1APRIL 2007 K ILINC AND BERINGER 1161

The Spatial and Temporal Distribution of Lightning Strikes and Their Relationship with Vegetation Type, Elevation, and Fire Scars in the

MUSA KILINC AND JASON BERINGER School of Geography and Environmental Science, Monash University, ,

(Manuscript received 17 October 2005, in final form 3 July 2006)

ABSTRACT In this paper the authors explore the spatial and temporal patterns of lightning strikes in northern Australia for the first time. In particular, the possible relationships between lightning strikes and elevation, vegetation type, and fire scars (burned areas) are examined. Lightning data provided by the Bureau of Meteorology were analyzed for a 6-yr period (1998–2003) over the northern, southern, and coastal regions of the Northern Territory (NT) through the use of Geographical Information Systems (GIS) to determine the spatial and temporal characteristics of lightning strikes. It was determined that the highest densities of lightning strikes occurred during the monsoon transitional period (dry to wet) and during the active monsoon periods, when atmospheric moisture is highest. For the period of this study, lightning was far more prevalent over the northern region (1.21 strikes per km2 yrϪ1) than over the southern (0.58 strikes per km2 yrϪ1) and coastal regions (0.71 strikes per km2 yrϪ1). Differences in vegetation cover were suggested to influence the lightning distribution over the northern region of the NT, but no relationship was found in the southern region. Lightning strikes in the southern region showed a positive relationship with elevations above 800 m, but no relationship was found in the northern region, which could be due to the low-lying topography of the area. A comparison of lightning densities between burned and unburned areas showed high variability; however, the authors suggest that, under ideal atmospheric conditions, large-scale fire scars (Ͼ500 m) could produce lightning strikes triggered by either enhanced free convection or mesoscale circulations.

1. Introduction great deal of research on the electrical structure of thunderstorms and the relationships between the po- Information concerning the characteristics of light- larities of the strikes (Latham 1991; Orville 1994; Von- ning strikes in different geographical regions is of in- negut et al. 1995; Orville et al. 2002). However, there is terest and can augment research on the interaction be- very little understanding of the relationship between tween the radiative properties of the surface and the lightning strikes and surface characteristics of eleva- atmosphere. The surface energy balance, albedo, sur- tion, vegetation type, or other surface inhomogeneities, face roughness, and the Bowen ratio are important fac- such as fire scars (burned areas). tors in determining available energy and the partition- Lightning strikes are produced mainly from cumu- ing of the energy fluxes over different surface types lonimbus, which are formed through four mechanisms: (Beringer and Tapper 2002), which are important char- buoyant warm air rising due to intense surface heating, acteristics in determining the microclimate and regional strong heating contrast between surfaces, frontal lifting, climate. The differential heating of two adjacent sur- or by the uplift of air parcels due to orographic lifting faces caused by radiative flux contrasts is likely to cause (Sturman and Tapper 1996). All of these processes may convective activity through uplift and generate meso- trigger convection and, hence, lightning activity. Light- scale circulations (Pielke and Avissar 1990). If the heat- ning originates around 3–4 km above sea level and is ing contrast is large, then enhanced convection may effectively caused by a charge separation that takes occur, leading to the formation of thunderclouds and place within the negatively charged reservoir of the lightning (Dissing and Verbyla 2003). There has been a cloud and the positive electric field of the ground sur- face (Cooray 2003). The resulting discharge is negative, positive, or a cloud-to-cloud stroke. Corresponding author address: Musa Kilinc, School of Geogra- phy and Environmental Science, Monash University, Victoria In the Northern Territory (NT) of Australia (Fig. 1), 3800, Australia. thunderstorms and lightning strikes are most common E-mail: [email protected] during convective periods, when there is a strong influ-

DOI: 10.1175/JCLI4039.1

© 2007 American Meteorological Society

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC

JCLI4039 1162 JOURNAL OF CLIMATE VOLUME 20

FIG. 1. Map of the Northern Territory (Australia) illustrating the three study areas used; northern, southern, and coastal regions. ence from the monsoon. The monsoonal influence dur- tween land and ocean have shown that there is high ing the wet season has a significant role in distributing variability between the distribution of lightning strikes moisture throughout the northern region; however the over the ocean and land (Boccippio et al. 2000; Wil- southern regions are very dry and are influenced by liams and Stanfill 2002; Williams et al. 2002). subtropical cold fronts (Beringer and Tapper 2000). Forced convection, or orographic lifting, triggers con- Thunderstorms develop in homogeneous air masses vection by transporting sensible and latent heat verti- that are associated with convergence zones and insta- cally into the atmosphere via uplift. The convection bility caused by monsoon onset and the active monsoon may then intensify the instability of an area by stronger period (wet season) (Sturman and Tapper 1996). The updrafts and cause lightning discharges. For example, contrast between the heating properties of the ocean the valley/mountain winds produced from differential and land is a trigger for strong sea-breeze development, heating between two sloping surfaces, combined with which can stretch several hundred kilometers inland orographic lifting, can be an instigator of lightning (Simpson 1994). Sea breezes are a form of mesoscale strikes and therefore produces a relationship between circulation and aid in the development of convective elevation and lightning strike density (López and Holle activity via frontal uplift and cause instability, and 1986; Lericos et al. 2002; Orville et al. 2002; Dissing and hence thunderstorms and lightning. Previous studies in- Verbyla 2003). Dissing and Verbyla (2003) found that a vestigating the contrast of lightning strike density be- positive correlation between lightning strike density

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1APRIL 2007 K ILINC AND BERINGER 1163 and elevation existed up to a maximum elevation of lift over a fire scar. An increased heat source over 1100–1200 m. Mesoscale circulations can aid in the de- burned areas may also lead to buoyancy and result in velopment of convective activity via uplift and cause local and regional instability. In a model used by convergence and instability over the affected area, Knowles (1993), a halving of the albedo over burned which may then trigger thunderclouds and hence light- vegetation areas increased convection and resulted in ning. The relationship between lightning strike distri- the formation of a mesoscale circulation system. bution and vegetation was also studied by Dissing and In this paper we explore the spatial and temporal Verbyla (2003), who found that mesoscale circulations patterns of lightning strikes in northern Australia, for triggered by the differential heating between two con- the first time, and the possible relationships between trasting vegetation types were likely to produce light- lightning strikes and elevation, vegetation type, and fire ning strikes. Surface inhomogeneities provide a heating scars. Through the use of Geographical Information contrast between two adjacent surfaces, which could Systems (GIS), we analyzed lightning data provided by produce mesoscale circulation patterns similar to a sea the Bureau of Meteorology for a 6-yr period (1998– breeze (Segal et al. 1988), though other examples have 2003) over the northern, southern, and coastal regions also been previously studied: snow breeze (Segal et al. of the NT. 1991), salt lake breeze (Tapper 1991), and lake breeze (Laird et al. 2003). 2. Study area Heating contrasts between surfaces arise due to dif- ferences in albedo, surface roughness, and the way in The Northern Territory (Australia) was selected as which energy is partitioned into sensible, latent, and the study area for this project since Christian et al. ground heat flux (Beringer and Tapper 2002). For ex- (2003) estimated that, on average, 44 Ϯ 5 lightning ample, Pielke and Vidale (1995) found that a larger flashes occur around the globe every second, which sensible heat flux over particular vegetation types in- strike between the geographic regions of 30°N and 30°S creased the air temperature and, as a result, triggered and account for approximately 75% of the global light- convection. In the Maritime Continent Thunderstorm ning count (Torancita et al. 2002). In addition, the NT Experiment (MCTEX), Beringer and Tapper (2002) has a variety of vegetation types from grassland to rain- showed that the sensible heat flux among various veg- forest and moderate terrain variability ranging from etation types (savannah, grasslands, forest, and shallow ϳ400 m in the northern region to ϳ1500 m in the south- tidal strait) was an important factor in the production of ern region. A high degree of burning, especially during a buoyant boundary layer that could produce conver- the dry season, occurs and previous studies suggest that gence as an uplift mechanism. up to 5.5% of the NT can undergo burning (Beringer et Almost one-third of Australia’s savanna region, in- al. 1995). The study area encompassed the NT coastal cluding the NT, is burned each year by pastoralists, waters and islands, including Groote Eylandt and the Aboriginal landholders, and conservationists (Russell- Tiwi Islands. Therefore, the study area was divided into Smith et al. 2000). Resultant fire scars provide large- three geographic zones: the northern region, the south- scale inhomogeneities that could potentially provide a ern region, and the coastal waters (Fig. 1). trigger for lightning. Lightning itself, is not a significant The northern region (Ϫ10° to Ϫ16°S) has a distinct source of fire ignition; however during the transition monsoonal climate with both a wet and dry season. The period from dry to wet, when fuel moisture is still rela- wet summer months, from December to March, are hot tively low, lightning strikes may ignite fires (Rorig and and humid with high rainfall, which is roughly 80% of Ferguson 2002). A study of the impact of savanna fires the annual rainfall (Sturman and Tapper 1996). The dry on the surface energy balance was recently completed season lasts from May to September and is dominated by Beringer et al. (2003), who found that the surface by fair sky conditions with little rainfall and low humid- energy properties varied considerably before and after ity from the southeasterly winds (Bureau of Meteorol- a fire event. Post fire albedo values were shown to de- ogy 1998). These two contrasting seasons are separated crease by half, from 0.12 to 0.06, while the latent heat by two periods of transition that occur around March flux over the surface decreased by 30%–75% (Beringer and October. et al. 2003). Consequently, fire scars were more in- On the other hand, the southern region is not af- tensely heated and therefore the sensible heat flux in- fected by the monsoonal flows since the intertropical creased by ϳ 36% when compared to the surrounding convergence zone (ITCZ) does not extend that far unburnt surfaces. This has the potential to trigger con- south, but instead is influenced by the southeasterly vection and lightning due to the development of a me- flows emanating from the subtropical continental air soscale circulation and associated convergence and up- mass in central Australia.

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1164 JOURNAL OF CLIMATE VOLUME 20

The vegetative landscape of the NT is diverse and year-to-year variability is small, suggesting consistent supports a range of vegetation types such as tropical patterns over that period. open woodland, eucalypt (Eucalyptus Tetrodonta and The dataset was incorporated into a GIS to allow for E. miniata) and acacia woodland (Acacia aneura), the exploration of relationships of lightning strikes with shrubland (A. georginae), and tussock and hummock surface characteristics, such as elevation, vegetation, grasslands (Williams et al. 1997). The distribution and and fire scars. The project was undertaken in the GIS composition of vegetation is strongly associated with laboratory at Monash University (Clayton) using Arc- the monsoonal climate and consequently rainfall (Hut- Map (ESRI, Inc., 1999–2002, version 8.3; http:// ley et al. 2001). For example, the eucalypt and acacia www.esri.com). woodlands are found in the north of the NT, where For the analysis of lightning strike distribution with annual rainfall is ϳ1700 mm, while much of the shrub- vegetation and elevation, only the northern and south- lands and grasslands lie in the southern regions, which ern regions were used. The vegetation dataset used in are subject to only ϳ250 mm. The NT has a moderately this project was the National Vegetation information variable landscape with large, flat sand plains and dis- System (NVIS) 2001—Major Vegetation Groups— connected ranges. Much of the northern region is flat Version 1.0 (Department of the Environment and Heri- with a maximum elevation of 400 m, with the main tage 2004). The vegetation classes were merged from geological feature of a sandstone plateau located in the original 27 classes to 4 classes: tropical open wood- Western Arnhem Land. There are extensive ranges in land, eucalypt and acacia woodland, shrubland, and the southern regions; the MacDonnell Range runs west grassland. The dataset had a cell resolution of 1 km and and east of Alice Springs; the Davenport and Murchi- was rescaled to a cell resolution of 250 m to be consis- son Ranges lie north; and the Harts and Dulcie are tent with the elevation dataset. Elevation data were northeast. There are also numerous ranges that are obtained from the GEODATA TOPO 250K Series 2 scattered around the region, in excess of 1000 m, and project created by Geoscience Australia. The original the highest peak is Mount Zeil at 1531 m. 50-m contour, vector tile map sheets at a scale of 1:250 000 were rasterized to a 250-m cell resolution with 3. Method 100-m contour intervals. The lightning strike dataset was rasterized to a 250-m To examine the spatial and temporal variation of cell size and grouped into two spatial datasets (northern lightning strikes we used a ground-based lightning de- and southern regions) and four temporal (seasonal) tection network for the Australasian region that was datasets: early wet (October–December), late wet made operational in 1998 by Global Positioning and (January–March), early dry (April–June), and late dry Tracking Systems (GPATS). The system uses the series (July–September), so that lightning strikes in relation III Lightning Positioning and Tracking System to elevation and vegetation could be analyzed. As a (LPATS). The basic principle of operation is to use the result of rasterizing the data from a feature dataset with time of arrival of the lightning discharge at three or a spatial accuracy of 200 m to grid cells of 250 m, the more receivers. The three sensor system can effectively Modifiable Area Unit Problem (MAUP) is created monitor large areas, 1 million km2. The detection effi- since analyzing different sized cells gives different re- ciency of the LPATS sensor is greater than 90% with a sults. However, this was unavoidable owing to the pro- spatial accuracy of 200 m (Global Positioning and cessing power of the computer and is likely to slightly Tracking Systems, 2004). The GPATS sensors are able underestimate lightning strike density. A possible to detect strokes as close as 0.5 ms apart and correctly model that would allow for a thorough investigation identify them into three stroke types: positive, negative, of MAUP and other sources of error would be in the and cloud-to-cloud strokes. The lightning data acquired form of spanned from October 1998 to December 2003; how- ever some months of lightning data were missing, spe- ͓strike density͔ ϭ ͓elevation term͔ ϩ ͓vegetation term͔ cifically January 2002, March 2003, April 2000, May 1999 and 2000, June 2000 and 2003, July 2003, August ϩ ͓spatial autocorrelation͔ ϩ ͓error͔. 2003, September 2003, October 2003, and November 2003. The gaps within the dataset effectively make this Only the northern and southern seasonal data for the investigation a 4-yr study period, which may not be early dry, early wet, and late wet seasons were used. representative of the broad patterns of lightning strike The late dry season was omitted because of the lack of density over the NT, and hence the conclusions pre- strikes. The hypothesis that fire scars could initiate me- sented in this paper are only preliminary. However, the soscale circulations producing lightning strikes was in-

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1APRIL 2007 K ILINC AND BERINGER 1165

TABLE 1. Summary of lightning strikes in the northern, southern, and coastal regions of the Northern Territory.

Stroke Total Percentage Northern Percentage Southern Percentage Coastal Percentage type strikes of strikes region strikes of strikes region strikes of strikes region strikes of strikes Positive 5 736 356 76 1 258 492 75 580 637 70 527 953 72 Negative 1 043 877 14 187 458 11 189 340 23 79 144 11 CC 769 353 10 232 945 14 58 565 7 125 015 17 Total 7 549 586 1 678 895 828 542 732 112 Density* 1.10 1.21 0.58 0.71

* Mean annual strikes kmϪ2 yrϪ1. vestigated by utilizing the fire scar datasets derived 4. Results from the Moderate Resolution Imaging Spectroradiom- Between October 1998 and December 2003, 7 549 586 eter (MODIS) and Advanced Very High Resolution positive cloud-to-ground, negative, and cloud-to-cloud Radiometer (AVHRR) (Department of Land Informa- (CC) stroke types were recorded in the study area. Of tion 2004). those, 76% of total lightning strikes were of positive Twelve fire scars that burned during October or No- polarity and 14% negative polarity while 10% were a vember were selected for the years 2000–02. We se- CC stroke (Table 1). We are uncertain as to why our lected fire scars to include a range of sizes from small study has such high rates of positive strikes when com- (Ͻ200 km2) to large fires (Ͼ500 km2). When choosing pared to other studies around the world (see, e.g., Pinto the fire scar samples, sites that were near the coast were et al. 1996; Orville et al. 1987); this may require further rejected because dynamical processes near coastal en- examination. A possible explanation for the high per- vironments differ from the dynamical processes occur- centage of positive strikes is the GPATS lightning sen- ring in the interior regions. Similarly, sites that were in sors’ ability to differentiate between negatively and close proximity to other fire scars were also rejected positively charged strikes, as has been documented with because fires surrounding the sample site may have ef- magnetic direction-finding networks (Stolzenburg fected the lightning distribution of the general region, 1994). through the production of a larger circulation system. Following Latham (1991), who investigated lightning a. Land–ocean lightning contrast strikes during an actively burning fire, we created a 50-km buffer zone around each fire scar with 10-km For the period of this study, lightning was far more 2 intervals. Control areas (same attributes as the fire scar prevalent over the northern region (1.21 strikes per km Ϫ1 2 Ϫ1 sample, i.e., area, vegetation, and elevation) were cre- yr ) than over the southern (0.58 strikes per km yr ) 2 Ϫ1 ated to provide a measure of lightning strike density in and coastal regions (0.71 strikes per km yr ). The an area, free of burning that could be compared to the early and late dry periods were characterized by very fire scar area itself during the same time period, since few strikes for the three regions (Fig. 2). The northern lightning activity increases rapidly during the transition from the dry to the wet season. These areas were cre- ated roughly 100 km away from the fire scar in order to negate the response of seasonality. The changes in al- bedo and sensible and latent heating caused by a sa- vanna fire can persist up to several weeks to months (Beringer et al. 2003): therefore, mesoscale circulations are most likely to develop most strongly soon after the fire event when there is the greatest contrast in the surface energy balance (Knowles 1993). We analyzed lightning strikes two weeks prior and two weeks after the fire event to observe whether or not there was a change in lightning strike density over the fire scar dur- ing the specified time period. We then compared these results with the control site to find if seasonality was a factor in influencing the lightning strike density over FIG. 2. Monthly mean distribution (1998–2003) of lightning the fire scar sample. strikes for the northern, southern, and coastal regions.

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1166 JOURNAL OF CLIMATE VOLUME 20 and southern regions both had maximum lightning strike density during November, which coincides with the southerly movement of the ITCZ. With the estab- lishment of the ITCZ, the frequency of lightning strikes begins to slowly decrease through the late wet season. However, the coastal region appears to not be grossly affected by the southerly shift and establishment of the ITCZ compared to the other two regions since there is little variability between November, December, and January. This study showed that the coastal region of the NT contained a greater concentration of lightning strikes over the coast (55%) than the southern region (45%), which differs from the studies of Christian (1999) who found that there was a considerable difference between FIG. 3. Storm height distribution observed by the TRMM pre- the lightning strikes detected between the land (82%) cipitation radar in January 1999 for northern Australia. Dark and the ocean (18%). However, this difference can patches around Darwin, Groote Eylandt, and the Gulf of Joseph probably be attributed to the scale of study. Christian Bonaparte indicate areas of elevated storm height (http:// investigated the lightning distribution on a global scale, www.eorc.jaxa.jp). whereas this study was based specifically on the North- ern Territory, and its coastal waters. Boccippio et al. ergy available but lacks a trigger mechanism to produce (2000) focused their study on the tropical zone and lifting and lightning strikes (Williams and Stanfill 2002). found that the land–ocean lightning contrast varied by CAPE alone may not influence lightning strike den- a factor of 2, which compares well with the northern sity but may also be dependant on cloud height. The region of our study, but not so well with the southern work presented by Ushio et al. (2001) and Mushtak region. et al. (2003) on storm height and lightning density sug- There are a number of factors explaining the land– gests that the cloud height (low over ocean and high ocean contrast. The most common explanation relates over land) influences the conversion of CAPE to to the differential heating between the ocean and the updraft kinetic energy. A larger cloud height would land. The land surface is asymmetrically heated faster allow for wider and stronger updrafts within the cloud than the ocean because of the lower thermal inertia and and thus would lead to a larger main negative charge higher sensible heat flux over the land, illustrated by region, hence higher strike rates over the areas (Mush- the higher Bowen ratio for land (0.2–1) compared to tak et al. 2003). The Gulf of Joseph Bonaparte, the ocean (0.1) (Williams et al. 2002). This heating differ- Darwin area, and Groote Eylandt all experience the ence causes the lower atmosphere to become vertically highest cloud-base storm height as seen from the Tropi- unstable and may cause deep convection and therefore cal Rainfall Measuring Mission (TRMM) radar image trigger thunderstorm activity (Williams and Stanfill (Fig. 3). 2002). One measure of the potential convective inten- sity is the convective available potential energy b. The spatial and temporal distribution of (CAPE). A thunderstorm with a larger CAPE is likely lightning strikes to produce a stronger updraft within the cloud, where the vertical motions are likely to affect the mixed phase On a temporal basis, there was a large seasonal con- region of the cloud and the charge separation, allowing trast across all regions with peak densities occurring in for the development of a more vigorous and electrically the early and late wet seasons. Seasonally averaged intensified storm (Cooray 2003). A larger CAPE can lightning strike densities for the study period are pre- either result from strong surface heating and hot sented. boundary layer air or by the presence of cold air aloft 1) EARLY WET SEASON (Williams and Stanfill 2002). CAPE values for oceans and the northern region of the NT are quite similar— During the early wet season (Fig. 4a), lightning was both experience values between 0 and 3000 J kgϪ1 (Wil- uniformly distributed across the whole region, though liams and Stanfil 2002). Furthermore, when there is a the southern region had less frequent strikes (0.00–0.50 greater CAPE over the ocean, the lightning activity is strikes kmϪ2). This is expected since the southern re- still low since CAPE only represents the potential en- gion is associated with southeasterly wind flow, while

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1APRIL 2007 K ILINC AND BERINGER 1167

FIG. 4. Mean (1998–2003) lightning strike density (per km2) showing the spatial and temporal variability of lightning strikes in the Northern Territory during the (a) early wet season, (b) late wet season, (c) early dry season, and (d) late wet season. Note that scales are not constant.

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1168 JOURNAL OF CLIMATE VOLUME 20 the northern region is associated with northwesterly strikes kmϪ2). However, lightning strikes in the north- monsoonal flow. The density values were much higher ern region were located around the coastal regions and along the base of the northern region, with some areas did not transgress inland, which may be associated with experiencing densities greater than 7.00 strikes kmϪ2 the retreating ITCZ. There was also a strong concen- (e.g., Groote Eylandt and the Gulf of Joseph tration of lightning strikes in the southeastern region, Bonaparte). The high densities over Limmen Bight which may have arisen as a result of cold front incur- may be caused by local convective systems and by the sions. interaction between the northwest monsoon flow and a 4) LATE DRY SEASON sea breeze influenced by the strong horizontal winds from the Gulf of Carpenteria. This area is also charac- The late dry season is characterized by little atmo- terized by a regular convergence line that produces spheric moisture, and thus fewer lightning strikes were morning glories as well as squall lines (Smith and observed when compared to other seasons (Fig. 4d). A Noonan 1998). Interaction of these atmospheric distur- majority of the strikes were located in the central inte- bances with monsoon flow may enhance deep convec- rior region with a slight bias toward the west (0.16 tion and produce an increase in lightning density. A strikes kmϪ2). As in the case of the early dry season, similar process is noted for the Gulf of Joseph many of these strikes can be attributed to cold front Bonaparte. However, the main difference between the incursions emanating from central Australia. The two is that the lightning density over Limmen Bight northern region was associated with very few to no appears to be localized while the density over the Gulf strikes at all (0.00–0.054 strikes kmϪ2), which was likely of Joseph Bonaparte is strengthened by the monsoonal caused by the limited moisture availability that re- westerlies and thus affects a broader area. stricted cloud formation and lightning strikes.

2) LATE WET SEASON c. Vegetation, elevation, and fire scar analysis The late wet season is characterized by a well- Previous studies have found that vegetation affects developed active monsoon, which produces a band of the distribution of lightning (Dissing and Verbyla highly concentrated strikes south and west of the north- 2003). Likewise, orographic lifting caused by elevated ern region (Fig. 4b). The southern region of the NT has areas was shown to influence lightning strike density low strike rates since the monsoonal flow does not (López and Holle 1986; Lericos et al. 2002; Orville et al. reach that far south. An interesting feature of the late 2002; Dissing and Verbyla 2003). To investigate the re- wet season is the low frequency of lightning strikes lationship between lightning strikes, vegetation, and el- (0.00–1.07 strikes kmϪ2) that occurred over Arnhem evation an exploratory data analysis was undertaken. Land and the Gove Peninsula, compared to the west Box plots were used to explore the data in this section coast of the NT. There may be a number of possible because they show the variability within the data and explanations for this phenomenon. The most likely ex- also provide an indication of the symmetry and skew- planation is that thunderstorms tend to be favored over ness of the data. A statistical analysis of the data, the western half of the northern region because of low- namely analysis of variance (ANOVA) and a Chi- level convergence affecting the area, whereas the east- squared test, was attempted, but the results are not ern part is under the influence of a divergent ridge presented in this discussion because some datasets pushing in from (J. Arthurs 2004, personal could not be normalized or showed inhomogeneity of communication). Also, sea-breeze convergence is ex- variance. The highly variable nature of the data, both pected to be stronger in the west than in the east, es- spatially and temporally, made it difficult to apply stan- pecially if there are low-level easterly flows, which dard statistical techniques. To use box plots, the data may trigger storms that penetrate a long way inland were first normalized by taking the logarithm of the (J. Arthurs 2004, personal communication). density so that the data would better fit a normal dis- tribution curve to decrease the affects of outliers and 3) EARLY DRY SEASON extremes. The early dry season is characterized by the end of The southern region vegetation analysis showed no the transition from the wet to the dry season and had relationship between vegetation type and the distribu- fewer strikes because the dry southeasterly wind flow tion of lightning strikes (Fig. 5a). The mean for the begins to dominate the landscape (Fig. 4c). Remnants three vegetation types lies within Ϫ3.65 and Ϫ3.80 (log of moisture from the late wet season over the northern annual strikes kmϪ2), though eucalypt/acacia woodland region were still evident, which affected the stability of and shrubland show greater variability from the mean, the area and hence lightning strike density (Ͼ0.49 probably because of the smaller area of these vegeta-

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1APRIL 2007 K ILINC AND BERINGER 1169

FIG. 5. Mean lightning strike density divided into vegetation classes for the (a) southern region and (b) northern region. Graphs show box plots with mean Ϯ standard error. Note that the less negative numbers translate to higher lightning strike density. tion types. In contrast, the northern region vegetation with elevation (Fig. 6b) showed no significant relation- results showed that lightning strikes were significantly ship with lightning density. However, elevation for the higher for grasslands (Ϫ2.8 log annual strikes kmϪ2), northern region extends only to 400 m, and therefore followed by shrublands, tropical woodland, and euca- this is consistent with the southern region. lypt/acacia woodland (Fig. 5b). However, the standard We mapped lightning strikes two weeks prior and error was quite variable. Shrublands had the greatest two weeks after 12 fire events (7 cases for the Novem- variability, while eucalypt/acacia woodland had the ber months and 5 cases for the October months) be- least. tween 1999 and 2003 to investigate the relationship be- Annual lightning strikes in the southern region, in tween fire-scarred areas and lightning activity. Out of relation to elevation, showed that lightning strikes were the 12 cases, 5 showed higher densities over the 50-km consistently low below a threshold elevation of 800 m buffer zone than the control after samples (Table 2), (Fig. 6a), after which lightning strike density increased which would suggest that under ideal atmospheric con- almost linearly with elevation. The highest lightning ditions an increase in lightning density can occur. Three strike density was at an elevation of 1200 m (Ϫ1.8 log of the five cases that showed an increase in lightning annual strikes kmϪ2), but was quite variable owing to strike density toward the fire scar were large-scale fires the small extent of high elevation areas. In the northern (ϳ500 km2). When considering the spatial pattern of region, the annual lightning strike density relationship lightning strikes over the fire scar, the wind flow and

FIG. 6. As in Fig. 5 but for mean lightning strike density divided into elevation intervals of 100 m.

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1170 JOURNAL OF CLIMATE VOLUME 20

TABLE 2. Effects of fire scar area on lightning strike density two 5. Discussion weeks before and after a fire. Differences in surface properties, such as roughness Case Area of fire Comparison of lightning strikes and the partitioning of energy (Bowen ratio) into sen- study scar (km2) before and after a fire scar sible and latent heat flux, may influence lightning strike 1 354 No effect density. The results from the box plots indicate that 2 1135 No effect vegetation does not influence the density of lightning 3 496 Increase 4 198 Increase strikes in the southern region. A likely explanation is 5 452 No effect the small variation in heating rates over the sparse veg- 6 341 No effect etation types since the southern region is homoge- 7 1857 Increase neously dry for most of the year. The main influence on 8 1197 No effect lighting activity over the southern region is likely to be 9 1213 Increase 10 665 No effect caused by cold front incursions rather than a variation 11 163 No effect in vegetation. Conversely, the northern region showed 12 276 Increase a significant difference in lightning density between vegetation types with grassland having the highest rates. These grasslands are surrounded by woodlands wind direction must also be accounted for. A possible that may cause mesoscale interactions to arise. How- mesoscale-induced thunderstorm can be conceived as ever, it is more than likely that the grasslands have a convergence from all sides of the fire scar, and a case of much higher sensible heat flux because they have se- this is illustrated in Fig. 7a. Large clusters of strikes nesced during the dry season. Woodlands continue to north, northwest, and southwest of the fire scar can be extract deep soil water and have much lower Bowen seen, with the highest densities of strikes occurring ratios compared with grasslands (Beringer et al. 2003). within 20 km. The other fire scars did not show consis- Therefore increased sensible heat flux may drive ther- tent patterns, but some showed a plume formation up- mal convection and increase lightning strike density. In wind or downwind of the fire scar (Fig. 7b). addition, grasslands are located partly on the fringes of

FIG. 7. Spatial distribution of lightning strikes over a fire scar two weeks after a fire event, with a 50-km buffer zone: (a) possible mesocale-induced circulation as the lightning strikes are clustered around the fire scar (b) showing a plume formation northwest of the fire scar.

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1APRIL 2007 K ILINC AND BERINGER 1171 the coast where they are probably influenced by meso- heating, but its impact on the heating of the boundary scale sea-breeze development, which is also a mecha- layer will be minimal. This hypothesis is consistent with nism for convection and, hence, lightning strikes. How- our results, which showed increased lightning strike ever, these speculative causes cannot be confirmed us- density over larger fire scars. The heating contrast over ing the current data. the large area is also likely to affect the atmospheric Atmospheric conditions are generally unsaturated pressure over the fire scar, perhaps causing an inflow through the atmospheric profile in the southern region from all sides of the scar to converge and uplift air over due to low atmospheric moisture availability. For light- the fire scar. Larger fire scars have also been associated ning strikes to occur, cloud formation is necessary via a with an increase in vertical air velocity causing an in- trigger mechanism (Sturman and Tapper 1996). Cloud crease in updrafts and a larger circulation center over formation in the southern region is most likely to occur or around the fire scar (Knowles 1993). Our results via orographic uplift of moist air incursions from south- supplement Knowles, who also hypothesized that a ern cold fronts, which occur infrequently (Beringer and larger burn area would lead to a more intense meso- Tapper 2000). The results from the analysis show that scale circulation around the fire scar. Our “plume” there is little variation in lightning strike density to 700 strikes are most likely associated with convection in- m, which may be around the point of the lifting con- duced by buoyancy and by a shifting of the wind. The densation level (LCL) during cold front incursions. plume strikes may be associated with or influenced by Lightning strike density then increased consistently other local convection systems and squall lines in the with higher elevation. This is due to the fact that the region. It may also be possible that lightning strikes higher elevation more often causes air to be lifted to the may be associated with pyrocumulus clouds from the LCL under a larger range of conditions compared to smoke plumes caused by the fire itself. Pollutants from the 800-m elevation, which only caused the LCL to be bushfires, on occasions, are able to generate increased reached under the most optimal conditions. The vertical updrafts by the heat from the fire; when these 1200-m contour had the highest strike density, consis- interact with the atmosphere under ideal conditions a tent with the results achieved by Dissing and Verbyla pyrocumulus cloud may develop (T. Bannister 2003, (2003). Other studies investigating the spatial distribu- personal communication). However, studies by Latham tion of lightning strikes have not quantified their data in (1991) and Vonnegut et al. (1995) showed that fire- relation to elevation but have, however, observed posi- induced clouds alone could not generate lightning tive relationships (López and Holle 1986; Lericos et al. strikes, though under the influences of other atmo- 2002; Orville et al. 2002). In the northern region it is spheric circulations occurring in the region an increase evident that elevation is not an important control fac- in lightning density may occur. tor. This is partly due to the shallow elevation range We have suggested some plausible mechanisms and (0–400 m); therefore convection via uplift is not neces- data to explain the interaction between fire scars and sarily expected. Additionally, the highest elevation lightning strike density; however there is scope for fur- zones in the north are located in Arnhem Land where ther work. For example, an analysis of high temporal there is generally a low distribution of lighting strikes resolution dynamics using LPATS and meteorological and thunderstorm days. radars could be undertaken. Field observations across The fire scar analysis showed that the differential fire scars using a network of automatic weather stations heating between unburned and burned areas could be a and profiling equipment could also elucidate mesoscale possible trigger mechanism for convection, cloud cover, circulations. and lightning activity. We suggest that buoyant warm air rising as a result of intense surface heating and a 6. Conclusions strong heating contrast between two surfaces may lead to the generation of a mesoscale circulation. However, This preliminary study attempted to identify the ef- the intensity of these convective processes is likely to fects of vegetation and elevation on lightning density by depend on the area of the surface: therefore, a larger selecting two contrasting climatic regions. GIS allowed effect on the atmosphere is expected over larger fire for data integration between the vegetation, elevation, scars because it will provide sufficient horizontal gradi- fire scar, and lighting strike datasets as well as quanti- ents for flow to occur. In addition, as air moves across fying the relationship between them. In the northern larger-sized scars the integrated heat input over the region there was a distinct increase in lightning strike larger fetch will be much greater, which may then trig- density from woodland to shrub to grassland. We sug- ger the uplift of buoyant air and enhance convection gest that the cause of this may be due to greater surface and lighting. A small fire scar will show an increase in heating of grasslands than the other vegetation types,

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1172 JOURNAL OF CLIMATE VOLUME 20 which would then allow for a heating differential to Bureau of Meteorology, 1998: Climate of the Northern Territory. arise between the two surface types and may act as a National Capital Printing, 38 pp. mechanism. In the southern region, vegetation was Christian, H. J., 1999: Optical detection of lightning from space. Proc. 11th Int. Conf. on Atmospheric Electricity, Guntersville, shown to have no significant effect on lightning strikes AL, International Commission on Atmospheric Electricity, because it is generally sparse and moisture is limited. 715–718. However, elevation in the southern region, which ——, and Coauthors, 2003: Global frequency and distribution of ranges from 0 to 1200 m, was found to play an impor- lightning as observed from space by the optical transient tant role in initiating convection via orographic uplift, detector. J. Geophys. Res., 108, 4005, doi:10.1029/ 2002JD002347. especially for elevations above 800 m. The influence of Cooray, V., Ed., 2003: The Lightning Flash. The Institution of elevation in the northern region did not show a rela- Electrical Engineers, 574 pp. tionship with lightning density, which may be due to the Department of the Environment and Heritage, cited 2004: Na- small range of elevation (0–400 m). tional Vegetation Information System. Australian Govern- The study of lightning density over fire scar affected ment. [Available online at http://www.deh.gov.au/erin/nvis/ index.html.] areas in the NT was largely based on the modeling Department of Land Information, cited 2004: Fire scar mapping. research conducted by Knowles (1993), who hypoth- Government of . [Available online at esized that the differential heating caused by the fire http://www.rss.dola.wa.gov.au/newsite/apps/firescarmap. scar and the adjacent vegetative area may produce a html.] mesoscale circulation that could generate lightning Dissing, D., and D. L. Verbyla, 2003: Spatial patterns of lightning strikes in interior Alaska and their relations to elevation and strikes. Our results suggest that the spatial distribution vegetation. Can. J. For. Res., 33, 770–782. of lightning strikes over fire scars may be induced by Görgen, K., A. H. Lynch, A. G. Marshall, and J. Beringer, 2006: mesoscale circulations or buoyancy due to intense heat- The impact of abrupt land cover changes by savanna fire on ing. The distribution of lightning strikes over fire scars northern Australian climate. J. Geophys. Res., 111, D19106, was greater over larger fire scars. This work illustrates doi:10.1029/2005JD006860. the importance of fire in influencing spatial heteroge- Hutley, L. B., A. P. O’Grady, and D. Eamus, 2001: Monsoonal influences on evapotranspiration of savanna vegetation of neities in surface properties and heating. Given the northern Australia. Oecologia, 126, 434–443. enormous extent of annual burning in northern Austra- Knowles, J., 1993: The influence of forest fire induced albedo lia it has been suggested that this could feed back to differences on the generation of mesoscale circulations. M.S. effect regional climate through changes in monsoon thesis, Dept. of Atmospheric Science, Colorado State Uni- strength (Beringer et al. 2003; Görgen et al. 2006). versity, 84 pp. Laird, N. F., D. A. R. Kristovich, and J. E. Walsh, 2003: Idealized model simulations examining the mesoscale structure of win- Acknowledgments. We thank Jim Arthur, Rob Por- ter lake-effect circulations. Mon. Wea. Rev., 131, 206–222. teous, Andrew Edwards, and the Bureau of Meteorol- Latham, D. J., 1991: Lightning flashes from a prescribed fire- ogy for providing data and support. This research was induced cloud. J. Geophys. Res., 96, 17 151–17 157. supported through Australian Research Council Lericos, T. P., H. E. Fuelberg, A. I. Watson, and R. L. Holle, 2002: (ARC) Grant (DP0344744). Warm season lightning distributions over the Florida penin- sula as related to synoptic patterns. Wea. Forecasting, 17, 83–99. REFERENCES López, R. E., and R. L. Holle, 1986: Diurnal and spatial variability of lightning activity in northeastern Colorado and Central Florida during the summer. Mon. Wea. Rev., 114, 1288–1312. Beringer, J., and N. J. Tapper, 2000: The influence of subtropical Mushtak, V., E. Williams, and D. Boccippio, 2003: Latitudinal cold fronts on the surface energy balance of a semi-arid site. variation of cloud base height and lightning parameters in the J. Arid Environ., 44, 437–450. Tropics. Proc. 12th Int. Conf. on Atmospheric Electricity, ——, and ——, 2002: Surface energy exchanges and interactions Versailles, France, International Commission on Atmo- with thunderstorms during the Maritime Continent Thunder- spheric Electricity, 9–13. storm Experiment (MCTEX). J. Geophys. Res., 107, 4552, Orville, R. E., 1994: Cloud-to-ground lightning flash characteris- doi:10.1029/2001JD001431. tics in the contiguous United States: 1989–1991. J. Geophys. ——, D. Packham, and N. J. Tapper, 1995: Biomass burning and Res., 99, 10 833–10 841. resulting emissions in the Northern Territory, Australia. Int. ——, R. A. Weisman, R. B. Pyle, R. W. Henderson, and R. E. J. Wildland Fire, 5, 229–235. Orville Jr., 1987: Cloud-to-ground lightning flash character- ——, L. B. Hutley, N. J. Tapper, A. Coutts, A. Kerley, and A. P. istics from June 1984 through May 1985. J. Geophys. Res., 92, O’Grady, 2003: Fire impacts on surface heat, moisture and 5640–5644. carbon fluxes from a tropical savanna in north Australia. Int. ——, G. R. Huffines, W. R. Burrows, R. L. Holle, and K. L. Cum- J. Wildland Fire, 12, 333–340. mins, 2002: The North American Lightning Detection Net- Boccippio, D. J., S. J. Goodman, and S. Heckman, 2000: Regional work (NALDN)—First results: 1998–2000. Mon. Wea. Rev., differences in tropical lightning distributions. J. Appl. Me- 130, 2098–2109. teor., 39, 2231–2248. Pielke, R. A., and R. Avissar, 1990: Influence of landscape struc-

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC 1APRIL 2007 K ILINC AND BERINGER 1173

ture on local and regional climate. Landscape Ecol., 4, 133– Stolzenburg, M., 1994: Observations of high ground flash densities 155. of positive lightning in summertime thunderstorms. Mon. ——, and P. L. Vidale, 1995: The boreal forest and the polar front. Wea. Rev., 122, 1740–1750. J. Geophys. Res., 100, 25 755–25 758. Sturman, A. P., and N. J. Tapper, 1996: Weather and Climate in Pinto, O., Jr., R. B. B. Gin, I. R. C. A. Pinto, O. Mendes, J. H. Australia and New Zealand. Oxford University Press, 476 pp. Diniz, and A. M. Carvalho, 1996: Cloud-to-ground lightning Tapper, N. J., 1991: Evidence for a mesoscale thermal circulation flash characteristics in southeastern Brazil for the 1991–1993 over dry salt lakes. Palaeogeogr. Palaeoclimatol. Palaeoecol., summer season. J. Geophys. Res., 101, 29 627–29 635. 84, 259–269. Rorig, M. L., and S. A. Ferguson, 2002: The 2000 fire season: Lightning-caused fires. J. Appl. Meteor., 41, 786–791. Toracinta, E. R., D. J. Cecil, E. J. Zipser, and S. W. Nesbitt, 2002: Russell-Smith, J., A. Grant, R. Thackway, T. Rosling, and R. Radar, passive microwave, and lightning characteristics of Smith, 2000: Fire management and savanna landscapes in precipitating systems in the Tropics. Mon. Wea. Rev., 130, Northern Australia. Fire and Sustainable Agricultural and 802–825. Forestry Development in Eastern Indonesia and Northern Ushio, T., S. J. Heckman, D. J. Boccippio, and H. J. Christian, Australia, J. Russell-Smith et al., Eds., ACIAR Proceedings, 2001: A survey of thunderstorm flash rates compared to Vol. 91, Australian Centre for International Agricultural Re- cloud top height using TRMM satellite data. J. Geophys. Res., search, 21–30. 106, 24 089–24 095. Segal, M., R. Avissar, M. C. McCumber, and R. A. Pielke, 1988: Vonnegut, B., D. J. Latham, C. B. Moore, and S. J. Hunyady, Evaluation of vegetation effects on the generation and modi- 1995: An explanation for anomalous lightning from forest fire fication of mesoscale circulations. J. Atmos. Sci., 45, 2269– clouds. J. Geophys. Res., 100, 5037–5050. 2292. Williams, E. R., and S. Stanfill, 2002: The physical origin of the ——, J. R. Garratt, R. A. Pielke, and Z. Ye, 1991: Scaling and land-ocean contrast in lightning activity. Appl. Phys., 3, 1277– numerical model evaluation of snow-cover effects on the gen- 1292. eration and modification of daytime mesoscale circulations. J. Atmos. Sci., 48, 1025–1042. ——, and Coauthors, 2002: Contrasting regimes over the Amazon: Simpson, J. E., 1994: Sea Breeze and Local Winds. Cambridge Implications for cloud electrification. J. Geophys. Res., 107, University Press, 234 pp. 8082, doi:10.1029/2001JD000380. Smith, R. K., and J. A. Noonan, 1998: On the generation of low- Williams, R. J., B. A. Myers, W. J. Muller, G. A. Duff, and D. level mesoscale convergence lines over northeastern Austra- Eamus, 1997: Leaf phenology of woody species in a north lia. Mon. Wea. Rev., 126, 167–185. Australian tropical savanna. Ecology, 78, 2542–2558.

Unauthenticated | Downloaded 09/25/21 12:47 AM UTC