Date

12-11-2025

Document Type

Thesis

Abstract

Urbanization is a growing threat to wildlife and poses unique challenges that animals must overcome to thrive in developed places. Temporal partitioning is one method that wildlife may employ to persist among the increased predation, competition, and human presence pressures of urban spaces. To evaluate diel patterns of mesocarnivores along an urban-to-rural (URG) gradient, data were utilized from camera traps established at 27 sites along a URG in southeastern Massachusetts as part of the Urban Wildlife Information Network (UWIN). Camera trap photos from summer and winter between 2023 to 2025 were used and filtered for coyote (Canis latrans), northern raccoon (Procyon lotor), Virginia opossum (Didelphis virginiana) and striped skunk (Mephitis mephitis) detections. Landscape covariates for the sites were calculated in ArcGIS Pro 3.6 (Esri 2025), but all other analysis was performed in R (R Core Team 2025). Six standardized covariates with low (< 5) variance inflation factor (VIF) values were chosen for subsequent analysis. Detections were binned into four diel categories (dawn, day, dusk, and night) using the “suncalc” package (Thieurmel and Elmarhraoui 2022). A generalized linear mixed model (GLMM) was performed by species to assess the influence of the fixed effects, including diel category and VIF-chosen covariates, on predicted detections per 100 trap nights. Tukey pairwise contrasts between diel category pairs were also performed. All four mesocarnivores exhibited predominantly nocturnal activity across the gradient, likely as a sustained human avoidance strategy. Night had significantly greater predicted detection rates than any other diel category for all four mesocarnivores. Only the raccoon saw a significant effect of landscape covariates on predicted detections, specifically for canopy cover within 100-m of the camera and log-transformed ratio of perimeter to area. Kernel density estimation (KDE) and overlap analysis was conducted to observe overlap in high and low covariate values for individual seasons, representing sites on either end of the gradient. By visual assessment, it appeared that there was a strong overlap between high and low values for most seasons and variables. KDE overlap could not be performed for the Virginia opossum and striped skunk due to insufficient data. Future recommendations include increasing the sample size, incorporating Bayesian hierarchical modeling for fixed effects, and evaluating the statistical significance of KDE analysis with a Watson-Wheeler test.

Department

Biological Sciences

Thesis Committee

Dr. Christopher Bloch, Thesis Advisor
Dr. Heather Marella, Committee Member
Dr. Thilina Surasinghe, Committee Member

Copyright and Permissions

Original document was submitted as an Honors Program requirement. Copyright is held by the author.

Download

Included in

Biology Commons

Share

COinS