Rainfall-Linked Megafires as Innate Fire Regime Elements in Arid Australian Spinifex (Triodia spp.) Grasslands

fevo-09-666241 July 5, 2021 Time: 15:55 # 1 ORIGINAL RESEARCH published: 05 July 2021 doi: 10.3389/fevo.2021.666241 Edited by: Eddie John Van Etten, Edith Cowan University, Australia Reviewed by: Ian Radford, Conservation and Attractions (DBCA), Australia Christopher Richard Dickman, The University of Sydney, Australia *Correspondence: Boyd R. Wright [email protected] Specialty section: This article was submitted to Biogeography and Macroecology, a section of the journal Frontiers in Ecology and Evolution Received: 09 February 2021 Accepted: 22 April 2021 Published: 05 July 2021 Citation: Wright BR, Laffineur B, Royé D, Armstrong G and Fensham RJ (2021) Rainfall-Linked Megafires as Innate Fire Regime Elements in Arid Australian Spinifex (Triodia spp.) Grasslands. Front. Ecol. Evol. 9:666241. doi: 10.3389/fevo.2021.666241 Rainfall-Linked Megafires as Innate Fire Regime Elements in Arid Australian Spinifex (Triodia spp.) Grasslands Boyd R. Wright1,2,3*, Boris Laffineur4, Dominic Royé5, Graeme Armstrong6and Roderick J. Fensham4,7 1Department of Botany, School of Environmental and Rural Science, The University of New England, Armidale, NSW, Australia, 2School of Agriculture and Food Science, University of Queensland, St. Lucia, QLD, Australia, 3The Northern Territory Herbarium, Department of Land Resource Management, Alice Springs, NT, Australia, 4Queensland Herbarium, Toowong, QLD, Australia, 5Department of Geography, University of Santiago de Compostela, A Coruña, Spain, 6NSW National Parks and Wildlife Service, Broken Hill, NSW, Australia, 7School of Biological Sciences, University of Queensland, St. Lucia, QLD, Australia Large, high-severity wildfires, or “megafires,” occur periodically in arid Australian spinifex (Triodia spp.) grasslands after high rainfall periods that trigger fuel accumulation. Proponents of the patch-burn mosaic (PBM) hypothesis suggest that these fires are unprecedented in the modern era and were formerly constrained by Aboriginal patch burning that kept landscape fuel levels low. This assumption deserves scrutiny, as evidence from fire-prone systems globally indicates that weather factors are the primary determinant behind megafire incidence, and that fuel management does not mitigate such fires during periods of climatic extreme. We reviewed explorer’s diaries, anthropologist’s reports, and remotely sensed data from the Australian Western Desert for evidence of large rainfall-linked fires during the pre-contact period when traditional Aboriginal patch burning was still being practiced. We used only observations that contained empiric estimates of fire sizes. Concurrently, we employed remote rainfall data and the Oceanic Niño Index to relate fire size to likely seasonal conditions at the time the observations were made. Numerous records were found of small fires during periods of average and below-average rainfall conditions, but no evidence of largescale fires during these times. By contrast, there was strong evidence of large-scale wildfires during a high-rainfall period in the early 1870s, some of which are estimated to have burnt areas up to 700,000 ha. Our literature review also identified several Western Desert Aboriginal mythologies that refer to large-scale conflagrations. As oral traditions sometimes corroborate historic events, these myths may add further evidence that large fires are an inherent feature of spinifex grassland fire regimes. Overall, the results suggest that, contrary to predictions of the PBM hypothesis, traditional Aboriginal burning did not modulate spinifex fire size during periods of extreme-high arid zone rainfall. The Frontiers in Ecology and Evolution | www.frontiersin.org 1July 2021 | Volume 9 | Article 666241 fevo-09-666241 July 5, 2021 Time: 15:55 # 2 Wright et al. Rainfall-Linked Megafires in Spinifex Grasslands mechanism behind this is that plant assemblages in seral spinifex vegetation comprise highly flammable non-spinifex tussock grasses that rapidly accumulate high fuel loads under favorable precipitation conditions. Our finding that fuel management does not prevent megafires under extreme conditions in arid Australia has parallels with the primacy of climatic factors as drivers of megafires in the forests of temperate Australia. Keywords: arid vegetation, fire ecology, grass-fire feedbacks, patch-burning, indigenous land management INTRODUCTION Large-scale high-intensity conflagrations occur during periods of meteorological extreme in many fire-prone ecosystems (Keeley and Fotheringham, 2001;Gill and Allan, 2008;Bradstock, 2010). Proponents of the patch-burn mosaic (PBM) hypothesis propose that such fires are anomalous in the modern era and are primarily the result of (1) the cessation of indigenous “patch-burning” practices that previously kept landscape-level fuel loads low; and (2) European fire-suppression practices that protect grazing lands and infrastructure but also promote widespread fuel accumulation (Bonnicksen, 1981;Minnich, 1983;Goforth and Minnich, 2007). A key assumption of the PBM hypothesis is that fuel age has a positive association with ecosystem flammability (Moritz et al., 2004;Goforth and Minnich, 2007). Accordingly, under the PBM hypothesis, when landscapes are broken up into mosaics of early and late successional vegetation, large highintensity wildfires should be rare due to the inability of early seral vegetation to carry fire (Minnich, 1983, 1998). Despite being an attractive hypothesis that has underpinned fire management activities in many fire-prone regions, the universal applicability of the PBM hypothesis is now being questioned (Gill, 2000;Moritz et al., 2004;Keeley and Zedler, 2009;Kimber and Friedel, 2015). Fire researchers in North America have demonstrated that for many American ecosystems, the assumed relationship between post-fire successional state and flammability does not hold (Keeley and Fotheringham, 2001;Moritz et al., 2009). For example, in ponderosa pine (Pinus ponderosa Douglas ex. C. Lawson) forests and chaparral shrublands, early seral states are characterized by fine surface layer fuels composed of highly flammable grasses and shrubs. Modeling has shown that these fuels carry fire well, even under relatively low wind situations, and are unlikely to act to as fire-retardant breaks under any fire weather (Moritz et al., 2004;Keeley and Zedler, 2009). In these systems, fire occurrence is driven more by extremes of meteorological variables such as temperature, humidity, and wind speed, rather than timesince-fire and associated characteristics of the fuel array (Keeley and Fotheringham, 2001). A widespread Australian ecosystem for which the applicability of the PBM hypothesis has also been questioned are the spinifex (Triodia R. Br.) grasslands of arid and semiarid Australia (Gill, 2000;Kimber and Friedel, 2015;Figures 1A–D). Contemporary fire cycles in these grasslands are characterized by periodic wildfires that are large in scale (e.g., occasionally exceeding 10,000 km2) (Allan and Southgate, 2002), high in intensity (e.g., up to c. 14,000 kW; Burrows et al., 1991), and driven by fuel accumulations that occur following exceptionally high rainfall years (Griffin et al., 1983;Allan and Southgate, 2002). A long-held assumption about fire in these grasslands is that large fires during the pre-European period were rare and were constrained by traditional Aboriginal burning that maintained fuel load heterogeneity in landscapes (Jones, 1969;Gamage, 2012). Evidence that supports this hypothesis comes from studies of aerial photos that clearly show small fire patch sizes under traditional burning regimes (Burrows and Christensen, 1990;Burrows and Chapman, 2018) and from satelliteimage-derived fire scar analyses indicating that contemporary Aboriginal burning buffers climate-driven increases in fire size (Bliege Bird et al., 2012). There is little doubt that in many regions, prior to European contact, skilled and purposeful burning by Aboriginal peoples maintained landscape mosaics comprised of fire scars of different post-fire ages. These fires were crucial for human survival in spinifex grasslands, as a means of increasing hunting capture rates and promoting the growth of food plants (O’Connell et al., 1983;Bird et al., 2008;Latz, 2018). Nevertheless, several lines of evidence suggest that big spinifex grassland fires did occur during the pre-contact period: 1. Anecdotal reports from formerly “traditional-living” Aboriginal peoples of large fires occurring when traditional patch burning was being practiced. These reports come from the spinifex-dominated Gibson and Great Sandy Deserts and include mention of fires that would “go for five nights” or that could be followed for “up to a few days” and of very large wildfires that “finish ’im up everything, right up . . ..” (Mutitjulu community and Baker, 1996;Yates and Morse, 2003; Burrows et al., 2006a). Such fires are reported to have forced people to abandon large areas for a time, as game and plant foods became scarce as burnt regions recovered (Yates and Morse, 2003). 2. Contemporary observations of large fires in recently burnt spinifex grassland subject to high post-fire rainfall (i.e., two large fires with an inter-fire interval of c. 2–3 years) (Latz, 1990, 2007;Gill, 2000;Allan and Southgate, 2002;Allan et al., 2003;Matthews, 2005; Wright, 2007;Wright and Clarke, 2007;Duguid et al., 2008). These observations reflect that post-fire seral state spinifex, which is usually dominated by highly flammable non-spinifex tussock grasses such as Aristida holathera Domin. (Figure 1C), can accumulate fuels quickly under high rain conditions and burn as readily, albeit at lower Frontiers in Ecology and Evolution | www.frontiersin.org 2July 2021 | Volume 9 | Article 666241 fevo-09-666241 July 5, 2021 Time: 15:55 # 3 Wright et al. Rainfall-Linked Megafires in Spinifex Grasslands FIGURE 1 | (A) Regenerating Triodia epactia grassland in central Gibson Desert during average rainfall conditions in 2015. (B) Long-unburnt Triodia basedowii-dominated grassland in eastern Gibson Desert during average rainfall conditions in 2017. (C) Early post-fire seral spinifex vegetation in Haasts Bluff Aboriginal Land Trust in far-eastern Western Desert during well-above average rainfall conditions in 2002. Aristida holathera is the main ground layer biomass component. (D) Lightning-ignited fire in Triodia pungens-dominated grassland on Deep Well station, central Australia. This fire occurred in November 2012 when widespread spinifex fuel accumulations occurred following exceptionally high rainfall years in 2010–2011. Triodia flowering culms double the height of the ground fuel layer. intensity, as old growth spinifex (see Table 1 for list of common seral spinifex species). 3. Incursions of large spinifex wildfires into central Australian conservation reserves where patch-burn fire management has been practiced. In these reserves, good rainfall seasons in 1983–1984, 2000–2002, and 2010– 2011 led to massive fuel accumulations and subsequent wildfires that either burnt over the top of previously instated patch mosaics or trickled around the patch burns and burnt large areas (Allan and Baker, 1990;Allan et al., 2003;Duguid et al., 2008;Horn and Liston, 2013). Given the broad-scale nature of prescribed management burning in Australian spinifex grassland reserves, further assessment of the efficacy of patch burning to constrain largescale wildfires during periods of climatic extreme is warranted. Here, we examined the records of arid Australian explorers, anthropologist’s reports, and remotely sensed data for evidence of a “rainfall–fire size” relationship in spinifex grasslands during the pre-contact period when traditional Aboriginal burning was being practiced. We restricted our search to the Western Desert, as the vegetation of this region is primarily spinifex grassland and several expeditions passed through it when traditional Aboriginal patch burning was still occurring. To ascertain the likely seasonal conditions associated with each of the expeditions, we examined (1) El Niño Southern Oscillation (ENSO) patterns derived from Oceanic Niño Index (ONI) data and (2) precipitation records from the late 1860s to present from long-term weather stations located on the periphery of the arid zone in southern South Australia (Orroroo), southwestern NSW TABLE 1 | Common seral spinifex species by % composition by weight and time-since-fire for maximum biomass levels according to Latz (1990) and TERN (2019). Species Family Reported % composition of post-fire seral vegetation by weight (range) Time (post-fire) for maximum biomass Chrysocephalum apiculatum Asteraceae 10 Uncertain Crotalaria eremea Fabaceae 25 1 Senna pleurocarpa Fabaceae 20–50 2–4 Tephrosia sphaerospora Fabaceae 50 1 Scaevola parviflora Goodeniaceae 35 1 Aristida holathera Poaceae 20–80 1–4 Aristida inaequiglumis Poaceae 8–12 2 Eragrostis eriopoda Poaceae 3–35 2–7 Eriachne helmsii Poaceae 13 2 Eulalia aurea Poaceae 15 Unclear Panicum decompositum Poaceae 5 1 Paraneurachne muelleri Poaceae 10–35 1–5 Yakirra australiense Poaceae 10 0.5 (Wentworth), and central QLD (Springsure). ENSO conditions are usually associated with continental-scale fluctuations in rainfall variability and are known to be primary drivers of interannual variability of fire weather (Harris and Lucas, 2019). Rainfall data from the various weather stations, although spatially distant from the Western Desert, were also considered useful, Frontiers in Ecology and Evolution | www.frontiersin.org 3July 2021 | Volume 9 | Article 666241 fevo-09-666241 July 5, 2021 Time: 15:55 # 4 Wright et al. Rainfall-Linked Megafires in Spinifex Grasslands as extreme-high-rainfall activity in the Australian arid and semiarid zones is often experienced at the continental scale (Denniston et al., 2015). During our literature search, we encountered several Western Desert mythologies that mention large-scale wildfires. As indigenous oral traditions can sometimes corroborate historic events (Nunn and Reid, 2015), we provide a very brief description of these fire myths in our “Results” section and outline the massive scale of the fires that are mentioned. Additionally, as a final exercise, we conducted a satellite imagery study from a spinifex-dominated region in the southwest of the Northern Territory where large areas (c. 500,000 ha) were reported of have burnt twice in 2 years in the early 1980s following successive years of high rainfall (Allan and Southgate, 2002). Doing this exercise would demonstrate unequivocally whether recently burnt patches of spinifex grassland can re-burn rapidly if high post-fire rainfalls occur. This, by corollary, would provide an examination of the tenet of the PBM hypotheses that a relationship exists between fuel age and flammability in the spinifex system. MATERIALS AND METHODS Study Area—The Western Desert The study area was the Western Desert, a large arid plateau occurring primarily in the Western Australian interior and extending into the far northwest of South Australia and the southwest corner of the Northern Territory. It covers an area of approximately 800,000 km2and encompasses the Gibson, the Great Victoria, the Great Sandy, and the Little Sandy Deserts. Geographically, the Desert is composed primarily of large sections of arid dunefield and sandplain interspersed with occasional salt lakes, clay pans, and mountain ranges. Climatically, its summers are hot, with mean maximum temperatures during the hottest month (January) reaching the mid-high 30s. Winters are cool, with mean maximum temperatures in the coolest month (July) rarely rising above 20◦C. Typical of other arid Australian regions, rainfall is unpredictable, with some years receiving little or (rarely) no precipitation and others experiencing monsoon-linked deluges. Annual average rainfall ranges from 180 mm in the southeast of the Desert (Cook BOM station, SA) to 360 mm in the northwest (Telfer Aero BOM weather station) (Australian Government Bureau of Meteorology, 2021). Study Community—Arid Spinifex Grasslands Triodia is a large genus of perennial hummock-forming grasses that dominate grassland and savannah-type systems over approximately 26.6% of the Australian continent (AUSLIG, 1990). As aridity generally limits fuel quantity and fuel continuity in these systems, large fires usually only occur after sufficient rainfall has occurred to promote lateral expansion of hummocks and growth of interstitial species (Allan and Southgate, 2002; Armstrong, 2011). Spinifex species themselves are coarse, occasionally resinous, and sometimes form rings as plants age and die out in the center. Tall flowering culms are highly flammable when dry and can raise the fuel heights of plants to 1.5 m (Jessop, 1981). Fuel loads in long-unburnt grassland range from 1.6 to >12 tons ha−1, depending on fuel age and whether the fuel array consists purely of Triodia spp. or includes woody species (Winkworth, 1967;Beadle, 1981;Burrows et al., 2006b). The vegetation of recently burnt spinifex grassland is quite different to that of long-unburnt grassland. In regions dominated by obligate-seeding Triodia, fire kills adult plants, and postfire rainfall triggers the proliferation of spinifex seedlings, nonspinifex grasses, forbs, and shrub and tree seedlings (Table 1; Latz, 1990;Wright and Clarke, 2007;TERN, 2019). This successional pattern reflects release from competition after adult spinifex death, with smaller-statured non-spinifex species taking advantage of increased light, nutrient, and water availability in the post-fire environment (Griffin, 1990). Non-spinifex species usually make up the bulk of early post-fire fuels in seederdominated grassland, with maximum biomass of these species generally peaking within 2–3 years post-fire. Spinifex seedlings may also contribute to the early post-fire fuel array (up to 50% of the fuel biomass under certain circumstances; Latz, 1990). However, the proportion of post-fire fuels comprised by spinifex seedlings is variable and depends on several factors, including (1) post-fire rainfall season and volume; (2) the species of spinifex involved; and (3) whether mature Triodia populations had masted recently prior to fire (masting “tops up” seed banks and maximizes seed availability for post-fire spinifex seedling recruitment) (Wright and Fensham, 2018). In spinifex grasslands dominated by resprouting Triodia, early postfire successional states are generally dominated by vegetatively regenerating spinifex hummocks rather than by short-lived nonspinifex grasses and forbs (Latz, 1990). Estimating Fire Sizes We located numerous explorer’s observations that allowed empiric estimation of pre-contact fire sizes in the Western Desert (Table 2 and Figure 2). Where possible, the approximate localities of the fires were identified. We followed the method used by Gill (2000) to produce a “conservative” estimate and a “less-conservative” estimate of fire size based on estimates of distances traveled over burnt ground. Conservative fire size estimates were calculated by taking the length of traverse over burnt country as the length of an elliptical fire with a length:breadth ratio of 2:1. The rationale for the use of ellipses here was that, in linear dunefield environments such as those in the Western Desert, fires frequently burn within dune valleys and are often stopped by high dune crests that act as fire breaks. This, in combination with windy conditions, often results in burn scars that are elliptical in shape. For each fire, a “less conservative” estimate of fire size was made by taking the reported length of traverse over burnt country as the diameter of a circular fire. As part of our review of pre-contact Western Desert fires, we examined fire scars on Landsat satellite imagery taken during the 1970s in the Lake Mackay region of Western Australia. In this region, the last group of never-before-contacted Aboriginal people (referred to in contemporary media as the “Pintupi Nine”) had been conducting traditional burning and living a huntergatherer lifestyle until 1984 (Guilliatt, 2010). In the mid-1970s, Frontiers in Ecology and Evolution | www.frontiersin.org 4July 2021 | Volume 9 | Article 666241 fevo-09-666241 July 5, 2021 Time: 15:55 # 5 Wright et al. Rainfall-Linked Megafires in Spinifex Grasslands TABLE 2 | Summary of observations of pre-contact fire sizes in Western Desert. Region Observation Estimated fire size (assuming circular fires) Seasonal conditions References 1873 Northern Great Victoria Desert 27 miles of burnt ground traveled over 2 days c. 148,000 ha Favorable Warburton, 1875 1873 Gibson Desert “Many miles of burnt, stony, gravelly undulations; at every four or five miles I obtained a view of similar country beyond . . .” Not clear, could be up to c. 45,000 ha fire if obtained a “similar view” of burnt country on at least three occasions (i.e., 3×4–5 miles) Favorable Giles, 1889a 1874 Gibson Desert Spinifex fires that “run on for scores of miles” (i.e., 40, 60 or more miles) c. 731,000 ha (if fires ran on for 60 miles) Favorable Giles, 1889b 1874 Great Victoria Desert “Four miles north were then traveled, over burnt country, . . ..” c. 3,250 ha Favorable Giles, 1889b 1889 Eastern Gibson Desert “. . . 3 or four miles these had been swept by fire, . . .”; “. . . five miles distant; all the country in that direction seems to have been recently burnt.” c. 1,800–5,077 ha Average Tietkens, 1890 1936 Eastern Gibson Desert A horseshoe shaped fire, 2 miles long, one mile wide. c. 500 ha Below average Finlayson (1936) in Gill, 2000 1953 Gibson Desert NA (satellite imagery) <10–1,744 ha Average Burrows and Christensen, 1990 1960/1961 Great Victoria Desert NA (satellite imagery) <1–3593 ha Average Burrows and Chapman, 2018 c. 1960 Warburton, Northern Great Victoria Desert (Referring to hunting fires) “. . . many hundreds of meters wide and extending for some kilometers.” 10–3,000 ha Average De Graaf (1975) in Gill, 2000 c. 1960 Great Sandy Desert Empiric description of hunting fire patch sizes given 800–2,500 ha Average Gould (1980) in Gill, 2000 Method for estimating fire size based on Gill (2000). one of the largest rainfall systems in history passed over northern Australia (Denniston et al., 2015;Figure 2 and online appendices 1, 2). We were interested to examine whether there was evidence of a shift in fire scar patterning in the region where the Pintupi Nine people had been burning before and after this high-rainfall period occurred. For this study, we examined imagery from 1972, a time of average to slightly above-average rainfall, and from 1979, which was when the next imagery was available after the mid-1970s rains. We mapped the fire scars for this study using Landsat 1 and 3 satellite images (WRS paths 113 and 114 and WRS rows 75 and 76; i.e., four scenes in total for each of the two time periods), and the satellite data were accessed from the USGS Earth Explorer (available at https://earthexplorer.usgs. gov/). The scenes were displayed with bands 3, 4, and 2 as red, green, and blue, respectively, to enhance the discrimination of fire scars. The 1979 scars were mapped in ArcGIS manually by drawing polygons around fire scars from a difference image created using imagery from 1972 and 1979. All scars on the 1972 image were mapped since no difference image could be created, as there was no previous Landsat image. The extent of the fires went well beyond the four chosen Landsat scenes. However, we limited the mapping to the extent of the four mentioned scenes. For the early 1980s “short-interval fire mapping” study, we used Landsat 4 and 5 imagery (WRS paths 104 and 105 and rows 76 and 77) downloaded from the USGS Earth Explorer (available at https://earthexplorer.usgs.gov/). Scenes from 1982 were compared with those from 1983 (an interval of about 6 months) to create a difference image. The scars of the fires were digitized in ArcGIS. The same methodology was repeated for the fires over the summer of 1984–1985. The fire scars of interest were all completely contained within the examined Landsat scenes. RESULTS Pre-contact Fire Several observations were made by explorers that indicate that large fires were occurring during the pre-contact period in the Western Desert. These observations were made in the earlymid 1870s during expeditions led by Col. P. Egerton-Warburton (hereafter referred to as Col. Warburton) and W.E.P. (Ernest) Giles (Table 2 and Figure 2). From the long-term weather data, seasonal conditions during these expeditions appear to have been highly favorable. The ONI for 1870 was −1.0 or lower across all weather stations, suggesting that La Niña conditions prevailed Frontiers in Ecology and Evolution | www.frontiersin.org 5July 2021 | Volume 9 | Article 666241 fevo-09-666241 July 5, 2021 Time: 15:55 # 6 Wright et al. Rainfall-Linked Megafires in Spinifex Grasslands FIGURE 2 | Map showing approximate locations of explorer’s observations of pre-contact fires in Western Desert. Circle size corresponds to relative fire sizes. Numbers associated with circles indicate estimated fire sizes (ha). Where observations give a range of fire sizes, the larger fire size is presented. Gray lines indicate Australian state and territory boundaries (South Australia, Western Australia, and Northern Territory). Triangles indicate large contemporary settlements and weather stations. and that higher-than-average rainfall would have occurred across much of northern and central Australia during the summer of 1870/1871 (Figure 3). Rainfall records of Springsure, Wentworth, and Ororroo weather stations provide further evidence that a continental-scale extreme wet period had occurred at this time (Figure 3 and Supplementary Appendix Figures 1,2). All stations showed extreme-high-rainfall years in the early 1870s, with 1870 being the highest rainfall year on record for the Wentworth and Orroroo weather stations and the fifth highest rainfall year on record for Springsure (Australian Government Bureau of Meteorology, 2021). Both Giles and Col. Warburton mentioned large-scale fires during their expeditions. While traveling through the Mt Wilson region of Western Australia in 1873, Col. Warburton traveled over 27 miles (43.5 km) of burnt ground over 2 days. This suggests a minimum fire size of c. 58,000 ha (assuming an elliptical fire with a 2:1 length:breadth ratio) and a maximum fire size of c. 148,000 ha (assuming a circular fire). Giles, while traveling through the southern Gibson Desert in 1876, estimated that under high-wind conditions, fires in spinifex habitat could “run on for scores of miles,” implying that he had observed fires that had burnt for distances of at least 40 miles (64 km) but possibly up to 60 miles (96 km) or more. If the fires he had witnessed had been 40 miles in cross section, then burnt areas could have been between 163,000 (assuming elliptical fires) and 325,000 ha (assuming fires were circular). If the fires had run for three score miles (96 km), then areas burnt could have been between 366,000 ha (if elliptical fires) and 731,000 ha (if circular fires). There were several observations of smaller fires (<1 to c. 5000 ha) during unexceptional rainfall years in the late nineteenth and early- to mid-twentieth centuries. These include observations made by Tietkins, Finlayson, Burrows and Christenson, De Graaf, and Gould (Table 2 and Figure 2). From our remote sensing study of the Lake Mackay region, numerous smaller fires were observed on the 1972 imagery, which may have been, at least partially, the result of traditional burning conducted by the Pintupi Nine (Figure 4). These small-scale fires were obliterated in the 1979 imagery, which showed c. 7.17 million ha (i.e., 66%) of the mapped area burnt. These large-scale fires presumably occurred in response to fuel accumulations after the recordbreaking rains of 1974/1975. It could be argued that these 1970s fire observations are irrelevant to the PBM hypothesis, as low population size would have limited the spatial extent of patch burning by the Pintupi Nine. By this time, other traditional-living Aboriginal peoples in the region had come out of the desert and were settled in communities such as Haasts Bluff, Papunya, and Balgo (Hansen, 1984;Heffernan, 1984). However, population densities in the extreme-low productivity Gibson Desert were always low, traditionally in the order of 1 person per 200 - 260 km2(Long, 1971), so the images may still be relevant. Frontiers in Ecology and Evolution | www.frontiersin.org 6July 2021 | Volume 9 | Article 666241 fevo-09-666241 July 5, 2021 Time: 15:55 # 7 Wright et al. Rainfall-Linked Megafires in Spinifex Grasslands FIGURE 3 | Twenty-four-month cumulative rainfall and rainfall anomaly data from Wentworth weather station in southwest NSW. Bottom panel shows Oceanic Niño Index (ONI) 3.4 from 1870 to present. Major rainfall events/periods associated with big arid zone wildfires are shown in gray bands. Aboriginal Fire Myths Several Western Desert myths were identified that mentioned extreme-large scale fires. In the Warlukurlangu myth (Kimber, 1983), a large blaze ignited by the lungkata (blue tongue) ancestral spirit began in Warlpiri tribal country close to present-day Yuendumu community and burnt in a southward direction to Pitjantjatjara country, some 400-km distant. The minimum dimensions of this fire, if circular, would have been c. 12.6 million ha, and if an ellipse with a 2:1 length:breadth ratio, would have been c. 6.3 million ha. In another myth, from the arid Lake Mackay region of Western Australia, a large fire that ignited close to the edge of Lake Mackay forced an ancestral being to flee underground to Nyinmi rock hole, some 150 km to the southwest (Tjupurrula, 2014). If this fire had burnt all the way to Nyinmi, then the approximate dimensions of this fire would have been c. 1.8 million ha if circular and c. 0.9 million ha if an ellipse with a 2:1 length:breadth ratio. There are also several mythologies from arid regions outside the Western Desert that report large spinifex fires. These include the “Ahiite” myth, where a large fire begins in Alyawarra country on the Sandover River and burns across 350 km in a southwest direction to Rubuntja, a fire dreaming site of the Aranda tribe in the western MacDonnell ranges (Moyle and Morton, 1986; Memmott, 1992). There is also an Alyawarra “traveling fire dreaming,” which concerns a fire that begins near the Plenty river in the southern Northern territory and burns across c. 400 km of the Triodia dominated Wakaya Desert to the Table Top region in western Queensland, c. 40 km southwest of Camooweal (Keys and Memmott, 2016; Memmott, P. pers. Comm. 2021). 1980s Short-Interval Fires Our mapping of the 1980s short interval fires from the southwest Northern Territory confirmed the report of Allan and Southgate (2002) that large areas of spinifex grassland had re-burnt in this region only 2 years after initial fires had occurred. The mapping showed that fires in late 1982/early 1983 burnt c. 1.95 million ha. These fires presumably occurred in response to rainfall-triggered fuel accumulations after exceptionally high rains in 1981–1982 (Figure 5). During this time, the Derwent meteorological station in the far NE of study area indicates 932-mm rainfall fell in the 18-m period prior to April 1982 (the Derwent has an annual average precipitation of 327 mm) (Supplementary Appendix Figure 4;Australian Government Bureau of Meteorology, 2021). Approximately Frontiers in Ecology and Evolution | www.frontiersin.org 7July 2021 | Volume 9 | Article 666241 fevo-09-666241 July 5, 2021 Time: 15:55 # 8 Wright et al. Rainfall-Linked Megafires in Spinifex Grasslands FIGURE 4 | 1970s fire scars from homelands of the Pintupi Nine, west of Lake Mackay in the Gibson Desert, Western Australia. Fires in light gray were evident on 1979 satellite imagery and are presumed to have occurred in response to fuel accumulations after high rainfall years of 1974/1975. Fires in black were burnt twice (1972 and 1979), and areas in dark gray were burnt in 1972 only. White areas were unburnt throughout study period. 2 years later, large fires in late 1984 burnt c. 1 million ha in the region, with c. 470,000 ha burning over areas that had already burnt in 1982/1983. These follow-up fires presumably occurred in response to fuel accumulations triggered by continued exceptionally high rainfall in 1983 and early 1984 (Figure 5 and Supplementary Appendix 3). During this time, the Derwent rainfall records indicate that 719-mm rainfall fell during the 18- month period prior to April 1984 (Supplementary Appendix Figure 4;Australian Government Bureau of Meteorology, 2021). DISCUSSION Pre-contact Fire Sizes The fires identified by Col. Warburton in 1873, Giles in 1873 and 1874, and the 1970s Lake Mackay mapping of this study are examples of massive rainfall-linked fires during the pre- European period when traditional Aboriginal patch burning was still being practiced. The size of these fires is similar in scale to larger arid Australian wildfires recorded following high-rainfall periods in the contemporary era (Griffin et al., 1983;Allan and Southgate, 2002;Allan et al., 2003;Verhoeven et al., 2020) and dwarfs modern fires observed during periods of average or below-average rainfall (Turner et al., 2008). Observations of large pre-contact fires directly contradict one of the central tenets of the PBM hypothesis—that indigenous patch burning should constrain large wildfires by fragmenting landscapes and breaking up fuel loads (Minnich, 1983, 1998). They are, however, consistent with paleoecological (dendrochronological) records from the United States that indicate episodic large-scale fires were a prominent feature in certain American pre-contact landscapes, despite frequent burning by indigenous people (Mensing et al., 1999). Additional evidence that large-scale spinifex fires occurred during the pre-European period may come from Western Desert Aboriginal fire mythologies. Previous authors have suggested that indigenous oral traditions can corroborate real events from the temporally distant past (i.e., euhemerism) (Vitaliano, 1973; Barber and Barber, 2004). Nunn and Reid (2015) have argued that because Australian Aboriginal cultural protocols place great importance on traditional knowledge, precise transgenerational transmission of oral traditions can occur over exceptionally long time periods. Extremely large wildfires in several Western Desert Aboriginal mythologies could therefore derive from real fires from the ancient pre-contact period. The scale of the fires mentioned in the myths is massive and, if they were actual fires, would have left large tracts of country uninhabitable for a period and presumably left long-lasting impressions on Aboriginal peoples living in the burnt regions. A highly referenced series of pre-contact Aboriginal fire scars from the Western Desert are those captured by the 1953 Blue Streak and 1960/1961 aerial photography (Burrows and Chapman, 2018;Clark, 2020). These fire observations have been held as strong evidence that the pre-contact fire Frontiers in Ecology and Evolution | www.frontiersin.org 8July 2021 | Volume 9 | Article 666241 fevo-09-666241 July 5, 2021 Time: 15:55 # 9 Wright et al. Rainfall-Linked Megafires in Spinifex Grasslands FIGURE 5 | Large-scale short-interval wildfires in west central Australia in the early 1980s. Patches in light gray show the extent of a single fire that occurred in 1982/1983. Patches that are dark gray burnt in 1984/1985 and patches that are black burnt in both periods. White areas within the 1982/1983 fire were unburnt throughout the study period. Rainfall during the study period at “The Derwent” weather station is given in Supplementary Appendix 4. regime of the Western Desert was characterized by patchy, small-scale, cool-season fires and that large summer fires in the contemporary era are anomalous. However, to properly determine if traditional burning constrained large rainfall-linked fires would require assessing imagery taken at a time of known high rainfall when widespread fuel accumulations would have been present. Although some parts of Australia were very wet in the early 1950s, especially western Queensland (see Springsure data—Supplementary Appendix Figure 2), weather records are available for the early 1950s from the Giles Metrological and Warburton weather stations. Both stations are within c. 150 km of the areas captured in the Blue Streak and 1960/1961 aerial photography, and both indicate only average to slightly belowaverage rainfall conditions during the years immediately prior to the time the photos were taken (Australian Government Bureau of Meteorology, 2021;Supplementary Appendix 4). Short-Interval Spinifex Fires Our mapping of large, short-interval fires during a highrainfall period in the Western Desert in the 1980s demonstrates unequivocally that the effectiveness of recently burnt areas to act as fire breaks can degenerate rapidly when rainfall conditions are good. Such observations of rainfall-triggered short-interval (2–3 years) fires in spinifex grasslands are not unique, with numerous other studies/authors having reported short-interval spinifex grassland fires on Aboriginal lands (Allan and Southgate, 2002;Wright, 2007), pastoral properties (Latz, 1990;Allan and Southgate, 2002), and fire-managed arid conservation reserves such as Uluru Kata-Tjuta (Ayer’s Rock) National Park (Allan et al., 2003) and Finke Gorge National Park (Matthews, 2005). The mechanism that allows post-fire seral spinifex to return to a flammable state quickly after high rainfall is rapid growth rates of highly flammable non-spinifex grass and forb species (Latz, 1990; Allan and Southgate, 2002). The C4tussock grass, kerosene grass (Aristida holathera), is one of the most important contributors by weight in seral spinifex vegetation, comprising up to 80% of vegetation biomass in the first 2 years after fire (Figure 1C and Table 1;Griffin et al., 1983;Latz, 1990; TERN, 2019). This grass species has seeds that are triggered to germinate by smoke (Wright, 2007;Wright and Clarke, 2009), and although generally an annual, it can behave as a perennial following successive good rainfall seasons that permit continued growth (Latz, 1990). Fires fueled primarily by A. holathera are fast burning (hence the common name “kerosene grass”), though the fuel loads of Aristida-fueled fires (c. 0.5 kg m−2) are lower than Triodia-fueled fires (c. 1.55 kg m−2) (Wright and Clarke, 2008) and hence generally burn at a lower intensity. The Evolutionary Fire Regime of Arid Spinifex Grasslands An obvious corollary to the suggestion that large rainfall-linked spinifex grassland fires occurred during the pre-European period Frontiers in Ecology and Evolution | www.frontiersin.org 9July 2021 | Volume 9 | Article 666241