Project Type

Poster

Publication Date

Spring 2026

Department or Program

Biological Sciences

College

College of Arts & Sciences

Faculty Mentor #1

Dr. Alan Griffith

Abstract

Understanding spatial patterns in biodiversity is important for identifying how ecological communities are structured across landscapes. Clustering methods provide a useful way to detect non-random species distributions and reveal potential environmental or geographic influences. In this study, we applied a clustering framework to a large long-term bee abundance dataset from the Mid-Atlantic United States to evaluate community structure across regions of Maryland, Delaware, and Washington, D.C.

The dataset included over 1,100 sites and more than 3,000 transects. Because sampling methods were not fully standardized, abundance data were converted to presence/absence and analyzed using the Jaccard index to measure site similarity. Hierarchical clustering was then used to identify natural groupings of sites without requiring preassigned cluster numbers. Non-metric multidimensional scaling (NMDS) was used to assess cluster separation, and mapped outputs were created to visualize geographic patterns.

Results showed clear regional differences in bee community structure. The two Plains regions were each dominated by one major cluster, suggesting relatively consistent species composition across sites. In contrast, the Northern Piedmont produced four more balanced clusters, indicating greater ecological heterogeneity and a higher proportion of unique communities. NMDS supported these findings by showing clear separation among clusters.

Overall, this study demonstrates that clustering methods can effectively simplify large ecological datasets and uncover meaningful spatial patterns in pollinator communities. The workflow used here offers a practical and reproducible approach for analyzing biodiversity data in other taxa and regions.

Download

Share

COinS