Upcoming Seminar
- Thursday, January 29, 2026
- 9:30-10:30 am · MSB 318
- B Sagar will speak on “Dynamics of a two-stage epidemiological model with post-infection mortality and transmission heterogeneity“
- Abstract: Many diseases naturally progress from an initial mild infectious phase to a more severe stage (e.g., Influenza and COVID-19). This motivates us to formulate a two-stage epidemiological model that incorporates key post-infection features, including reinfection and post-infection mortality (PIM). The model emphasizes the role of transmission heterogeneity in shaping disease dynamics, influencing both endemic levels and oscillatory behaviors. Numerical simulations show that late-stage hyper-infectivity leads to higher endemic infection levels with long-period oscillations, while early-stage hyper-infectivity results in lower infection levels with shorter-period oscillations.
Previous Seminars
- Wednesday, November 5, 2025, 1:00–2:00 pm · MSB 110
- Christen Fleming: Kernel Density Estimation in Spatial Ecology
- Abstract: Kernel Density Estimation (KDE) is a cornerstone technique in statistics for estimating a probability density function when its parameterized form is unknown. KDE is widely used in spatial ecology for calculating animal home ranges and species distributions where the assumptions behind conventional KDE are often pushed beyond their limits. In this seminar I will present over a decade of research that now allows for accurate kernel density estimates to be computed from irregularly sampled times-series data that are sampled from heterogeneous populations.
- Wednesday, October 1, 2025, 1:00–2:00 pm · MSB 110
- Erika Lin: Improving statistical methods for wildlife corridor estimation
- Nozomu Hirama: Methods for quantifying nocturnality and human impacts using animal tracking data
- Abstracts
Methods for quantifying nocturnality and human impacts using animal tracking data
Diel activity patterns often reflect an animal’s adaptations and strategies to optimize their fitness. However, these patterns can be disturbed by human influence, potentially forcing individuals to be active during less optimal times of the day. To better understand such effects, we introduce a new method for estimating nocturnality from tracking data, along with a new metric of the overlap and distance between the home range and Human Footprint Index (HFI). First, we have developed a continuous-time movement model (ctmm) that switches between high and low levels of movement according to solar time. Using this model, we can accurately estimate the proportion of nighttime activity from irregularly sampled tracking data. Second, we have developed a novel approach to quantifying HFI levels, or other index, both within an individual’s home range and in the proximal surrounding areas by using log expectation values. Importantly, this metric increases with increasing disturbance within the home range, with increasing disturbance in the neighborhood of the home range, and with increasing proximity to said disturbances, which makes this a suitable predictor for examining any response to human influence. With our new metrics, we explore the impacts of human disturbance on nocturnality for the common raccoon (Procyon lotor), coyote (Canis latrans), and Temminck’s ground pangolin (Smutsia temminckii). Our work is implemented in the ctmm R package, providing accessible tools to better inform conservation and management in an increasingly anthropized environment.
Improving statistical methods for wildlife corridor estimation
Habitat connectivity is essential to conserving biodiversity, by allowing animals to search across landscapes for resources and mates. This is particularly important for migration and dispersal, which can be impeded by human activities that degrade and fragment habitats. To maintain connectivity, it is imperative that we identify wildlife corridors—areas where animals traverse frequently to move between suitable environments. Traditionally, connectivity has been modeled via two main methods: resistance circuit models and Brownian bridges. However, these methods do not target a “well-calibrated” probabilistic corridor distribution—i.e., where 95% of animals pass through the 95% cross-section. We have developed a new statistical method for corridor estimation, using cross-sectional kernel density estimation (KDE) to model corridors as “range distributions” from animal tracking data. We define the corridor by the distribution of repeated passages between two areas in a landscape. To ensure that this method is robust and offers improvements, we conducted a comparative sensitivity analysis across parameters that impact performance, such as sampling frequency and passage count, using GPS-tracked mule deer (Odocoileus hemionus) and jaguars (Panthera onca). We found that cross-sectional KDE is insensitive to the sampling frequency and produces consistent distribution estimates regardless of the location-recording interval, thus demonstrating that our method provides more rigorous estimates of corridor space needed by migrating animals even with few tracking points available. Our research, implemented in the ctmm R package, contributes a novel statistical tool that ecologists can directly apply to conservation management, through designating new wildlife corridors and evaluating the impact of existing ones.
Seminars listed on the old website