Facilitation and niche theory
Facilitative niche theory proposes that whereas competitive interactions contract an organism’s fundamental niche (inherent environmental requirements) to its realized niche (available environmental requirements not excluded by presence of a superior competitor), the assistance of a mutualist can extend the realized niche to at least the boundaries of the fundamental niche (Bruno, Stachowicz & Bertness 2003). This necessitates that facilitators operate throughout or beyond the fundamental niche requirements of the facilitated species, and where niche requirements overlap, the environmental conditions are unlikely to be optimal for both (Bronstein 1989). The large-scale distribution of myrmecochorous plants and their key dispersers, Aphaenogaster spp., shows a remarkable amount of large-scale overlap in the deciduous forests of the eastern United States (Ness, Morin & Giladi 2009). We show that this facilitative interaction can fail at finer scales, however, and this can be linked to environmental conditions. Specifically, Aphaenogaster spp.-mediated dispersal services approximate zero at higher soil moisture levels (Fig. 1) where H. arifolia adults are highly clumped, indicating multiple generations of surviving seedlings (Fig. 2c, d). Moreover, transplant experiments indicate that H. arifolia adults and seedlings survive and grow where soil moisture exceeds the conditions observed in these seed-dispersal experiments (Fig. S1). For example, ant-mediated seed dispersal decreases to 0 where soil moisture is >30%, yet adult transplant survival and growth are highest in such conditions (Fig. S1a,b) and seed recruitment remains relatively steady across soil moisture gradients ranging from 0 to 51%, although it drops somewhat when soil moisture is >41% (Fig. S1c). By failing to facilitate H. arifolia’s ability to exploit its full range of niche requirements, the limits of a positive biotic interaction contract the fundamental to the realized niche in a pattern similar to the classic contraction driven by competition or predation. Similarly, Kjellberg & Valdeyron (1990) found that species-specific pollination limited rather than enhanced range expansion of Ficus carica due to the niche limits of its pollinator.
Our results do not preclude the possibility that seed dispersal by ants contributes to the expansion of H. arifolia’s realized niche at the lower end of the soil moisture gradient or along other environmental gradients. Nevertheless, ant-mediated dispersal may instead most benefit H. arifolia by alleviating small-scale density-dependent burdens such as competition and disease (Cain, Damman & Muit 1998; Giladi 2006). We find a correlation between increased seed removal by ants and decreased plant aggregation (consistent with decreased clustering of undispersed seeds around reproductive adults), and this may alleviate localized density-dependent effects (Fig. 2). At higher soil moisture, where the mutualism breaks down, H. arifolia seeds remained undispersed and aggregation increases, which is maladaptive (Anderson 1988; Higashi et al. 1989; Boyd 2001; Giladi 2004, 2006).
In the southern Appalachian Mountain region (USA), where our work was performed, ant diversity and myrmecochorous seed removal decreases with elevation, putatively as a function of decreasing temperature (Sanders et al. 2007; Zelikova, Dunn & Sanders 2008). We also found temperature shifts important in explaining dispersal services at the warmest site (MO), but the decreased seed removal with higher temperatures found here is the inverse of that seen at the elevational gradients where seed removal increases with temperature (Sanders et al. 2007; Zelikova, Dunn & Sanders 2008). This may be a function of spatial scale as our investigation occurs along metre transects separated, at most, by 1 km, whereas the elevational research occurred at the scale of kilometres, and thus, more than spatial gradients, these approaches addressed very different temperature gradients. We observed a temperature gradient of approximately 17–20 °C while Sanders et al. (2007) observed a gradient of 12–25 °C. Notwithstanding, precipitation also increases with elevation (and evaporation decreases), and the results reported here suggest that soil moisture also must be investigated as a driver of ant-mediated dispersal services. We found no relation between dispersal services and diffuse light. In addition, although the density of inter- or intraspecific myrmecochores is speculated to affect seed removal rates by ants (Smith, Forman & Boyd 1989b; Smith et al. 1989a), we found no evidence that the abundance or identity of background myrmecochores had any confounding impact on seed removal at either site.
The dominant ant dispersers from our bait stations were from the Aphaenogaster genus, and this is consistent with a growing body of research suggesting a specialization by this genus in association with myrmecochores (Mitchell, Turner & Pearson 2002; Zelikova, Dunn & Sanders 2008; Ness, Morin & Giladi 2009). Research into the impact of soil moisture on ant behaviour is not common, but Wang, Strazanac & Butler (2001a) found increased soil moisture to be a negative predictor of ant diversity, and Levings (1983) suggested that increased soil moisture reduces ant foraging and decreases the suitability of nest sites. Higher soil moisture also causes increased fungal infections in ant colonies (Clark & Prusso 1986; Chen, Hansen & Brown 2002); and soil inundation from periodic flooding appears calamitous for ground-dwelling ant species (Cogni, Freitas & Oliveira 2003; Ballinger, Lake & Mac Nally 2007). However, it has also been shown that many ant species avoid desiccation (Kaspari 1993; Kaspari & Weiser 2000). Because our research targeted H. arifolia, which does not occur in the driest habitats within its range (Warren 2008, 2010), the bait stations may have only captured one end of the ant niche, and the discrepancy between our and other research is simply due to sampling at different points along the gradient (seeVan Horne 2002).
The soil moisture limit on the ants may occur via nesting restrictions, but, at minimum, the results here indicate it curtails foraging. The increased aggregation of H. arifolia plants (Fig. 2d) indicates that myrmecochore propagules occasionally reach the wetter areas of the landscape, possibly by means other than ant dispersal (Fig. 2a), such as gravity and/or overland water flow, and subsequently produce undispersed propagules. An additional possibility is that the plants are dispersed into typically wet habitat during drought conditions. Aphaenogaster ants and their colonies are highly mobile and move their nests – found in soil, folded leaves and below rocks or bark – as often as every 30 days (Smallwood 1982; McGlynn et al. 2004). As such, the ants may migrate into otherwise wet habitats during drought years. Our finding that the negative relationship between seed removal and soil moisture broke down during the extreme MO drought in 2008 (Fig. 1a) supports this hypothesis. The top three seed removal sites in 2008, during the drought, were not the top three the next year when soil moisture increased. Further, at MO, there is no significant relationship between seed removal per individual plot in the drought year (2008) and subsequent non-drought year (2009) (linear regression: coeff. = 0.263, t = 1.199, P = 0.242), whereas at BC, where rainfall was similar between years, seed removal per individual plot in 2008 correlated significantly with seed removal per individual plot in 2009 (linear regression: coeff. = 1.113, t = 4.631, P = 0.001). This suggests that ant-dispersers moved into plots during the drought at MO that they would otherwise avoid in typically wet years.
Species distributions and climate
Considering linkages between conceptual and practical niche ideas, our data are important for evaluating the likelihood that predictive models of plant distribution will yield reliable forecasts. For example, a criticism of species distribution models is that they fail to incorporate biotic interactions (Davis et al. 1998; Araujo & Luoto 2007) – although such models typically make predictions at geographic scales where species are theorized to be more influenced by climate than biotic interactions (Soberon 2007). Indeed, the impact of resource availability on species interactions and the impact of those interactions on populations and communities remain a longstanding theoretical dilemma in ecology. If biotic interactions between species remain constant across resource gradients, mapping the abiotic variables defining the organism’s niche will provide a reasonable approximation of potential distributions. However, the nature and intensity of competitive interactions can vary with resource gradients (Leathwick & Austin 2001; Pearson & Dawson 2003), and we show that the intensity of facilitative interactions between ant-dispersers and H. arifolia vary along soil moisture and, somewhat, temperature gradients (Fig. 1). Studies of facilitative interactions assessing resource enhancement and habitat amelioration suggest that facilitative interaction strengths vary with abiotic conditions (Maestre, Valladares & Reynolds 2003; Freestone 2006; Maestre et al. 2009), and our data on recruitment enhancement suggest this may be a general pattern for facilitative interactions.
If the strength of facilitative interactions changes in a systematic way with an environmental gradient, the inclusion of the environmental parameter as a surrogate for the biotic interactions in predictive models (without explicitly modelling the interactions) should not severely affect the model efficacy. Problems will arise, however, if the facilitative relationships can not be generalized or are not transferable. The similarity in dispersal services as a function of soil moisture between two study sites (MO and BC) located 100 km apart and in different biomes is encouraging for the use of soil moisture as a proxy for the dispersal interaction; however, the increase in dispersal services from 2008 to 2009 in the outlier sites at MO is quite disproportional to the change in soil moisture across years (Fig. 1). This suggests that ant dispersal services may shift in an unpredictable manner during drought years and this makes the omission of this biotic interaction problematic in predicting potential distributions based on current patterns. The strength of ant seed-dispersal services varies with geographic and environmental gradients (Mitchell, Turner & Pearson 2002; Giladi 2004; Kersch & Fonseca 2005; Zelikova, Dunn & Sanders 2008), and reliable niche models for myrmecochores will need to capture ant as well as plant responses to changing climate. This may be especially true given the asymmetry of the mutualism (i.e. the ant is not particularly reliant on the plant for food, yet the plant is predominantly reliant on the ant for dispersal (Kersch & Fonseca 2005; Zelikova, Dunn & Sanders 2008; Ness, Morin & Giladi 2009).