- Top of page
1 Both fire regimes and the conditions under which fires occur vary widely. Abiotic conditions (such as climate) in combination with fire season, frequency and intensity could influence vegetation responses to fire. A variety of adaptations facilitate post-fire recruitment in mediterranean climate ecosystems, but responses of other communities are less well known. We evaluated the importance of climate by comparing sites with mediterranean and subtropical climates.
2 We used paired burned and mature sites in chamise chaparral, mixed chaparral and coastal sage scrub (California), and rosemary scrub, sand pine scrub and sandhill (Florida), to test whether (i) patterns of pre-fire and post-fire seedling recruitment are more similar between communities within a region than between regions, and (ii) post-fire stimulation of seedling establishment is greater in regions with marked fire-induced contrasts in abiotic site characteristics.
3 Post-fire seedling densities were more similar among sites within climatic regions than between regions. Both seedling densities and proportions of species represented by seedlings after fires were generally higher in California.
4 The only site characteristic showing a pre-fire–post-fire contrast was percentage open canopy, and the effect was greater in California than in Florida. Soil properties were unaffected by fire.
5 Mediterranean climate ecosystems in other regions have nutrient-poor soils similar to our subtropical Florida sites, but show post-fire seedling recruitment patterns more similar to the nutrient-rich sites in California. Climate therefore appears to play a more major role than soil characteristics.
- Top of page
Although natural fires occur in virtually all terrestrial ecosystems (Whelan 1995), variations in predictability, frequency and intensity of fire lead to quite different community responses (Bond & van Wilgen 1996). Environments with unpredictable, infrequent fires that function as catastrophic disturbances (e.g. tropical rainforests; Mueller-Dombois & Goldammer 1990) contrast with landscapes with more frequent and predictable fires, where fire has played a selective role in the evolution of life-history characteristics, as seen in mediterranean climate ecosystems (Christensen 1985; Keeley 1986).
Natural fires are essential for maintaining species diversity in those mediterranean climate scrub communities in which a large percentage of plant species (at least 70%) have seedlings that establish after fires (Vogl & Schorr 1972; van Wilgen & Forsyth 1992). Many such species are fire-dependent, and seedling recruitment, which is nil in the absence of fire, is triggered by immediate post-fire conditions as a result of stimulation of germination of a persistent soil seed bank by heat, smoke or charred wood (Keeley et al. 1985; Bond & van Wilgen 1996; Keeley & Fotheringham 1998). It is generally assumed that this is an evolutionary response to improved conditions for survival after fire, such as the creation of large light gaps for recruitment (Keeley 1986). Similarity in post-fire response among mediterranean climate ecosystems is commonly cited as an example of evolutionary convergence between these climatically similar regions (Cody & Mooney 1978; Naveh & Whittaker 1979; Shmida 1981), although the degree of climatically driven convergence has been questioned (Barbour & Minnich 1990).
Non-mediterranean communities such as Florida scrub can be used to assess the importance of climate. Although Florida scrub is physiognomically similar to mediterranean-type vegetation, this scrub persists under a summer rain/winter drought (rather than winter rain/summer drought) climate. Florida scrub is subjected to fires of varying frequency (Myers 1990; Brenner 1991), with Florida having the highest frequency of lightning-ignited fires in the United States (Komarek 1964).
This study evaluated the role of fire in altering vegetation structure in three plant communities in inland peninsular Florida, and compared these patterns with three mediterranean communities in southern California. We tested two predictions: (i) that patterns of pre-fire and post-fire seedling recruitment are more similar between communities within a region than between regions; and (ii) that post-fire seedling establishment is more prevalent in regions with larger fire-induced contrasts in abiotic site characteristics (e.g. canopy cover, soil nutrients). We determined pre-fire and post-fire seedling densities and species richness by sampling paired burned and unburned sites. We also assessed site characteristics and used canonical analysis to relate site variables to seedling establishment patterns, to suggest processes underlying the observed patterns.
- Top of page
Seedling densities in mature sites did not differ between Florida and California, although densities tended to be higher in California (Fig. 3). Among recently burned sites, seedling densities were lowest in Florida sandhill and highest in California coastal sage scrub, and, overall, were higher in California communities (planned comparison test, P < 0.001; Fig. 3). Recently burned sites tended to have higher numbers of seedling species in California, even though species were quantified there over a smaller area per site (Table 1 and Fig. 3). Annuals comprised large proportions of the seedling species in California, but were virtually absent in Florida communities (Fig. 3).
Figure 3. Seedling density in mature (M) and recently burned (B) sites in Florida and southern California, and proportion of species represented by seedlings, number of species of seedlings of annuals and perennials, and total number of seedling species. Sites consisted of rosemary scrub (RmS), sand pine scrub (SPS) and sandhill (SaH) in Florida; and chamise chaparral (ChC), mixed chaparral (MxC) and coastal sage scrub (CSS) in California. Bars represent means and standard errors from five sites of each community type. Letters above bars show results of planned contrasts among recently burned sites after a Kruskal–Wallis test (for seedling densities) and an analysis of variance (for proportions of species represented by seedlings). Different letters denote significant differences (P < 0.05). Asterisk shows a difference between mature and recently burned sites (P < 0.05 for RmS only). Seedling species numbers were not counted over the same area at all sites, and differences in means were therefore not tested statistically.
Download figure to PowerPoint
Table 1. Mean proportions of species represented by seedlings, total numbers of seedling species and numbers of these species thought to be dependent on fire for recruitment in recently burned sites in southern California and Florida
|Location (n = 15 sites)||Proportion of species represented by seedlings (mean ± standard error)||Total number of seedling species||Number of species dependent on fire for recruitment*|
|California||0.62 ± 0.04||74||25|
|Florida||0.30 ± 0.03||25||7|
Total species richness (i.e. including adult plants) varied from 21.2 ± 3.1 (mean and standard error; Florida sand pine scrub) to 38.8 ± 6.3 (Florida sandhill). Mean percentage of total species represented by seedlings varied from 18.0 ± 2.1% in Florida sandhill to 71.4 ± 3.4% in California chamise chaparral, and was higher in California communities (F1,24 = 70.8, P < 0.001; Table 1 and Fig. 3).
California and Florida sites were very different with respect to most of the site variables quantified. California soils had higher percentage total nitrogen, percentage clay and pH, while Florida soils had higher percentage sand (stepwise discriminant analysis, P < 0.05 for nitrogen, clay and pH; Fig. 4).
Figure 4. Means and standard errors for percentage open canopy and percentage total soil nitrogen, percentage total soil phosphorus, soil pH and percentage soil sand and clay in burned and mature sites in Florida and California (abbreviations as in Fig. 3). n = 5 for each community type.
Download figure to PowerPoint
Differences between burned and mature sites
Among Florida community types, only rosemary scrub had higher percentage open canopy, seedling density or soil pH in burned areas (Figs. 3 and 4). An unusually long time (151 days) between collection dates within a pair of sites did not affect the contrasts. Although seedling densities did not differ within site pairs in sand pine scrub and sandhill, seedling species composition did vary. Coefficients of community similarity between paired sites (Mueller-Dombois & Ellenberg 1974) were 0 and 0.23 for sand pine scrub and sandhill, respectively (data not shown). Burning did not affect soil phosphorus or soil nitrogen at any Florida sites (Fig. 4).
Percentage open canopy alone differed between paired California sites, with burned sites having greater values (Fig. 4). Seedling species were more similar between burned and mature sites in mixed chaparral and coastal sage scrub than they were in chamise chaparral or in any of the communities in Florida. Coastal sage scrub had the highest coefficient of community similarity (0.37), although almost half of the species shared by burned and mature sites were exotics.
Patterns of seedling and site variables in burned sites
The canonical discriminant analysis comparing Florida community types yielded one significant axis that accounted for 80% of the variability in the data. Rosemary scrub sites, with more open canopies and sandier, higher-pH, lower-nitrogen soils, were clearly separated from sand pine scrub and sandhill sites along this axis. Sand pine scrub and sandhill were less clearly separated from each other, but the former tended to have less open canopies, higher soil nitrogen and lower pH (Fig. 5).
Figure 5. The first and second canonical axes from canonical discriminant analyses of recently burned sites in (a) Florida and (b) California (abbreviations as in Fig. 3). Coordinates of vectors represent direction and strength of loading on canonical axes for six site variables (see Table 3), and axis lengths are proportional to the amount of variation in site variables that is explained. The first canonical axes explain 80% and 71% of the variation in site variables in Florida and California, respectively (see Table 2 for significance levels).
Download figure to PowerPoint
Although seedling densities and proportions of species represented by seedlings were higher in rosemary scrub sites (Fig. 3), canonical correlation did not demonstrate a significant relationship between these seedling variables and site variables. Neither of the canonical variates was significant (Table 2), and both site canonical variates combined explained only 41% of the variance in seedling density and proportion of species represented by seedlings.
Table 2. Significance tests of variation in site characteristics explained by canonical axes, and of variation among seedling-related variables (i.e. seedling density, proportion of species represented by seedlings) and site-related variables (i.e. percentage open canopy and soil phosphorus, nitrogen, sand, clay and pH) explained by canonical variates in recently burned sites in Florida and southern California
|Canonical axis or variate||Likelihood ratio||Approximate F||Numerator d.f.||Denominator d.f.||P|
|Variation in site characteristics|
|Variation among seedling and site variables|
|Variation in site characteristics|
|Variation among seedling and site variables|
One significant axis accounted for 71% of the variability in the data and defined a gradient from less open sites with relatively sandy low-pH soils to more open sites with less sandy higher-pH soils (loadings shown in Table 3). Although sites intergraded somewhat, the general order of community types along the gradient was mixed chaparral, coastal sage scrub and chamise chaparral (Fig. 5).
Table 3. Loadings of seedling and site variables from recently burned sites in three southern California communities on canonical variates, and canonical correlations. Seedling-related variables are seedling density (DENSITY) and proportion of species represented by seedlings (PROPSPP); site-related variables are percentage open canopy (OPENCAN), percentage total soil phosphorus (PHOS), percentage total soil nitrogen (NITR), percentage soil sand (SAND), percentage soil clay (CLAY) and soil pH (PH)
|% variance of seedling variables explained||46||32|
Although neither seedling densities nor proportions of species represented by seedlings differed between community types (Fig. 3), canonical correlation suggested a relationship between seedling-related characteristics of sites (i.e. seedling density, proportion of species represented by seedlings) and site-related characteristics (i.e. percentage open canopy and soil phosphorus, nitrogen, sand, clay and pH). The canonical correlation produced two canonical variates, with the first variate significant (Table 2); the second, although only barely significant, nevertheless substantially improved the correlation. The two site canonical variates combined explained 78% of the variance in seedling density and proportion of species represented by seedlings.
Seedling density and the proportion of species represented by seedlings were not correlated with each other (r = –0.17, P > 0.05), and were correlated with different canonical variates. The first statistically significant variate was associated with high values for proportion of species represented by seedlings, and with high soil pH and low phosphorus, while the second was associated with high seedling densities, high soil clay contents and low soil sand and canopy openness values (Table 3).
- Top of page
Although numerous comparisons of post-fire responses between mediterranean-type ecosystems exist (Cody & Mooney 1978; Naveh & Whittaker 1979; Barbour & Minnich 1990), this study is the first to compare responses with other scrub communities (but see Whelan 1985). We predicted that because of differences in climate between California (mediterranean) and Florida (subtropical), seedling establishment patterns would be more similar within regions than between regions. This prediction was borne out for recently burned sites, in which both post-fire seedling densities and proportions of species represented by seedlings were higher in California than in Florida.
In contrast, most previous comparisons among communities have quantified post-fire seedling species richness and/or relative cover of species recruiting by seed (Naveh & Whittaker 1979; Shmida 1981; Barbour & Minnich 1990). Our two parameters gauge the response of the community to fire, because together they approximate a measure analogous to the species’ importance values (Spurr & Barnes 1973) used in other community-level studies. Different processes may operate at different spatial and temporal scales: values for seedling density suggested that they reflected immediate post-fire conditions for germination and establishment, whereas the proportion of species represented by seedlings largely resulted from longer-term evolution of species’ characteristics in response to particular disturbance regimes and climates. Both parameters contribute to the characteristic post-fire ‘flush’ of recruitment (e.g. Barbour & Minnich 1990).
Seedling densities in mature sites may reflect recruitment between fires and so help to characterize communities in different regions. Although means were not statistically different, values were slightly higher and had much higher variances in California than in Florida. This difference suggests that establishment conditions between fires were occasionally favourable in California, but only rarely so in Florida.
Among recently burned sites, California sites had higher seedling densities and proportions of species represented by seedlings than Florida sites, and also had more species dependent upon fire for seedling recruitment (Table 1 and Fig. 3). Florida densities were similar to previous reports (Johnson 1982; Hawkes & Menges 1995; Carrington 1996) but those for chaparral were modest. For example, 90 post-fire sites (including chaparral and coastal sage scrub) averaged 68 seedlings m–2 (Keeley 1997), and higher densities are not uncommon (Keeley 1998).
Greater post-fire recruitment in California communities and Florida rosemary scrub was associated with greater canopy openness, supporting the suggestion that fire-recruiting species are gap-dependent (Keeley & Zedler 1978; Christensen 1985; Malanson & O’Leary 1985). We hypothesize that the observed change in canopy coverage may be the cause of post-fire seedling establishment: the closed canopy in mature chaparral may inhibit recruitment via limited levels of light, water and nutrients, as well as biotic factors such as increased predation and allelopathy (Keeley & Keeley 1989). Post-fire recruitment capitalizes on improved biotic and abiotic conditions, and in California is particularly marked in sites with relatively mesic conditions (e.g. high soil clay contents, higher percentage canopy cover).
Two factors lead to the diminished importance of post-fire seedling recruitment in Florida: gaps in mature vegetation are larger, allowing some inter-fire recruitment (Menges & Kimmich 1996), and rapid resprouting limits the opportunities for seedling establishment, even in burned sites. Rosemary scrub, the only Florida community with seedling densities comparable to chaparral, had, as would be expected from this analysis, the largest fire-induced contrast in percentage open canopy (Fig. 4). Its seemingly harsh, xeric conditions for seedling establishment (i.e. sandy, low-nutrient soil) may, however, reduce competition by slowing regrowth of competitively superior adult plants (Menges & Kohfeldt 1995). The physiognomic and floristic differences between Florida sand pine scrub and sandhill, which have similar soil conditions and low seedling densities, are probably due to different fire regimes (Kalisz & Stone 1984; Myers 1985).
Sorting out selective factors responsible for seedling establishment patterns in California and Florida communities requires careful analysis, because climate, soils and fire histories differ between the two regions. In California, summer drought may limit post-fire cover by reducing survivorship of resprouting species and by delaying resprouting until winter rains, whereas summer rains in Florida may enhance survivorship. Greater canopy coverage in mature chaparral results in more intense fires and thus reduced survivorship (Borchert & Odion 1995), whereas in Florida the more open canopy results in lower fire intensity and increased survival (Hawkes & Menges 1995; Menges & Kimmich 1996).
Florida and California are similar in that most seedling establishment occurs in spring (Carrington 1996; Lambert & Menges 1996; Keeley 1998). In Florida, however, natural fires typically occur in spring and early summer, and seedling establishment is thus delayed until 9–10 months post-fire. Summer rains following the fire season allow resprouts to grow for a full season before seedlings can establish. In California, fires typically occur in autumn and resprout growth is limited initially by drought and later by winter temperatures. Seedlings and resprouts thus start growth at roughly the same time, and opportunities for successful recruitment are therefore greater in chaparral.
Fire season was consistent for all sites within each region, and therefore should not have contributed to differences in seedling recruitment patterns. Fire histories, however, undoubtedly varied among sites, especially in Florida where fire suppression has been practised for most of the last six decades. The extremely long fire return interval, particularly in sand pine scrub sites, probably contributed to low post-fire seedling recruitment (Abrahamson & Abrahamson 1996).
Despite very marked differences in soil texture and fertility (Fig. 4) and in fire history, we believe that seedling recruitment patterns in California and Florida are largely determined by climate. Two other mediterranean climate ecosystems, South African fynbos and Western Australian kwongan, occur on sandy infertile soils comparable to the subtropical Florida scrub sites, but their post-fire seedling recruitment (Kruger et al. 1983) matches that observed at our California sites.
Climatic processes may influence seedling recruitment patterns in California and Florida by differentially modifying levels of competition and/or allelopathy from adult plants (Williamson et al. 1992). Because this study only identified patterns in seedling recruitment and site characteristics that suggest underlying processes, experimental studies are needed to determine causal mechanisms. Empirical field research is particularly needed on the effects of competition and allelopathy on post-fire seedling establishment, and on the influence of fire regime on characteristics of species- and community-level post-fire recruitment. Research in these areas will elucidate further between-region differences in community responses to fire.