Fire on a tropical floodplain: a fine-scale fire history of coastal floodplains in the Northern Territory, Australia
Robin Leppitt A B * , Jay Evans A C D , Luke Einoder A E , Peter M. Kyne A , John C. Z. Woinarski A B and Stephen T. Garnett A BA Research Institute for the Environment and Livelihoods, Charles Darwin University, Darwin, NT 0909, Australia.
B Threatened Species Recovery Hub, National Environmental Science Program, Australia.
C Darwin Centre for Bushfire Research, Research Institute for Environment and Livelihoods, Charles Darwin University, Darwin, NT 0909, Australia.
D Bushfire and Natural Hazards Cooperative Research Centre, East Melbourne, Vic. 3002, Australia.
E Kakadu National Park, Jabiru, NT 0886, Australia.
International Journal of Wildland Fire 32(8) 1200-1211 https://doi.org/10.1071/WF22173
Submitted: 4 August 2022 Accepted: 16 May 2023 Published: 20 June 2023
© 2023 The Author(s) (or their employer(s)). Published by CSIRO Publishing on behalf of IAWF. This is an open access article distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND)
Abstract
Background: The coastal floodplains of northern Australia are fire-prone, but the impact of fire on floodplain biota is not well understood.
Aims: In this study, we sought to characterise the fire history of six adjacent floodplains in coastal Northern Territory, Australia.
Methods: We built a fine-scale 31-year fire history (1988–2018) to compare fire regimes on floodplains across land tenures and floodplain fire regimes with the surrounding savanna fire regime, determine the extent to which current fire regimes are meeting existing ecological fire thresholds, and investigate the relationship between rainfall totals and the extent of burning.
Key results: Floodplains in conservation reserves burnt more frequently than those on pastoral lands, and savannas burnt more often than floodplains. Current floodplain fire regimes comfortably meet existing ecological fire thresholds. The proportion of floodplain burning is inversely proportional to the amount of rain in the previous wet season.
Conclusions: Floodplain fire regimes vary markedly between land-use types, and floodplain fire regimes differ to those of savannas. The current management thresholds for floodplain fire regimes would benefit from further evidence of conservation outcomes.
Implications: For more effective floodplain fire management, research is needed to generate floodplain-specific thresholds that best conserve their considerable conservation value.
Keywords: fire, fire management, floodplain, invasive grasses, Kakadu National Park, northern Australia, traditional burning, tropics.
Introduction
Changes to the fire regimes of northern Australia as a result of European settlement in the last two centuries are one of the primary drivers in the decline of many components of its vertebrate fauna and some plant species (Fraser et al. 2003; Franklin et al. 2005; Woinarski et al. 2011; Bowman et al. 2022). In an attempt to halt or reverse this decline, there has been a concerted effort to quantify the effect of various fire regimes on biodiversity to improve fire management practices. Most northern Australian fire research has focused on savanna woodlands (Andersen et al. 1998; Russell-Smith et al. 2003; Evans and Russell-Smith 2020), which are the dominant vegetation type of monsoonal northern Australia and are the most fire-prone biome on earth (Andersen et al. 2012). To a lesser extent, other fire-sensitive habitat types in northern Australia have also been the subject of research (e.g. monsoon forest, Bowman and Fensham 1991; Callitris intratropica, Bowman and Panton 1993; sandstone plateau, Russell-Smith et al. 1998). In contrast, very little is known about the effects of fire on northern Australia’s tropical floodplains, notwithstanding their substantial extent and considerable ecological, cultural, and economic value (Lucas et al. 1997; Dyer et al. 2001; Finlayson et al. 2006).
The World Heritage-listed Kakadu National Park (NP) incorporates floodplains including Australia’s second largest Ramsar-listed wetland as well as multiple globally recognised Key Biodiversity Areas (KBAs). Much of the conservation value placed upon floodplains stems from their unusually diverse and abundant waterbird, amphibian, and reptile populations, and extremely productive wetland systems (Bayliss and Yeomans 1990; Finlayson 2005; Finlayson et al. 2006; Madsen et al. 2006a, 2006b). The main land-uses of these floodplains are pastoralism, conservation, and Indigenous cultural practices. Outside conservation areas, the primary industry on the floodplain is beef production, with the floodplains’ water-holding capacity providing greener pastures for a greater part of the year than that of the surrounding savanna.
Northern Australian floodplains are seasonally inundated in the wet season (broadly November–May) when high rainfall and temperatures trigger rapid production of grassy biomass. The floodplains subsequently dry out through the dry season (July–October), when the biomass cures and often burns. Wet season rainfall totals can affect the extent of burning in the subsequent dry season in some environments (Heinl et al. 2006; Archibald et al. 2010), but the extent to which they affect fire on northern Australian floodplains is not well understood. As with the surrounding savanna, fire can have important effects on the structure and floristics of floodplain vegetation, and hence biodiversity more generally, over short and long temporal scales (Heim et al. 2019).
The vegetation of the northern Australian floodplains has probably been influenced by human-induced fire regimes since their formation. These floodplains formed when rising sea levels, following thawing from the last ice age, stabilised around 6000–7000 years ago (Woodroffe et al. 1987). The fire regimes practised by Indigenous peoples over millennia are considered responsible for northern Australian floodplains being a largely treeless, grassy expanse – as are the floodplains’ geomorphological origins (Lucas et al. 1997). Indigenous fire regimes seek to improve access to natural resources through patchy, seasonal burning (Roberts 1993). Indigenous burning is still practised on some floodplains of northern Australia (McGregor et al. 2010). Fire continues to shape floodplain vegetation (Cowie et al. 2000), with the timing and size of fire being one of the dominant drivers of the highly variable annual floodplain vegetation dynamic (Kleindl et al. 2015). Fire regimes influence perennial/annual vegetation competition and vegetation homogeneity, particularly of the grass community (Cowie et al. 2000; Boyden et al. 2003), and the cessation of fire management may lead to increased woody vegetation on floodplains (Lucas et al. 1997; Crowley et al. 2009).
Floodplain fire regimes in northern Australia have undergone significant changes since European settlement. Pastoral and tourism ventures have displaced, and continue to displace, Indigenous peoples from their ancestral lands, preventing or supplanting traditional burning (McGregor et al. 2010). The introduction and subsequent spread of the Asian water buffalo (Bubalus bubalis Linnaeus, 1758) has led to widespread degradation of the floodplains, particularly during their population peak in the second half of the 20th century (Skeat et al. 1996; Werner 2014). The reduction in grassy biomass associated with high density Asian water buffalo populations led to a reduction in the frequency and extent of floodplain fire (Roberts 1993). Similarly, over-grazing by domestic cattle can lead to a reduction in fuel loads (Sharp and Whittaker 2003). Although widespread culling of Asian water buffalo in the late 1980s saw fire return to the floodplains (Russell-Smith et al. 1997), Asian water buffalo numbers have increased again in some areas in recent decades (Robinson and Whitehead 2003; Petty et al. 2007; Campbell et al. 2021). Exotic pasture grasses introduced on pastoral stations now occur in some floodplain areas as extensive monocultures contributing to more intense and larger fires compared with native grasses (Cowie et al. 2000; Douglas and O’Connor 2004). Fire can also be used to break up monocultures of some invasive plant species, with varying success (Lonsdale and Miller 1993; Grice et al. 2010; McGregor et al. 2010). Despite these changes, the extent to which land tenure, vegetation composition, and invasive weeds and animals have influenced floodplain fire regimes is not well understood.
There have been few studies that have addressed fire management objectives for northern Australian floodplains. One study of the habitat preferences of the Alligator Rivers Yellow Chat (Epthianura crocea tunneyi Matthews, 1912), an endangered bird endemic to the floodplains of the Van Diemen Gulf coast of the Northern Territory, found that the chats prefer floodplain habitat that has not burnt for five or more years (Leppitt et al. 2022). This represents the only fire metric recommendation for the conservation of northern Australian floodplain biodiversity of which we are aware, and it pertains to only one floodplain species. In northern Australia, several studies of the fire history of Kakadu NP have included floodplains (Russell-Smith et al. 1997; Gill et al. 2000; Edwards et al. 2003). Other than the recommendations on fire management for the Yellow Chat, and in the absence of any detailed floodplain-specific interpretations of the fire metrics presented or knowledge of the relationships of floodplain conservation values with fire, the only published assessment of floodplain fire performance thresholds pertained to fire-driven habitat heterogeneity: (1) at least 25% of floodplains remain unburnt for 3+ years; and (2) at least 5% of floodplains remain unburnt for 10+ years. These thresholds were taken from a recommendation originally developed for savanna habitats (Woinarski and Legge 2013; Woinarski and Winderlich 2014; Russell‐Smith et al. 2017). The extent to which savanna fire studies can be applied to floodplain fire management is unknown, but these environments have obvious differences in vegetation structure (floodplains being largely treeless) and hence fuel loads, and hydrology (floodplains undergo significant inundation in the wet season, which delays the curing of vegetation, and therefore become flammable later in the dry season than savannas). They also differ markedly in species composition and in conservation values.
In this study, we use fine-scale fire-mapping techniques to build a 31-year fire history (1988–2018) across six adjacent floodplains of varying land tenure in coastal Northern Territory, Australia. Using this approach, we provide a broad-scale assessment of floodplain fire regimes in the region to address the following aims: (1) compare fire regimes on floodplains managed for conservation with those floodplains managed for pastoralism; (2) compare fire regimes in floodplains and the contiguous savannas; (3) determine the extent to which current fire regimes are meeting existing thresholds for the floodplain vegetation community (Russell‐Smith et al. (2017) and key species (e.g. Yellow Chat; Leppitt et al. (2022)); and (4) investigate the relationship between wet season rainfall totals and the subsequent extent of fire on the floodplains. The results of this study will be valuable for managing fire on these globally significant floodplains.
Methods
Study area
Floodplains of six adjacent northern Australian rivers were selected for fine-scale fire mapping. These floodplains were chosen because they encompass a relatively even proportion of conservation and pastoral tenures and contain high-value conservation areas (Ramsar-listed Wetlands of International Importance and Key Biodiversity Areas). These floodplains are also fire prone due to the high volumes of grassy fuel loads, a product of the extremely high rainfall they receive (approximately 1600 mm annually; Bureau of Meteorology rainfall data, http://www.bom.gov.au/climate/data/). The study area encompassed 4150 km2 of the floodplains of the Adelaide River, Mary River, Wildman River, West Alligator River, South Alligator River, and the western side of the East Alligator River – all of which lie between Darwin (12.46°S, 130.85°E) and Gunbalanya (12.21°S, 133.02°E), Northern Territory, Australia (Fig. 1). The boundaries of these floodplains were mapped manually from satellite imagery because published floodplain boundaries often represent historical maximum flood extent, therefore incorporating adjacent woodland (Ward et al. 2014).
Map of the study area showing the floodplains (light green area) of the Adelaide River (A), Mary River (M), Wildman River (W), West Alligator River (WA), South Alligator River (S), and East Alligator River (E), by land tenure (pastoral or conservation). The dark green represents the lowland savanna of Kakadu National Park (NP).
Land tenure of these floodplains comprises areas allocated to pastoral (47%) and conservation (53%) use, with the floodplain managed for conservation incorporating parts of Kakadu NP (managed by Parks Australia) and several smaller national parks and reserves (managed by the Parks and Wildlife Commission of the Northern Territory [PWCNT]). Other uses of the study area include recreational fishing, hunting, and wildlife tours, which have associated accommodation and infrastructure. Much of the conservation reserve network is also owned or co-managed by Indigenous peoples.
To enable comparison with the fire history of the floodplain study area, we compiled a fire history of the lowland savanna of Kakadu NP (land tenure entirely conservation) (Fig. 1). We chose this area of savanna because fire scar mapping was available for all landscapes within Kakadu NP, and because it is immediately adjacent to a large portion of the surveyed floodplains.
Fire mapping
Fire maps of both the floodplains and savannas of Kakadu NP were provided by the Darwin Centre for Bushfire Research (DCBR). To characterise the fire history over the rest of the study area, we obtained all available Landsat (30 m by 30 m resolution) satellite imagery from 1988 until 2018, sourced from the United States Geological Survey (USGS) database. Because suitable Landsat imagery was not available for the entire floodplain for 10 of the 31 years, fire maps outside of Kakadu NP were partially supplemented with data from the Northern Australia Fire Information (NAFI) service (MODIS satellites, 250 m by 250 m resolution; https://firenorth.org.au/nafi3/).
We viewed annual sets of true colour images in the Geographical Information System (GIS) program ArcMap (ESRI 2015) to select those that were largely cloud free. We applied a two-step process to map fire. The first step involved manually thresholding difference indices calculated from consecutive image dates, then editing commission errors on-screen from the vectorised outputs. Indices utilised were the normalised burn ratio and mid-infrared burn index. These indices incorporate the near infrared and short-wave infrared spectral regions and are sensitive to variations in soil moisture and leaf water content (Xu 2006; Wang and Qu 2007, 2009) so are effective in highlighting burned areas (López García and Caselles 1991; Pereira et al. 1999; Pleniou and Koutsias 2013). These indices are often confounded by factors that include changes in the presence of surface water (e.g. inundation or drying of the floodplains, tidal movement), vegetation changes as the floodplains dry out, and clouds and their shadows. We therefore applied a second step to cross-check mapped burn scars against corresponding true-colour images and then manually discarded areas that were not burn scars. Differenced image pairs were a maximum of 48 days apart, because longer time periods made burn scars too difficult to discern visually. Burn scars were then converted to polygons, attributed with month of burn (median of image date pairs) and combined with all the other polygons for that year, resulting in 31 fire maps representing where and when fires had burnt in the study area across each of 31 years.
Floodplain fire ecological thresholds
Two recommended thresholds specific to conservation reserves or for conservation outcomes in northern Australian floodplains (Table 1) have been proposed in an assessment of the fire management of Kakadu NP (Russell‐Smith et al. 2017), and one in a study of the habitat use of the Alligator Rivers Yellow Chat, a bird endemic to the floodplains of this study area (Leppitt et al. 2022). These three thresholds take the form of a percentage of floodplain unburnt over a given time horizon (3, 5, and 10 years). As far as we are aware, these are the only floodplain-specific thresholds for northern Australia. The fire history we generated for the floodplains was used to determine how often these thresholds were met during the study period.
| Threshold | Assessed | Reference |
|---|---|---|
| At least 25% of floodplain should remain unburnt for 3 years | Achieved consistently >40% for 2006–2015 (Kakadu National Park only) | Russell‐Smith et al. (2017) |
| At least 5% of floodplain should remain unburnt for 10 years | Achieved consistently >15% for 2006–2015 (Kakadu National Park only) | Russell‐Smith et al. (2017) |
| Floodplain unburnt for 5+ years, no spatial threshold provided | Not assessed | Leppitt et al. (2022) |
Assessments shown are assessments completed in the reference provided.
Data analysis
The fire mapping performed for this study, as well as the fire-mapping provided by the DCBR, was ground-truthed by visiting a total of 183 sites (78 in the area mapped by this study, 105 in the area mapped by the DCBR) within the study area during the late dry seasons of 2017 and 2018. At each of these survey points we recorded whether or not the site had burned and then compared the result with the fire maps for that year.
We generated four fire metrics from fire mapping: (1) total area of floodplain burnt each year; (2) patch size of each fire scar; (3) fire frequency; and (4) seasonality of each fire scar using ArcMap (ESRI 2015). The differences in average patch size between pastoral and conservation floodplain, floodplain and savanna, and early and late dry season fires were assessed by applying the Kruskal–Wallis test, a nonparametric analysis of variance (Kruskal and Wallis 1952). Interpretations of the estimates of patch size were carefully examined to ensure adjacent fires were not mistaken for single large contiguous burnt areas, which was possible when lower resolution imagery was used or there were longer intervals between mapped image pairs, because this increased the chances that more than one fire had occurred within the time period of interest. Seasonality of all fire scars was determined by classifying fires as early dry season (EDS) if they occurred before 1 August, or late dry season (LDS) if they occurred after 31 July (following Russell-Smith and Edwards 2006). Fire frequency was calculated as the number of years an individual 30 m2 pixel from the fire scar raster file was mapped as burnt by the fire scar mapping over the entire survey period. We also analysed the frequency of fire when areas of floodplain that did not burn for the entire survey period were excluded, because these areas may be entirely non-flammable (e.g. permanently inundated areas, saltpans), to determine how much of the floodplain that can burn remains unburnt for the 3-, 5-, and 10-year time horizons used by Russell‐Smith et al. (2017) and Leppitt et al. (2022).
We applied least-squares regression using the package ‘Stats’ in the statistical environment R (R Core Team 2017), with calendar year as the independent variable to examine the trends of the percentage of area burned each year over time for the whole floodplain, pastoral floodplain, conservation floodplain, and Kakadu NP savanna (R Core Team 2017):
where ŷ is the regression estimate of the percentage burnt, x is the calendar year, β0 is the regression intercept, and β1 is the regression slope. The statistical significance of the regression trends over time was assessed using the standard error of the slope and its associated P-value.
We plotted the percentage of area burned on the whole floodplain against the total rainfall for the preceding wet season. Because there were no weather stations on the floodplains, monthly rainfall data were obtained from the Bureau of Meteorology for Darwin Airport, Jabiru Airport, and the South Alligator Ranger Station weather stations (situated approximately on the western, eastern, and southern edges of the study area, respectively). Annual rainfall totals were calculated by adding monthly rainfall totals from July of one year to June of the next, to capture the entire wet season preceding the fire season. Where rainfall data from a particular weather station were not available for a given month, that weather station was omitted from the rainfall calculations.
Results
Fire-mapping accuracy
The fire scar mapping performed in this study was 96.2% accurate when ground-truthed at the 78 sites visited on the ground, and mapping provided by the DCBR was 81.0% accurate for the 105 sites visited, for an overall estimated accuracy of 87.4%.
Area burned
On average, 17.1 ± 1.0% (standard deviation) of the floodplains burnt annually, with 7.3 ± 0.6% in the LDS (Fig. 2). A greater proportion of conservation floodplain (26.3 ± 1.6%) was burnt annually than was pastoral floodplain (7.0 ± 0.7%). Seasonality of fire varied between tenures, with a greater proportion of fires on conservation floodplain occurring in the EDS (61.2%) than in the LDS (38.8%). In contrast, on pastoral floodplain, fires more commonly occurred in the LDS (58.6%). Within Kakadu NP, fire was more extensive in the adjacent savanna, with more than twice the area burned on average each year (56.9 ± 1.8%) compared with the floodplain (26.5 ± 1.4%). There was a significant positive trend in the percentage of Kakadu NP savanna burnt annually over the 31-year study period (P = 0.002), but no significant trend across time in fire extent on pastoral or conservation floodplain, or floodplain collectively (Fig. 3).
Mean (±standard deviation) percentage of the study area burned annually during 1988–2018 by season, as well as by land tenure (pastoral or conservation), and the surrounding savanna of Kakadu National Park.
The percentage of the study area burned each year for the entire surveyed floodplain, each land tenure, and surrounding savanna (Kakadu National Park). The black bars represent the extent of late dry season fires, and the black and grey bars combined represent the total area burned. The solid lines are the trend of total fire, and the dashed line is the trend line of late dry season fire. The R2 value is calculated for the total trend of fire (solid black bar).
Fire frequency
In total, 2970 km2 (71.3%) of the surveyed floodplains burnt at least once from 1988 to 2018 (Fig. 4), meaning that 1180 km2 (28.7%) remained unburned throughout the entire survey period. After removing unburnt areas of floodplain, fire frequency varied markedly between land tenures: 89.6% of conservation floodplain burnt at least once compared with 62.4% of pastoral floodplain, and 88.4% of pastoral floodplain burned five times or less, compared with 47.3% of conservation floodplain. For areas that burnt at least once, average fire frequency over the entire survey period varied. For the entire floodplain it was 5.0 times over the 31-year fire history, for pastoral floodplain 1.9 times, and for conservation floodplain 7.7 times, and the average interval between fires was 6.0 years for the entire floodplain, 15.5 years for pastoral floodplain, and 3.9 years for conservation floodplain. Within pastoral floodplain, >99% of floodplain that burned more than three times was on the Adelaide River, whereas within conservation floodplain, the frequency of fire was more evenly spread, but was highest on the Adelaide River and South Alligator River floodplains (Fig. 5).
The percentage of the study area burned at a range of different fire frequencies for the entire surveyed floodplain, each land tenure, and surrounding savanna (Kakadu National Park), over the 31-year period. Black bars represent all fires, and white bars represent late dry season fires only. All fire frequencies except for no fires (zero) have been binned into 5-year intervals.
Map of the surveyed floodplains showing the frequency of all fires and fires in the late dry season only during 1988–2018, across each land tenure (pastoral and conservation). The colour gradient reflects the number of times that each 30 m2 pixel has been burned for the specified fire type. Note that colour scale is adjusted for late dry season fire to better illustrate areas that burned frequently at that time.
The average patch size of all fires that burnt on the floodplain during the study period was 45.9 ha (Table 2). Floodplain fire patches were significantly larger on pastoral floodplain compared with conservation floodplain but were significantly smaller than fires in adjacent savanna (Tables 2 and 3).
| All fires | Early dry season | Late dry season | |
|---|---|---|---|
| All floodplain | 45.9 | 44.1 | 48.6 |
| Pastoral floodplain | 56.6 | 55.9 | 57.1 |
| Conservation floodplain | 43.8 | 42.4 | 46.2 |
| Surrounding savanna | 127.8 | 125.7 | 133.5 |
| Difference in average patch size | P |
|---|---|
| All floodplain < Surrounding savanna | <0.001 |
| Conservation floodplain < Pastoral floodplain | <0.001 |
| All floodplain EDS < All floodplain LDS | <0.001 |
| Pastoral floodplain EDS < Pastoral floodplain LDS | <0.001 |
| Conservation floodplain EDS < Conservation floodplain LDS | <0.001 |
EDS, early dry season; LDS, late dry season; land tenures, conservation and pastoral.
Floodplain fire thresholds
The percentage of unburnt floodplain over the three time horizons was relatively consistent over the course of the survey period, with the most variability in the 3-year time horizon (Fig. 6). A minimum of 59.8% and a maximum of 77.1% of floodplain was unburnt over the previous 3 years, a minimum of 53.3% and a maximum of 63.8% over the previous 5 years, and a minimum of 43.2% and a maximum of 50.1% over the previous 10 years. After excluding unburnt areas (possibly non-flammable), an average of 38.5% of floodplain was unburnt over the preceding 3 years, 29.6% over the preceding 5 years, and 17.6% over the preceding 10 years. The thresholds outlined by Russell‐Smith et al. (2017) were therefore met (i.e. the observed extent of longer-unburnt floodplains was greater than the threshold value), whether or not unburnt floodplain was included.
The percentage of floodplain that had remained unburnt over the preceding indicated time horizon (3, 5, and 10 years). The 3- and 10-year time horizons are from the ecological performance thresholds for floodplain burning suggested in Russell‐Smith et al. (2017), whereas the 5-year time horizon is the floodplain fire history associated with the floodplain endemic bird, the Alligator Rivers Yellow Chat (Epthianura crocea tunneyi; Leppitt et al. 2022) (Table 1).
Relationship with rainfall
There was a significant negative relationship between proportion of floodplain burnt and the preceding wet season’s rainfall, meaning less rain often resulted in more floodplain fire. In contrast, on savanna, less rain resulted in a reduced overall area burnt, although this was a weak relationship (Fig. 7).
Discussion
This study represents the first floodplain-specific assessment of fire regimes in northern Australia and the first assessment of the fire regimes of northern Australian floodplains outside of Kakadu NP. The results of this study demonstrate the significant differences in fire history between conservation and pastoral floodplain land tenures, and between the surveyed floodplains and adjacent Kakadu NP savanna. This study found that the fire regimes of the surveyed floodplains have comfortably met the thresholds set out by Russell‐Smith et al. (2017) and Leppitt et al. (2022) every year for at least the last three decades. This study also found that rainfall in the previous wet season explains more of the variation in the extent of burning on the surveyed floodplains than in Kakadu NP savanna, and that the relationship between the previous wet season’s rainfall total and the extent of floodplain burning is negative. These results confirm that northern Australian floodplains require floodplain-specific fire management, and that significantly more research needs to be undertaken on the effects of various fire regimes on floodplain biota to better inform and establish ecological fire thresholds.
Floodplain tenure and fire
We found a major difference between land tenures in floodplain fire regimes, with dramatically less fire on pastoral floodplain than on conservation floodplain. This result is unsurprising – northern Australian pastoral fire management generally seeks to exclude wildfire from the floodplains to maintain grassy pastures throughout the year (Dyer et al. 2001). Fire can have an immediate impact on the availability of grass for cattle, particularly if soil moisture is low, making the pastures unusable until the following wet season. Fire management for conservation of savannas in northern Australia generally encourages fire that attempts to reproduce pre-colonial fire regimes, which is assumed to have consisted of frequent patchy burning in the early dry season that reduced the incidence of large, high-intensity late dry season fires (Fraser et al. 2003; Woinarski et al. 2005; Yates et al. 2008). Fire managers of conservation floodplains appear to have been at least partially successful in achieving this fire regime for the survey period, with the majority of conservation floodplain fire being in the early dry season.
Pastoral floodplains may also burn less frequently or extensively than conservation floodplains because of the high grazing pressure exerted by cattle (Bos indicus Linnaeus, 1758), which reduce grassy fuel loads, and hence fire probability. Feral herbivores are present on floodplains inside Kakadu NP (such as horse (Equus caballus Linnaeus, 1758), donkey (Equus asinus Linnaeus, 1758), pig (Sus scrofa Linnaeus, 1758), Asian water buffalo, and cattle), but management of feral animals over several decades has probably kept densities and therefore grazing pressure lower within the park than on the pastoral floodplains (Australian Government 2016). The majority of burning of floodplains under pastoral use occurred on the Adelaide River floodplain, which is the only floodplain with both pastoral and conservation floodplain directly adjacent (i.e. not separated by a river). This may be because fires moved from the more frequently burnt conservation floodplain to pastoral floodplain. It is also the only floodplain that is in close proximity to some urban areas, which may also make it more prone to fires started accidentally or deliberately by people.
Comparison between floodplain and savanna burning
We also found a major contrast in fire regimes between habitat types, with the surveyed floodplains burning much less frequently and extensively than Kakadu NP savanna. This is probably because pastoralists aim to exclude fire (Dyer et al. 2001), whereas fire is integral to management across both savanna and floodplains within Kakadu NP (Australian Government 2016). Even when pastoral floodplains were excluded from analysis, all observed fire metrics showed considerably more burning in savanna than floodplains. The surveyed floodplains likely burnt less than Kakadu NP savanna due to their higher water holding capacity: northern Australian floodplains are often inundated for extensive periods of the year (see Ward et al. 2014), reducing the amount of time they are dry enough to burn. However, a study of the Okavango Delta in Botswana found the opposite: floodplains burnt more frequently than adjacent drylands due to increased vegetation growth from streamflow on the floodplains (Heinl et al. 2006, 2007). In global comparisons, the relationship between floodplain and surrounding landscape burning may therefore be influenced by human use, climate, and fuel loads.
The average patch size of fires in Kakadu NP savanna is considerably higher than those on the floodplain (Table 2). In northern Australia, floodplains dry gradually, with water receding to channels and depressions and then backwater swamps as the dry season progresses (Ward et al. 2014). Fires that burn during this drying process often extinguish when they encounter wet areas (McGregor et al. 2010), thereby limiting their size. Additionally, the smaller size of the floodplains when compared with the Kakadu NP savanna and their often-elongated narrow shape means that fires spread less comprehensively than across large contiguous patches of savanna.
Burning for conservation – ecological thresholds
Notwithstanding some well documented traditional ecological knowledge (Russell-Smith et al. 1997; McGregor et al. 2010), there is a paucity of published research investigating the responses of floodplain taxa to fire. Thresholds represent monitoring criteria that are often used to evaluate fire regimes, and are most successful when they are based upon the fire requirements of the biodiversity they seek to protect (van Wilgen et al. 2014). The thresholds for floodplain burning set out in Russell‐Smith et al. (2017) were largely based upon studies of savanna burning. However, the species composition and conservation values of floodplains and savannas are markedly dissimilar (Woinarski et al. 1988), and the extent to which savanna burning techniques are applicable to floodplains is unknown. For example, the maintenance of long-unburnt habitat is a metric of fire management in northern Australia savanna because key components of northern Australian biota appear to favour long-unburnt habitat (that is, habitat that had not burnt for 5 years or longer) (Andersen et al. 2005), but it may not directly translate to a floodplain setting. The 28.7% of the floodplains that remained unburnt for the entire 31-year survey period is likely non-flammable, either because it is permanently inundated, or because it is a non-flammable environment (e.g. saltpan, samphire). Therefore, the inclusion of non-flammable areas of floodplain into calculations of unburnt floodplain, as was done in Russell‐Smith et al. (2017), does not give an accurate measure of floodplain that is long unburnt, because it is also measuring floodplain that is seemingly un-burnable. The number of years unburnt threshold may therefore be more useful if it is only applied to floodplain that is flammable. When only considering areas of floodplain that burnt at least once during the survey period, all three published thresholds (Russell‐Smith et al. 2017; Leppitt et al. 2022) were still met, but that does not mean there is empirical evidence that they meet management objectives of maintaining the biological diversity of Kakadu NP (Australian Government 2016), or that of other floodplains in the region. Determining more appropriate thresholds for floodplain burning will require further research to establish how different fire regimes affect floodplain habitat structure and diversity.
Fire management for the conservation of biodiversity should be informed by the responses to fire of key floodplain plant and animal species, and the potential for use of fire to manage floodplain pest and weed species, but the influence that floodplain fire regimes have on those species is mostly unknown. In addition to the floodplain-endemic Yellow Chat, the floodplains are used by the nationally listed Plains Death Adder (Acanthophis hawkei) and globally significant numbers of waterbirds (Finlayson et al. 2006). More research is needed on the responses to fire of floodplain species of conservation or management concern. However, in the current absence of robust evidence of biodiversity responses to fire, floodplain burning thresholds for the purpose of conservation should ideally encourage a return to floodplain burning regimes preceding European settlement, because floodplain biota likely adapted to these fire regimes over the millennia they were practised (Yates et al. 2008). These fire regimes were likely practised on the floodplains since their formation (Lucas et al. 1997). Indigenous floodplain burning was gradual and patchy, and used the matrix of wet and dry areas of floodplain present in the early dry season to keep fire patch sizes small and iteratively build up protective firebreaks (including burning the same area several times) that would restrict the spread of intense LDS fire used to promote habitat heterogeneity (Haynes 1985; McGregor et al. 2010). Therefore, more apt thresholds may incorporate more nuanced patch size and frequency requirements, rather than seasonality or a broad proportional area of floodplain unburnt over a given time horizon as is used in Russell‐Smith et al. (2017).
Although this study found that the majority of the floodplain meets the preferred fire regime of the Alligator Rivers Yellow Chat, the proportion of the floodplain that needs to meet this criterion is unclear. Alligator Rivers Yellow Chats appear to inhabit quite a narrow floodplain niche (Leppitt et al. 2022), so fire management for this threatened taxon may need to target specific chat ‘hot-spots’, or areas of floodplain that meet other chat habitat criteria (as per Leppitt et al. 2022).
Wet season rainfall and fire
This study found that high-rainfall wet seasons generally result in less floodplain being burnt during the subsequent dry season. This is probably due to greater levels of rainfall and floodplain inundation reducing the amount of time floodplains are dry enough to burn. In contrast, this study found that the relationship between wet season rainfall total and the extent of burning in Kakadu NP savanna in the preceding dry season was much less pronounced. Previous studies have found that savannas generally have a positive relationship between rainfall and subsequent extent of burning; generally, rainfall promotes the growth of readily burnable grassy biomass, which fuels more extensive fires when the landscape dries (Archibald et al. 2010). However, this relationship becomes less pronounced at very high levels of annual rainfall (Archibald et al. 2010) because extended wet seasons can reduce the amount of time that landscapes are dry enough to burn, as well as the growing season of the grasses that emerge after the water recedes.
Conclusions
This study has quantified the fire regimes prevalent under two of the dominant land tenures on the northern Australian floodplains – pastoralism and conservation. It has demonstrated that in the Van Diemen Gulf coastal floodplains of the Northern Territory, pastoralists have been largely successful at excluding fire from their floodplain pastures. In conservation reserves, where management objectives are more complex, burning has been much more frequent, with most fires occurring in the early dry season. Further research is required to fully understand the influence of fire regimes on floodplain vegetation, conservation values, and potential uses for the control of pest and weed species, because this can inform thresholds which land managers can strive to achieve. These thresholds should enable the continued use of the floodplains as an important food source for Indigenous peoples, as well as allowing floodplain biota to thrive, with an emphasis on floodplain-dependent species. Better-informed floodplain fire management will increase the resilience of floodplain biodiversity to the threats posed by weeds, feral animals, and sea-level rise.
Declaration of funding
This research received support from Charles Darwin University and the Australian Government’s National Environmental Science Program through the Threatened Species Recovery Hub. PMK was supported through the Marine Biodiversity Hub, a collaborative partnership supported through funding from the Australian Government’s National Environmental Science Program.
Acknowledgements
The authors acknowledge the Traditional Owners of the lands on which the research was conducted, most notably Victor Cooper, Sean Nadji, and the Larrakia people, on which Charles Darwin University’s Casuarina campus stands. We also acknowledge Roy Tipiloura and Harold Swan, who contributed to the collection of data.
References
Andersen AN, Braithwaite RW, Cook GD, Corbett LK, Williams RJ, Douglas MM, Gill AM, Setterfield SA, Muller WJ (1998) Fire research for conservation management in tropical savannas: introducing the Kapalga fire experiment. Australian Journal of Ecology 23, 95-110.
| Crossref | Google Scholar |
Andersen AN, Cook GD, Corbett LK, Douglas MM, Eager RW, Russell-Smith J, Setterfield SA, Williams RJ, Woinarski JCZ (2005) Fire frequency and biodiversity conservation in Australian tropical savannas: implications from the Kapalga fire experiment. Austral Ecology 30, 155-167.
| Crossref | Google Scholar |
Andersen AN, Woinarski JCZ, Parr CL (2012) Savanna burning for biodiversity: fire management for faunal conservation in Australian tropical savannas. Austral Ecology 37, 658-667.
| Crossref | Google Scholar |
Archibald S, Nickless A, Govender N, Scholes RJ, Lehsten V (2010) Climate and the inter‐annual variability of fire in southern Africa: a meta‐analysis using long‐term field data and satellite‐derived burnt area data. Global Ecology and Biogeography 19, 794-809.
| Crossref | Google Scholar |
Bayliss P, Yeomans K (1990) Seasonal distribution and abundance of Magpie Geese, Anseranas semipalmata Latham, in the Northern Territory, and their relationship to habitat, 1983-86. Wildlife Research 17, 15-38.
| Crossref | Google Scholar |
Bowman DMJS, Fensham RJ (1991) Response of a monsoon forest‐savanna boundary to fire protection, Weipa, northern Australia. Australian Journal of Ecology 16, 111-118.
| Crossref | Google Scholar |
Bowman DMJS, Panton WJ (1993) Decline of Callitris intratropica RT Baker & HG Smith in the Northern Territory: implications for pre-and post-European colonization fire regimes. Journal of Biogeography 20, 373-381.
| Crossref | Google Scholar |
Bowman DMJS, Williamson GJ, Johnston FH, Bowman CJW, Murphy BP, Roos CI, Trauernicht C, Rostron J, Prior LD (2022) Population collapse of a Gondwanan conifer follows the loss of Indigenous fire regimes in a northern Australian savanna. Scientific Reports 12, 9081.
| Crossref | Google Scholar |
Boyden J, Bayliss P, Kennett R, Christophersen P, Lawson V, McGregor S, Begg G (2003) ‘Vegetation change analysis on Boggy Plain, South Alligator River, using remote sensing: Report on preliminary analyses.’ (Australian Government, Department of the Environment and Heritage, Supervising Scientist: Darwin, NT)
Campbell HA, Loewensteiner DA, Murphy BP, Pittard S, McMahon CR (2021) Seasonal movements and site utilisation by Asian water buffalo (Bubalus bubalis) in tropical savannas and floodplains of northern Australia. Wildlife Research 48, 230-239.
| Crossref | Google Scholar |
Crowley G, Garnett S, Shephard S (2009) Impact of storm‐burning on Melaleuca viridiflora invasion of grasslands and grassy woodlands on Cape York Peninsula, Australia. Austral Ecology 34, 196-209.
| Crossref | Google Scholar |
Douglas MM, O’Connor RA (2004) Weed invasion changes fuel characteristics: para grass (Urochloa mutica (Forssk.) TQ Nguyen) on a tropical floodplain. Ecological Management & Restoration 5, 143-145.
| Crossref | Google Scholar |
Edwards A, Kennett R, Price O, Russell-Smith J, Spiers G, Woinarski J (2003) Monitoring the impacts of fire regimes on vegetation in northern Australia: an example from Kakadu National Park. International Journal of Wildland Fire 12, 427-440.
| Crossref | Google Scholar |
Evans J, Russell-Smith J (2020) Delivering effective savanna fire management for defined biodiversity conservation outcomes: an Arnhem Land case study. International Journal of Wildland Fire 29, 386-400.
| Crossref | Google Scholar |
Finlayson CM (2005) Plant ecology of Australia’s tropical floodplain wetlands: a review. Annals of Botany 96, 541-555.
| Crossref | Google Scholar |
Finlayson CM, Lowry J, Bellio MG, Nou S, Pidgeon R, Walden D, Humphrey C, Fox G (2006) Biodiversity of the wetlands of the Kakadu Region, northern Australia. Aquatic Sciences 68, 374-399.
| Crossref | Google Scholar |
Franklin DC, Whitehead PJ, Pardon G, Matthews J, McMahon P, McIntyre D (2005) Geographic patterns and correlates of the decline of granivorous birds in northern Australia. Wildlife Research 32, 399-408.
| Crossref | Google Scholar |
Fraser BF, Lawson V, Morrison S, Christophersen P, McGreggor S, Rawlinson M (2003) Fire management experiment for the declining partridge pigeon, Kakadu National Park. Ecological Management & Restoration 4, 94-102.
| Crossref | Google Scholar |
Gill A, Ryan P, Moore P, Gibson M (2000) Fire regimes of World Heritage Kakadu National Park, Australia. Austral Ecology 25, 616-625.
| Crossref | Google Scholar |
Heim RJ, Hölzel N, Heinken T, Kamp J, Thomas A, Darman GF, Smirenski SM, Heim W (2019) Post-burn and long-term fire effects on plants and birds in floodplain wetlands of the Russian Far East. Biodiversity and Conservation 28, 1611-1628.
| Crossref | Google Scholar |
Heinl M, Neuenschwander A, Sliva J, Vanderpost C (2006) Interactions between fire and flooding in a southern African floodplain system (Okavango Delta, Botswana). Landscape Ecology 21, 699-709.
| Crossref | Google Scholar |
Heinl M, Frost P, Vanderpost C, Sliva J (2007) Fire activity on drylands and floodplains in the southern Okavango Delta, Botswana. Journal of Arid Environments 68, 77-87.
| Crossref | Google Scholar |
Kleindl WJ, Rains MC, Marshall LA, Hauer FR (2015) Fire and flood expand the floodplain shifting habitat mosaic concept. Freshwater Science 34, 1366-1382.
| Crossref | Google Scholar |
Kruskal WH, Wallis WA (1952) Use of ranks in one-criterion variance analysis. Journal of the American Statistical Association 47, 583-621.
| Crossref | Google Scholar |
Leppitt R, Einoder L, Kyne PM, Woinarski JCZ, Garnett S (2022) Understanding habitat use of the Endangered Alligator Rivers Yellow Chat Epthianura crocea tunneyi to inform monitoring and management. Bird Conservation International 32, 423-438.
| Crossref | Google Scholar |
Lonsdale WM, Miller IL (1993) Fire as a management tool for a tropical woody weed: Mimosa pigra in Northern Australia. Journal of Environmental Management 39, 77-87.
| Crossref | Google Scholar |
López García MJ, Caselles V (1991) Mapping burns and natural reforestation using Thematic Mapper data. Geocarto International 6, 31-37.
| Crossref | Google Scholar |
Madsen T, Ujvari B, Shine R, Buttemer W, Olsson M (2006a) Size matters: extraordinary rodent abundance on an Australian tropical flood plain. Austral Ecology 31, 361-365.
| Crossref | Google Scholar |
Madsen T, Ujvari B, Shine R, Olsson M (2006b) Rain, rats and pythons: Climate‐driven population dynamics of predators and prey in tropical Australia. Austral Ecology 31, 30-37.
| Crossref | Google Scholar |
McGregor S, Lawson V, Christophersen P, Kennett R, Boyden J, Bayliss P, Liedloff A, McKaige B, Andersen AN (2010) Indigenous wetland burning: conserving natural and cultural resources in Australia’s World Heritage-listed Kakadu National Park. Human Ecology 38, 721-729.
| Crossref | Google Scholar |
Petty AM, Werner PA, Lehmann CER, Riley JE, Banfai DS, Elliott LP (2007) Savanna responses to feral buffalo in Kakadu National Park, Australia. Ecological Monographs 77, 441-463.
| Crossref | Google Scholar |
Pleniou M, Koutsias N (2013) Sensitivity of spectral reflectance values to different burn and vegetation ratios: A multi-scale approach applied in a fire affected area. ISPRS Journal of Photogrammetry and Remote Sensing 79, 199-210.
| Crossref | Google Scholar |
Roberts SJ (1993) Fire on the Gapalg (floodplain): contemporary aboriginal and other burning patterns in Kakadu National Park, northern Australia. Fire in wetlands: a management perspective. In ‘Proceeding of the Tall Timbers Fire Ecology Conference’. pp. 31–38. (Tall Timbers Research Station: Florida)
Robinson CJ, Whitehead P (2003) Cross-cultural management of pest animal damage: a case study of feral buffalo control in Australia’s Kakadu National Park. Environmental Management 32, 445-458.
| Crossref | Google Scholar |
Russell-Smith J, Edwards AC (2006) Seasonality and fire severity in savanna landscapes of monsoonal northern Australia. International Journal of Wildland Fire 15, 541-550.
| Crossref | Google Scholar |
Russell-Smith J, Ryan PG, Durieu R (1997) A LANDSAT MSS-Derived fire history of Kakadu National Park, monsoonal northern Australia, 1980–94: seasonal extent, frequency and patchiness. Journal of Applied Ecology 34, 748-766.
| Crossref | Google Scholar |
Russell-Smith J, Ryan PG, Klessa D, Waight G, Harwood R (1998) Fire regimes, fire‐sensitive vegetation and fire management of the sandstone Arnhem Plateau, monsoonal northern Australia. Journal of Applied Ecology 35, 829-846.
| Crossref | Google Scholar |
Russell-Smith J, Whitehead PJ, Cook GD, Hoare JL (2003) Response of Eucalyptus‐dominated savanna to frequent fires: lessons from Munmarlary, 1973–1996. Ecological Monographs 73, 349-375.
| Crossref | Google Scholar |
Russell‐Smith J, Evans J, Edwards AC, Simms A (2017) Assessing ecological performance thresholds in fire‐prone Kakadu National Park, northern Australia. Ecosphere 8, e01856.
| Crossref | Google Scholar |
Sharp BR, Whittaker RJ (2003) The irreversible cattle‐driven transformation of a seasonally flooded Australian savanna. Journal of Biogeography 30, 783-802.
| Crossref | Google Scholar |
van Wilgen BW, Govender N, Smit IPJ, MacFadyen S (2014) The ongoing development of a pragmatic and adaptive fire management policy in a large African savanna protected area. Journal of Environmental Management 132, 358-368.
| Crossref | Google Scholar |
Wang L, Qu JJ (2007) NMDI: A normalized multi‐band drought index for monitoring soil and vegetation moisture with satellite remote sensing. Geophysical Research Letters 34, L20405.
| Crossref | Google Scholar |
Wang L, Qu JJ (2009) Satellite remote sensing applications for surface soil moisture monitoring: A review. Frontiers of Earth Science in China 3, 237-247.
| Crossref | Google Scholar |
Ward DP, Petty A, Setterfield SA, Douglas MM, Ferdinands K, Hamilton SK, Phinn S (2014) Floodplain inundation and vegetation dynamics in the Alligator Rivers region (Kakadu) of northern Australia assessed using optical and radar remote sensing. Remote Sensing of Environment 147, 43-55.
| Crossref | Google Scholar |
Woinarski JCZ, Legge S (2013) The impacts of fire on birds in Australia’s tropical savannas. Emu-Austral Ornithology 113, 319-352.
| Crossref | Google Scholar |
Woinarski JCZ, Tidemann SC, Kerin S (1988) Birds in a tropical mosaic - the distribution of bird species in relation to vegetation patterns. Wildlife Research 15, 171-196.
| Crossref | Google Scholar |
Woinarski JCZ, Williams RJ, Price O, Rankmore B (2005) Landscapes without boundaries: wildlife and their environments in northern Australia. Wildlife Research 32, 377-388.
| Crossref | Google Scholar |
Woinarski JCZ, Legge S, Fitzsimons JA, Traill BJ, Burbidge AA, Fisher A, Firth RSC, Gordon IJ, Griffiths AD, Johnson CN, McKenzie NL, Palmer C, Radford I, Rankmore B, Ritchie EG, Ward S, Ziembicki M (2011) The disappearing mammal fauna of northern Australia: context, cause, and response. Conservation Letters 4, 192-201.
| Crossref | Google Scholar |
Woodroffe CD, Thom B, Chappell J, Wallensky E, Grindrod J (1987) Relative sea level in the South Alligator River region, north Australia, during the Holocene. Search (Sydney) 18, 198-200.
| Google Scholar |
Xu H (2006) Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery. International Journal of Remote Sensing 27, 3025-3033.
| Crossref | Google Scholar |
Yates CP, Edwards AC, Russell-Smith J (2008) Big fires and their ecological impacts in Australian savannas: size and frequency matters. International Journal of Wildland Fire 17, 768-781.
| Crossref | Google Scholar |
