Victorian fire weather trends and variability

Abstract

Information on the trends and variability of fire weather indices and weather variables has a wide variety of applications to the development of systems to assist fire management activities including prevention, preparedness, response and recovery. Such information based on the observation record is constrained to the few observation sites that have sufficiently long record, while climate model produced climate datasets tend to lack the spatial resolution required by fire agencies. A bias-corrected, downscaled reanalysis climatology over Victoria (VicClim) ameliorates these limitations. Using the latest version of the VicClim gridded fire weather dataset, the spatial (~4km grid) and temporal (hourly) variability of fire weather for Victoria for 1972-2017 is explored, with McArthur's Forest Fire Danger Index (FFDI) representing fire weather. Various metrics are calculated: The 99th percentile of FFDI, probability of days with an FFDI over defined fire danger thresholds (High, Very High and Severe), and the decadal differences in these thresholds for all grid points. We also determine the Victorian averaged diurnal, seasonal and annual daily maximum FFDI, annual 90th percentile FFDI, and cumulative FFDI. The annual number of days with an FFDI above the fire danger thresholds are calculated, and finally we determine the variability of fire season length (FFDI days > 25) and changes to the start and end of the fire season. It is shown that there is a significant positive relationship between fire activity (annual number of fires, total area burned) and annual accumulated FFDI. There is large spatial variability across Victoria in extreme (99th percentile) FFDI with the highest values in the north west and the lowest values in the Alpine region. The variability also differs across seasons with the biggest spatial contrasts in FFDI occurring in spring and summer. There is broad interdecadal variability in the 90th percentile FFDI averaged across Victoria through the study period, with a neutral period through the 1970s to the early 1980s, a negative anomaly from the mid 1980s to the end of the 1990s, and a positive anomaly from the early 2000s through to 2017, with a short-lived but strong negative anomaly in 2010-11. The broad interdecadal variability is also evident when daily maximum FFDI is separated by seasons for summer, autumn and spring but the upward trend in the most recent decades found in those seasons is not evident in the winter. Diurnal differences in trend and variability were observed with greater year-to-year variability and an upward trend since the early 2000s in summer 3pm and 9pm FFDI values but less variability and no upward trend in the summer 3am and 9am FFDI. There are spatial variations in decadal changes in the number of days with an FFDI over 25 with some decades having one half of the state with above average number of days and the other half having below average number of days. Overall, the findings correspond with the Victoria averaged time series of below average number of days for each threshold in the 1990s and an above average number of days for each threshold in the 2000s. We find that the number of days with an FFDI over 25 for over 10% or more of the state each year (July to June) from 1972-73 to 2001-02 is on average 66 days. Since 2002-03 through to 2016-17 the average has increased to 94 days per year. There are only six fire seasons that have over 100 days with an FFDI>25 and these all occur between 2002-03 and 2016-17. Finally, when the first and last day of the season with an FFDI of 25 is compared with a selected calendar date (1st Sep and 1st May, respectively) we find that from 1972-73 through to 2001-02 there were five occurrences of the fire season starting earlier, whereas since 2002-03 there have been 10 years that have had an earlier start. This trend to an earlier start to the fire season is also supported by an increase in the number of days with an FFDI over 25 in the transition month at the beginning of the season (September) being found. There is no increase evident in the number of days from the end date of the season or in the transition month at the end of the season (April). The results from this study provide hitherto unavailable understanding of regional variability and trends in fire weather over Victoria for fire agency personnel to incorporate in the long term planning of resource allocation and landscape management.