When the wild things are: Defining mammalian diel activity and plasticity

Abstract

Circadian rhythms are a mechanism by which species adapt to environmental variability and fundamental to understanding species behavior. However, we lack data and a standardized framework to accurately assess and compare temporal activity for species during rapid ecological change. Through a global network representing 38 countries, we leveraged 8.9 million mammalian observations to create a library of 14,587 standardized diel activity estimates for 445 species. We found that less than half the species' estimates were in agreement with diel classifications from the reference literature and that species commonly used more than one diel classification. Species diel activity was highly plastic when exposed to anthropogenic change. Furthermore, body size and distributional extent were strongly associated with whether a species is diurnal or nocturnal. Our findings provide essential knowledge of species behavior in an era of rapid global change and suggest the need for a new, quantitative framework that defines diel activity logically and consistently while capturing species plasticity.

Fig. 1.. Visual summary of the consolidated dataset with project countries color-coded by taxonomic richness along with a phylogeny of species in the dataset.

(A) Summary of terrestrial coverage of the dataset we have compiled based on more than 8.9 million trail camera photo records spanning 38 countries color-coded by country-specific species richness. The semitransparent black dots on the map represent the average location of each camera trap project, which was summarized from the spatial coordinates of each project’s camera trap locations. (B) Phylogeny of 445 nonvolant mammal species from 67 families and 21 orders, with major clades denoted by silhouettes of representative taxa. A subset of this dataset was used for subsequent analysis after data checks. NA, not applicable.

Fig. 2.. A comparison of species reference phenotypes with empirical data.

Large dots for each reference phenotype (y axis) represent the among-species average probability of support for a given diel phenotype (x axis), horizontal bars are ±1 SD of species-specific estimates, and smaller dots are species-specific estimates across all of their respective analysis units. For each species’ analysis unit, the traditional hypothesis set was fitted with equal prior weight on each diel phenotype. Crepuscularity (n = 24) was least accurate at 0%, while nocturnality (n = 202) was found to be about 58% accurate. Diurnality (n = 159) had the best accuracy at 82%, followed by cathemerality at 57% (n = 60) when compared to empirical data.

Fig. 3.. Diel phenotype hypotheses and their associated probability space for the traditional and general hypothesis sets, as well as empirical support under each hypothesis set for two sample species (raccoon and African savanna elephant).

The Traditional hypotheses are composed of diurnal, nocturnal, crepuscular, and cathemeral phenotypes while the General hypotheses encapsulates more phenotypes (crepuscular-nocturnal, diurnal-crepuscular, and diurnal-nocturnal). Subplot axes indicate the probability of activity in twilight (x), daytime (y), and nighttime (z).For the species results, each circle represents an analysis unit’s posterior median probabilities of activity, colored by the supported diel phenotype. For both species we see the loss in information when the Traditional hypotheses are considered while the General provides more specific insights, making a clear separation of biphasic activity (e.g., diurnal-nocturnal) from triphasic activity (e.g., General cathemeral). Note that “Reference” refers to the literature reference classification and “Units” is the number of unique analysis units.

Fig. 4.. Probability of agreement for each species with reference phenotype category mapped to phylogeny.

Only families with five or more species are plotted (n = the number of species per family that were included in our analysis). Lower probability values mean that, for the corresponding species, there are many analysis units that do not agree with their reference classification. Large dots for each reference phenotype represent the among-species average, horizontal bars are ±1 SD of species-specific estimates within a family, and smaller dots are species-specific estimates across their respective analysis units. For each species’ analysis unit, the traditional hypothesis set was fit with equal prior weight on each diel phenotype.

Fig. 5.. The predicted probability of a species being nocturnal, diurnal, or cathemeral depending on a species’ level trait value, average environmental, or anthropogenic gradient.

The solid line and shaded ribbon for each plot respectively represent the median estimate across species and its associated 95% credible interval. Points are species-specific probabilities of nocturnality, diurnality, or cathemerality that were either estimated at their trait value (e.g., mass or distributional extent) or their among analysis-unit average for a given environmental or anthropogenic gradient (e.g., mean distance from the equator, mean hours per day, and mean global human footprint).

Fig. 6.. Roughly a third of the species analyzed demonstrated plasticity in their diel phenotypes along gradients of global human footprint.

This plot is a representative example of species that became more nocturnal (y axis) with increasing global human footprint (x axis), which includes the striped skunk (Mephitis mephitis), snowshoe hare (Lepus americanus), gray fox (Urocyon cinereoargenteus), North American porcupine (Erethizon dorsatum), and the crab-eating fox (Cerdocyon thous). Lines represent median species-specific estimates. The thinner line represents nocturnality predictions for global human footprint values that fall either above or below the range observed across a species’ analysis units, whereas the thicker line represents predictions that fall within that range.

Fig. 7.. Examples of how analysis units—which represent a subset of a camera trap project’s data—could be generated under different camera trap deployment scenarios and the steps we followed to create them.

We created these analysis units to discretize a camera trap project’s sampling effort, calculate the frequency of species detections that occurred during the twilight, daytime, and nighttime, and connect a species detection data to spatiotemporal covariates (e.g., global human footprint and mean hours of daylight per day).