Degree Name

BS

Department

Biology

College

Life Sciences

Defense Date

2025-07-17

Publication Date

2025-12-12

First Faculty Advisor

Stephen Piccolo

First Faculty Reader

Matthew H. Bailey

Honors Coordinator

Steven L Peck

Keywords

Random Forest, PPS

Abstract

The Antarctic McMurdo Dry Valleys are some of the harshest environments on our planet. They may not support higher fauna and vegetation, but the environment has a fairly diverse soil microbiome. The soil microbiome consists mainly of invertebrate nematodes (Plectus, Scottnema, and Eurydiolaimus), rotifers, and tardigrades. In this study, we examined ecological modeling approaches to enhance our understanding of the processes contributing to the overall population. In this process, two extensive data sets were used: the Biotic Effects Experiment (BEE) and the Elevational Transect (ET) data set compiled by the McMurdo Dry Valleys Long Term Ecological Research program were combined. This study differs from other studies because of the focus on rigorous data-driven analysis of model performance, variable influence, and the limits of ecological prediction in polar ecosystems. We employed Predictive Power Score (PPS) analysis and Random Forests modeling in order to contrast the explanatory power of a range of environmental and biological factors in predicting patterns in counts of total, living, and dead soil organisms. This allowed us to compare the explanatory power of abiotic factors such as moisture and conductivity with biotic indicators such as measures of invertebrate population of major invertebrates. Through comparative testing with other statistical approaches, we highlight the importance of preprocessing, variable selection, and model validation in ecological forecasting. The modeling framework presented here is a valuable tool for testing species interactions, community structure, and ecological responses in data-poor, extreme environments. This study is one demonstration of how integrated, exploratory modeling can be used to enable hypothesis generation and improve forecasting activity in climate-sensitive environments like the McMurdo Dry Valleys.

Download

Share

COinS