Abstract

Half-hourly airport weather observations have been used to construct high-temporal-resolution datasets of McArthur Mark V forest fire danger index (FFDI) values for three locations in Tasmania, Australia, enabling a more complete understanding of the range and diurnal variability of fire weather. Such an understanding is important for fire management and planning to account for the possibility of weather-related fire flare ups—in particular, early in a day and during rapidly changing situations. In addition, climate studies have hitherto generally been able to access only daily or at best 3-hourly weather data to generate fire-weather index values. Comparison of FFDI values calculated from frequent (subhourly) observations with those derived from 3-hourly synoptic observations suggests that large numbers of significant fire-weather events are missed, even by a synoptic observation schedule, and, in particular, by observations made at 1500 LT only, suggesting that many climate studies may underestimate the frequencies of occurrence of fire-weather events. At Hobart, in southeastern Tasmania, only one-half of diurnal FFDI peaks over a critical warning level occur at 1500 LT, with the remainder occurring across a broad range of times. The study reinforces a perception of pronounced differences in the character of fire weather across Tasmania, with differences in diurnal patterns of variability evident between locations, in addition to well-known differences in the ranges of peak values observed.

1. Introduction

Over much of the forested area of Australia, a forest fire danger index (FFDI) is calculated using the McArthur Mark V forest fire danger meter (McArthur 1967). McArthur noted that, depending on the density of forest cover, most eucalypt forests would achieve minimum fuel moisture (and therefore maximum capacity to sustain fire activity) during the early afternoon to midafternoon. In the case of many manual observation sites with records stretching over some decades, 0900 and 1500 local clock time (LCT) are the only times during the day at which readings were made.1 Of these, only the observations made at 1500 LCT generally provide observations representative of fire-weather conditions during the day. McArthur (1967) indicated, and other researchers (Ryan 1977; Tolhurst and Cheney 1999; Beck et al. 2002) have felt, that in most cases surface heating has not been sufficient by 0900 LCT to allow mixing to the surface of airmass properties that will influence the day’s fire-weather conditions, and in particular the moisture content of potential fuels.

While it is certainly useful for monitoring and fire-weather warning purposes to indicate either forecast or observed conditions at a set time of the day, and in particular when fire danger at that time is commonly high relative to other times, awareness of potential fire-weather conditions at other times is critical for safe and effective fire-control operations. When fire danger values are computed from all (eight 3-hourly) synoptic observations at sites for which a record is available throughout the day there is some evidence (Fox-Hughes 2008) that peak fire danger frequently occurs at times of the day other than midafternoon.

A number of studies have examined the fire-danger history at sites in Australia (Long 2006; Fox-Hughes 2008; Lucas 2010; B. Santos et al. 2011, unpublished manuscript), and some studies have linked fire-weather behavior to broadscale atmospheric and ocean processes such as ENSO (Verdon et al. 2004; Williams and Karoly 1999). Many of these studies examine fire danger index values computed from weather parameters observed at 1500 LCT. It is widely recognized by these and other authors who have used 1500 LCT data, as the only available data, in fire-weather climatological studies (Lucas 2010) that FFDI may peak at times other than 1500 LCT. As a result, some potentially significant fire-weather events are missed in such studies. Even when a full 3-hourly synoptic dataset is considered, however, short-lived events may be missed [as discussed in Mills (2008)]. Here, “significant” is used for those events for which peak FFDI reaches at least 25 (very high fire danger in the McArthur rating system), above which fire managers would not attempt a direct attack on an ongoing fire.

With the increasing deployment of automatic weather stations (AWS) throughout Australia since the mid-1990s, hourly or half-hourly observations have become available from an increasing number of locations. This has demonstrated that, in certain conditions, short-period, large-amplitude variations in FFDI can occur. The Climate Section, Tasmania and Antarctic Regional Office of the Australian Bureau of Meteorology, has digitized archived Tasmanian airport weather reports since October of 1990. The length of the record is now such as to allow characteristics of these frequent observations to be determined and to assess the impact of more frequent data on fire-weather climates.

The current study aims to describe a more comprehensive fire-danger climate at sites in Tasmania during periods of up to 20 yr by computing fire-danger indices from half-hourly airport AWS records. This task is of interest in its own right by allowing, for example, better planning by fire agencies. In addition, it can inform climatic-change studies that calculate future fire danger based on current and past levels of those quantities (e.g., Williams et al. 2001). Many such studies make use of fire-danger indices calculated from midafternoon weather observations. This is, of course, very reasonable because midafternoon is broadly regarded as the time of peak fire danger diurnally. In addition, many sites have very limited data available at other times, certainly over longer periods. This study aims to quantify the extent to which the assumption of a midafternoon fire peak is valid, at least in the Tasmanian context, and thereby to suggest the extent to which such climate studies have sampled the full range of fire-weather behavior of the studies’ domains; B. Santos et al. (2011, unpublished manuscript) are undertaking a similar study using Western Australian data.

2. Data/method

FFDI values were calculated using observations of recent rainfall (in the form of a “drought factor”), temperature, dewpoint temperature, and wind speed using the McArthur Mark V forest-fire danger meter (McArthur 1967), as encoded for numerical calculation by Noble et al. (1980). Drought factors were calculated using soil-dryness index values (Mount 1972), using the modified drought-factor calculation detailed in Griffiths (1999). The FFDI values resulting from these calculations are integers that are greater than or equal to 0. FFDI values have been grouped into categories for ease of public understanding and classification of fire risk, following McArthur (1967). The ratings used in Tasmania currently are “low–moderate” (LM: FFDI 0–11), “high” (H: 12–24), “very high” (VH: 25–49), “severe” (SV: 50–74), “extreme” (EX: 75–99), and “catastrophic” (CAT: 100+).2 In this paper, an additional distinction is made to highlight FFDI values that reach VH 38, representing the upper extent of the VH range, where fire-weather warnings are issued by the Bureau of Meteorology in Tasmania and fire authorities declare a day of “total fire ban.” A total fire ban is a legally enforced prohibition by the Tasmania Fire Service (TFS) on the lighting of fires and a restriction on other activities that might lead to an unintended fire ignition. Such a ban is generally imposed when the FFDI is forecast to exceed 38 at several locations. TFS reserves the right to declare a total fire ban at other thresholds, however, depending on factors such as resource availability.

As noted above, digitized half-hourly or hourly AWS reports are available from a number of locations from October of 1990 onward. Observations from Hobart, Launceston, and Devonport Airports were chosen for examination in this study because all have a generally stable observation history with 24-h observation coverage.

The locations of Hobart, Launceston, and Devonport Airports are indicated in Fig. 1. Devonport Airport is located on the central north coast of Tasmania, within approximately 500 m of the coast, with Bass Strait to the north separating Tasmania from continental Australia. Launceston Airport lies at the southern end of the Tamar Valley, with Tasmania’s Central Plateau to the west and the northeastern highlands to the east. Hobart Airport is in the southeast of the state, sheltered by the Wellington Range from the prevailing westerly winds that flow over Tasmania. To its immediate south lies Frederick Henry Bay, with Storm Bay and the Tasman Sea beyond.

Fig. 1.

Map of Tasmania, showing locations and topographic details mentioned in the text. Lighter shading over land indicates higher topography.

Fig. 1.

Map of Tasmania, showing locations and topographic details mentioned in the text. Lighter shading over land indicates higher topography.

Airport observations are recorded at routine intervals, in which case they are known for brevity in operations and in the remainder of this paper as “METARs.” Nonroutine observations are recorded on the exceedance of criteria critical for aviation operations, in which case a “SPECI” is issued. Of the phenomena for which a SPECI might be issued, only the occurrence of a wind gust is of relevance in this study. In particular, gusty winds are a feature of many days of dangerous fire weather. In the normal course of operations, the only occasions on which a scheduled METAR is not issued is when a SPECI has been issued within the preceding 10 min. In addition, there are rules governing the frequency with which SPECI messages are issued.

The dates selected for analysis were those at each airport for which digitally archived routine half-hourly AWS observations were available. These are

  1. Hobart Airport: 2 October 1990–30 June 2010, uninterrupted for any prolonged period;

  2. Launceston Airport: 8 May 1992–29 July 2004; and

  3. Devonport Airport: 21 July 1991–17 July 1998 and 19 October 2006–30 June 2010.

Thus, approximately 19 yr of half-hourly data are available for Hobart Airport and 12 yr of data are available at the other two sites. Because the purpose of this paper is not a direct comparison among the three sites, it is not critical that the three sites have identical periods of data availability.

For the three airports, routine METARs are generated half-hourly, resulting in 48 METARs in a 24-h period of normal operation. There may be many more SPECIs reported—in particular, during windy days with frequent gusts. So that such days are not overrepresented in this analysis it is useful to employ at least one summary measure of the fire danger experienced during a day. For much of the following analysis, the peak FFDI reported between midnight and the following midnight is used as an indicator of the overall level of fire-weather activity for a day. It is recognized that there are many other possibilities, and some of these alternatives are discussed later in the text. In brief, alternatives to this measure include

  1. length of time that FFDI continuously exceeds a trigger value such as 24 or 50;

  2. total time over a 24-h period, or subset thereof, that FFDI exceeds a threshold value; and

  3. an integration of the FFDI over 24 h, or subset thereof.

Data were quality controlled to remove occasional spurious observations that had slipped through the real-time quality checking. Observations that clearly exceeded climatological extremes or that were grossly inconsistent with those close in time were discarded. Thus, temperatures in the dataset that exceeded the record maximum temperature at each site were removed. Spikes in temperature or dewpoint that clearly were not associated with physical processes (such as frontal passages) were also removed when detected. Such spikes and anomalous values frequently manifested as values that were different by 10°C or more from those 30 min before or after. In addition, the count of observations per day was examined. It was deemed that there had been substantial problems with either the AWS itself or with communications on days for which fewer than 36 observations were stored. On this basis, 145 days of observation were removed from the Hobart Airport dataset, 73 days of observations were removed from the Launceston Airport dataset, and 146 days of observations were removed from the Devonport Airport dataset. Subsequent analysis was conducted on 349 421 observations on 7036 days for Hobart Airport, 212 332 observations on 4363 days for Launceston Airport, and 183 195 observations on 3710 days for Devonport Airport.

Routine 3-hourly synoptic reports from the same stations for the same time periods were also gathered. In most cases, these were a subset of the METAR/SPECI reports (except in events for which a SPECI was reported within the 10 min preceding a synoptic report and no routine METAR was then issued). In contrast to METAR/SPECI reports, only eight synoptic observations are made per day. Nonetheless, to allow as similar as possible a comparison between the two observation sets, maximum daily values of FFDI were also prepared for synoptic observations. Note that maximum daily FFDI values computed from the (eight) synoptic observations are not the same quantity as FFDI computed solely from the 1500 LCT synoptic observation (as noted in the introduction). To highlight this difference and to provide an indication of the degree to which studies using only 1500 LCT data may underestimate the true occurrence of fire-weather events, the latter are included in this report. It is worth noting here that it is important to be aware of changes to and from daylight saving time when examining synoptic observations—hence, the use of LCT in this report. It is a less significant issue for METAR reports, however, at hourly or half-hourly frequency.

Although the three airports have enjoyed relative observational stability over the last 20 yr, some changes have occurred. An AWS was installed at Hobart Airport in April of 1990. This was shortly prior to the period for which digitally archived data are available and does not, therefore, affect the data examined here. In a similar way, Devonport Airport AWS was installed in December of 1990, prior to the commencement of half-hourly METAR reports.

An AWS was installed at Launceston Airport on 19 May 1992. In addition, the observation site at Launceston Airport was moved in July of 2004 in response to increased security provisions, although the old site was maintained for several years at a reduced reporting frequency for comparison purposes. The latter change at Launceston Airport is the more fundamental. Although the broad environment remained the same, the observation site was moved approximately 1350 m to the south-southeast and 0.5 m higher in elevation, which affected the microclimate to some extent. A comparison of the available monthly summary data from the two sites reveals that the new station is slightly warmer during the day and colder at night than the old site (by fractions of a degree Celsius), and that it receives a little less rainfall—during 2007, for example, the old station received 541.2 mm as compared with the new station’s 471.8 mm. Both wind speed and dewpoint temperature are, on average, fractionally higher at the new site. It is acknowledged, however, that the period of overlap for the two Launceston Airport sites is relatively short, and any conclusions regarding differences between them are tentative. The dataset used in this study, however, has been restricted to the old site, for the period during which half-hourly reports are available, to rule out any possible effects of the site change on the results of this study.

3. Results

a. Broad comparison of METAR and synoptic observation datasets

Percentile values of FFDI are presented in Fig. 2 for each station for both synoptic observations and METAR reports, on the basis of maximum daily FFDI, in stacked, clustered columns. Several indicative percentiles (50th, 90th, and 99th) are given in each case, using all data. Percentiles calculated for the period of October–March within each dataset time period are indicated by an incremental increase on each column. Thus, for example, the 90th percentile of daily maximum FFDI computed for METAR observations at Hobart Airport throughout the year is 17, whereas the 90th percentile of METAR observations for the same site during October–March is 22. October–March is generally regarded as the peak fire-weather period (the fire season), although occasional fire-weather events do occur in September or April. In addition, percentile values of the FFDI measured at 1500 LCT values are included for comparison. A number of results are clear from Fig. 2. For all stations and for all percentile values, the METAR value for a percentile exceeds that for synoptic observations. Further, percentiles calculated from METAR and most synoptic observation values exceed those for 1500 LCT (synoptic) observations only. This is the case whether the whole year is considered or only the (nominally October–March) fire season is considered. These summary statistics confirm that consideration of 1500 LCT data alone will underestimate the level of fire danger experienced at a site. In a similar way, consideration of only synoptic observation data will also underestimate the true level of FFDI at a site. In particular, for higher percentile values the gap between METAR, all synoptic observations, and solely 1500 LCT observations tends to widen, especially during the fire season. Thus, the more dangerous the fire weather is, the more synoptic observations and 1500 LCT–only observations underestimate the true situation. The manner in which this difference increases with FFDI (or percentile of FFDI) has not been investigated, however.

Fig. 2.

Stacked, clustered column plots of 50th, 90th, and 99th percentiles of FFDI values recorded at (a) Hobart Airport, (b) Launceston Airport, and (c) Devonport Airport for each of METAR reports, synoptic reports, and 1500 LCT reports. Each column is split into the percentile obtained using all data for the year (“all” in the figure legend), capped with the increment obtained when the period considered is restricted to the usual fire-season months of October–March (“fs” in the legend).

Fig. 2.

Stacked, clustered column plots of 50th, 90th, and 99th percentiles of FFDI values recorded at (a) Hobart Airport, (b) Launceston Airport, and (c) Devonport Airport for each of METAR reports, synoptic reports, and 1500 LCT reports. Each column is split into the percentile obtained using all data for the year (“all” in the figure legend), capped with the increment obtained when the period considered is restricted to the usual fire-season months of October–March (“fs” in the legend).

Figure 3 shows, for both METARs and synoptic observations at Hobart and Launceston Airports, the number of days per year (July to the following June) on which the maximum FFDI was at least 25 (Devonport Airport is similar but with fewer counts of very high FFDI). Years are labeled in the figure by their first calendar half-year—for example, the 2006 year consists of the period 1 July 2006–30 June 2007. It is clear that there will be at least as many days of VH FFDI when considering METARs as compared with synoptic observations since the latter are (in general) a subset of the former. Figure 3 demonstrates that for most seasons there are between 1 and 2 times as many days of VH FFDI evident on METARs as synoptic observations, further illustrating that an assessment of fire-danger levels based on synoptic (or less frequent) observations will underrepresent the true levels of FFDI at a site.

Fig. 3.

Counts of FFDI ≥ 25 (very high fire danger) by season for (a) Hobart Airport and (b) Launceston Airport. Note the different y-axis scales on the plots. The same x scale is used on both plots, to better allow comparison of activity between the stations over different seasons.

Fig. 3.

Counts of FFDI ≥ 25 (very high fire danger) by season for (a) Hobart Airport and (b) Launceston Airport. Note the different y-axis scales on the plots. The same x scale is used on both plots, to better allow comparison of activity between the stations over different seasons.

There is a well-recognized, substantial interannual variability in the number of recorded very high FFDI events. In the case of Hobart Airport, there is a range of 1–28 days per season on which VH FFDI was recorded over the study period. At Launceston Airport, the range is 0–14 days per season. There is also considerable variability in the ratio of events recorded as METARs to those recorded as synoptic observations. In general, active seasons result in a higher ratio of METAR events to those computed from synoptic observations. This is very likely a result of a higher number of SPECIs reported during the course of such seasons, given that fire-weather events are often characterized by gusty winds for which a SPECI is likely to be issued.

Further, a comparison of Figs. 3a and 3b suggests that there is only a moderate correlation between the two plots. Some seasons are particularly active or inactive at both locations, but there are other seasons in which there are very different levels of activity. For example, the 2002 (i.e., 2002/03) season was active at both locations, with 17 days on which VH FFDI occurred from METAR reports at Hobart Airport and 12 at Launceston Airport. The following season, however, was relatively inactive at Launceston Airport, with 6 days of VH FFDI from METARs, but it was the most active season in the dataset at Hobart Airport, with 28 METAR events. These differences are discernible from synoptic or even 1500 LTC–only observations, but they are highlighted by the availability of more frequent observations and are of some significance for studies of future climate, suggesting that data from a single station will be of use in projections for only very limited regions.

b. Duration of events

In many cases, episodes of VH FFDI are short lived and are not resolved by synoptic observations (as suggested in Figs. 2 and 3). This assertion can be examined by checking the duration between the first and last observation of VH FFDI on days of very high fire danger. On marginal VH FFDI days in particular, the FFDI can dip below VH rating for periods. Nonetheless, it is valid to define the VH event duration as the time interval between first and last VH FFDI observation, even if that period is not continuous. Of course, a number of different measures of event duration are possible, each of which may be informative in particular contexts. For example, the longest period of continuous VH FFDI and the total time above VH FFDI are both reasonable indicators of the level of fire danger for a day. The time interval between first and last observation of VH FFDI, however, provides a stringent upper limit to event duration and therefore the likelihood of detection of the VH FFDI event by a synoptic or 1500 LTC–only observation schedule.

Figure 4 plots the time interval between first and last VH FFDI observation for both Hobart (Fig. 4a) and Launceston Airports (Fig. 4b). VH FFDI events occurred very infrequently at Devonport Airport and are not included here. Duration of events was binned into half-hourly categories. Those with a duration of less than 0.5 h, most commonly single VH FFDI observations, were placed in the “0 hour” bin. Also plotted are the durations of events for which the FFDI exceeds 38. Again, a single observation of VH 38+ was placed in a “0 hour” duration bin, together with, rarely, two or more such observations within a 0.5-h period (with no subsequent observation of FFDI greater than or equal to 38). In the case of both Hobart and Launceston Airports, there is a very rapid decrease in the number of VH FFDI events as duration increases, with a long tail of longer-duration events—in particular, at Hobart Airport. Of interest is that in both locations the decline in the number of VH 38+ events with increasing duration is much less abrupt.

Fig. 4.

Duration of fire danger at (a) Hobart Airport and (b) Launceston Airport. “Number of events” refers to the total number of events recorded at each station for the respective study periods, binned by 0.5 h. Axes are again scaled differently.

Fig. 4.

Duration of fire danger at (a) Hobart Airport and (b) Launceston Airport. “Number of events” refers to the total number of events recorded at each station for the respective study periods, binned by 0.5 h. Axes are again scaled differently.

Of considerable interest for both locations is the proportion of events that last for fewer than 3 h, because these could potentially escape detection by a synoptic observing program. For Hobart Airport, of the 301 VH FFDI events, 175 (58%) were of less than 3 h in duration and thus were at risk of not being detected by a synoptic-only observation schedule. In the case of Launceston Airport, 48 of a total 73 VH FFDI events (66%) lasted for fewer than 3 h.

Of the 100 cases of FFDI VH 38+ at Hobart Airport, 62 reached or exceeded FFDI 38 for fewer than 3 h. When the criterion for FFDI was set at FFDI 70 (not displayed), some 17 events were detected in the METAR database for Hobart Airport, of which 15 (88%) were at or above FFDI 70 for fewer than 3 h. At Launceston Airport, of the 10 VH 38+ events, 8 lasted for fewer than 3 h. The above suggests that the higher the value at which FFDI peaks is, the less likely the peak is to be resolved by a synoptic observing network.

Further, of the 100 days on which Hobart Airport FFDI equaled or exceeded 38, the first observation of FFDI ≥ 38 occurred after 1500 LCT on 16 days; on 10 of those days, FFDI was still below 24 at 1500 LCT. In a similar way, there are days in the Hobart Airport dataset on which the latest observation of FFDI of ≥38 occurred prior to 1500 LCT. Of the 100 days on which FFDI reached or exceeded 38, conditions had eased by 1500 LCT on 37 days. Of these 37 events, conditions had eased below FFDI 24 by 1500 LCT on 15 days.

As an example of the rapid variability of fire-weather conditions around southeastern Tasmania, the event of 22 January 2006 is examined. On this day, Hobart Airport recorded a peak FFDI value of 64 at 1825 LCT. This FFDI was the highest of the 16 days in the Hobart Airport dataset that recorded FFDI below 38 until after 1500 LCT. The FFDI was 13, barely in the high range, at that time and had been lower than that until 1430 LCT. A weak sea breeze was pushed offshore after 1630 LCT as northwesterlies strengthened. By 1700 LCT, the FFDI had climbed to 44, but by 1901 LCT a cooler, moist southeasterly change had moved through Hobart Airport and the FFDI fell to 15. The time that the FFDI spent above 38 was less than 2 h, and the observation record indicates that FFDI was in the range above 24 for only a few minutes longer than that. On this occasion, the synoptic observation schedule captured the event but not the peak value. Had the cold front traversed the Hobart Airport region an hour earlier, however, the event would not have been captured at all by the synoptic schedule. Furthermore, even in this scenario, the event would not have been registered by an examination of 1500 LCT–only observations.

Note that the brevity of this event was not simply a function of the microclimate of Hobart Airport as a particular location in a complex coastal environment. Bushy Park, located approximately 55 km to the west–northwest of Hobart Airport in the inland Derwent Valley, also recorded elevated fire danger on this day. The FFDI at Bushy Park peaked at 52 and exceeded 38 for about 75 min, spending about 3.5 h with FFDI above 24. Thus, although the transition to and from elevated fire danger was not as abrupt at Bushy Park, it experienced conditions that were comparable to those at Hobart Airport. Other locations in southern Tasmania experienced broadly similar fire weather on this day, although Hobart Airport did record the highest FFDI. During this time, a number of fires were burning across a wide swathe of Tasmania as a result of widespread lightning on 20 January. Most of these fires became more difficult to control as a result of the weather conditions on 22 January.

Some events have occurred with very pronounced peaks early in the day but with the 1500 LCT observation being below the high range. A peak FFDI of 115 was recorded at 1200 LCT 11 January 1991, yet the passage of a cool, moist southeasterly change resulted in a 1500 LCT FFDI value of 9. On the other hand, early onset of sea-breeze conditions, as on 22 January 2006, or the persistence of a maritime boundary layer until well into the afternoon can mask the approach of a change, and hot, dry north-to-northwesterly winds are typically not then experienced until after a 1500 LCT reading has been made.

These real and not atypical events indicate strongly that climatological descriptions that are based solely on 1500 LCT observations will miss many such significant fire-weather events. To resolve the full range of possible fire-weather conditions, vital in an attempt to characterize current and future fire-weather climatological behavior, weather conditions throughout the day need to be considered.

c. Relationship between duration and severity of events

The event that had the longest recorded period of FFDI in excess of 24, some 20.5 h, occurred on 12 October 2006 at Hobart Airport. An accidental fire ignition occurred during this day that resulted in an urban-interface fire that burned 800 ha. This event also recorded one of the highest FFDI values in the entire Tasmanian fire-weather database and raises the question of whether long-duration fire-weather events are necessarily associated with unusually elevated FFDI and vice versa.

An examination of events with FFDI of at least 24 and lasting more than 10 h (eight events at Hobart Airport during the study period) indicates that all but one had peak FFDI ratings that were greater than or equal to 50. The exception, 6 December 2006, had FFDI 25 at 0100 LCT, after a day on which FFDI exceeded 50 on 5 December. Late in the afternoon of 6 December, FFDI again increased above 24. This event, then, was of long duration only by virtue of a set of unusual conditions. (The definition of event duration was not amended at this point, however, because it affected only a very small proportion of the total number of events in the dataset.) Five of the remaining six events had peak FFDI of 69 or more.

To examine the possibility that significant events are necessarily, or usually, long lived, days at Hobart Airport with a peak FFDI that reached at least 70 were identified, together with the duration of each event. The 17 resulting events (Table 1) had a surprisingly diverse range of durations, from 2 to 20.5 h. These indicate that severity is usually, but certainly not always, associated with a duration of at least 5 h. Of interest is that two of the most dramatic counterexamples, on 11 January 1991 (peak FFDI 115; discussed above) and 21 January 1997 (peak FFDI 97), recorded FFDI that was ≥24 for only 2.5 h. Both events were associated with the development of small but intense mesoscale circulations over southeastern Tasmania. Mills and Pendlebury (2003) examined the latter event in detail.

Table 1.

Occasions within the period of October 1990–June 2010 on which FFDI at Hobart Airport reached 70.

Occasions within the period of October 1990–June 2010 on which FFDI at Hobart Airport reached 70.
Occasions within the period of October 1990–June 2010 on which FFDI at Hobart Airport reached 70.

d. Time of maximum FFDI

The high temporal resolution of the METAR database allows an investigation of the time of greatest FFDI associated with fire-weather events that may be of use for planning purposes by fire agencies and in the design and interpretation of fire-weather climate studies. The time of maximum FFDI of an event is defined here as the first occurrence during a day of the peak value. This is not a trivial definition, because the peak FFDI can occur on a number of occasions. For example, one event in the Launceston Airport METAR dataset recorded peak FFDI of at least 24 on six occasions. A reasonable definition of “peak time” might be the midpoint of the times of peak FFDI. The time of first occurrence of the peak FFDI, however, was chosen because it was felt that this would be of particular significance for fire managers.

Ranges of values were examined for Hobart, Launceston, and Devonport Airports. In the case of Hobart Airport (Fig. 5a), four sets were plotted. These were the range of times of peak FFDI for days with that peak ≥12, 25, 38, and 50. Because Launceston Airport had fewer days of more severe fire weather (and no days of FFDI ≥50), only three ranges were plotted (Fig. 5b), these being days with peak FFDI that was greater than or equal to 12, 25, and 38. In a similar way, two ranges were plotted for Devonport Airport, those days with peak FFDI of ≥12 and ≥25.

Fig. 5.

Frequency analysis of times of peak FFDI for various FFDI ranges at (a) Hobart Airport, (b) Launceston Airport, and (c) Devonport Airport, plotted with a logarithmic y scale (to resolve better the plots displaying higher and lower ranges of FFDI).

Fig. 5.

Frequency analysis of times of peak FFDI for various FFDI ranges at (a) Hobart Airport, (b) Launceston Airport, and (c) Devonport Airport, plotted with a logarithmic y scale (to resolve better the plots displaying higher and lower ranges of FFDI).

In all cases, peak FFDI occurred in the afternoon. At Launceston Airport, the most common time of peak FFDI of ≥12, 25, and 38 was 1500 LCT in all cases but with a tail skewed slightly toward earlier in the day and no events with peak FFDI after 2000 LCT. Devonport Airport was very similar to Launceston Airport in behavior but with fewer days of high or very high FFDI, experiencing a broad peak for both ranges between 1200 and 1600 LCT.

Hobart Airport experienced a much wider range of peak FFDI times in corresponding ranges of values, essentially at any time during the day but with the most common time being 1400 LCT for FFDI of at least 12 and a slightly broader peak between 1300 and 1500 LCT for FFDI of ≥25. Of interest is the dip in the plot for FFDI ≥12 at 0800, for which there is no obvious solely meteorological reason. An examination of the raw data shows a decline from 0000 until 0800 in the number of days with peak FFDI during the early morning. It seems likely that this at least partly a consequence of the way peak daily FFDI is selected. As noted above, if there is more than one occurrence of the day’s highest FFDI, the earliest such occurrence is selected as the daily peak. During the early morning, in the absence of insolation and the resultant enhanced mixing and heating, it is less likely that a later time will record a higher FFDI. It is only after significant insolation has occurred that peak daily FFDI is likely to be recorded during any particular time period.

A trend toward earlier maxima with higher FFDI is suggested by the set of days with peak FFDI of at least 38, where time of peak FFDI was most commonly at 1300 LCT. A broad peak between 1200 and 1400 LCT, with a secondary peak at 1600 LCT, occurred for days with highest FFDI of ≥50, lending some support to the suggestion of earlier peak times for higher FFDI values.

4. Discussion and summary

Operational meteorologists and their fire and land management colleagues have generally been aware of the variability of fire-weather conditions during the course of a day, an awareness that this study goes some way toward quantifying. On some occasions in the past, however, failures to control fire outbreaks have resulted from fire managers being faced with difficult weather conditions at “unusual” times of the day. The outcomes of this study have highlighted to fire and land managers the potential in particular for “early” starts to a day’s fire weather, an awareness that has been translated into both training and operations. Tasmanian fire agencies have ensured, as a result, an increased weight of response to incidents earlier in a day in recent seasons than was previously the case (M. Chladil, Tasmania Fire Service, 2011, personal communication).

Information on the duration of fire-weather events is valuable in an ongoing discussion about the appropriate length of time that conditions can persist above warning thresholds before a fire-weather warning or alert is issued. At this time in Tasmania, any observation of FFDI above the warning threshold of 38 is considered to be justification for a warning. It is useful, though, to have an indication of the change in the number of warnings potentially issued should the warning criterion be amended to include a requirement for conditions to persist for a specified length of time.

This study raises the question of how, in forecasts for fire-control purposes, to express fire-weather variability most effectively. Fire-weather forecasts using the McArthur fire-danger system in particular have traditionally provided details of the conditions at the time of maximum temperature in the expectation that this will not be far from the period of peak FFDI. As an alternative, such forecasts have detailed explicitly the conditions at the time of peak fire danger. It is clear that both of these presentation methods offer a picture that is far from complete.

Hourly output from numerical weather models can give a very detailed forecast of fire weather over an area of responsibility but can present a very large amount of information for users to assimilate. One or, of more use, several summary measures of forecast fire weather may still be the most realistic way of communicating forecast conditions for as broad an area as an Australian state or territory for many users. Such summary measures can be computed easily from available hourly fields of fire weather from numerical weather prediction models. Perhaps the simplest such summary field is the peak FFDI over an area. Another field that the above discussion immediately suggests is that of time of peak FFDI. It is clear that, although the most common time of occurrence of peak fire weather is midafternoon, there is a wide variability in the timing of the peak, depending on location and the meteorological conditions of the day. A “time of maximum fire danger index” chart is very likely therefore to be of use to fire and land managers.

A potentially useful field that could be computed using hourly numerical weather prediction output is the integration of FFDI with time over an area. This is analogous to the ΣFFDI concept discussed by Beer et al. (1988) but with FFDI accumulated over 24 h (or a subset thereof) rather than annually. The value of the measure lies in highlighting areas where the FFDI is forecast to be elevated for substantial periods of time, summarized in a single image. These summary data can be presented at individual sites or, with full implementation of the Next Generation Forecasting System currently being deployed across Australia, as areal plots.

The availability of high-resolution time series of fire-weather observations has allowed examination of a number of characteristics of fire-weather events, including duration and peak times, in addition to expanding on earlier climatological summaries. The calculation of forest fire–weather conditions at high temporal resolution for three Tasmanian locations has also demonstrated that climatological summaries of fire weather using data solely at 1500 LCT, or even at all synoptic hours, are insufficient to resolve the true variability of fire weather at a location. At Hobart Airport, for example, examination of observations made at 1500 LCT alone will only resolve some 53% of days on which peak FFDI exceeded 38. This is important to bear in mind not only for safe fire management but for interpretation of climate studies. Although not the primary purpose of the study, the substantial and complex interannual variability of FFDI in Tasmania was also noted. This is an important issue in its own right for fire managers and forecasters and will be the subject of future research.

Acknowledgments

Mark Chladil of the Tasmania Fire Service provided helpful feedback on early results of this study and in the review process—in particular with regard to operational implications of the study for fire managers. Graham Mills of the Centre for Australian Weather and Climate Research and Kelvin Michael of the Institute for Marine and Antarctic Studies at the University of Tasmania insightfully reviewed drafts of the document. The editor and anonymous reviewers made a number of suggestions that helped to clarify and improve the focus of the paper. Ian Barnes-Keoghan and Doug Shepherd (retired) of the then Climate and Consultative Services Section, Tasmania and Antarctic Regional Office of the Australian Bureau of Meteorology, commenced digital archival of Tasmanian airport weather reports in 1990, well before this activity was undertaken nationally in Australia. In doing so, they enabled this study to be undertaken. Figure 1 was generated using the Jules Map Server of UNAVCO (http://jules.unavco.org/).

REFERENCES

REFERENCES
Beck
,
J. A.
,
M. E.
Alexander
,
S. D.
Harvey
, and
A. K.
Beaver
,
2002
:
Forecasting diurnal variations in fire intensity to enhance wildland firefighter safety
.
Int. J. Wildland Fire
,
11
,
173
182
.
Beer
,
T.
,
A. M.
Gill
, and
P. H. R.
Moore
,
1988
:
Australian bushfire danger under changing climatic regimes
.
Greenhouse. Planning for Climate Change, G. I. Pearman, Ed., CSIRO, 421–427
.
Fox-Hughes
,
P.
,
2008
:
A fire danger climatology for Tasmania
.
Aust. Meteor. Mag.
,
57
,
109
120
.
Griffiths
,
D.
,
1999
:
Improved formulae for the drought factor in McArthur’s forest fire danger meter
.
Aust. For.
,
62
,
202
206
.
Long
,
M.
,
2006
:
A climatology of extreme fire weather days in Victoria
.
Aust. Meteor. Mag.
,
55
,
3
18
.
Lucas
,
C.
,
2010
:
On developing a historical fire weather data-set for Australia
.
Aust. Meteor. Oceanogr. J.
,
60
,
1
13
.
McArthur
,
A. G.
,
1967
:
Fire behaviour in eucalypt forests
.
Forestry and Timber Bureau Leaflet 107, 25 pp
.
Mills
,
G. A.
,
2008
:
Abrupt surface drying and fire weather. Part 2: A preliminary synoptic climatology in the forested areas of southern Australia
.
Aust. Meteor. Mag.
,
57
,
311
328
.
Mills
,
G. A.
, and
S.
Pendlebury
,
2003
:
Processes leading to a severe wind-shear incident at Hobart Airport
.
Aust. Meteor. Mag.
,
52
,
171
188
.
Mount
,
A. B.
,
1972
:
The derivation and testing of a soil dryness index using run-off data
.
Forestry Commission of Tasmania Bulletin 4, 30 pp
.
Noble
,
I. R.
,
G. A. V.
Bary
, and
A. M.
Gill
,
1980
:
McArthur’s fire-danger meters expressed as equations
.
Aust. J. Ecol.
,
5
,
201
203
.
Ryan
,
B. C.
,
1977
:
A mathematical model for diagnosis and prediction of surface winds in mountainous terrain
.
J. Appl. Meteor.
,
16
,
571
584
.
Tolhurst
,
K.
, and
N. P.
Cheney
,
1999
:
Synopsis of the knowledge used in prescribed burning in Victoria
.
Department of Natural Resources and Environment, Melbourne, Rep., 97 pp
.
Verdon
,
D. C.
,
A. S.
Kiem
, and
S. W.
Franks
,
2004
:
Multi-decadal variability of forest fire risk—Eastern Australia
.
Int. J. Wildland Fire
,
13
,
165
171
.
Williams
,
A. A. J.
, and
D. J.
Karoly
,
1999
:
Extreme fire weather in Australia and the impact of the El Niño–Southern Oscillation
.
Aust. Meteor. Mag.
,
48
,
15
22
.
Williams
,
A. A. J.
,
D. J.
Karoly
, and
N.
Tapper
,
2001
:
The sensitivity of Australian fire danger to climate change
.
Climatic Change
,
49
,
171
191
,
doi:10.1023/A:1010706116176
.

Footnotes

1

Times used throughout the analysis are local clock time, to prevent any difficulties resulting from changeovers to and from daylight saving time, which occurred in Tasmania every October and March during the period covered by this study. Local clock time is Australian eastern standard time (10 h ahead of UTC) outside of daylight saving time and 11 h ahead of UTC during daylight saving time.

2

Following the Victorian bushfire disaster of 7 February 2009, Australian fire authorities updated the classification of McArthur (1967) at the beginning of the 2009/10 fire season, largely to distinguish degrees of severity above FFDI 50.