4. Discussion
We observed different patterns in species assemblage distribution over small-scale vertical and horizontal gradients in a complex habitat. The high local diversity of arboreal ants can be explained by high environmental and habitat heterogeneity along both horizontal and vertical dimensions. We found consistently greater horizontal turnover in ant assemblages than that observed at comparable vertical distances. However, a distance-decay pattern of ant turnover was only detected vertically, but not horizontally, with the high species turnover between horizontal positions being independent of distance. The observed patterns confirmed our hypotheses, and are likely associated with continuous directional changes in environment (i.e., microclimate) and resources vertically, and stochastic variation in environment and resources, as well as poor connectivity horizontally.
We found the spatial distribution of ant assemblages was associated with heterogeneity in air temperature and relative humidity within the three-dimensional structure of the forest. The spatial pattern of microclimate variables in our study plot is in line with other studies in tropical rain forest: air temperature increased with height above ground, while relative humidity decreased (Davis et al. 2019, De Frenne et al. 2019). Horizontal variation in microclimate was also high but without directional change, presumably relating to variation in vegetation structure within and between trees (Fetcher et al. 1985, Scheffers et al. 2017). Such non-directional horizontal variation in microclimate may contribute to the high turnover of ant assemblages and lack of distance-decay pattern in this dimension at the scales we sampled. Given that turnover at very short horizontal distances was already nearly maximal, it would not have been possible for turnover to then be even greater with increasing horizontal distance. Tropical arboreal ants show thermal adaptation to their vertical habitat use through their physiology (Kaspari et al. 2015), morphology (Law et al. 2020) and nesting site selection (Plowman et al. 2019). The high variance in microclimate generates diverse thermal niches for ant species with different thermal tolerances, and hence can facilitate co-existence of multiple species at small spatial scales (Lessard et al. 2009). For tiny ectotherms like ants, fine-scale environmental heterogeneity can play an important role in defining their distributions, probably due to the small foraging range and small body size of ants, and the thermal diversity of the environment (Ribas and Schoereder 2007, Fayle et al. 2010, Klimes et al. 2012, Kaspari et al. 2015, Bütikofer et al. 2020). The association between microclimate pattern and ant assemblage composition may result not only from direct effects through environmental preferences of ant species, but can also be a consequence of indirect biotic effects, for example if microclimate influences ant food availability. However, the relative importance of direct and indirect microclimate influences on ant assemblage composition remains to be investigated.
In addition to microclimate, biotic influences such as resource limitation and vegetation structure may also contribute to the high horizontal turnover we observed. The decline in ant richness and abundance with height in the canopy could be due to reduction in leaf area, which limits foraging range and nest site availability (Adams et al. 2019, Plowman et al. 2019). Ant assemblage composition can be affected by vegetation structure such as tree size, number of branches, and cavity diversity (Powell et al. 2011, Yusah and Foster 2016, Adams et al. 2019, Plowman et al. 2019). Hence, we can expect distinct ant communities to be hosted by different individual trees and that the distance between trees per se might not be driving assemblage dissimilarity. The lack of distance effects on high horizontal turnover arboreal ants in our study is consistent with that found in canopy ant assemblages across greater horizontal distances (100-700 m) in Mexican rain forest, whereas a distance-decay pattern was observed in ground ant assemblages (Antoniazzi et al. 2021). For social insects like ants where workers are wingless, vertical movement of workers within the colony tree is likely to be less challenging than movement between trees, especially without vegetation connections such as lianas (Yusah and Foster 2016, Adams et al. 2019), which may be less common in the Asian tropics compared with the American tropics (Dial et al. 2004a). Such lack of connectivity between individual trees means that each tree canopy might function as an island within the forest (Southwood and Kennedy 1983, Adams et al. 2017).
Our findings offer important insights into the way in which biodiversity is maintained at fine scales in complex three-dimensional habitats. The high species richness discovered within the 130 m long by 70 m high vertical plot in our study site represents a high proportion of ant diversity at larger scales. The number of species that we detected using precision fogging across vertical strata among 11 trees reached about 40% of the number of species sampled in a similar forest habitat elsewhere in Sabah from 99 trees (Floren et al. 2014). This finding of a relatively large proportion of regional biodiversity being sampled from small plots within rain forest is in line with patterns for herbivorous insects (Novotny et al. 2007), birds (Huang and Catterall 2021) and butterflies (Daily and Ehrlich 1995) and is likely driven by high structural complexity at small scales (Basset et al. 2012).
Our findings need to be interpreted with the caveats of our limited sampling time and sample size. Microclimate information was only monitored for 24 hours for each transect, although we feel that this is representative of longer-term patterns (see above). Future studies involving long term monitoring of microclimate at fine spatial scales are needed to understand the distribution of microclimate with the complex three-dimensional structure of tropical rain forest. In addition, although our sampling intensity per tree was high, we sampled across a relatively limited spatial area and some samples were missing. Future research using greater numbers of replicates at larger spatial scales are needed to confirm the generality pattern we report here.