4 DISCUSSION
Anthropocene fire regimes are catalyzing major shifts in plant community structure, function, and diversity across the world (Bowman, 2020). We detected changes in woody plant species diversity in Chiricahua National Monument from 2002-2003 to 2017-2018, a time span intersected by the Horseshoe Two Megafire of 2011 and characterized by increasing aridification throughout the southwestern United States. The key results were that gamma species richness (γ) increased by nearly 1/3, alpha richness (α) declined by 1/5, and beta diversity (β) increased nearly two-fold. As a result, the contributions of α and β to γ reversed their order from the first sample (higher for α) to the second (higher for β).
Several lines of evidence tie these temporal changes directly to the Horseshoe Two Fire for all three scales of species diversity. γ, α species richness, Inverse Simpson, total β, and β turnover all shifted from before to after the fire in burned but not unburned plots. The increase in γ, a common result for ecosystems with fire-adapted species (Perry et al., 2011; Romme et al., 2016; Pausas & Ribiero, 2017; He et al., 2019), is not surprising for Chiricahua National Monument, given that wildfires have been largely excluded for more than a century in vegetation that historically experienced frequent fire. In our 138 plots, only three species disappeared, while 13 newcomers appeared—primarily cacti indicative of drier conditions and shrub species that thrive in exposed areas, both of which became more prevalent in the wake of the wildfire. The negative impact of the Horseshoe Two Fire on α species richness differs from most studies of plants, which have usually detected positive effects of fire (e.g., Burkle et al., 2015 [forbs and graminoids]; He et al., 2019; Miller & Safford, 2020; but see Collins et al., 2007; Burkle et al., 2015 [woody plants]). The significant increase in the Inverse Simpson Index aligns more closely with the bulk of those studies, but that result is difficult to interpret given that this index incorporates not just richness but also evenness, which increased in both burned and unburned plots (i.e., independent of the fire).
The generally positive effect of the Horseshoe Two Fire on β is similar to results from studies across a wide range of taxa (Myers et al., 2015; He et al., 2019; Miller & Safford, 2020), although neutral or negative outcomes are common as well (Reilly et al., 2006; He et al., 2019; Richter et al., 2019; Miller & Safford, 2020). β in our study stemmed chiefly from species turnover rather than nestedness, as did significant spatial and temporal patterns. In other words, along fire and topographic gradients, as well as across the two sample periods, woody plant communities tended to exhibit wholesale replacement of species (turnover) rather than shifts in which smaller local assemblages were subsets of larger ones (nestedness). The role of species turnover across spatial gradients accords well with the famously high level of terrain ruggedness and tightly packed communities of Chiricahua National Monument (Poulos et al., 2007) and throughout the Sky Islands (Whittaker et al., 1968 Niering & Lowe, 1984; Barton, 1994; Coblentz & Ritters, 2004; Villarreal et al., 2019). That the Horseshoe Two Fire depressed α species richness but amplified β suggests that fire did not act uniformly in decreasing local richness, but instead filtered out species differentially among plots, promoting variation in post-fire assemblages, nearly entirely through the enhancement of species turnover.
The intermediate disturbance hypothesis (IDH) in the context of fire has focused largely on α species richness (Connell, 1978; Sousa, 1979; Huston, 2014), but has also been proposed for β and γ (He et al., 2019). We found support for the IDH for fire severity but only for γ, which peaked at low severity. This pattern appears to have stemmed from both higher species extirpation and lower establishment of novel species associated with higher severity fire, probably a result of the extreme conditions in those plots during and after the Horseshoe Two Fire. Outcomes vary widely regarding γ and the IDH, with Richter et al. (2019) supporting the IDH for understory plant communities in the Sierra Nevada, some studies finding no relationship (Miller & Safford, 2020), and others arguing that the positive effect of fire on landscape-scale plant diversity generally increases linearly along fire gradients (Pausas & Ribiero, 2017; He et al., 2019).
In contrast to γ, neither α nor β exhibited unambiguous support for the IDH. β increased for all fire severity classes, although the continuous relationship with dNBR exhibited a slight decline at very high severity. Figure 5 appears to reveal a hump-shaped relationship between α and fire severity, but the same pattern occurs for the results for before the fire. These parallel relationships stem not from any impact of fire, of course, but instead from the sharp decline in fire severity with elevation (r=0.43) and an underlying hump-shaped relationship of diversity with elevation. Only by examining changes in diversity from pre- to post-fire in burned vs. unburned plots were we able to establish that alpha diversity and fire severity exhibited no clear relationship. This pattern demonstrates the importance of coupling pre- and post-fire sampling, but also reveals a weakness of our study: the lack of experimental treatments in which factors such as fire severity and elevation can be investigated independently.
Although the IDH has been rarely tested for β (see Burkle et al., 2015; Richter et al., 2019 for mixed results), many studies support the proposition for α (DeSiervo et al., 2015; Morgan et al., 2015; Stevens et al., 2015 Heydari et al., 2017; He et al., 2019; Richter et al., 2019; Strand et al., 2019; Miller & Safford, 2020; but see Schwilk et al., 1997). Three factors might explain the negative impact of fire on α and lack of support for the IDH in our results. First, Huston (1979, 2014; see also Burkle et al., 2015) argued on theoretical grounds that a hump-shaped diversity-disturbance relationship should occur primarily in more productive sites, where disturbance alleviates interspecific competition and promotes species co-existence, rather than in unproductive areas, such as the dry, hot sites in our study area, where interspecific competition may be less pronounced to begin with. The few tests of this IDH-productivity hypothesis for fire have produced mixed results (Burkle et al., 2015 Strand et al., 2019; Miller & Safford, 2020). Second, more than a century of fire exclusion in Chiricahua National Monument may have reduced the population sizes of less common, fire-associated species, resulting in lower and more variable seed rain of these species at the local scale after the fire. Moreover, regeneration of these species after the Horseshoe Two Fire, especially from seed, might have been constrained by the long-term drought that started in the 1990s and was especially extreme post-fire in 2011 (Williams et al., 2014). These three scenarios together could translate into the input of relatively few seeds with low establishment probabilities, leading to deficient replacement of species extirpated from local assemblages by the Horseshoe Two Fire, and a lack of positive response to the reduction of local competition assumed by the IDH.
We found mixed support for the hypothesis that pyrodiversity promotes β, a relationship confirmed in many studies of plants (Perry et al, 2011; Burkle et al., 2015; Myers et al., 2015 Heydari et al., 2017; Freeman et al., 2019; He et al., 2019)—but not all (Reilly et al., 2006; Masunga et al., 2013). While our results do not explain whether variation in fire severity led to higher β because of differential fire-induced mortality or post-fire regeneration, the number of novel species after the fire far outstripped the number of extirpated species, suggesting a role for newfound species regeneration in the wake of fire. The decline in α species richness, however, points to the potential importance of differential species mortality across plots. Plots subject to high fire severity, in particular, were markedly transformed by fire-induced mortality of nearly all stems, massive resprouting of oaks and shrubs, and minimal regeneration of fire-resistant conifers that rely on establishment by seed. These plots diverged significantly from those experiencing lower fire severities. In Cave Creek Canyon, on the east side of the Chiricahua Mountains, Barton and Poulos (2018; see also Minor et al. 2017) documented such conversion of structurally complex Madrean pine-oak forests to sprouting shrublands after high severity fire—a pattern found increasingly across the Southwest (Falk, 2013; Coop et al., 2020). The same process appears in part to have significantly shifted species diversity patterns in Chiricahua National Monument after the Horseshoe Two Fire.
A recent review by Miller & Safford (2020) demonstrated that both α species richness and β are generally highest at fire severities typical of the historic fire regimes of plant communities rather than strictly following the IDH. Ecosystems characterized historically by frequent, surface fire tended to exhibit the highest plant diversity at low to moderate fire severity, whereas diversity for areas typically experiencing infrequent, stand replacing fires often peaked at higher fire severity. Ecological filtering for compatibility of species to fire regime is the assumed driving force underlying these patterns. In our study, pine-oak forest was historically characterized by frequent, low severity surface fires (Swetnam et al., 1989; Swetnam et al. 1996; Barton et al., 2001), whereas juniper and piñon woodlands experienced a mixed-severity fire regime (Baisan & Morino, 2000). If historic fire regime shaped the responses of species diversity to fire, we would thus expect divergent responses to fire between pine-oak forest and the other two vegetation types. The results do not support this prediction: the three vegetation types exhibited similar decreases in α species richness and increases in total β over time when burned plots were analyzed separately (P <0.01; data not shown). Moreover, species richness and pre-/post-fire dissimilarities showed significant continuous relationship with dNBR only for pine-oak forest, for which β peaked at nearly the highest fire severity, contrary to what would have been predicted given its historic low severity fire regime.
At least two factors might have mitigated the role of historic fire regime in the responses of these vegetation types. First, although Miller & Safford (2020) cite evidence for the homogenizing effect of high severity fire, a century without fire may well have had the same impact on the vegetation of Chiricahua National Monument, especially pine-oak forest, which was transformed over a century of fire exclusion from an open woodland to a much denser, even light-limited, forest. It is not surprising that fire of any severity may have injected heterogeneity onto this vegetative canvas, promoting β. Second, the species pool governing pine-oak forest includes typical fire-resistant species that regenerate only from seed (e.g., Pseudotsuga menziesii and P. arizonica ), and thus, as assumed by Miller & Safford (2020), respond poorly to high severity fires (Barton, 2002; Barton & Poulos, 2018). Other species common in that vegetation type, however, exhibit multiple traits, such as thick bark, resprouting capacity, and serotinous cones (e.g., P. leiophylla ), that are adaptive in the context of multiple fire severities. Poulos et al. (2018) argued that such multiple strategies might indicate a complex evolutionary history with respect to the fire regimes experienced by these species. An important assumption of the historic fire regime hypothesis that bears further inspection, therefore, is the extent to which species traits are the legacy of recent historic fire regime versus a longer, more complex evolutionary heritage.
The evenness component of α and the nestedness component of β exhibited temporal changes that were independent of fire, that is, these metrics changed in both burned and unburned plots over the 15-year period. While this study documents the effects of wildfire on woody plant diversity in Chiricahua National Monument, long-term drought stress may also be influencing contemporary woody plant diversity dynamics at this site. The region has experienced severe moisture deficits since the 1990s (Cook et al, 2014; Cook et al., 2015; Ault et al. 2016). Restricted soil moisture and pronounced vapor pressure deficit have caused a wide range of recent ecological changes in the region, including larger, more intense wildfires (Abatzoglou & Kolden, 2013; Abatzoglou & Williams, 2016; Singleton et al, 2016; Williams et al., 2019), pronounced tree mortality (Allen et al., 2010; Williams et al., 2013), and shifts in community composition (Falk, 2013; Coop et al., 2020). We cannot rule out the possibility that moisture stress amplified fire-associated impacts even in burned plots (Barton and Poulos, 2018 Poulos et al., 2020; Poulos et al., in press). The diversity effects of fire comprise the sum of mortality imposed by burning and post-fire plant regeneration from seed and resprouting. Intensified moisture stress acting especially on post-fire seed germination and establishment may well have acted synergistically with fire in shaping the diversity patterns described in this paper, an argument made also for changes in species composition after the Horseshoe Two Fire (Barton & Poulos, 2018; Poulos et al., 2020). The possible mechanisms connecting fire, drought, and diversity are unclear at this point and deserve further investigation.
Independent of time and fire, topography strongly influenced species diversity in Chiricahua National Monument. In fact, the relationships of diversity indices with topographic variables changed little from before to after the fire, suggesting that topography is an intrinsic regulator of diversity at this site, regardless of the effects of fire. Elevation is a complex master environmental variable controlling the structure, composition, and processes of Sky Island ecosystems (Shreve, 1915; Marshall, 1957; Whittaker & Niering, 1975; Sawyer & Kinraide, 1980; Niering & Lowe, 1984; Barton, 1994; Poulos et al., 2007, 2010). From lower to higher elevation, temperature decreases and moisture increases; other key environmental variables change as well along this gradient (Shreve, 1915; Whittaker et al., 1968; Barton, 1994; Vivoni et al., 2007). α peaked at intermediate elevations and exhibited sharp reductions towards both lower and higher elevation in our study, which is similar to past studies of Chiricahua National Monument (Poulos et al., 2007) and other Sky Island ranges (Whittaker & Niering, 1975). In contrast, β decreased with elevation, suggesting higher levels of habitat heterogeneity at lower elevations. All three α metrics increased with increasing terrain ruggedness (TRI), which measures the degree of topographic complexity at the plot scale. Higher levels of α in more rugged plots likely arises from increased microhabitat heterogeneity and favorable conditions for a wider array of species than in more homogeneous terrain. Coblentz and Riitters (2004) found a similar relationship at the regional scale, attributing the pronounced biodiversity of the Sky Island ranges of the Southwest USA and northern Mexico to the physical complexity of the mountains (see also Felger & Wilson, 1995)