Comparison of δ13C and δ15N in ant species at the drought and control plots
The 18 ant species sampled covered a broad range of δ15N and δ13C values. Both δ15N and δ13C varied significantly among species at the control plot (ANOVA, δ15Ncontrol, F17,61=5.817, P < 0.001; δ13Ccontrol, F17,61=5.817, P < 0.001) as well as the drought plot (ANOVA, δ15Ndrought, F17,62=3.212, P < 0.001; δ13Cdrought, F17,62=9.832, P < 0.001) (Fig. 1). Species mean isotope signatures differed by as much as 5.8‰ in δ15N and 3.8‰ in δ13C within either the control or drought sites, suggesting that the diet of the ant community ranges across two trophic levels within each plot.
We found a significant effect of the drought treatment on ant δ15N values (paired t-test, t(17)= -4.46, p < 0.001) with δ15N being higher at the drought plot (mean ± SD: 6.52 ± 1.11) compared to the control plot (mean ± SD: 5.57 ± 1.74). Fourteen species had higher δ15N values under drought conditions (Fig. 1), and the ant species that demonstrated the greater differences in δ15N between drought and control, were those species that had low δ15N values under control conditions (Spearman correlation, rs=-0.81, p < 0.001). The effect of the drought treatment on mean δ13C signatures was more mixed, and overall no significant difference was found between plots (paired t-test, t(17) = 0.12, p = 0.91). Overall, we found those trends in δ15N and δ13C occurred both during dry and wet seasons. There was a significant effect of drought treatment on δ15N both during the dry season (paired t-test, t(8)=-2.86, p < 0.05) and wet season (paired t-test, t(8)=-4.03, p < 0.005). There was no effect of drought treatment on δ13C for both seasons.
Furthermore, we observed significant differences in the mean isotope signatures between drought and control for the two common ant species, O. smaragdina and A. gilberti (Fig. 2). For these species, samples from the drought plot had significantly higher δ15N than the control (ANOVA, O. smaragdina, F1,44=22.12, P < 0.0001; A. gilberti, F1,36=54.89, P < 0.0001). In contrast, no significant difference was found for δ13C values (ANOVA, O. smaragdina, F1,44=0.10, P = 0.75; A. gilberti, F1,36=0.03, P = 0.88).
Seasonal changes in δ13C and δ15N
We examined seasonal trends in ant isotope composition (δ15N, δ13C) and soil moisture in the experiment using smoothing functions in GAMs. Smoothers significance and percent of deviance explained by each model indicate that time of the year was a good predictor for δ15N variations at the control plot, but not the drought plot (Table 1). The overall trend in δ15N values was seasonal for both dominant species at the control plot, but not the drought plot (Fig. 3). For the control plot, peaks in δ15N values occurred in August-October, towards the end of the dry season for both ant species, whereas no peaks in isotope values occurred at the drought plot. There were no seasonal trends in ants’ δ13C in either plots (Fig. 3).
Table 1
Seasonal patterns in soil moisture and ant isotope values of O. smaragdina and A. gilberti as expressed by the strength of the Generalised Additive Models
| Variable | edf | F | P | Explained deviance (%) |
Control plot | δ15N O. smaragdina | 6.998 | 11.00 | < 0.001 | 87.1 |
A. gilberti | 7.366 | 13.29 | < 0.001 | 90.3 |
δ13C O. smaragdina | 1.408 | 0.19 | 0.77 | 4.9 |
A. gilberti | 1.130 | 1.73 | 0.17 | 12.7 |
Soil moisture | 7.248 | 12.82 | < 0.01 | 96.8 |
Drought plot | δ15N O. smaragdina | 1 | 0.01 | 0.92 | 0.1 |
A. gilberti | 1 | 0.10 | 0.76 | 0.6 |
δ13C O. smaragdina | 1 | 1.03 | 0.32 | 4.3 |
A. gilberti | 1.712 | 0.63 | 0.62 | 11.8 |
Soil moisture | 6.965 | 51.77 | < 0.001 | 99.1 |
Smoothers fitted to all soil moisture data were highly significant (Table 1) and could explain similar levels of deviance in the drought and control datasets (96.8% of deviance explained for the control plot, 99.1% for the drought plot). However, comparison of the GAMs for the two plots reveal different soil moisture profiles, with the peak magnitude in soil-water availability not as pronounced in the drought plot as in the control plot (Fig. 3). The exceptionally wet year in 2018 resulted in a significant recharge of the soil moisture in the drought experiment from lateral water movement outside the experimental area. This was detected from soil depths of 50 cm to 150 cm (but not at the surface) from mid-wet season through to mid-year dry season.
At the control plot, we observed distinct patterns in isotope signature that correlated significantly with soil moisture seasonality (Table 2). Peaks in δ15N coincided with troughs in soil moisture (Fig. 3). At the drought plots, however, ant δ15N isotope signatures were not correlated with variations in soil moisture. δ13C was not correlated with soil moisture in either the drought or control plot.
Table 2
Strength and magnitude of the Spearman-rank correlations between GAM predictions of soil moisture and isotope values in O. smaragdina and A.gilberti.
δ15N Control plot | δ15N Drought plot |
O. smaragdina | A. gilberti | O. smaragdina | A. gilberti |
rs = 0.366 P < 0.001 | rs = 0.684 P < 0.001 | rs = 0.013 P = 0.417 | rs = 0.053 P = 0.269 |
δ13C Control plot | δ13C Drought plot |
O. smaragdina | A. gilberti | O. smaragdina | A. gilberti |
rs = -0.0193 P = 0.763 | rs = 0.0846 P = 0.395 | rs = 0.00127 P = 0.321 | rs = 0.0197 P = 0.618 |