Fire in the bog: responses of peatland vegetation in the Australian Alps to fire

Keywords: bog, climate change, feral animals, fire, grazing, obligate seeder, peatland, resprouter, weeds.

There is scattered evidence of fire in alpine and sub-alpine treeless vegetation in Australia throughout the Holocene (Singh et al. 1981; Dodson et al. 1994; Kershaw et al. 2002). For instance, sub-alpine peatlands contain significant levels of charcoal throughout their profiles and preliminary work suggests that fire was prominent around 11–9000 years ago and has become even more frequent over the last four millennia (Theden-Ringl 2018; Hope et al. 2019). A study of charcoal preserved in Bogong Moth (Agrotis infusa) deposits in northern Kosciuszko National Park and the Australian Capital Territory revealed evidence of burning in the sub-alpine zone over the past millennium (Keaney 2016). Little is known about the use of fire by humans in the high country prior to European settlement (Banks 1989). There is evidence that an increase in fire frequency in the Australian Alps between 1000 and 1600 years ago was concurrent with a contraction of the alpine area and expansion of adjoining eucalypt woodland (Thomas et al. 2022). An increasing fire frequency is therefore likely to exacerbate the threat to alpine plant communities from expected climatic changes.

For most of the 20th century, landscape-scale fires were rare at high elevations because of the low incidence of extreme fire weather conditions, leading to the view that fire was unusual there (Zylstra 2006). Fires, however, have become increasingly common in south-eastern Australia during periods of extended hot, dry weather in the last four decades (Zylstra 2006; Bradstock 2008; Nolan et al. 2020; Canadell et al. 2021), resulting in large areas of alpine and sub-alpine vegetation being burnt, including many peatlands (Whinam et al. 2010; Tolsma 2020). Several landscape-scale fires have occurred in the Australian Alps since 1985, and fire modelling predictions indicate that fire frequency and severity are likely to increase in the coming century (Hennessy et al. 2005; Fox-Hughes et al. 2014). Given the likelihood of fire recurring in high mountain ecosystems, an understanding of the way in which vegetation there responds to fire is critical to its future management. This is particularly important for peatland ecosystems that play a role in high mountain catchment function (Cartwright and Morgenstern 2016), are a significant local store of carbon (Wilson et al. 2022), are known to be sensitive to disturbance (Clarke and Martin 1999; McDougall 2007) and are a feature of the alpine region of mainland south-east Australia (covering about 520 km²; Grover et al. 2005). Their distinctiveness, restricted geographic distribution, importance to ecosystem function and the threats they face are the reasons for their listing as endangered under the Commonwealth Environment Protection and Biodiversity Act 1999 (Department of Environment, Water, Heritage and the Arts 2009).

Despite a growing global recognition that fire will become more common with increasing drought and temperature in high- to mid-latitudes where most peatlands occur (e.g. Moritz et al. 2012), there appear to have been relatively few studies that document the effects of fire on peatland vegetation. Most Australian peatland species resprout after fire, re-emerging quickly (Walsh and McDougall 2004; Clarke et al. 2015). Similarly, in a New Zealand peatland, resprouters were found to dominate 2 years after fire, with floristic composition mostly recovering by 14 years post-fire (Wilson 2020). There, species with serotinous seed, or seed stored in soil seed banks that germinated in response to high temperature, may have been favoured by the fire. In another New Zealand peatland, rhizomatous species were found to have regained their pre-fire cover 50 months after fire (Norton and De Lange 2003). Regeneration of dominant Sphagnum L. mosses in peatlands lags behind that of resprouting species in both Australia and New Zealand (Timmins 1992; Johnson 2001; McDougall 2007). In addition to the impacts on vegetation structure and composition, fire can also disrupt hydroseral processes (Ashton and Hargreaves 1983; Campbell 1983), increase sedimentation and cause the loss of peat and nutrients, while hydraulic changes and lowered water tables can result from stream incision (Tallis 1983; Whinam and Chilcott 2002). Few studies have followed the effects of recurrent fires on peatlands, either through repeated measurements or chronosequences (but see Kirkpatrick and Dickinson 1984; Coates et al. 2006; Benscoter and Vitt 2008; Coates and Walsh 2010; Coates et al. 2012 for exceptions).

As the incidence of fire in high mountain peatlands has grown in recent decades so too has research into its effects on vegetation and peatland processes. Unsurprisingly, much of this research has sought to address local conservation imperatives, such as the development of revegetation techniques and the assessment of natural recovery of a threatened community. However, there has been no attempt yet to bring this disparate research together. In this paper, we use case studies of available literature and previously unpublished work to identify generalisations about fire effects on peatlands in the Australian Alps, and to inform future planning for their management. The four case studies cover much of the range of peatland vegetation in the Australian Alps (Fig. 1).

Fig. 1.  Localities of sample sites for the four case studies (CS) highlighted in the paper. NP, national park.

Peatland vegetation of the Australian Alps occurs mostly on high plateaux, between about 1100 and 2100 m above sea level, stretching from the Baw Baw plateau in southern Victoria to the mountains of Namadgi National Park in the Australian Capital Territory (ACT). The largest extent occurs within the Alpine National Park in Victoria and Kosciuszko National Park in New South Wales (NSW). McDougall and Walsh (2007) described three plant communities, characterised by hummocks of Sphagnum cristatum and deep peats; these are the focus of this paper. The floristic differences between the three communities are minor, relating to latitude and elevation (McDougall and Walsh 2007) and we do not differentiate between them in this paper. These communities have collectively been called bogs (e.g. in legislation), mossbeds and peatlands, the latter being the term we use to encompass all such vegetation in this paper. Peatland vegetation typically has a high cover of S. cristatum, through which rhizomatous graminoids (monocots not in the Poaceae family, such as Empodisma minus (Hook.f.) L.A.S.Johnson & D.F.Cutler, Carex gaudichaudiana Kunth, Carpha spp. and Oreobolous distichus F.Muell.), and the shrub Dracophyllum continentis (formerly Richea continentis) protrude. Tall shrubs such as Baeckea gunniana Schauer and Epacris paludosa R.Br. are scattered within peatlands at high elevations but may dominate at lower elevations with other shrub species such as Callistemon pityoides. The low cover of Sphagnum in some peatlands may be a function of past grazing disturbance as this moss is known to be very sensitive to trampling (Costin et al. 1959; Whinam et al. 2003). Frequently-occurring species (B. gunniana, C. gaudichaudiana, D. continentis, E. minus, E. paludosa and S. cristatum) are abbreviated to their genus name in the text.

A grazing exclosure was established in 1946 on a tributary of Rocky Valley Creek, east of Falls Creek in Victoria, to investigate the effects of excluding cattle grazing on a range of sub-alpine plant communities. The fenced and adjoining unfenced areas had been burnt in 1939 but the extent and severity of burning at that time were not documented. While data were collected on dry shrub- and grass-dominated vegetation, peatland vegetation proved difficult to monitor (Carr and Turner 1959). However, by 1977 the ungrazed peatland was observed to have expanded greatly whereas no change was apparent in the grazed peatland (Carr 1977).

In 1979, monitoring grids were established in both peatlands using a theodolite and corners were permanently marked with steel posts at 20 m intervals; there were 39 20 × 20 m grid squares in the ungrazed peatland and 28 grid squares in the grazed peatland (McDougall 1989). Each grid square was subdivided into 25 4 × 4 m plots using nylon tapes. In each plot, Sphagnum cover and the position of pools and drainage features were mapped onto scaled graph paper. Five contiguous 20 × 20 m grid squares in the centre of each peatland were re-sampled in 1999, 2004, 2006, 2009 and 2018. In alternate plots, all species with a cover of 5% or more were recorded and assigned a Braun-Blanquet cover value (giving a total of 13 plots/grid square), which was subsequently converted to the following scale: 1 = 0 to <5% (for species absent or with <5% cover in a plot at one or more sample times but with >5% cover at other times), 2 = 5–25%; 3 = 25–50%; 4 = >50%. In 2003, wildfire burnt most of the re-sampled grid squares in both peatlands. Grazing by cattle ceased after the fire. The significance of differences in the cover of key species between years of measurements was assessed using Wilcoxon rank sum tests.

The covers of all frequently-occurring species except Carex were significantly lower in 2004 and 2006 than pre-fire, regardless of grazing status, but most species had returned to pre-fire cover by 2018 (Table 1). However, the changes were not the same in the two peatlands, the most notable difference being in Sphagnum cover. In general, the ungrazed peatland became shrubbier with much less Sphagnum, while the grazed peatland became less shrubby with more Sphagnum and Carex cover. The cover of Carex was unchanged in the ungrazed peatland across all measurements before and after the fire; in the grazed peatland it was unchanged until 2009, after which it was significantly higher than in 1999 (Table 1). The post-fire increase in Carex cover in the grazed peatland may have resulted from release from grazing pressure as the species is known to be highly palatable to cattle (van Rees 1984). Baeckea cover was lower post-fire until 2009 in both peatlands. Empodisma attained its pre-fire cover by 2009 in the ungrazed peatland but then declined again when measured in 2018; in the grazed peatland its cover had returned to pre-fire level by 2018. Epacris returned to its pre-fire cover in the grazed peatland by 2018 and in the ungrazed peatland by 2009; however, in the ungrazed peatland its cover was significantly higher in 2018 than in 1999. Dracophyllum cover had not returned to its pre-fire cover by 2018 in the grazed peatland, while in the ungrazed peatland, its cover, though reduced post-fire until 2004, was not significantly different from its pre-fire cover by 2009. Sphagnum returned to pre-fire cover in the grazed peatland by 2009 but had not reached its pre-fire cover in the ungrazed peatland by 2018. Indeed, Sphagnum had significantly less cover in 2018 than in 2009 in the ungrazed peatland (P < 0.01); a decrease in cover was recorded in that period in 22 of 48 plots and no increase in any plots. The large but non-significant changes for some comparisons in cover between years for Carex and Dracophyllum may be a consequence of small sample size and the coarseness of the Wilcoxon rank sum test for small sample sizes.

Table 1.  Net percentage of plots that changed in cover between 1999 and subsequent years of measurement for frequently-occurring species in grazed and ungrazed peatlands in Rocky Valley (i.e. the number of increases minus the number of decreases divided by total number of comparisons, as a %).

Given the keystone role of Sphagnum in peatlands, the contrast in its response to fire in the two adjoining peatlands is noteworthy. The difference may relate partly to the nature of the fire and local hydrology post-fire, rather than grazing history. In the ungrazed peatland, the fire burnt in patches, leaving unburnt areas fragmented. Pools and rivulets shifted to new parts of the peatland (Fig. 2), a trend that was already evident in 1999 (McDougall 2007). By 2018, Sphagnum had also disappeared from unburnt parts of the peatland. In contrast, the fire in the grazed peatland left a connected, central area of Sphagnum unburnt, which may have aided colonisation of burnt areas. In addition, the grazed peatland became much wetter after the fire, judging from the large increase in the cover of pools and rivulets in 2018 (Fig. 3), and new areas of Sphagnum established while burnt areas recovered.

Fig. 2.  Map of the ungrazed peatland in Rocky Valley near Falls Creek, Victoria showing Sphagnum, water (pools and rivulets), areas with both and areas with none in 1979, 1999 and 2018. The mapped area is 2000 m².

Fig. 3.  Map of the grazed peatland in Rocky Valley near Falls Creek, Victoria showing Sphagnum, water (pools and rivulets), areas with both and areas with none in 1979, 1999 and 2018. The mapped area is 2000 m².

Although rarely-occurring species were not recorded prior to the fire, some changes were evident. For instance, two herbaceous species (Scaevola hookeri, Viola fuscoviolacea) were locally common after the fire but had not been observed before the fire (McDougall 2007). By 2018, both species had again become scarce: S. hookeri was present in 17 plots in 2009 but only three in 2018; V. fuscoviolacea was present in 24 plots in 2009 but only one in 2018. Four perennial non-native species (Juncus articulatus, J. effusus, J. tenuis, Salix cinerea) were recorded in the grazed peatland in 2006 but had not been recorded in 2004 (McDougall 2007) and were not thought to occur there prior to the fire. All but J. tenuis were recorded in 2018 despite efforts to remove these weeds using targeted herbicide application.

A floristic survey of alpine and sub-alpine treeless areas of Kosciuszko National Park (NSW) was conducted in late 2002 to classify the vegetation of the Australian Alps (McDougall and Walsh 2007). Many of the plots were burnt in January 2003 and some, especially in peatland vegetation, could be relocated and permanently marked for repeated measurements post-fire. Fourteen plots in peatlands were measured for floristic composition in 2002 (pre-fire), 2003, 2013 and late 2019 (almost 17 years post-fire); all were completely burnt in 2003. Here, we present new data collected in 2019, a few weeks before the fires of the 2019/20 summer.

At each measurement date, all vascular plant species in 5 × 5 m plots were recorded and given an estimate of cover on the following cover scale: 1 = < 5% (but present); 2 = 5–25%; 3 = 26–50%; 4 = 51–75%; 5 = 76–100%. Floristic differences between years of sampling for each plot were assessed by calculating Bray–Curtis similarity in Primer v6 (Clarke and Gorley 2006). No transformation was applied to the data. Similarity data were not normally distributed as determined by a Shapiro-Wilks test and transformations did not produce normality. Accordingly, a non-parametric Wilcoxon signed rank test with continuity correction in the base version of R (R Core Development Team 2019) was used to test the significance of differences in similarities between years and between burnt and unburnt areas. Diversity measures were calculated for each year of measurement in Primer v6 (Clarke and Gorley 2006). The significance in differences between years for the measures was assessed using one-way ANOVA. Wilcoxon signed rank tests were also used to assess the significance of differences in summed cover of life form/fire regeneration mode between all years of measurement. For this purpose, numeric cover values were randomly assigned within the range of the ordinal cover values. The randomisation was performed 10 times and the mean cover of fire regeneration modes in each plot then calculated. Fire regeneration mode was derived from Walsh and McDougall (2004) and McDougall et al. (2015).

Mean species richness (species/plot) and Hill-Shannon diversity were not significantly different between years (P > 0.05; Table 2). Evenness was significantly different between years (P < 0.001), peaking 1 year after the fire. This is reflective of the temporary reduction in cover of dominant species after the fire; in 2002, 68 ± 3% of species were given a cover value of 1, whereas in 2003 this had risen to 85 ± 4% of species.

Table 2.  Diversity metrics (mean ± s.e.) for 14 plots measured at four times in peatland vegetation in Kosciuszko National Park.

The Bray–Curtis similarity of floristic composition between 2002 and subsequent measurements was significantly higher for plots measured in 2013 (10 years post-fire, P < 0.05) than for plots measured in 2003 and 2019 (1 year and c. 17 years post-fire) (Fig. 4). That is, 10 years after the fire, the plots were closer to pre-fire floristic composition than 1 year and 17 years post-fire. For the latter case, this may have been due to the drought in the years prior to 2019.

Fig. 4.  Box plot of Bray–Curtis similarities between 14 peatland plots measured on four occasions and compared between: A = 2002 and 2003; B = 2002 and 2013; C = 2002 and 2019.

The covers of herbaceous resprouters, Sphagnum, shrubs and all species combined decreased significantly (P < 0.01) between 2002 (pre-fire) and 2003 (almost 1 year post-fire); herbaceous obligate seeder species had low cover in all years of measurement and there were no significant differences between years (Fig. 5). After 2003, total cover and the cover of both seeder and resprouter shrubs returned to pre-fire levels (P > 0.05). Sphagnum cover was still significantly below the pre-fire level in 2013 and 2019 (P < 0.01), although significantly greater than 1 year after the fire in 2003. Conversely, the cover of resprouter herbs (e.g. Empodisma) had increased significantly by 2019 (P < 0.01).

Fig. 5.  Box plots for the cover of all species (Total), Sphagnum, herbs (resprouters and seeders) and shrubs (resprouters and seeders). Cover was square-root transformed prior to plotting to facilitate comparisons between years for all species groups.

The mean number of non-native species/plot differed significantly between years (P < 0.001) with the highest number of species and species/plot occurring 1 year after the fire in 2003 (Table 3). The most frequent non-native species in plots in 2003 was the biennial forb Cirsium vulgare. It was not recorded before the fire or at other times post-fire. The seeds of this species are not generally dispersed far by wind but a small proportion of seed produced may be long-dormant if sufficiently buried (Michaux 1989). This suggests that C. vulgare was probably present in many peatlands pre-fire and will re-appear after each fire event if the fire interval does not exceed seed longevity.

Table 3.  Mean and total numbers of non-native species in 14 peatland plots in Kosciuszko National Park that were burnt in 2003; data collected in 2002 are pre-fire.

The results from this case study show that some changes following fire were short-lived (e.g. the cover of resprouting graminoids and the richness of non-native species) while other changes were enduring, though approaching pre-fire levels after 17 years (e.g. the reduction in the covers of Sphagnum and Dracophyllum). Importantly, the lower similarity to pre-fire floristic composition 17 years post-fire compared with 10 years post-fire shows that recovery may be affected by other factors, in this case possibly drought.

Four major fires since 1972 have burnt large tracts of sub-alpine vegetation at Mt Buffalo, in north-east Victoria. Subalpine peatland vegetation on the Mt Buffalo plateau was assessed in 2008, with plots having been burnt at a range of frequencies in the previous 80 years. The study aimed to determine whether floristic differences between plots were related to fire frequency and/or site characteristics.

Mt Buffalo consists of an extensive granite massif with a mildly dissected plateau. Its vegetation has experienced more frequent burning in the past century than high-elevation vegetation elsewhere on the mainland. Wildfires occurred in parts of the plateau in 2009, 2006, 2003, 1985, 1972, 1939, 1926 and 1918 (Rowe 1970; Zylstra 2006). Prescribed burns were also conducted in some areas in 1984, 1991 and 1994 (Department of Sustainability and Environment, unpublished data). Peatland vegetation occurs in low-lying areas within sub-alpine woodland and heathland at elevations between about 1300 and 1550 m above sea level.

Data were obtained from 39, 4 × 5 m plots in valley peatlands at the headwaters of streams, springs and gently sloping seepage lines in March 2008. These plots had not been burnt in the 2006 fire. Percentage live cover of all vascular plant species and Sphagnum were recorded in each plot.

Explanatory variables used in analyses were:

• Fire frequency. Obtained by superimposing plot locations over fire mapping (Department of Sustainability and Environment, unpublished data) using GIS (geographic information system, Arcview 3.3). Plots that were burnt or unburnt within the 1985 and 2003 fire perimeters were known from previous studies (N. Walsh unpublished data, Coates et al. 2006, respectively).

• Percent moisture (determined by comparing fresh weight with oven-dried soil weight). While soil moisture can vary greatly in time (especially after rain), peatland soils tend to be perennially wet and moisture will be less prone to large fluctuations. In addition, the peatlands were sampled within a week, and were at similar elevations and close enough to each other so that differences in soil moisture from rainfall would be small.

• Percentage organic content (measured using loss on ignition, Dean 1974) at 10 cm depth.

• Bulk density at 2 cm (ratio of dry weight to volume, g/cm³). Samples (1 cm³) were weighed, oven-dried at 95°C for 24 h, re-weighed and dry weight (g/cm³) calculated. Bulk density is a measure of the degree of decomposition of organic matter within a peat soil – the higher the value, the more decomposed the organic material (Boelter 1968).

• Humification was measured by determination of the nature and amount of material passing through the fingers on squeezing (Kershaw 1997). Samples were assigned to four categories: 0 = fresh peat yielding clear water; 1 = slightly decomposed peat yielding dark coloured, turbid water; 2 = decomposed peat yielding about half its mass; 3 = very decomposed peat yielding about three-quarters of its mass; 4 = totally decomposed peat yielding almost all its mass. Samples were taken at 2 cm depth (just below the root mat) using a D-section corer.

• Mean soil depth was calculated from five points in each plot using a steel probe (3 mm × 1 m).

The effects of the explanatory variables on floristic composition were assessed using distance-based linear models in the Permanova+ add-on to Primer v.6 (Clarke and Gorley 2006; Anderson et al. 2008). The relationships of the explanatory variables with total cover of shrubs, grasses, graminoids, forbs, and native and non-native species richness were determined by fitting generalised linear models using a poisson distribution for the response of richness variables and a normal distribution otherwise. Prior to analysis, cover values were square-root transformed and all explanatory variables were standardised (with mean = 0 and standard deviation = 1) so that model estimates could be interpreted as relative effect sizes. The correlation coefficients for all variable pairs were <0.8.

In marginal tests of the distance-based linear model, fire frequency (P = 0.01), soil moisture (P = 0.001), organic content (P = 0.01), and bulk density (P = 0.002) were significant in explaining variation in floristic data; but across all sequential tests, only soil moisture (P = 0.001) and fire frequency (P = 0.024) added significantly to the overall model (which explained 26% of total variation); bulk density was near significant (P = 0.10).

Fire frequency and soil moisture tended to have opposite effects on species composition (Fig. 6); i.e. the higher the fire frequency, the lower the soil moisture. B. gunniana, Pultenaea tenella, Caesia alpina and Celmisia tomentella tended to favour sites with higher fire frequency while S. cristatum, D. continentis, E. minus, E. paludosa, Asperula gunnii, Oreobolus distichus and C. gaudichaudiana favoured sites with higher moisture.

Fig. 6.  Biplot from principal coordinates analysis of species with vector length >0.25 and three significant or near-significant explanatory variables: Fire, fire frequency; Moisture, % soil moisture, BD, bulk density; As, Aciphylla simplicifolia; Am, Agrostis muelleriana; Ag, Asperula gunnii; Bg, Baeckea gunniana; Ca, Caesia alpina; Cb, Carex breviculmis; Cg, Carex gaudichaudiana; Cj, Carex jackiana; Ct, Celmisia tomentella; Cc, Craspedia gracilis; Dc, Deyeuxia carinata; Di, Deyeuxia innominata; Em, Empodisma minus; Ep, Epacris paludosa; Ha, Hydrocotyle algida; Hr, Hypochaeris radicata; Js, Juncus sandwithii; Oo, Oreobolus oxycarpus; Pc, Poa costiniana; Pt, Pultenaea tenella; Rg, Ranunculus graniticola; Rc, Dracophyllum continentis; Sh, Scaevola hookeri; Sg, Senecio gunnii; Sc, Sphagnum cristatum; Vf, Viola fuscoviolacea; Wc, Wahlenbergia ceracea. Axis scale at top and right is for fire, moisture and BD; axis scales at bottom and left are for the species shown.

Bulk density had a significant positive effect on grass cover and the number of non-native species, fire frequency had a significant positive effect on shrub cover, and soil moisture had a negative effect on the number of non-native species (Fig. 7). Frequently burnt peatlands had higher shrub cover. This could mean that peatlands with substantial shrub cover are more likely to burn or that fire leads to greater shrub cover, or both. If both are true, burning is likely to promote more shrub cover and make the peatlands more vulnerable to fire in the future. High elevation peatlands are noted for their small number of non-native species and minimal cover of grass (McDougall and Walsh 2007). The significant positive relationship between bulk density and non-native species and grass cover, and a high positive correlation between bulk density and fire frequency (with both being inversely related to soil moisture on Fig. 7) suggests that increasing fire frequency may also promote greater grass and non-native species establishment.

Fig. 7.  Effect size (estimate and 95% confidence interval of generalised linear mixed models) of bulk density, fire frequency and % soil moisture on graminoid cover (square root %), grass cover (square root %), forb cover (square root %), shrub cover (square root %), native species richness and non-native species richness, for 39 plots sampled in 2008 on Mt Buffalo. 95% confidence intervals that do not cross 0 can be considered significantly different (P < 0.05).

The results of this study suggest that the floristic composition of vegetation in peatlands on Mt Buffalo is influenced by fire history and soil moisture. The likelihood of recurrent fire in peatlands is exacerbated by their locations within heathlands on the Mt Buffalo plateau, and colonisation of native species from these flammable communities onto drying peat at peatland margins. Ubiquitous species widespread in these drier vegetation types accounted for some of the diversity in more frequently burnt peatlands where grasses and herbs were prominent. Gradual attrition of peatland vegetation from invasion by dryland species and the depletion of slow-growing peat endemics can be expected if saturated areas continue to contract under the types of conditions observed in 2003 and 2006 following years of drought.

After the 2003 fires, one hundred and twenty 0.5 × 0.5 m plots were established at seven fire-impacted sub-alpine peatlands in Namadgi National Park and adjacent Kosciuszko National Park to test the effect of a range of shade and transplant treatments on peatland recovery (Hope et al 2005; Whinam et al 2010). Many of these peatlands were severely burnt and there was concern that recovery would not occur without management intervention, especially for those that may still have been recovering from decades of grazing by domestic stock in the 20th century (Costin et al. 1959; Good 2000).

The fires also provided an opportunity to investigate the rate of recovery of peatland vegetation, as many areas had not been burnt in the era of ecological research in the Australian Alps (post-1950s) and, in 2003, little was known about the effect of fire in these systems. The post-fire recovery rate without management intervention was expected to be slow, with reduced Sphagnum moss cover, loss of peat, changes in species diversity and, in places, a transition to grasslands and/or Carex fens. Post-fire recovery of peatland vegetation in the ACT may also have been different because, unlike NSW and Victoria, domestic grazing was not a feature of its early land use. In this case study, we document ecosystem recovery in the 10 years post-fire in seven peatlands in the ACT and NSW using our control (no treatment) plots.

We chose 32 plots from our data that met the following criteria: no treatment, >20% Sphagnum cover pre-fire, severely burnt (i.e. in burn categories 3–5 of Whinam et al. 2010), and measured at the same intervals post-fire (i.e. 2 months post-fire, and 2, 4 and 10 years post-fire). The only pre-fire data for the plots were Sphagnum cover.

For each measurement, a 0.5 × 0.5 m frame gridded at 5 cm intervals to give 100 squares/plot, was positioned using fixed stainless steel pins in plot corners. Counts of the numbers of grid squares within the frame for each species were used as a surrogate for cover. A conservative estimate of pre-fire Sphagnum cover was obtained 2 months after the fire as the sum of dead and live Sphagnum. The heights above the ground surface of two fixed pins at each plot for three sites (Ginini Flats, Rotten Swamp, Snowy Flats) were recorded to assess relative surface change through time.

Sphagnum cover was well below pre-fire levels 10 years after the fire of 2003 (Fig. 8). In only one plot out of 32 did Sphagnum return to pre-fire cover 10 years post-fire. There was no regeneration of Sphagnum in 10 plots after 10 years; plots with no regeneration occurred at all seven sites. Even with low replication at the site level, Sphagnum regeneration is clearly highly variable at small scales. After 10 years, Sphagnum cover as a percentage of pre-fire cover ranged from 0–88% at Boggy Plain, 0–126% at Delaneys Bog, 0–35% at Ginini Bog, 0–5% at Pengilleys Bog, 0–76% at Rotten Swamp, 0–85% at Snowy Flat and 0–5% at Tom Gregory Bog. Observations showed that Sphagnum expansion tended to occur in favourable micro-habitats such as moist, shaded hollows. Conversely, Sphagnum tended not to regenerate on the exposed tops of dead moss hummocks or on sites that dried out with the loss of surface water flow.

Fig. 8.  Boxplot of Sphagnum cover in 32 sample plots in the northern Australian Alps; pre-fire in 2002, and post-fire in 2003, 2005, 2007 and 2013.

The Pearson correlation coefficients comparing Sphagnum pre-fire cover with cover at all other times were not significant (ranging from r = 0.19 to 0.30 across all sites). That is, pre-fire cover was not indicative of post-fire cover up to 10 years post-fire.

Shrubs were totally removed by the fire in the plots and mostly had < 10% cover in the first 4 years (Fig. 9). After 10 years, shrub cover had increased in most plots but was highly variable. The shrubs with highest cover in most plots were Baeckea and Epacris, both resprouting species. Dracophyllum, an obligate seeder, was not found in any plots after 10 years, though we observed burnt, dead individuals post-fire at all sites. Elsewhere in the peatlands, the first seedlings of Dracophyllum were observed 3 years post-fire and the first flowering of this species was recorded in 2019/20, 13 years after seedling establishment. Graminoid cover increased greatly in the first 2 years after fire and again between 4 and 10 years when mean cover was close to 60%. The main graminoid of sample plots was Empodisma. Forb cover reached a peak 2 years post-fire when species such as Asperula and Veronica spp. were locally abundant; some plots had up to 25% cover of forbs at that time. Forb cover declined in many plots thereafter.

Fig. 9.  Mean species richness and mean cover of forbs, graminoids, shrubs and Sphagnum (±s.e.) in 32 plots in seven peatlands in the ACT and NSW, measured in 2003, 2005, 2007 and 2013.

The peat surface was expected to gradually rise as the vegetation stabilised and produced surficial litter, or else suffer a precipitous decline as peat erosion and oxidation took hold. However, the heights of the fixed pins, on average, changed little over 10 years (Fig. 10). This was the result of both falls and rises across sampled plots, some by as much as 55 mm. There were few rises and falls in the period between 2 and 4 years after fire.

Fig. 10.  Violin plots of changes in surface level at three time intervals (for pooled data from three sites) in the northern Australian Alps. The width of each plot reflects the frequency of records with particular change.

There were many findings in common between the four case studies. The recovery of resprouting species was rapid after the 2003 fire, with most species generally reaching their pre-fire cover between six and 10 years post-fire (Case Studies 1 and 2), whereas the main obligate seeder of peatlands, Dracophyllum, was either absent post-fire (Case Study 4 observation) or took up to 15 years to reach its pre-fire cover (Case Study 1). The dominant moss of the peatlands, Sphagnum, had not attained pre-fire cover 10 years after the 2003 fire in Case Study 4 and was approaching pre-fire cover 15–17 years after the fire in Case Studies 1 and 2. Non-native species richness increased in Case Studies 1 and 2 and the invading species tended to be persistent where largely resprouters (Case Study 1) and transitory where largely annuals (Case Study 2). Soil moisture was found to be an important predictor of floristic composition in peatlands (Case Study 3) with species most characteristic of the community favouring sites with high soil moisture. The influence of soil moisture on the rate and pattern of peatland recovery (and in particular Sphagnum expansion and survival) was evident in Case Study 1. Native species richness was unchanged 1 year after the fire and for the following 16 years, and post-fire floristic composition was closest to pre-fire composition 10 years after the fire (Case Study 2). Clarke et al. (2015) also found that floristic composition of peatlands in Kosciuszko largely returned to pre-fire composition after a decade.

The case studies also demonstrated how vegetation recovery after fire can be idiosyncratic. In part, the idiosyncrasies may relate to scale. For instance, at the scale of plots in Case Study 4 (0.25 m²), 10 years post-fire Sphagnum was not recorded at all in 10 plots but had increased in cover in one plot. However, at the scale of plots in Case Study 2 (25 m²), Sphagnum was recorded in all but one plot 1 year post-fire though at reduced cover, while at the scale of connected plots in Case Study 1 (2000 m²), Sphagnum recovery could be very different in two peatlands only 200 m apart. Idiosyncrasy may also be related to the legacy of disturbance from grazing in the late 19th and much of the 20th Century because trampling by livestock can cause stream entrenchment and lower water tables (Wimbush and Costin 1983; Ashton and Williams 1989). Soil moisture was found to be a key predictor of floristic composition in peatlands of Case Study 3 and so the variable intensity of grazing by cattle and sheep across the Australian Alps has perhaps led to differences in the rate of recovery from recent fires. A variable response to fire at the site level was similarly reported from peatlands in Canada, where the trajectory of change could depend on micro-topography and fire severity (Benscoter and Vitt 2008). At that scale, clear differences were observed between regeneration on hummocks and the hollows between them, relating to water availability. Similarly, in Case Study 4, regeneration of Sphagnum was observed to be greater in shaded, moist hollows, and in Case Study 1, hummocks of Sphagnum died out when they apparently became separated from the water table, regardless of whether they had been burnt. The variable vegetation recovery of plots within Case Studies 1, 2 and 4 show how important it is to avoid making assumptions about fire response based on a single site.

Sphagnum recovery was slow in comparison with most other peatland species but may be delayed further if the fire also causes hydrological changes in the peatland system, as seems to have been the case for one of the peatlands in Case Study 1. A fire frequency less than 15 years is likely to put the community at considerable risk of permanent change. Over time, the more a peatland is burnt, the higher the likelihood of its water source being diverted and the habitat becoming unsuitable for Sphagnum. This is supported by Case Study 3, where fire frequency was significantly related to species composition – peatlands experiencing more fires were characterised by less Sphagnum and more species characteristic of surrounding dryland communities.

The case studies demonstrate that most peatland species return quickly following fire, with some reaching their pre-fire cover within 10 years. Indeed, Williams et al. (2012) found that some species in sub-alpine peatlands in Victoria had reached their pre-fire cover 3 years after burning. This is not surprising as regeneration from basal buds that resprout after fire is the dominant strategy of peatland species. Conversely, for the obligate seeder shrub Dracophyllum, germination did not occur until the third growing season after fire and, based on observations of plots in Case Study 4, flowering does not occur until around 13 years after recruitment; i.e. 16 years after the fire. Unless this species possesses a dormant seed bank, fire with a return period of less than 16 years is likely to lead to its local extinction. Obligate-seeder shrubs were found to be rare or absent from burnt alpine vegetation in Tasmania (Kirkpatrick and Dickinson 1984). Norton and De Lange (2003) suggested that some rare forbs in New Zealand peatlands may require fire for persistence by briefly removing competition and promoting seed production. This may also be the case in the Australian Alps where species such as S. hookeri and V. fuscoviolacea briefly became common in burnt peatlands of Case Study 1 that had not experienced fire since 1939. They presumably persist as a long-lived seed bank.

Based on the most fire-sensitive species in peatlands and the large variability in time to recovery, a fire interval of less than 20 years will often be a threat to peatland function and species composition. The risks of fire occurring more frequently than every 20 years, and the subsequent damage that may cause, are possibly greater at low elevations where shrub cover is naturally greater. As was found in Case Study 3, shrubbier peatlands tended to be associated with more frequent fire, whether by cause or effect. So, how should the surviving peatlands of the Australian Alps be managed in a climate future that is more variable in its rainfall, hotter and more prone to high intensity and high frequency wildfire? The risk to vegetation from fire is typically managed by fuel reduction or asset protection. However, these approaches may not be practicable or effective in the Australian Alps.

In the United Kingdom, fuel reduction by burning in peatlands has traditionally been used to promote habitat for grazing and hunting animals by removing the shrub layer while keeping the moss layer intact, but is increasingly being viewed as a management tool for reducing the risks to these habitats from wildfire (Grau-Andrés et al. 2019). However, the value of fire in habitat management there has been contested because of variable outcomes depending on intensity (e.g. Grau-Andrés et al. 2019; Noble et al. 2019; Whitehead et al. 2021). Differences in plant regenerative strategies between Australian and UK peatlands suggest that prescribed burning in Australia would not achieve an aim of reducing the risk of damage to peatlands from wildfire; rather increased fire frequency may actually escalate risk by promoting establishment and growth of shrubs and grasses in some peatlands (Case Study 3). Many species in UK peatlands regenerate from seed and so the vegetation is reset after fire from seed present on site and that arriving from surrounding communities (Shepherd et al. 2021). Here, vegetative resprouting is the dominant regenerative strategy and such species regain structural dominance within a few years. A high fire frequency would be needed to achieve a goal of limiting shrub dominance but that would likely risk the elements of the peatland that such intervention was intended to protect, in particular, Sphagnum and the peat layer beneath. Fire suppression in areas surrounding peatlands, to reduce the likelihood of wildfire, is also likely to be ineffective. Many peatlands on the Bogong High Plains are surrounded by a dense grassy sward, which does not carry fire well. We observed that the fire there in 2003 jumped these grasslands and ignited the shrubby cover in the peatlands, then burnt through the Sphagnum into the peat.

Nelson et al. (2021) reviewed the factors that contribute to resilience in peatlands. Hydrologically connected, moist, unaltered peatlands with low bulk density tend to experience less severe fires. Conversely, drier, hydrologically fragmented peatlands with high bulk density will experience higher intensity burns perpetuating the dryness and vulnerability of the systems, while also releasing carbon. Our measurements and observations suggest that this model of northern hemisphere peatland resilience holds true for peatlands in the Australian Alps. Indeed, a link between bulk density and fire likelihood is consistent with the findings of Case Study 3. If peatlands are wet at the time of the fire and remain so afterwards, recovery seems to be quicker and more complete. Disturbances that cause peatlands to dry out by altering their hydrology, such as entrenchment of the streams that supply peatlands with water caused by trampling and damage of soils by horses, deer, pigs, vehicles and bushwalkers (e.g. Robertson et al. 2019; Tolsma 2020), are likely to make peatlands less resilient and more vulnerable to future large fire events. In a peatland in Canada, a slight reduction in the water table of a peatland, caused by disturbance and followed by fire, facilitated the replacement of peatland species by species from surrounding vegetation, making the peat soils more vulnerable to subsequent fire (Kettridge et al. 2015). Slowing water movement through peatlands by placing impediments in streams that have been damaged by trampling may improve resilience in some cases. Sphagnum growth can be promoted post-fire with shading and its recovery boosted by planting Sphagnum plugs into severely burnt peatlands (Hope et al 2005; Whinam et al. 2010), further promoting water retention. While the former might only be practicable in a few important or easily accessible peatlands after a fire, the planting of plugs might be done more broadly to speed recovery of the Sphagnum layer. Establishing a wild supply of hydroponic Sphagnum during severe droughts when the risk of wildfire is great would enable the faster recovery of peatlands and provide nuclei for Sphagnum expansion, as most observed recovery is vegetative.

Case Studies 1, 2 and 4 documented the recovery of peatland vegetation after a single fire in 2003. Post-fire colonisers, including non-native species, exploited the initial increase in bare peat and lack of competition. These were quickly replaced by resident resprouters whereas obligate seeders and Sphagnum were slowest to recover. Differences in the rate or extent of recovery may be related to the inherent variable intensity of fire and the floristic variability across the sub-alpine Sphagnum peatlands of the Australian Alps, as well as the frequency of burning, past land use, the wetness of the peat at the time of fire and climatic conditions post-fire. Despite these complexities, most peatlands had recovered floristically within a decade and in the cover of dominant species within two decades. Case Study 3 demonstrated how vegetation can change following multiple fires, with typical peatland species becoming less common and species from surrounding dryland communities more common. Peatlands of the Australian Alps can recover from fire if the interval between fires is not too small. The fire frequency threshold that leads to community change is unknown but likely to be at least 20 years.

The future of Australian ecosystems will not be shaped solely by the effects of a single fire but by the patterns of fire in space and time, and by the intensity of other threatening processes (Keith et al. 2022) such as grazing and trampling by feral animals (Robertson et al. 2019; Tolsma 2020), the invasion of woody weeds (e.g. Moore 2020), the introduction of pathogens (McDougall et al. 2018), and predicted increasing temperature and decreasing precipitation (especially snowfall) associated with climate change. Continued monitoring on the recovery and/or change to these communities due to the impact of the 2003 fires will provide valuable information for their management. Future management of peatlands has to be different from past management, when there was an assumption of low risk because threats were infrequent. Greater intervention is likely in the future, with active restoration being required at times. However, ecosystem resilience can be built now by reducing manageable threats such as weeds and trampling by feral animals.

Data from 1979 and 1999 for Case Study 1 are published in McDougall (1989). Other data used in the paper for Case Studies 1, 2 and 4 will be shared upon reasonable request to genevieve.wright@environment.nsw.gov.au. Data for Case Study 3 are available in the Victorian Biodiversity Atlas and are included in Coates et al. (2012).

The authors declare no conflicts of interest.

No funding was obtained for the additional data collection reported in this paper. Support for previously published work associated with the case studies is reported in the relevant acknowledgement sections.

KM collected data for Case Studies 1 and 2, analysed data and prepared a first draft of the manuscript; JW and GH collected data for Case Study 4; FC collected data for Case Study 3; JM collected data for Case Study 1; NW collected data for Case Studies 2 and 3; GW collected data for Case Studies 2 and 4; all authors contributed to the preparation of the final manuscript.

We dedicate this paper to the memory of our friend and colleague Emeritus Prof. Geoff Hope, who died during its preparation. Geoff had a life-long passion for field work, particularly in the mountains of West Papua, Papua New Guinea, the Pacific and south-east Australia, endeavouring to comprehend and explain landscapes over geologic time. Geoff was an enthusiastic advocate for peatlands, promoting understanding of their evolution, their special place in the Australian landscape, and the need for protection and conservation management - as generations of students, colleagues, family, friends and land managers will attest to. Vale Geoff.