Skip to main content
Dryad logo

Climate seasonality drives ant-plant-herbivore interactions via plant phenology in an extrafloral nectary-bearing plant community

Citation

Calixto, Eduardo et al. (2020), Climate seasonality drives ant-plant-herbivore interactions via plant phenology in an extrafloral nectary-bearing plant community, Dryad, Dataset, https://doi.org/10.5061/dryad.sbcc2fr45

Abstract

  1. Interactions between ants and plants bearing extrafloral nectaries (EFNs) are among the most common mutualisms in Neotropical regions. Plants secrete extrafloral nectar, a carbohydrate-rich food that attracts ants, which in return protect plants against herbivores. This ant-plant mutualism is subjected to temporal variation, in which abiotic factors can drive the establishment and frequency of such mutualistic interaction. However, studies investigating how abiotic factors (e.g., climate) directly and indirectly influence ant-plant-herbivore interactions are incipient.
  2. In this study, we investigated direct and indirect (via plant phenology) effects of temperature and rainfall on ant-plant-herbivore interactions. To address these goals, we estimated six plant phenophases (newly flushed leaves, fully-expanded leaves, deciduousness, floral buds, flowers, and fruits) monthly, the activity of EFNs and abundance of ants and herbivores in 18 EFN-bearing plant species growing in a markedly seasonal region (the Brazilian Cerrado) during a complete growing season.
  3. Our results showed that (i) there were marked seasonal patterns in all plant phenophases, EFN activity, and the abundance of ants and herbivores; (ii) the peak of EFN activity and ant and herbivore abundance simultaneously occurred at the beginning of the rainy season, when new leaves flushed; and (iii) rainfall directly and indirectly (via changes in theproduction of new leaves) influenced EFN activity and this in turn provoked changes in ant abundance (but not on herbivores).
  4. Synthesis: Overall, our results build toward a better understanding of how climate drives seasonal patterns in ant-plant-herbivore interactions, explicitly considering plant phenology over time.

Methods

Study area

We performed this study in the Ecological Reserve of the Clube de Caça e Pesca Itororó de Uberlândia (CCPIU: 48º18’01”W; 18º59’05”S, Uberlândia, Brazil). The study area had ~200 ha of cerrado sensu stricto vegetation (Oliveira & Marquis, 2002). Open areas consisted of shrubs and small trees, whereas more enclosed areas consisted of trees reaching up to 15 meters in height (Del-Claro, Rodriguez-Morales, Calixto, Martins, & Torezan-Silingardi, 2019). The climate is markedly seasonal with the rainy season from October to March, and the dry season from April to September. The annual mean temperature varies from 18 to 28 ºC and the rainfall from 800 to 2,000 mm (Alvares, Stape, Sentelhas, De Moraes Gonçalves, & Sparovek, 2013).

Data collection

For this study, we systematically established ten 50 m transects at least 50 m apart from each other. Transects started perpendicular to a reserve trail (2.5 m wide) going through the vegetation. In order to diminish edge effects, we set transects separated at least 10 m from the trail. All transects were parallel to each other and soil characteristics, light availability and humidity of each transect were similar.

We selected the 18 most common and abundant plant species (Appolinario & Schiavini, 2002; Lange, Calixto, & Del-Claro, 2017; Lange & Del-Claro, 2014) bearing EFNs in our field site: Banisteriopsis malifolia (Nees & Mart.) B. Gates (Malpighiaceae), Banisteriopsis stellaris (Griseb.) B. Gates (Malpighiaceae), Bauhinia rufa (Bong.) Steud. (Fabaceae), Bionia coriacea (Nees & Mart.) Benth. (Fabaceae), Caryocar brasiliense Cambess. (Caryocaraceae), Eriotheca gracilipes (K.Schum.) A.Robyns (Malvaceae), Heteropterys pteropetala A. Juss. (Malpighiaceae), Lafoensia pacari A.St.-Hil. (Lythraceae), Manihot caerulescens Pohl (Euphorbiaceae), Ouratea hexasperma (A.St.-Hil.) Baill.(Ochnaceae), Ouratea spectabilis (Mart.) Engl. (Ochnaceae), Qualea grandiflora Mart. (Vochysiaceae), Qualea multiflora Mart. (Vochysiaceae), Qualea parviflora Mart. (Vochysiaceae), Senna rugosa (G.Don) H.S.Irwin & Barneby (Fabaceae),Senna velutina (Vogel) H.S.Irwin & Barneby (Fabaceae), Smilax polyantha Griseb. (Smilacaceae), and Stryphnodendron polyphyllum Mart. (Fabaceae). In total, we tagged 180 plant individuals equally divided among the 10 transects. Each transect included one individual of each species. We selected these individuals based on similar phenological characteristics within each species (e.g., total height and number of branches).

Each month, we collected data of plant phenology for each tagged individual of all species from March 2017 to February 2018. In particular, we evaluated six phenophases: newly flushed leaves, fully-expanded leaves, deciduousness, floral buds, flowers, and fruits. For each phenophase, we used the following semi-quantitative scale: 0 = absence of plant phenological event; 1 = sporadic event (1-25% of the branches); 2 = infrequent event (26-50% of the branches); 3 = frequent event (51-75% of the branches); and 4 = very frequent event (76-100% of branches) (Fournier, 1974). For each plant, we also evaluated EFN activity by estimating the number of plants with active EFN monthly. We considered that EFNs were active when they were bright and light in colour, and generally producing extrafloral nectar. By contrast, we considered that EFNs were inactive when they were dark with a necrotic aspect and did not produce nectar (Calixto, Lange, & Del-Claro, 2015). The main reason to choose EFN activity instead other EFN characteristics (e.g., volume, concentration) was that the circular statistics used (see details in Statistical analyses) only work with presence or absence of specific event over time. Finally, to quantify the number of ants and herbivores we conducted field observations on two consecutive days per month during the same period as the plant phenology observations. Each day we censused half of transects (five transects) in both daytime (8:00 a.m. to 12:00 p.m.) and nighttime (7:00 p.m. to 11:00 p.m.). In each plant individual, we first collected ants and other insects without touching the plant. Next, using a white inverted umbrella placed under the plant, we shook the branches and collected those insects who fell on the umbrella. We subsequently stored all insects in 70% ethanol. Then we identified insect herbivores by using insect field guides. At night, we used flashlights covered with red cellophane to lessen the deterrence of insects.

Climatic data

We obtained climatic data (namely daily averages of rainfall, humidity, and temperature) of our fieldwork season from Climate Station of the Instituto de Geografia da Universidade Federal de Uberlândia. For statistical analyses, we calculated mean humidity and temperature, and total rainfall for each month.

Statistical analyses

We performed all statistical analyses in R software 4.0.0 (R Core Team, 2020). Packages used are detailed in each analysis below.

First, we investigated whether there was seasonal variation (i.e., peaks in activity or abundance) in plant phenophases (newly flushed leaves, fully-expanded leaves, deciduousness, floral buds, flowers, and fruits), EFN activity, ant and herbivore abundance throughout the year. For this, we used circular statistical analysis (Morellato, Alberti, & Hudson, 2010; Novaes et al., 2020). Circular statistics evaluate recurring biological events associated to climatic factors, where specific units of time (e.g., months) are converted in angles (Morellato et al., 2010; Novaes et al., 2020). Basically, it uses the frequency or abundance of a specific event over time, identifying peaks of activity (events concentrated in specific times). We converted months into angles, and used the abundance of each factor to calculate the mean vector (µ), mean vector length (r), median, standard deviation, Rayleigh test Z, and Rayleigh test p. Rayleigh test Z with significant values (p <0.05) and mean vectors (r) close to 1 indicate seasonality of the data, that is, phenological activity was concentrated around one single date or mean angle (Morellato et al., 2010; Novaes et al., 2020). We conducted circular analyses with the package ‘circular’ (Agostinelli & Lund, 2017). We performed the Watson’s goodness of fit to examine unimodality before conducting circular analyses. In the case of EFN activity and abundance of ants and herbivores, we used the total number of plants with active EFNs, ants and herbivores per month for analysis at community level.

Second, we investigated whether the peak in the activity of EFNs, and the abundance of ants and herbivores simultaneously occurred in the same period and when plants produce new leaves. To corroborate our hypothesis, the peak (mean vector) of EFNs, ants and herbivores should occur at the same time (or shortly after the appearance of leaf flushing), always before any other phenophase, and should not overlap with other phenophase mean vector. Since there is a sequence of plant phenological events over time, we expected that EFNs peak when newly flushed leaves appear, and not during other plant phenological events. For this, we used two statistical approaches: circular statistics and overlap analysis. With these two statistical approaches, we were able to test for differences between peak activity of plant phenological events (e.g., EFN activity and newly flushed leaves) and also to analyze the overlap of events occurrence over time. We firstly carried out Watson’s two-sample test of homogeneity to test whether peaks were different, and compared the mean vector in the circular diagrams of EFNs activity with newly flushed leaves, fully-expanded leaves, deciduousness, floral buds, flowers, fruits, ants and herbivores. We secondly conducted niche overlap analyses by using temporal abundance matrices, which were analysed with the coefficient of overlapping (Ridout & Linkie, 2009) and niche overlap null models. The coefficient of overlapping uses Kernel density estimates fitted to the data to approximate two density functions, that is, it represents quantitative measure of overlap (0 = no overlap, 1 = identical peak). We used the estimator Dhat1, since it is the best option for small samples sizes (Ridout & Linkie, 2009). In the case of niche overlap null model, we used two randomization algorithms (1000 simulated null assemblies), RA3 and RA4, and the Pianka’s metric with the package “EcoSimR” (Gotelli, Hart, & Ellison, 2015). RA3 rearranges the rows and RA4 rearranges the non-zero row values; both retain “niche breadth” of each species.

Finally, we investigated whether climatic variables directly and indirectly influenced the ant-plant-herbivore interaction. For this, we carried out a Structural Equation Modelling (SEM) based on monthly data at plant individual level using ‘lavaan’ package (Rosseel, 2012). SEM is a multivariate statistical analysis that estimate direct and indirect relationships among variables. Specifically, we investigated direct associations among climatic factors (temperature and rainfall), newly flushed leaves, EFN activity and the abundance of ants and herbivores, as well as the indirect associations between climate and the abundance of ants (via newly flushed leaves and EFN activity) and herbivores (via newly flushed leaves, EFN activity and ants). We only chose newly flushed leaves as plant phenophase for these analyses because their peak simultaneously occurred with the peak in ant and herbivore abundance (see Results). The hierarchical order of the variables included in the SEM (i.e., EFNs > Ants > Herbivores) was chosen because EFNs attract ants and these in turn prey on herbivores. We estimated direct effects in the SEM as standardized partial regression coefficients, whereas we obtained indirect effects by combining the specified coefficients for direct effects on both the predictor and the response. We assessed the significance of direct and indirect coefficients with z-tests. We evaluated the goodness of fit of the general model using the Chi-square statistic, Comparative Fit Index (CFI), the Root Mean Square Error of Approximation (RMSEA), and the Standardized Root Mean Square Residual(SRMR).

Funding

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior, Award: Finance code - 001