upper-level moisture (Zelinka and 294 Hartmann 2009). This could be a probable reason behind the higher number of events showing 295 negative anomaly during the very next day of convection in our present analysis. 296 Understanding the change in the humidity at different altitudes during different stages of 297 evolution of the convective events is very important. In the present study, we tried to delineate 298 the behaviour of low level water vapour (850 mb) expressed in terms of water vapour density. 299 Studies revealed that even small level change in the boundary layer moisture can greatly 14 300 influence the initiation of convective activity (Keil 2008; Sherwood 2010). It has been stated by 301 Weckwreth et al. (2005) and Fabry et al. (2006) from their radar reflectivity observations that 302 enhanced low level moisture occurs prior to convection initiation. Most of the analysis are 303 carried out by taking individual convective events like thunderstorm, hail, squall lines etc. It has 304 to be noted that in our case, we are defining a convective event on a daily basis rather than 305 considering each and every individual convective events. Due to the same reason, we did not 306 extend our analysis to upper level water vapour, since it varies more significantly during deep 307 convection and is short lived. 308 For the current analysis, we took central 1β° x 1β° grid from each 5β° x 5β° SR, totally ten 309 1β° x 1β° grid from ten 5β° x 5β° SRs and attempted to observe the variation of water vapour density 310 during individual convective (satisfies the criteria of vertical wind velocity at 500mb, OLR, 311 rainfall and SST) and non-convective (doesn't obey any of the four criteria) days. The number of 312 individual convective events occurred during SWM season of the study period (1998 - 2010), 313 mean and SD of water vapour density (WVD) during convective and non-convective events for 314 each individual 1β° x 1β° SR are calculated. 315 From the calculated values, we noticed that the variability of low level WVD is observed 316 to be high for non convective events for all the SRs and number of convective events is found to 317 be more than non convective events in SR 4 where as in other SRs the non convective events are 318 high. Fig. 5(a) shows the plot for number of convective events and mean WVD at 850 mb along 319 with their SD for the period 1998 – 2010. The data on primary Y-axis represent the cumulative 320 number of convective days of the central 1β° x 1β° grids from all SRs. The secondary Y- axis 321 denotes the mean WVD at 850 mb along with its SD of the respective boxes. From Fig. 5(a) it is 15 322 observed that the entire study period has shown a WVD of > 12 g/m3 in all years. The 323 corresponding number of convective days varied from 28 days (in the year 2002) to a maximum 324 of 113 (in the year 1999). Similarly from Fig. 5(b) it is understood that the WVD at 850 mb <12 325 g/m3 in all the years. It is also noticed from Fig .5(a and b) the WVD at 850 mb has more SD of 326 about 0.6 during the non-convective days when compared to convective days (SD=0.2). 327 In the process of finding the threshold value of WVD at 850 mb, we analysed the water 328 vapour separately for convective and non-convective days of the entire study period and we 329 inferred the frequency count of WVD at 850 mb with intervals ranging from 1 to 4 g/m3. This 330 analysis helped us to understand the variability of water vapour density along with the other test 331 variables. Deriving a mathematical construct for defining a threshold is a difficult task. However, 332 we considered all the convective and non-convective events for the entire study period and 333 performed the frequency count analysis. 334 Figure 6(a and b) illustrates the percentage contribution of a) convective and b) non- 335 convective events with WVD lesser/greater than 12 g/m3 respectively. From Fig.6(a and b), it is 336 clear that upon the total 1008/1359 convective/non-convective events WVD at 850 mb of 86% / 337 77 % of the events are found to be above/below 12 g/m3. We carried out the same analysis by 338 eliminating the upper limit in the SST criterion, in view of recent study of Roxy (2014). Though 339 the number of convective events increased to 1653 there is no change in the percentage of days 340 having WVD above 12 g/m3. 341 The recent review of Sherwood et al. (2010) on understanding the convective interaction 342 with water vapour and changes associated with water vapour in warmer climate have revealed a 343 very important point that it is the relative humidity which has to be focused rather than specific 344 humidity since, it is closely associated with the change in surface and environmental 16 345 temperature. To be precise, relative humidity is more dependent on lapse rate. Based on the work 346 of Tomasi (1984), we did a preliminary analysis and tabulated the temperature, relative humidity 347 and WVD at different altitudes in the free troposphere with regular interval of 0.5 km during the 348 selected convective and non-convective days. 349 Table 3 shows the mean temperature, relative humidity and WVD in the free troposphere 350 for the average of all the ten selected central 1β° x 1β° grids of each SR at different altitudes during 351 convective and non-convective days. From Table 3 it has been observed that the temperature 352 variability is very small even though it influences the relative humidity and WVD varies greatly 353 during convective and non-convective days. During convective days, from 850 mb to the levels 354 of boundary layer (2- 3 km) the relative humidity is found to be decreasing and then starts 355 increasing till 5.5 km where as in non-convective days it shows a similar behaviour of increase 356 up to 4 km after a lull in the boundary layer and in addition to that a decrement is observed 357 further if we go up to 5 .5 km. 358 We also examined the water vapour scale height by using also the surface level water 359 vapour density during convective and non convective days. This analysis not only helped to 360 derive the ranges of scale height during convective and non-convective days but also to 361 understand the variations in WVD with the scale height. The scale height of WVD ranged from 362 1.3 km to 2.8 km during the non-convective days, in case of convective days, the scale height 363 varied between 3 km to 3.5 km. Hence, it may be noticed that a marginal scale height of 2.9 km 364 can be considered to distinguish convective and non-convective days over BoB during the study 365 period. 366 17 367 6. Conclusions 368 In the present analysis the authors have attempted to understand the features of convection using 369 satellite derived SST and other atmospheric variables over Bay of Bengal. The study concludes 370 the following 371 i) The mean SST and OLR over the study region are following a trend of decrement 372 (0.5 °C) from the month of May to June and increment (0.1 °C) from the month of 373 September to October 374 ii) SRs of spatial average of 5° x 5°, SR 5 and SR 10 (Fig.1) are observed to be the 375 coldest and warmest over the study region from their SST values and SR 4 to be the 376 most convective one since it has recorded maximum number of convective events 377 (266) in the study period. 378 iii) ISV of SST and other atmospheric parameters studied through wavelet analysis 379 showed the dominance of the major periodicities of SWM season, 30 – 60 day (MJO) 380 and 10 – 20 day (QBWO) with significant variability in the power of each parameter 381 during different years. 382 iv) From the lead-lag analysis, it is observed that the CWV and SSTs maintained a 6 to 383 10 days lead or lag during the study period. In the cold SR, most of the years have 384 witnessed CWV leading the SST and in the remaining most convective and warm 385 box, neither a lead nor lag behaviour is observed in the study period. 18 386 v) Decrease in the CWV just after the convective event is illustrated by our analysis of 387 observing the anomaly of CWV immediately after a convective rainfall event in the 388 SR 1 which is located in the head Bay. 389 vi) In our main attempt to explore the existence of a threshold for WVD (850 mb) in 390 lower troposphere, we have arrived at a conclusion that WVD of 86% / 77% of the 391 convective / non convective events are found to be above/below 12 g/m3 by applying 392 the convective criteria to the central 1° x 1° grid of each sub region. Such a distinct 393 demarcation in scale height is found to be at 2.9 km. 394 vii) From the mean vertical profile of temperature, relative humidity and WVD in the free 395 troposphere, for the average of all the ten selected central 1β° x 1β° grids of each 5β° x 5β° 396 SRs, we observed though the temperature variability is very small, it influences the 397 relative humidity and WVD greatly during convective and non-convective days. 398 During convective days from 850 mb to the level of boundary layer (2- 3 km) the 399 relative humidity is found to be decreasing and then starts increasing till 5.5 km 400 where as in non-convective days it shows a similar behaviour of increase upto 4 km 401 after a lull in the boundary layer and in addition to that a decrement is observed 402 further if we go upto 5.5 km. 403 404 viii) The variability of lower tropospheric WVD is found to be high for non-convective days (0.6 g/m3) when compared to convective days (0.2 g/m3). 405 19 406 Acknowledgement 407 This work was carried out with the support of project funded by ISRO RESPOND. We 408 gratefully acknowledge Dr. B.V. Krishnamurthy, former director of SPL, VSSC, Trivandrum for 409 his useful suggestions and encouragement. We thank D.Selvaraj, Senior Research Fellow, SRM 410 University for programming assistance. Our sincere thanks to the anonymous reviewers for their 411 constructive comments. 412 References 413 Bielli S and Hartmann DL (2004) On Wind, Convection, and SST Variations in the 414 Northeastern Tropical Pacific Associated with the Madden–Julian Oscillation. J. Clim, 17: 415 4080–4088. 416 417 Bretherton C S, Peters ME, and Back LE (2004) Relationships between water vapor path and 418 precipitation over the tropical oceans, J. Clim., 17:1517– 1528. 419 420 Daubechies I (1992) Ten lectures on wavelets, SIAM, 357. 421 422 Fabry F (2006) The spatial variability of moisture in the boundary layer and its effect on 423 convection initiation: Projectβlong characterization. Mon. Weather Rev., 134:79–91. 424 425 Gadgil S, Joseph PV and Joshi NV (1984) Ocean-atmosphere coupling over monsoon 426 regions. Nature.312:141–143. 427 428 Goswami BN and Ajaya Mohan RS (2001) Intraseasonal Oscillations and interannual 429 variability of the Indian Summer monsoon. J. Clim, 14:1180-1198. 430 431 432 20 433 Han W, Timothy Liu W and Lin J (2006) Impact of atmospheric submonthly oscillations on 434 sea surface temperature of the tropical Indian Ocean. Geophys. Res. Lett.,33, L03609, 435 doi:10.1029/2005GL025082. 436 437 Hareesh Kumar PV, Prasada Rao CVK, Swain J and Madhusoodanan P (2001) Intra-seasonal 438 oscillations in the central Bay of Bengal during summer monsoon 1999. Curr. Sci.,80(6):786 439 – 790. 440 441 Holloway CE and Neelin JD (2009) Moisture vertical structure, column water vapor, and 442 tropical deep convection. J. Atmos. Sci., 66:1665–1683. 443 444 Inamdar AK and Ramanathan S (1994) Physics of Greenhouse effect and convection in 445 warm oceans. J.Clim,. 7:715-731. 446 447 Keil C, Rpnack A, Craig GC, and Schumann U (2008) Sensitivity of quantitative 448 precipitation forecast to height dependent changes in humidity. Geophys. Res. Lett., 35, 449 L09812, doi:10.1029/2008GL033657. 450 451 Krishnamurti TN, and Ardanuy P (1980) The 10 to 20 day westward propagating mode and 452 “breaks in the monsoon.” Tellus.32:15–26. 453 454 Lafrenière M and Sharp M (2003) Wavelet analysis of inter-annual variability in the runoff 455 regimes of glacial and nival stream catchments, Bow Lake, Alberta. Hydrol. 456 Process.,17(6):1093 – 1118. 457 458 Lanzante JR (1996) Lag relationships involving Tropical Sea Surface Temperatures. J.Clim., 459 9:2568–2578 460 21 461 Lau KM and Chan PH (1988) Intraseasonal and Interannual Variations of Tropical 462 Convection: A Possible Link between the 40–50 Day Oscillation and ENSO?. J. Atmos. 463 Sci., 45:506–521. 464 465 Lawrence DM, and Webster PJ (2000) Interannual variations of the intraseasonal oscillation 466 in the south Asian summer monsoon region. J Clim 14:2910–2922. 467 468 Mooley D and Shukla J (1989) Main features of the westward moving low pressure systems 469 which form over the Indian region during the summer monsoon season and their relation to 470 the monsoon rainfall. Mausam.40:137–152. 471 472 Roxy M and Tanimoto Y (2007) Role of SST over the Indian Ocean in influencing the 473 Intraseasonal Variability of the Indian summer monsoon, J. Meteor.Soc.Japan, 85(3):349- 474 358. 475 476 Roxy M, Tanimoto Y, Preethi B, Pascal T and Krishnan R (2013) Intraseasonal SST- 477 precipitation relationship and its spatial variability over the tropical summer monsoon region. 478 Clim Dyn 41(1):45–61. 479 480 Roxy M (2014) Sensitivity of precipitation to sea surface temperature over the tropical 481 summer monsoon region—and its quantification. Clim Dyn. 43:1159–1169. 482 483 Satheesan K and Krishnamurthy BV (2005) Modulation of Tropical tropopause by wave 484 disturbances. J. Atmos. Sol-Terr. Phy., 67:878-883. 485 486 Shankar D, Shetye SR and Joseph PV (2007) Link between convection and meridional 487 gradient of Sea Surface Temperature in the Bay of Bengal. J. Earth Syst. Sci. 116 (5):385- 488 406. 489 22 490 Shenoi SSC, Shankar D and Shetye SR (2002) Differences in heat budgets of the near- 491 surface Arabian Sea and Bay of Bengal: Implications for the summer monsoon. J. Geophys. 492 Res. 107 doi:10.1029/2000JC000679. 493 494 Sherwood SC (1999a). Convective precursors and predictability in the tropical western 495 Pacific, Mon. Weather Rev., 127:2977–2991. 496 497 Sherwood SC, Roca R., Weckwerth TM and Andronova NG (2010) Tropospheric Water 498 Vapor,Convection And Climate.Rev. Geophys., 48(2), doi:10.1029/2009RG00301. 499 500 Sifuzzaman M, Islam SR and Ali MZ (2009) Applications of Wavelet Transform and its 501 advantages compared to Fourier Transform. J. Phy. Sci. 13:121-134. 502 503 504 Sud YC, Walker GK and Lau KM (1999) Mechanisms regulating sea surface temperatures 505 and deep convection in the tropics. Geophys. Res. Lett., 26:1019–1022. 506 507 Tomasi C (1984) Vertical distribution features of atmospheric water vapor in the 508 Mediterranean, Red Sea and Indian Ocean. J. Geophys. Res., 89 (D2):2536 – 2566. 509 510 Torrence C and Compo GP (1998) A Practical Guide to Wavelet Analysis. Bull. Amer. 511 Meteor. Soc.,79:61–78.4 512 513 Vinayachandran PN, and Shetye SR (1991) The warm pool in the Indian Ocean, Proc. Indain 514 Acad Sci (Earth Planet Sci.),100 (2):165-175. 515 516 Vinayachandran PN, Neema CP, Simi Mathew and Remya R (2012) Mechanisms of summer 517 intraseasonal sea surface temperature oscillations in the Bay of Bengal. J. Geophys. Res.: 518 Oceans (1978–2012) 117(C1). 23 519 520 Waliser DE and Graham NE (1993) Convective cloud systems and warm pool seas surface 521 temperatures : Coupled interactions and self regulation. J. Geophys. Res., 98 (12):12,881- 522 12,893 523 524 Wang B and Wang Y (1996) Temporal Structure of the Southern Oscillation as Revealed by 525 Waveform and Wavelet Analysis. J. Clim., 9:1586–1598. 526 527 Weckwerth TM, Pettet CR, Fabry F, Park S, LeMone MA and Wilson JW (2005) Radar 528 refractivity retrieval: Validation and application to shortβterm forecasting, J. Appl. Meteorol., 529 44:285–300. 530 531 Zelinka MD and Hartman DL (2009) Response of humidity and clouds to tropical 532 convection. J.Clim., 22:2389-2404. 533 534 Zhang C (1993) Large-scale variability of atmospheric deep convection with respect to sea 535 surface temperature in the tropics. J. Clim.,6:1898 – 1912. 536 537 Zhang GJ, Ramanathan V, McPhaden, MJ (1995) Convection-Evaporation Feedback in the 538 Equatorial pacific. J. Clim, 8:3040–3051. 539 540 Uma R., Lakshmi Kumar TV, Narayanan MS, Jyothi Bhate, Rajeevan M and Niranjan 541 Kumar K (2013) Large scale features and assessment of spatial scale correspondence 542 between TMPA and IMD rainfall data sets over Indian landmass. J. Earth Syst. Sci.122(3): 543 573-588. 24