I ntr o d u cti o n: t h e pr o bl e m of r ai nf or e st e d g e s

O nce distri b ute d o ver m ost of t he c o nti ne nt ( Hill, 1 9 9 4 ), A ustr ali a's re m ai ni n g r ai nf orests n o w c o ver o nl y c. 0. 5 % of t he l a n d are a, fr a g me nte d i nt o a ' mesic arc hi pel a g o' i n t he e aster n p orti o n of t he c o u ntr y ( Fi g. 1 ; K a n o ws ki et al., 2 0 0 9 ; B yr ne et al., 2 0 1 1 ; K o o y m a n et al., 2 0 1 3 ). Fre q u e ntl y a dj oi ni n g a n d i nter mi xi n g wit h t he r ai nf orest are wet, t all e uc al y pt f or ests ( T E Fs, ofte n k n o w n i n A ustr ali a as ' wet scler o p h yll f orests'; > 3 0 m i n hei g ht). T he T E Fs are d o mi n at e d b y trees i n E uc aly pt us s.l., c ollecti vel y k n o w n as 'e uc al y pts' ( E uc aly pt us , A ng o p h or a a n d C ory m bi a ), a n d ot her M yrt ace a e, a n d t he y occ ur i n h u mi d cli m ates si mil ar t o r ai nf orests. T he y occ u p y c. 0. 7 5 % of A ustr ali a, i ncl u di n g a versi o n of t he T E F i n Wester n A ustr ali a wit h o ut nei g h b ori n g r ai nf orest n ot disc uss e d f urt her here (e. g. W ar dell-J o h ns o n et al., 2 0 1 7 ). T he r ai nf orest-T E F i nterf ace is me di ate d b y fire, cli m at e, t o p o gr a p h y, e d a p hic f act ors, a n d pl a nt tr aits ( Be a dle & C osti n, 1 9 5 2 ; As ht o n, 1 9 8 1 ; As h, 1 9 8 8 Alt h o u g h n u mer o us eff orts are i n pl ace t o pr ote ct t he m, A ustr ali a n r ai nf orests are f ar fr o m 's a ve d'. R ai nf orest hist oric al cle ari n g ( 2 8 -3 5 %) w as a p pr o xi m atel y t wice t h at of t he T E Fs ( 1 5-1 8 %; Fi g. 1 ; Br a ds h a w, 2 0 1 2 ). Re m ote se nsi n g a n al yses i n dic ate d 1. 3 % a n d 2. 4 % net l osses f or t he 2 0 0 0-2 0 1 2 i nter v al tropical TEFs and that 'leaving it to nature is not an option'. Bradford (2018) recommended burning in Rose Gum (E. grandis; Fig. 1) forests of tropical Queensland every 3-5 yr, 'regardless of the fire regimes that shaped these communities'. Australian rainforest 'expansion' (Fig. 1) has been characterized as an indicator of potentially catastrophic global carbon cycle imbalances with 'the potential to significantly alter the earth system within a relatively short time frame' (Tng et al., 2012a). Chapman & Harrington (1997) stated that 'to maximize biological diversity in the wet tropics of north Queensland, it is necessary to maintain the full spectrum of natural habitats. Fire management is therefore required to maintain the wet eucalypt forest and its dependent fauna'. Chapman & Kofron (2010) wrote that 'unless fire has a central role in managing tropical wet sclerophyll forest, then this forest type and its dependent species will cease to exist'.

Although we acknowledge the detailed spatial analyses required to document changing forest boundaries, the 'invasion' argument has numerous issues. It often omits historical context, uses post-clearing baselines of forest boundaries, labels native plants as 'invasive,' emphasizes fire interventions, and selectively focuses on protecting a few tropical TEF vertebrate species (Table 1) at the apparent expense of the greater vertebrate diversity, endemism, and threatened species concentrations in adjacent rainforests (Nix & Switzer, 1991;Williams et al., 2010;WTMA, 2016). The Wet Tropics contain 45% of all Australian vertebrate species, and critical conservation areas for rainforest animals are concentrated on the western edge (WTMA, 2016), where the effects of human modification of the rainforest-TEF interface are the highest. Despite the urgent warnings, little evidence indicates that rainforest edge movements or insufficient burning has caused any biodiversity catastrophe, and there are few monitoring studies (Bradford & Harrington, 1999;Tng, 2019). By contrast, excessive fire, logging, agriculture, clearing, and invasive predators are well-documented threats to tropical TEF (and rainforest) species, including, ironically, the vertebrates most commonly stated to require increased rainforest suppression through burning for survival (Table 1). The logging and burning of large trees remove vital tree-hollow habitats, a little-studied factor that has detrimental effects on several TEF species (Woinarski et al., 2016;Tng, 2019). The core idea that active fire promotion is required Major Vegetation Groups (MVGs) 1 and 2 defined at https://www.dcceew.gov.au/environment/land/publications/nvis-fact-sheet-series-4-2; the NVIS pre-1750 estimates and extant data are available at https://www.dcceew.gov.au/environment/land/native-vegetation/national-vegetation-information- system. The NVIS data indicate losses since 1750 of 28% of Australian rainforests and vine thickets (MVG1) and 18% of TEFs (MVG2), similar to the findings of Bradshaw (2012: fig. 3). Areas of TEFs in central Victoria are missing because they were classified (incorrectly) in MVG3 by NVIS, with negligible influence on the overall loss estimates because they were absent from both the historical and extant datasets.

New Phytologist (2025) 245: 1355-1365 www.newphytologist.com Ó 2024 The Author(s). New Phytologist Ó 2024 New Phytologist Foundation.

Viewpoint

Forum

New Phytologist to maintain TEFs lacks experimental support (Fensham & Fairfax, 2002).

Historical decreases in Aboriginal fire management have been invoked as the principal justification for fire suppression in many of the listed papers. However, Aboriginal practices provide questionable equivalence for justifying programmed burning because of their vastly different objectives, seasonal timing, and scale (Hill et al., 2000). Australia's tropical rainforests also have high biomass, often exceeding that of the TEFs (Unwin, 1989;Bradford et al., 2014;Wardell-Johnson et al., 2017). Suppressing them with fire puts Australia at odds with the world-wide consensus for conserving biodiverse, carbon-dense rainforests (e.g. Muthee et al., 2022;Mo et al., 2023) to prevent extinction and forest conversion from carbon sinks to sources (Baccini et al., 2017), a process accelerated by excessive fire (Datta & Krishnamoorti, 2022).

In contrast to the 'invasive' narrative, other literature characterizes Australian rainforests more broadly to include the TEFs not as a distinct, adversarial forest type but as an intermittent disturbance zone dominated by rainforest pioneer plants, which in Australia overwhelmingly consist of the charismatic, frequently enormous eucalypts (van Steenis, 1958;Smith & Guyer, 1983;Adam, 1992;Tng et al., 2012bTng et al., , 2014;;Lindenmayer et al., 2024). Long-lived, opportunistic rainforest pioneer trees world-wide, including Australian eucalypts, were well defined by van Steenis (1958) as ecological 'nomads,' which take advantage of bare soils exposed by processes including volcanic flows, erosion, fires, landslides, and anthropogenic clearing. Leaf traits analyzed across TEF-rainforest transitions showed no significant differences (Tng et al., 2013), further reducing the distinctiveness of TEFs from rainforests regarding functional ecology. In this context, tropical TEFs have been proposed as rainforest fringe habitats dominated by eucalypt nomads that evolved pioneer-like traits under various nonfire disturbance regimes (Smith & Guyer, 1983).

Advancing the debate about tropical TEFs requires historical context from the fossil record, which has often been invoked as a missing or data-deficient element (e.g. Hopkins, 1981;Unwin, 1989;Adam, 1992;Wardell-Johnson et al., 1997) but now has much to offer from recent discoveries. We now turn to fossil data to address fundamental questions. Were eucalypts associated with rainforests through geologic time? If so, is human intervention necessary today? Have most eucalypts always been fire-dependent, or were they exapted (preadapted) to fire in earlier settings? What can we learn from the fossil record to improve forest conservation? We review the fossil history of eucalypt-rainforest interactions and show that Australian tropical rainforests and TEFs represent a modern iteration of a single interacting system that can be traced back to at least 50 million years ago (Ma) in Gondwana, with no evidence of significant fire mediation until geologically recent times. Based on the fossil data, we reject the 'invasive' narrative and conclude by highlighting the need for improved modern conservation and management of the demonstrably ancient, evolutionarily significant rainforest-TEF interface.

Eucalypts in rainforests through time

Eocene Patagonia

Until recently, the eucalypt fossil record was sparse and confined to Australia and New Zealand, limiting direct insights into the group's early history and biogeography (e.g. Hill et al., 2016). This situation changed with the recent discovery of the oldest and best-preserved Eucalyptus fossils at the early Eocene (52 Ma) Laguna del Hunco fossil lake site in Chubut, Patagonian Argentina (Fig. 2). Fossil Eucalyptus frenguelliana leaves are preserved there in great abundance (Fig. 3), along with extraordinary additional evidence for the genus from complete infructescences, flower buds, and a flower with in situ pollen (Fig. 2; Gandolfo et al., 2011;Hermsen et al., 2012;Zamaloa et al., 2020). The combined evidence is sufficient to place the fossils in the Eucalyptus crown group, and they preserve diagnostic features of the large subgenus Symphyomyrtus.

Eucalyptus leaves, preserving characteristic oil bodies and occasionally still attached to twigs, were the second most abundant (12% of specimens) of the > 200 plant macrofossil species collected from several thousand fossils at the site (Fig. 3; Wilf et al., 2005). These abundance data were obtained from quantitative field counts, and modern leaf litter studies support the idea that the ranked abundance of fossil leaves reflects the relative biomass of their respective source trees (Burnham et al., 1992). Therefore, E. frenguelliana trees most likely were abundant and reached large sizes. Their ecological importance is also evident from the diverse insect-feeding damage observed in the leaf fossils, of which every fossil damage type has been found in living Australian wet-forest eucalypt species, suggesting remarkable conservatism and host tracking (Giraldo et al., 2024). Laguna del Hunco is the only South American location that unambiguously preserves fossils of Eucalyptus (reviewed by Hermsen et al., 2012) and many other genera. The site's floristic distinctiveness presumably reflects both its outstanding fossil preservation and its status as the only well-sampled paleobotanical site in Patagonia (and South America) from the early Eocene, illuminating the long (c. 16 Myr) regional gap between the early Paleocene Salamanca/Las Flores and middle Eocene R ıo Pichileuf u floras (e.g. Wilf et al., 2013).

The 170 m thick section of fossil lake beds exposed at Laguna del Hunco belongs to the Tufolitas Laguna del Hunco of the Huitrera Formation and consists mainly of tuffaceous clastic sediments (Arag on & Mazzoni, 1997;Wilf et al., 2003;Gosses et al., 2021;Hajek et al., 2024). The Tufolitas filled a caldera lake of considerable size (c. 30 km N-S) that existed for c. 200 Kyr. Plant material was deposited principally through slope failures of the surrounding, steep caldera walls due to earthquakes, overwet soils, and overgrown vegetation; subsequent rapid transport to the lake bottom allowed excellent preservation of diversity and New Phytologist (2025) 245: 1355-1365 www.newphytologist.com Ó 2024 The Author(s). New Phytologist Ó 2024 New Phytologist Foundation.

Viewpoint

Forum

New Phytologist morphological details (Hajek et al., 2024). The age of the section is tightly constrained to the 52.2-52.0 Ma window by multiple 40 Ar-39 Ar and U-Pb analyses on primary tuffs from several levels interbedded with the fossiliferous strata and associated igneous rocks (Wilf et al., 2003;Gosses et al., 2021;Hajek et al., 2024). This precise determination dates the fossils to the early Eocene climatic optimum and the terminal phase of Gondwana, the last time a moist forest corridor existed from Patagonia across frost-free Antarctica to Australia (Wilf et al., 2013).

An extraordinary diversity of other well-preserved plant and animal life occurs in the same beds as the Eucalyptus fossils, providing many additional connections to living forests. Notably, many Laguna del Hunco fossil plant taxa now inhabit Australian rainforests and interact directly with Eucalyptus, including the fern Todea; the conifers Agathis, Araucaria, and Podocarpus; angiosperms such as Ceratopetalum, Gymnostoma, Orites, Macaranga, Akania, and Ripogonum; and others closely related to Daphnandra, Wilkiea, Zygogynum, and Ackama (for summaries and primary references, see Wilf et al., 2005Wilf et al., , 2019Wilf et al., , 2023;;Barreda et al., 2020;Matel et al., 2022). Additional Laguna del Hunco genera are also well known as fossils from Australia but survive elsewhere, including the abundant (Fig. 3) water-demanding conifers Papuacedrus, Acmopyle, Dacrycarpus, and Retrophyllum, whose extinction in Australia is attributed to increasing water stress (Brodribb & Hill, 2004). These taxa comprise some of the numerous examples of how Australia served as the 'life raft' enabling the survival of multiple Gondwanan paleo-Antarctic rainforest lineages that persisted there or eventually dispersed to island Australasia, Malesia and beyond (Kooyman et al., 2014;Wilf et al., 2019). The climate was frost-free and everwet (perhumid), inferred from the drought intolerance of many living relatives of the plants and confirmed by the extraordinary preservation of highly drought-sensitive accessory transfusion tissue in Acmopyle leaves (Andruchow-Colombo et al., 2023) and vessel-less Winteraceae wood (Brea et al., 2021).

Reconstructing the landscape ecology of the Patagonian fossil eucalypts is challenging because the remains were transported into a large lake basin with no information regarding their original locations. However, they co-occurred on the same bedding planes as all the taxa listed, which means that they were deposited during the same transport events and grew close to or intermingled with traditional rainforest elements. The environment was dynamic, and the steep volcanic habitat provided significant disturbances (Hajek et al., 2024); we suggested earlier (in Gandolfo et al., 2011) that the ancient Eucalyptus trees were opportunists that readily colonized areas with freshly exposed mineral soils. Today, several non-Australian rainforest eucalypts specialize in nonfire disturbances, including volcanic clearance, fluvial erosion, landslides, and clear-cuts, such as the Rainbow Gum (E. deglupta) in New Guinea (Carr, 1972;Paijmans, 1976;Johns, 1986). Even in Australia, eucalypts recruit in gaps, and cyclones frequently damage large areas of tropical rainforest, a process thought to facilitate eucalypt recruitment at the margins, although direct documentation is scarce (Unwin et al., 1988;Tng et al., 2012aTng et al., , 2012b)). As in New Guinea's volcanic province and similar modern environments (Ganteaume & Syphard, 2018), fires at Laguna del Hunco would have been rare and limited to lightning strikes and volcanic ignition during active eruptions, as corroborated by the lack of definite charcoal in the site sediments (Barreda et al., 2020;V. Barreda, pers. comm. 2024). The caldera slopes probably featured a complex Abundance 600 400 200 0 Dobineaites ameghinoi Eucalyptus frenguelliana Castaneophyllum patagonicum 'Schmidelia' proedulis Fabaceae sp. TY117 'Myrica' mira Agathis zamunerae 'Coprosma' incerta Rhamnaceae sp. TY25 Lauraceae sp. TY84 Malvales sp. TY23 Lomatia preferruginea Lauraceae sp. TY122 Cunoniaceae sp. TY116 dicot sp. TY56 Araucaria huncoensis Podocarpus sp. TY9 Dacrycarpus puertae Lomatia occidentalis 'Cupania' grosse-serrata Araucaria huncoensis Retrophyllum oxyphyllum Papuacedrus prechilensis Racemofructus fasciculatus 'Myrcia' deltoidea Ginkgoites patagonicus dicot sp. TY30 Acmopyle grayae dicot sp. TY106 dicot sp. TY38 Angiosperm leaves Angiosperm reproductive Gymnosperm leaves Gymnosperm reproductive Fig. 3 Top 30 most abundant plant organs (of > 200 known) at the early Eocene Laguna del Hunco site in Chubut, Argentina, based on raw counts of 4303 total fossils from all quarries combined, nomenclature updated from Wilf et al. (2005). Abundant families include Anacardiaceae (Dobineaites), Myrtaceae (Eucalyptus and 'Myrcia'), Fagaceae (Castaneophyllum), Fabaceae, Araucariaceae (Agathis and Araucaria), Rhamnaceae, Lauraceae, Proteaceae, Cunoniaceae (sp. TY116 and Racemofructus, considered related to Weinmannia s.l.), Podocarpaceae (Podocarpus, Dacrycarpus, Retrophyllum, Acmopyle), and Cupressaceae (Papuacedrus). Ó 2024 The Author(s). New Phytologist Ó 2024 New Phytologist Foundation. New Phytologist (2025) 245: 1355-1365 www.newphytologist.com

mosaic of regenerating forest patches from mixed disturbance cohorts, as seen today on New Guinea landslide surfaces (Johns, 1986).

From the high abundances of both Eucalyptus and standard rainforest elements and their persistence throughout the 200 Kyr window of fossil deposition (Wilf et al., 2005; Fig. 3), we see that neither rainforests nor eucalypts extinguished the other. Instead, natural expansion, contraction, disturbance, and regeneration maintained a single interacting system. The caldera lake margin forests included diverse life-history strategies embodied by Eucalyptus and a few probable others (Macaranga, Physalis, and Gleicheniaceae) as disturbance specialists, whereas most taxa represented closed forests. However, the closed forest category is more heterogeneous than is often assumed; disturbance opportunism in Australia, New Guinea, and elsewhere is well known in several conifer and angiosperm genera represented by fossils at Laguna del Hunco (e.g. Johns, 1986Johns, , 1990;;Enright & Hill, 1995). Pole (2019) recently suggested the alternate hypothesis that Laguna del Hunco eucalypts inhabited closed forests. This explanation is intriguing but less likely because of the direct sedimentological evidence for multiple disturbance types, including frequent slope failures associated directly with fossil deposition events (Hajek et al., 2024), which would have limited the area of closed forests and provided abundant fresh mineral soil exposures for recruitment. The vegetative and reproductive traits in the Eucalyptus fossils are virtually indistinguishable from their living counterparts (including foliar oil bodies, operculate corollas, and valved, capsular fruits; Hermsen et al., 2012), implying conserved life-history strategies.

Australia and New Zealand

Despite a large number of paleobotanical localities (e.g. Hill, 1994;Kooyman et al., 2014), the eucalypt macrofossil record from Australia and New Zealand remains limited, especially for Eucalyptus s.s. However, Eucalyptus s.l. leaves and a few fruit capsules have been recorded since the early Eocene in Australia, and several incompletely studied reproductive fossils hold promise for detailed investigations using new techniques (Rozefelds, 1996;Hermsen et al., 2012;Hill et al., 2016;Pole, 2019;Thornhill et al., 2019;Rozefelds et al., 2024). The oldest well-accepted occurrence of Eucalyptus s.s. consists of leaves from Berwick Quarry (Victoria, late Oligocene/early Miocene; Pole et al., 1993;Hill et al., 2016). Neogene and younger Eucalyptus reproductive fossils are known from Chalk Mountain, Miocene of New South Wales (Holmes & Anderson, 2019), Bacchus Marsh, late Miocene of Victoria (Ladiges, 1997), Stony Creek, early Pleistocene of Victoria (e.g. Sniderman et al., 2013), and various Cenozoic fluvial and basaltic silcrete deposits (Rozefelds et al., 2024).

There are several informative patterns in the scarce record. First, even the best-studied Paleogene sites lack definite evidence of eucalypts, such as Anglesea, Golden Grove, and Cethana (e.g. Hill, 1994). Second, unlike Laguna del Hunco, Eucalyptus macrofossils are not reported to be abundant at any time in Australia, except for a leaf mat from Berwick Quarry (Pole et al., 1993). The excellent preservation of abundant Eucalyptus at Laguna del Hunco in Argentina (Figs 2, 3) indicates no intrinsic preservation bias against this genus, in contrast to some taphonomic studies (e.g. Hill & Gibson, 1986). Thus, we consider the overall rarity of Paleogene Eucalyptus macrofossils in Australia to reflect low original abundance, at least locally, although palynological evidence for eucalypts is more common (Thornhill et al., 2019).

Third, eucalypt macrofossils from Australia and New Zealand almost always co-occur with typical rainforest elements. For example, Berwick Quarry preserves Dacrycarpus, Gymnostoma, and Nothofagus, leading to the interpretation of a mixture of rainforests and disturbance-mediated open forests in a volcanic landscape with no evidence of fire (Pole et al., 1993;Hill et al., 2016), a setting comparable to Laguna del Hunco. Several other Australian fossil sites with eucalypts have associated volcanic strata (Rozefelds et al., 2024), and volcanism was widespread in eastern Australia from the Eocene to the Miocene, continuing into the Pleistocene in tropical Queensland (Jones et al., 2017). Like Laguna del Hunco, volcanic eruptions, extrusive flows, ash falls, and landslides were presumably important factors in forest disturbance, whereas fire ignition was mostly limited to lightning strikes (Enright & Thomas, 2008). Rainforest elements are also present at Neogene-Pleistocene sites that preserve Eucalyptus capsules, including Chalk Mountain (Agathis, Ceratopetalum, Dendrocnide; Holmes & Anderson, 2019) and Bacchus Marsh (Nothofagus, Araucaria, Dacrycarpus; Greenwood, 1994). Nearly all these associated genera (except Dendrocnide) are reported from Laguna del Hunco (Barreda et al., 2020). The combined evidence from the austral records suggests that Eucalyptus has spent much of its history as a rainforest disturbance specialist in largely nonfire disturbance regimes.

Fourth, it is striking that Eucalyptus is, thus far, not known from nearly all pre-Quaternary assemblages with well-characterized sclerophyll or open forest signatures, including several Paleocene and younger sites that preserve well-vetted fossils of the classical sclerophyll element Banksia (e.g. Carpenter et al., 2014Carpenter et al., , 2016;;McCurry et al., 2022). Some localities preserve direct evidence of fire disturbance in charcoal records, such as the Eocene-Miocene Latrobe sequence in Victoria (Korasidis et al., 2019), or xeromorphic adaptations, as seen in fossil Banksia leaf anatomy (Carpenter et al., 2014).

Eucalyptus reached high abundances in Australian pollen assemblages by the late Miocene to early Pliocene (e.g. Nullarbor Plain, southwestern Australia; Sniderman et al., 2016), approximately coeval with increased fire, as indicated by charcoal counts (Kershaw et al., 1994). Eucalyptus involvement in fire-prone sclerophyll assemblages is well-captured in a geologically young and well-preserved early Pleistocene deposit at Stony Creek, Victoria. (Sniderman & Haberle, 2012;Sniderman et al., 2013). The Stony Creek macrofossil and pollen floras preserve a nonanalog, diverse, humid sclerophyll forest type, wherein charcoal frequency is correlated positively with periods of Eucalyptus pollen dominance and negatively with rainforest intervals. Thus, Stony Creek is the oldest well-preserved example of interacting fire, rainforests, and sclerophyll vegetation with confirmed Eucalyptus macrofossils, demonstrating natural fire response and survival of Australian rainforest and sclerophyll vegetation together in a prehuman, c. 1.5 Ma, mixed-forest setting.

New Phytologist (2025) 245: 1355-1365 www.newphytologist.com Ó 2024 The Author(s). New Phytologist Ó 2024 New Phytologist Foundation.

Viewpoint

Forum

New Phytologist

There are no relevant Eucalyptus macrofossils from dryland biomes because of well-known preservation biases. However, a calibrated molecular study detected robust signals for species diversification across five sections of subgenus Symphyomyrtus, beginning only c. 3 Ma (Thornhill et al., 2019). This result aligned well with isotopic evidence from leaf waxes in marine cores off northwest Australia for an increase in Australian C 4 plant abundance, considered related to increased rainfall seasonality and fire, at c. 3.5 Ma, later than most other continents (Andrae et al., 2018). Similarly, evidence from the sedimentary geochemistry of river muds from Western Australia showed significant fluctuations in moisture since 5.3 Ma and evidence for increasing aridity and rainfall seasonality since c. 3.8 Ma (Stuut et al., 2019).

Detailed information about the eucalypt-rainforest-fire system since the late Pleistocene primarily comes from pollen, charcoal, and other proxy data from volcanic maar lakes in northern Queensland and the spectacular discovery of subfossil Eucalyptus charcoal in the Daintree rainforest (Hopkins et al., 1993(Hopkins et al., , 1996)). These datasets show significant contractions, fragmentations, and re-expansions of rainforest and eucalypt forests through time (e.g. Kershaw, 1994;Dimitriadis & Cranston, 2001;Haberle, 2005;Kershaw et al., 2007;Haberle et al., 2010). However, there has been little assessment of volcanic disturbance to the paleovegetation in this area, which includes extensive Quaternary basalt deposits and maar lakes (Dimitriadis & Cranston, 2001). The potential effects of Aboriginal populations on forest dynamics have been extensively discussed, especially regarding the promotion of fire (Cremer, 1960;Kershaw, 1994;Hopkins et al., 1996;Kershaw et al., 2007). However, there is scarce evidence that Aboriginal fire management had transformative impacts in wetter regions of the continent (Enright & Thomas, 2008;Roberts et al., 2021;Lindenmayer et al., 2024). Most large Aboriginal fires observed by early settlers were set for resistance purposes, not to clear land (Hill et al., 2000). In the Wet Tropics, some Aboriginal fire practices that conserve localized resources in closed and open forests remain (Hill et al., 2000;Roberts et al., 2021).

The profound impacts on Australian forests since European colonial activities include those of agriculture, grazing, logging, introduced species, mining and programmed burning (e.g. Hill et al., 2000;Bradshaw, 2012;Fig. 1). Queensland has lost c. 50% of its tropical forests and 74% of its lowland and tableland rainforests since European arrival, mainly for agriculture, and it currently loses more forest annually than any other Australian state (Bradshaw, 2012;Neldner et al., 2023). Queensland's rainforests were heavily cleared because they were thought to hold the most fertile soils, and their foliage was considered a food source for beetles that fed on sugarcane crops; some supposedly harmful rainforest 'expansion' represents natural recovery from massive sugarcane clearing and mining (Hill et al., 2000(Hill et al., , 2001)). Today, many examples of purported rainforest 'encroachment' that are linked to current fire-management policies (Harrington & Sanderson, 1994;Russell-Smith et al., 2004;Tng et al., 2012a;Stanton et al., 2014;Krishnan et al., 2019) appear to be measured against postclearing baselines. It is more likely that the 'encroachment' represents small recoveries from earlier, large-scale deforestation (Hill et al., 2000) that removed twice the coverage of rainforests vs TEFs (Fig. 1).

Cattle and horse grazing, often combined with fire management, are significant factors that increase the loss rate of TEFs (by nine times in one study) through the removal of grass at the rainforest margin, expediting recruitment by rainforest plants (Unwin, 1989;Hill et al., 2001;Campbell & Clarke, 2006).

Discussion

Large, emergent eucalypts in or around rainforest assemblages can be explained either by reclassifying these areas as nonrainforest due to eucalypt presence or by recognizing that eucalypts reflect interactions between shade-tolerant, long-lived rainforest trees, and opportunistic, long-lived, shade-intolerant pioneers that regenerate on disturbed mineral soils. The strategy of long-lived opportunists (nomads; see the Introduction section), with numerous analogs globally, is to arrive early under high-light conditions with mineral soil exposure, with initially low levels of competition, and to leave late, persisting in the stand as tall emergent trees (Fig. 1).

Postdisturbance rainforest pioneers, other than eucalypts, are often described as having a mix of traits associated with short lifespans, such as low wood density, low stature, small seeds, shade intolerance, and low leaf mass per area; however, other early secondary and opportunist nomad-type species (e.g. some Cecropia species in the Neotropics; Dendrocnide and Flindersia in Australia) can germinate quickly after disturbance, grow rapidly, and persist as large trees. Eucalypts have traits that allow them to become some of the largest plants world-wide, reflecting similar, opportunistic ecological strategies. Forest dynamics can favor plant traits and strategies differentially, including the effects of long-time-interval disturbances and expansion-contraction driven by climatic oscillations. Under some circumstances, these processes lead to the intermixing of plant ecological strategy types, reflecting short and long regeneration intervals and long-timescale biogeographic patterns.

The fossil evidence shows that Eucalyptus has been an integrated disturbance specialist in rainforest assemblages for most of its known history, suggesting that Australia's TEFs are a modern, firemediated iteration of ancient rainforest disturbance zones dominated by eucalypts that extended far beyond Australia. Until perhaps the past 5 Myr, the disturbances largely came from nonfire causes, and eucalypts apparently were absent from sclerophyll vegetation. Disturbances that create opportunities by allowing natural contraction and expansion of rainforests have resulted from volcanic activity and dormancy, as seen in ancient Patagonia, modern New Guinea, and Eocene-Miocene Australia, as well as landslides, fluvial processes, and climate change. The emergence of fire disturbance in the late Cenozoic of Australia was associated with the natural contraction and expansion of rainforests and eucalypt forests, but not with the extirpation of either (Sniderman & Haberle, 2012).

The close affinities of eucalypts with rainforest Myrtaceae (e.g. Eucalyptopsis, Stockwellia; Ladiges et al., 2003;Thornhill et al., 2019) have long suggested that they originated in humid habitats with limited fire. However, a molecular phylogenetic study found that epicormic resprouting, a signature fire adaptation in living eucalypts, maps deeply onto the phylogeny of living Australian Eucalyptus species, supporting the view that fire ecology has been prevalent in the crown lineage since the Cretaceous (Crisp et al., 2011). This conclusion rested on the assumptions that fire adaptation in deep time can be inferred, without fossils, from mapping one trait onto molecular phylogenies and that fire has always been the singular stimulus for the trait. Nevertheless, epicormic resprouting also expedites regenerative responses in eucalypts and other plants to nonfire disturbances (Burrows, 2013;Kenefick et al., 2024), all of which are applicable to early eucalypts, including mechanical injuries (landslides, wind, and herbivores), droughts, and lava flows.

Overall, the evidence indicates that early wet-forest eucalypts were nomads adapted to nonfire disturbance, enabling them much later in their history to take remarkable advantage of novel, expanding fire regimes by exaptation (van Steenis, 1958;Smith & Guyer, 1983). The Eocene Laguna del Hunco fossils occur in a highly disturbed paleoenvironment with frequent mineral soil exposure (Hajek et al., 2024) and without evidence of fire. They preserve foliar oil bodies, which are implicated today in herbivore deterrence and increased flammability, and thick-walled capsular fruits that confer seed protection, whose living analogs allow postfire seed release (e.g. Hermsen et al., 2012;Hern andez et al., 2022).

We view the tropical TEF-rainforest interface as an ancient evolutionary system that provides ample but underappreciated fulfillment of World Heritage criteria viii and ix for conserving evolutionary processes and stages in the evolution of life (https://whc.unesco.org/en/criteria). These valuable forests should be restored and conserved (Tng et al., 2014;Lindenmayer et al., 2024) to support species fitness and resilience via evolution through natural selection at the rainforest edge. We find no historical or experimental basis for the established but longdisputed (Gilbert, 1959) idea that fire intervention is required to maintain rainforest-TEF boundaries (see Fensham & Fairfax, 2002). This contention is directly refuted by charcoal and pollen data from Pleistocene Australia, which show that well before human influence, rainforests and eucalypt forests intermingled and regenerated over time in response to fire regimes (Sniderman & Haberle, 2012). We see no justification for the widespread burning of the margins of Australia's remaining rainforests (Fig. 1) to make even more space for Eucalyptus, a single genus that already dominates 77% of native Australian forests (ABARES, 2018), than it would occupy as part of the natural expanding and contracting rainforest systems that existed for tens of millions of years before humans.

We partly agree with Chapman & Harrington (1997) that 'to maximize biological diversity in the wet tropics of north Queensland, it is necessary to maintain the full spectrum of natural habitats,' but this should involve restoring and not intensively burning the specific habitat of dynamic rainforest margins. Oxymoronic terminology that labels native tropical rainforests as 'invasive' and 'encroaching' is deeply ingrained in the literature and forestry practice. This viewpoint reflects a shortterm perspective that overlooks deep time and human history, natural forest dynamics, and the disproportionate historical and continuing losses of Australian rainforests compared to TEFs (Fig. 1; Ash, 1988;Fensham & Fairfax, 2002;Bradshaw, 2012;Hansen et al., 2013;Tng, 2019;Bauman et al., 2022). Little evidence suggests that rainforest expansions have reduced biodiversity or population sizes except at limited spatial scales (Baker et al., 2020). By contrast, clearing, logging, increased fire, exclusion fencing, barbed wire, invasive weeds, and carnivores are well-documented sustained threats to tropical wildlife over large areas (Table 1).

Outlook

We recommend revised management policies to re-establish the natural dynamics and demographics of the interface zone, prioritizing both natural rainforest recovery and the integrity of TEFs, with reduced logging to allow mixed forests with charismatic and ecologically significant giant eucalypts to reemerge. Precise measures require discussion, development, monitoring, economic planning, and interdisciplinary investigations, and we avoid being prescriptive. However, additional actions could prioritize minimized grazing at the interface and emphasize strategic, small-scale fire management, in close consultation with the First Nations, to protect forest dynamics, essential wildlife habitats, and traditional knowledge and ways of life. Rainforests should not be used as firebreaks; burning should instead be directed away from the rainforest. Promising scientific avenues include intensified gathering of historical data and modeling to improve the precision and widespread use of presettlement, not postclearing baselines (Fig. 1); increased monitoring studies for endangered species at the interface; assessments of multiple disturbance types on fire characteristics, stand demographics, and biotic interactions; and precise quantification of changes in three-dimensional burned and unburned forest structures over time using remote sensing and terrestrial lidar scanning (Forbes et al., 2022;Lindenmayer et al., 2022). The rewards will be considerable, including survival and significantly increased health of ancient rainforests and TEFs, thereby increasing the value of Australian contributions to biodiversity and World Heritage.