Without management interventions, endemic wet‐sclerophyll forest is transitioning to rainforest in World Heritage listed K’gari (Fraser Island), Australia

Abstract Wet‐sclerophyll forests are unique ecosystems that can transition to dry‐sclerophyll forests or to rainforests. Understanding of the dynamics of these forests for conservation is limited. We evaluated the long‐term succession of wet‐sclerophyll forest on World Heritage listed K'gari (Fraser Island)—the world's largest sand island. We recorded the presence and growth of tree species in three 0.4 hectare plots that had been subjected to selective logging, fire, and cyclone disturbance over 65 years, from 1952 to 2017. Irrespective of disturbance regimes, which varied between plots, rainforest trees recruited at much faster rates than the dominant wet‐sclerophyll forest trees, narrowly endemic species Syncarpia hillii and more common Lophostemon confertus. Syncarpia hillii did not recruit at the plot with the least disturbance and recruited only in low numbers at plots with more prominent disturbance regimes in the ≥10 cm at breast height size. Lophostemon confertus recruited at all plots but in much lower numbers than rainforest trees. Only five L. confertus were detected in the smallest size class (<10 cm diameter) in the 2017 survey. Overall, we find evidence that more pronounced disturbance regimes than those that have occurred over the past 65 years may be required to conserve this wet‐sclerophyll forest, as without intervention, transition to rainforest is a likely trajectory. Fire and other management tools should therefore be explored, in collaboration with Indigenous landowners, to ensure conservation of this wet‐sclerophyll forest.

One such fire-dependent system is wet-sclerophyll forests that form an ecotone between dry-sclerophyll forest and rainforest (Peeters & Butler, 2014;, and the focus of our study. Australian wet-sclerophyll forests occur over a large latitudinal range in Australia and are typically dominated by canopy trees in the Eucalypteae tribe and wider Myrtaceae family (Ashton & Attiwill, 1994;Donders, Wagner, & Visscher, 2006) with varying understorey depending on climate and disturbance histories (Donders et al., 2006;Harrington & Sanderson, 1994;King, 1985).
With low fire frequencies, for example, once in 100-350 years, wet-sclerophyll Eucalyptus-dominated forest in southern Australia favors an understorey of rainforest species and is able to regenerate and maintain canopy tree composition and structure (Ashton & Attiwill, 1994;Attiwill, 1994b;. A minimum of 100 years between fires is proposed as the required interval to maintain wet-sclerophyll forests in Tasmania (Gilbert, 1959;Hickey, 1994). Without logging or fire, these forests can transition to rainforest (Hickey, 1994), but whether appropriate fire regimes should be implemented to maintain wet-sclerophyll forests remains debated. Fire is often negatively perceived by the public and some land managers, and with limited knowledge of forest trajectories, attempts to reintroduce fire regimes for conservation purposes can prove difficult . Finding appropriate management regimes is pertinent for numerous reasons including that the loss of wet-sclerophyll habitats may lead to the decline of endangered fauna such as glider possums (P. australis var. reginae), powerful owl (Ninox strenua), and northern bettong (Bettongia tropica; Peeters & Butler, 2014;. Motivated by the need to direct conservation efforts, we analyzed long-term experimental plots of wet-sclerophyll forest. Such long-term research sites are rare (Franklin, 1989), but when present, allow examining natural and anthropogenic disturbances over larger temporal scales (Pretzsch, 2010). We follow a successional sequence over 65 years in Syncarpia hillii-Lophostemon confertus dominated wet-sclerophyll forests on K'gari, Australia. This forest (Regional Ecosystem 12.2.4) is an "of concern" regional ecosystem with ~10,000 hectares remaining (Queensland Government, 2017).
Prior to European interference from the mid-19th century, traditional Butchulla landowners managed K'gari's ecosystems.
Fires were thought to be frequent (i.e., every two months) and of low intensity, which resulted in a comparatively open understorey (Fensham, 1997). Fire management changed with the takeover of K'gari by timber loggers and led to less frequent and higher intensity fires (Spencer & Baxter, 2006). Today, the larger fires (>100 ha) on K'gari are generally linked to increased fuel load rather than periods of low rainfall (Srivastava et al., 2013). Logging commenced on K'gari in 1863 with the extensive removal of Araucaria cunninghamii (hoop pine) and Agathis robusta (kauri pine). With the discovery of marine borer resistant Syncarpia hillii in the late 1870s, logging intensified with the timber used inter alia for the construction of the Suez Canal (Lennon, 2012). Logging continued until December 1991 when the island was recommended for UNESCO World Heritage status (Lennon, 2012).
Here, we used three 0.4-ha plots that have been studied from 1952 to 2017 to examine how wet-sclerophyll forest (with Syncarpia hillii and Lophostemon confertus as main overstorey species) changes in tree species composition after selective logging and intermittent disturbances by fire and cyclone. We hypothesize that the current no/low disturbance regime will lead to (a) the eventual decline of focus species Syncarpia hillii and Lophostemon confertus as competition by rainforest species increases; and (b) changes in tree species diversity, composition, and relative basal area.

| Research plots and methodology
We studied forest growth plots ("yield plots") that were established by the Queensland Forestry Department in 1952(Joint Conservation Group, 1991. Three plots in Syncarpia hillii-Lophostemon confertus (Satinay-Brushbox) forest were studied with each plot measuring 100 × 40 m (0.44 ha) and mapped with latitudes and longitudes (Table 1). Plots were divided into subplots of 25 × 10 m. Within these subplots, trees with a stem ≥10 cm in diameter at breast height (DBH) were identified to the species level, location within plots mapped, and DBH recorded. The Forestry Department documented plot histories including selective logging, silvicultural treatments, and natural disturbances that varied between plots (Table 1, Joint Conservation Group, 1991). Tree mortality and DBH measurements were made in 1952, 1954, 1958, 1967, 1968, 1975, 1983, and 1989 by the Forestry Department with records of tree recruitment (≥10 cm DBH). The 1989 survey only measured trees with DBH >20 cm, and we excluded this record from the analysis.
In late 2016-early 2017 (2017 in the following), we measured all trees ≥10 cm DBH in the three plots. Trees with DBH <10 cm had not been recorded previously, but were measured by establishing five 5-m-radius plots located randomly within each 0.44-ha plot to track the recruitment of tree species.

| History of the three studied plots
Disturbance regimes, logging frequency and intensity, and basal area differed between plots (Table 1). Fire affected all plots pre-1935, and plot 10 in 1952 reducing undergrowth. Cyclones affected plots 9 and 10 (severely damaged) in 1975 (Joint Conservation Group, 1991).
Only plot 9 was subjected to ringbarking with intentions to convert the plot to a hoop pine plantation but most trees survived. All plots were logged (estimated since 1915) prior to commencement of regular monitoring in 1952 with selective removal of old growth trees (Joint Conservation Group, 1991). Plot 10 was the heaviest logged of the three plots pre-and post-1952 (Table 1).

| Species diversity
Analyses were performed using the program R (R Core Team, 2017).

| Changes in species composition
Tree stem maps were generated for each plot as a graphical representation of tree location and DBH sizes (≥10 cm) relative to each plot at the start of the survey period (1952) and at the latest measurement (2017). Non-metric multidimensional scaling (nMDS) based on the Bray-Curtis similarity index was used to examine the pattern F I G U R E 1 Location of Fraser Island in Australia and study plots in plant species composition in 1975, 1983, and 2017 among plots with trees ≥10 cm DBH. These three years were chosen to examine the changes between plots after the cyclone in 1975 (Table 1). Data from the 48 subplots (16 (25 × 10 m) subplots in each 0.4-ha plot) were used for analysis due to a lack of adequate replication.
To visualize the composition of the understorey among plots, nMDS was performed for trees <10 cm DBH in 2017. This was conducted using functions metaMDS() and ordiellipse() in the vegan package (Oksanen et al., 2016) and plotted using ggplot2 (Wickham, 2009). This ordination method is preferred in community ecological studies to determine patterns in multivariate datasets (McCune & Grace, 2002). The following analyses were performed to determine compositional differences between plots. Homogeneity of variance between plots for trees ≥10 cm and <10 cm DBH was assessed using the betadisper() function in vegan, an analogue of Levene's test for homogeneity of variance (Anderson, Ellingsen, & McArdle, 2006). Differences in tree species composition ≥10 cm and <10 cm DBH among plots were determined by using permutational multivariate analysis of variance (PERMANOVA) with the adonis() function in vegan (Anderson, 2001). If adonis() returned significant results, a post hoc multilevel pairwise analysis was performed using the pairwise.

| Tree species composition changes in plots from 1952 to 2017
The trajectory of change in tree composition over 65 years was mapped with tree stems (≥10 cm DBH; Figure 2). The focus species S. hillii and L. confertus were dominant in all plots in 1952, with an increasing number of rainforest trees recruiting by 2017. In 1952, S. hillii was more abundant (20 trees) than L. confertus (3 trees) in plot 9 and S. hillii trees were on average larger (> 80 cm DBH) than in plots 10 and 11. While plot 10 had smaller individuals of S. hillii than plot 9, it had higher abundances of S. hillii (60 trees) and L. confertus

Plot
In 2017, all plots had significantly different species composition (

| Tree recruitment and mortality from 1952 to 2017 (≥10 cm DBH)
Tree mortality varied across plots. Plot 9 had the lowest mortality of the focus species over 65 years (Table 3). One S. hillii and one L. confertus was removed in plot 9 in logging operations in 1977 and 1975, respectively; the other seven trees presumably died of natural causes. Four out of the seven S. hillii trees that had died of presumed natural causes had DBH >100 cm (Table 4).
In plot 11, twelve S. hillii and eleven L. confertus trees were removed by logging in 1975 (Table 3). They ranged from 40 to>100 cm in DBH. Eleven L. confertus and 21 S. hillii trees were destroyed during logging. The majority of these trees were recruits in the 10-20 cm DBH class (Table 4).
From 1954 to 2017, plot 9 had no recruitment of S. hillii, but 36 and 53 S. hillii recruited in plots 10 and 11, respectively. The majority of recruitment occurred in the 34 years from 1983 to 2017 (Table 5).
Similarly, recruitment of L. confertus was low in plot 9 totaling three trees from 1958 to 2017. In plots 10 and 11, 32 and 48 L. confertus trees recruited, respectively, between 1954 and 2017, most of which were recorded in 2017 (Table 5). A larger number of rainforest tree species had recruited in plots 9 (65) and 10 (27) in 1983 compared to plot 11 (6). Recruitment of rainforest species had expanded in 2017 to a total number of 95, 94, and 155 trees of ≥10 cm DBH in plots 9, 10, and 11, respectively.

| Species diversity and basal area changes of trees ≥10 cm DBH
Tree species richness (≥10 cm DBH) remained relatively low from 1952 to 1975 ( Figure 4). Plot 10 contained five tree species during this period; plots 9 and 11 had three and two species, respectively. By 1983, plots 9 and 10 had increased in species richness and totaled 12 tree species, while plot 11 increased to five species. Surveyed in 2017, all plots featured 14 to 15 tree species (DBH ≥10 cm; Figure 4).
Tree species evenness had similar trends to Shannon's indices over the survey period. In 1952, plots 9 and 10 had an evenness index of 0.47 and 0.57, respectively, and plot 11 of 0.78 (Table 6).

| Survey and analysis of trees <10 cm DBH
In 2017, trees <10 cm were surveyed in five 78.5 m 2 subplots within each of the three 0.4-ha plots to gain a more complete picture of tree recruits. Across the 15 subplots, we measured 260 live trees (<10 cm DBH) with 27 species from 14 families (Table 7).
No S. hillii recruits were recorded at any plot, and only three and two L. confertus recruits were found in plots 10 and 11, respectively. The remaining recruits were species typically associated with rainforests, with all plots containing Backhousia myrtifolia, Syzygium oleosum, Cryptocarya macdonaldii, and Cryptocarya glaucescens. Between six and 11 species were common across plots.
Backhousia myrtifolia recruits were common in plots 9 and 11 with 31 and 63 individuals, respectively, but only five B. myrtifolia recruited in plot 10. Nineteen Syzygium oleosum recruits were recorded in plot 10, eight in plot 9, and five in plot 11. Schizomeria ovata recruits were present in plots 9 and 10 with 11 and eight individuals, respectively, but were absent from plot 11. All three plots contained  future projections remain unclear as rainforest tree species recruit at considerably faster rates than wet-sclerophyll species (Figure 8).

| D ISCUSS I ON
Throughout Australia, the deliberate use of fire by Indigenous landowners, and subsequent cessation with the displacement of the Indigenous population, has led to widespread change in landscapes and their ecosystems (Gammage, 2012). With the cessation of burning regimes by Butchulla landowners for over a century, and end to silvicultural disturbances since 65 years, it appears likely that this wet-sclerophyll forest transitions to rainforest. We discuss the implications, limitations, and opportunities resulting from this study.

| Mortality and recruitment of focus species Syncarpia hillii and Lophostemon confertus
Considerable post-logging mortality of trees (≥10 cm DBH) can occur in tropical and subtropical forests (Nebel, Kvist, Vanclay, & Vidaurre, 2001;Smith & Nichols, 2005). Post-logging mortality was minor in our plots with two individuals of the focus species and one E. pilularis dying as the result of the selective logging within eight years of logging (Joint Conservation Group, 1991). Reasons include that non-merchantable stems were not poisoned, a common silvicultural practice elsewhere (Smith & Nichols, 2005 (King, 1985). In our study, established L. confertus had a mean annual diameter increment of 0.2-0.4 cm. Growth rate depends on resource availability, competition, and facilitation, and is often not linear (Florence, 2004).
Recruiting focus species may have commenced their life in the past two decades (assuming 0.4 cm annual increment) or have been seedlings in 1952 (assuming slow initial growth). The least disturbed of the plots (plot 9), only recruited three individuals of L. confertus (≥ 10 cm DBH) and no S. hillii. Likely reasons include a lower abundance of smaller individuals of the focus species at the start of the survey (7 individuals between 10 and 40 cm vs. an average of 27 individuals in plots 10 and 11), lower logging intensity (~80 m 3 timber removed ha −1 vs. ~340 and 150 m 3 /ha removed from plots 10 and 11), and no fire for at least 80 years. Only three to five L. confertus recruited in the <10 cm DBH class, suggesting that more recent recruitment of the focus species may have been prevented by a lack of seeds and/or unfavorable conditions for seedling establishment and growth.

Syncarpia hillii Lophostemon confertus
Collectively, our observations present some evidence that focus species recruitment occurs proportional to disturbance regimes, with selective logging having enabled recruitment of the focus species, although further research is required to assess this notion. If confirmed that the logging regimes rather than fire enables regeneration of wet-sclerophyll species, it would contrast observations elsewhere where a lack of fire has prevented regeneration of wet-sclerophyll species (Ashton, 1981;Ashton & Attiwill, 1994;Harrington & Sanderson, 1994;. Silvicultural regeneration practices in Tasmanian wet-sclerophyll forests use a combination of clear felling, burning, and sowing of Eucalyptus seeds (Baker & Read, 2011;Hickey, 1994). This approach, practised since the 1960s, regenerates Eucalyptus species by removing competition from a dense understorey and creating an environment conducive for seedling establishment (Attiwill, 1994b;Baker & Read, 2011;Bassett, Edwards, & Plumpton, 2000). In the mid-north coast of New South Wales (Australia), wet-sclerophyll species regenerated irrespective of post-logging burning, but regeneration was significantly higher after a post-logging burn (King, 1985). Fire is a requirement for maintaining wet-sclerophyll forests across broad climate gradients in North Queensland and Tasmania (Hickey, 1994;, and the lack of post-logging burning at our study plots may have limited the recruitment of the focus species. High frequency burning every 2-4 years in wet-sclerophyll forests (dominated by Eucalyptus pilularis) in south-east Queensland increased the mortality of L. confertus and Syncarpia glomulifera, while a 30-year absence of fire enhanced their recruitment (Guinto, House, Xu, & Saffigna, 1999). This indicates that regenerative fire intervals ought to exceed decadal scales for wet-sclerophyll species to reach a size that enables survival.
As outlined above, we observed recruitment of S. hillii and L. confertus in the more intensively logged plots 65-80 years post-fire.
L. confertus is relatively shade tolerant (King, 1985), and Syncarpia glomulifera resprouts epicormically after high-intensity fires (Benson & McDougall, 1998). The focus species in our study are thought to regenerate naturally in a disturbed understorey with increased light availability, but knowledge on germination physiology is lacking (Fitzgerald, 1990). In general, wet-sclerophyll Eucalyptus species are unable to germinate and survive at undisturbed sites, especially where a substantial leaf litter layer covers the soil (Florence, 2004;Florence & Crocker, 1962). Improving knowledge on seed TA B L E 5 Recruitment of main tree species (≥10 cm diameter at breast height) in plots that were established in 1952 and monitored over 65 years

| Recruitment of rainforest tree species and fire
The fast recruitment of rainforest species relative to the focus species raises concern that rainforest may eventually replace the wetsclerophyll forest. While the composition of tree species remained similar prior to 1975, it diverged significantly following the heavy recruitment of rainforest species in more recent decades. This shift in species composition is supported by an increase in Shannon's and evenness indices, with higher diversity attributed to recruitment of rainforest trees. It is possible that recruitment of rainforest trees was triggered by a cyclone disturbance in 1975 that affected plots 9 and 10, with gaps in the canopy allowing rainforest species to establish within eight years. Only minor recruitment of rainforest species occurred in plot 11 during this period. Regardless whether the cyclone played a role in rainforest tree recruitment, the 2017 survey detected the highest rate of rainforest species recruitment (DBH <10 and ≥10 cm) in plots, indicating favorable growth conditions for rainforest trees. The two most common species were pioneer rainforest trees Schizomeria ovata and B. myrtifolia which occur in the upper strata of rainforests on K'gari (Lennon, 2012). While S. hillii remains the dominant species (113 m 2 BA ha −1 in total across plots in 2017), the gradual increase in the basal area of rainforest species from 20 m 2 /ha (1983) to 43.3 m 2 /ha (2017) indicates increased presence (and competition) from rainforest recruits in all plots.  Wet-sclerophyll forests in north Queensland experience similar encroachment by rainforest species with a narrow wet-sclerophyll ecotone between Eucalyptus-dominated forest and rainforest (Harrington & Sanderson, 1994;Russell-Smith & Stanton, 2002;Unwin, 1989). The required fire intensity to maintain wet-sclerophyll forests in north Queensland rarely occurs (Ash, 1988) as low to moderate fires extinguish upon reaching rainforest boundaries Unwin, Stocker, & Sanderson, 1985).
Rainforest species in north Queensland demonstrate fire resilience similar to most sclerophyllous species, and only frequent fire can inhibit their establishment (Williams, Parsons, Jensen, & Tran, 2012). The absence of fire over 20 years appears to provide an irreversible transition to rainforest at these sites (e.g., Mt Fox and Taravale; . In southern Queensland, however, high-intensity fire does not appear to be required for maintenance of wet-sclerophyll species; instead, irregular low-intensity fire or mechanical disturbances seem to suffice (Peeters & Butler, 2014

| Managing Syncarpia hillii-Lophostemon confertus wet-sclerophyll forest
Our study provides some insight into the fate of the two focus species over a 65-year survey period and indicates that both species are unlikely to sustain their presence over time in an unmanaged forest, that is, in the absence of fire or other interventions. Our findings are in line with the observed decline in overstorey Eucalyptus species in unburnt wet-sclerophyll forests (Ellis, 1985;Harvest, Davidson, & Close, 2008 (Harvest et al., 2008). In the same forest, removal of rainforest species (logging and burning) can reverse the deterioration of Eucalyptus species (Ellis, Mount, & Mattay, 1980).
Fire exclusion confers a competitive advantage to rainforest species as they develop quickly and subsequently reduces the likelihood of fire . It appears that the absence of fire promotes faster recruitment of rainforest species than of S. hillii and L. confertus, transitioning the wet-sclerophyll forest to a closedcanopy rainforest. While the pros and cons of rainforest expansion remains debated in Australia   (Srivastava et al., 2013). This amplifies the fire management issue as certain sclerophyllous species may not be able to recover after intense wildfires (Peeters & Butler, 2014). Evaluating the effects of fire regimes, including those of Indigenous owners, should be a priority.
It is more difficult to justify logging to allow recruitment of S. hillii and L. confertus. Logging intensities and the effect of logging as a surrogate for natural disturbance remain debated (Fitzgerald, 1991).
Whether selective logging could be a substitute for fire is unclear, while potential benefits include revenue to assist in managing the forests.
Selective logging as a management tool would need to have strong stakeholder agreement of Indigenous landowners, government regulators, and the public, as well as considering K'gari's World Heritage status and forest stewardship certification. Selective logging may be viewed as resurrecting K'gari's past logging regime and a step into the (Continues) TA B L E 7 Species list and abundance of trees <10 cm diameter at breast height in the three surveyed plots of Syncarpia hillii-Lophostemon confertus forest. Gray shading represents low to higher number of individuals past when the forests were exploited for their timber. Logging has not occurred on K'gari since the cessation of commercial logging 27 years ago. Selective logging is used to manage Tasmania's wet-sclerophyll forests together with other silvicultural practices (Baker & Read, 2011;Hickey, 1994).
Another factor to consider in the management of these forests is climate variability. Changes in temperature and rainfall may affect tree population densities and competitive abilities of species in the wet-sclerophyll ecosystem (Peeters & Butler, 2014). Such changes are difficult to predict and their influence on species is not well understood (Peeters & Butler, 2014), but maintaining wetsclerophyll forests remains a priority as it provides the maximum carbon and biodiversity benefits at sites suited for this vegetation.
Taken together, our study suggests that the survival of S. hillii is dependent on its regenerative capabilities following disturbance.

Its endemism to K'gari and adjacent mainland Great Sandy National
Park makes this species a conservation priority, and its loss would adversely affect the biodiversity of this ecosystem. K'gari's World Heritage status is based on the unique nature of the island with dynamic soils and vegetation. It is a tourist hotspot which makes it more difficult to implement fire regimes.
Several steps could be taken to better understand the dynamics of the forest ecosystem studied here. Additional plots in the S. hillii-L. confertus forests could be studied to evaluate whether the higher recruitment of rainforest than focus species occurs across the island. Long-term trial plots testing various fire regimes could be implemented to assess the most appropriate fire intensity and interval for management of these forests in collaboration with Indigenous landowners. A thorough assessment is warranted on K'gari and the adjacent mainland region to study this forest type.
A broader range of disturbance regimes and environmental conditions should be examined to disentangle the contribution of biotic and abiotic variables on the future of this wet-sclerophyll forest.

ACK N OWLED G M ENTS
We are grateful to the staff from the Tropical Forests and People Research Centre who assisted with re-measuring the experimental plots. We thank student volunteers Michael Purcival and Tahlia Kinrade for assistance with field work. We wish to thank B. Hogg (DAF Queensland) and D. Lee (USC and DAF Queensland) for their invaluable assistance in allowing access to the data files associated with plots 9, 10, and 11. We are grateful for help from Lui Weber and Bill MacDonald with tree species identification. We also recog- Raw data containing tree number, tree species and identity, survey year, diameter at breast height (≥ 10 cm and < 10 cm), coordinate within plots, and associated R codes used for analyses have been deposited to Dryad: https://doi.org/10.5061/dryad.