One of the core goals of ecologists is identifying how species interact. Does species A, for instance, have a positive, negative, or neutral impact on species B? Elucidating such relationships is key to understanding community structure and stability.
We know that interactions can be competitive, predatory, parasitic, or mutualistic – but it can be extremely challenging to determine which of these interactions two species have in nature. Many interactions are cryptic, and numerous factors can influence their direction and strength, especially the abundance of those species. Ecologists have used different modeling approaches to help uncover these relationships, but due to the challenges of obtaining accurate estimates of wild species’ populations, most models that examine species interactions focus on species occupancy (presence or absence) and do not account for abundance. A paper published in the January 2025 issue of the journal Ecology looks to address that shortcoming.
A team of researchers from Cornell University and the U.S. Geological Survey, led by Joshua P. Twining, developed a model for examining species interactions involving two or more species that incorporates the abundance of the focal species. As part of this modeling process, the authors employed a novel method for estimating species abundance using detection/non-detection data (for example, presence/absence data from camera traps). They tested the efficacy of their model in a few ways. First, using a series of computer simulations, they compared their model (with its abundance estimates) to existing species-interactions models (those relying on occupancy data alone) to evaluate how well the different models performed under different ecological conditions (for example, whether species vary in abundance over space and time) and data-collection scenarios (for example, varying levels of precision in the detection data). Next, they tested the new model using detection/non-detection data of a community of predators in northeastern New York. The authors examined whether estimated coyote (Canis latrans) abundance influenced the abundance of two mustelid species – fisher (Pekania pennanti) and American marten (Martes americana) – as well as whether fishers had an effect on the abundance of martens.
Researchers collected predator data across the Adirondack and Tug Hill regions of New York between 2016 and 2018. They established 195 plots, each measuring 15 square kilometers, and deployed a single camera-trap within each plot, with a bait station on a tree opposite each camera. Cameras recorded which species visited the bait station over the course of three weeks. Researchers assessed several additional parameters, including forest type (deciduous, coniferous, mixed), snow depth, and date to help understand what other factors might influence each species’ presence and abundance.
For the simulation studies, the authors found that the existing modeling approaches (those relying primarily on occupancy) had higher rates of error in the interaction parameter estimates compared to the new model, meaning that the strength of the species interaction was more likely to be over- or underestimated and to result in potential misinterpretations. The authors state that “[t]his [new] model represents an advancement over prior methodologies” that only account for occupancy, or that utilize less appropriate estimates of abundance.
Results from the camera-trap study indicate that coyotes had a positive impact on fishers, meaning there were more fishers in areas with more coyotes, but a negative impact on martens. They also found that fishers did not appear to have an effect on martens. These results were somewhat surprising because prior studies had suggested that coyotes have a negative effect on fishers and martens, and that fishers have a negative effect on martens. In addition, a previous paper by Twining, using the same dataset but relying on a traditional modeling approach focused solely on occupancy, had failed to detect an effect of coyotes on martens.
The authors did not speculate on the possible causes of the abundance-mediated species interactions, but they did report on the other parameters associated with the presence and abundance of the three predators. Coyotes were positively associated with forest edge and deer availability, while fishers were positively linked to all three forest types and negatively linked to snow depth. Marten abundance was also positively associated with all three forest types and snow depth.
The authors conclude that this new modeling framework has significant potential as a “tool in a wide range of real-world conservation and management contexts” and may ultimately lead to better-informed decisions in wildlife management and conservation.