Spatial distribution and frequency of lightning activity and lightning flash density maps for Australia

Authors

  • Yuriy Kuleshov,

    1. National Climate Centre, Australian Bureau of Meteorology, Melbourne, Victoria, Australia
    Search for more papers by this author
  • David Mackerras,

    1. School of Information Technology and Electrical Engineering, University of Queensland, Brisbane, Queensland, Australia
    2. Also at Lightning and Transient Protection Pty Ltd, Brisbane, Queensland, Australia.
    Search for more papers by this author
  • Mat Darveniza

    1. School of Information Technology and Electrical Engineering, University of Queensland, Brisbane, Queensland, Australia
    2. Also at Lightning and Transient Protection Pty Ltd, Brisbane, Queensland, Australia.
    Search for more papers by this author

Abstract

[1] The spatial distribution and frequency of lightning activity in Australia have been analyzed using lightning data obtained by ground-based lightning detection instruments denoted CIGRE-500 and CGR3 and by NASA satellite-based instruments denoted OTD and LIS. The geographical distribution of lightning incidence is described by a map of total lightning flash density, Nt (i.e., cloud-to-ground and intracloud flashes). A high level of lightning activity, Nt > 10 km−2yr−1, is observed in the northern parts of Australia, and a decrease in total flash density occurs southward to Nt < 5 km−2yr−1 in the central and southern parts of Australia. The peak lightning occurrence is in the northwestern part of the Australian continent with Nt values up to about 35 km−2yr−1 centered around 16°S 126°E. A reduction in Nt by a factor of about 10 for a change in latitude from 10°S to 40°S was found, which is in agreement with the earlier studies. The data from all the sources were used to estimate the cloud flash-to-ground flash ratio, Z, which at the studied localities was found to be in a range of values from 0.75 to 7.7. We concluded that for the range of latitude over Australia the most representative value of Z is about 2 ± 30%, and it is relatively independent of latitude. We used this to develop a map of average annual lightning ground flash density, Ng, the first for Australia. Ng varies from over 6 km−2yr−1 in the northern parts of Australia to about 1 km−2yr−1 and below in the southern parts.

1. Introduction

[2] The climatological distribution of thunderstorm days prepared by the World Meteorological Organization (WMO) is the most readily available source of information about the worldwide geographical distribution of thunderstorms [World Meteorological Organization (WMO), 1953]. It does not provide reliable data for many locations, however, particularly those locations where thunderstorm data have not been recorded systematically. Nevertheless, since the 1950s, there have been advances in observational practice, which have made it possible to improve the accuracy and the resolution of the WMO thunderstorm climatology. Recently, using the Australian thunderday records, the thunderstorm distribution across the country was analyzed in detail and an updated average annual thunderday map of Australia was prepared at the National Climate Centre, the Australian Bureau of Meteorology [Kuleshov et al., 2002]. This map was included in the Australian Lightning Protection Standard AS/NZS 1768(Int)-2003 [Lightning Protection, 2003]. Thunderday records are commonly used as proxy information for describing lightning activity. In the absence of better information, the lightning ground flash density, Ng, may be estimated from thunderdays, Td, using an equation of the form Ng = aTdb, in which a and b are empirically derived constants that depend on the meteorological conditions at a given location. The earliest estimates of the equation for Australia were by Mackerras [1978], who derived values of a = 0.01 and b = 1.4. However, lightning flash densities derived from such empirical relationships to thunderday records are less accurate than those derived from direct measurements of Ng.

[3] To obtain instrumental records of lightning incidence, many lightning flash counters (LFC) have been installed at meteorological and other sites worldwide. On the basis of earlier work in the UK by Pierce [1956], two types of LFC were developed that have been used extensively around the world. These are the 500 Hz CIGRE LFC (Comité Internationale des Grands Réseaux Electriques), (that is, International Committee on Large Electric Systems) developed at the University of Queensland, Australia, and in Pretoria, South Africa [Anderson et al., 1979]. The 500 Hz LFC initially referred to as the RBQ 500 Hz LFC [Barham, 1965; Barham and Mackerras, 1972] was developed and tested in SE Queensland, Australia. The CIGRE 10 kHz counter was developed and tested in South Africa by Anderson and his colleagues in Pretoria. It used a 10 kHz filter instead of the 500 Hz filter and was initially referred to as RSA 10 kHz [Anderson et al., 1979]. CIGRE counters were intended primarily to register ground flashes, but also respond to some cloud flashes. With the 10 kHz counter, less than 10% of the total registrations are caused by cloud flashes.

[4] Another type of counter, primarily developed for research purposes, is the CGR3 (Cloud-to-Ground Ratio #3) instrument [Mackerras, 1985]. The CGR3 counter has separate registers for negative ground flashes, positive ground flashes, and cloud flashes. CGR3 counters were installed in 11 countries and were used for the derivation of total, intracloud and ground flash density estimates around the world [Mackerras and Darveniza, 1994; Mackerras et al., 1998].

[5] In Australia, a network of 39 CIGRE-500 LFCs has been operated by the Australian Bureau of Meteorology (34 LFCs) and by Electric Power Authorities (5 LFCs) since 1973. LFCs have proved to be useful and reliable sources of lightning data. However, their sparse spatial distribution makes it difficult to prepare a map of lightning distribution without supplementary information. Recently, remotely sensed lightning data gathered by the National Aeronautics and Space Administration (NASA) satellite-based instruments became available. Satellite observations provide better spatial coverage of lightning which is, by its nature, widely distributed geographically. The remotely sensed data are a valuable source of information for the areas with little or no local observation data (i.e., LFC or thunderday observations). Using five years of the satellite data, a world map of average annual distribution of total lightning activity was constructed [Christian et al., 2003] and it was demonstrated that the spatial distribution of flash density is in broad general agreement with the world thunderday map [WMO, 1953].

[6] The availability of long-term records from LFCs and thunderday records, as well as recently acquired NASA satellite-based lightning data, allowed us to analyze in detail geographical distribution of thunderstorm and lightning activity for the Australian continent and to derive conclusions about variations in lightning flash density. The lightning flash density, defined as the number of flashes of a specific type occurring on or over unit area in unit time, is commonly used to describe lightning activity. We denote the lightning flash densities as Ng, Nc and Nt for cloud-to-ground (CG, or simply ground), intracloud (IC, or simply cloud) and total (CG + IC) flashes, respectively, and express them as a number of flashes per square kilometer per year (km−2 yr−1). In this paper, we present spatial and latitudinal variations of total and ground flash densities around Australia. We also examine the cloud flash-to-ground flash ratio (denoted by Z), where

equation image

[7] One measure of lightning total flash density, NtSAT, was derived for the Australian continent from satellite-based data (the v1.0 LIS/OTD 0.5° × 0.5° gridded climatology) for the observation period of the space instruments from 1995 to 2002. A second measure of lightning total flash density, NtLFC, and ground flash density, NgLFC, were calculated from the data obtained with the ground-based lightning flash counters (CIGRE-500 and CGR3) at 39 sites. The values of Ng were also estimated from thunderdays, Td, using an empirically derived equation. The values of NgLFC and Td were long-term, typically based on more than 10 years of data. On the basis of a polynomial expression for the latitudinal variation of the ratio NtLFC/NtSAT as a function of latitude, λ, we modified the NtSAT values to optimize the agreement between the LFC and the satellite data, and produced a total flash density map that agreed well with the LFC data. From this map and an estimate of the cloud-to-ground flash ratio (Z), we produced the first ground flash density map for Australia.

2. Sources of Information

2.1. Instruments of Lightning Registration

[8] Data from two types of ground-based LFCs (CIGRE-500 at 39 sites and CGR3 at 2 sites) and two NASA satellite-based lightning detectors (OTD and LIS) were used in this study. The LFC site locations are shown in Figure 1. The CGR3 instruments were at Brisbane and Darwin, and were used to obtain estimates of lightning flash occurrence that were independent of the results from the CIGRE-500 flash counters.

Figure 1.

Geographical distribution of the CIGRE-500 counters used in the analysis.

2.1.1. Australian Bureau of Meteorology Lightning Flash Counter Network

[9] The CIGRE-500 counter is the instrument used by the Australian Bureau of Meteorology for its network of LFCs. It was designed primarily to register ground flashes [Barham, 1965; Prentice, 1972]. However, the electric field changes caused by cloud flashes and positive ground flashes often exhibit positive steps (K changes) and truncated negative ramps that may cause the instrument to register such an event [Bunn, 1968]. Therefore the determination of Nt and Ng values from annual counter registrations, K, require estimates of the effective ranges of the instrument for ground and cloud flashes and estimates of Z. This study used effective ranges of the counter of Lg = 30 km for ground flashes [Prentice and Mackerras, 1969] and Lc = 18 km for cloud flashes (this is a recent unpublished corrected estimate by the authors replacing the earlier estimate of 20 km). Data from the CIGRE-500 counters were used to obtain estimates of total flash density, Nt, and ground flash density, Ng, at 39 Australian sites, with length of records ranging from 5 to 23 years and covering the time period from 1981 to 2003. The 39 sites were separated into 17 sites with longer periods of reliable records, and 22 sites with shorter periods of records. This data set, described in detail in section 3.1, is given in Table 1.

Table 1. Values of the Average Annual Ground Flash Density as Derived From the CIGRE-500 Counter Registrations, NgLFC, and Thunderdays, Td, Total Flash Density as Derived From the Satellite Data, NtSAT, and the Computed Cloud Flash-to-Ground Flash Ratio, Z
LocationLatitude, °SLongitude, °EYears of DataNgLFC, km−2 yr−1Td, day yr−1NtSAT, km−2yr−1Z
Darwin12.3131.0216.679720.072.01
Gove12.3136.852.35454.240.80
Centre Island15.7136.8101.80607.623.24
Kununurra15.8128.7106.006020.592.43
Mareeba17.1145.5141.56234.982.19
Sth Johnstone17.5145.9141.19233.071.57
Townsville19.1146.5130.69195.166.45
Tennant Creek19.6134.2211.47343.681.50
Port Hedland20.4118.6231.80233.150.75
Mt Isa20.6139.5202.30375.471.38
Hughenden20.8144.2111.10245.974.44
Blackwater23.6149.1100.76216.237.24
Emerald23.6148.1141.66205.442.29
Callide Dam24.2150.6142.19197.272.33
Meekatharra26.6118.5200.92222.912.17
Brisbane27.5153.0151.43258.294.82
Mt Gravatt27.6153.1161.24258.305.67
Geraldton28.8114.7220.23111.997.69
Lismore28.8152.3171.75267.683.55
Grafton29.8153.0152.31327.412.20
Three Springs30.0116.0140.6191.321.15
Coffs Harbour30.3153.1212.89366.871.38
Moora30.6116.0170.2871.283.61
Bowraville30.7152.8132.15397.032.26
Kalgoorlie30.8121.5230.94214.093.35
Cobar31.5145.8191.03274.823.68
Perth31.9116.0220.39151.392.56
Taree31.9152.571.43385.472.82
Ceduna32.1133.7190.36140.781.18
Port Augusta32.5137.8171.05101.05N/A
Waikerie34.2140.0130.43131.041.41
Albany34.9117.8210.48100.860.81
Nowra35.0150.5130.56237.4712.34
Mulwala36.0146.0121.37172.841.08
Cooma North36.2149.1140.97256.545.72
Whitlands36.9146.3211.27173.121.45
Ballarat37.5143.8220.54131.832.39
Melbourne37.7144.8220.87133.352.85
Mt Burnett38145.5210.70143.233.59

2.1.2. Satellite-Based Instruments

[10] Data from two NASA satellite-based instruments, the MicroLab-1 Optical Transient Detector (OTD) [Christian et al., 1996] and the Tropical Rainfall Measuring Mission (TRMM) Lightning Imaging Sensor (LIS) [Christian et al., 1999], were used in this study. These lightning detectors were designed by the Global Hydrology Resource Centre Lightning Team and were manufactured at the Marshall Space Flight Centre in Alabama, USA. A combined OTD and LIS climatology (a 0.5° × 0.5° gridded composite of total lightning, expressed as a flash density (km−2 yr−1)) was used to obtain the NtSAT total flash density values over the Australian region. This grid resolution corresponds to approximately 60 km spatial averaging, which is similar to the spatial averaging of the CIGRE-500 counter for ground flashes. The combined data archive for OTD and LIS covers the period of observations from 1995 to 2002.

2.1.3. CGR3 Lightning Flash Counter

[11] The CGR3 (Cloud-to-Ground Ratio #3) counter is a modification of the original forms of the instrument (CGR1 and CGR2) developed at the University of Queensland (Brisbane, Australia). The CGR3 instrument was used for lightning research for a number of years and it is described in detail by Mackerras and Darveniza [1992, 1994] and by Baral and Mackerras [1992, 1993]. The CGR3 counter has separate registers for negative ground flashes (NGF), positive ground flashes (PGF), and cloud flashes (CF). The effective ranges are estimated as 12 km for CF, 14 km for NGF, and 16 km for PGF. In Australia, these counters have been installed in Brisbane (27.5°S, 153.0°E) and Darwin (12.3°S, 131.0°E), and were used for estimating ground, cloud, and total flash densities, and Z at those localities. In determining the effective ranges, the position of ground flashes was considered to be the strike point on earth, and the position of cloud flashes was considered to be the point on earth vertically below the middle of the discharge channel.

[12] Long-term measurements using CGR3 instruments with supporting electric field change records in Brisbane have supplied supplementary information concerning lightning occurrence that was used in this study as a check on the validity of other methods. For the 9 years July 1995 to June 2004 (for reporting purposes the years are from July to June, covering complete thunderstorm seasons), the following results were obtained.

equation image

(Bracketed numbers are minimum and maximum annual values.)

[13] It will be noticed that the mean annual value of Z (2.0) differs from the ratio of long-term flash density values, Nc/Ng = 1.5. The reason is that Z tends to be high during years of low activity and low during years of high activity. In 3 years with mean Nt = 2.7 km−2 yr−1, the mean Z was 3.2, whereas in 3 years with mean Nt = 9.5 km−2 yr−1, the mean Z was 1.2. These observations confirm the large interannual variability in lightning occurrence characteristics that has been observed elsewhere. They also show the difficulty in arriving at reliable long-term mean values, and show that there are two ways of calculating Z, depending on the application. The uncertainty in the flash density values is about ±20%, and in Z values about ±30%.

[14] For Darwin, Mackerras and Darveniza [1994] using CGR3 instruments, reported Nt = 10 km−2 yr−1 and Z = 1.4, from which Ng = 4.2 km−2 yr−1, for a 3-year period between 1987 and 1991. The mean ratio of positive to negative ground flashes was 0.02. For the period August 2003 to June 2004, a CGR4 instrument [Farnsworth et al., 2004] has been in use at Darwin Airport. The CGR4 has the same response to lightning as the CGR3, and differs only in electronic circuit details. The following results were obtained: Nt = 10.5 km−2 yr−1, Ng = 5.6 km−2 yr−1, Nc = 4.9 km−2 yr−1, Z = 0.9, and proportion of positive ground flashes = 0.04, in reasonable agreement with the earlier results, and with the long-term (21 yr) CIGRE-500 value: NgLFC = 6.7 km−2 yr−1 (Table 1).

2.2. Method of Thunderstorm Registration

[15] Thunderstorm occurrence at a particular location is traditionally expressed in terms of thunderdays. A thunderday, Td, is defined as [Huschke, 1959, p. 582] “an observational day (any 24-hour period selected as the basis for climatological or hydrological observations) during which thunder is heard at the station. Precipitation need not occur.” The 24-hour period selected at the Australian Bureau of Meteorology for the phenomenon registration is midnight to midnight (local standard time). The requirement that thunder should actually be heard limits the area covered by each observing point, under favorable listening conditions, to a circular area with a radius of some 20 km [WMO, 1953]. The thunderday records for this study were obtained from the National Climate Centre archive.

2.3. Methods of Estimating Lightning Flash Density

2.3.1. Cloud Flash-to-Ground Flash Ratio, Z

[16] As stated in section 2.3.2 the cloud flash-to-ground flash ratio, Z, is required when estimating total and ground flash density from CIGRE-500 LFC annual registrations. Earlier estimates of the variation of Z with latitude by Pierce [1970] and Prentice and Mackerras [1977] suggested that Z was high (about 6 to 9) in the tropics, was about 3 to 4 in the subtropics and temperate regions, and was about 1 to 1.5 at high latitudes. Earlier observations [Williams et al., 1996, Figure 1] also indicated an increase in Ng away from the equatorial zone, and it was suggested [Williams et al., 2005; Mushtak et al., 2005] that this increase in Ng is an important contributor to the latitudinal decline in the IC/CG ratio, Z, away from the equator as it was reported by Pierce [1970] and Prentice and Mackerras [1977]. However, Mackerras and Darveniza [1994], using a worldwide network of 14 CGR3 counters in 11 countries, demonstrated that Z is largely independent of latitude, with a mean Z value of 1.9. Boccippio et al. [2001] analyzed variation of the Z values derived from four years of OTD and National Lightning Detection Network (NLDN) data over continental United States. The value of Z over the studied area was found to be 2.64–2.94, with anomalies as low as 1 and as high as 8–9; however little evidence was found to support a systematic latitudinal variation. In a more recent study, 5 years of satellite data were used and the Nt values as derived from the OTD data, together with Ng values as derived from NLDN data, were presented for the six United States stations located between approximately 29°N and 38°N [Christian et al., 2003, Table 1]. Computing the Z values from these Nt and Ng estimates, one can find that the mean value of Z is 3.6, and ranges from 1.5 to 5.

[17] The continental United States studies confirmed the conclusion of Mackerras and Darveniza [1994] that Z is largely independent of latitude, at least between 0 and 40°. However, the estimates of mean Z value are different in these three studies: it is 2.2 for latitudes between 20° and 40° (data for the sites both in Northern and Southern Hemispheres) in the work by Mackerras and Darveniza [1994], 2.64–2.94 for 20–40°N in the work by Boccippio et al. [2001], and 3.6 for 20–40°N as it was computed from the data in the work by Christian et al. [2003]. It appears that local variability (spatial and temporal) has greater impact on variability of the Z values than latitudinal variation. For the latitude range 10° to 40°, we selected the Z values from Table 2 of Mackerras and Darveniza [1994] for six sites with at least 1 year of records and found that the mean value of Z was 2.1. It was demonstrated in 2.1.3 that long-term measurements in Australia (1995 to 2004) using CGR3 instruments gave an average annual value of Z = 2 ± 30%. We use this value as the most appropriate for Australian localities in the study period. The use of Z = 2 also gives a good match between the Australian data and CIGRE 10-kHz LFC results for South Africa [Anderson and Eriksson, 1980; Kuleshov and Jayaratne, 2004] which gives us additional confidence in the selected Z value.

[18] We estimated values of Z for 39 Australian sites, using Nt values as derived from OTD and LIS combined climatology and estimates of Ng values as derived from the CIGRE-500 counter registrations, and applying the equation

equation image

[19] The computed values of Z are presented in Table 1, column 8. The range of the Z values is from 0.75 to 7.7 with an average of 2.8, without indication of a systematic latitudinal variation. On the basis of the results of previous studies and these findings, we conclude that for the range of latitude over Australia the most representative long-term value of Z is about 2 ± 30%, and it is largely independent of latitude.

2.3.2. Lightning Flash Density Estimates From CIGRE-500 Registrations

[20] The total annual registrations, K, of CIGRE-500 LFCs are a combination of registrations caused by ground flashes and by cloud flashes. The procedure for calculating flash densities from annual registrations is given in Appendix A.

[21] During 1997 to 1999, all Australian LFC sites were checked for the state of the installation and calibration of the LFCs so that a site correction factor could be obtained. There are several reasons why annual registrations differ from those that would be obtained from an ideal installation. These include aerials partly shielded by nearby elevated objects, incorrect instrument adjustment, flat batteries, poor aerial insulation resistance under wet conditions, and spurious counts caused by a nearby interference source. With the exception of the last, these factors usually cause the annual recorded counts to be below the value that would be given by a flawless installation. Using the site correction factor, F (range 0.8 to 1.0 for 95% of sites), the counter registrations, K, were first adjusted accordingly, and then Ng and Nt were estimated from the counter registrations by using equations (A2) and (A6) respectively. The uncertainty in flash density values resulting from uncertainty in the effective range values and the Z value is about ±30%.

[22] Despite the correction procedure described above, annual registrations at individual sites may be greater or less than the value that would have been obtained by a flawless installation, with factors such as aerial shielding and flat batteries making it more likely that the registrations were below the true value.

2.3.3. Lightning Flash Density Estimates From CGR3 Registrations

[23] The method of estimating lightning flash density from CGR3 counter registrations was described by Mackerras and Darveniza [1994] and by Baral and Mackerras [1992, 1993]. Using the separate registers for negative (NGF) and positive (PGF) ground flashes, and cloud flashes (CF), estimates of lightning flash density were calculated from the number of registrations, Kng, Kpg, and Kcf corresponding to NGF, PGF, and CF, respectively. The effective ranges of the counter are: Lng = 14 km, Lpg = 16 km, and Lcf = 12 km, corresponding to NGF, PGF, and CF, respectively. Then, the average annual flash densities are as follows.

equation image

Thus

equation image

and Z is calculated by using equation (1).

2.3.4. Ng Estimates From Td

[24] Mackerras [1978] used Australian data to produce an empirical formula for estimating lightning ground flash density, Ng, from thunderdays, Td, of the form Ng = a Tdb with a = 0.01 and b = 1.4. This study was based on results from 26 sites for the period 1965 to 1977 and the Australian Bureau of Meteorology thunderday map based on data from 1954 to 1963. Anderson and Eriksson [1980], based on 120 observations over two years in South Africa, derived values of a = 0.023 and b = 1.3. The corresponding equation has since become known as “Eriksson's Formula.” A subsequent study using 62 stations over a longer period of five years from 1976 to 1980 yielded the values a = 0.04 and b = 1.25 [Anderson et al., 1984]. The equation using these two values of the constants is generally known as the “CIGRE Formula.” Both the CIGRE and Eriksson's equations are used in the literature, although they were both derived at the same location. However, there are very few systematic studies enabling one to compare their results with those from other parts of the world. Recently, using long-term (up to 22 years) LFC registrations and thunderday observations for 17 Australian localities, Kuleshov and Jayaratne [2004] updated Mackerras' empirical formula and compared the results with Eriksson's and CIGRE formulas. It was concluded that the empirical formula,

equation image

gives the best estimate of Ng in Australia. Consequently, equation (5) was used in this study to derive estimates of lightning ground flash density, Ng, from annual thunderdays, Td.

3. Results and Discussion

3.1. Summary of Lightning Flash Counter Measurements

[25] The CIGRE-500 counter records are available for the period 1981–2003. From this data set, 39 sites across Australia (Figure 1) which had a period of observation of at least 5 years were selected for this study. A summary of the counter measurements covering the range of latitudes over Australia from 12°S to 38°S is given in Table 1 (entries are arranged in order of increasing magnitude of latitude, λ). The location, latitude and longitude of each site, and the number of years of data used in our analysis are in columns 1–4, respectively. Values of the average annual ground flash density, NgLFC (column 5), were derived from the annual counter registrations, K, using equation (A2) after applying the site correction factors, F. Values of the average annual number of thunderdays, Td (column 6), were obtained from the annual Td recorded at each site, or estimated from the average annual thunderday map (Figure 2) for sites where Td records were not available from the archive.

Figure 2.

Map of average annual thunderdays for Australia (1990–1999) [Kuleshov et al., 2002].

3.2. Summary of Satellite-Based Measurements

[26] The v1.0 gridded satellite lightning data (LIS and OTD 0.5° high-resolution full climatology) produced by the NASA LIS/OTD Science Team were used in this study. Best available calibrations for instrument detection efficiency had been applied in the v1.0 product; the procedure is described by Boccippio et al. [2002] and Christian et al. [2003]. Observations in the LIS/OTD v1.0 reanalysis were corrected by the LIS/OTD Science Team using estimated flash detection efficiency, applied as a function of sensor, local hour, date of mission, and (for the OTD) geographic location. For the entire data set, these corrections correspond to average flash detection efficiencies of 47% (OTD) and 82% (LIS). The adjustments were derived from a combination of laboratory calibration, ground validation, and cross normalization between the two instruments. Additional uncertainty due to local variations may arise from undersampling of a given grid location, and its relative impact may increase as the actual local climatological flash rate decreases. Thus the minimum uncertainty in the gridded data was estimated by the NASA LIS/OTD Science Team as ±10%.

[27] The spatial distribution of average annual total flash density, NtSAT, over Australia was derived from OTD and LIS combined climatology (1995–2002) after applying an adjustment described below, and is presented in Figure 3. As part of the mapping process, a 3 × 3 (box) binomial smoother was applied to the high-resolution full climatology data (0.5° × 0.5° gridded composite of total lightning flashes). The resultant grid was then resampled to 0.025° × 0.025° using cubic convolution [Richards, 1993].

Figure 3.

Map of average annual total lightning flash density over Australia as derived from OTD and LIS combined climatology (1995–2002), adjusted to conform to the LFC data.

[28] For each of the 39 Australian LFC stations, where a total flash density, NtLFC, had been calculated from the annual registrations, the average annual total flash density, NtSAT, was also calculated from the satellite data (Table 1, column 7) by interpolation between adjacent NtSAT grid values. We examined the latitudinal variation of the ratio

equation image

[29] We grouped the 39 values into 3 latitude bands, 10–20°S, 20–30°S, and 30–40°S, arbitrarily weighted the 17 more reliable longer-term sites (mean observation period 19 yr) with twice the weight of the 22 shorter-term sites (mean period 14 yr), and calculated the weighted mean values of RTL-TS in each band, Table 2.

Table 2. Variation of Mean Weighted Values of RTL-TS With Latitude
Latitude Range, °SMean Weighted RTL-TSNumber of Sites UsedAnnotations
10 to 200.748actual range 12.3 to 19.6°S
20.1 to 300.6113 
30.1 to 401.1518actual range 30 to 38°S
10 to 400.8839the complete data set

[30] Table 2 indicates good overall agreement between the two methods for the whole Australian continent, within about ±12%, which is well within the overlap of uncertainties for the two methods. However, the satellite data indicated higher Nt values than the LFC data for latitudes from 10°S to 30°S, and lower values from 30°S to 40°S.

[31] We propose that a possible reason why the satellite data for latitudes from 30°S to 40°S are below the values from 10°S to 30°S arises from the combination of data from two instruments launched on different platforms. OTD was launched aboard the MicroLab-1 satellite into a near polar orbit at an inclination of 70° with respect to the equator. Therefore the OTD data cover the whole range of Australian latitudes from 10°S to 40°S. The LIS was launched aboard the TRMM satellite. The orbit of the TRMM has an inclination of 35°. As a result, the LIS instrument can observe lightning activity only between 35°S and 35°N latitudes. This difference in spatial coverage of the instruments leads to potential undersampling for data beyond 35°S and 35°N latitudes in the combined OTD and LIS archive.

[32] As we wished to adjust the NtSAT values to agree as well as possible with the LFC values, we derived an empirical polynomial to give the best fit to the variation of RTL-TS with λ, as follows.

equation image

with λ in magnitude of degrees. A satisfactory fit, within 1%, was found with A = 1.2339, B = 2.8108E-3, C = 1.3661E-5, and N = 3.5.

[33] We adjusted all the NtSAT values by multiplying each NtSAT value by RLNP(λ) and used this new set to plot the “adjusted” total flash density map for Australia, see Figure 3.

[34] Applying the assumption that the Z value over the range of latitude for the Australian continent is uniform (Z = 2), we derived the Ng values from the Nt values (adjusted satellite data) by using the equation

equation image

[35] These derived Ng values allowed us to prepare the first map of lightning ground flash density for Australia (Figure 4). The map is contoured in units of flash density (km−2 yr−1). We recommend the map as a guide for estimating Ng values for applications, e.g., lightning protection system design.

Figure 4.

Map of average annual lightning ground flash density for Australia derived from the Nt map in Figure 3 using Z = 2 (determined from satellite and LFC data, see text).

[36] An alternative method would be to derive a polynomial to adjust the NASA values to optimize agreement with the LFC ground flash density values, and to plot the ground flash density map directly from the adjusted total flash density map. This method would be equally valid, and is mathematically equivalent to the method we have adopted. The overall uncertainty in the ground flash density map values is estimated to be about ±30%.

[37] Having produced the ground flash density map directly from the total flash density map based on adjusted NASA satellite data, it was found that discrepancies existed in a few places between the Ng values indicated by the map and the Ng values based on long-term LFC results, or on evidence from lightning damage rates on electric power systems and industrial installations. Consequently, the contours defining the boundaries between adjacent Ng regions were slightly shifted to minimize the discrepancies. This was done in the vicinity of Gove (12.3°S, 136.8°E), Mt. Isa (20.6°S, 139.5°E), Mulwala (36.0°S, 146.0°E), and Three Springs (30.0°S, 116.0°E).

3.3. Variation of Thunderstorm and Lightning Distribution

3.3.1. Geographic Distribution

[38] The spatial distribution of average annual total flash density, Nt, (Figure 3) is in general qualitative agreement with the thunderstorm climatology (Figure 2) as represented by the average annual thunderday map of Australia [Kuleshov et al., 2002], which is the standard reference on thunderstorm frequency in this country [Lightning Protection, 2003]. Both maps demonstrate high annual thunderstorm (Td > 40 day yr−1) and lightning activity (Nt > 10 km−2 yr−1) in the northern parts of Australia, and decreases in annual thunderdays and total flash density southward.

[39] The peak lightning occurrence is in the northwestern part of the Australian continent with Nt values above 15 km−2 yr−1 and it is in the region of the peak in thunderstorm occurrence (Td > 50 day yr−1) as it appears in the thunderday map. The two peaks do not exactly coincide, however this finding is not unexpected. Discussing the world distribution of lightning as observed by the OTD, Christian et al. [2003] reported values of total flash density of about 23 km−2 yr−1 at Entebbe (Td = 206) and Kampala (Td = 242), Uganda, while the planet's “lightning hot spot,” Kamembe, Rwanda, had 83 km−2 yr−1 and 221 thunderdays recorded.

[40] The maximum average annual number of Td > 80 is in the vicinity of Darwin, where storms triggered over the Arnhem Land ranges and over Bathurst Island are very common. High frequency of thunderstorms (40–60 thunderdays per year) is observed at the King Leopold and Durack Ranges in the north of Western Australia. This approximately corresponds to the area of the maximum average annual total lightning flash density with Nt > 30 km−2 yr−1 centered around 16°S, 126°E. The highest value of Nt as derived from the satellite data is about 35 km−2 yr−1.

[41] The secondary maximum of thunderstorm and lightning occurrence is over the northeastern part of the Northern Territory and northern part of Queensland, with values of Td between 40 and 60 days per year, and with Nt values between 5 and 15 km−2 yr−1. Another secondary maximum is evident over eastern part of Australia, with local maxima over the Great Dividing Range, and it is generated by a combination of moisture from the Coral and Tasman Seas, and low-level convergence into low-pressure troughs. Average annual number of thunder days is between 20 and 40 in this area, and lightning activity is also high with Nt values up to 10 km−2 yr−1. Lightning occurrence generally decreases southward, and number of total lightning flashes is small in Victoria, South Australia, the southern parts of Western Australia and New South Wales, and Tasmania, with the values of Nt < 5 km−2 yr−1.

3.3.2. Seasonal Variations

[42] Analysis of annual distributions of lightning occurrences, as registered by the CIGRE-500 LFCs at the selected sites, demonstrates dominance of lightning in Australia in summer months. On average, 90% of annual lightning registrations have been recorded between October and April, with 55% recorded in three months December to February. The maximum monthly lightning occurrences were recorded in December at 30% of sites and in January at 50% of sites. Comparison of seasonal distributions of standardized monthly mean lightning registrations for Kununurra (10 years of data for 1987–1992 and 1999–2002), Brisbane (15 years of data for 1980–1994) and Perth (21 years of data from 1982 to 2002) in Figure 5 demonstrates this summer peak occurrence of lightning. Atmospheric conditions (high boundary layer moisture levels and lower surface pressure) are favorable for thunderstorm and lightning development in this part of the continent in wet season months. The seasonal distributions for the sites located in the northern half of Australia show clear seasonality of the phenomenon: 99.8% of annual lightning registrations in Kununurra and 97.5% in Brisbane were recorded in warmer months October to April. Seasonal distribution of lightning activity in higher latitudes is more uniform. The distribution for Perth (Figure 5) demonstrates that in higher latitudes lightning is still more frequent in warmer months (65% of annual lightning registrations in Perth), but it also occurs in cooler months (May–September) in association with active frontal systems.

Figure 5.

Seasonal distributions of monthly mean lightning ground flashes in Kununurra, Brisbane and Perth, expressed as a fraction of the annual mean.

3.3.3. Variation of Total Flash Density With Latitude

[43] Latitudinal dependence of lightning characteristics is a subject of long-term investigations. Orville and Spencer [1979] reported satellite-based optical observations of total lightning flash density at local dusk and midnight over a latitude range 60°N to 60°S and demonstrated a reduction in flash frequency by a factor of 10 for a 40° increase in latitude in the Southern Hemisphere. Results of another satellite-based observations using radio-frequency detection of lightning indicated a similar rate of change of total flash density, a reduction in lightning flash rate by a factor of about 10 for a change in latitude from 20°S to 60°S [Kotaki et al., 1981]. Mackerras and Darveniza [1994] analyzed latitudinal variation in lightning activity using data from ground-based lightning flash counters recorded at 14 sites in 11 countries covering latitudes from 60°N to 27°S, and concluded that total flash density falls by a factor of 10 for every 30° increase in latitude.

[44] To compare our results with the results of previous studies, values of total flash density, NtSAT, as recorded by satellite detectors were averaged over the Australian continent in 2.5° latitudinal strips (only data gathered over the land were included in the analysis). These estimates of Nt are plotted against magnitude of latitude, λ, in Figure 6, and they demonstrate a reduction in lightning total flash density by a factor of about 10 for a change in latitude from 10°S to 40°S. This reduction rate is similar to the reduction rate in lightning total flash density (an order of magnitude per 30° latitude) found in earlier studies [Mackerras and Darveniza, 1994]. Using these estimates of average annual total flash density, we derived the best fit linear approximation for ln(Nt) as a function of latitude, λ, as

equation image
Figure 6.

Estimates of average annual total flash density, NtSAT, plotted against magnitude of latitude, λ, derived from OTD and LIS combined data.

[45] The values of coefficients α and β were 3.85 and 0.088, with a correlation coefficient of −0.97. These values are in agreement with values of the coefficients α = 3.7 and β = 0.07 as found by Mackerras and Darveniza [1994] using data from worldwide survey for latitudes from 60°N to 27°S. In this study, we specifically used the data gathered by the satellite detectors over the landmass of Australia. Consequently, the variation of Nt for the range of latitudes over the Australian continent (10°S to 40°S) is approximately expressed by the empirical relationship

equation image

where λ is the magnitude of the latitude in degrees.

[46] Lightning is a highly variable phenomenon, both in spatial and temporal terms. The acquired satellite data cover 8 years of records, from 1995 to 2002. In deriving lightning climatology, it is recommended to use long-term records (preferably 10 years or longer) to get statistically reliable conclusions [Lightning Protection, 2003]. Therefore we examined the relationship between estimates of lightning activity, NtLFC, as derived from long-term data (the LFC registrations) versus estimates, NtSAT, calculated from short-term observations (the satellite data). Long-term estimates of Nt were derived from the CIGRE-500 counter registrations, as explained in section 2.3.2. A linear relationship between short-term and long-term estimates of Nt was found with correlation coefficient of 0.95 (Figure 7).

Figure 7.

Estimates of NtSAT values derived from the satellite data (short-term data set) versus estimates of NtLFC values derived from the LFC data (long-term data set).

[47] We also compared the NgLFC values with the Ng values derived from Td using equation (5) (Figure 8). Overall, the two data sets (NgLFC and Ng) are in good agreement (correlation coefficient 0.86), apart from data for two sites, Kununurra and Centre Island. The Ng values as estimated from thunderdays are the same for these localities, while LFC data show that lightning ground flash density is approximately three times higher in Kununurra (NgLFC = 6.0 km−2 yr−1) compared with Centre Island (NgLFC = 1.8 km−2 yr−1). Significantly higher lightning activity in Kununurra, compared with Centre Island, is also confirmed by comparing NtSAT values in Table 1.

Figure 8.

Estimates of NgLFC values derived from CIGRE-500 LFC data versus estimates of Ng values derived from the thunderday data, Td, by applying equation (5).

[48] General good agreement among the data derived by three different methods gives us confidence in the statistical reliability of conclusions about lightning activity derived from the analyzed satellite data set and data from the ground-based LFCs.

3.4. Possible Physical Basis for the Variation of Lightning Flash Density

[49] One of the explanations of variation in lightning rate is traditionally sought in convective available potential energy (CAPE) which is a driving force for a thunderstorm development [Williams and Renno, 1991; Williams, 1995]. CAPE determines the updraft velocity in deep convective systems and therefore determines electrical charge separation rates and rate of lightning occurrence. On the basis of results of the Down-Under Doppler and Electricity Experiment (DUNDEE) conducted during the periods of November 1988 to February 1989 and November 1989 to February 1990 near Darwin, Rutledge et al. [1992] and Williams et al. [1992] demonstrated a strong increase in lightning activity with CAPE as well as nearly linear relationship between CAPE and wet bulb potential temperature. In view of the observations by Rutledge et al. [1992], Williams [1992] and Williams et al. [1992], Mackerras and Darveniza [1994] suggested that the physical basis for the annual variation in monthly total lightning flash density in Darwin is to be sought mainly in the annual variation in wet bulb temperature and CAPE. Investigating further the relationship between lightning activity and surface wet bulb temperature and its variation with latitude in Australia, Jayaratne and Kuleshov [2006] examined data from ten LFCs gathered over a sufficiently long period (ranging from 15 to 21 years of records) and widely distributed across continental Australia. It was demonstrated that at each of the stations the monthly total of lightning ground flashes increased with the monthly mean daily maximum wet bulb temperature. The dependence was most pronounced in the tropics (in Darwin, a modest 3–4°C increase in wet bulb temperature increased the lightning activity by over two orders of magnitude) and decreased in temperate latitudes (in Melbourne, an increase of about half an order of magnitude in the monthly total of ground flashes within a 10°C range of wet bulb temperature was observed). The study by Jayaratne and Kuleshov [2006] was based on one of the world longest records of lightning data, and quantified for the first time the relationship between lightning and surface wet bulb temperature. It appears that the main explanation for the annual variation in monthly flash density and latitudinal variation in annual flash density across the Australian continent is found in the variation in wet bulb temperature and CAPE. However, other factors such as variation in freezing level [Williams et al., 2005; Mushtak et al., 2005], in aerosol loading [Williams and Stanfill, 2002], as well as variation in topography [Boccippio et al., 2001], may also influence the geographical and seasonal variations of lightning parameters. This remains a topic for further investigation.

4. Summary and Conclusions

[50] We have used the lightning data obtained by lightning flash counters (CIGRE-500 and CGR3) over a period of up to 23 years at 39 localities around Australia, as well as lightning data gathered by the NASA satellite-based instruments (OTD and LIS) over a period of eight years, in order to analyze spatial distribution and frequency of lightning activity in Australia. The lightning data were used to produce lightning flash density maps, the first for Australia. Long-term records of thunderdays registered at the Australian stations were also used in this study.

[51] In general, the geographical distribution of lightning incidence expressed as total lightning flash density, Nt, (i.e., cloud-to-ground and intracloud flashes) demonstrates high level of lightning activity (Nt > 10 km−2 yr−1) in the northern parts of Australia and a decrease in total flashes southward (Nt < 5 km−2 yr−1 in Victoria, the southern parts of Western and Southern Australia, and Tasmania). The peak lightning occurrence is in the northwestern part of the Australian continent with Nt values above 30 km−2 yr−1 centered around 16°S 126°E. Secondary maxima are evident over the northeastern part of the Northern Territory and northern part of Queensland, and over the eastern parts of Queensland and New South Wales, with Nt values up to 10 km−2 yr−1. Spatial distribution of lightning occurrence is in general agreement with the spatial distribution of thunderstorms as it is presented in the map of average annual thunderdays for Australia.

[52] Seasonal distributions of lightning occurrence at the selected localities demonstrate predominance of lightning in Australia in warmer months, with peak occurrence of lightning in January. In the northern half of the country, development of thunderstorms and lightning is enhanced by high boundary layer moisture levels and lower surface pressure in the wet season months (October–April). Seasonal distribution of lightning activity in higher latitudes is more uniform; lightning is still more frequent in warmer months, but it also occurs in cooler months (May–September) in association with active frontal systems.

[53] A reduction in lightning total flash density, Nt, by a factor of about 10 for a change in latitude from 10°S to 40°S was found, and this is in agreement with the earlier studies. The variation of Nt for the range of latitudes over the Australian continent is approximately expressed by the empirical relationship Nt = exp(3.85 − 0.088λ) km−2 yr−1, where λ is the magnitude of the latitude in degrees.

[54] The computed value of the cloud flash-to-ground flash ratio, Z, using equation (2) at the studied localities was found to be between 0.75 and 7.7. No supportive evidence for a systematic latitudinal variation of the Z value was found. On the basis of the results of previous studies and these findings, we conclude that for the range of latitude over Australia the most representative value of Z is about 2, and it is largely independent from latitudinal variation. The overall uncertainty in the flash density values indicated on the flash density maps is estimated to be about ±30%.

Appendix A

[55] The procedure for calculating flash densities is [Prentice and Mackerras, 1969]

equation image

where Lg and Lc are the effective ranges for ground and cloud flashes, respectively, so

equation image

whence

equation image

Hence

equation image

or

equation image

where

equation image

Since

equation image

and

equation image

so

equation image

where

equation image

Thus both ground flash and total flash densities can be calculated, using (Lc/Lg) = 18/30 = 0.6, and Z = 2.

Acknowledgments

[56] Satellite data were taken from the v1.0 LIS/OTD gridded climatology provided by the Global Hydrology Resource Center, NASA, USA. The CIGRE-500 lightning flash counter registration data were provided by the Observations and Engineering Branch, Australian Bureau of Meteorology. The calibrations of CIGRE-500 lightning flash counter installations were carried out using funds provided to the University of Queensland by several members of the Energy Supply Association of Australia. The analyses of both CIGRE 500 and CGR3 LFCs were also supported by Lightning and Transient Protection Pty Ltd. G. de Hoedt and S. Chitty, National Climate Centre, Australian Bureau of Meteorology, helped with the mapping process.

Ancillary