Many meteorological phenomena display distinctive diurnal patterns that vary spatially and throughout the annual cycle; the patterns are particularly interesting for phenomena related to convective activity. While dozens of studies have been conducted on diurnal patterns in convective activity for many areas of the Earth, including studies on precipitation amount, frequency, hail, lightning, and tornadoes, we focus in this investigation on lightning activity across a large area in tropical Asia. Others (Haldar et al., 1991; Jou, 1994; Basu, 2007) have examined diurnal patterns in the amount and frequency of precipitation events in the region and found that strong diurnal cycles in the data were largely determined by the interaction of local topography such as distance from the coast and highlands areas.
More specifically in the study area, a strong diurnal cycle associated with convective processes leading to late afternoon to early evening precipitation events was found in the interiors of the Indian subcontinent (Pathan, 1994; Basu, 2007; Sen Roy and Balling, 2007). The time of maximum precipitation events also coincided with the time of maximum surface temperatures. Additionally, along the west coast of India, the time of maxima occurred predominantly between midnight and the early morning hours as a result of moist air striking the Western Ghats and rising abruptly from the coast up to a height of 1–2 km (Sen Roy and Balling, 2007). The role of local topography in the form of ‘orographic land–atmosphere interaction’ was revealed in a study by Barros et al. (2004) on the south-facing slopes of the Himalayas where precipitation tended to occur in the nocturnal hours. A similar role of topography was also found in Taiwan, where the initiation of thunderstorms occurred on the mountain peaks, which moved downslope leading to heavy rain in the low lying areas during the afternoon hours (Jou, 1994). Zhou et al. (2008) revealed a late-afternoon maximum over southeastern and northeastern China, while a near-midnight maximum was observed over the eastern section of the Tibetan Plateau due to intense convection and/or large mesoscale convective systems. The diurnal phases of amount and frequency were similar in most of China, except the Yangtse River valley, where two peaks were observed including one in the morning and the other later in the afternoon. Similar influences of local topography on the resulting diurnal patterns of precipitation were observed over the island of Borneo, where a progression of rainfall activity to the leeward side of the island was observed between midnight and morning hours (Ichikawa and Yasunari, 2006). Takahashi et al. (2010) highlighted the importance of topography in the Indochina peninsula, where early afternoon maximum was observed along the mountain ranges and coastal sections. Near the foothills of the mountain ranges the time of maximum rainfall occurred in the evening hours, while early morning showers were concentrated around the coastal areas over eastern Gulf of Thailand and Bay of Bengal. Late evening convective showers were also found in the interior of Sumatra (Mori et al., 2004).
Given the strong diurnal signal in rainfall patterns in different parts of south and southeast Asia, we focus in this investigation on the diurnal cycle in the occurrence of lightning events during the active summer season. There are a limited number of regional-level studies on the diurnal patterns of lightning activity specifically in tropical Asia, mainly due to the scarcity of detailed lightning data for the region. Recently, with the availability of satellite-based high-temporal resolution lightning data, such as the Tropical Rainfall Measuring Mission (TRMM) dataset, it is now possible to examine diurnal patterns of lightning strikes in less explored regions of the tropics. A few regional-level studies examining lightning patterns include South Korea (Hyun et al., 2010), Chongqing, China (Li et al., 2012), and eastern Tibetan plateau (Xu and Zipser, 2011), as well as for the entire tropics (Liu and Zipser, 2008). In case of South Korea, lightning data from station level observations revealed predominantly night-time lightning strikes in the south, while an afternoon peak was observed in the midland areas (Hyun et al., 2010). In another regional study over the Tibetan plateau involving the use of TRMM lightning data, Xu and Zipser (2011) showed an early morning peak in lightning activity over the central and eastern Tibetan Plateau foothills. More recently, there have been several studies examining lightning activity specifically over south and Southeast Asia in view of the concentration of active convection over the landmasses as a result of availability of TRMM data. Kumar and Kamra (2012) analyzed the variability in lightning activity during 1997–1998 and 2002–2003 El Niño/La Niña events, and found a significant increase in the number of flashes and average flash rate during the 1997–1998 El Niño Southern Oscillation (ENSO) event compared with the La Nina event in 2002–2003. In another study, Yuan et al. (2011) examined the lightning activity occurring over the western Pacific Ocean and its relationship with aerosol loadings, and found an approximately 60% increase in aerosol loading leading to 150% increase in lightning flashes. The results of their study suggested the modification of cloud microphysics due to increased aerosol loading which resulted in enhanced lightning activity. Additionally, the analyses of diurnal variations of extreme convection in South Asia by Romatschke et al. (2010) revealed the preferred time of formation of convective cores in the evening hours over land, as a result of near surface moist flow capped by dry air aloft. Furthermore, the results of their study revealed a nocturnal peak in the occurrence of convective cores along the foothills of the Himalaya. Similarly, late afternoon and midnight peaks were also observed in western China plain, and midnight to early morning peaks were observed in eastern China plain (Yuan et al., 2012). These patterns were attributed to mountain–valley and land–sea breeze interactions. However, there are no specific studies examining the detailed spatial patterns of lightning activity over an extended period of time in tropical Asia. Therefore, in this study we have used harmonic analysis to examine the spatial patterns of lightning activity across tropical Asia.
2. Data and methods
Lightning data from 1998 to 2011 were obtained from the TRMM dataset collected through the lightning imaging sensor (LIS) instruments, version 4.1 (Boccippio et al., 1998). It is inclined at 35° and at an altitude of about 350 km. This sensor is comprised of a 128 × 128 charge coupled device pixel array, with individual pixel resolutions from 3 to 6 km across, and a total field of view of 550 × 550 km2, and 90–95% detection efficiency. Furthermore, the LIS sensor array does not rotate, and almost all geographic locations observed by LIS during a given orbit are viewed for about 90 s. This sensor measures total intra-cloud and cloud-to-ground lightning strikes with minimal regional bias in the tropics, with very high spatial accuracy, and therefore ideal for storm scale applications and tropical climatological usage. Further detailed information about the instrument and specific characteristics is available from Boccippio et al. (1998). Additionally, the LIS detection efficiency ranges from about 69% near local noon to 88% overnight (Boccippio et al., 2002). The spatial span of the LIS orbits is between 35° north and south (Cecil et al.,2012). In this study the analyses of lightning activity were limited to the two convectively active months of July and August in most of the tropical Asia when lightning activity is at a maximum.
The dataset we analyzed was comprised of lightning events that occurred during 1998–2011, during the two predominantly rainy months of July and August specifically in Asia. The TRMM lightning data are stored in a hierarchical data format, with the lightning data stored in various groupings of lightning optical pulses, which match with physical features including thunderstorms, flashes, and strokes. In this study, the lightning flash activity was obtained specifically from the area file, which are defined as distinct regions of the Earth that have one or more flashes in a given orbit (Boccippio et al., 1998). The file had records for number of flashes in specific locations, along with the time and duration of the observations. There were 16 orbits per day, with lightning flash times and locations recorded with about 5 km resolution.
We developed a matrix with one row for each observation including the critical measurement of the latitude, longitude, and the time of the lightning flash recorded in International Standard Atomic Time (TAI93). We converted the atomic time values to month, day, year, and true solar time at the location of the lightning flash ranging from 0.00 to 24.00. As seen in Figure 1, there were a total of over 200 000 recorded flashes in July and August over the 1998–2011 time period with considerable variation from year to year. The spatial distribution of combined lightning flash rates in the study area during 1998–2011 is shown in Figure 2. We ultimately developed a lightning frequency matrix with 24 column, one for each hour from midnight to midnight, and 5566 rows, one for each 1° latitude × 1° longitude grid cell between 0°N and 45°N and 60°E–180°E. However, we used the criterion that if any grid cell had more than four hourly intervals with a frequency count of zero, the cell was eliminated. Because harmonic analysis basically fits trigonometric curves to the data, we would ideally want no more than one of the 24 frequency counts for any grid cell to have a value of zero. Using that as a criterion, we would eliminate far too many grid cells. We experimented with this decision and found that limiting the analyses to grid cells with no more than four hourly frequency counts of zero resulted in both meaningful harmonic analysis results and a sufficient number of grid cells to reveal spatial patterns in the timing of lightning activity. This left a final matrix of 621 rows and 24 columns containing the lightning frequency data in each of the 24 hourly intervals.
A broad review of the methodology used for analyzing diurnal patterns in atmospheric processes reveals harmonic analysis as a popular, widely used, and effective method (Schwartz and Bosart, 1979; Basu, 2007; Sen Roy, 2009). Therefore, in this study we also have used harmonic analysis to detect the diurnal signal in lightning activity over tropical Asia. The basic form of the harmonic equation is:
where f(x) is the estimated value in each interval, is the average value over the N observations, Ar is the amplitude of the rth harmonic wave, r is the frequency or number of times the harmonic wave is repeated over the fundamental period, θ is derived as 2πx/N where x represents the intervals through the fundamental period, and Φr is the phase angle of the rth harmonic often reinterpreted as the time of maximum. The basic form is expanded to:
where the Fourier coefficients, ar and br, are calculated as:
The amplitude, Ar, is calculated as (ar2 + br2)0.5, the standardized amplitude is calculated as Ar/2 , the phase angle, πr, equals tan−1 (ar/br), and the portion of variance explained by the rth harmonic wave, and Vr, is determined as Ar2/2s2 where s is the standard deviation of the N values.
The results of the harmonic analysis were next mapped to reveal the spatial patterns of the diurnal cycle of lightning activity in the study area. The time of maximum and variance explained were plotted using rotated wind direction symbols to show the time of maximum during the 24 h of a day by the wind direction line (Figure 3). The spatial variations in the strength of the diurnal cycle, in the form of the variance explained by the first harmonic were plotted using graduated symbol technique (Figure 4).
3. Results and discussion
The overall results of the harmonic analysis revealed distinct spatial patterns in the time of maximum lightning activity driven largely by the local topography and distance from the coast. In general, the interior low lying areas exhibited significantly different spatial patterns to the highland areas, including the Himalayas. Generally, the time of maxima for lightning activity was concentrated in the late afternoon to evening hours, which may be associated with the peaking of the local convective activity associated with heating of the surface in the interiors of the landmasses (Figure 3). A phase progression from east to west was observed in East Asia, and from southeast to northwest in south Asia from early afternoon hours to evening and later nocturnal hours in the interiors. This progression took place from the afternoon hours (near the coastal areas) to late afternoon to early evening hours in the interiors. The early afternoon peaking lightning activity near the coastal areas is caused by the interaction of land and sea breezes, resulting in the formation of thunderstorm clouds. The land–sea breeze interaction has been well documented in previous studies examining diurnal precipitation patterns in East Asia and South Asia, which is driven by the differential heating of land and water. This results in greater sensible heat flux over land, leading to enhanced vertical circulation along the coast (Kumar and Kamra, 2010; Yuan et al., 2012). The time of maximum lightning activity was close to midnight in high altitude areas with undulating topography, which included the Himalayas and the Western Hills in mainland China. The nocturnal peak in lightning activity along the foothills of the Himalaya is caused by the interaction of mountain and valley breezes, as well as diurnal heating of the surface. This is in conformity with the findings of previous studies by Houze et al. (2007) and Romatschke et al. (2010), which indicate the formation of convective cores along the foothills of the Himalaya, Karakoram Range, and Hindu Kush mountains. The suggested formation process of these deep convective cores were suggested to be caused by moist air flow from the Arabian Sea capped by dry westerly or north westerly midlevel flow blowing down from adjacent higher elevation regions (Sawyer, 1947; Houze et al., 2007). An in-depth analysis of National Centers for Environmental Prediction (NCEP) reanalysis fields (Romatschke et al., 2010) and results from field experiments conducted at the foothills of the Himalaya (Egger et al., 2000; Barros and Lang, 2003) revealed the change in wind direction from upslope during the day to downslope around midnight and early morning hours. This pattern shows the clear evidence of diurnal heating of land surface during day and radiative cooling during night-time. Furthermore, the role of nocturnal low level jets on the occurrence of nocturnal precipitation related processes have been highlighted in previous studies (He and Zhang, 2010; Huang et al., 2010).
The peak lightning activity over the Tibetan plateau was consistently in the afternoon hours. The time of maximum lightning activity was more haphazard and varied over short distances in the archipelagos of southeast Asia, southern Honshu island in Japan, and southern part of South Korea. Over Malaysia and Indonesia specifically the time of maximum was determined by the leeward or windward side of the highland areas, ranging from later in the afternoon to close to midnight hours. The overall predominance of lightning activity during the afternoon hours over the land areas was also reported in a previous study by Liu and Zipser (2008) using TRMM data. The afternoon peak in lightning activity is caused by the development of intense convection in the form of thunderstorm clouds over land. Similar afternoon maximum in lightning activity was also found in specific regional level studies such as in case of Chongqing (Li et al., 2012). Similarly, late afternoon peak in lightning activity associated with deep convection was also observed in the adjacent Tibetan plateau during the monsoon season (Xu and Zipser, 2011).
Furthermore, the close relationship between peak time of lightning activity and peak time of precipitation events have been the focus of several published literature (Liu and Zipser, 2008; Hyun et al., 2010). Overall, a phase-wise progression was observed for most of the tropics with varying lag time from the analysis of TRMM data by Liu and Zipser (2008). The results of their study specifically identified the longest life cycle of convective processes that were observed over Indonesia. In case of the Indian subcontinent, the time of maximum for lightning activity versus rainfall events across the Gangetic basin and the Deccan plateau (Sen Roy and Balling, 2007) showed a phase lag of about 2–3 h. However, the phase lag was longer specifically along the foothills of the Himalaya. Along the eastern coastal strip, the difference in the time of maximum between rainfall and lightning activity was minimal. Additionally, the analysis of 3 years TRMM PR (Precipitation Radar) data revealed a mid-afternoon peak in precipitation over the continental land areas and islands of Southeast Asia (Nesbitt and Zipser, 2003). Interestingly, the time of maximum for lightning activity in Southeast Asia was predominantly in the evening to midnight hours. Finally, the comparison of the diurnal patterns of lightning activity to that of precipitation patterns over the Tibetan plateau revealed similar lags of peak lightning activity (late afternoon hours) occurring a few hours before the peak rainfall activity (late afternoon and midnight hours) (Yuan et al., 2012). Similar lags were also found over eastern China plain, where the peak lightning activity occurred in early afternoon hours and double peak in precipitation activity in late afternoon and early morning hours.
The strength of the diurnal cycle was stronger in the interior low lying areas of Indian subcontinent and mainland China, above 15°N (Figure 4). Some of the strongest signals were observed over the Tibetan plateau, eastern mainland China, and along the Gangetic basin. Parts of Southeast Asia, including Thailand, Vietnam, and northern part of Cambodia also experienced moderately strong diurnal signals. The strength of the diurnal cycle was relatively less near the Equator, and along the foothills of the Himalayas and in central mainland China near the Yangtse river basin. The oceanic areas in general experienced weaker diurnal signals, thus highlighting the differential heating of land and sea, in the resulting convective processes and low overall lightning activity. This is further validated by Christian et al. (2003) who reported that the majority of the lightning strikes occur over land, with a mean annual land to ocean flash ratio of 10 : 1.
The active convection over South Asia, particularly during the summer pre-monsoon and monsoon months has been widely documented in previous studies (Kodama, et al., 2005; Zipser et al., 2006; Houze et al., 2007). More specifically, Zipser et al. (2006) highlighted the occurrence of some of the deepest intense convection in the arid northwest. The results of their study further mention about the occurrence of most intense convective atmospheric processes in low rainfall regions. Therefore, the presence of stronger diurnal signal in the occurrence of lightning activity in northwestern India caused by the development of an intense low pressure area here further is consistent with the findings from previous studies.
The greater strength in the diurnal cycle over central part of South Korea is related to the orographic uplift of westerly and southwesterly flows of moist air into this region. In conformity with the findings of Hyun et al. (2010) the strongest diurnal signals were observed in the central part of South Korea, which also experiences substantially higher number of thunderstorms in comparison to the rest of the country. Similar role of topography on the moist onshore blowing air leading to strong diurnal strength of lightning activity was also observed in Southeast Asia, mainly Thailand and along the Arakan Yoma ranges in Myanmar.
In this study the diurnal timing in the occurrence of lightning activity over tropical Asia was analyzed using over a decade of satellite-based TRMM lightning sensor data. The availability of high spatial and temporal resolution lightning data over an extended period of time has enabled the examination of complex convective atmospheric processes in the form of lightning activity in tropical Asia. Given the relative abundance of published literature on the diurnal timing of precipitation events, from both satellite and surface-based station level observations, the results of our study further extend our understanding of atmospheric processes in this region. The analysis was limited to the two main summer rainfall months of July and August in most of Asia. The main findings of our study are summarized below:
Most of the lightning activity in tropical Asia occurs over land, with significantly lesser proportion occurring over the ocean areas.
A phase-wise progression was observed in the time of maximum lightning activity from the low lying areas in the southeast to the foothills of the Himalayas. In general, the lightning activity peaked during the afternoon hours in the interior of the continents and low lying areas. This may be associated with the peaking of convective processes over land areas.
The interaction between orography and weather systems was highlighted in the case of major mountain ranges including the Himalayas, Western Hills in southeast mainland China, and the Arakan Yoma ranges in Myanmar. The time of maximum for lightning activity over the mountains was mostly in the nocturnal hours as a result of the interaction between mountain and valley breezes.
In the case of the coastal areas the lightning activity usually peaked in the early evening hours, as a result of the interaction between the sea and land breeze.
The strength of the diurnal cycle was strongest in the interior of the continents, particularly over northwestern India and the Tibetan plateau. The cycle was weaker in general near the Equator and along the foothills of the Himalaya.
Finally, the comparison of time of maximum for rainfall (analyzed in previously published literature) and lightning activity revealed a lag of several of hours for the lightning maximum after the rainfall maximum. This lag increased over the high altitude regions of Himalaya, while the lag was minimal in the coastal areas.
The results of this study provide valuable information about convective processes in tropical Asia. The availability of this detailed dataset makes it possible for future studies to focus on diurnal patterns in specific characteristics associated with lightning flashes in the tropics.