Fossil evidence for open, Proteaceae-dominated heathlands and fire in the Late Cretaceous of Australia

Geological background

The Bundey Basin, which hosts the HUC11 sediments, is a small basin located ∼140 km northeast of Alice Springs, in the Northern Territory of Australia (∼22°39′25″S, 135°15′04″E; Fig. 1). During the Late Cretaceous, this region was at ∼45°S, and Australia was broadly juxtaposed with Antarctica (Fig. 2). The Bundey Basin forms eastward continuity with the larger (1200 km²) Waite Basin (Edgoose and Ahmad, 2013) and underlies the valley of a small tributary of the (usually dry) Bundey River, the headwaters of which drain the Harts Range to the south (see Freeman, 1986). These basins onlap the northern margins of the Paleozoic to Mesoproterozoic Arunta Inlier, one of two uplifted, fault-bounded crustal blocks that dominate the landscape of central Australia. The Bundey Basin is infilled with less than ∼127 m of interbedded claystones, siltstones, sandstones and conglomerates, with chalcedonic limestone at shallow depths, a sequence considered equivalent to the Waite Formation (Senior et al., 1995).

Much of central Australia was flooded by marine incursion in the middle Aptian and early Albian, and trace numbers of Early Cretaceous marine dinocysts in the Bundey Basin indicate that saline and paralic environments occurred along the northern margin of the Arunta Inlier during this period. Final regression of the seaway from central Australia during the Cenomanian left a low-lying area of shallow lakes and mires. Sedimentary evidence, including that from previous studies of carbonaceous claystones at ∼95–108 m depth in HUC11, indicates that these environments persisted into the Maastrichtian (Truswell, 1987; Senior et al., 1995; Macphail, 1997).

The current paleodrainage networks in central Australia are believed to have originated during the Late Mesozoic (Woodgate et al., 2012), and by the end of the Cretaceous, the fundamental physiography and paleovalley architecture of the region was established (Fujioka and Chappell, 2010). Subsequent deep weathering and erosion have largely destroyed whatever Late Cretaceous organic deposits had accumulated in the Late Mesozoic paleovalleys, but the thin carbonaceous siltstone in HUC11 and the lignite at the base of the Cenozoic succession in the Ayers Rock Basin (Freeman, 1986; Truswell, 1987; Macphail, 1997) are known exceptions.

Being within Australia's arid center, the Bundey River region now has a mean annual rainfall of only ∼300 mm, maximum summer temperatures frequently >40°C, and mean annual potential evaporation of ∼2500 mm (Neave et al., 2006). The typical vegetation of the region is mulga (Acacia aneura Benth.) shrubland, with patches of eucalypt open woodland and spinifex (Triodia spp.) grassland.

Recovery and examination of microfossils

Core chip samples of dark gray to carbonaceous shale from the interval 96.00–108.14 m in HUC11 were originally obtained from the NTGS Core Library, Alice Springs, Northern Territory, in 1996 (Macphail, 1997). These samples were processed for palynological examination by Laola Pty Ltd, Perth, Australia. Extra samples within 96.00–108.14 m were obtained from NTGS in 2013, and subsamples at 102.65 m and 105.65 m were processed and microscope slides prepared for the current study by Core Laboratories Australia Pty Ltd, P. O. Box 785, Cloverdale 6985, Australia using standard techniques of oxidation and filtration through 10 µm sieve cloth to recover palynomorphs. All slides from the initial (1996) and recent preparations were examined. Palynotaxa were photographed using a Zeiss (Jena, Germany) Axiophot microscope fitted with a Zeiss (Jena, Germany) Axiovision digital image capturing camera and software. The time distributions of microfossil taxa within the sampled interval of HUC11 are shown in Table 1, with taxonomic authorities following the International Plant Names Index (http://www.ipni.org/). Table 2 summarizes the distributions of age-diagnostic taxa. HUC11 microfossil taxa are illustrated within Figs. 3, 6 and 7.

Dating of microfossils

Microfloras preserved in central Australian basins are dated using the time distribution of palynotaxa shared with the continental margin basins in southeastern Australia, in particular in the major reference basin for the Late Cretaceous and Cenozoic in southern Australia, the Gippsland Basin (charts 3 and 4 of Partridge, 2006; see also Stover and Partridge, 1973). In this palynostratigraphy, the late Late Cretaceous time is subdivided into three biozones, (1) the middle to basal Late Campanian Tricolporites (al. Tricolpites) lilliei Zone, (2) the Late Campanian to early Late Maastrichtian Lower Forcipites longus Zone, and (3) the middle Late to latest Maastrichtian Upper Forcipites longus Zone. The boundaries between these zones are defined by the First Appearance Datum (FAD) of T. lilliei, F. longus and/or Quadraplanus brossus and Stereisporites (al. Tripunctisporis) maastrichtiensis, respectively in southeastern Australia. Here we use the FAD of S. maastrichtiensis to subdivide the HUC11 sequence into Upper and Lower Forcipites longus Zone Equivalent intervals. The upper boundary of the Upper Forcipites longus Zone is marked by a major extinction horizon defined by the last appearance of morphospecies including F. longus, Q. brossus, Ornamentifera sentosa, Propylipollis crotonoides and/or T. lilliei. This horizon coincides with the K/T Event at 65.5 Ma. Caveats to the use of Gippsland Basin age range data in central and northwestern Australia have been discussed previously (Macphail and Stone, 2004; Macphail, 2007), and include that, as might be expected, at least some taxa show apparent diachronisms. However, there are no reasons to doubt the applicability of the key age-diagnostic taxa.

Recovery and examination of mesofossils

Subsamples of ∼5 g each from the HUC11 core samples were disaggregated in warm water and washed through a 300 µm sieve. The captured material was then searched for intact plant material under a dissecting microscope. The bulk remaining material was then soaked in household bleach (sodium hypochlorite 42 g/L) for several hours, and then rinsed, stained with safranin O solution (1% w/v safranin O in 20% methylated spirits), placed in a petri dish with water and further searched. Cuticular fragments were picked out using fine forceps and mounted on glass slides in glycerin jelly for light microscopy and photography (Nikon [Tokyo, Japan] E200 microscope with Tucsen [Fuzhou, China] 5.0 megapixel digital camera). Some of the same and further pieces of cuticle were placed on aluminum stubs and carbon or platinum coated for scanning electron microscopy (SEM) using a Philips (Eindhoven, Netherlands) XL 30 FEGSEM operated at 10 kV at the University of Adelaide and an Hitachi (Tokyo, Japan) SU-70 FESEM operated at 10 kV at the University of Tasmania. All microscope slides and SEM stubs are housed at the School of Biological Sciences, University of Adelaide. Leaf fossil material is illustrated in Figs. 4, 5 and 7.

Identification of leaf remains

Leaf remains were assessed with respect to descriptions in the literature and especially to Proteaceae cuticle preparations housed at the University of Adelaide. This collection comprises 600 species and includes all genera and multiple species of all large genera. Fossils were identified to Proteaceae based on the combined presence of brachyparacytic stomata and trichome bases that appear as donut-like rings (Carpenter and Jordan, 1997; Carpenter et al., 2005; Carpenter, 2012), and proposed to belong to taxa within the family based on a range of distinctive features, including apparent synapomorphies for subfamily Grevilleoideae, tribe Banksieae, including Banksia (Carpenter et al., 2010a) and subfamily Persoonioideae, including tribe Persoonieae (Carpenter et al., 2010b). Although we do not attempt formal assignment of any leaf fossil taxa at genus level in the current study, we note that close matches between some of the fossils and extant Proteaceae can readily be found.

Age of the sediments

Microfossils from the HUC11 samples were generally well-preserved, abundant and diverse (Table 1), providing an extensive resource for dating the sediments. The presence of Quadraplanus brossus (Fig. 3A), a more sparsely ornamented variant of Q. brossus referred to here as Quadraplanus sp. cf. Q. brossus and Tricolporites lilliei (Fig. 3B), demonstrates that the sediments from 96.00 to 108.14 m were deposited during the Late Campanian to Maastrichtian (Forcipites longus Zone Equivalent time). Within this interval, the section from 96.00 to 98.04 m is dated as late Late Maastrichtian (Upper Forcipites longus Zone Equivalent time; ∼67–65.5 Myr old) and the section from 99.82 to 108.14 m as Late Campanian to early Late Maastrichtian (Lower Forcipites longus Zone Equivalent; ∼75–67 Myr old), based on the presence or absence of Stereisporites (al. Tripunctisporis) maastrichtiensis (Fig. 3C), respectively (Table 2).

The HUC11 samples yielded low numbers of poorly preserved Early Cretaceous dinoflagellates which are likely to have been reworked from sediments deposited during marine transgression of the region during the Aptian and Albian. Several spore taxa (e.g., Neoraistrickia truncata and Vallizonosporites tegmentus) that are more typical of the Late Jurassic and Early and mid-Cretaceous were also recorded. The presence of these taxa could also reflect reworking, or perhaps, later extinction in Central Australia than in the Gippsland Basin.

Fossil composition

In addition to the extensive range of microfossil taxa (Table 1), leaf and cuticular fragments, megaspores, charcoalified tissues and other structures were recovered from the Late Campanian to early Late Maastrichtian interval (99.82–108.14 m) in HUC11. Foliar remains comprising very small leaves and cuticular fragments of <1 mm² were sparse (most sediment samples yielded <20 such fragments), but there was no evidence of meaningful floristic change within the interval.

Proteaceae

There was strong evidence from both foliar (Table 3; Figs. 4, 5A, 5B) and pollen (Table 1, Fig. 6) fossils for Proteaceae being the most diverse taxonomic group in the HUC11 sediments, and presumably this family was a dominant component of the local vegetation during the latest Cretaceous. At least 12 foliar taxa were recovered, including at least two that are represented by amphistomatic cuticular sleeves, each only ∼1–1.5 mm wide (Fig. 4A), which we reconstruct as being derived from filiform leaves or segments of small, divided leaves. Sclerophylly, and possibly xeromorphy, was evident in the form of sunken guard cells and thick cuticles, including evidence of subepidermal cuticularisation (Fig. 4C–E) (see Hill, 1998; Jordan et al., 2005). Many of these fossils (Fig. 4A–J) show all the features characteristic of subfamily Proteoideae (Table 3), and some (especially Proteaceae taxa 1–4; Fig. 4A–G) conform closely to leaves of open-habitat genera such as Stirlingia, Isopogon, and Conospermum (Fig. 5C–F). Other HUC11 cuticle taxa (Table 3) are consistent with subfamily Grevilleoideae (Fig. 4K–O), including Banksia (tribe Banksieae) (compare Fig. 4K, L with Fig. 5G; Fig. 5B with 5H; see also Carpenter et al., 2010a), and subfamily Persoonioideae, tribe Persoonieae (compare Fig. 5A with 5I; see also Carpenter et al., 2010b).

Table 3. HUC11 foliar fossils that have diagnostic characters of Proteaceae.

	Character states
Taxon	1	2	3	4	5	6	7	Proposed nearest living relative
1	+	+	+	+				Subfamily Proteoideae, cf. Conospermum
2	+	+	+	+				Subfamily Proteoideae, cf. Stirlingia
3	+	+	+	+				Subfamily Proteoideae, cf. Conospermum
4	+	+	+					Subfamily Proteoideae, cf. Isopogon
5	+	+	+					Subfamily Proteoideae
6	+	+	+					Subfamily Proteoideae
7	+	+	+					Subfamily Proteoideae
8	+	++			+	+		Subfamily Grevilleoideae, Banksia
9	+	+						Subfamily Grevilleoideae
10	+	++						Subfamily Grevilleoideae
11	+	+						Subfamily Grevilleoideae
12	+	+	+				+	Subfamily Persoonioideae, Persoonieae

• Notes: Character states are as follows: (1) brachyparacytic stomata; (2) donut-like trichome bases (++ indicates some bases associated with more than one epidermal cell); (3) stomata aligned more-or-less parallel with long axis of leaf; (4) definitely amphistomatic; (5) thickened cylindrical trichome bases; (6) raised subsidiary cells; (7) mean guard cell length > 50 µm.

Fossil pollen attributed to Proteaceae was very diverse in HUC11, and usually amounted to more than half the total abundance of angiosperm taxa recorded at each depth (Table 1, Fig. 6). Many of these >30 pollen taxa have not been formally described, but at least 10 could be referred to, or are closely comparable with, taxa from Late Cretaceous–Cenozoic deposits in the Gippsland and Otway Basins of southeastern Australia (Stover and Partridge, 1973; Dettmann and Jarzen, 1996, 1998). Several of these described taxa closely resemble pollen of modern genera of Proteaceae (Pocknall and Crosbie, 1988; Martin, 1995; Dettmann and Jarzen, 1996, 1998; Milne, 1998), and/or have been accepted as belonging to extant clades of Proteaceae after rigorous phylogenetic assessment of morphological traits (Sauquet et al., 2009a,b; Barker et al., 2007). In particular, Cranwellipollis palisadus (Fig. 6A), with its clavate sculpturing, has been accepted as belonging to the subfamily Proteoideae lineage now only represented by the open-habitat SWAFR genus, Franklandia (Sauquet et al., 2009a, b), and Beaupreaidites spp. (Fig. 6B, C) are indicative of Beauprea (Pocknall and Crosbie, 1988; Milne, 1998), or more conservatively, the subfamily Proteoideae clade now comprising Beauprea, Faurea and Protea (Sauquet et al., 2009a, b). Protea is a flagship genus of the CFR, while Beauprea and Faurea have species in both open and more mesic habitats in New Caledonia and Africa/Madagascar, respectively. Propylipollis ambiguus (Fig. 6D) has the synapomorphy (Barker et al., 2007) of spinulose ornamentation (Feuer, 1990) and has accordingly been placed in tribe Embothrieae, subtribe Embothriinae. Subtribe Embothriinae is now represented by the South American genera Embothrium and Oreocallis and the southeastern Australian genus Telopea, which are found in both open woodland/montane habitats and rainforests, and the Australian rainforest genus Alloxylon. Propylipollis crotonoides (Fig. 6E) belongs to the mainly rainforest tribe Macadamieae (Sauquet et al., 2009a). Interestingly, given our finding of Banksia-like cuticular fragments, we also recorded rare diporate pollen (Fig. 6F) not inconsistent with that of tribe Banksieae (see Cookson, 1950; Johnson and Briggs, 1975; Barker et al., 2007). However, in the absence of detailed study, this fossil pollen could represent an unusual form of an otherwise triporate proteaceous taxon (Feuer, 1990), or indeed, an angiosperm of unknown affinities.

Other taxa

As shown in Table 1, freshwater aquatic and wet habitat fossil taxa were common in HUC11 and included algal cysts (e.g., Saeptodinium), cryptogam spores (e.g., Cyatheaceae, Gleicheniaceae and Sphagnum [Fig. 7A]) and megaspores (Fig. 7B) attributable to Marsileaceae and lycophytes including Selaginellaceae. Gymnosperm pollen was overall not abundant and was mostly represented by Araucariaceae and Podocarpaceae, with rare grains of Cycadales and the extinct conifer family Cheirolepidiaceae. Some minute, keeled, scale-leaves possibly assignable to Araucariaceae (Fig. 7C, D) were present, as well as cuticles of several Podocarpaceae taxa (Fig. 7E), a cuticle fragment indistinguishable from the extant Australian cycad Bowenia (Zamiaceae) (Fig. 7F) and several other gymnosperms of unknown affinities.

Angiosperm cuticles other than those attributed to Proteaceae were generally scarce, but at least one type of monocot cuticle was recovered (Fig. 7G). Monocot pollen was well-represented and included several taxa of Liliacidites (Liliales) and Arecaceae (palm) pollen (Fig. 7H, I). Ericaceae pollen tetrads (Fig. 7J) were also diverse, and other dicot taxa represented by pollen included Anacolosa (Olacaceae), Ilex (Aquifoliaceae), Winteraceae and Gunneraceae. Several other angiosperm taxa, including Dicotetradites clavatus, have no known close living relatives.

The nature of latest Cretaceous vegetation and environment in Australia

Very diverse fossils attributed to Proteaceae, including several pollen types that belong to open-habitat lineages and leaves that conform to very small/fine-leaved, sclerophyllous extant species of subfamily Proteoideae (Fig. 7M, N), provide compelling evidence for the Late Cretaceous presence of open vegetation in central Australia, and great antiquity of the sclerophyll biome. Growth in the open is especially indicated by amphistomaty (Jordan et al., 2014). Similarly, all HUC11 leaf fossils show more-or-less straight epidermal walls (e.g., Figs. 4, 5A, 5B), a condition associated with relatively sunny habitats (Dunn et al., 2015). Further, as elsewhere in the Mesozoic world, including Australasia (e.g., Townrow, 1967; Pole, 1995, 2000; Pole and Douglas, 1999), minute-foliaged gymnosperms occur in HUC11. By analogy to extant scale or needle-leaved conifers, which are now mostly limited to relatively open sites where they can be photosynthetically competitive with angiosperms (Brodribb et al., 2007; Biffin et al., 2012), these fossil gymnosperms probably also occupied open habitats.

Open habitats within the ancient landscapes of Central Australia are likely to have been related to oligotrophy. In particular, high abundances of freshwater algal cysts and spores of Sphagnales suggest that the HUC11 sediments accumulated in a Sphagnum fenland setting, an exemplar oligotrophic habitat in modern ecosystems (Rydin and Jeglum, 2006). Spores and pollen of a range of damp-habitat taxa that are typically associated with such fens were fossilized, including gleicheniaceous ferns and some herbaceous angiosperms, e.g., Gunneraceae, Callitrichaceae, and diverse Liliales. Less frequently inundated (e.g., raised) areas within the fen probably supported sclerophyllous heath dominated by Proteaceae, but also with various other Liliales, prominent and diverse Ericaceae, and low-growing gymnosperms, perhaps including Microcachrys. Local saline influence, as indicated by low numbers of the dinocyst, Cerodinium obliquipes (Deflandre & Cookson) J.K.Lentin & G.L.Williams in many HUC11 samples, may have influenced the degree of sclerophylly. We also predict from the palynological evidence that individual taller trees or palms and/or closed thickets with more fire-sensitive, shade-tolerant species, e.g., other conifers and Proteaceae, Chloranthaceae and Anacolosa, were present within the pollen source area.

Our data support previous pollen-based inferences for Proteaceae being the most successful angiosperm “invaders” of the Mesozoic austral vegetation that was dominated by podocarps and araucarian conifers (Dettmann, 1994; Dettmann and Jarzen, 1996, 1998). Although abundances of taxa vary, latest Cretaceous palynofloras of similar morphological diversity to those in the central Australian HUC11 core have been reported from southern Australia (Dettmann, 1994; Dettmann and Jarzen, 1996, 1998; Wagstaff et al., 2006; Gallagher et al., 2008) and northwestern Australia (Macphail, 2007). Relatively open vegetation types with gymnosperms and Proteaceae were therefore probably widespread and possibly locally dominant across the Australian landscape, albeit with some evidence of phytogeographical regionalism, including the greater presence of thermophilous taxa at lower latitudes (Macphail, 2007).

Cretaceous remnants in the Southwest Australian Floristic Region?

Central Australian environments are now extremely arid and they do not support Sphagnum-dominated communities, nor any species of Proteoideae (Fig. 1). The main concentrations of Proteaceae-dominated sclerophyll communities are now in the Mediterranean-climate Cape Floristic Region and Southwest Australian Floristic Region (Fig. 7O) (Cowling et al., 1996; Hopper and Gioia, 2004; Linder, 2005; Hopper, 2009; Allsopp et al., 2014; Mucina et al., 2014), where the SWAFR is of special significance as a center of ancient phylogenetic diversity for the family (Johnson and Briggs, 1975; Sauquet et al., 2009a). In fact, our diverse Cretaceous fossils of at least Proteaceae and Ericaceae raise the possibility that the current SWAFR vegetation may represent a vestige of a more widespread Australian flora, and this proposal receives some support from molecular studies of other taxa. For instance, dated phylogenies suggest that in addition to Franklandia and several other Proteaceae clades that are strongly represented in the SWAFR (Sauquet et al., 2009a, b), many more SWAFR plant lineages (Magallón et al., 2015) also have Australian Cretaceous origins, including the monotypic, presumably relict families Anarthriaceae, Cephalotaceae, Dasypogonaceae, and Eremosynaceae (Mucina et al., 2014). Not only does the SWAFR harbor these families, it is also a clear center of diversity for other ancient groups, including Boryaceae, Centrolepidaceae, Ericaceae, Hydatellaceae and Restionaceae (Mucina et al., 2014). Apart from the retention of sufficiently moist climates in parts of the SWAFR, the persistence of ancient lineages there is probably testament to the long and continuous stability of the SWAFR landscape (Hopper and Gioia, 2004; Hopper, 2009; Mucina et al., 2014) and to the early evolution of strategies that enable plant growth in the world's most phosphorus (P)-impoverished soils (Lambers et al., 2010). A competitive advantage in extremely P-limited environments may even have been a trait of the mid-Cretaceous common ancestor of all Proteaceae: clusters of short, lateral “proteoid” cluster roots that are highly efficient at extracting key nutrients are found widely in the family (Johnson and Briggs, 1975; Weston, 2007; Lambers et al., 2010), and a large component of the rich diversity in the SWAFR is explained by the presence of these and other specialized, nonmycorrhizal structures in other lineages (Lambers et al., 2010).

HUC11 taxa that are now obviously absent from the SWAFR include lineages that became extinct at the end of the Cretaceous (e.g., Cheirolepidiaceae and other gymnosperms) and lineages whose extant relatives are mostly confined to the subtropics–tropics (e.g., Anacolosa and palms) or to consistently humid, cool habitats (e.g., certain cryptogams, Lagarostrobos, Microcachrys, Ascarina, Gunneraceae and Winteraceae). We therefore envisage that the central Australian Cretaceous vegetation also had similarities to that now found in ultramafic regions of New Caledonia, where diverse Proteaceae (including Beauprea), ferns, palms, and conifers occur within open maquis or adjacent forest thickets. Other similar extant environments include subalpine regions of Tasmania, where sclerophyllous Proteaceae co-occur with Winteraceae, Gunneraceae, Ericaceae and diverse Podocarpaceae including Microcachrys and Lagarostrobos, and the Sydney sandstone region of eastern Australia, where sclerophyllous communities with Telopea and other Proteaceae intergrade with more mesic vegetation.

Although gymnosperms are now present in southwestern Australia (Podocarpus, callitroid Cupressaceae and Macrozamia), these gymnosperms belong to clades that, on the basis of molecular evidence, are probably much younger than Cretaceous age (Piggin and Bruhl, 2010; Biffin et al., 2011; Ingham et al., 2013). However, HUC11 gymnosperm taxa including Araucariaceae, Dacrydium, Dacrycarpus, Phyllocladus and a range of ferns are known from Eocene macrofossils in southwestern Australia (Hill and Merrifield, 1993; McLoughlin and Hill, 1996; Carpenter et al., 2014), and palynostratigraphic evidence suggests that these taxa persisted there until the mid-Pliocene, when they occurred as rare elements within local, fire-affected sclerophyll vegetation not unlike that in the SWAFR today (Bint, 1981; Atahan et al., 2004; Dodson and Macphail, 2004). Banksia leaves were also prominent in the Eocene of southwestern Australia and are suggestive of a very long, continuous history in the region (Carpenter et al., 2014). Indeed, some pollen evidence indicates that forms of Proteaceae-dominated, open sclerophyll vegetation existed during the Eocene in southern Australia (Milne, 1998; Itzstein-Davey, 2004). This evidence is consistent with the hypothesis of refugia (recently termed “edaphic ghettos” [Bond, 2015]) for such vegetation in the absence or scarcity of fire, probably within oligotrophic, perhaps swampy regions that could not support the presumably more closed rainforests that appear to have largely replaced the Cretaceous heathlands by the latest Paleocene to middle Eocene (Macphail, 2007). The Australian sclerophyll biome then undoubtedly re-expanded during the Neogene as the climate became overall drier, and probably reached its Cenozoic acme prior to the Pleistocene glacial–interglacial cycles in regions of high rainfall, summer wet (non-Mediterranean) climates (Sniderman et al., 2013). The current rich diversity of the SWAFR can be explained at least in part by this region being relatively less affected by these climatic oscillations (and the development of the contemporary Mediterranean climate) than at least the sclerophyll floras of the southeastern Australian coastal margins, where fossil evidence shows that numerous extinctions occurred (Sniderman et al., 2013).

The role of fire

The presence of common, anatomically preserved, charcoalified material in the HUC11 core shows that fire occurred in central Australia, just as it appears to have done elsewhere across the “high-fire” Late Cretaceous Earth, when high atmospheric oxygen levels combined with periodically dry climates were conducive to burning, including within peatlands (Belcher et al., 2010; Bond and Midgley, 2012; Brown et al., 2012). Ignition sources in central Australia were almost certainly lightning strikes, because Late Cretaceous volcanism is unknown in this region (Quilty, 1994). Widespread burning is likely to have driven a net loss of P from terrestrial environments in the Late Cretaceous (Brown et al., 2012), and we therefore propose that wildfires promoted the radiation of Australian sclerophyllous lineages, including Proteaceae and perhaps epacrid Ericaceae (subfamily Styphelioideae), by exacerbating the depletion of P-levels in Australian soils.

Our fossil data strongly support molecular evidence that Proteoideae already occurred in fire-prone habitats in the mid–Late Cretaceous (Lamont and He, 2012). Also, our evidence of Banksia suggests that this genus may be much older than estimated from molecular evidence (Sauquet et al., 2009a) and therefore may help to resolve the problem that fire-related traits in Banksia were estimated to have originated during the Paleogene (He et al., 2011), when climates were probably wet, and burning must have been uncommon (Macphail et al., 1994). However, the contentious question (Hopper, 2009; Bradshaw et al., 2011; Keeley et al., 2011; Bowman et al., 2014) of whether traits such as serotiny are adaptations to fire or exaptations remains unanswered, and it is unknown to what extent Cretaceous angiosperms were fire-promoting or requiring. Certainly, we detected no evidence of Eucalyptus (or other Myrtaceae) in the HUC11 sediments, consistent with previous fossil (Thornhill and Macphail, 2012) and phylogenetic (Crisp et al., 2011) evidence that this genus, which now dominates flammable woodlands in Australia, originated somewhat later.

Finally, whatever combination of climatic factors, fire intervals and fire intensities prevailed in the Late Cretaceous, it is apparent that diverse gymnosperms and sclerophyllous Proteaceae coexisted in the landscape over many millions of years.

Open vs. rainforest biomes

Australian closed rainforests are often regarded as representing the ultimate ancestral source of other vegetation types, and are popularly portrayed as 100-Myr-old remnants of a flora that covered Gondwana. Recent syntheses of available molecular phylogenetic data for Australian biota have been used to infer the rainforest habitat in Australia as being ancestral to the sclerophyllous habitat (Byrne et al., 2011), and in particular, that the SWAFR heathland flora had its origins within, or proximal to, rainforest communities during the Eocene (Byrne et al., 2014). These views are broadly consistent with classic interpretations of the evolution of the Australian flora, where sclerophyll elements were envisaged to have evolved in the early Cenozoic from rainforest stock, and to have occurred at the lower end of the nutrient gradient on rainforest margins (e.g., Beadle, 1966, 1981; Johnson and Briggs, 1975; Specht, 1981). However, combined with the evidence of a concentration of other ancient lineages in the SWAFR (Mucina et al., 2014) and in the montane heathlands of Tasmania (Jordan et al., in press), our evidence of sclerophyllous Proteoideae in the Cretaceous challenges the ancestral rainforest paradigm and places new emphasis on the antiquity of the open sclerophyll biome. With respect to Proteaceae, we also note that Byrne et al. (2011) recognized the results of an analysis by Jordan et al. (2005), that although the rainforest environment optimized as ancestral in subfamily Grevilleoideae, it did not for the family as a whole.

Although the especially poor Australian record of Campanian (∼83.5 Ma) to mid-Paleocene (∼60 Ma) plant macrofossils (Dettmann et al., 1992; Douglas, 1994) limits discussion, the known Australasian mid–Late Cretaceous leaf floras that predate the substantial radiation of Proteaceae during the Santonian to Maastrichtian (∼85–70 Ma) (Dettmann and Jarzen, 1998) also do not show clear evidence of closed rainforests: rather, they share rich gymnosperm components and mostly Fagalean-like angiosperm leaves that are reminiscent of temperate, deciduous forests of the northern hemisphere (McLoughlin et al., 1995; Pole, 2015). Palynotaxa belonging to lineages that are now well represented in rainforests, including palms, Chloranthaceae, Trimeniaceae and Winteraceae, do occur in HUC11 and other Australian Campanian–Maastrichtian deposits, some of which also have Nothofagus pollen (Dettmann, 1994; Dettmann and Jarzen, 1996, 1998; Wagstaff et al., 2006; Macphail, 2007; Gallagher et al., 2008). Flowers and foliar material attributable to Lauraceae are also known from Maastrichtian assemblages in New Zealand (Pole and Douglas, 1999; Kennedy, 2003; Cantrill et al., 2011). However, our results imply that at least some of the Cretaceous representatives of the lineages listed above did not belong to complex, closed rainforests, or that the extent of rainforest vegetation was less than often assumed. Similarly, we suggest caution in using ancestral state reconstructions based on palm phylogeny to infer mid-Cretaceous rainforests in Laurasia (Couvreur et al., 2011). Such reconstructions also cannot properly account for environmental drivers that have been lost or profoundly altered over time. For instance, it is probable that because of low incident light angles, vegetation structure at very high latitudes could not have been closed, regardless of whether lineages now largely confined to rainforests were present (Specht et al., 1992; Dettmann, 1994; Hill, 1994).