Lightning activity in the dry environment of northwest India and Pakistan (NW) and in the moist environment of northeast India (NE) has been examined from the Optical Transient Detector and Lightning Imaging Sensor data obtained from the Tropical Rainfall Measuring Mission satellite during 1995–2010. In the NW region, seasonal variation of flash rate is annual with a maximum in July but is semi-annual with a primary maximum in April and a secondary maximum in September, in the NE region. On diurnal scale, flash rate is the maximum in the afternoons, in both the NE and NW regions. The correlation of flash rate with convective parameters, viz. surface temperature, convective available potential energy (CAPE) and outgoing long-wave radiation is better with convective activity in the NW than in the NE region. Mean value of aerosol optical depth at 550 nm is ~ 26% higher and is highly correlated with flash rate in NW as compared to that in NE. Results indicate that CAPE is ~ 120 times more efficient in NW than in the NE region for production of lightning. The empirical orthogonal function analysis of flash rate, surface temperature, and CAPE shows that variance of lightning activity in these regions cannot be fully explained by the variance in the surface temperature and CAPE alone, and that some other factors, such as orographic lifting, precipitation, topography, etc., may also contribute to this variance in these mountainous regions. Further, the increase in CAPE due to orographic lifting in the Himalayan foothills in the NE region may contribute to ~ 7.5% increase in lightning activity. Relative roles of the thermally induced and moisture-induced changes in CAPE are examined in these regions. This study merely raises the questions, and that additional research is required for explaining the fundamental reasons for the reported observations here.
 Thunderstorms, the deepest convective clouds in the Earth's atmosphere, are formed by buoyancy forces which are initially set up when solar radiation heats the Earth's surface and the air in the adjacent atmospheric boundary layer. Buoyancy is controlled primarily by surface air temperature. Convection is deeper and more frequent in the tropics than at higher latitudes [Williams, 1992]. Formation of thunderstorm is based on thermodynamic conditional instability of this boundary layer. The energy that feeds this process is the convective available potential energy (CAPE). The initial finite vertical displacement for conditional instability can be caused by several mechanisms such as boundary layer thermals, orographic lifting, frontal surfaces, etc. Irrespective of the nature of initiator, the buoyant air parcels rise and cause the condensation of water vapor. If the ascent of air extends to the altitudes above freezing level, all three phases of water coexist. The conditions in this mixed-phase lead to the formation of ice crystals and graupels, which are generally believed to be fundamental ingredients for strong electrification and lightning in thunderstorms [Latham, 1981; Illingworth, 1985; Krehbiel, 1986; Williams, 1989, 2001; Saunders, 1995]. The electrical activity in thunderstorms results because of a strong and complex interaction among CAPE, vertical air motion, microphysics, and vertical extent of ice particles. Size and vertical distribution of particles that contribute to charge separation are influenced by both the magnitude and vertical distribution of CAPE [Williams, 1995]. Results of laboratory simulation experiments [Takahashi, 1978; Saunders et al., 1991] showing the charge transfers in the particle collisions and the behavior of lightning discharges seem to be consistent with the large-scale observations of dynamics and microphysics of clouds. For example, lightning activity is closely linked with the appearance of ice phase and depth of the cloud [e.g., Williams, 2001, and references therein].
 Thunderstorm depth, generally measured as radar cloud top height is closely related with the lightning flash rate [Vonnegut, 1963; Ushio et al., 2001; Boccippio, 2001; Williams, 1985]. In the tropics, lightning responds sensitively to surface temperature at many sites [Williams, 1992], and flash rate tends to be maximum at the time of peak cloud height [Frost, 1954; Williams, 1991]. Recent study of Penki and Kamra [2012a] shows that lightning flash rate is highly correlated with surface temperature in the Himalayan foothills of the northwestern India. This association in specific locations and meteorological regimes is attributed to the gravitational energy of large ice particles in the upper regions of the storm. Such accumulations of ice particles have also been reported to be positively correlated with CAPE and the updraft speed [Williams et al., 1992; Rutledge et al., 1992; Zipser, 1994; Petersen et al., 1996]. Our results of the empirical orthogonal function (EOF) analysis show that 22% changes in flash rate in the Himalayan foothills can be associated with variability in CAPE [Penki and Kamra, 2012a].
 Wide ranges of meteorological and environmental conditions and their interactions with topography and landforms lead to the formation of clouds with great diversity in their microphysical and dynamical characteristics. Therefore, the thunderclouds developing in different seasons exhibit a large variety of lightning activity in different regions around the globe. Localized case studies and field campaigns have suggested diagnostic and sometimes even predictive relationships of the lightning activity with updraft development [Goodman et al., 1988], rainfall rate [Petersen and Rutledge, 1998; Tapia et al., 1998], cloud top height [Williams, 1985; Cherna and Stansbury, 1986; Price and Rind, 1992], and mesocyclone occurrence [MacGorman et al., 1989; Williams et al., 1999]. A major complication in such studies is that many relationships are not robust or unique, particularly when applied to convection in different regions. Even in the comparatively “simple” convective environment of the tropics, the globally invariant relationships between lightning and other observed parameters are rare. As examples of the lack of regional invariance in such relationships, Petersen and Rutledge  found that the lightning-local rain yield relationships for various regions around the globe, though of strong diagnostic value, were in some manner dependent on the local “convective regime,” presumably a mix of the frequency, vigor, depth, and level of organization of local storms.
 Because of the dominating importance of the global warming issue in the climate research, there have been numerous studies dealing with the relationship between lightning activity and surface temperature. For example, for every 1° of surface warming, lightning increases anywhere between 10 and 100% [Price, 2008]. Several other studies also show a high positive correlation between surface temperature and lightning activity [Williams et al., 2005; Price and Asfur, 2006a; Sekiguchi et al., 2006; Penki and Kamra, 2012a, 2012b]. Large number of relationships between temperature and lightning activity have been explored on various time scales at different places [Price, 1993; Markson and Price, 1999; Markson, 2003; Anyamba et al., 2000; Williams et al., 1999; Satori and Zieger, 1996; Füllekrug and Fraser-Smith, 1998; Nickolaenko et al., 1998; Manohar et al., 1999; Nickolaenko et al., 1999; Christian et al., 2003; Williams, 1994; Nickolaenko and Rabinowicz, 1995; Heckman et al., 1998; Williams et al., 2000]. Changes in global monthly lightning activity on land are well correlated with changes in global monthly land wet-bulb temperatures except in the tropics [Reeve and Toumi, 1999]. Africa has high lightning activity with the same cold cloud tops because of larger Bowen ratio in this region [Williams and Satori, 2004], a stronger diurnal cycle in temperature, and more powerful but less frequent deep convection [Williams, 2004]. In comparison, maritime continent shows less lightning activity [Christian et al., 2003] for the same area of high level cirrus [Kent et al., 1995; Wang et al., 2003]. This is because maritime continent has very frequent, high top convection with modest updraft strength because the wet surface suppresses the low level instability [Williams and Satori, 2004].
 In the tropics, cloud base height is dominated by the dry-bulb temperature over the wet-bulb temperature as the lightning-regulating temperature in regions characterized by moist convection [Williams et al., 2005]. In the extratropics, an elevated cloud base height may enable larger cloud water concentrations in the mixed-phase region, a favorable condition for the positive charging of large ice particles that may result in thunderclouds with a reversed polarity of the main cloud dipole [Williams et al., 2005].
 In this paper, we study the diurnal and seasonal variations of flash rate in two Himalayan regions in the northeast and northwest of India, marked as NE and NW in Figure 1 from the Optical Transient Detector (OTD) and the Lightning Imaging Sensor (LIS) data for the 1995–2010 period. These regions experience extreme convective activity in the dry and moist environmental conditions and have the highest density of lightning flashes in South Asia in some part of the year. We examine the correlations of flash rate with convective parameters and aerosol optical depth (AOD) and apply the empirical orthogonal function (EOF) analysis on the monthly averaged values of flash rate, atmospheric surface temperature, and convective available potential energy (CAPE) in these regions. The main objective of this paper is to examine the regional differences in lightning activity occurring in thermodynamic conditions of contrasting environments of NE and NW. However, we have also discussed our results on the basis of the earlier investigations of the differences in convective and microphysical characteristics and the height of radar echo and vertical structures of the radar echo profiles in these regions.
2 Data Set
 The lightning data products, used here, are obtained from the Optical Transient Detector (OTD) launched in April 1995 and the Lightning Imaging Sensor (LIS) launched in December 1997 aboard the Tropical Rainfall Measuring Mission (TRMM) satellite. Both sensors measure total [Intra-Cloud (IC) and Cloud-to-ground (CG)] lightning with 90% detection efficiency during both day and night, with little regional bias. The following data sets from the website http://ghrc.msfc.nasa.gov are used in this study.
 LIS/OTD 0.5 Degree High-Resolution Full Climatology (Annual Flash rate) of Version 2.3. The product is a 0.5° × 0.5° gridded composite of total (IC + CG) lightning bulk production expressed as a flash rate (fl km−2 yr−1).
 LIS/OTD 2.5 Degree Low-Resolution Diurnal Climatology of version 2.3. The product is a 2.5° × 2.5° gridded composite of climatological total (IC + CG) lightning bulk production as a function of local hour, expressed as a flash rate(fl km−2 h−1).
 LIS/OTD 0.5 Degree High-Resolution Monthly Climatology of version 2.3. The product is a 0.5° × 0.5° gridded composite of total (IC + CG) lightning bulk production as a function of month of year, expressed as a flash rate (fl km−2 day−1). This product is a 30 day average, centered on the 15th of each month.
 LIS/OTD 2.5 Degree Low-Resolution Monthly Time Series. The product is a 2.5° × 2.5° gridded composite of monthly time series of total (IC + CG) lightning bulk production, expressed as a flash rate density (fl km−2 day−1).
 In all the above four data products, climatologies from the 5 year OTD (April 1995 to March 2000) and 8 year LIS (January 1998 to December 2010) missions as well as a combined OTD + LIS climatology and supporting base data (flash counts and viewing times) are included. The data sets used in this analysis are corrected for the TRMM orbit altitude change that occurred in August 2001.
 Lightning flash radiance from the TRMM satellite from the website http://ghrc.msfc.nasa.gov for the period 1998–2005. The flash optical radiance observed by LIS is a measure of lightning discharge energy or the discharge energy or the discharge current [Koshak et al., 2000]. Opacity of the cloud/atmosphere, however, adds some uncertainty to such an inference.
 The parameters, surface temperature, CAPE, outgoing long-wave radiation (OLR), with 0.5/2.5° grid resolution, and SST OIv2 data from the Climate Forecast System Reanalysis data developed by NOAA's National Center for Environmental Prediction (NCEP) (http://nomads.ncep.noaa.gov). The model-generated monthly mean values of the CAPE using assimilated data are used. Hence, the CAPE is instantaneous because it uses instantaneous temperature and humidity profiles at any given forecast time. In this model, the parcel to lift is selected by locating the parcel with warmest equivalent potential temperature in the lowest 70 mb above the model surface and lifting it from its lifting condensation level to the equilibrium level [Zhang and McFarlane, 1991]. During the lifting process, positive area in each layer is summed up as CAPE and negative area as the convective inhibition energy. No attempt to study diurnally resolved CAPE is made, in this study.
 In this investigation, we have selected two regions, denoted by NW and NE in Figure 1, of extreme dry/wet convection over the western (31°N–36°N, 70°E–77°E, mean elevation of 2918 m) and eastern (25°N–27°N, 86°E–90°E, mean elevation of 762 m) sides of the Himalayan range. In the NW region, relatively barren land of the Thar desert is located and the Indus plains rise sharply to the Himalayas. On the other hand, the Ganges delta followed by the moist vegetated land surface characterized by irrigated crops and wooded wetlands at low elevations and forest over the higher terrain in the NE region rise gently to the Himalayas.
4 Convective Activity in the NE and NW Regions
 The location of occurrence of the most extreme convection is closely related to the topography of the region, and the form taken by extreme monsoon convective systems differs between the dry northwestern subcontinent and the moist eastern Ganges region [Houze et al., 2007; Romatschke et al., 2010]. The TRMM data analysis of these authors shows that the western indentation of the Himalayas has a maximum frequency of convective systems containing intense convection echo (40 dbz radar reflectivity > 10 km in height) that occurs over the lower elevations of the Himalayan barrier. Since the environmental shear is relatively weak, these deep intensive echoes are vertically erect, sometimes extending above 17 km indicating that exceptionally strong updrafts loft graupel to high altitudes. Also occurring in these regions are the wide intense convective echoes (40 dbz echo > 1000 km2 in the horizontal dimension) exhibiting mesoscale structures. On the other hand, the eastern partitions of the Himalayan region have maximum frequencies of the broad stratiform echoes (> 50,000 km2 in area) that occur in connection with the Bay of Bengal's depressions. These depressions provide the moist maritime environment for convection to develop layer stratiform echoes associated with the mesoscale systems. The moist air flows advancing from the Arabian sea and the Bay of Bengal in the NW and NE regions, respectively, encounter dramatically different land surfaces, and hence, the surface heat fluxes are different in the two regions [Niyogi et al., 2002]. While the surface latent heat flux has a local maximum over the irrigated cropland, wetlands, and forests, it has a local minimum near the Thar desert where the surface sensible heat flux dominates.
Medina et al.  showed that the low moisture content observed in Afghanistan and West Pakistan is part of a large-scale feature that extends throughout North Africa and the Middle East. Therefore, strong moisture gradients appear near the India-Pakistan border that separate moist monsoonal air over the Indian Peninsula from the dry air that prevails over North Africa and the Middle East.
5 Seasonal and Diurnal Variations of Lightning Activity in the NW and NE Regions
 Figure 2 shows the monthly variations of average flash rate for the period of 1995–2010 in the NW and NE regions. In the NW region, seasonal variation in flash rate is annual. The flash rate starts increasing from January and reaches a maximum of 0.13 fl km−2 day−1 in the monsoon month of June. Flash rate remains high throughout the monsoon season (June to September) and is low from October to February. On the other hand, in the NE region, seasonal variation in flash rate is biannual. Low values of flash rate in the winter sharply rise and reach a maximum of 0.13 fl km−2 day−1 in the premonsoon months of April–May itself. It also shows a secondary maximum of flash rate equal to 0.04 fl km−2 day−1 in September. The result that the average flash rate is maximum in the premonsoon season in the NE region but in the monsoon season in the NW region is important and will be discussed further in section 8.
 In diurnal variability of flash rate in both the NW and NE regions, the low values of flash rate during morning hours, increase by noon and remain high throughout the afternoon and night (Figure 3). However, the flash rate in general, is higher in the NW than in the NE region.
6 Response of Lightning Activity to the Convective Activity
 In the tropics, differences in meteorological regimes are very influential in the distribution of buoyancy and radar reflectivity in different seasons, and, therefore, the relationships between radar cloud top height and lightning flash rate are not unique in different meteorological regimes [Rutledge et al., 1992; Williams et al., 1992; Zipser, 1994; Watson et al., 1995; Petersen et al., 1996; Williams et al., 2002]. These studies demonstrate that the peak flash rates in the thunderstorms embedded in large-scale monsoon airflow are an order of magnitude less than in more vigorous storms in the more unstable “break period” regime. Such comparative studies emphasize the need for caution in using single parameter relationships to predict lightning activity. The relationship between the cloud top height and lightning activity at a location in a particular meteorological regime is generally attributed to the gravitational energy of the large ice particles accumulated in upper portion of the cloud. Such ice particle accumulations are positively correlated with CAPE and the updraft speed [Williams et al., 1992; Rutledge et al., 1992; Zipser, 1994; Petersen et al., 1996], and this correlation possibly relates the electrical activity of the cloud to the environmental thermodynamic conditions in which the storm develops. However, it must be emphasized that CAPE is not the only factor in controlling the updraft and microphysics of convective cores. Williams et al.  point out that land-ocean updraft contrast is inconsistent with the land-ocean contrast in CAPE. For example, CAPE over warm tropical oceans is of comparable magnitude [Williams and Renno, 1993; Lucas et al., 1996], yet the lightning activity differs by more than an order of magnitude [Williams and Stanfill, 2002; Christian et al., 2003; Penki and Kamra, 2012b]. Many other factors such as cloud base height, Bowen ratio, storm morphology, and aerosol loading combine in determining updraft intensity and lightning production [Rosenfeld and Lensky, 1998; Williams and Stanfill, 2002; Williams, 2002; Williams et al., 2005; Zipser, 2003]. Linking the response of all these factors or a combination of them to lightning activity is not within the scope of a single paper and requires many more studies in this direction. The main aim of this paper is to examine the regional difference in lightning activity in contrasting environments (dry and moist) of NW and NE. However, in section 9, we discuss our results in view of the earlier studies for understanding the behavior of convective activity to some other factors in South Asia, especially in the same Himalayan regions.
 We have examined the correlations of monthly flash rates with three indicators of convective activity, namely average surface temperature, convective available potential energy (CAPE,) and outgoing long wave radiation (OLR) in both the NE and NW regions. OLR is inversely related to the cloudiness and may significantly influence both the incoming and outgoing radiations and thus affect the surface temperature and the vertical thermal structure and convective activity of the lower troposphere. Since cloudiness undergoes a remarkable change, especially during the onset and withdrawal phases of the monsoon in our areas of investigation and changes in OLR and convection feedback each other, influence of OLR is being investigated to better understand the convection-lightning relationship on large spatial scale. Figure 4 shows that monthly variations of these three parameters are drastically different in the NE and NW regions. In general, average values of surface temperatures and CAPE are higher in the NE than in the NW region. However, in the NW region, while the average surface temperature is maximum in June, CAPE reaches its maximum value only in August. In the NE region, while the average surface temperature is maximum in May, CAPE reaches its maximum value only in September with a secondary peak in June. In both regions, OLR dips down to low values during the monsoon months of July–August. However, while it is maximum in the pre- and postmonsoon months of May and October in the NW, it remains around maximum values throughout the November–April period in the NE.
 Figure 4 also shows the correlations of the average flash rate with the average values of surface temperature, CAPE, and OLR. While the correlations of the three parameters with flash rates are comparatively good in the NW region than in the NE region, thereby indicating that flash rate is better correlated with convective activity in the NW than in the NE region. Also plotted in Figure 4 are the monthly variations of AOD as derived from MODIS data at 550 nm and its correlations with flash rate in the NW and the NE. The AOD data will be discussed in section 7.
 Table 1 shows the mean values of altitude, surface temperature, flash rate, radiance, CAPE, and flash per CAPE for the period 1995–2010 for both the NE and NW regions. Mean values of altitude and surface temperature of the NW region are about 2.5 and 0.75 times, respectively, of those for the NE region. However, mean values of flash rate and radiance are approximately equal, indicating that flashes are equally energetic in both regions. Noticeable feature of Table 1 is that in spite of roughly one-sixth value of CAPE, the flash per CAPE is ~ 19 times of that in the NE region. This result indicates that CAPE is ~ 120 times more efficient in the NW than in the NE region in producing lightning.
Table 1. Mean Values of Altitude, Surface Temperature, Flash Rate, Outgoing Longwave Radiation, Radiance, CAPE and Flash per CAPE in the NW and NE Regions
Category and Region Latitudes, Longitudes
Mean Altitude m
Mean Surface Temperature 0C
Mean Flash Rate fl km-2yr-1
Mean OLR W m -2
***Mean Radiance Kj m-2 sr-1
***Flash Per CAPE
NW 700E -770E 320N-360N
NE 860E -900E 230N-270N
7 Response of the Lightning Activity to Aerosol Loading
 Aerosols act as another mediator of cloud microphysics, precipitation, electrification, and lightning [Rosenfeld and Lensky, 1998; Rosenfeld and Woodley, 2003]. Increase in aerosol concentration reduces mean droplet size, a suppression of warm rain coalescence, and an enhancement of the cloud water reaching the mixed-phase region [Williams et al., 2002]. Enhancement in lightning activity as predicted by the aerosol hypothesis was found to be doubtful from the observations of aerosol and electrical parameters in Amozene Basin in Brazil in both aerosol-rich and aerosol-poor conditions [Williams et al., 2002]. Aerosol can also invigorate the convection [Andreae et al., 2004].
 Figure 5 shows the distributions of the mean AOD at 550 nm from MODIS data for the 2002–2010 period in the NW and NE regions. Comparatively much higher values of AOD in the NW region, especially over the northwestern part of Pakistan, are noticeable. Mean values of AOD are ~ 26% higher over the NW than the NE for this period (Table 1).
 Also plotted in Figure 4 are the curves for seasonal variations of the mean AOD in the NW and NE regions. In the NW region, the mean AOD is maximum in June with a secondary maximum in November. AOD in the NW has high correlation with the flash rate (Figure 4), which is also maximum in June. In the NE region also, AOD is maximum in June with a secondary maximum in January, but its correlation with the flash rate is much lower as compared to that in the NW region.
8 The EOF Analysis of Flash Rate, Surface Temperature, and CAPE
 We have applied the empirical orthogonal function (EOF) analysis to the monthly averaged data of the flash rate, surface temperature, and CAPE for the period of 1998–2005 to examine the behavior of lightning distribution from the storms developing in the dry and moist environment conditions of the NW and NE regions, respectively. Figure 6 shows the first two EOF modes and the corresponding time series of the flash rate in the NW and NE regions. In the spatial distributions of the first mode (EOF1) in the NW region, its presence in March over the northwest India and Pakistan is shown. The time series of the first mode shows that it peaks during March–April and declines to its minimum in October–November. Thus, the combination of these two data sets implies that the first EOF mode is characterized by high flash rates over the northwest India and Pakistan in March–April, and it accounts for ~ 92% of variance. In the NE region, the first mode of the flash rate appears ~ 200 km north of the mouth of the Bay of Bengal. Time series of the first mode in the NE region shows that it also peaks in March. The high flash rate variations in this area account for ~ 90% of variance. It can be concluded from Figure 6 that the first EOF modes of flash rate in the NW and NE regions are in phase with each other. The second EOF modes (EOF2), accounting for 6–7% variability in flash rate in these regions, are almost in phase in time but appear nearly perpendicular in their orientations, i.e., in the north-south direction in the NW region and in the east-west direction in the NE region.
 Figure 7 shows the first two EOF modes and the corresponding time series of surface temperature in the NW and NE regions. In the NW region, the first EOF mode extends over a wide area from West Pakistan to northwest India and occurs in the monsoon-onset months of June–July. This mode accounts for ~ 91% variance in surface temperature in this region. The second component in EOF mode in surface temperature is not in phase, both spatially and temporally with the first component, and accounts for the 7% variance in surface temperature. In the NE region, the first EOF mode of surface temperature appears a few hundred kilometers north of the mouth of Bengal and extends over a wide area from central India to the Burmese mountain range with a tongue extending to the Bay of Bengal. The first mode in the NE region peaks in the monsoon-onset months of June–July and accounts for 81% variance in the surface temperature. The second component in the EOF mode (EOF2) in surface temperature appears at almost the same place and time as the first mode and accounts for 15% variance in the surface temperature. But contrary to the case of flash rate, the EOF2 in the surface temperature is oriented in the north-south direction in both the NW and NE regions.
 Figure 8 shows the first two EOF modes and the corresponding time series of the CAPE in the NW and NE regions. The first EOF mode of CAPE, though roughly anchored at the same place as the first modes of flash rate and surface temperature, is temporally, as in the case of surface temperature, out of phase with the first mode in flash rate in the NW region. In the NE region, although the first mode in the CAPE is roughly anchored at the same place, the first modes in CAPE and surface temperature are spatially oriented opposite to each other. The second mode in CAPE in the NW region is roughly anchored at the same place as its first mode. The second mode in CAPE in the NE region shows the most interesting feature. Axis of the orientation of the maximum variation in this mode, which accounts for 7.5% of the variance is along a line which roughly coincides with the Himalayan foothills. This result shows the role of the Himalayas in increasing the CAPE by orographic lifting and thus the lightning activity in this region.
 The results presented in Figures 6-8 bring out three important features of the lightning activity in the NW and NE regions. First, in the NW region, the first EOF modes in the flash rate, surface temperature, and CAPE show very good coexistence in their spatial distributions. However, the maximum variance in surface temperature and CAPE lags by 2–3 months from the maximum variance in flash rate. Second, in the NE region, although the first modes of both the surface temperature and CAPE appear 100–200 km north of the mouth of the Bay of Bengal, their orientations are opposite in direction to each other, i.e., north to south in case of atmospheric temperature and south to north in case of CAPE. Third, the second EOF mode of CAPE in the NE region provides a good evidence of the role of the Himalayas in increasing the CAPE and the flash rate by orographic lifting in the foothills of the Himalayas.
 Extremely high values of average flash rates in the NW region during the monsoon season and in the NE region during premonsoon season call for a comparative study of the type of clouds and environmental conditions of the two regions. The seasonal and diurnal variations of flash rate in the NE region (Figures 2 and 3) are in conformity with the severe thunderstorm occurrence in this area, which are accompanied with the intense lightning activity. The eastern and northeastern parts of India, including the NE region, get affected by severe thunderstorms (locally called as Nor'wester's or Kalabaisakhi's) during premonsoon months, particularly during the period of April and May, when about 28 thunderstorms occur during this period in this region. Intense convection is initiated in this area due to strong heating of land mass during mid-day. The convection moves southeast and gets intensified by mixing with warm moist air mass from the Bay of Bengal. The upper airflow over this region has a shallow layer of southerlies/southeasterlies from the Bay of Bengal near the ground and dry westerlies of continental origin above. The layer of transition between these two airstreams is stable in which moisture decreases very rapidly with height. At times, it develops an inversion in temperature. Although, low-level inversion inhibits growth of convective clouds, it prevents the penetration of convection into the layers above. The moist layer increases in moisture and warmth, thus increasing CAPE which is a favorable condition for a severe nor'wester outbreak. Superposition of these lower tropospheric conditions with areas of upper tropospheric positive vorticity advection in association with troughs in westerlies provides the large-scale vertical velocity situation favorable for triggering thunderstorms activity [Koteswaram and Srinivasan, 1958; Krishna Rao, 1966; Srinivasan et al., 1973].These severe thunderstorms are often arranged in long lines, 150 to 250 km in length, in which adjacent clouds are very close to each other. These squall lines are frequent in North and Northeast India and are accompanied with strong winds, lightning, thunder, and hail. About 72% of tornadoes in India have occurred in northeast India and Bangladesh, and about 76% of the tornadoes in India occurred during March to May, the most favored period being the afternoon and evening hours during the month of April [Bhattacharya and Banerjee, 1980; Mandal and Saha, 1983; Asnani, 1985].
 In the NW region, a semipermanent system of heat low appears at the beginning of monsoon season. The Himalayas provide an orographic flow barrier, and the configuration of mountain ranges produces the frictional convergence in the region. These conditions coupled with relatively high surface temperatures over northwest India and central Pakistan produce a weak ascent in the atmosphere. Availability of sufficient precipitable water content over West India and adjoining Pakistan provides favorable conditions for the genesis of thunderstorms. Very high values of flash rate in this region observed throughout the whole monsoon season in Figure 2 are associated with such thunderstorm activity. Moreover, the pattern of high flash rates in May is pushed toward the Hindu Kush mountains in the NW region during the monsoon season.
 The thunderclouds developing in the NW region are dry with low moisture contents and produce very little rainfall. They develop in environment of low relative humidity (< 25%) and have high cloud bases at 3 to 4 km altitude. The rain falling from these storms often evaporates before reaching the ground, because of their high bases and low relative humidity of the air below. Because of these reasons, these thunderstorms often generate dust storms. However, they grow to large heights extending up to tropopause and consequently have large depths. On the other hand, the thunderstorms occurring over the NE region produce heavy rain showers, hail, severe squalls, and at times tornadoes.
 The environment in which thunderstorms grow perhaps provides the most important single distinction to their severity. During the spring and summer, the South Asian region exhibits a range of extremes weather systems ranging from deep convective towers to Mesocale convective Systems (MCSs) with extremely large stratiform regions. Deep convective cores change their preferred location from the east coast in the premonsoon to the western foothills in the Himalayan foothills in the monsoon season [Webster et al., 1998; Romatschke et al., 2010]. Zipser et al.  conclude that some of the deepest intense convection in the world occurs near the mountains of arid northwestern part of the subcontinent. This deep intense convection can occur either as isolated cells or may be embedded in large MCSs [Hauze et al., 2007]. Timing and spatial distributions of convective echo structures analyzed by Romatschke et al.  suggest that occurrence of most lightning in the NE and NW regions can be attributed to the occurrence of deep convective echoes formed by the buoyancy produced due to sensible heat flux from heated land surfaces in these regions. In both seasons, such cores develop in regions of strong surface specific humidity gradients where near-surface airflow is capped by dry air aloft. Triggering of the intense convection in these regions occurs when the low-level flow is orographically lifted to a level where it can break through the capping stable layer. Both spatial and temporal variations in the extreme convective events in their study suggest differences in convective activity in rainy and dry regions. Timing and spatial distribution of AOD (Figures 4 and 5) in our analysis indicate a strong influence of aerosol loading on the lightning activity in the dry environment of the NW region. High value of correlation coefficient between flash rate and AOD (Figure 4) supports such a relationship in NW.
 Our results in Table 1 show that the mean flash rate and mean radiance in the NW and NE regions are not much different from each other indicating that flash rate in both the regions is approximately equally energetic within some reservations as pointed out by Koshak [2010, 2011]. It is worth noting that while the CAPE is about five to six times greater, the flashes per CAPE are about 19 times less in the NE than in the NW region. Thus, the occurrence of lightning in the storms that develop in the NW region is more sensitive to CAPE changes and can more readily generate lightning compared to the storms in the NE region. This may be associated with several factors such as the higher cloud bases, lower moisture contents, lower rainfall, steeper orographic lifting, comparatively lower wind shear in the upper troposphere in thunderstorms, and higher aerosol concentrations that occur in the NW region. For example, higher cloud bases in the NW region will provide larger depth of the mixed-phase region where larger availability of liquid and solid phase particles may participate in the cloud electrification processes [Williams et al., 2002, 2005]. Also, steeper orography may contribute to more vigorous updrafts and higher cloud tops of clouds, which have been frequently reported to be nonlinearly related to the increase of flash rate in thunderclouds [Williams, 2001 and many references given therein]. Further, lower wind shear in the upper troposphere, a distinction dictated by the environment of the region, helps the updraft to grow nearly vertical and develop thunderclouds which stand erect and grow to higher altitudes. Higher flash rate in such thunderclouds may also result because of higher probability for intracloud flashes as the altitude of charge separation process and region of high electric field rises to regions of weaker dielectric strength in the atmosphere [Rutledge et al., 1992]. Another plausible reason, as discussed in section 7 for higher efficiency of lightning production in the NW region, may be higher concentration of aerosols in this arid environment.
 Our analysis shows that lightning activity in the NW and NE regions rapidly increases as the surface temperature increases in the premonsoon season from winter to summer. It also shows how the progress of the monsoon current arrests the increase in lightning activity in these regions. The Bay of Bengal branch of monsoon current progresses from the mouth of the Bay of Bengal in the northeast to the northwest of India through the Himalayan foothills and plains of central India. The monsoon arrival in the NE region in May–June arrests the growth of surface temperature and lightning activity because of the decrease in shortwave radiation reaching the ground on account of the increasing cloudiness in this season. It is worth noting that CAPE reaches its maximum only in September (Figure 4), just before the monsoon withdrawal when flash rate again shows a secondary maximum in this region (Figure 2). In the NW region, however, the surface temperature and flash rate keep increasing up to June–July when the Bay of Bengal branch of the monsoon current reaches there and arrests their growth. It is significant to note that CAPE is also maximum during July–August in this region. The above results indicate that on seasonal scale, the changes in CAPE induced by the changes in the vertical thermal structure of the atmosphere are necessary but not sufficient conditions for changing the lightning activity in these regions. Changes in moisture content of the boundary layer are essentially required for variance in lightning activity in a region. Such changes in moisture content are provided by the arrival of the Bay of Bengal component of monsoon current in the month of May–June in the NE region and in the month of June–July in the NW region.
 Seasonal variation of flash rate is annual in the dry convection of the NW region but is semi-annual in the moist convection of the NE region. On diurnal scale, flash rate is the maximum in the afternoons in both the NW and NE regions. Flash rate is better correlated with the surface temperature, CAPE, OLR, and AOD in the NW than the NE region. For the production of lightning, the CAPE is ~120 times more efficient in the NW than in the NE region. The EOF analysis shows that variance of flash rate in the NW and NE regions cannot be fully explained by the variance in surface temperature and CAPE and that some other additional factors such as orographic lifting, rainfall, topography, etc., may contribute to the variance of lifting in these mountaneous regions. In the Himalayan foothills in the NE region, variance in CAPE due to orographic lifting may contribute to 7.5% of variance in lightning activity. Further, in addition to the changes in CAPE due to change in the vertical thermal structure of the atmosphere, availability of moisture content in the boundary layer is essentially required in a region. This moisture is supplied on the arrival of the Bay of Bengal branch as monsoon current in May–June in the NE region and, a month later, in June–July in the NW region.
 P.R.K. is thankful to the Director, Indian Institute of Tropical Meteorology, Pune for a research fellowship and necessary facilities to do this work. A.K.K. acknowledges the support under the INSA Senior Scientist program. We acknowledge the NASA for making the data available on the global lightning data on the websites http://ghrc.msfc.nasa.gov, the aerosol optical depth data on the website http://gdata.1.sci.gsfc.nasa.govt, NOAA, and ECMWF for making the data of different parameters available through their websiteshttp://nomads.ncep.noaa.gov and http://data-portal.ecmwf.int/data.