Discussion
Longer-term (ca . 5 year) invasion by P. megacephalareduced carbon fixation in A. drepanolobium by 69% during the growing (wet) season, consistent with other studies demonstrating that invasive species can alter fundamental ecosystem processes (a review of ecosystem functions affected by invasions: Hooper et al. 2005; invasive insects reducing hardwood forest productivity: Kurz et al. 2008; Albani et al. 2010; Clark et al. 2010; disruption of pollination by invasive insects: Morales et al.2017). The combination of shorter-term small or neutral effects of recent invasion on leaf-level photosynthesis with longer-term large reductions in canopy photosynthesis for trees that had been invaded for 5 years highlight that some ecologically relevant effects of invasive species can lag behind their initial appearance in a habitat (Crooks 2005; Strayer et al. 2006; Simberloff 2011). Finally, the large positive effect of vertebrate herbivore exclusion on canopy photosynthesis in invaded habitats demonstrates how invasive insects can indirectly interact with native fauna, including vertebrates, to substantially change carbon fixation for foundational plants.
Our results from Transition sites suggest that P. megacephalatriggers minimal (if any) benefits for newly invaded trees, thus indicating that the loss of native ant mutualists has little direct effect on leaf-level photosynthesis in the short term. Wet season leaf-level photosynthesis and transpiration were nearly identical immediately before and after invasion for Transition trees, while the significant decline for both variables over the same period for Control trees may simply be driven by a decline in water availability for Control trees between the 2017 and 2018 wet seasons. Recently invaded trees experienced declines in dry-season leaf-level photosynthesis that were almost twice as large as declines for Control trees, perhaps because mixed feeders that forage on woody species during the dry season (Illius & O’Connor 1999) began to target their large and unprotected leafy canopies. The recent extirpation of C. mimosae by P. megacephala does not appear to cause increases in leaf-level photosynthesis, addressing our first research question about how the recent removal of native ant mutualists affects leaf photosynthetic rate. Crematogaster mimosae imposes a low net cost for the host tree compared to other ant species like C. nigriceps andC. sjostedti (Stanton & Palmer 2011), which may explain why their extirpation has no immediate effect on leaf physiology. Instead, our BACI experiment suggests that recently invaded trees become targets of herbivory during dry seasons, which may rapidly reduce leaf-level photosynthesis.
The longer-term effects of P. megacephala invasion on wet season carbon fixation is likely to affect how surviving invaded A. drepanolobium produce biomass and interact with other savanna species. In savannas that have been invaded for > 5 years, a 2-meter-tall tree has a canopy photosynthetic capacity of onlyca . 31% of the carbon fixation per hour as a comparable uninvaded tree during the wet season, the primary growing period for many African acacias (Gourlay 1995). While ca . 20% of longer-term invaded trees at OPC experience complete destruction by large herbivores, ca . 60% experience mild to moderate levels of canopy damage (Riginos et al . 2015): our findings apply to this large majority of trees that survive for years after invasion. Reduced canopy-level photosynthesis can limit the availability of dissolved sucrose in aboveground tissue for A. drepanolobium (P. Milligan, unpublished manuscript), which is an important plant resource for growth (e.g., Wiley & Helliker 2012) and response to biotic and abiotic stress (e.g., Hummel et al. 2010; Engelsdorf et al. 2013). Low carbon availability can reduce population growth (McDowell 2011; Wiley & Helliker 2012), and likely contributes to A. drepanolobiumpopulation declines in longer-term invaded areas predicted by empirically derived demographic models of OPC trees (B. Hays, unpublished manuscript). Carbon limitation can prevent trees from producing aboveground biomass (Allen et al. 2010; McDowell 2011), which would reduce the influence of A. drepanolobium on non-woody plant growth in its understory (Riginos et al. 2009) and on the visibility of predators and prey in black cotton savannas dominated byA. drepanolobium (Riginos 2015). Also, carbon limited trees likely have fewer carbohydrates to support endosymbionts, which could affect nitrogen-fixing bacteria in A. drepanolobium roots that input N into black cotton soil (Fox-Dobbs et al. 2010).
The interactive effect of invasion and vertebrate herbivory is likely the primary driver of leaf- and canopy-level photosynthesis and transpiration decline for longer-term invaded trees. The displacement of aggressive native C. mimosae by P. megacephala results in more intense browsing of A. drepanolobium by elephants (Riginoset al. 2015), which suppresses canopy size for A. drepanolobium and for other savanna trees (Pellew 1983; Biggs & Jacobs 2002; Goheen & Palmer 2010), and thus reduces canopy-level photosynthesis. At the leaf-level, P. megacephala may function similarly to the native C. sjostedti , which provides negligible defense (Martins 2010) and is associated with low leaf-level photosynthesis when large herbivores are present (King & Caylor 2010). directly result from damage by herbivores (e.g., Delaney et al.2008) or indirectly result from the tree’s production of antiherbivory compounds in leaves as a response to damage (in this system: Ward & Young 2002; across plants: Ishida et al. 2008; Sumbele et al. 2012; Keenan & Niinemets 2016, see Note S5). While we did not observe reductions in leaf- or canopy-level photosynthesis on C. mimosae -removal trees in uninvaded areas, those trees may have benefitted from “associational defense” (Barbosa et al. 2009) with neighboring trees that were occupied by defensive native ants (see Note S5).
Compared to vertebrate herbivores, invasive P. megacephala had smaller but significant direct negative effects on A. drepanolobium carbon fixation. This may be driven by P. megacephala ’s habit of nesting around tree roots (P. Milligan, pers. observation), which can have large negative impacts on carbon fixation and leafy growth for A. drepanolobium saplings (Milligan et al . unpublished manuscript). Pheidole megacephala may also affect tree physiology through facultative interactions with other insect pests in the canopy, which were present in both ant-present and ant-excluded tree canopies >8 months after canopies were initially fogged with insecticide (P. Milligan, pers. observation). While we did not explicitly examine these relationships here, P. megacephala tends lycaenid caterpillars (pers. observation T. Palmer; a widespread association in Australia, Eastwood & Fraser 1999) and tolerates phloem-feeding cerambycid larvae (pers. observation P. Milligan) in the canopy. Infestations of phloem-feeding insects can affect the photosynthetic rates and water use efficiency of many plant species (Cockfield et al. 1987; Meyer & Whitlow 1992; Haaviket al. 2008; Golan et al. 2015), which may further limit tree above- and belowground productivity (Wiley & Helliker 2012).
While our finding of strong interactive effects of an invasive ant and vertebrate herbivores on carbon fixation of a dominant plant is novel, there are some parallels to be found with other systems. Herbivores are often more selective for high quality forage, such as the highly nutritious leaves produced by A. drepanolobium (Rubanza et al. 2007), in other savanna systems (e.g., Roques et al. 2001; Kos et al. 2012; Abraham et al. 2019). Thus, plants likeA. drepanolobium with highly nutritious leaves may experience large changes in herbivore pressure if their deterrents against herbivory are disrupted by invasive insects. Chronic herbivory often imposes cumulative aboveground growth costs for plants in other systems (e.g., Kozlov & Zvereva 2017; Wilson et al. 2018), and those increased metabolic costs can negatively affect leaf- and canopy-level photosynthesis (Wiley & Helliker 2012). McDowell (2011) reviewed how invasive insects contribute to carbon limitation and tree mortality during periods of environmental stress, but it is unknown whether interactive effects between invasive insects and native herbivores may be a more general phenomenon in terrestrial ecosystems.
The longer-term loss of photosynthetic capacity for ant-plants in invaded habitat reduces A. drepanolobium ’s carbohydrate pool, which may affect other ecosystem processes to which this foundational tree contributes. For example, A. drepanolobium imports nitrogen into N-limited black cotton vertisols via N-fixing root endosymbionts (Fox-Dobbs et al. 2010). If host plants reduce resource allocation to roots in invaded habitats, this could in turn reduce both N-fixing symbiont activity and soil respiration (seen in a large-scale girdling experiment in boreal forests, Högberg et al. 2001). Potentially compounding this effect, elephants may reduce tree cover within invaded savannas over the longer term, further reducing N inputs into the system, as has been shown in other areas of East Africa where removal of acacia species reduces both the total content and mineralization of C and N in soils (Glaser et al. 2001). Finally,Acacia drepanolobium has density-dependent effects on the productivity of understory plants (Riginos et al. 2009), and thus the resource availability and productivity of invaded trees may also be linked to understory productivity. By increasing the mortality (Riginoset al. 2015) and decreasing the carbon fixation of this monodominant tree species, P. megacephala invasion may fundamentally alter carbon cycling and connected ecosystem processes in these savannas.
Acknowledgements: We thank the Kenyan government (NACOSTI/P/18/4376/9459) for their permission to conduct this work. Gabriella Mizell, Nelly Maiyo, Jackson Ekadeli, Gilbert Buseinei, Isaac Kipkoech, and John Mosiany provided excellent field assistance. Elizabeth Pringle provided helpful comments during manuscript preparation, and three anonymous reviewers provided valuable feedback. Mpala Research Centre administration and the Ol Pejeta Conservancy management team (particularly Samuel Mutisya and Benard Gituku) provided substantial logistical support. This research was supported by a University of Florida International Center RADS grant to PDM, a Smithsonian ForestGeo grant to PDM, a National Geographic Society Young Explorer grant to PDM, and a National Science Foundation grant (NSF DEB 1556905) to TMP, CR and JRG.