A Network Model for M.ovi Transmission in Bighorn Sheep and Aoudad
(Angela’s Version, The Ten-Minute Version)
By Angela Patrick, PhD student
Preface: Dr. Conway and Courtney dared me to be bejeweled and describe my research entirely in Taylor Swift song titles (and a few lyrics). Are you ready for it...?
Many recent studies into big game ecology have examined sociality and migration, demonstrating that wildlife are highly connected through landscapes but, as we saw in COVID, interactions (aka transmission events) among individuals also have connections too. Bighorn sheep in the Trans-Pecos region are native, highly social, and vulnerable to disease outbreaks. Add in introduced aoudad, a highly gregarious, invasive ungulate, and things have only gotten crazier. For part of my dissertation, this is me trying to build a network model that examines how Mycoplasma ovipneumoniae (hereafter, M.ovi) moves between bighorn sheep and aoudad populations and herd groups. Competition between bighorn sheep and aoudad have been a “long time coming” as the central focus of part of three iterations of what’s affectionately known as the "BAM" tri-species project (bighorn, aoudad, mule deer).
BAM is a longstanding collaborative effort between Borderlands Research Institute (BRI) at Sul Ross State University, Texas Parks and Wildlife Department (TPWD), and Texas Tech (Us). This latest iteration of BAM, BAM 3.0 hits different because we’ve added the overarching question of “how does M.ovi move between these two species?” as an objective. Texas Tech’s role throughout BAM has been focused on disease surveillance, while BRI has been taking the I know places route and focusing on GPS collars and examining competition and movement ecology.
My task is to be the mastermind of the epidemiological models, using as much available data as possible (which is bigger than the whole sky, if we’re being honest). Using the GPS collar data from BRI and pairing it with the disease surveillance data that is ours, we can examine group transmission events using a network model and later, I’ll examine transmission on an individual level using an agent-based model (stay tuned for that).
Network models are nothing new but can allow us to characterize connectivity and transmission potential among herd groups, species, and mountain ranges in a formal risk assessment framework. The Trans-Pecos region is now a system where everything has changed and an introduced invasive species (aoudad) vastly outnumbers the native species (bighorn sheep) in their native habitat. Fundamentally, network models are connected by an invisible string because of the social importance of the individual.
In other words, you might have individuals who are the mirrorball of the group, and the extent to which they dominate the network, is probably closely linked to their role in disease spread, further emphasizing network model relevance to epidemiology. Some individuals may be Mr. Perfectly Fine and either not get sick or don’t transmit disease to other individuals; others may be treacherous and run all over the place, spreading disease as they go.
All individuals in a network model are assigned a finite set of contacts, capturing the permanence of interactions and network structure impact on disease dynamics, therefore highlighting how they may or may not ruin the friendship, specifically how clustering, heterogeneity, and long-range connections can impact dynamics. In my network model, nodes will represent the 1, individual animals or herds, and edges are the strength and/or frequency of the interaction. To identify individuals who are highly connected and could act as “superspreaders” (the epitome of you're not sorry), I also look for individuals who are like a good Taylor Swift bridge, connecting otherwise disconnected herds. I can use metrics like:
eigenvector centrality - the measure of influence or importance in a network
Freeman betweenness centrality - how often an individual lies on the shortest path between other individuals within a network, also quantifies bridging roles in the network
modularity - how strongly a network divides into distinct communities
By using these metrics, I’ll be able to highlight herd-to-herd or mountain-range connectivity, which lets me say I knew you were trouble to problematic individuals in terms of epidemiology and apply their data to the ABM.
Where I am right now is in the production stage and have definitely had to fight the great war when it came to making sure we could access the database where our collar data are stored. I would be remiss to say that learning the theory and math behind network models has been a bit like navigating a labryinth, and I’m sure coding all of this in R will have me trying to shake it off for bad model runs or the anticipated endless errors I know I will get. Naturally, some of my data cleaning has had me muttering look what you made me do as I find problem children who pop up singing “It’s me, hi, I’m the problem, it’s me.”
In working with partners at Borderlands Research Institute at Sul Ross State University and Texas Parks and Wildlife Department, I’ve identified that the outcome of both models I’ll be developing is to project which animals or even which areas are more likely to experience transmission of M.ovi. Essentially, I hope that these models can help us imagine wildest dreams scenarios, like what happens if we have individuals that foray long distances? What happens if individuals stay in one area? What happens if the foraying individual encounters the group staying in one place? With more iterations, we can revisit scenarios and know all too well what could happen.
Long story short, by creating this network model and eventually the ABM, we can hopefully check off an item from a wildlife managers’ wish list by providing a tool that can provide additional decision support. As we’ve observed in the past 10 years, there has been a lot of Change in our bighorn sheep populations because of aoudad and because of M.ovi. We don't want our current herd of bighorn to be the last great American dynasty. Hopefully, understanding the connections between herds and individuals from a M.ovi perspective will allow us to maybe one day, say Long Live to healthier herds and maybe we can get out of the woods with overpopulated aoudad.
The model is still a bit of a blank space right now, and I’m certain I’ll encounter some champagne problems as I work through it. But I hope by the end of this, we will no longer be dancing with our hands tied and will prevent bighorn death by a thousand cuts.
So, how many Taylor Swift references did you catch from Angela's post?
Comments