Methods
Study Site - Fieldwork was conducted from July 2017 to September 2018 at Ol Pejeta Conservancy (“OPC”; 0°0’52.62”N, 36°51’58.64”E, 1800 m above sea level). This 360 km2 conservancy receives ca. 250-300 mm of rainfall in wet seasons (March to May; October to December, and intervening periods are typically dry and hot with monthly rainfall of ca . 30-50 mm (OPC records). The OPC elephant population (ca . 130-300 depending on forage availability, OPC records) disproportionately imposes heavy damage onA. drepanolobium in areas where P. megacephala has invaded (Riginos et al. 2015). Ground-dwelling P. megacephala ants have expanded from human habitation areas on OPC into black-cotton savannas for the past ca . two decades (Riginos et al.2015), where they occupy trees and soil. During this study, P. megacephala extended each monitored invasion front by ca . 50 m/yr (Pietrek et al. in revision).
Survey Regime - We measured leaf gas exchange (photosynthesis and transpiration) in concurrent Before-After-Control-Impact (BACI) and factorial experiments (Fig. S3) during rainy and dry seasons. For each surveyed tree, we measured leaf water potential at mid-day and before dawn to 1) confirm assumptions that all sites had similar soil water status within each ca. 2-week survey period, 2) to confirm that our designations of “wet” and “dry” seasons were appropriate relative to studies of other East African acacias (Gebrehiwot et al. 2005; Gebrekirstos et al. 2006), and 3) to calculate leaf water potential range, which can be compared with leaf gas exchange rates to indicate changes in water management by the plant.
BACI experiment – To assess short-term impacts of P. megacephala invasion, we measured gas exchange rates and leaf water potential on the same trees before and after invasion, and compared those to concurrent measurements on uninvaded trees that were protected by native C. mimosae . We collected data from trees in plots near the invasion front (“Transition” sites) before and after invasion, and also surveyed non-manipulated trees <1 km from each Transition site (“Control” sites) that remained unaffected by P. megacephala range expansion over the course of the study. All sites were accessible to large herbivores. In the July 2017 dry season and November 2017 wet season, we surveyed 20-24 adult trees (1.5-2 meters tall) at each Transition (pre-invasion) and Control sites.Pheidole megacephala workers expanded into Transition sites in December 2017, and we repeated surveys at each site in the May 2018 wet season and September 2018 dry season. Five trees were destroyed (evidently by elephants) between December 2017 and May 2018 and excluded from analyses.
Factorial experiment comparing long-term (>5 years) impact of invasion In the factorial experiment, we tested direct and indirect effects of invasive P. megacephala , nativeC. mimosae, and vertebrate herbivores on leaf and canopy gas exchange. We measured leaf water potential and gas exchange rates in two dry (July 2017 and September 2018) and two wet seasons (November 2017 and May 2018). Treatment factors were large herbivores (present vs. absent) and ants (present vs. absent), resulting in four treatments (Fig. S3). We conducted our experiment in three sites where acacias had been invaded for ca. 5 years (“Invasion”), and in 3 neighboring (< 2 km away) uninvaded sites with comparable tree density (“Uninvaded” sites). We constructed an electric fence exclosure at each site to exclude large herbivores (>20 kg) from a 50 x 50 m plot (0.25-ha) containing ca . 40 adult trees (1.5-2 meters tall). We marked 40 adult trees (1.5-2 m tall) in a plot of similar area and tree density ca. 200 m from each fenced plot to serve as unfenced controls. Each site comprised two plots, with a total ofca . 80 marked trees at each site. We fogged canopies with 0.6% alpha-cypermethrin (2–3 day half-life in full sunlight; World Health Organization public health specifications for insecticides) and applied sticky barriers (Tanglefoot ® Insect Barrier, Contech Enterprises, Victoria, BC, Canada) to the trunks of 20 trees (e.g., see Stanton and Palmer 2011) in each plot to remove and exclude ants, and reapplied both as needed.
Tree physiological measurements - We conducted all plant physiology measurements on fully-expanded leaves growing from non-lignified shoots in the unshaded sections of the upper canopy. Leaf-level light-saturated photosynthetic and transpiration rates [henceforth “leaf photosynthetic rate” (Amax-leaf ) and “leaf transpiration rate” (Eleaf )] were measured using a LI-6400XT Portable Photosynthesis System (Li-Cor Biosciences, Lincoln, NB) during sunny or partly cloudy days from 07:30-11:30. In long-term Invaded and Uninvaded sites, we also visually estimated canopy leaf area for a random subset of 7-17 acacias per treatment in our factorial experiment in 2018 ( N = 102 in wet season, N = 61 in dry season; means ± SEM in Table S1), and multiplied estimated leaf area by leaf photosynthetic and transpiration rates to estimate idealized light-saturated whole-canopy photosynthesis and transpiration [henceforth “canopy photosynthetic capacity” (Amax-canopy ) and “canopy transpiration capacity” (Ecanopy )].
Amax-leaf andEleaf are extrapolated from gas exchange rates measured in the ideal environment within a controlled cuvette, and likely are higher than net photosynthesis and transpiration of a tree in naturally variable conditions (McGarvey et al. 2004). “Amax is an idealized estimate of photosynthesis and “Eleaf ” is an idealized estimate of transpiration; the suffix leaf refers to the leaf-level rate of gas exchange per unit leaf area, while the suffixcanopy refers to estimates of the idealized canopy gas exchange. We therefore compare these approximations to estimate relative differences in gas exchange for trees in both BACI and factorial experiments, but they do not estimate the absolute effect of invasion on carbon fixation.
We measured pre-dawn (ψPD ) and mid-day leaf water potential (ψMD) on the same day as the gas exchange measurements for each study site using a Model 610 Plant Pressure Chamber (PMS Instruments, Corvallis, OR). Treatment means (± SEM) of ψPD andψMD are in Tables S2 and S3. Wet seasonψPD ranged from ca . -1.0 to -1.5 MPa and dry season ψPD ranged fromca . -1.9 to -2.1 MPa; studies of related tree species in the region recorded ψPD of ca . -2.0 MPa in dry conditions (Gebrekirstos et al. 2006).
For each tree we calculated diurnal leaf water potential range (∆ψleaf ) as the difference between pre-dawn and mid-day leaf water potentials (∆ψleaf = ψPDψMD). ∆ψleafdemonstrates the range of viable water conditions that a leaf will experience (Gebrehiwot et al. 2005; Gebrekirstos et al.2006): that range is fundamentally created by stomatal water loss (Henryet al. 2019) and increased by loss of vascular hydraulic conductivity (Lambers et al. 2008; Scoffoni et al. 2017). Plants will often remain within a species-specific ∆ψleaf (e.g., Gebrekirstos et al.2006), while photosynthesis and transpiration can vary without affecting ∆ψleaf as a result of osmotic or stomatal adjustments (Inoue et al. 2017; Martínez‐Vilalta & Garcia‐Forner 2017; Hochberg et al. 2018; Zhang et al.2019). Note S1 further describes tree physiology methods.
Statistical Analysis - We used generalized linear models (GLMs) to analyze data in the BACI and factorial experiments. For the BACI experiment, we constructed individual GLMs for each season (wet/dry) forAmax-leaf ,Eleaf , and ∆ψleaf . In the BACI GLMs, sampling year (2017, 2018) and site type (Transition, Control) and their interaction term were fixed effects. We specifically report significant interaction terms for the BACI analysis results, to identify differences in leaf physiological traits for trees that were invaded between the 2017 and 2018 surveys, relative to interannual differences for the paired Control trees that experienced similar environmental conditions but no invasion. For the factorial experiment, we constructed separate GLMs for each season (wet/dry) for Amax-canopy ,Ecanopy ,Amax-leaf ,Eleaf , and ∆ψleaf . In the factorial experiment GLMs, ant identity (C. mimosae or P. megacephala ) was a fixed effect, the exclusion of herbivores and ant occupants were fixed effects nested within ant occupant identity, and data from sampling periods were pooled. For both experiments, we also tested models that included site as a random effect, but this term was non-significant for all models and resulted in higher AICc scores, so we removed this term from our final analyses. Analyses were conducted using JMP Pro 14.1.0 (SAS Institute, Cary, North Carolina, USA). Further details on GLMs are in Note S2.