This blog post is provided by Brian D. Gerber and Mason Fidino and tells the #StoryBehindthePaper for the paper “A model-based hypothesis framework to define and estimate the diel niche via the `Diel.Niche’ R package”, which was recently published in Journal of Animal Ecology. In their paper, they present the ‘Diel.Niche’ R package and demonstrate how this package can be used to evaluate hypotheses of diel phenotypes based on empirical data and estimate the probability of activity during the twilight, daytime, and night-time periods.

How individuals distribute their activity throughout a diel period, otherwise known as the light-dark cycle of a single day, is a fundamental animal behaviour and important niche dimension. Animals are well known to have morphological and physiological adaptations that fit to when they are most active over diel time. Primates such as the Tarsiers and Galagos, for example, have disproportionately large eyes for their body size which allows more light in, enabling them to see better when it is dark. As such, it is not surprising that these species are largely nocturnal.

When animals are active over a diel period it influences what resources are available to them, what environmental conditions they must contend with (e.g., temperature and humidity), and which species they will interact with. A hawk and an owl, for example, may prey on the same food resource but do so at different times of day, which decreases direct interspecific competition between them. As such, variation in animal activities (e.g., herbivory, nutrient deposition, predation) at different times of the day influences ecosystem functioning by means of changing the abiotic environment and timing of species interactions and energy flow through the system.

In the last several decades, technologies (e.g., camera-traps and GPS-telemetry devices) have advanced dramatically to allow researchers to observe activity levels of wild animals continuously throughout the day. This has led to new opportunities to better understand species’ activity and especially how their activity changes due to biotic and abiotic factors, such as climatic season, landscape context, human activity and infrastructure, and species interactions. The development of our work, “A model‐based hypothesis framework to define and estimate the diel niche via the `Diel.Niche’ R package”, was motivated by what we saw as a missing conceptual piece to make inference on animal activity comparable across studies. Specifically, we saw a disconnect between animal activity inference (e.g., probability of activity during certain hours of the day) and terms like diurnal, nocturnal, crepuscular, and cathemeral, which ecologists and evolutionary biologists have long used to describe the expression of observable diel behaviour. What proportion of time should a species be active at night to be considered nocturnal? If a species spends 60% of their time active during the day, does that mean they are diurnal? Further, can we make probability statements about how certain we are of a species’ diel classification? It is questions like these that we set out to answer when we made the `Diel.Niche` R package.

In our paper, we outline three main hypothesis sets (Maximizing, Traditional, and General; Figure 1) that translate the probability of activity during the three fundamental periods of light availability (twilight (low light), daytime (full light), and night-time (no light)) into different diel phenotypes. The three hypothesis sets differ in the objective they aim to accomplish. More so, we implement these hypothesis sets within a model-based framework in the Diel.Niche R package. Researchers can use this package to either classify estimates they have produced from a separate model using the diel phenotypes we have described, or use data directly to estimate support for diel phenotype hypotheses and estimate the probabilities of activity in twilight, daytime, and night-time. Our goal in developing this conceptual and modelling framework is to standardize the language of the diel phenotypes so that we can be confident in describing animal activity across studies.

For example, if two studies describe Coyotes (Canis latrans) as being nocturnal and diurnal, respectively, is this difference due to a real observed shift in diel phenotype, or is it because the operational definitions being used for diurnal/nocturnal are simply different? We have found that many studies use qualitative or eye-balling of diel activity density plots to describe activity patterns (Figure 2), which can easily lead to unnecessary disagreement among studies and thus potentially create spurious inference on shifts in diel phenotypes. The solution is standardization of the definitions of the diel phenotypes and quantification, which is what the Diel.Niche R package solves. Furthermore, you can use the tools we developed within Diel.Niche for the entirety of a research project, or to simply add an additional sentence to the results so that you share your results in a simple and standardized way. For example: “We found that under the traditional hypothesis set, there was a 0.90 probability that coyotes were nocturnal in our study area.”

When developing the model framework in this project, we approached the problem of estimating a species diel phenotype with the goal that it had to be useful across a large range of sample sizes. We did not want to create a methodology that could only be applied to common or highly detectable species. All the authors on this project have many years of experience working with camera-trap data and know well that it is common that rare species and hard to detect species are not studied because of small sample sizes. In our paper, we present extensive simulations for all diel phenotype hypotheses to provide researchers with guidance on what they can expect under sample sizes ranging from 10 to 1080 observations. Our findings are encouraging that in many scenarios 10-80 observations of a species (depending on the hypothesis) can lead to a high-level of confidence in supporting the correct diel phenotype.

In addition to the paper and the R package, we have provided many online vignettes (https://github.com/diel-project/Diel-Niche-Modeling) to help users get started using the Diel.Niche package. We demonstrate how to make inference using data on a single species, multiple-species, how to estimate the diel phenotype from circular kernel density analyses (e.g., the `overlap` R package) and state-dependent animal movement models. We also demonstrate how users can create their own diel hypotheses within this framework. We hope that our work will not only encourage discussions about definitions of diel phenotypes, but also encourage people to reach out to us if they would like help implementing these within Diel.Niche.

Read the full paper

Gerber, B. D., Devarajan, K., Farris, Z. J., & Fidino, M. (2024). A model-based hypothesis framework to define and estimate the diel niche via the ‘Diel.Niche’ R package. Journal of Animal Ecology, 00, 1–1 5 . https://doi.org/10.1111/1365-2656.14035