I am broadly interested in understanding pathogen spillover and establishment in new hosts. Model systems are critical for developing basic science surrounding disease spread, but my goal is to operationalize findings from lab systems so that they can be applied in real time during emergent epidemics. I’m also interested in understanding whether the drivers of pathogen transmission change as a disease event transitions from between-host-species spillover to persistence within the new host species. At the moment, I primarily work on pneumonia dynamics of bighorn sheep living in the Hells Canyon region of ID, WA, and OR. Hells Canyon sits at a nice intersection between the rigor of a model system, and the challenge of a fully wild one: we have information on many marked individuals over a substantial timespan, along with particularly high-resolution summer data on pneumonia epidemics among lambs, yet most of the specific epidemiological information (for example, exactly when an individual was infected, exactly what contacts are sufficient for pathogen transmission) remain unknown.
That said, the challenges of collecting wildlife disease data are massive, even for a super-charismatic and well-studied species like bighorns. Sheep live a long time (so gathering survival data is a slow process), they’re hard to spot even when they’re collared, and harder still to collect post-mortem for necropsy. Furthermore, sample sizes in this system are huge for wildlife, but very small relative to the intensive data streams available for human or lab-controlled systems. Finally, experimental manipulation and replication pose major ethical and logistical challenges. My work aims to leverage all the data simultaneously, so that we can harness the entire range of available information and synthesize it into a single informative framework.
Severity of lamb pneumonia varies quite a lot between populations and years (see Cassirer et al. 2013, Journal of Animal Ecology), but it’s not clear why some populations lose all their lambs to pneumonia, while others are subject to minimal disease impacts. Understanding the processes underlying this variation is a crucial first-step in developing management strategies to improve summer lamb survival in the presence of pneumonia-causing pathogens.
The team I work with is studying several hypothesized drivers of epidemic severity, including epidemiological history, pathogen and host genetics, coinfection status, and environmental conditions. A full synopsis of our research is located at bighornhealth.org. My current project focuses on understanding the role of one potential driver, variation in bighorn behavior over summer. The idea is that some populations might be very well-mixed during summer, so that one or a small number of infectious ewes could potentially produce exposure among all new lambs, whereas other populations might be more segregated, so that an infected ewe would only have the opportunity to infect lambs in her local group, but not the entire population. Most of my work is based out of the Asotin Creek area, about twenty miles west of Asotin, WA. Below is a rough map of the study area.
Existing data from Hells Canyon suggests that ewe group membership accounts for a lot of the variation in summer lamb survival during years with pneumonia events (Manlove et al. 2014 Proceedings of the Royal Society B). My field work follows up on this hypothesis through an intensive study of social contact patterns (both within groups and across whole populations) in four different study herds. Operationally, this means locating every radiocollared animal every day (or as often as possible), recording each ewe’s lamb status, as well as body condition and molt status, observed symptoms, and the demographic composition of her current group. We also collect focal follows and scan samples documenting individual behaviors and pair-wise interactions.
The field season runs continuously from April 20 to July 15, with occasional week-long follow-up trips in the fall.
My work regularly employs mixed effects models, multivariate methods, survival analysis, and state-space modeling. I’m very interested in methods that simultaneously leverage information from multiple data streams to address questions about mechanistic processes (for example, integrative population models in population biology). I’m also curious about the role of sampling variability in network-based metrics. Behavioral ecologists have amassed extensive data on animal contact networks in the last ten years… but what inferences can we really draw from them, and which metrics are most effected by incomplete observation?
Since I am more scientist than theoretician, my statistical daydreams tend to track my current biological questions. Lately, these have focused on issues pertaining to the trade-offs surrounding decisions about how to project animal movement data onto contact networks (e.g., as spatial networks vs. social networks), and about means of differentiating between separate stuttering chains of transmission and temporally consistent carrier / non-carrier status in sporadically sampled animals.
I’ve occasionally run tutorials and written on code management practices; here are a few links to offshoots of those projects.
Introduction to R tutorial: StatsInRTutorial
Code management in R and RStudio: Code Management in R(1)