Rainforest persistence and recruitment after Australia’s 2019–2020 fires in subtropical, temperate, dry and littoral rainforests

Keywords: basal resprouting, fire, fire-cued seeding, persistence niche, plant resilience traits, rainforest, resilience, succession, woody encroachment.

Fire plays a major role in controlling rainforest boundaries, with rainforest communities usually limited to topographic positions that are sheltered from frequent fire (Bowman 2000; Wood et al. 2011; Ondei et al. 2017; Bond 2019). Yet the response of rainforest flora to fire is complex, and many rainforest species from tropical to temperate latitudes are capable of surviving occasional fires by resprouting and fire-cued seedling recruitment (e.g. Unwin 1983; Hill and Read 1984; Floyd 1990; Bowman 1991; Williams 2000; Marrinan et al. 2005; Campbell and Clarke 2006; Williams et al. 2012; Tolsma et al. 2019). Postfire resprouting and seedling recruitment are key life-history traits that confer resilience, profoundly influencing community composition and ecosystem function through differential species persistence after fire (Bond and Midgley 2001; Pausas et al. 2004). Together, these persistence traits provide a valuable basis for predicting postfire community assembly, and resilience of species and ecosystems to changing climate and fire regimes (Pausas et al. 2004; Pausas and Bradstock 2007).

Under typical weather conditions, fires originating in fire-prone ecosystems typically extinguish along rainforest margins (Wood et al. 2011; Collins et al. 2019), where the closed rainforest canopy reduces ecosystem flammability through a combination of reduced grassy fuels, lower temperatures, higher relative humidity, and higher moisture content and decomposition rates of litter fuels (Hoffmann et al. 2012b; Little et al. 2012; Alexander and Arthur 2014). However, these fuel and microclimate property differentials between open- and closed ecosystems are muted under extreme fire weather and drought (Collins et al. 2019), increasing the likelihood of fire encroachment into rainforest (Marrinan et al. 2005; Clarke et al. 2014; Nolan et al. 2020). In many regions of Australia, especially the south-east, climate change is projected to increase the number of extreme fire days, lightning ignitions, and the severity of drought (Bradstock 2010; Clarke and Evans 2019; Canadell et al. 2021), potentially leading to increased incidence of fires encroaching into rainforest. Understanding the resilience of rainforest flora to projected regional increases in fire activity is therefore crucial to inform future rainforest conservation and management (Bradstock 2016).

Conversely, in many high rainfall regions of Australia, rainforest pioneers (rainforest taxa capable of colonising non-rainforest communities) are expanding into fire-prone open forests due to declining fire frequency (Gilbert 1959; Russell-Smith et al. 2004; Stanton et al. 2014; Tasker et al. 2017; Fletcher et al. 2021). Global change is predicted to favour further rainforest expansion in some regions, especially northern and north eastern Australia where fire danger conditions are forecast to decrease or remain stable (e.g. Lim and Roderick 2009; Clarke et al. 2011). Elevated atmospheric carbon dioxide (CO₂) is commonly invoked to explain rainforest expansion by favouring tree cover at the expense of flammable grasses (e.g. Bowman et al. 2010; Wigley et al. 2010). Other potential climate drivers include increased regional rainfall (Bowman et al. 2010), increased atmospheric humidity (Willett et al. 2007), and improved forest water-use efficiency in elevated CO₂ atmospheres (Keenan 2015). While there is increasing evidence that invading rainforest pioneers displace open forest flora and fauna (Chapman and Harrington 1997; Woinarski et al. 2004; Baker et al. 2020a, 2020b) and disrupt ecosystem function (Hoffmann et al. 2012a; Baker et al. 2021), the threshold at which these invading rainforest plants become resistant to removal by subsequent fires is not well understood (Baker et al. 2020a). Resprouting capacity in woody plants often increases from a low level at the seedling and sapling stage, and peaks at the young to mature stage, before declining during late maturity to senescence (Bond and Van Wilgen 1996; Vesk 2006; Clarke et al. 2013; Jaureguiberry et al. 2020), although there is variation between growth forms and ecosystem type (Clarke et al. 2013).

During the unprecedented drought conditions of 2019–2020, extensive areas of both rainforest and non-rainforest vegetation were burnt within the Gondwana Rainforests of Australia World Heritage Area, posing a potentially major threat to these globally significant ecosystems (Nolan et al. 2020). We took advantage of this historically unprecedented event to examine postfire resprouting and seed germination among woody rainforest taxa across four rainforest classes. Postfire resprouting and seedling germination of rainforest plants has been widely reported in Australian rainforests (e.g. Williams 2000; Campbell and Clarke 2006; Clarke et al. 2015), yet to our knowledge, no study has quantified the impacts of fire on the subtropical, dry or littoral rainforest classes of Keith (2004), limiting understanding of community scale responses to changing climate and fire regimes. Further, fire-response data is lacking for most rainforest flora species (Kenny et al. 2004; OEH 2014), hampering efforts to define appropriate domains of fire frequency for open forests with rainforest plants in the understorey. Understanding postfire recovery traits of rainforest flora is crucial for predicting shifts in rainforest distribution and composition in response to changing climate and fire regimes. Motivated by the need to better understand the resilience of rainforests to fire, the aims of the present study were to (1) quantify the mortality, resprouting and seedling recruitment of woody rainforest taxa following wildfire in subtropical, temperate, dry and littoral rainforests on the North Coast of New South Wales; (2) compare prefire and postfire stem density and species richness; and (3) examine the effect of prefire stem diameter on postfire resprouting rates.

The study was undertaken in and adjacent to reserves of the Gondwana Rainforests of Australia World Heritage Area in the North Coast Bioregion of NSW. To examine patterns of postfire resprouting and seedling recruitment, 17 sites were surveyed across four rainforest classes (Keith 2004; Supplementary Table S1) burnt during the extensive 2019–2020 bush fires of eastern Australia (Fig. 1), including subtropical (6 sites), warm temperate (7), dry (1) and littoral (3) rainforest. Burnt cool temperate rainforest was not sampled due to prohibitively difficult access. Fire severity at the sites was generally low to moderate, resulting in substantial scorching of the midstorey and subcanopy, but with variable degrees of canopy scorch (0–90%). Near-complete canopy scorch occurred on one site each of subtropical and littoral rainforest. Sites were located in burnt areas of mature rainforest (lacking Eucalypts or sclerophylls; Lynch and Neldner 2000) where possible, although sclerophyll species occurred as scattered individuals on many sites, usually as emergent trees above a well-developed, closed rainforest canopy. Five sites were located in burnt rainforest ecotones with a canopy of sclerophyll trees (>30% crown cover) above a well-developed rainforest subcanopy, where fire had self-extinguished at the rainforest boundary without encroaching stands of mature rainforest. The minimum distance between any two plots was 200 m (mean = 4889 m, SE = 1727 m).

Fig. 1.  Study area and sampling locations in relation to the Gondwana Rainforests of Australia World Heritage Area in the North Coast Bioregion, NSW, Australia. NP – National Park, SCA – State Conservation Area, NR – Nature Reserve (Source of wildfire extent: DPIE 2020).

Sampling was undertaken over a 6-month period from June to November 2020, 6–12 months after fire, thus corresponding to the immediate postfire period (≤ 1 year; Clarke et al. 2015). Data for subtropical, warm temperate and dry rainforest was derived from sampling by the NSW National Parks & Wildlife Service (NPWS; Nicholson et al. 2020). At each site, vegetation was sampled once in ∼20 m × 20 m quadrats, with plot dimensions modified as necessary to maximise sampling of narrow ecotones or areas of consistent burn severity (mean plot area = 421 ± 16.3 m²). Within each plot, all woody stems (rainforest, sclerophyll, trees, shrubs, vines) were identified to species and examined for evidence of postfire resprouting, before assignment to one of three postfire response classes: fire killed (no postfire resprouting), live resprout (generation of new shoots from dormant buds after 100% crown scorch), and alive escape (crown survived with only minor scorching). For living stems, species identification was based on examination of live foliage and stem features. In the case of dead stems, most individuals retained dead leaves at the time of sampling, and these were used for identification to species. Identification of dead stems without foliage was frequently possible by examination of bark features, branching architecture and comparison to adjacent stems with live foliage. Stems lacking sufficient features to allow confident identification were assigned as unknown and excluded from analyses.

Resprouting location was identified as aerial (apical, stem), basal (arising from stem at ground level) or underground (lignotubers, rhizomes, root suckers). Root suckers were differentiated from seedlings by the general absence of cotyledons, juvenile leaf form and visible rootlets, and excavation of surrounding soil to confirm connection to a pre-existing root system. Where possible, root suckers were assigned to identified pre-existing stems, by identification of the proximal roots (pointing toward the parent plant), which are typically thinner than distal roots and allow determination of the direction of the parent plant. Prefire plant size was quantified by recording the height and diameter at breast height (DBH, 1.3 m above ground) of original stems (dead, alive). For finer stems burnt below 1.3 m, diameter was measured at the highest available location.

In addition to resprouting response, identification of fire-cued seedling recruitment (sensu Clarke et al. 2015; germination in the immediate postfire period; ≤ 1 year) was also assessed at each site. Seedlings were differentiated from root suckers by the presence of cotyledons, juvenile leaf form and visible rootlets, and excavation of surrounding soil to confirm a lack of connection to a pre-existing root system. For groups of plantlets of uniform size, occasional seedlings (∼1 in 10) were pulled out to confirm seedling status. All seedlings were identified to species and tallied. In rare cases where plants could not be assigned confidently to suckers or seedlings, they were excluded from further analysis.

All identified stems were classified into functional groups based on descriptions of the Royal Botanical Gardens Trust (RBGT 2021): rainforest (rainforest or its margins), sclerophyll (predominantly open forest or woodland). Rainforest taxa were also classified as early successional (pioneer, early secondary; Kooyman 1996) or late successional (later secondary, mature phase; Kooyman 1996). All analyses were limited to native woody rainforest taxa, except for a comparison of mean seedling densities between rainforest and sclerophyll taxa.

To examine fire-resilience traits, taxa with at least five individual records of 100% crown scorch were classified to postfire resprouting and seedling response classes according to Table 1. Resprouting and seedling traits were also combined for each species to allow assignment to the plant–fire functional groups (Table 1).

Table 1.  Postfire response classes and plant–fire functional groups to which taxa were assigned in this study in Gondwana Rainforests of Australia on the NSW North Coast (adapted from Gill and Bradstock 1992; Clarke et al. 2015).

The prepared data were analysed in terms of the proportional abundance of resprouting and seeding responses of woody rainforest taxa across rainforest classes. We used the Kruskal–Wallis rank sum test and Conover–Iman post hoc tests with Bonferroni correction (P ≤ 0.05) to determine whether there were differences among fire-resilience traits between rainforest type and rainforest successional stage. Statistical analyses were processed using R software (v.3.4.2.; R Foundation for Statistical Computing, Vienna, Austria) and the package PMCMR (Pohlert 2018). Change in overall abundance of taxa across all sites was examined by comparing pre- and postfire counts of taxa with ≥ 10 prefire records.

To explore the effect of fire on the community structure and composition, we compared counts of prefire individuals (fire killed + live resprout + alive escape) and postfire individuals (live resprout + root suckers + alive escape + seedlings) to calculate changes in plant density, species richness and size class distribution. Root suckers were counted as individuals in the postfire individual total, as root suckers are a common means of postfire vegetative reproduction among rainforest taxa (Hoffmann 1998). To account for variation in plot area, we standardised data as follows: stem density (stems m⁻²) and species richness (number of taxa per plot, 421 m²). Pairwise differences between prefire and postfire stem density and species richness were determined using Conover–Iman post hoc tests with Bonferroni correction (P ≤ 0.05). Kruskal–Wallis rank sum tests were performed on prefire and postfire stem class distribution using the ‘kruskal.test’ function in R. To examine pathways of species turnover (persistence, immigration, extinction), all fire responses observed for each species at a site were grouped into combined classes (e.g. resprouting + seed germination), before calculating the proportion of individuals in each combined class across all species at a site prefire. Taxa with both postfire seedlings and prefire standing plants as potential parents were classed ‘local seed’, while those with postfire seedlings but no prefire plants were classed ‘new seed’.

To examine the influence of plant size on resprouting capacity, stem diameter and height data across all woody rainforest plants were grouped into six DBH classes (0–1 cm, 1–2 cm, 2–3 cm, 3–5 cm, 5–10 cm, and >10 cm) and five height classes (0–1 m, 1–2 m, 2–5 m, 5–10 m and >10 m) respectively for analysis. The largest DBH class (> 10 cm) was not further subdivided as > 95% of stems in this class were <50 cm DBH. To determine proportion of taxa capable of resprouting with a DBH of ≤ 1 cm, taxa with at least five individual records of 100% crown scorch and resprouting rates ≥ 30% were classified as capable of resprouting in this size class.

Conover–Iman post hoc tests with Bonferroni correction (P ≤ 0.05) were used to determine whether there were differences between classes of stem diameter and height. All means are reported as mean ± s.e. (s.e.m.).

We recorded a total of 4 254 fire-affected standing plants and 12 719 seedlings from 228 taxa of woody rainforest plants. Growth forms included trees/shrubs (∼84%) and vines (∼16%). Approximately 9% of all fire-affected standing plants could not be identified. Overall response of standing woody rainforest plants to wildfire included individuals that were either killed by fire (25.7 ± 2.7%), resprouted after 100% canopy scorch (53.7 ± 4.4%) or survived with their crowns largely unscorched (20.7 ± 3.0%). Survival as live crowns varied by stem size (DBH), including >10 cm (63.5 ± 5.8%; H₁ = 3.53, P = 0.06), 2–10 cm (27.4 ± 3.96%; H₁ = 20.83, P < 0.001) and <2 cm (11.9 ± 3.08%; H₁ = 24.43, P < 0.001). Of all individuals subject to complete canopy scorch, 66.4 ± 3.7% survived by resprouting (Table 2). Resprouting rates were not significantly different between rainforest classes (P > 0.05; Table S3). The most common position of resprouting was basal (52.3 ± 9.1%), followed by underground (34.8% ± 7.7%) and aerial stems (12.9 ± 2.2%). Postfire seedling germination of woody rainforest plants was observed at a mean density of 1.8 ± 0.4 seedlings m⁻². Seedling densities were not significantly different between rainforest types (P > 0.05). Mean seedling density of sclerophyll taxa (e.g. Lophostemon, Allocasuarina, Eucalyptus) was 0.06 m⁻², of which 56% were L. confertus, a common dominant of both sclerophyll and rainforest communities in the study region. Although excluded from the analysis, postfire weeds were absent from all but three sites with very low weed stem abundance of 0.2–1.7%.

Table 2.  Mean values (± s.e.) for woody rainforest plant response to wildfire (resprouting, seedling germination) in Gondwana Rainforests of Australia on the NSW North Coast.

Of the 228 woody rainforest taxa assessed, 213 (93.4%) displayed a capacity to persist in situ through wildfire by either resprouting (188 taxa) or seed germination (118 taxa); 85 (37.2%) both resprouted and germinated from seed, while seven taxa (3.1%) showed no evidence of persistence mechanisms (Table S2). Among the 126 taxa of plants with ≥ 5 records of 100% crown scorch, resprouters (R+) comprised the highest proportion of attributed taxa (63%), intermediate resprouters (Ri) comprised 22% and non-resprouters (R−) comprised 15% (Fig. 2). Of these taxa, 62% persisted in situ through seed germination, with or without resprouting (seeders, S+).

Fig. 2.  Proportion of woody rainforest plants by rainforest type across postfire resprouting and seedling recruitment traits, shown as stand-alone traits (a) and response combinations (b) in fire-affected Gondwana Rainforests of Australia on the NSW North Coast. Fac Res, facultative resprouter; Obl Res, obligate resprouter; Obl see, obligate seeder; * denotes marginal class due to intermediate resprouting capacity (31–69%); DRF, dry rainforest; LRF, littoral rainforest; STRF, subtropical rainforest; WTRF, warm temperate rainforest. Shown are mean (upper limit of bar) and standard deviation (error bars).

Across all sites and rainforest types, facultative resprouters (R+ S+) comprised the highest proportion of taxa (43%); obligate resprouters (R+ S−) comprised 21%; marginal facultative resprouters (Ri S+) comprised 11%; marginal obligate resprouters (Ri S−) comprised 11%; obligate seeders (R− S+) comprised 8%; and fire avoiders (R− S−) comprised 6% (Table 2, Fig. 2). Among taxa attributed as early successional, 83% were seeders (S+), 72% resprouted via suckering; 67% were resprouters (R+), 50% were facultative resprouters (R+ S+), and no taxa were fire avoiders (R− S−). Among late-successional taxa, resprouters comprised 61%, 59% resprouted via suckering, seeders comprised 48%, facultative resprouters comprised 32%, and fire avoiders comprised 11%.

Eight woody rainforest taxa listed as threatened species under the Commonwealth Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act) and/or the NSW Biodiversity Conservation Act 2016 (BC Act) were found to have postfire regeneration mechanisms (Table 3). Among these, seven taxa had the ability to resprout after 100% crown scorch by fire (≥ 1 record), three taxa showed postfire germination and one taxon increased stem density by root suckers.

Table 3.  Fire response of rainforest taxa listed as threatened under NSW and Commonwealth biodiversity conservation legislation in Gondwana Rainforests of Australia on the NSW North Coast.

The number of standing plant crowns declined by 79.3 ± 3.0% immediately postfire, being either returned to ground level by complete canopy scorch or killed. Including postfire seedlings and resprouts, the density of living woody rainforest plants was significantly higher (∼four times higher, P < 0.001) after wildfire (2.47 ± 0.36 plants m⁻²) than before wildfire (0.61 ± 0.05) (Table 2, Fig. 3). Postfire stem densities comprised ∼74% of prefire standing plants that remained alive after the fire (0.45 ± 0.04 plants m⁻²) plus the postfire emergence of additional seedlings (1.79 ± 0.35 m⁻²) and root suckers (0.23 ± 0.05 m⁻²). Postfire stem densities increased significantly in the 0–2 cm DBH class (H₁ = 23.75, P < 0.001) and decreased in the 2–10 cm class (H₁ = 22.45, P < 0.001), but did not change significantly in the >10 cm class (H₁ = 1.334, P = 0.23) (Fig. 4). Species richness of woody rainforest plants was also significantly higher (21.8% higher, P = 0.006) after wildfire (49.44 ± 2.12) than before wildfire (40.59 ± 1.75) (Fig. 3). Species turnover following wildfire comprised ∼5% of taxa becoming locally extinct from plots, ∼95% of taxa persisting through resprouting, germination from local seed or remaining unscorched, and the addition of new taxa to standing plant populations through postfire seedling germination (∼11 new taxa per plot) increasing species richness by ∼28% (Fig. 5). Seedling density and resprouting rates were not significantly affected by fire severity (P > 0.05; Table S3).

Fig. 3.  Plant density (a) and number species per plot (b) of woody rainforest plants before and after the 2019–2020 wildfires in Gondwana Rainforests of Australia on the NSW North Coast. Shown are median (horizontal line within box), mean (filled circle within box), 25th- and 75th-percentile (box boundaries), highest and lowest values (whiskers), outliers (filled circle outside box), individual site data (hollow circles). Lowercase letters indicate significance between classes using Conover–Iman post hoc tests with Bonferroni correction (P ≤ 0.05).

Fig. 4.  Size class distribution of woody rainforest plants before (a) and after (b) the 2019–2020 wildfires in Gondwana Rainforests of Australia on the NSW North Coast. Shown are mean (upper limit of bars) and 95% confidence interval (error bars). DBH, diameter at breast height.

Fig. 5.  Pathways of species turnover for woody rainforest taxa after wildfire in the Gondwana Rainforests of Australia on the NSW North Coast, including the proportion of prefire standing taxa that persisted on site through various mechanisms (blue) or became locally extinct (white), plus new taxa introduced to aboveground assemblages through germination of seedlings (yellow). Shown are mean (upper limit of bars) and 95% confidence interval (error bars).

Moderate to high proportions of rainforest plants resprouted across all classes of stem height (∼64–100%) and stem diameter (∼66–100%) (Table 2, Fig. 6). There were no significant differences in resprouting response between classes of stem height or stem diameter.

Fig. 6.  Proportion of resprouting stems across different classes of (a) stem height and (b) stem diameter (DBH) in fire-affected Gondwana Rainforests of Australia on the NSW North Coast. Shown are median (horizontal line within box), mean (filled circle within box), 25th- and 75th-percentile (box boundaries), highest and lowest values (whiskers), outliers (filled circles outside box), individual site data (hollow circles). Lowercase letters indicate significance between classes using Conover–Iman post hoc tests with Bonferroni correction (P ≤ 0.05).

Although numerous studies have documented postfire resprouting and seeding in Australian rainforests (e.g. Gilbert 1959; Floyd 1990; Williams 2000; Marrinan et al. 2005; Campbell and Clarke 2006; Williams et al. 2012), none have examined fire effects in subtropical, dry or littoral rainforests. By including previously unstudied rainforest classes in our study, our findings extend this literature to show that subtropical, dry and littoral rainforests have high fire resilience – comparable to that reported for other rainforest classes. Together with earlier studies, our findings suggest that high fire resilience, at least to infrequent fires, is a universal property of Australian rainforests.

In the short-term, single wildfires greatly reduced live above-ground biomass (∼80% of stems), resulting in profound structural changes in the rainforest communities examined. Despite this, most larger rainforest trees (> 10 cm DBH) survived beyond the scorch height of the fires and high proportions of woody rainforest stems and taxa have persisted through a pulse of resprouting and seed germination, resulting in a surprising net increase of stem density and species richness in the first year after fire. While the long-term trajectory of these forests remains unclear, the high density and richness of persisting woody rainforest taxa indicate a high potential for structural and compositional recovery in the longer term.

Previous studies show rainforest structure can regenerate quickly in the initial years after fire. Tropical Amazonian rainforest can develop a dense 4 m high canopy from basal shoots within 20 months after fire (Kauffman 1991). In Queensland wet sclerophyll forests, basal shoots of resprouting rainforest trees returned to prefire height (4–8 m) between 2 and 3 years after fire and postfire rainforest seedlings reached 8–12 m within 13 years (Williams et al. 2012). Further, in Eucalypt forest ∼15 km from the littoral rainforest sites in this study, postfire resprouts and seedlings of the rainforest trees Glochidion ferdinandi and Alphitonia excelsa formed a 12 m tall, closed canopy 16 years after severe wildfire (Baker et al. 2020a). Further insights into rainforest canopy recovery can be gained from rainforest plantings for ecological restoration. Canopy closure is routinely achieved within 3–5 years (Catterall and Harrison 2006; Freebody 2007) using industry standard planting densities of 0.31 plants m⁻² (1.8 m spacings) (Freebody 2007; Parkes et al. 2012) – less than half the postfire density of resprouting woody rainforest plants and remote suckers observed in this study (0.68 plants m⁻²). Furthermore, canopy regeneration via resprouting is faster than from seedlings due to well-established root systems and starch reserves (Kauffman 1991; Bond and Midgley 2001), allowing resprouting saplings to rapidly refill their own gaps and those left by adjacent non-resprouters (Marrinan et al. 2005). Including seedlings, total postfire stem densities observed in the present study (2.47 plants m⁻²) are eight times higher than standard rainforest restoration planting densities.

The high postfire seedling densities we observed (∼1.8 seedlings m⁻²) are consistent with previous studies. Hill and Read (1984) found seedling densities of several rainforest taxa were higher in recently burnt than in unburnt wet sclerophyll forests. Williams et al. (2012) found rainforest plant recruitment in rainforest ecotones was most abundant (∼2.3 seedlings m⁻²) in the first year after fire. The long-term survival rate of seedlings recorded in this study is unknown, however studies of rainforest seedling recruitment in warm temperate rainforests and wet sclerophyll forests in northern NSW (Campbell et al. 2012, 2016) suggest seedling survival rates may be lower in the more open conditions found after fire, due to postfire herbivory and increased moisture stress. In wet sclerophyll forest in far north Queensland, Williams et al. (2012) found high postfire densities of Alphitonia petriei seedlings had declined by 97% within 13 years after fire, although they still maintained a subcanopy above 50% cover that was sufficient to shade out the grassy understorey. However, dendrochronological analysis in Victorian Mountain Ash forest (Simkin and Baker 2008) revealed that a rainforest subcanopy of Nothofagus cunninghamii resulted from widespread germination in the wake of the major 1939 ‘Black Friday’ fires, further demonstrating the capacity for postfire rainforest seedlings to make substantial long-term contributions to forest structure. Irrespective of long-term seedling survival rates on our study sites, the density of resprouting individuals and suckers (0.68 plants m⁻²) exceeds prefire stem density (0.61 m⁻²) and is therefore likely to enable canopy closure within ∼5 years after fire (Kauffman 1991; Freebody 2007; Williams et al. 2012).

The increased postfire species richness, comprising high proportions of late-successional taxa, provides a strong basis for the long-term maintenance of floristic diversity in the rainforests examined. Assessment of species turnover pathways identified very high rates of taxa persistence (∼95%) through resprouting, local seed germination and/or escaping unscorched, while postfire seedling germination also added new taxa to standing plant populations, increasing species richness by ∼28%. Additionally, species richness is likely to increase over time as recovering structural complexity facilitates subsequent establishment of additional late-successional taxa. For example, Williams et al.(2012) reported a second suite of rainforest taxa that began to recruit additional seedlings around 8 years after fire, presumably as leaf litter and canopy closure increased. Although our results show a strong overall capacity for recovery of rainforest structure and diversity generally, differential postfire abundance among taxa will likely lead to compositional shifts, at least in the short term. The high rates of postfire resprouting and seeding recruitment among late-successional taxa contradicts the definition of these taxa as being restricted to ‘undisturbed rainforest’ (Kooyman 1996).

Very few differences were found in postfire response between the rainforest classes examined. A notable exception was the unique absence of fire avoiders from littoral rainforest. One possible explanation is that, unlike other classes, littoral rainforest plots were located in fire-prone Eucalypt forest with a closed rainforest subcanopy, where the periodic fires inherent to these forests may select for a higher proportion of fire-resilient rainforest taxa. Fire avoiders occurred in all other classes, albeit in low proportions (5–13%). Fire avoiders can neither survive fire by resprouting nor recruit by seed immediately postfire, so rely on seed dispersal from unburnt populations to re-colonise postfire (Clarke et al. 2015). Among the 11 fire avoiders identified in this study, eight were still capable of resprouting at low rates (16–29%), while nine were capable of long-distance seed dispersal by birds or wind, as indicated by seed morphology (RBGT 2021). Fires in the study locations were typically patchy within sites (as indicated by high proportions of unscorched crowns) and across the broader landscape (DPIE 2020), increasing the likelihood of fire-avoiding taxa persisting in unburnt refuges in the landscape and recolonising by bird- or wind-dispersed seed over time.

We found no evidence of a shift in floristic composition towards a more-flammable sclerophyll forest (e.g. Barker 1994; Tolsma et al. 2019). Firstly, scattered individuals of sclerophyll taxa (e.g. Allocasuarina, Corymbia, Eucalyptus, Lophostemon and Xanthorrhoea) already existed on or immediately adjacent to 14 of the 17 sites prefire, with the remaining three devoid of sclerophyll taxa postfire. Secondly, postfire sclerophyll seedling densities were extremely low, at around 1% of the density of rainforest plants, including on sites where Eucalypts dominated the canopy. Finally, regeneration of flammable ground-layer plants (e.g. graminoids) was negligible in all plots, and any minor regeneration is likely to be quickly eliminated as regenerating rainforest plants achieve canopy closure (Williams et al. 2012; Hoffmann et al. 2012a; Baker et al. 2021).

Beyond postfire recovery of rainforest, the fire resilience of rainforest plants also affects the persistence of invading rainforest pioneers in long-unburnt open forests (e.g. dry-, wet- and swamp-sclerophyll forest) when fire eventually returns. Invading rainforest pioneers rapidly outcompete shade-intolerant understorey plant communities in long-unburnt open forest (Rose and Fairweather 1997; Woinarski et al. 2004; Lewis et al. 2012; Baker et al. 2020a), although some displaced plant taxa persist as soil-stored seed (Baker et al. 2020a), awaiting the return of fire to restore open conditions through removal of the dense rainforest midstorey (Campbell et al. 2012). However, fire-resilient rainforest plants may quickly reassert competitive pressure postfire (e.g. Williams et al. 2012), potentially limiting long-term re-establishment of typical sclerophyll understorey plant communities, fauna habitat (Laurance 1997; Tasker et al. 2017) and understorey fuels that underpin open-forest flammability (Hoffmann et al. 2012a; Baker et al. 2021). Our results demonstrate resprouting potential among a wide diversity of rainforest taxa known to colonise long-unburnt open forests adjacent to the major rainforest classes of eastern Australia.

Additionally, the high rates of postfire resprouting (∼60–80% individuals) we observed in small rainforest plants (1 cm DBH, 1 m tall) suggests invading rainforest pioneers may become resilient to removal by fire soon after recruiting into open forests. This is consistent with Williams et al. (2012) who found 37% of seedlings from four rainforest taxa resprouted basally when burnt only 2 years after germination, including seedlings of Polyscias elegans, a facultative resprouter identified in this study. We found resprouting in small plants (DBH ≤1 cm) occurred in 14 of 17 (83%) woody rainforest taxa for which we had sufficient data, indicating that rapid attainment of this trait is widespread among different taxa. However, in the context of rainforest invasion, significant biodiversity declines can be driven by only one or two fire-resilient species (Rose and Fairweather 1997; Baker et al. 2020a).

The high proportion of early-successional taxa (83%) capable of exploiting the postfire environment through seedling recruitment recorded in this study is also highly relevant to rainforest-invasion processes in open forests. While resprouting provides a mechanism for the postfire persistence of invading-rainforest pioneers, seedling recruitment provides a potential mechanism for rainforest pioneer populations to increase after fire. Many taxa with postfire seedling germination are recognised pioneers, adapted to establishing in open conditions, so seedling survival rates are expected to be relatively high. We observed Euroschinus falcatus densities to increase over 100-fold due to seedling recruitment on sites that appear to be rainforest-invaded dry sclerophyll forest, based on the canopy species and their habit. These seedlings were observed around 1 year postfire and showed no evidence of postfire herbivory or moisture stress. Although Williams et al. (2012) reported a very low (3%) survival rate for Alphitonia petriei seedlings 13 years after fire, this still resulted in around one tree every 10 m⁻², most of which were > 8 m tall and situated among other rainforest trees that also recruited postfire. More research is needed to determine long-term seedling survival rates in this context.

Historical rainforest fire intervals in the study area (∼50–1000 years) provide a baseline for understanding tolerable fire regimes, with charcoal evidence of pre-European wildfires in subtropical rainforest (including with Brush Box emergents) in Nightcap National Park every ∼300–1000 years (Turner 1984), and intervals <50 years reported for warm temperate rainforests (Floyd 1990). Despite encroachment of the 2019–2020 fires into rainforests, there are no records to our knowledge of recurrent wildfire affecting any stand of mature rainforest of the Gondwana WHA since European colonisation, and thus no evidence of a shift from multi-centennial to decadal fire frequency. Although the postfire-regeneration capacity of rainforest observed in this study is likely to confer ecosystem resilience to at least centennial-scale, and potentially longer decadal-scale fire frequencies, fire frequency and resilience under future climate change remains uncertain. While this and previous studies demonstrate the resilience of rainforest to occasional fires, it is generally understood that there is a threshold frequency which can prevent rainforest resprouts and seedlings from maturing, and ultimately restrict rainforests to landscape positions where the topography creates fire refugia (Bowman 2000; Fensham et al. 2003). These thresholds urgently require further study. Conversely, in open forests, low fire frequency can facilitate invasion and displacement by rainforest, and in high rainfall regions, ultimately restrict open forest to landscape positions where frequent landscape fires persist (Russell-Smith et al. 2004; Stanton et al. 2014). Thus, where rainforest sits within a matrix of open forest, active management of fire frequency, such as through the restoration of Aboriginal burning, is needed to maintain rainforest boundaries and ensure the long-term conservation of the values inherent to both rainforest and open-forest plant communities (Bowman 2000; Bradstock 2016). Rainforest areas need to be protected as much as possible from potential increases in fire frequency with climate change. Planned fires in open forest immediately adjacent to rainforest have been shown to mitigate the spread of unplanned fires into rainforest (King et al. 2008), and may be needed to prevent the increased fire encroachment into rainforest predicted with climate change (Bradstock 2016).

More specifically, our findings prompt review of existing fire interval guidelines for the maintenance of open-forest biodiversity near rainforest. Firstly, our findings demonstrate that rainforest pioneers which invade dry open forests can become resistant to removal by fire, posing a major threat to the postfire recovery of shade-intolerant plant communities and associated fauna habitat. Therefore, to maintain dry sclerophyll open-forest biodiversity, fires should be of sufficient frequency to prevent widespread establishment of fire-resistant rainforest plants in forests where they were historically absent or rare.

Secondly, recommended minimum fire intervals for shrubby wet sclerophyll forests are partly based on assumptions that most understorey rainforest taxa are killed by fire and that to maintain their populations, minimum fire intervals need to be long enough to allow recolonisation by seed, and for recruits to complete their primary juvenile period before replenishing seed banks (Noble and Slatyer 1980; Kenny et al. 2004). Our findings demonstrate that a high proportion of rainforest plants are not removed by fire, but persist by resprouting, and may attain this capacity at a relatively young age. Such taxa may therefore be capable of maintaining resilient populations under shorter maximum fire intervals than is typically recommended (e.g. 25 years, Kenny et al. 2004). This study has generated fire-response data for a substantial number of additional rainforest flora (∼147; NSW Flora Fire Response Database v 2.1; OEH 2014), and their integration into future plant–fire vital attribute analyses (Noble and Slatyer 1980; Kenny et al. 2004) would enable more accurate identification of tolerable fire intervals for open forests with understorey rainforest plants.

We are mindful that the use of postfire counts (fire killed + live resprout + alive escape) as a proxy measure for prefire stem density and species richness may underrepresent non-resprouting stems that were completely incinerated or disintegrated postfire. Related findings should therefore be interpreted with caution. However, we argue that any effects would be limited to very small size classes, given that the diameter of live stems consumed by flaming combustion is typically < 3 mm (Hines 2010; Cruz et al. 2013) and small fire-killed stems (< 1 cm DBH) in our plots typically remained standing and retained scorched foliage, suggesting low postfire disintegration rates. Further, we found that the percentage change in density and species richness were not significantly altered by fire severity, indicating that stem loss may not have strongly influenced the overall trends observed. In rainforests, the seedling cohort typically comprises the highest stem densities, but lowest species richness (e.g. Green et al. 2014), indicating larger potential uncertainty in our results for stem density than for species richness.

This study adds to the large body of field evidence from other rainforest classes across Australia, which together demonstrate the high resilience of rainforest to occasional fire (Bowman 2000). We show that high fire resilience extends to previously unstudied major rainforest classes (subtropical, dry, littoral) and late-successional rainforest taxa. Although we found no evidence to indicate long-term floristic or structural decline in rainforests exposed to a single fire, the frequency at which fires prevent maturation of rainforest resprouts and seedlings or increase community flammability is unknown. In regions where fire frequency is predicted to increase with climate change, preventing recurrent fire incursions into rainforest may require fuel management in adjacent open forests, including through the restoration of more frequent Aboriginal burning.

In relation to rainforest-invaded open forests, our findings suggest that most rainforest pioneers can become resistant to removal by occasional fire, and that postfire seedling recruitment provides a potential mechanism for population increases after occasional fires. In high rainfall regions where global change factors are predicted to favour rainforest expansion (e.g. increased rainfall, increased CO₂, reduced fire activity) our findings suggest that the maintenance of dry sclerophyll open forests within a matrix of rainforest may require relatively high fire frequencies.

Supplementary material is available online.

The data that support this study were obtained from the NSW National Parks & Wildlife Service by permission. Data will be shared upon reasonable request to the corresponding author with permission from NSW National Parks & Wildlife Service.

The authors declare no conflicts of interest.

This research was supported by an Australian Government Research Training Program (RTP) Scholarship. This research did not receive any additional specific funding.

AB conceived and designed the research, performed the experiments and data analysis and wrote the manuscript; CC and MW supervised the project and provided critical feedback on research design, analysis and manuscript.