Atmospheric and Fuel Moisture Characteristics Associated with Lightning-Attributed Fires [Journal of Applied Meteorology and Climatology]
(Journal of Applied Meteorology and Climatology Via Acquire Media NewsEdge) ABSTRACT
A systematic examination is presented of the relationship between lightning occurrence and fires attributed to lightning ignitions. Lightning occurrence data are matched to a database of fires attributed to lightning ignition over southeastern Australia and are compared with atmospheric and fuel characteristics at the time of the lightning occurrence. Factors influencing the chance of fire per lightning stroke are examined, including the influence of fuel moisture and weather parameters, as well as seasonal and diurnal variations. The fuel moisture parameters of the Canadian Fire Weather Index System are found to be useful in indicating whether a fire will occur, given the occurrence of lightning. The occurrence of ''dry lightning'' (i.e., lightning that occurs without significant rainfall) is found to have a large influence on the chance of fire per lightning stroke. Through comparison of the results presented here with the results of studies from other parts of the world, a considerable degree of universality is shown to exist in the characteristics of lightning fires and the atmospheric conditions associated with them, suggesting the potential for these results to be applied more widely than just in the area of the study.
Lightning is the ignition source of many wildfires throughout the world. A greater understanding of the processes through which lightning-ignited fires occur can therefore be expected to produce benefits such as reductions in the response time to these fires and thus a reduction in the damage that they cause, as well as a better understanding of natural fire regimes and their range of ignition parameters and conditions.
This paper investigates atmospheric states that could potentially influence the chance that lightning will cause a fire, including the occurrence of ''dry lightning''- lightning that occurswithout significant rainfall.Alightning strike on the ground with little accompanying precipitation (dry lightning) can occur in a number of ways: if a thunderstorm is high based with relatively dry air at lower levels such that rain evaporates before reaching the ground (i.e., virga), if a thunderstorm is fast moving such that the rainfall is spread thinly on the ground, or if lightning occurs outside of the rain shaftof a thunderstorm (commonly known as a bolt from the blue).
Rorig and Ferguson (1999, hereinafter RF99) showed that the occurrence of dry lightning, defined as lightning that occurs with rainfall less than 0.1 in. (2.54 mm), in the U.S. Pacific Northwest was related to high values of lower-tropospheric instability (represented by a large temperature difference between 850 and 500 hPa) combined with low atmospheric moisture levels (represented by high dewpoint depression at 850 hPa). RF99 found that the number of lightning-caused fires corresponded more closely to these two parameters than to the total number of lightning flashes that occurred. Their method correctly classified between 56% and 80% of days on which thunderstorms occurred as either ''dry'' or ''wet,'' using their rainfall threshold.
Fires attributed to lightning ignitions (hereafter referred to as lightning fires) have been the subject of a number of recent Northern Hemisphere studies [examples of these include Lutz et al. (2009), Reineking et al. (2010), Drobyshev et al. (2010), Kilpeläinen et al. (2010), and Garcia-Ortega et al. (2011)]. In contrast, a relatively limited number of such studies have been undertaken for Southern Hemisphere locations. This imbalance is not ideal as there are some characteristics of lightning fires for which considerable variation occurs, even between geographic regions that are relatively close to each other. For example, Wierzchowski et al. (2002) report that the average probability of fire from a single lightning flash for adjacent provinces in Canada is 0.07% for Alberta and 2% for British Columbia.
The probability of fuel ignition by lightning has been shown to be relatively independent of fuel moisture, with some fuels igniting even though they may be very wet (e.g., Latham et al. 1997). In contrast, the probability that an ignition is sustained is highly dependent on fuel moisture (e.g., Rothermel 1972; Wotton and Martell 2005). The high dependency of ignition survival on fuel moisture is the reason why the concept of dry lightning is of importance. The influence of the preexisting fuel moisture is investigated in this study using the three fuel moisture components of the Canadian Forest Fire Weather Index (FWI) System (Van Wagner 1987; Dowdy et al. 2009).
This study is the first systematic examination undertaken in Australia of the relationship between lightning occurrence and lightning fires. Nine years of lightningfire occurrence data for fires occurring on public lands in the state of Victoria (Fig. 1) in Australia's southeast are used together with lightning occurrence data to examine the atmospheric conditions associated with lightning fires. The occurrences of lightning, dry lightning, and lightning fires are compared with tropospheric atmospheric states (the RF99 phase space of instability and dryness) and surface atmospheric and fuel characteristics that might influence ignition survival using a combination of observational and gridded products.
2. Data and methodology
a. Data sources
In this study, data from a variety of sources are compared, including fire occurrence and characteristics data, lightning data, rainfall data, atmospheric profile data obtained from NWP models, and modeled fuel moisture data.
The Department of Sustainability and Environment in Victoria maintains a fire database that includes an attribution of ignition source to each fire, and those fires attributed to lightning ignition provide the dataset for this study. The data also include information such as the location of the fire ignition and the time that the fire was first observed. The dataset includes fires that were ignited on public land (i.e., land managed by the state government of Victoria), as well as a small number of fires that were ignited on private land and spread onto public land. In this study fires ignited on private land have not been used, as to do so would produce significant inconsistencies in the data between different locations. This subset of the fire dataset, consisting of lightningattributed fires that were ignited on public land, composes our lightning-fire dataset.
The lightning fires are most commonly first observed during the afternoon [between about 1300 and 1900 local time (LT)]. The fires are observed primarily from their smoke seen during daylight hours, which is likely to account for the jump in the number of fires observed after 0600 LT. About 96% of the lightning fires occur between November and March, with January being the most common month for their occurrence. Considerable interannual variability is apparent in the number of lightning fires, ranging from about 60 during the year 2000 up to about 380 during the year 2007.
Lightning data were obtained from the commercial provider Global Position and Tracking Systems Pty., Ltd., (GPATS) Australia. The GPATS data are based on the time of arrival of the lightning discharge at a network of three ormore radio receivers (e.g., Cummins and Murphy 2009). This technique can detect the multiple return strokes that can be contained within a single lightning flash, as well as distinguish between cloud-to-cloud and cloud-to-ground lightning. This study uses a broadscale approach for combining the lightning data with the fire data (i.e., only using a daily temporal resolution, as well as an effective spatial resolution of ;10 km as detailed in the following section; see section 2b) and so imperfections in the lightning detection efficiency (such as if they are not 100% efficient at distinguishing between cloud-to-ground and cloud-to-cloud lightning) are not expected to significantly influence the results presented in this study.
To classify whether a particular lightning stroke is wet or dry, rainfall data have been obtained from a gridded analysis of rainfall observations (Jones et al. 2009). The rainfall data represent the total rainfall for the 24-h period up to 0900 LT. This is 11 h ahead of UTC when daylight saving time is in effect. The grid spacing of the data is 0.058 in both latitude and longitude-approximately 5 km3 5 km-throughout Victoria.
NWP forecasts from a mesoscale version of the Australian Bureau of Meteorology's operational Limited Area Prediction System (MESOLAPS; Puri et al. 1998) have been used to provide the horizontal, vertical, and temporal resolutions of parameters describing the atmospheric state. For the period of the study, MESOLAPS forecasts were produced each day at analysis times of 0000, 0600, 1200, and 1800 UTC. Forecasts are available on 29 terrain-following vertical sigma (i.e., pressure scaled by surface pressure) levels from sigma 0.9988 to 0.05, on a 0.1258 latitude-longitude grid at 3-hourly intervals to 48 h after initialization. A known issue with the MESOLAPS forecasts is that the 10-m wind speed is underestimated in comparison with the observations. To reduce this bias, wind speed is calculated here as the average of the 10-m wind speed and the gust speed (calculated as the peak wind speed in the mixed layer). This wind speed has been shown to provide more realistic forecasts of fire weather based on MESOLAPS forecasts, following the analysis of meteograms as described by Mills (2005).
he FWI System is used to examine fuel moisture conditions. The FWI System was selected, in preference to the McArthur forest fire danger index (FFDI) (McArthur 1967) that is widely used in Australia, since the FWI includes multiple fuel moisture components: the Fine Fuel Moisture Code (FFMC), which represents the moisture content of fine fuels and litter on the forest floor (based on rainfall, temperature, relative humidity, and wind speed); the DuffMoisture Code (DMC), representing the moisture content of loosely compacted decomposing organicmatter (based on rainfall, temperature, and relative humidity); and the Drought Code (DC), representing the moisture content of deep compact organic matter of moderate depth (based on rainfall and temperature). These three components indicate fuel moisture conditions for three different fuel sizes/depths- a level of detail that is valuable for this study. The fuel moisture components of the FWI System have been calculated from noon (LT) MESOLAPS forecast values of temperature, wind speed, and relative humidity and gridded analyses of rainfall observations (Weymouth et al. 1999). The FFMC, DMC, and DC data consist of daily values, with a gridded resolution of 0.258 in both latitude and longitude (approximately 25 km325 km) throughout Victoria.Details can be found in Dowdy et al. (2009, 2010).
The various datasets used for this study each have different periods of data availability. The time period that has been used throughout this study is from 1 January 2000 until 31 January 2009: this is the maximum period for which all the datasets are available.
b. Determining fire ignition time
The lightning-fire data contain the time that a fire was first observed, not the time that a fire was ignited. The lightning-fire data can be combined with the lightning data to estimate the ignition time of each lightning fire. The lightning stroke closest in time prior to the first observation of a lightning fire is used as the most likely fire ignition time, provided that the lightning stroke occurred within an area of 60.058 (approximately 65 km) in both latitude and longitude of the fire ignition location, resulting in an effective spatial resolution of ;10 km.
It is found that the vast majority of the fires grow large enough to be observed within the first few days after their ignition by lightning (e.g., 72% are observed within 3 days). It is estimated that more than 99% of the ignition days are correctly determined using this method for smolder periods up to 3 days, based on modeling lightning occurrence as a Poisson distribution (Dowdy and Mills 2009), and so a maximum smolder period of 3 days is used to ensure the quality of the lightning-fire ignition times.
c. The chance of fire per lightning stroke
The average chance of fire per lightning stroke is calculated by dividing the number of lightning fires in the dataset by the number of lightning strokes that occurred. There are 1797 lightning fires in the dataset used for this study, representing all fires attributed to lightning ignitions on public land in Victoria during the period 1 January 2000-31 January 2009. The number of lightning strokes that occurred on public land during this period is calculated as follows.
To provide spatial consistency with the lightning-fire dataset, lightning data are used in this study only at public land locations where a lightning fire occurred at least once during the available period of data, with the lightning data and lightning-fire data gridded to the nearest grid point of the 0.058-resolution rainfall dataset. The grid points resulting from the above conditions represent 12%of the area of Victoria, with a total of 131 600 lightning strokes occurring at these locations during the available period of data. Given that about 39% of Victoria is publicly managed land, the total number of lightning strokes that occurred for all of the public land in Victoria during the available period of data is estimated here to be about 427 700 (i.e., 131 600 3 39/12).
Scaling the number of lightning strokes using this method is expected to produce reasonably accurate results: for example, this method suggests that there should be about 1.1 million strokes (i.e., 131 600 3 100/12) throughout the entire area of Victoria during the period of available data, which is the same as the actual value that occurs. The average chance of fire per stroke is thus estimated to be 0.4%, corresponding to 1797 fires resulting from 427 700 lightning strokes. A strong annual variation is apparent in the chance of fire per lightning stroke, with the maximum chance of fire per stroke (0.7%) occurring during January. More details are presented in Dowdy and Mills (2012).
a. The relationship between dry lightning and fire occurrence
To examine the relationship between dry lightning and the occurrence of lightning fires, Fig. 2 shows lightning strokes distributed according to their accompanying rainfall (using 1-mm bins to categorize the rainfall data). Distributions are shown for strokes that were matched to fires (Fig. 2a), and for strokes that were not matched to fires (Fig. 2b). Note that the chance of fire per stroke is shown only when the chance of fire per stroke multiplied by the total number of strokes is greater than 5 to avoid showing data based on very small sample proportions. Statistically significant differences from the average chance of fire per lightning stroke are indicated in Fig. 2c by a square box around the value shown, with statistical significance calculated here and throughout this study using a two-sided binomial test with a 95% confidence interval.
The strokes matched to fires are most commonly accompanied by the lowest rainfall (i.e., from 0 to 1 mm), with a very rapid decrease in the number of fires that occur as rainfall amounts increase (Fig. 2a). In contrast, the distribution for all strokes exhibits a more gradual decrease with increasing rainfall (Fig. 2b). The difference between the two distributions is clearly illustrated by the chance of fire per stroke shown in Fig. 2c: dry lightning has a much greater chance of causing a fire than wet lightning. If less than 1 mmof rainfall occurs, the chance of fire per stroke is about 4 times higher than average.
b. Atmospheric conditions associated with dry lightning
There have been relatively few studies determining the atmospheric conditions associated with the occurrence of dry lightning. The crossover rainfall amount from greater than average to less than average chance of a fire for a given lightning strike (Fig. 2c) is very similar to the 2.54 mmfound by RF99 for the Pacific Northwest, although it should be noted that rainfall data used in this study are based on a gridded analysis of rainfall observations, whereas RF99 used station-based rainfall observations. For the remainder of this section this 2.54-mm threshold has been used to determine whether a lightning stroke is dry or wet.
Following RF99, lightning strokes are categorized by the 850-hPa dewpoint depression (hereinafter DPD) and the 850-500-hPa temperature lapse (hereinafter TL) for 1700 LT (obtained from the 0600 UTC MESOLAPS NWP forecast), smoothed with a threepoint running mean in both latitude and longitude (Fig. 3). Plots are shown separately for dry strokes (Fig. 3a) and wet strokes (Fig. 3b), as well as the ratio of dry strokes to all strokes, to provide an indication of the chance of dry lightning (Fig. 3c). Lightning data are used only for the period 1500-1900 LT from the start of November until the end of March. This time restriction produces a more targeted analysis, since this is when the majority of the lightning-fire ignitions occur. RF99 used a similar targeted analysis period (from May to September, with DPD and TL data from 1700 local daylight time corresponding to 0000 UTC), although they did not determine ignition times of the fires and so fire data were used for all ignition times (whereas Fig. 3 uses fires thought to be ignited only during the period 1500-1900 LT).
There is a higher chance of dry lightning occurrence for higher values of both DPD and TL, although there are a few outliers in regions of low data availability. This is also the case if fires for all ignition times throughout the day are used, which would more closely resemble the methodology of RF99.
c. Atmospheric conditions associated with lightning fires
We now extend the above analysis by examining atmospheric conditions (i.e., theDPD and TL) in relation to fires caused by lightning. In Fig. 4, distributions in TL-DPD phase space are shown of lightning strokes matched and not matched to fire ignitions, together with the ratio of strokesmatched to fires to all strokes (i.e., the sum of the number of strokes matched to fires and the number of strokes not matched to fires-providing an indication of the variation in the chance of fire per stroke). It is apparent that higher values of DPD and TL typically indicate a higher chance of fire per stroke.
The dependence of the chance of fire per stroke on DPD and TL (Fig. 4) is very similar to that for the occurrence of dry lightning (Fig. 3). To examine the relationship between the increased chance of dry lightning and the increased chance of lightning fires, Table 1 shows the number of lightning strokes classified using threshold values of 108C for DPD and 308C for TL, as well as 2.54 mm of rainfall for categorizing the lightning as either dry or wet. Conditional probabilities based on the data from Table 1 are shown in Table 2, with high values of DPD and TL producing statistically significant differences in the probabilities in each case.
The probability of fire given the occurrence of a dry lightning stroke is 3.9% (Table 2). However, the probability of fire given the occurrence of dry lightning and high values of DPD and TL is 4.5%. A similar result can be seen given the occurrence of wet lightning, with the chance of fire being higher when the DPD and TL are high than when they are not high. These results suggest that high DPD and TL values may possibly be related to other factors (in addition to dry lightning) that also produce an increased chance of fire given the occurrence of lightning, and this is examined further in the next section. It is estimated that about two-thirds of the increased chance of fire per lightning stroke for high values ofDPDand TL is attributable to the increased chance of dry lightning for high values of DPD and TL (for details see Dowdy and Mills 2009).
To examine the range of DPD values shown in Fig. 4 that produce a statistically significant change in the average chance of fire per lightning stroke, the distribution of DPD values (using 48C bins to categorize the DPD data) for strokes that were matched to fire ignitions and for strokes that were not matched to fire ignitions is shown in Fig. 5. The ratio of strokes matched to fire ignitions to all strokes is also shown, including an indication of the DPD values for which a statistically significant difference from the average chance of fire per stroke occurs.
The distribution for lightning strokes matched to fire ignitions (Fig. 5a) is skewed toward higher DPD values than is the case for the distribution of lightning strokes notmatched to fire ignitions (Fig. 5b).An average chance of fire per stroke occurs for DPD values in the range 88- 128C, with higher DPD values corresponding to a statistically significant increased chance of fire per lightning stroke (Fig. 5c).
The distribution of TL (using 28C bins to categorize the TL data) for strokes matched and not matched to fires is shown in Fig. 6. There is a clear difference in the shape of the distributions for strokes matched to fire ignitions, which are skewed toward higher values (Fig. 6a) as compared to strokes not matched to fire ignitions (Fig. 6b), with TL values of 308C or higher corresponding to a statistically significant increased chance of fire per lightning stroke and values lower than this corresponding to a statistically significant decreased chance of fire per lightning stroke (Fig. 6c).
d. The influence of surface weather conditions on ignition survival
In addition to dry lightning, there are other factors that might influence ignition survival, including surface weather conditions (temperature, relative humidity, and wind speed) and the preexisting state of fuel moisture.
Lightning strokes that were matched to fires, as well as those strokes not matched to fires, are shown distributed by 1700 LT (obtained from the 0600 UTC MESOLAPS analysis) temperature, relative humidity, and wind speed in Figs. 7-9, respectively. Lightning strokes are used for all times of day and for all times of the year. The chance of fire per stroke is also shown, together with the average chance of fire per stroke.
The temperature distribution (Fig. 7) has a maximum at 288-308C for strokes that were matched to fires, as compared with 248-268C for strokes that were not matched to fires. In general, the higher the temperature is, the higher is the chance of fire per stroke, with temperatures of 268-288C and above having a higher-thannormal chance of fire per lightning stroke. Conversely, temperatures of 228-248C and lower correspond to a lower-than-normal chance of fire per lightning stroke.
Even though relative humidity is partly dependent on temperature, the distributions for relative humidity (Fig. 8) are considerably different to those for temperature (Fig. 7). The relatively humidity distributions are less symmetric (skewed toward low values of relative humidity) than the temperature distributions, particularly for strokes matched to fires. In general, the lower the value of relative humidity, the higher the chance of fire per stroke, with the chance of fire per stroke being close to the average value for relative humidities in the range of about 35%-50%. In addition, there is a much greater proportion of strokes associated with higher humidities in the nonfire distribution, reflecting the need for moisture to sustain convection.
Wind speed does not appear to have a very large influence on the chance of fire per stroke within the range 20-30 km h21 (Fig. 9). Wind speeds above 30 km h21 generally correspond to an increased chance of fire. Wind speeds between 5 and 15 km h21 correspond to a slightly reduced chance of fire, although there is some indication there is an increased chance of fire per stroke for wind speeds less than 5 km h21.
Relative humidity appears to be a better indicator of the chance of lightning fires than wind speed or temperature. For example, for the 100 fires with the lowest relative humidity (,22%), the chance of fire per stroke is about 3.2 times the average value, whereas the chance of fire per stroke is only about 2.5 times the average value for the 100 fires with the highest wind speeds (.37 km h21) and 2.2 times the average value for the 100 fires with the highest temperatures (.338C). This order of importance relates to the chance that an ignition will be sustained, whereas for fire weather conditions in general (as represented by fire weather indices such as the FFDI and FWI) wind speed tends to have the largest influence on severe fire weather conditions in Australia followed by relative humidity and then temperature (Dowdy et al. 2010).
e. The influence of fuel moisture on ignition survival
The distributions of the FFMC, DMC, and DC for lightning strokes that caused fires and for those that did not are shown in Figs. 10-12. Lightning strokes are used for all times of day and for all times of the year. The values of the FFMC, DMC, and DC represent the fuel moisture state prior to the lightning occurrence since they are based on 1200 LT conditions. The formulation of the FFMC is limited to values from 0 to 101, while the DMC and DC can range from 0 upward with no upper limit, with higher values of the fuelmoisture components representing drier fuel conditions.
High values of the FFMC (.90), indicative of dry fine fuels, correspond to a higher than average chance of fire per stroke (Fig. 10). This is also generally the case for high values of the DMC that indicate dry fuels of medium size or depth (Fig. 11), and for high values of the DC that indicate dry fuels of large size or depth (Fig. 12). For the 100 fires with the highest FFMC values (.95.4), the chance of fire per stroke is about 3.4 times the average value. In comparison, the chance of fire per stroke is only about 1.3 times the average value for the 100 fires with the highest DMC values (.81) and 1.1 times the average value for the 100 fires with the highest DC values (.713). This indicates that dry fine fuels (high FFMC values) are the best indicator of a high chance of fire from lightning, followed by dry fuels of moderate size or depth (high DMC values), and then the dry fuels of large size or depth (high DC values).
Low values of the fuel moisture components (moist fuels) shown in Figs. 10-12 indicate relatively low probabilities that a fire will occur. However, there are numerous cases where a fire occurred even though a particular fuel moisture code was low, suggesting that the lightning ignition is being sustained by fuels of a different size or depth to the fuel type represented by that particular fuel moisture code. To examine this possibility, the conditional probability was examined: that when one particular fuel type has high moisture content, an ignition can sometimes be sustained by one of the other fuel types (with high levels of fuel moisture defined here as fuel moisture components less than their 10th percentiles). It was found that when the fine fuel is wet (FFMC , 67), the chance of fire per stroke is 0.14% when the fuel ofmediumsize-depth iswet (DMC,10.5) and 0.37% when it is not wet (DMC $ 10.5). When the fuel of large depth-size is wet (DC ,165), the chance of fire appears to be somewhat dependent on the moisture content of both the fine fuels and the fuels of medium size-depth. However, when the fuels of medium size- depth are wet (DMC , 10.5), fires are unlikely to occur regardless of the values of the other two fuel moisture components. This suggests that although it may be possible for an ignition to survive in one particular fuel layer, even though another fuel layer may be too wet, if fuels of medium size or depth have high moisture content (as represented by low DMC values) the chance that a fire will occur is very small regardless of the moisture content of the other fuel layers.
Of the three fuel moisture components, the chance of fire is lowest when the DMC is low, with only four fires having low DMC values. In contrast, there are 12 fires that have low FFMC values and 18 fires that have low DC values. This suggests that for high moisture content (i.e., low values of the fuel moisture components), it is the fuels of medium size or depth (as represented by the DMC) that are the most critical for determining whether or not an ignition will survive.
It is possible that some lightning-fire ignitions could smolder for a number of days before a change in conditions occurs that extinguishes the smolder without it ever growing large enough to be observed, or that holdover fires could occur (Anderson 2002; Wotton and Martell 2005). To examine the conditions favorable for the survival of a smoldering fire, time series (not shown) of the fuelmoisture components (FFMC,DMC, andDC) and meteorological parameters (temperature, relative humidity, and wind speed) were examined throughout the smolder periods of the fires, but a repeatable pattern could not be identified. The time series did not appear to provide any additional insight than was obtained from Figs. 7-12 into the factors that determine whether or not a fire will result from the occurrence of lightning.
f. The relationship between a fire weather index and ignition survival
Fire weather indices such as the FWI represent the combined influence of fuel moisture and weather parameters on aspects of fire behavior. It might therefore be expected to some degree that the formulations of these indices would represent a reasonable methodology for combining the influence of the various different fuel moisture and weather parameters so as to forecast the chance of ignition survival. To examine this hypothesis, the chance of fire per stroke was investigated in relation to the actual value of the FWI. The value of the FWI is calculated based on a combination of the FFMC, DMC, DC and wind speed, representing the peak daily intensity of the spreading fire as the energy per unit length of fire front (Van Wagner 1987).
Very high values (e.g., ;200) of the FWI better indicate a higher chance of fire per lightning stroke than any of the individual fuel moisture or weather parameters examined previously (from Figs. 7-12). However, these high FWI values occur so infrequently that they would not be particularly useful in applications such as daily operational forecasts of the chance of fire per lightning stroke. For example, for the 100 fires with the highest FWI values (.94), the chance of fire per stroke is about 3.3 times the average value, which is about the same as occurs individually for relative humidity (3.2 times the average value) and the FFMC (3.4 times the average value). This suggests that the way in which the FWI formulation combines all of the various weather and fuel moisture parameters does not result in a large improvement over merely using relative humidity by itself (or the FFMC), indicating that combining the different factors that influence the chance of fire from the occurrence of lightning is not a straightforward matter.
g. The influence of lightning characteristics on ignition survival
There are various different characteristics of a single lightning flash that could potentially have an influence on whether or not a lightning-induced ignition will be sustained. The lightning dataset includes information about each lightning stroke, including the polarity and magnitude of its current. However, because of the inherent limitations of the geographic precision of the datasets, it was not possible to determine whether or not these characteristics have an influence on the chance of fire per lightning stroke.
The multiplicity of the lightning data, shown in Fig. 13, was determined by grouping strokes together for consideration as a single flash if they occurred within 0.5 s and 0.078 (;7 km) in both latitude and longitude of the earliest stroke in a potential grouping [following the method of Richard and Lojou (1996)]. However, the fact that the exact lightning stroke that caused the fire cannot be unambiguously determined in many of the cases means that it was not possible to determine whether or not lightning flashmultiplicity (i.e., the number of strokes per individual lightning flash) has a significant influence on the chance of fire per stroke.
4. Discussion and conclusions
Atmospheric conditions associated with the occurrence of lightning fires were examined for the state of Victoria in southeast Australia. Factors influencing lightning fire occurrence were investigated, with a particular focus on dry lightning, as was the influence of fuel moisture and weather parameters on the chance of fire per lightning stroke.
It was found that if lightning occurs, the amount of accompanying rainfall has a large influence on whether or not a fire will result. For example, if less than 1 mm of rainfall accompanies the lightning, then the chance of fire per stroke was found to be about 4 times higher than average, with the chance of a fire resulting from lightning becoming less than average once the accompanying rainfall exceeds 2-4 mm (Fig. 2c). This is very close to the 2.54-mm (0.1 in.) rainfall threshold determined by RF99 for the U.S. Pacific Northwest. Applying the RF99 method for discriminating between dry and wet lightning- based on the 850 hPa dewpoint depression and the 850- 500-hPa temperature lapse-to southeast Australian conditions showed results consistent with those of RF99. Higher values of both the DPD and the TL indicate a higher chance that lightning will be dry.
The southeast Australian climate is generally drier and less humid than that of the Pacific Northwest, although it is similar in that lightning predominantly occurs in the midafternoon during the summer months. The fact that the relationship between dry lightning and the DPD and TL is reasonably similar in both regions indicates that the physics behind dry lightning may be somewhat universal; that is, high-based thunderstorms with low atmospheric moisture at lower levels produce favorable conditions for the occurrence of dry lightning.
Higher values of the DPD and TL were also found to indicate a higher chance that a fire will result given the occurrence of lightning. Dowdy and Mills (2009) showed that the increased chance of dry lightning for high DPD and TL is likely to account for most (64 out of 87) of the estimated increase in the number of fires that occurred for high DPD and TL. However, this is still 23 fires short of the 174 lightning fires that occurred for high DPD and TL, indicating that high values of these parameters may be related to factors other than low rainfall that also produce an increased chance of fire given the occurrence of lightning.
Fuel moisture and weather parameters were also found to have an influence on the chance of fire per lightning stroke. Temperatures above 268C, relative humidities below 35%, and wind speeds above 30 km h21 all indicate a statistically significant higher than normal chance of fire given the occurrence of lightning. Lower temperatures and higher relative humidities corresponded to a low chance of fire, although this was not found to consistently be the case for lower values of wind speeds. Relative humidity was found to be a better indicator of a high chance of fire per lightning stroke than either temperature or wind speed. This is in contrast to fire weather indices such as the FFDI and FWI for which wind speed tends to have the largest influence inAustralia (Dowdy et al. 2010), and a combined index (the FWI) did not provide an improved indication of a higher chance of lightning fire than did the individual parameters.
The DPD and TL parameters are very similar to those used in the Haines index (HI) (Haines 1988), and thus the HI may be a potential candidate for explaining the increase in fire occurrence that is not accounted for by the increased dry lightning occurrence for high DPD and TL. The HI is a fire weather index used frequently in the United States and has been shown to be a valuable indicator of the potential rapid growth of fires in the western region of the country, althoughRorig and Ferguson (2002) suggest that the HI is of limited use for indicating dry lightning and lightning fires as its categories are too coarse. The HI has also been proposed as a useful measure of potential fire activity in some regions of southern Australia, including Tasmania (Bally 1995) and southwest Western Australia (McCaw et al. 2007). Mills and McCaw (2010) show that statistically extreme values of their extended ''continuous Haines index'' occurred in several cases of lightning-ignited fire outbreaks across southern Australia that they documented.
Dry fine fuels (high FFMC values) were found to be the best indicator of a high chance of fire from lightning, followed by dry fuels of moderate size or depth (high DMC values), with the dry fuels of large size or depth (high DC values) indicating only a slightly higher than average chance of fire per stroke. For high moisture content (i.e., low values of the FFMC, DMC, and DC), the fuels of medium size or depth (as represented by the DMC) were found to have the most influence in determining whether or not an ignition will survive.
There was some indication of ignitions surviving in one particular fuel layer, even though another fuel layer may be too wet. This phenomenon has been reported previously in other studies (e.g., Muraro and Lawson 1970). However, the results presented here also suggest that if fuels of medium size or depth have high moisture content (as represented by lowDMCvalues), the chance that a fire will occur is very small regardless of the moisture content of the other fuel layers. Wotton et al. (2005) showed that a modified version of the DMC was better than the DMC itself in predicting lightning-fire ignition in Ontario, Canada. They defined a Sheltered DuffMoisture Code that represents the amount of moisture in the upper part of the organic layer (the upper;8 cm) in very sheltered locations near the boles of overstory trees. For the Canadian province of Alberta and northwestern Ontario they use a threshold ofDMC520, below which it is assumed lightning will generally result in unsustained ignitions, while in southern Ontario a threshold of DMC 5 30 is used (Wotton et al. 2005) that shows similarities to results for southeast Australia (Fig. 11c).
It has been shown that there are a number of different factors that influence whether or not lightning will cause a fire, including DPD and TL, weather conditions (such as temperature, relative humidity, and wind speed), and fuel moisture content. NWP-based forecasts are available for all of these factors, suggesting considerable potential exists for forecasting the chance of fire given the occurrence of lightning. These forecasts would need to be combined with forecasts of thunderstorm potential (i.e., the chance of lightning) based on predictions of thunderstorm likelihood, such as from the National Thunderstorm Forecast System (Deslandes et al. 2008), if estimations of the chance of lightning-fire occurrence were to be forecast. Such a method could possibly also be used for statistical downscaling of lightning-fire occurrence from climate datasets (e.g., NWP climate simulations, reanalyses or global climate model data). A study by Price and Rind (1994) based on monthly averages of lightning-fire occurrence in the U.S. Southwest suggests a 44% increase in the number of lightningcaused fires for a doubling in atmospheric carbon dioxide level. Vazquez and Moreno (1998) report that the number of lightning fires in peninsular Spain has been increasing. However, there is currently a relative lack of Southern Hemisphere studies relating to how lightningfire occurrence could be expected to change in the future.
The method of predicting dry lightning examined in this study (based on the DPD and TL) provides an indication of high-based thunderstorms with dry air at lower levels such that precipitation evaporates before reaching the ground. However, there are a number of other processes through which dry lightning can occur. These other processes provide scope for further research, including whether fast-moving thunderstorms are more likely to cause fires than slow-moving thunderstorms, as well as investigating the chance of fire from ''bolts from the blue'' where lightning occurs outside of the rain shaftof a thunderstorm.
The results presented here for the state of Victoria in southeast Australia have shown strong similarities to results presented for other locations throughout the world, including the relationship between the amount of rainfall that accompanies lightning and the chance that it will result in a fire, the atmospheric conditions associated with the occurrence of dry lightning, and some aspects of the influence of the preexisting fuel moisture on the chance that a fire will result from the occurrence of lightning. Although some differences were apparent, in general there is a considerable degree of universality between different parts of the world in the characteristics of lightning fires and the atmospheric states associated with them, and this suggests the potential to extend the conclusions of this study to areas beyond those covered by the data used.
Acknowledgments. The lightning-attributed fire data were made available for this study by the Victorian Government Department of Sustainability and Environment. Lightning data were made available from GPATS. The authors would also like to thank JeffKepert (CAWCR) and Bertrand Timbal (CAWCR) for their useful suggestions in improving earlier versions of this document.
Anderson, K., 2002: A model to predict lightning-caused fire occurrences. Int. J. Wildland Fire, 11, 163-172.
Bally, J., 1995: The Haines index as a predictor of fire activity in Tasmania. Proc. Bushfire '95: Australian Bushfire Conf., Parks, and Wildlife Service, Hobart, TAS, Australia, 22 pp.
Cummins, K. L., and M. J. Murphy, 2009: An overview of lightning locating systems: History, techniques, and data uses, with an in-depth look at U.S.NLDN.IEEE Trans. Electromagn.Compat., 51, 499-518.
Deslandes, R., H. Richter, and T. Bannister, 2008: An end-to-end severe thunderstorm forecasting system in Australia: Overview and training issues. Aust. Meteor. Mag., 57, 329-343.
Dowdy, A. J., and G. A. Mills, 2009: Atmospheric states associated with the ignition of lightning-attributed fires. Centre for Australian Weather and ClimateResearch Tech.Rep. 19, 34 pp. [Available online at http://www.cawcr.gov.au/publications/ technicalreports/CTR_019.pdf.]
_____, and _____, 2012: Characteristics of lightning-attributed wildland fires in south-east Australia. Int. J. Wildland Fire, 21, 521- 524, doi:10.1071/WF10145.
_____, K. Finkele, G. A. Mills, and W. de Groot, 2009: Australian fire weather as represented by the McArthur forest fire danger index and the Canadian forest fire weather index. Centre for Australian Weather and Climate Research Tech. Rep. 10, 84 pp. [Available online at http://www.cawcr.gov.au/ publications/technicalreports/CTR_010.pdf.]
_____,_____, _____, and _____, 2010: Index sensitivity analysis applied to the Canadian forest fire weather index and the McArthur forest fire danger index. Meteor. Appl., 17, 298-312.
Drobyshev, I., M. D. Flannigan, Y. Bergeron, M. P. Girardin, and B. Suran, 2010: Variation in local weather explains differences in fire regimes within a Québec south-eastern boreal forest landscape. Int. J. Wildland Fire, 19, 1073-1082.
Garcia-Ortega, E., M. T. Trobajo, L. López, and J. L. Sánchez, 2011: Synoptic patterns associated with wildfires caused by lightning in Castile and Leon, Spain. Nat. Hazards Earth Syst. Sci., 11, 851-863.
Haines, D. A., 1988: A lower atmospheric severity index for wildland fires. Natl. Wea. Dig., 13, 23-27.
Jones, D. A., W. Wang, and R. Fawcett, 2009: High-quality spatial climate data sets for Australia. Aust. Meteor. Oceanogr. J., 58, 233-248.
Kilpeläinen, A., S. Kellomäki, H. Strandman, and A. Venäläinen, 2010: Climate change impacts on forest fire potential in boreal conditions in Finland. Climatic Change, 103, 383-398, doi:10.1007/ s10584-009-9788-7.
Latham, D., R. Burgan, C. Chase, and L. Bradshaw, 1997: Using lightning location in the Wildland Fire Assessment System. Intermountain Research StationGeneral Tech. Rep. INT-GTR- 349, Department of Agriculture, 6 pp.
Lutz, J. A., W. J. W. van Wagtendonk, A. E. Thode, J. D. Miller, and J. F. Franklin, 2009: Climate, lightning ignitions, and fire severity in Yosemite National Park, California, USA. Int. J. Wildland Fire, 18, 765-774.
McArthur, A. G., 1967: Fire behaviour in eucalypt forests. Department of National Development Forestry and Timber Bureau Leaflet 107, Canberra, ACT, Australia, 36 pp.
McCaw, L., P. Marchetti, G. Elliott, and G. Reader, 2007: Bushfire weather climatology of the Haines index in south-western Australia. Aust. Meteor. Mag., 56, 75-80.
Mills, G. A., 2005: On the subsynoptic-scale meteorology of two extreme fire weather days during the eastern Australian fires of January 2003. Aust. Meteor. Mag., 54, 265-290.
_____, and L. McCaw, 2010: Atmospheric stability environments and fire weather in Australia-Extending the Haines index. Centre for Australian Weather and Climate Research Tech. Rep. 20, 151 pp. [Available online at http://www.cawcr.gov.au/ publications/technicalreports/CTR_020.pdf.]
Muraro, S. J., and B. D. Lawson, 1970: Prediction of duffmoisture distribution for prescribed burning. Pacific Forest Res. Centre Info. Rep. BC-X 46, Canadian Forestry Service, 23 pp.
Price, C., and C. Rind, 1994: The impact of a 2 3 CO2 climate on lightning-caused fires. J. Climate, 7, 1484-1491.
Puri, K., G. D. Dietachmayer, G. A. Mills, N. E. Davidson, R. A. Bowen, and L. W. Logan, 1998: The newBMRCLimited Area Prediction System (LAPS). Aust. Meteor. Mag., 47, 203-223.
Reineking, B., P. Weibel, M. Conedera, and H. Bugmann, 2010: Environmental determinants of lightning- v. human-induced forest fire ignitions differ in a temperate mountain region of Switzerland. Int. J. Wildland Fire, 19, 541-557.
Richard, P., and J. Y. Lojou, 1996: Assessment of application of storm cell electrical activity monitoring to intense precipitation forecast. 10th Int. Conf. on Atmospheric Electricity, Society of Atmospheric Electricity of Japan, Osaka, Japan, 284-287.
Rorig, M. L., and S. A. Ferguson, 1999: Characteristics of lightning and wildland fire ignition in the Pacific Northwest. J. Appl. Meteor., 38, 1565-1575.
_____, and _____, 2002: The 2000 fire season: Lightning-caused fires. J. App. Meteor., 41, 786-791.
Rothermel, R. C., 1972: A mathematical model for predicting fire spreads in wildland fuels. U.S. Forest Service Res. Paper INT- 115, 40 pp.
Van Wagner, C. E., 1987: Development and structure of the Canadian Forest Fire Weather Index System. Canadian Forestry Service Tech. Rep. 35, Ottawa, ON, Canada, 37 pp.
Vazquez, A., and J. M. Moreno, 1998: Patterns of lightning- and people-caused fires in peninsula Spain. Int. J. Wildland Fire, 8, 103-115.
Weymouth, G., G. A. Mills, D. Jones, E. E. Ebert, and M. J. Manton, 1999: A continental-scale daily rainfall analysis system. Aust. Meteor. Mag., 48, 169-179.
Wierzchowski, J., M. Heathcott, and M. D. Flannigan, 2002: Lightning and lightning fire, central cordillera, Canada. Int. J. Wildland Fire, 11, 41-51.
Wotton, B. M., and D. L. Martell, 2005:A lightning fire occurrence model for Ontario. Can. J. For. Res., 35, 1389-1401.
_____, B. J. Stocks, and D. L. Martell, 2005: An index for tracking sheltered forest floor moisture within the Canadian Forest Fire Weather Index System. Int. J. Wildland Fire, 14, 169-182.
ANDREW J. DOWDY AND GRAHAM A. MILLS
Centre for Australian Weather and Climate Research, Docklands, Melbourne,
and Bushfire Cooperative Research Centre, East Melbourne, Victoria, Australia
(Manuscript received 12 October 2011, in final form 16 May 2012)
Corresponding author address: Andrew J. Dowdy, Bureau of Meteorology, 700 Collins St., Docklands, Melbourne, VIC 3008, Australia.
(c) 2012 American Meteorological Society
[ Back To Technology News's Homepage ]