Natural and anthropogenic sources of habitat variation influence exploration behaviour, stress response, and brain morphology in a coastal fish.

Evolutionary ecology aims to better understand how ecologically important traits respond to environmental heterogeneity. Environments vary both naturally and as a result of human activities, and investigations that simultaneously consider how natural and human-induced environmental variation affect diverse trait types grow increasingly important as human activities drive species endangerment. Here, we examined how habitat fragmentation and structural habitat complexity affect disparate trait types in Bahamas mosquitofish Gambusia hubbsi inhabiting tidal creeks. We tested a priori predictions for how these factors might influence exploratory behaviour, stress reactivity and brain anatomy. We examined approximately 350 adult Bahamas mosquitofish from seven tidal-creek populations across Andros Island, The Bahamas that varied in both human-caused fragmentation (three fragmented and four unfragmented) and natural habitat complexity (e.g. fivefold variation in rock habitat). Populations that had experienced severe human-induced fragmentation, and thus restriction of tidal exchange from the ocean, exhibited greater exploration of a novel environment, stronger physiological stress responses to a mildly stressful event and smaller telencephala (relative to body size). These changes matched adaptive predictions based mostly on (a) reduced chronic predation risk and (b) decreased demands for navigating tidally dynamic habitats. Populations from sites with greater structural habitat complexity showed a higher propensity for exploration and a relatively larger optic tectum and cerebellum. These patterns matched adaptive predictions related to increased demands for navigating complex environments. Our findings demonstrate environmental variation, including recent anthropogenic impacts (<50 years), can significantly affect complex, ecologically important traits. Yet trait-specific patterns may not be easily predicted, as we found strong support for only six of 12 predictions. Our results further highlight the utility of simultaneously quantifying multiple environmental factors-for example had we failed to account for habitat complexity, we would not have detected the effects of fragmentation on exploratory behaviours. These responses, and their ecological consequences, may be complex: rapid and adaptive phenotypic responses to anthropogenic impacts can facilitate persistence in human-altered environments, but may come at a cost of population vulnerability if ecological restoration was to occur without consideration of the altered traits.

Digitally deconstructing leaves in 3D using X-ray microcomputed tomography and machine learning.

PREMISE: X-ray microcomputed tomography (microCT) can be used to measure 3D leaf internal anatomy, providing a holistic view of tissue organization. Previously, the substantial time needed for segmenting multiple tissues limited this technique to small data sets, restricting its utility for phenotyping experiments and limiting our confidence in the inferences of these studies due to low replication numbers. METHODS AND RESULTS: We present a Python codebase for random forest machine learning segmentation and 3D leaf anatomical trait quantification that dramatically reduces the time required to process single-leaf microCT scans into detailed segmentations. By training the model on each scan using six hand-segmented image slices out of >1500 in the full leaf scan, it achieves >90% accuracy in background and tissue segmentation. CONCLUSIONS: Overall, this 3D segmentation and quantification pipeline can reduce one of the major barriers to using microCT imaging in high-throughput plant phenotyping.