Fire and functional traits: Using functional groups of birds and plants to guide management in a fire‐prone, heathy woodland ecosystem

Many dry forests and woodlands worldwide are fire‐prone and support bird and plant communities shaped by fire. Changes in fire regimes, including the time between fires, have important implications for population trajectories. We studied the responses of bird and plant communities of heathy woodlands to time since the last fire, a key measure underpinning fire management, to evaluate whether current management strategies will enhance conservation of multiple taxa.

evidence that fire regimes are changing in Mediterranean-type ecosystems-including sclerophyll vegetation of Australia, chaparral in the United States, fynbos in South Africa, and matorral in Chile-with substantial consequences for the biota (González et al., 2018;Keeley & Syphard, 2019;Msweli et al., 2020;Nolan et al., 2020;Vitolo et al., 2019). Shifts in fire regimes can threaten species and ecological communities with risk of local extinction, including by compromising their ability to survive changing conditions Giljohann et al., 2015;He et al., 2016;Stephens et al., 2019). There is an urgent need to understand how fire, and fire management practices, affect multiple taxa and biological levels.
For animals, persistence in fire-prone environments is closely linked to habitat structure and depends on their ability to capitalize on food and shelter resources associated with post-fire vegetation changes (Fox, 1982;Sitters et al., 2014;White et al., 2016).
For example, the Dartford Warbler (Sylvia undata) in southern Europe relies on dense low shrub cover that results from past fires in Mediterranean woodlands (Pons et al., 2012). A focus on postfire vegetation succession and accompanying changes in habitat structure has translated into management for animals that aims for spatially diverse fire mosaics (Kelly et al., 2018). In south-eastern Australia, for example, post-fire age classes, determined by compositional and structural changes to vegetation, have been defined to represent key "growth stages" in post-fire succession in an ecosystem (e.g., Cheal, 2010). Even though much of the fauna of temperate south-east Australia has not shown strong preferences for particular successional stages (e.g., Rainsford et al., 2021;Swan et al., 2015), the assumption that such categories represent distinct successional communities has rarely been tested for animal taxa.
For plants, persistence in fire-prone landscapes depends on the ability of populations to recover from fire events, either through germination of seeds, resprouting of survivors or a combination of both (Keith et al., 2007;Keeley et al., 2011). As such, the interval between fires is critical for both reproductive and resprouting success, but the required interval depends on species' traits such as time to reproductive maturity and the life span of established plants and seedbanks Enright et al., 2015;Menges, 2007).
An emphasis on the species traits of plants has translated into management that aims for variation in fire intervals over time (Kraaij et al., 2013;Menges, 2007;Noble & Slatyer, 1980;van Wilgen et al., 2011). An influential approach is the concept of minimum and maximum tolerable fire intervals (TFIs) (closely related to the concepts of "thresholds of particular concern" and "bounded ranges of variation"; Kelly et al., 2018). The minimum tolerable fire interval is based on the number of years required post-fire for key plant species to successfully reproduce and set seed, while the maximum tolerable fire interval represents the age post-fire at which these key species begin to senesce (Bradstock & Kenny, 2003;Kraaij et al., 2013). For example, in Victoria, Australia, fire managers quantify the proportion of the landscape that is either within, or outside, the recommended tolerable fire interval to decide where and when to carry out prescribed burns (York & Friend, 2016). This approach, however, does not explicitly recognize the needs of other taxa, such as animals (Clarke, 2008), although there has been progress in recent years. Diversity metrics based on species' abundances have been used to test the effect of the spatial structure of post-fire age classes on biota (Giljohann et al. 2015;Kelly et al., 2015;Di Stefano et al., 2013).
However, a key knowledge gap remains how such practices shape ecological communities.
Recognizing functional groups that share common traits, and doing so for multiple taxonomic groups, has the potential to improve fire management for biodiversity. Recurrent fire has driven the evolution of a variety of traits that enable species to persist in fire-prone landscapes (Pausas & Parr, 2018). For animals, important traits determining post-fire occurrence relate to habitat-use (e.g., foraging behaviour, diet specialization, nest type) (Gosper et al., 2019;Jacquet & Prodon, 2009). Relating these traits to species' fire responses may help to reveal the processes that drive patterns of biodiversity across fire-prone landscapes, which can inform management strategies. Generalizations based on analyses of species' life history traits will build understanding of how ecosystem structure and function change following fire (Gosper et al., 2019;Sitters et al., 2016). This knowledge will help identify those species and ecosystem processes that may be affected by more frequent fires, or benefit from more widespread prescribed burning, and thereby help improve decision making in fire management.
Here, we examine how approaches to fire management may affect bird and plant communities in a heathy woodland ecosystem dominated by epicormic-resprouting Eucalyptus trees in southeastern Australia. Specifically, we aimed to determine how (a) individual species, (b) functional groups of species and (c) the composition of bird and plant communities respond to time since fire. We then discuss these outcomes in the context of the potential impacts of fire management as guided by (a) tolerable fire intervals, based on plant life history traits (e.g., reproductive age), and (b) the spatial arrangement of post-fire age classes (based on vegetation growth stages), on the bird and plant communities of heathy woodlands.

| Study area
The study area is part of the Great Otway National Park and Forest Park in the Otway Plains bioregion of southern Victoria, Australia ( Figure 1). The climate is temperate with mean annual rainfall ranging from 540 to 895 mm. The highest rainfall occurs in winter (August) and the hottest month is February (mean daily maximum 28°C) (Mount Gellibrand, station no. 090035l, Cape Otway lighthouse, station no. 090015; http://www.bom.gov.au/). The topography is gently undulating with elevation from ~40 to 250 m above sea level.

| Vegetation
Heathy eucalypt woodlands in south-eastern Australia mostly are confined to coastal areas, with some inland occurrences on nutrientpoor, deep sandy soils. In the study area, the canopy is low (≤ 10 m) and dominated by Brown Stringybark (Eucalyptus baxteri) and Western F I G U R E 1 Study area and heathy woodland vegetation. (a) Map of the study area showing the extent of heathy woodlands, the location of study sites and distribution of fire age classes (darker tones represent younger age classes). White areas have been cleared or heavily disturbed, including for human settlement and agriculture. Heathy woodland vegetation: (b) ~1 year after prescribed fire, showing a scorched canopy of Eucalyptus species resprouting epicormically, Austral Grass-tree resprouting apically; and (c) 51 years after fire with a welldeveloped mid-storey dominated by Monotoca glauca and Leptospermum species following disturbance ( Figure 1c) (Rainsford et al., 2020).

| Fire regime and fire history
In heathy woodlands, wildfires typically occur in summer months while prescribed burning is carried out by government agencies during autumn and early spring. Wildfires occur at ~20-to 100-year intervals (Murphy et al., 2013). In the study area, a wildfire occurred in 1939 and several wildfires also burned patches during the 1960s. Prescribed fire is employed to achieve objectives relating to fuel reduction and, less often, biodiversity conservation. The timing of prescribed burns in Victoria is guided by the designation of minimum and maximum tolerable fire intervals (TFIs) for a particular vegetation type, based on the number of years required for key plant species to set seed (minimum TFI) and begin to senesce (maximum TFI) (Cheal, 2010). Prescribed fires typically burn more patchily than wildfires, but in heathy woodlands both fire types generally scorch the canopy (Figure 1b).

| Study design
Potential study sites were selected to meet four main criteria. First, sites were selected to sample a single vegetation type (heathy woodlands Ecological Vegetation Class, Victorian Government Department of Environment & Sustainability, 2004). Second, we used fire history maps (in a GIS) to identify sites spanning a chronosequence from 1to 79 years post-fire. Sites were selected to evenly cover a range of post-fire "age classes," successional states based on the vegetation growth stages described by Cheal (2010). These were 0.5 -2.5 years (AC1, 8 sites), 2.5 -8.5 years (AC2, 10 sites), 8.5 -33.5 years (AC3, 8 sites), >33.5 years since fire (AC4, 12 sites). Third, sites were selected to be away from gullies to avoid the influence of inherent differences in productivity and structure of vegetation between gullies and slopes.
Last, sites were located at least 1 km apart to enhance sample independence. Potential sites were checked in the field in relation to these criteria. The mapped time since fire was ground-truthed by looking for signs of charring on eucalypt bark, epicormic resprouting and other structural features. In total, 38 sites were selected for study.

| Data collection
To survey bird and plant communities at each site, we established a 250-m transect, commencing at least 50 -100 m from a road edge.
The start-point of the transect was randomly selected in a desktop GIS.

| Birds
To sample the bird community at each site, a 2-ha plot was centred over the 250-m transect and surveyed by a single experienced observer (FR) a total of six times: three times during the austral autumn/winter and three times during spring/summer, between 2017 and 2018. Surveys were conducted in clear weather within four hours of dawn, except for two winter survey rounds during which sites were each surveyed once in the morning and once in the afternoon. Each survey was undertaken over a 20-min period, and all individuals either heard or seen were identified to species level and recorded. The perpendicular distance (m) to each detection from the transect line was estimated to test for differences in detectability between sites. Nocturnal birds, raptors, and swifts were recorded but excluded from all analyses as these groups of birds are not reliably detected using these surveys methods.

| Individual species
To account for potential issues of detectability of birds, we first used linear regression to test for a relationship between the distance to detection of species (or groups of similar species) and mid-storey vegetation cover (see Appendix S1). For several bird taxa, there was a weak negative relationship between mid-storey vegetation cover and distance to detection, suggesting some individuals of these species may not have been detected at sites with high mid-storey vegetation cover. To control for potential errors due to detectability, we used a presence/absence-based index (reporting rate) to compare the relative abundance of species between sites. Reporting rate is the number of survey rounds during which a species was detected (here, from 0 to 6). Because it does not rely on counts of individuals, reporting rate is less prone to biases caused by differences in detectability or flocking behaviour.
This approach is a robust alternative to model-based approaches (i.e., distance analysis) for which modelling assumptions cannot be met (Hutto, 2016) and is a reliable proxy for relative abundance (Royle & Nichols, 2003).
For individual species of birds and plants that occurred at ≥7 sites, we used generalized additive models (GAM) (Wood, 2017) to predict changes in reporting rate and cover, respectively, with time since fire. Models for species with fewer records failed to converge.
For birds, models were fitted using the Poisson error distribution.
An observation-level random factor was used in this mixed-model framework if overdispersion of data was detected, following Harrison (2014).
For plants, the mean projected foliar cover was modelled by using the beta error distribution. The beta distribution can overcome inherent problems with proportion data (i.e., bounding at zero and one) that violate the assumptions of other distributions (Douma & Weedon, 2019). For one species, silver banksia, using a beta distribution, was problematic and a better model was fitted by using the Gaussian error distribution. The distribution of data supported this decision. GAMs were built using the mgcv package in r (Wood, 2017). Classification of plant regeneration traits follows Clarke et al. (2015). The classification system for plant growth forms was based on Meers et al. (2010) and . The number of species detected within each group is given. Individual species classifications can be found in Appendices S2 and S3. a Food types include fruit, nectar or pollen, seeds, foliage or herbs, corms or tubers, terrestrial invertebrates, terrestrial vertebrates, carrion (Garnett et al., 2015).

TA B L E 1 Functional groups of birds and plants in heathy woodlands
By assessing fitted response curves, species were each assigned to a generalized response curve as described by Watson et al. (2012).
We then calculated the percentage of species with a significant relationship with time since fire (p < .05) that resembled each response shape. Four response shapes were detected: "irruptive" (abundance highest in the first few years following fire), "bell" (initial increase followed by a decrease with time since fire), "incline" (gradual increase with time since fire) and "plateau" (initial increase followed by stability in later years post-fire). Non-significant relationships were classed as "NS."

| Community composition
To test the influence of time since fire on the composition of bird and plant communities, we used the four post-fire age classes (AC1 to AC4, see above) based on vegetation growth stages (Cheal, 2010).
AC4 included two growth stages, "waning" and "senescence," because there were fewer sites in these categories and their vegetation structure is similar. We then used reporting rate (for birds) and relative cover (for plants) matrices and non-metric multidimensional scaling (NMDS) ordination analysis. NMDS represents ecological communities in lower-dimensional space, based on their dissimilarity (Legendre & Legendre, 1998 (Garnett et al., 2015), was used to classify birds into functional groups.
We used a database of plant vital attributes for the Victorian flora (Cheal, 2011) to determine fire regeneration traits of plant species.
If a species' regeneration trait was not listed, it was classified based on congeneric species unless this was not available or there was inconsistency within the genus, in which cases the species was not included in the analyses.
We then used GAMs to model (

| RE SULTS
We made 3,975 detections of 44 species of diurnal bird (Table S2).
We did not detect a significant effect of location or withingroup dispersion on the differences between post-fire age classes in either bird (p = .968) or plant (p = .418) community composition (Appendix S8).
Functional group analyses revealed a strong influence of life history traits on bird species' relationships with time since fire, in some cases ( Figure 4). The summed reporting rate of ground-foraging species was greatest in recently burnt vegetation, then declined rapidly in the first 20 years after fire (p < .01, deviance = 35% Figure 4a).
The reporting rate of birds that forage in lower-mid-storey vegetation increased with time since fire and plateaued at ~50 years postfire (p < .001, deviance = 57%, Figure 4b). For birds that forage in the upper-mid-storey, reporting rate increased linearly with time since fire (p < .05, deviance = 13%, Figure 4b) and species that forage throughout the vertical strata did not respond to time since fire. The reporting rate of open-nest species increased with time since fire (p < .01, deviance = 22%, Figure 4c). Hollow-nesting species were not strongly associated with time since fire. Reporting rate of birds

| D ISCUSS I ON
In this study, we examined how communities of birds and plants Tree hollows are a limiting resource for many Australian faunal species because they take decades to develop (Gibbons et al., 2000) and may be destroyed by fire . There are two plausible hypotheses to explain this contrast (a) because the stems of canopy trees in heathy woodlands are not killed by fire (c.f., the obligate-seeding trees), the presence of tree hollows in this ecosystem is not strongly associated with time since fire and so this nesting resource is not a limiting factor post-fire; and (b) hollow-nesting birds may forage within heathy woodlands but nest in adjacent vegetation types (e.g., wet forest, foothill forest). Further studies to determine the nesting behaviour of birds in heathy woodlands and the surrounding landscape would benefit fire management.
For plants, key changes in community structure over time since fire are attributed to (a) increasing cover of facultative-resprouting and shrub species and (b) declining species richness of obligateseeding, low shrub and shrub species. A decline in plant diversity over time following fire has been observed in several ecosystems (e.g., Fournier et al., 2020;Gosper et al., ,2012Keeley et al., 2005;Penman et al., 2009). When fire consumes aboveground biomass, light, nutrients and space become more accessible, facilitating germination of seeds and/or growth from resprouting buds, depending on species (Safford & Harrison, 2004). Consequently, aboveground species richness often is high soon after fire. Then, over time, some species become dominant (e.g., Austral Grass-tree,  (Safford & Harrison, 2004), climate and weather (Burrows et al., 2019;Parra & Moreno, 2018), and other attributes of the fire regime (Keeley et al., 2005;Kelly et al., 2017). In this study, it is likely that between-site differences in productivity influenced variation within age classes, due to differences in topography (e.g., position on slope) and soils (e.g., depth of sand).
A key difference between the responses of bird and plant communities was the decline in aboveground species richness of most plant functional groups with time since fire versus the increase and plateau response of lower-mid-storey foraging birds. This presents a challenge for conservation management in heathy woodland ecosystems, due to a risk of (a) potential loss of floristic diversity in the absence of fire, and (b) reduction in bird species abundances if fire is too frequent. We note that patterns of seedbank diversity of plants in heathy woodlands may show alternative patterns to that of aboveground mature plants (Chick et al., 2019). Knowledge of the length of seed viability for obligate-seeding plants, and other groups, is useful for determining the maximum fire intervals tolerated before significant loss of species occurs. Nevertheless, there will likely be loss of adult obligate-seeding plants in the complete absence of fire.

| Implications for fire management
Fire management strategies, including the timing and placement of prescribed burns, will have greatest benefit if they incorporate biodiversity responses to fire along with reduction of risk to human life and property from wildfire (Driscoll et al., 2010). Our results have implications for two approaches to fire management that largely are based on plant responses to fire.
First, this study highlights a particular challenge to the timing of prescribed fire based on tolerable fire intervals (TFIs) because these differentially affect taxa. The peak in species richness of obligate seeders, shrubs and low shrubs and the cover of facultative resprouters and shrubs, coincides with the minimum tolerable fire interval for heathy woodlands (i.e., 12 years; Cheal, 2010), whereas the peak in abundance of lower-mid-storey foraging birds was later than the maximum TFI (i.e., 45 years) ( Figure 6). Other work on plant responses to fire in heathy woodland vegetation (Chick et al., 2019;Duff et al., 2013) also suggests that introducing more fire into the landscape would benefit plant diversity. This could be  It is important to note that several components of bird and plant communities were not related to time since fire (e.g., birds that forage throughout the strata, hollow-nesting birds, herb and low shrub species). The distribution of these groups is likely influenced by landscape heterogeneity not related to fire (e.g., topographic variation), other components of the fire regime (e.g., between-fire interval, spatial configuration) or their generalist traits. Incorporating further complexity, in addition to time since fire (e.g., Swan et al., 2018, Hutto et al., 2020, into a landscape mosaic approach to fire management will further benefit bird and plant diversity.

| CON CLUDING REMARK S
By identifying mechanisms that shape bird and plant communities post-fire, a deeper understanding is gained of how manipulating fire regimes influences components of biodiversity. Fire management guided by measures based solely on plant functional traits (e.g., TFI) can disadvantage faunal communities. Rather, incorporating the responses of both plant and animal communities to fire by spatial representation of post-fire age classes of greatest value to different plant and animal groups Di Stefano et al., 2013) will help conserve multiple taxa in fire-prone landscapes. However, incorporating further complexity in landscape planning in addition to time since fire is needed to represent those components of ecosystems not strongly related to time since fire. This includes explicitly incorporating environmental gradients and topographic variation, as well as considering other temporal attributes of the fire regime (e.g., between-fire interval), the spatial context of fire (e.g., amount or diversity of fire) and landscape context (e.g., surrounding vegetation types, connectivity). Finally, incorporating species' functional traits into fire management frameworks can also help guide management actions to achieve explicit conservation targets, based on how ecosystems change over time following fire. This may be especially useful in instances where data and knowledge of particular species are lacking.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13278.

F I G U R E 6
The response of bird and plant functional groups to time since fire and tolerable fire intervals (TFI) in heathy woodlands. (a) The number of species of ground-foraging birds and obligate seeder plants. (b) The reporting rate of lower-mid-storey-foraging birds and the relative cover of shrubs. Lines are fitted generalized additive models. Shaded areas represent 95% confidence intervals. Values on the y-axis are relative to each taxon