In Crystal Lake, British Columbia, small fry (≤15 mm SL) of the threespine stickleback (Gasterosteus aculeatus) are concentrated in vegetation while larger fry are not. Because fry in all size classes feed primarily on zooplankton, even when in vegetation, we hypothesized that size-limited predation was responsible for the observed shift in habitat use with size. The major predators on fry in Crystal Lake are adult threespine stickleback, the water scorpion, Ranatra sp. (Hemiptera: Nepidae), backswimmers, Notonecta spp. (Hemiptera: Notonectidae), and dragonfly naiads of the genus Aeshna (Odonata: Aeschnidae). On the basis of distribution and hunting behavior we excluded the insects Ranatra sp., and Notonecta sp. as causal agents for this shift in resource by fry in water >0.25 m deep. Ranatra was found almost exclusively near the shoreline in water <0.25 m deep, and both insects hunted primarily as ambush predators within vegetation. Such predators seemed more likely to drive vulnerable fry from vegetation than to restrict them to it. In contrast, Aeshna naiads and adult stickleback frequently hunted outside of vegetation. In prey preference experiments the naiads did not show the decline in predation efficiency on fry >15 mm SL that would be expected if size-limited predation by this insect was responsible for the observed shift in resource use by fry. Adult stickleback only fed on fry <15 mm SL, and in an experimental situation, consumed fry at a rate 10 times greater than that exhibited by any of the insects. Predation experiments demonstrated that small fry (11-15 mm) spent more time in vegetation in the presence of adult conspecifics than they did in control pools, as would be expected if size-limited cannibalism caused small, vulnerable fry to be restricted to vegetation. Fry >15 mm SL were found outside of vegetation more often than in control treatments. The probable cause of this result is that adults become aggressive toward fry at this size, and often could be seen chasing large fry from vegetation during the experiments. Dragonfly naiads (Aeshna spp.) spent most of their time in vegetation in the experimental pools. Both size classes of fry spent less time in vegetation in the presence of dragonfly naiads than they did in control treatments, an apparent reflection of their similar vulnerabilities to these naiads. The presence of vegetation in pools reduced predation rates by adult stickleback on small fry. Because the experiments presented here indicate that fry are capable of rapidly assessing predation risk and of altering their behavior adaptively, we conclude that small fry occupy vegetation as a refuge from cannibalism. Once fry have reached the size-threshold at which they are no longer vulnerable to adult conspecifics they are able to forage farther from vegetation thereby reducing risk of predation by insects in vegetation and possibly acquiring more abundant food resources.

Increasing urbanisation has led to a greater use of artificial structures by bats as alternative roost sites. Despite the widespread presence of bats, roost availability may restrict their distribution and abundance in urban environments. There is limited quantitative information on the drivers of bat roost selection and roosting preferences, particularly in African bats. We explore the factors influencing roost selection in the Mauritian tomb bat (Taphozous mauritianus), within an urban landscape in Lilongwe city, Malawi. Eight building and five landscape features of roosts were compared with both adjacent and random control buildings throughout the city. Bat occupied buildings were situated closer to woodland (mean 709m) compared to random buildings (mean 1847m) but did not differ in any other landscape features explored. Roosts were situated on buildings with larger areas and taller walls, suggesting bats select features for predator-avoidance and acoustic perception when leaving the roost. Bats preferred buildings with exposed roof beams which may provide refuge from disturbance. Whilst roosts are situated more often on brick walls, this feature was also associated with landscape features, therefore its importance in roost selection is less clear. These results are indicative that T. mauritianus selects roosts at both the building and landscape level. The selectivity of T. mauritianus in relation to its roost sites implies that preferred roosts are a limited resource, and as such, conservation actions should focus on protecting roost sites and the woodland bats rely on.