Australian fire weather variability using calibrated long-term reanalysis data

doi:10.21203/rs.3.rs-2477059/v1

Download PDF

Research Article

Australian fire weather variability using calibrated long-term reanalysis data

https://doi.org/10.21203/rs.3.rs-2477059/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Spatiotemporal variability on multi-decadal time scales can be observed in weather conditions associated with wildfires. The climatic factors that drive this variability in fire weather conditions have been studied using bias-corrected 20th Century Reanalysis data of the mean state and extreme values of the McArthur Forest Fire Danger Index (FFDI) from 1876 to 2014. In particular, the association of this time series with the climate drivers of El Niño-Southern Oscillation (ENSO), Indian Ocean Dipole (IOD), Southern Annular Mode (SAM) and Interdecadal Pacific Oscillation (IPO) are analysed using several techniques. The potential for long-term changes in the relationships between the climate drivers and seasonal FFDI values is also explored by breaking the total time period into two epochs. Composite analysis was performed to separate the phases of ENSO, IOD and SAM and determine which phase combinations influence severe fire weather conditions the most. The most severe fire weather conditions were observed during a specific combination of ENSO, IOD and SAM phases, with results mapped to show spatial and seasonal variations in this over Australia. Long-term changes in the relationships between fire weather conditions and the three climate drivers (ESNO, IOD, SAM and IPO) are documented. These findings are based on a longer time period than similar previous studies, intended to help contribute to the range of guidance available for long-range fire weather outlooks and enhanced decision-making for fire agencies.

wildfire

El Niño Southern Oscillation (ENSO)

Indian Ocean Dipole (IOD)

Southern Annular Mode (SAM)

Interdecadal Pacific Oscillation (IPO)

climate

Wildfires (also known as bushfires in Australia) have been part of environmental systems for millions of years, and our understanding of them is rooted in the concept of a fire regime (Gill, 1975; Krebs et al., 2010). Since its inception, this concept of a fire regime has developed considerably and now typically incorporates the prevailing size, severity, type, frequency and seasonality of wildfires at a specific location. Australia (Fig. 1) is one of the most wildfire-prone continents of the world, though the fire regime varies internally (Gill and Catling, 2002; Williams, 2002), relating to variation in the different components that contribute to bushfire occurrence.

According to Bradstock (Bradstock, 2010), there are four hypothetical "switches" that must all be activated simultaneously for bushfires to occur. These four switches are biomass abundance (i.e., fuel load), fuel availability (i.e., suitably low moisture content to sustain a fire), ignition sources, and fire weather conditions that can influence the rate of fire spread and intensity. The primary focus of our paper is on fire weather conditions (including wind speed, humidity, temperature and rainfall, noting that these can also influence fuel availability in addition to fire spread).

In recent years, considerable efforts have been made to improve methods and techniques for long-term monitoring of fire weather conditions in Australia, including for use in climate analysis. These efforts have focussed primarily on environmental variables such as those relating to the 'switches' as described above, including creating long-term historical fire weather data records. Daily fire weather indices are often used to describe the combined influence of various meteorological factors and associated relevant fuel information on the risks associated with forest fires. These indices often combine rainfall, relative humidity, temperature, and wind speed, such as for the McArthur FFDI (McArthur, 1967), the United States National Fire Danger Ratings System (NFDRS) (Bradshaw et al., 1983) and the Canadian FWI System (Wagner, 1987) used in many different regions of the world. The FFDI is chosen for the purposes of this study as it is more widely used in Australia than other indices such as FWI or NFDRS. It is considered a useful metric that combines several weather variables known to influence fire behaviour, noting that the relative weighting of these variables may vary between different regions and fuel types, so the results are interpreted accordingly here. For example, FFDI values are interpreted in a general sense to indicate potentially dangerous combinations of weather conditions.

Many Australian regions have observed an increase in the severity of fire weather conditions over recent decades (Clarke et al. 2013; Dowdy, 2018; Harris & Lucas, 2019). This increased severity of fire weather conditions has been attributed, at least in part, to an overall increasing trend of hotter conditions, particularly since the mid-20th century, when the influence of anthropogenic warming has become more evident and rapid (Chiew et al., 2022). These hotter temperatures can also be associated with lower relative humidity, influencing fuel moisture content and ultimately leading to increased severity of fire weather conditions and longer fire seasons in recent decades (Dowdy, 2018; Harris and Lucas, 2019; Canadell et al., 2021).

Various modes of natural climatic factors have an influence on fire weather conditions in Australia, particularly the El Niño-Southern Oscillation (ENSO), Indian Ocean Dipole (IOD) and Southern Annular Mode (SAM). Several studies have shown that these are key climatic drivers that influence many aspects of the Australian climate, including bushfire conditions (Harris and Lucas, 2019; Abram et al., 2021; Squire et al., 2021). The primary aim of our study is to investigate the influence of each of the climatic drivers on Australian fire weather conditions since the late – nineteenth century (i.e., the 1870s) and also their synergistic relationship with fire weather conditions throughout Australia. In addition, we also examine the effects of longer-term variability due to Interdecadal Pacific Oscillation (IPO) on fire weather conditions over Australia (Power et al., 1999; Verdon et al., 2004; Harris and Lucas, 2019). The study aims to help contribute to the broader knowledge already existing around this topic of how climate modes can influence fire weather conditions based on using complementary methods and data to those used in previous studies (such as Harris and Lucas, 2019; Abram et al., 2021; Squire et al., 2021). In particular, it is acknowledged that considering a range of different methods and data sources can be beneficial when considering climate influences on extreme phenomena, including previous studies as well as future studies that could potentially use a range of other complementary data sets and approaches.

2.1 Observation-based Reference Dataset

A combination of datasets from the Australian Water Availability Project (AWAP) (Jones et al., 2009) and the National Centers for Environmental Prediction - National Center for Atmospheric Research (NCEP–NCAR) (Kalnay et al., 1996) is used in this study as a reference for calibrating the reanalysis product for long term analysis. This approach provides consistency with various other available datasets and studies, such as those detailed in Dowdy (2020). From the AWAP dataset, we have used the daily accumulated rainfall at 0900 local time (LT), daily maximum temperature and the relative humidity field calculated from the daily maximum temperature and vapour pressure at 1500 LT (which is the nearest available data around the approximate time of daily maximum temperature). The network of stations used for creating the gridded AWAP dataset for rainfall, temperature, and dew point temperature is sparse in certain regions across Australia (Jones et al., 2009). Therefore, care must be exercised when interpreting results from data-sparse areas (e.g. central-western Australia) (King et al. 2013). Wind speed data are derived from the NCEP–NCAR reanalysis products at 0600 UTC (Kalnay et al., 1996), bilinearly interpolated to 0.05 × 0.05 degrees for consistency with fire weather forecast data produced by the Australian Bureau of Meteorology as detailed in Dowdy (2018).

2.2 The 20th Century Reanalysis Dataset

The 20th Century Reanalysis dataset version 2c (20CRv2c), available at a spatial resolution of 2 x 2 degrees, is used for the time period 1851 to 2014. The use of this dataset for this study is intended to be a complementary line of evidence to consider together with other studies on how large-scale climate modes can influence fire weather in Australia, such as those previous studies discussed in the Introduction section based on analysis of observations and other reanalysis datasets (Harris and Lucas, 2019; Abram et al., 2021; Squire et al., 2021). The 20CRv2c is a reconstructed historic weather dataset generated by NOAA's Physical Sciences Laboratory (PSL) and Cooperative Institute for Research in Environmental Sciences (CIRES) at the University of Colorado, supported by the U.S. Department of Energy. Unlike many other reanalysis datasets which are constructed using climate modelling methods incorporating data from various sources such as observational data, satellite-based data, etc., the 20CRv2c is produced using modelling that only includes sea level pressure observations (Compo et al., 2011). In comparison to other recent reanalysis products, this approach has helped produce a more temporally uniform dataset (Compo et al., 2011), and is so considered optimal for long-term climate analysis.

The daily accumulated rainfall is calculated from the 20CRv2c using the 3-hourly data, while the daily maximum temperature is computed from the 6-hourly data. As for the wind speed, the daily values are calculated from the u and v components at 0600 UTC (in sync with the observation-based reference dataset). The relative humidity data were taken at 0600 UTC to be consistent with the reanalysis wind speed data timing.

2.3 El Niño-Southern Oscillation (ENSO)

ENSO characterises the interaction between the ocean and atmosphere in the tropical and subtropical Pacific, often described using spatio-temporal variations between below and above normal sea surface temperatures (SSTs). It has a wide range of influences on different components of our environment, including leading to a series of dry and wet conditions over Australia and elsewhere around the globe. It is usually described as having three phases: El Niño, La Niña and Neutral. El Niño refers to the anomalous warming of the central and eastern tropical Pacific Ocean, while La Niña is associated with below-average SST in that region (Trenberth, 1997). During the Neutral phase, there are little, or no anomalous warmings of SST and the conditions are similar to that of the mean climatology. Each phase has been reported to occur without any regular basis periodically, with occurrences on average approximately every 2–7 years. Previous studies (Dowdy, 2018; Harris and Lucas, 2019) have shown that the positive and negative phases of ENSO strongly influence the fire weather conditions in Australia, particularly from the spring to summer seasons. The Niño-3.4 index (Bamston et al., 1997) has been used in this study as a representation of ENSO conditions. Three-month averages are used here for DJF (December to January), MAM (March to May), JJA (June to August) and SON (September to October), with the Niño-3.4 data obtained from monthly indices (January 1870 to August 2022) (Smith, 2018) based upon the HadISST1 dataset as described in Rayner et al. (2003).

2.4 Indian Ocean Dipole (IOD)

The difference in SST anomaly gradient between the western and eastern parts of the Indian Ocean near the equator is defined as the Indian Ocean Dipole (Saji et al., 1999; BoM, 2012). A positive phase of the IOD deprives the western and southeastern parts of Australia of their winter rainfall quota (Ashok et al., 2003) and can lead to consecutive dry months before the next bushfire season in southern Australia (i.e., a relatively dry spring leading into the summer fire season). During its negative phase, the IOD can cause increased rainfall over the southern Australian region (Ummenhofer et al., 2009). The IOD influence on fire weather can combine with ENSO to have a compounding effect in some regions (Harris and Lucas, 2019; Abram et al., 2021). When these two phases occur simultaneously, it significantly affects southeast Australia, particularly during the spring season (Cai et al., 2009; Harris and Lucas, 2019). A positive IOD phase severely decreases the chances of rainfall over southeast Australia during spring and thus increases the risk of forest fire by lowering the moisture content in forest biomass (Cai et al., 2009). To study the influence of IOD on fire weather, we have used the Dipole Mode Index (DMI) (Saji et al., 1999). IOD is primarily relevant during winter and spring for influencing fire weather conditions in Australia (e.g., Harris and Lucas 2019), but for completeness here we used three-month average values for all four seasons (for DJF, MAM, JJA and SON) of DMI, obtained from the monthly indices (January 1870 to January 2022) (Smith, 2022) based on the dataset as described in Rayner et al. (2003).

2.5 Southern Annular Mode (SAM)

Southern Annular Mode (SAM) is associated with the large-scale meridional movement of atmospheric mass from the mid-latitudes to the high latitudes in the Southern Hemisphere (Gong and Wang, 1999). The influence of the SAM on Australian rainfall is dependent on season and region. During the main fire seasons of spring and summer the main signal is of increased rainfall in southern Australia during the positive phase of SAM, except for western Tasmania (Hendon et al., 2007; Meneghini et al., 2007; Hendon et al., 2014). During the austral spring and summer seasons, the negative phase of SAM promotes warm conditions over southeastern Australia (Abram et al., 2021), exacerbating the already dry conditions, which in turn can further influence the region's fire weather conditions. To establish the teleconnections between forest fire conditions in Australia and SAM, we have used the Antarctic Oscillation Index (AOI) as defined by Gong and Wang, (1999). Empirically, it is defined as the difference between zonal sea level pressure at 40°S and 65°S latitude bands (Gong and Wang, 1999). Three–month averages of AOI were calculated from the monthly indices (January 1851 to December 2011) (PSL, 2022) using data from20CRv2c.

2.6 Interdecadal Pacific Oscillation (IPO)

The Interdecadal Pacific Oscillation (IPO) is an indicator of interdecadal ocean-atmosphere climate variability in the Pacific. It is measured as the difference between the mean sea surface temperature anomaly (SSTA) near the central equatorial regions in the Pacific Ocean and the average of the SSTA in the northwest and southwest regions of the Pacific Ocean (Power et al., 1999; Henley et al., 2015). The Tripole Index (TPI) (Henley et al., 2015) has been used in this study to examine the potential for a relationship between IPO and forest fire conditions across Australia. This study used TPI data obtained from the National Oceanic and Atmospheric Administration (NOAA), using the HADISST1.1 filtered version of TPI (available from https://psl.noaa.gov/data/timeseries/IPOTPI/tpi.timeseries.hadisst11.filt.data, accessed January 2022). Just like the previous three instances, the three – months averages (DJF, MAM, JJA and SON) of the TPI were obtained from monthly indices (July 1876 to February 2016).

3.1 Calibration of the 20CRv2c dataset for FFDI computation

The 20CRv2c data was first bilinearly interpolated to the 0.05 × 0.05-degree grid, to be consistent with the observation-based reference dataset. The interpolated data needs to be bias-corrected first with respect to the observation-based input dataset before computing FFDI. Bias correction is done to help minimise the propagation of error from the input climatic variables into the calculated FFDI dataset. In this study we have used a non-parametric approach based on the cubic spline for quantile-quantile matching as it not based on any assumptions of linearity or any particular statistical distribution, with details as documented previously in Biswas et al. (2022).

The non-parametric probability distribution functions (PDF) were fitted with two sets of data obtained from the 20th Century reanalysis data and observed quantiles (Gudmundsson et al., 2012). Following Biswas et al., (2022), this bias correction approach can be defined as

$$\bigcup\limits_{{i={M_S}}}^{{{M_e}}} {{D_C}} =FUN\left( {\bigcup\limits_{{i={M_S}}}^{{{M_e}}} {{D_M},Quantile\left( {\bigcup\limits_{{i={O_S}}}^{{{O_e}}} {{D_O}} } \right)} ,Quantile\left( {\bigcup\limits_{{i={O_S}}}^{{{O_e}}} {{D_M}} } \right)} \right)$$

where D_C, D_O and D_M are the bias-corrected dataset, observation-based dataset and the 20th-century reanalysis dataset, respectively, while M_e, M_s, O_e and O_s are the 20CRv2c dataset's end date, 20CRv2c dataset's start date, observation-based dataset's end date and start date, respectively. The FUN represents the monotone piecewise cubic Hermite spline function we will use for the interpolation (Biswas et al., 2022).

Let the set of quantiles for the observation-based and 20CRv2c dataset over the training period to be used in the function FUN be X and Y, respectively, and the total 20CRv2c dataset to be corrected be x. A bin size of 450 has been used here for the quantile-quantile matching as evaluated in Biswas et al., (2022). The interpolating function used in the quantile-quantile matching for bias-correction is based on a monotone Hermite cubic spline, as demonstrated by Fritsch and Carlson (1980). In order to form the Hermite polynomial functions for a particular set of data, the monotonicity of X and Y needs to be checked first. Before proceeding further, let us define the slope of the line segment between the dataset X and Y to be ${\Delta _k}={{\left( {{Y_{k+1}} - {Y_k}} \right)} \mathord{\left/ {\vphantom {{\left( {{Y_{k+1}} - {Y_k}} \right)} {\left( {{X_{k+1}} - {X_k}} \right)}}} \right. \kern-0pt} {\left( {{X_{k+1}} - {X_k}} \right)}}$, where k = 1 to n, where n is the number of values in X and Y (Fritsch and Carlson, 1980). The derivatives ${d_k}$ were initialised such that $sign\left( {{d_k}} \right)=sign\left( {{d_{k+1}}} \right)=sign\left( {{\Delta _k}} \right)$k = 1 to n (Fritsch and Carlson, 1980; Biswas et al., 2022). To initialise the derivatives, we have used Heun's approximation method to calculate ${d_2}$, ${d_3}$, …,${d_{n - 1}}$ (Chapra and Canale, 2010), and for the end derivatives (i.e., ${d_1}$and${d_n}$), we have set the value equal to the slope (Biswas et al., 2022). If all the following three conditions are met, the function is outside the monotone region, and the derivative values need to be modified.

1. $\left( {2{\alpha _k}+{\beta _k} - 3} \right)>0$

2. $\left( {{\alpha _k}+2{\beta _k} - 3} \right)>0$

3. $3{\alpha _k}\left( {{\alpha _k}+{\beta _k} - 3} \right)<{\left( {2{\alpha _k}+{\beta _k} - 3} \right)^2}$

where,

$\alpha =\frac{{{d_k}}}{{{\Delta _k}}}$ and $\beta =\frac{{{d_{k+1}}}}{{{\Delta _k}}}$

The derivative values outside the monotone region were modified as follows:

${d_k}=3{\Delta _k}\frac{{{\alpha _k}}}{{\sqrt {{\alpha ^2}_{k}+{\beta ^2}_{k}} }}$ and ${d_{k+1}}=3{\Delta _k}\frac{{{\beta _k}}}{{\sqrt {{\alpha ^2}_{k}+{\beta ^2}_{k}} }}$

To perform the piecewise cubic interpolation using cubic Hermite Spline (Fritsch and Carlson, 1980) polynomials, the intervals between the dataset x and X were calculated first. Let this calculated interval be j. As for the cubic Hermite Spline function, we have used the 0-th derivative, i.e., the function itself for interpolation.

$$h(x)=\left[ {\frac{{{d_j}+{d_{j+1}} - 2{\Delta _j}}}{{{{\left( {{x_j} - {x_{j - 1}}} \right)}^2}}}} \right]{\left( {x - {X_j}} \right)^3}+\left[ {\frac{{ - 2{d_j} - {d_{j+1}}+3{\Delta _j}}}{{{x_j} - {x_{j - 1}}}}} \right]{\left( {x - {X_j}} \right)^2}+{d_j}\left( {x - {X_j}} \right)+{Y_j}$$

3.2 Calculation of FFDI

The daily values of FFDI (McArthur, 1967; Noble et al., 1980) have been calculated using bilinearly interpolated bias-corrected 20CRv2c data from 1851 to 2014, applying the same computing approaches of FFDI and its components as detailed in Dowdy (2018).

The daily FFDI is calculated here based on temperature t_max (℃), relative humidity rh (%), wind speed ws (km h^− 1), and the dimensionless drought factor df (Griffiths, 1999), as shown in Eq. (3). The FFDI formulation used here in Eq. (3) is a rearranged version to reduce computing time (Dowdy, 2018) compared to the more widely used version (Noble et al., 1980).

$$FFDI=\exp (0.0338 \times {t_{\hbox{max} }} - 0.0345 \times rh+0.0234 \times ws+0.243147) \times d{f^{0.987}}$$

The drought factor (df) (Griffiths, 1999) is calculated using the Keetch–Byram Drought Index (KBDI) (Keetch and Byram, 1968) and from the past 20 days' rainfall data (Finkele et al., 2006). It is formulated as

$$df=10.5 \times \left( {1 - \exp \left( {\frac{{ - KBDI - 30}}{{40}}} \right)} \right) \times \frac{{41{x^2}+x}}{{40{x^2}+x+1}}$$

where the variable x is a function of the past rainfall (p) and the number of days since the last rainfall (n),

$$x=\left\{ {\begin{array}{*{20}{c}} {\frac{{{n^{1.3}}}}{{{n^{1.3}}+p - 2}} n \leqslant 1 and p>2} \\ {\frac{{{{0.8}^{1.3}}}}{{{{0.8}^{1.3}}+p - 2}} n=0 and p>2} \\ {1 p<2} \end{array}} \right.$$

and KBDI is calculated using Eq. (6).

$$KBDI=KBD{I_{n - 1}} - {p_{eff}}+ET$$

In Eq. (6) $KBD{I_{n - 1}}$ denotes the previous days KBDI, while ${p_{eff}}$ (mm/day) signifies the effective rainfall. The effective rainfall is calculated by subtracting the surface runoff from the total daily rainfall. The surface runoff is considered to be the first 5 mm of rainfall between successive days with non-zero rainfall (Sullivan, 2001). The Daily Evapotranspiration (ET) in Eq. (6) is calculated as follows (Sullivan, 2001).

$$ET=\frac{{\left( {203.2 - KBD{I_{n - 1}}} \right)\left( {0.968 \times \exp \left( {0.0875 \times {t_{\hbox{max} }}+1.5552} \right) - 8.3} \right)}}{{1+10.88 \times \exp \left( { - 0.00173 \times {R_{annual}}} \right)}} \times {10^{ - 3}}$$

where, ${t_{\hbox{max} }}$ is the daily maximum temperature and ${R_{annual}}$ is the mean annual rainfall. Next, we repeated the bias-correction of the input variables and the subsequent FFDI calculation using different ensemble members from the 20CRv2c dataset. The mean of the calculated FFDI from these different ensembles is used in the following sections for further analysis to better remove any inherent statistical uncertainties in the FFDI data.

3.3 Correlation Analysis

The influence of various modes of natural climate variability (represented by their respective indices) on the Australian fire weather conditions is examined using a time series analysis of the seasonal FFDI values from 1876 to 2011 for SAM and 1876 to 2014 for others. We have examined various climate drivers' influence on the mean-state and extreme fire weather conditions across Australia. As for the mean-state fire weather conditions, we used seasonal means of FFDI for each year to determine the relationship with FFDI. While for the extreme fire weather conditions, the seasonal 90th percentile peak-over-threshold (POT) values (Lang et al., 1999; Villarini et al., 2011) of FFDI have been chosen. By peak over threshold (POT) values, we mean the number of days per season that the FFDI is above the seasonal 90th percentile (hereafter referred to as the extreme FFDI days), noting that the seasonal 90th percentile is computed just for an individual season (rather than based on data from all time throughout the year). Sample Pearson's correlation coefficient, r, has been used to study the statistical association between the seasonal indices and fire weather. Critical values for 0.05 significance level and n-2 degrees of freedom have been used to examine the significance of the correlations between FFDIs and ENSO, IOD and SAM. Similarly, for IPO we used critical values at a 0.05 significance level, but the degrees of freedom were adjusted to correctly account for the strong autocorrelation present in the IPO time series, based on scaling the degrees of freedom using the ratio of the time interval between data points to the decay time when autocorrelation drops to 1/e as detailed in Panofsky and Brier (1958).

4.1 Climatological characteristics of Australian bushfire conditions

The mean-state and the 90th percentile extreme FFDI values, calculated at each grid point across Australia for the entire study period 1876–2014 under consideration, are shown in Fig. 2.

Overall, this shows that the austral spring and summer seasons are when dangerous fire weather conditions generally occur throughout most of Australia, though the Northern Territory and parts of Western Australia and Queensland can also experience high FFDI values during the austral autumn and winter seasons. The coastal regions around the eastern parts of Australia are slightly more vulnerable to bushfire threats during austral spring than in summer, consistent with previous studies (Luke and McArthur 1978; Dowdy 2020).

However, it is important to note that year-to-year variability in the mean-state and extreme FFDI days across Australia can be considerably large. These spatio-temporal variations can be attributed to the variability in various climatic drivers that influence the Australian climate. In the next sections, we will look at the effects of various natural climate drivers that modulate FFDI conditions over Australia for different seasons.

4.2 Effects of natural climate variability

We examine the effects of various climate drivers, namely ENSO, IOD, SAM and IPO, on both the mean FFDI conditions as well as on the extreme conditions for all seasons across Australia. As dangerous fire weather conditions in Australia often occur during the austral spring and summer seasons, we primarily focus on these two seasons in this study. Also, to better understand how the relationship between climate drivers and FFDI has potentially evolved over time, we divided the study period into two epochs: the first epoch comprises the period from 1876–1945 and the second epoch, 1945–2014. This particular approach is intended to help enable statistical analysis of differences between the climatological averages of the first and second halves without making assumptions about physical changes in climate at a particular point.

4.2.1 ENSO-FFDI

ENSO plays a dominant role in modulating the extreme FFDI days (Figs. 3 and 4) and monthly mean FFDI conditions (Supplementary Figs. 1 and 2) in both spring and summer seasons, as well as in other seasons (Supplementary Figs. 3–6). In other words, a positive phase of ENSO increases the likelihood of dangerous fire weather conditions. However, when we examined the ENSO-FFDI relationship for the two separate epochs (i.e., 1876–1945 and 1945–2014), we found a considerable strengthening of the relationship in the second epoch for all seasons compared with the first epoch in many regions, such as shown in Figs. 3 and 4 for the spring and summer seasons, respectively.

During the austral spring season of the first epoch, a statistically significant positive correlation pattern could be observed over southeastern Australia (Fig. 3a). However, as we move into the second epoch, we see a statistically significant positive correlation over a much larger region of Australia (Fig. 3b). For the austral summer (DJF), a negative correlation with ENSO for parts of western Australia is the main feature apparent in the first epoch, while a positive correlation is apparent over eastern Australia and northern tropics during the second epoch (Fig. 4b). However, we note that the strength of the ENSO-FFDI relationship is relatively weak during the DJF season (Figs. 4a, b) compared with the SON season (Figs. 3a, b).

For the ENSO-FFDI relationship during the austral autumn, regions with a significant positive correlation are only around northeastern Australia (Supplementary Fig. 3b). Here we can again observe a similar strengthening of the ENSO-FFDI relationship during the second epoch compared to the first epoch. For the winter season (Supplementary Figs. 4a-4b), there is a substantial increase in the areas with a significant positive correlation in the southeastern parts of Australia during the second epoch compared with the first epoch. Though these areas have a significant positive correlation pattern, the values at these locations in the FFDI climatological maps (Figs. 2c – 2f) are the lowest among the four seasons.

4.2.2 IOD-FFDI

The IOD can influence fire weather conditions across Australia, though the magnitude of variability differs between seasons and regions (Figs. 3 and 4; Supplementary Figs. 1–6).

During the spring season (i.e., SON months), there is a positive correlation pattern between IOD and both mean FFDI and extreme FFDI days over most of Australia, noting that the statistically significant relationship is mainly around the southeastern Australian region (Fig. 3c and Supplementary Fig. 1c). This positive correlation strengthens and expands into south and western Australia in the second epoch (Fig. 3d and Supplementary Fig. 1d). For the DJF season, in the first epoch, there are significant positive FFDI-IOD correlations in parts of northern Australia, but during the second epoch, those regions are no longer indicated as significant, with the main region of significant correlations occurring in eastern Australia, with a negative sign (Figs. 4c, d and Supplementary Figs. 2c, d). This is likely to be associated with IOD having a more robust influence on Australian weather conditions during the austral spring in contrast to summer when the decaying phase of the IOD occurs (Zhao and Hendon, 2009; Hendon et al., 2012; Lim and Hendon, 2017).

During the first epoch of the austral autumn season, we observe a strong positive correlation between IOD and extreme FFDI days (Supplementary Fig. 3c), as well as between IOD and mean FFDI (Supplementary Fig. 5c), over entire Australia except for the western region around the coast. In the second epoch, a general weakening of the FFDI-IOD relationships could be observed (Supplementary Fig. 3d, 5d).

In the winter season, we see a similar positive correlation pattern over southern and western Australia (Supplementary Figs. 4c,4d, 6c, 6d) as we observed in the spring season (Figs. 3c, 3d and Supplementary Figs. 1c, 1d). This is anticipated as IOD is active during the austral winter and spring seasons (Zhao and Hendon, 2009; Hendon et al., 2012; Lim and Hendon, 2017). Note that even though we observe a strong significant positive correlation over the southern half of Australia during the winter, it does not necessarily indicate dangerous bushfire conditions. As we have previously highlighted, the mean extreme FFDI values over the aforementioned regions during the winter months are very low (Figs. 2c-2f), and so the results need to be interpreted in this context accordingly.

4.2.3 SAM-FFDI

SAM is another key climatic driver that affects fire weather conditions, particularly in the southern parts of Australia, where the negative phase of SAM is associated with reduced rainfall during austral spring and summer, and hence enhanced bushfire conditions (Harris and Lucas 2019).

For the spring season in the first epoch, there is a strong negative correlation between SAM and FFDI values (for both extreme and mean-state FFDI values) across most of Australia, noting exceptions around Victoria and Northern Territory (Fig. 3e and Supplementary Fig. 1e). In the second epoch, there is a significant negative correlation over most of the eastern half of Australia (Fig. 3f and Supplementary Fig. 1f). During the austral summer, there is a strong negative correlation in eastern and northern Australia, particularly during the first epoch and to a somewhat lesser degree during the second epoch (e.g., Fig. 4e, 4f).

For autumn (MAM), the SAM-FFDI correlations are negative in parts of eastern Australia during the first epoch (Supplementary Fig. 3e) and over WA in the second epoch (Supplementary Fig. 3f). For winter, there is only a relatively small region in northern Australia with significant positive correlations during the first epoch, while in the second epoch, there is a relatively large region with significant negative correlations across the eastern and central regions of Australia (Supplementary Fig. 4f).

4.2.4 IPO-FFDI

The IPO is examined here in relation to fire weather conditions. It is important to note that measures such as the TPI index used for monitoring IPO can have a considerable association with ENSO indices due to the nature of their derivations, such that we have first removed ENSO influences on the IPO using a linear regression approach based on seasonal values of TPI and Niño-3.4. The residual IPO data, referred to here as TPI*, can then be used to evaluate the relationship between IPO and FFDI that is not associated with the IPO-ENSO relationship. Moreover, due to the low frequency of variability, the TPI itself has very high autocorrelation (r = 0.96 for spring and summer time series used in this study), such that statistical degrees of freedom are adjusted here when evaluating the significance of the statistical relationship between IPO and FFDI as described in Section 3.3 (Panofsky and Brier, 1958).

Figure 5 shows correlations between TPI* and FFDI during the austral spring and summer presented individually for the first and second epochs of the study period. During the first epoch (Fig. 5a, c), we see a moderately high positive correlation pattern in eastern Australia with weaker or negative correlations in some western regions, noting that these correlations are not statistically significant (at the 95% confidence level used in this study). This lack of statistical significance relates to the strong autocorrelation of the IPO, which reduces the degrees of freedom, as discussed earlier. On the contrary, we notice a considerable weakening in the relationship in the correlations during the second epoch for spring and summer. It shows little resemblance to those of the first epoch for each of those seasons, with certain regions now having a weak positive correlation pattern in NSW and WA (Fig. 5b, d). The correlations are once again not statistically significant. Though the primary focus here is on the relationship between FFDI values and IPO during the austral spring and summer, we also included the results for autumn and winter (Supplementary Fig. 7).

4.3 Synergetic relationship between drivers

In this section, we focus on the combined effects of ENSO, IOD and SAM on Australian fire weather conditions. In particular, we look at eight different combinations resulting from two phases each for these three modes of variability (Fig. 6 and Supplementary Figs. 8–10 for the second epoch and Supplementary Figs. 11–14 for the entire period under consideration). Note that there are no exact definitions for different phases of these climate modes, with all approaches being somewhat arbitrary, with a number of approaches considered in this study. For the aims of this analysis, around examining different combinations of multiple modes of variability using a systematic approach, a large sample size is beneficial to help populate the various combinations as well as possible. Given this focus for the analysis here, we have used 'zero' threshold to separate each mode into their respective 'negative' and 'positive' phases. However, even after following this criterion, we find some combinations of climatic modes still have very few samples, and so the following results must be interpreted accordingly, particularly for cases with relatively low sample sizes. We also examined results using a more common approach for defining phases of these climate drivers using higher magnitude threshold values, as shown in Supplementary Fig. 15. Using that more restrictive approach, with fewer cases with data available for the phase combinations, the available results are broadly similar in sign to those shown using the more inclusive approach in Fig. 6.

During the austral spring season, extreme FFDI days in southeast Australia are shown to be more likely during positive IOD conditions, with a somewhat surprising result being that this is largely independent of the ENSO conditions (see Figs. 6). That result is based on 11 seasonal average values in total for positive IOD and SAM conditions (comprising 8 positive ENSO conditions and 3 negative ENSO conditions), such that results should be considered accordingly given that sample size and acknowledging that Australia can have substantial interannual variability in fire weather conditions. ENSO conditions have a larger influence on the FFDI values in central eastern Australia, with more frequent extreme FFDI occurring for positive ENSO conditions (based on the Niño-3.4), which corresponds to conditions more similar to El Niño rather than La Niña. These extreme FFDI conditions become even more likely if the combination of positive IOD and positive ENSO conditions occurs during the negative phase of SAM (Fig. 6e). We also see very high values of extreme bushfire conditions over Australia during a combination of negative ENSO, positive IOD and negative SAM (Fig. 6g), noting that there are only three realisations for this combination. We also notice considerably high extreme FFDI days over Australia during the co-occurrence of positive ENSO, positive IOD and negative SAM phases (Supplementary Fig. 8e) in the summer season. High extreme FFDI days are also noticed over eastern Australia during positive ENSO, negative IOD and negative SAM (Supplementary Fig. 8f). This result is somewhat consistent with the observed positive ENSO – FFDI, negative IOD – FFDI and negative SAM – FFDI correlation patterns during the summer months across eastern Australia (Fig. 4), as discussed in the previous section.

As for the austral autumn season, we again see a considerably high frequency of extreme FFDI days across Australia during positive ENSO, positive IOD and negative SAM combination (Supplementary Fig. 9e), particularly over the western regions. Similarly, the combination of a negative IOD and negative SAM also shows a high frequency of extreme FFDI days during summer across many regions of Australia (Supplementary Fig. 9f and Supplementary Fig. 9h). However, during the winter months, we see the highest frequency of FFDI extreme days during a combination of positive ENSO, positive IOD and positive SAM (Supplementary Fig. 10a) across western and southeastern Australia. This is closely followed by positive ENSO, negative IOD and negative SAM combination in terms of extreme FFDI values during the winter season (Supplementary Fig. 10f).

In conclusion, it can be said that the highest likelihood of dangerous bushfire conditions across Australia is during a combination of positive ENSO, positive IOD and negative SAM modes of variability. The regional features shown here can also be useful for practical fire management applications, including as guidance material for upcoming seasons, given that these modes of variability can sometimes be predictable several months in advance. A key feature of this study is that the long time period of the reanalysis data used here enables a large sample size to help produce small subsets for different combinations of modes of variability, with figures such as Fig. 6 helping to provide new insight on the combined influences of these key modes of variability on Australian fire weather conditions.

4.4 Independent Influence of each driver on the mean FFDI

There are interdependencies between different modes of climate variability, even to the extent that the climate signal of one mode may be contaminated by the other (e.g., as noted for the IPO-ENSO interdependencies detailed in Section 4.4). Therefore, in order to assess the relative roles between mean fire weather conditions and ENSO, IOD and SAM independently, multiple linear regression has been used (Snecdecor and Cochran, 1989; Lim et al., 2021). We first detrended the mean seasonal FFDI time-series and the key climate indices (i.e., Niño-3.4, DMI and AOI) before creating a multiple linear regression model for each grid point to prevent any contamination in the analysis arising from its inherent trend. This detrended data was normalised before creating the regression model so that all the input variables have unit variance (Snecdecor and Cochran, 1989) for easy comparison.

Multiple linear regression model during the spring season shows positive coefficients associated with ENSO (i.e., Nino 3.4 index) across parts of northern and inland-eastern Australia (Fig. 7a), indicating enhanced FFDI values in those regions during El Niño conditions. As for IOD (using DMI), positive coefficients occur across parts of southern Australia, particularly for Victoria in southeast Australia and in southern South Australia, with some regions of negative correlations in parts of central Queensland and eastern NSW (Fig. 7b). Negative correlations occur for SAM in some parts of eastern Australia and through central Australia into southern Western Australia, with positive correlation shown for parts of weather Australia and in the southeast, including Tasmania, as well as a thin strip along the central east coast into Queensland (Fig. 7c).

For summer, a positive coefficient of Nino 3.4 could be seen across many regions of Australia (Fig. 8a). For IOD in summer, the correlations are positive in western Australia and negative in parts of eastern Australia (Fig. 8b). For SAM in summer, the correlations are positive in sign for parts of South Australia and negative in sign across parts of eastern, northern and western Australia (Fig. 8c).

These patterns shown in Figs. 7 and 8 are broadly similar to our correlation patterns for mean seasonal FFDI and the respective climatic drivers (Supplementary Figs. 1–2). Correlation patterns between mean seasonal fire weather conditions and key climatic drivers during austral autumn and winter (Supplementary Figs. 5–6) are also found to be similar with multiple linear regression plots for the respective seasons (Supplementary Figs. 24–25).

Concluding from the multiple linear regression analysis across all four seasons, ENSO is an important climatic driver influencing the fire weather conditions over Australia, with IOD and SAM having significant independent influences too. In particular, our results demonstrate this quantitatively, providing details on how these influences vary spatially through Australia and between different seasons. This new insight is intended to be useful guidance for operational season prediction of fire danger in Australia by providing new insight from this long data set to help understand which climate driver (or drivers) are most important for influencing the fire weather conditions for a given region and season.

4.5 Environmental controls

The seasonal relationship between each of the input climatic variables (daily maximum temperature, wind speed, rainfall and relative humidity) for FFDI and the three key climatic drivers (i.e., ENSO, IOD and SAM) are analysed using a similar correlation technique (Figs. 9–12 and Supplementary Figs. 18–29). This enables us to understand the changing relationship between FFDI and the respective climatic drivers between the two epochs. But first, we validated the correlation patterns in the relationships with previous studies (Risbey et al., 2009; Abram et al., 2021) for similarity and consistency. The second emphasis is on how the relationship between these variables with the climatic driver changes in the second epoch and, consequently, their influence on FFDI.

During the austral spring and summer season, we notice stronger magnitude correlations in the second epoch compared to the first for temperature (Figs. 9a, 9b and Supplementary Figs. 18a, 18b), rainfall (Figs. 11a, 11b and Supplementary Figs. 20a, 20b), and relative humidity (Figs. 12a, 12b and Supplementary Figs. 21a, 21b). This is consistent with the enhancement of the FFDI-ENSO relationships from the first to the second epochs (Figs. 3a, 3b, 4a, 4b and Supplementary Figs. 1a, 1b, 2a, 2b), noting that the stronger magnitude positive correlations for temperature and stronger magnitude negative correlations for relative humidity and rainfall tend to contribute to the enhanced FFDI-ENSO relationships. There is little change indicated between the two epochs for the wind speed-ENSO relationships (Figs. 10a, 10b and Supplementary Figs. 19a, 19b).

During the spring season, a stronger positive correlation is seen for the IOD-temperature relationship from the first to the second epochs (Figs. 9c and 9d) and stronger negative correlations for the IOD relationships with wind speed (Figs. 10c and 10d) and relative humidity (Figs. 12c and 12d).

For SAM, a negative correlation is shown with temperature (Figs. 9e and 9f) over eastern Australia and a positive correlation with rainfall (Figs. 11e and 11f) and relative humidity (Figs. 12e and 12f). We also notice an enhancement of these relationships over Eastern Australia, along with the negative FFDI-SAM (Figs. 3e and 3f) relationship in the second epoch as compared to the first epoch. For wind speed-SAM relationships, the correlations are negative in southeast Australia, with some regions of positive correlations in northeast Australia near the coast (Figs. 10e, 10f), which is not similar to the correlations seen for FFDI-SAM in Figs. 3e and 3f, indicating that the other weather variables have a more dominant influence than wind speed on the FFDI-SAM relationship.

We have then repeated our analysis on autumn (Supplementary Figs. 22–25) and winter (Supplementary Figs. 26–29) as well. We again find temperature (Supplementary Figs. 22 & 26), rainfall (Supplementary Figs. 24 & 28) and relative humidity (Supplementary Figs. 25 & 29) have an influence on the relationship between the climatic drivers and FFDI in many cases (Supplementary Figs. 1–2 & 5–6). Similarly, here also, wind speed does not seem to influence the change between the epochs in fire weather for many locations across Australia (Supplementary Figs. 23 & 27).

This study explores the long-term variability of Australian fire weather in terms of its association with various climatic drivers. This has been achieved through the study of the relationship between seasonal mean values, extreme FFDI days and various climatic drivers across Australia. Several past studies examining the influence of ENSO, SAM and IOD on FFDI have also reported similar correlation patterns to some of those shown here (Cai et al., 2009; Dowdy, 2018; Harris and Lucas, 2019; Abram et al., 2021; Squire et al. 2022). Previous studies with overlapping time periods have also shown similarities in the relationship between the above-mentioned climatic drivers and the input climatic variables (Hendon et al. 2007; Riseby et al. 2009; Min et al., 2013; Marshall et al., 2014; White et al., 2014; Abram et al., 2021). However, the long time period provided by the reanalysis data used here allowed several novel aspects to be examined in this study to provide new insight into fire weather variability in Australia. Our key conclusions are as follows, summarised for the spring and summer periods individually.

Results for spring throughout Australia

High FFDI values are most likely during a combination of positive IOD with negative SAM in southeast Australia, as well as through Australia in general, when those conditions occur together with positive ENSO (e.g., Fig. 6), with these relationships associated with hotter and drier weather conditions.
IOD is the dominant driver for Victoria, South Australia and Tasmania, with positive correlations (e.g., Fig. 7).
ENSO is the dominant driver for NSW and QLD to the west of the Great Dividing Range and in the more northern regions of tropical Australia, with positive correlations.
SAM is the dominant driver in far western regions of the continent and for western Tasmania, with positive correlations, as well as regions of negative correlations for southeast Western Australia and along the Great Dividing Range in eastern Australia.
The drivers show relatively weak relationships for the Eastern Seaboard in general, but with some notable points. SAM has positive correlations along the QLD coast and negative correlations for the southern NSW coast; IOD has negative correlations for NSW coast; ENSO has positive correlations for the NSW and QLD coast (i.e., opposite in sign to the case on the other side of the Great Dividing Range).
IPO does not have a significant influence when the influence of ENSO is accounted for. This is consistent with a previous study, exploring the relationship between IPO and FFDI (Harris and Lucas, 2019).
Some indication of a strengthening of the relationships between FFDI and the drivers (e.g., Fig. 3), providing scope for research on long-term trends in fire weather conditions based on these data.

Results for summer in extratropical regions (rather than the tropics during the wet season)

High FFDI values are most likely through Australia for positive ENSO, associated with hotter and drier weather conditions, with an exception to this indicated for western Australia during negative SAM (e.g., Supplementary Fig. 8).
ENSO is the dominant driver in most of Victoria and western NSW with positive correlations, as well as for coastal NSW with negative correlations (e.g., Fig. 8).
SAM is the dominant driver for eastern Victoria as well as central and southeast NSW and parts of southern WA, with negative correlations. SAM is also the dominant driver for regions in SA, Tasmania and the NSW north coast, with positive correlations.
IOD shows relatively weak correlations through extratropical Australia in summer.
IPO does not have a significant influence when the influence of ENSO is accounted for. A previous study (Harris and Lucas, 2019), exploring the relationship FFDI-IPO relationship also reported that IPO does not have any significant influence on FFDI across Australia.

Further detailed work on the detection and attribution of long-term climate change trends in bushfire conditions across the Australian landscape is intended to be covered in subsequent research using this long-term bias-corrected FFDI data. The results presented here are intended to help provide a foundation for that subsequent research on climate trends. For example, this study found no significant relationship between IPO and FFDI in the first or second epoch after the ENSO influence was accounted for, which could help provide some confidence in long-term trends being associated with physical processes other than the IPO (e.g., human-caused climate change trends in weather conditions (IPCC 2021)).

The relationships with various climate drivers shown in this study are intended to help enhance guidance information for long-range forecasting of Australian fire weather, given that various climate drivers can be predicted several months in advance in some cases (Latif et al., 1998; Zhao et al., 2020; Liang et al., 2021; Weisheimer et al., 2022). Understanding the relationships between Australian fire weather and various climate drivers is one step forward toward enhanced long-range fire weather predictions, including at multi-week to season time scales, ultimately aiming to help enable more effective resource planning for fire risk management and increased resilience to fire impacts on society and the environment.

Acknowledgments

This research is supported by Henry Sutton PhD scholarship through Federation University, as well as through the funding from the Earth Systems and Climate Change Hub of the Australian Government's National Environmental Science Programme (NESP) Climate Systems Hub. Support for the Twentieth Century Reanalysis Project version 2c dataset is provided by the U.S. Department of Energy, Office of Science Biological and Environmental Research (BER), and by the National Oceanic and Atmospheric Administration Climate Program Office. Thanks are also due to Evan Dekker for his constant technical support.

Statements and Declarations

Funding

Competing Interests

The authors have no relevant financial or non-financial interests to disclose.

Author Contributions

Dr. Andrew J. Dowdy, Dr. Savin S. Chand and Soubhik Biswas contributed to the study conception and design. Material preparation, data collection and analysis were performed by Soubhik Biswas. The first draft of the manuscript was written by Soubhik Biswas, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Data Availability

The 20^th century Version 2c (20CRv2c) data are available from the NERSC Science Gateway https://portal.nersc.gov/project/20C_Reanalysis/. The datasets generated during the current study are not publicly available on any servers but are available from the corresponding author on reasonable request.

Abram NJ, Henley BJ, Sen Gupta A, Lippmann TJR, Clarke H, Dowdy AJ, Sharples JJ, Nolan RH, Zhang T, Wooster MJ, Wurtzel JB, Meissner KJ, Pitman AJ, Ukkola AM, Murphy BP, Tapper NJ, Boer MM (2021) Connections of climate change and variability to large and extreme forest fires in southeast Australia. Communications Earth & Environment 2:8. https://doi.org/10.1038/s43247-020-00065-8.
Ashok K, Guan Z, Yamagata T (2003) Influence of the Indian Ocean Dipole on the Australian winter rainfall. Geophysical Research Letters 30. https://doi.org/10.1029/2003GL017926.
Bamston AG, Chelliah M, Goldenberg SB (1997) Documentation of a highly ENSO‐related sst region in the equatorial pacific: Research note. Atmosphere-Ocean 35:367-383. https://doi.org/10.1080/07055900.1997.9649597.
Biswas S, Chand SS, Dowdy AJ, Wright W, Foale C, Zhao X, Deo A (2022) Statistical Calibration of Long-Term Reanalysis Data for Australian Fire Weather Conditions. Journal of Applied Meteorology and Climatology 61:729-758. https://doi.org/10.1175/jamc-d-21-0034.1.
BoM (2012) An analysis of the La Niña life cycle and the impacts and significance of the 2010–11 and 2011–12 La Niña events in Australia. Melbourne, Australia.
Bradshaw LS, Deeming JE, Burgan RE, Cohen JD (1983) The 1978 National Fire-Danger Rating System: Technical Documentation. General Technical Report Ogden, Utah. p 49.
Bradstock RA (2010) A biogeographic model of fire regimes in Australia: current and future implications. Global Ecology and Biogeography 19:145-158. https://doi.org/10.1111/j.1466-8238.2009.00512.x.
Cai W, Cowan T, Raupach M (2009) Positive Indian Ocean Dipole events precondition southeast Australia bushfires. Geophysical Research Letters 36. https://doi.org/https://doi.org/10.1029/2009GL039902.
Canadell JG, Meyer CP, Cook GD, Dowdy A, Briggs PR, Knauer J, Pepler A, Haverd V (2021) Multi-decadal increase of forest burned area in Australia is linked to climate change. Nature Communications 12:6921. https://doi.org/10.1038/s41467-021-27225-4.
Chapra SC, Canale RP (2010) Heun's Method. Numerical Methods for Engineers, 6 edn. McGraw-Hill. pp 720-724.
Chiew F, Costello MJ, Hennessy K, Lansbury N, Nidumolu UB, Pecl G, Rickards L, Tapper N, Woodward A, Wreford A (2022) Australasia. Climate Change 2022: Impacts, Adaptation and Vulnerability. Cambridge University Press, Cambridge, UK and New York, NY, USA. pp 1581–1688.
Clarke, H., Lucas, C. and Smith, P., 2013. Changes in Australian fire weather between 1973 and 2010. International Journal of Climatology, 33(4), pp.931-944.
Compo GP, Whitaker JS, Sardeshmukh PD, Matsui N, Allan RJ, Yin X, Gleason BE, Vose RS, Rutledge G, Bessemoulin P, Brönnimann S, Brunet M, Crouthamel RI, Grant AN, Groisman PY, Jones PD, Kruk MC, Kruger AC, Marshall GJ, Maugeri M, Mok HY, Nordli Ø, Ross TF, Trigo RM, Wang XL, Woodruff SD, Worley SJ (2011) The Twentieth Century Reanalysis Project. Quarterly Journal of the Royal Meteorological Society 137:1-28. https://doi.org/10.1002/qj.776.
Dowdy AJ (2014) Long-term changes in Australian tropical cyclone numbers. Atmospheric Science Letters 15:292-298. https://doi.org/https://doi.org/10.1002/asl2.502.
Dowdy AJ (2018) Climatological Variability of Fire Weather in Australia. Journal of Applied Meteorology and Climatology 57:221-234. https://doi.org/10.1175/JAMC-D-17-0167.1.
Dowdy AJ (2020) Seamless climate change projections and seasonal predictions for bushfires in Australia. Journal of Southern Hemisphere Earth Systems Science 70:120-138. https://doi.org/https://doi.org/10.1071/ES20001.
Finkele K, Mills GA, Beard G, Jones DA (2006) National gridded drought factors and comparison of two soil moisture deficit formulations used in prediction of Forest Fire Danger Index in Australia. Australian Meteorological Magazine 55:183-197,
Fritsch FN, Carlson RE (1980) Monotone Piecewise Cubic Interpolation. SIAM Journal on Numerical Analysis 17,
Gill AM (1975) Fire and The Australian Flora: A Review. Australian Forestry 38:4-25. https://doi.org/10.1080/00049158.1975.10675618.
Gill AM, Catling PC (2002) Fire regimes and biodiversity of forested landscapes of southern Australia In: Bradstock RA, Williams JE, Gill MA (eds). Flammable Australia : the fire regimes and biodiversity of a continent. Cambridge University Press. pp 49-75.
Gong D, Wang S (1999) Definition of Antarctic Oscillation index. Geophysical Research Letters 26:459-462. https://doi.org/10.1029/1999GL900003.
Griffiths D (1999) Improved Formula for the Drought Factor in McArthur's Forest Fire Danger Meter. Australian Forestry 62:202-206,
Gudmundsson L, Bremnes JB, Haugen JE, Skaugen TE (2012) Downscaling RCM precipitation to the station scale using statistical transformations – a comparison of methods. Hydrology and Earth System Sciences 16:3383–3390. https://doi.org/10.5194/hess-16-3383-2012.
Harris S, Lucas C (2019) Understanding the variability of Australian fire weather between 1973 and 2017. PLoS ONE 14. https://doi.org/10.1371/journal.pone.0222328.
Henley BJ, Gergis J, Karoly DJ, Power S, Kennedy J, Folland CK (2015) A Tripole Index for the Interdecadal Pacific Oscillation. Climate Dynamics 45:3077-3090. https://doi.org/10.1007/s00382-015-2525-1.
Hendon, H.H., Thompson, D.W. and Wheeler, M.C., 2007. Australian rainfall and surface temperature variations associated with the Southern Hemisphere annular mode. Journal of Climate, 20(11), 2452-2467.
Hendon HH, Lim E-P, Nguyen H (2014) Seasonal Variations of Subtropical Precipitation Associated with the Southern Annular Mode. Journal of Climate 27:3446-3460. https://doi.org/10.1175/jcli-d-13-00550.1.
IPCC (2021). Summary for Policymakers. In: Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)].Jones DA, Wang W, Fawcett R (2009) High-quality spatial climate data-sets for Australia. Australian Meteorological and Oceanographic Journal 58:233-248,
Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo KC, Ropelewski C, Wang J, Leetmaa A, Reynolds R, Jenne R, Joseph D (1996) The NCEP/NCAR 40-Year Reanalysis Project. Bulletin of the American Meteorological Society 77:437-472. https://doi.org/10.1175/1520-0477(1996)077<0437:Tnyrp>2.0.Co;2.
Keetch JJ, Byram GM (1968) A Drought Index for Forest Fire Control. Forest Service Research Paper 38, https://www.fs.usda.gov/treesearch/pubs/40.
King AD, Alexander LV, Donat MG (2013). The efficacy of using gridded data to examine extreme rainfall characteristics: a case study for Australia. International Journal of Climatology, 33(10), 2376-2387.
Krebs P, Pezzatti GB, Mazzoleni S, Talbot LM, Conedera M (2010) Fire regime: history and definition of a key concept in disturbance ecology. Theory in Biosciences 129:53-69. https://doi.org/10.1007/s12064-010-0082-z.
Lang M, Ouarda TBMJ, Bobée B (1999) Towards operational guidelines for over-threshold modeling. Journal of Hydrology 225:103-117. https://doi.org/10.1016/S0022-1694(99)00167-5.
Latif M, Anderson D, Barnett T, Cane M, Kleeman R, Leetmaa A, O'Brien J, Rosati A, Schneider E (1998) A review of the predictability and prediction of ENSO. Journal of Geophysical Research: Oceans 103:14375-14393. https://doi.org/10.1029/97jc03413.
Liang XS, Xu F, Rong Y, Zhang R, Tang X, Zhang F (2021) El Niño Modoki can be mostly predicted more than 10 years ahead of time. Scientific Reports 11:17860. https://doi.org/10.1038/s41598-021-97111-y.
Lim E-P, Hendon HH (2017) Causes and Predictability of the Negative Indian Ocean Dipole and Its Impact on La Niña During 2016. Scientific Reports 7:12619. https://doi.org/10.1038/s41598-017-12674-z.
Lim E-P, Hendon HH, Butler AH, Thompson DWJ, Lawrence ZD, Scaife AA, Shepherd TG, Polichtchouk I, Nakamura H, Kobayashi C, Comer R, Coy L, Dowdy A, Garreaud RD, Newman PA, Wang G (2021) The 2019 Southern Hemisphere Stratospheric Polar Vortex Weakening and Its Impacts. Bulletin of the American Meteorological Society 102:E1150-E1171. https://doi.org/10.1175/bams-d-20-0112.1.
Luke, R. H., and McArthur, A. G. (1978). Bushfires in Australia. Australian Government Publishing Service, ISBN 0642039909.
Marshall AG, Hudson D, Wheeler MC, Alves O, Hendon HH, Pook MJ, Risbey JS (2014) Intra-seasonal drivers of extreme heat over Australia in observations and POAMA-2. Climate Dynamics 43:1915-1937. https://doi.org/10.1007/s00382-013-2016-1.
McArthur AG (1967) Fire behaviour in eucalypt forests. Australia Forestry and Timber Bureau : Leaflet ; no 107. Forestry and Timber Bureau, Canberra. p 36.
Meneghini B, Simmonds I, Smith IN (2007) Association between Australian rainfall and the Southern Annular Mode. International Journal of Climatology 27:109-121. https://doi.org/10.1002/joc.1370.
Min S-K, Cai W, Whetton P (2013) Influence of climate variability on seasonal extremes over Australia. Journal of Geophysical Research: Atmospheres 118:643-654. https://doi.org/https://doi.org/10.1002/jgrd.50164.
Noble IR, Gill AM, BARY GAV (1980) McArthur's fire‐danger meters expressed as equations. Australian Journal of Ecology 5:201-203. https://doi.org/10.1111/j.1442-9993.1980.tb01243.x.
Panofsky HA, Brier GW (1958) Some Applications of Statistics to Meteorology. Mineral Industries Extension Services, College of Mineral Industries, Pennsylvania State University
Piani C, Weedon GP, Best MJ, Gomes SM, Viterbo PA, Hagemann S, Haerter JO (2010) Statistical bias correction of global simulated daily precipitation and temperature for the application of hydrological models. Journal of Hydrology 395:199-215. https://doi.org/10.1016/j.jhydrol.2010.10.024.
Power S, Casey T, Folland C, Colman A, Mehta V (1999) Inter-decadal modulation of the impact of ENSO on Australia. Climate Dynamics 15:319-324. https://doi.org/10.1007/s003820050284.
PSL N (2022) Southern Annular Mode (SAM). NOAA Physical Sciences Laboratory https://psl.noaa.gov/data/20thC_Rean/timeseries/monthly/SAM/ Accessed July 23 2022.
Rayner NA, Parker DE, Horton EB, Folland CK, Alexander LV, Rowell DP, Kent EC, Kaplan A (2003) Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. Journal of Geophysical Research: Atmospheres 108. https://doi.org/10.1029/2002JD002670.
Risbey JS, Pook MJ, McIntosh PC, Wheeler MC, Hendon HH (2009) On the Remote Drivers of Rainfall Variability in Australia. Monthly Weather Review 137:3233-3253. https://doi.org/10.1175/2009mwr2861.1.
Saji NH, Goswami BN, Vinayachandran PN, Yamagata T (1999) A dipole mode in the tropical Indian Ocean. Nature 401:360-363. https://doi.org/10.1038/43854.
Smith C (2018) Niño 3.4 SST Index. NOAA-ESRL Physical Sciences Laboratory https://psl.noaa.gov/gcos_wgsp/Timeseries/Nino34/ Accessed 23 July 2022.
Smith C (2022) Dipole Mode Index (DMI). NOAA-ESRL Physical Sciences Laboratory https://psl.noaa.gov/gcos_wgsp/Timeseries/DMI/ Accessed 23 July 2022.
Snecdecor GW, Cochran WG (1989) Multiple Linear Regression. In: Wiley-Blackwell (ed). Statistical Methods,
Squire DT, Richardson D, Risbey JS, Black AS, Kitsios V, Matear RJ, Monselesan D, Moore TS, Tozer CR (2021) Likelihood of unprecedented drought and fire weather during Australia's 2019 megafires. npj Climate and Atmospheric Science 4:64. https://doi.org/10.1038/s41612-021-00220-8.
Sullivan A (2001) Review of the Operational Calculation of McArthur's Drought Factor. Canberra, A.C.T. p 46.
Teutschbein C, Seibert J (2012) Bias correction of regional climate model simulations for hydrological climate-change impact studies: Review and evaluation of different methods. Journal of Hydrology 456-457:12-29. https://doi.org/10.1016/j.jhydrol.2012.05.052.
Trenberth KE (1997) The Definition of El Niño. Bulletin of the American Meteorological Society 78:2771-2778. https://doi.org/10.1175/1520-0477(1997)078<2771:Tdoeno>2.0.Co;2.
Ummenhofer CC, England MH, McIntosh PC, Meyers GA, Pook MJ, Risbey JS, Gupta AS, Taschetto AS (2009) What causes southeast Australia's worst droughts? Geophysical Research Letters 36. https://doi.org/10.1029/2008GL036801.
Verdon DC, Kiem AS, Franks SW (2004) Multi-decadal variability of forest fire risk - eastern Australia. International Journal of Wildland Fire 13:165-171. https://doi.org/10.1071/WF03034.
Villarini G, Smith JA, Ntelekos AA, Schwarz U (2011) Annual maximum and peaks-over-threshold analyses of daily rainfall accumulations for Austria. Journal of Geophysical Research: Atmospheres 116. https://doi.org/10.1029/2010JD015038.
Wagner CEV (1987) Development and structure of the Canadian Forest Fire Weather Index System. Forestry Technical Report, Ottawa. p 35.
Weisheimer A, Balmaseda MA, Stockdale TN, Mayer M, Sharmila S, Hendon H, Alves O (2022) Variability of ENSO Forecast Skill in 2-Year Global Reforecasts Over the 20th Century. Geophysical Research Letters 49:e2022GL097885. https://doi.org/https://doi.org/10.1029/2022GL097885.
White CJ, Hudson D, Alves O (2014) ENSO, the IOD and the intraseasonal prediction of heat extremes across Australia using POAMA-2. Climate Dynamics 43:1791-1810. https://doi.org/10.1007/s00382-013-2007-2.
Williams RJ (2002) Fire regimes and biodiversity in the savannas of northern Australia. In: Bradstock RA, Williams JE, Gill MA (eds). Flammable Australia : the fire regimes and biodiversity of a continent. Cambridge University Press. pp 49-75.
Zhao M, Hendon HH (2009) Representation and prediction of the Indian Ocean dipole in the POAMA seasonal forecast model. Quarterly Journal of the Royal Meteorological Society 135:337-352. https://doi.org/https://doi.org/10.1002/qj.370.
Zhao S, Stuecker MF, Jin F-F, Feng J, Ren H-L, Zhang W, Li J (2020) Improved Predictability of the Indian Ocean Dipole Using a Stochastic Dynamical Model Compared to the North American Multimodel Ensemble Forecast. Weather and Forecasting 35:379–399. https://doi.org/10.1175/WAF-D-19-0184.1.

No competing interests reported.

SupplementaryFigures.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Australian fire weather variability using calibrated long-term reanalysis data

Status:

Version 1

Abstract

Figures

1 Introduction

2 Data

2.1 Observation-based Reference Dataset

2.2 The 20th Century Reanalysis Dataset

2.3 El Niño-Southern Oscillation (ENSO)

2.4 Indian Ocean Dipole (IOD)

2.5 Southern Annular Mode (SAM)

2.6 Interdecadal Pacific Oscillation (IPO)

3 Methodology

3.1 Calibration of the 20CRv2c dataset for FFDI computation

3.2 Calculation of FFDI

3.3 Correlation Analysis

4 Results

4.1 Climatological characteristics of Australian bushfire conditions

4.2 Effects of natural climate variability

4.2.1 ENSO-FFDI

4.2.2 IOD-FFDI

4.2.3 SAM-FFDI

4.2.4 IPO-FFDI

4.3 Synergetic relationship between drivers

4.4 Independent Influence of each driver on the mean FFDI

4.5 Environmental controls

5 Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1