Thesis
Application of Machine Learning Algorithms to Predicting Insect Assemblages in Grass Communities
Washington State University
Master of Science (MS), Washington State University
2022
DOI:
https://doi.org/10.7273/000005171
Abstract
Machine learning algorithms are increasingly used in ecological application to identify species images, categorize habitats and predict populations. These modeling techniques are relatively new to the field of entomology and have rarely been applied outside of image classification. In cereal crop systems, insects primarily use grasses as their alternative hosts. In this thesis, I demonstrate the application of machine learning models to predicting insect communities within grass ecosystems of the Palouse region of northern Idaho and eastern Washington.
First, I apply machine learning models to three common pest insect families (Cicadellidae, Miridae, Aphididae). Applying both classification and regression techniques, I determined the feasibility of making insect predictions using alternative host plants as predictor variables. In classification models, data balanced between presences and absences contributed greatly to model performance while regression models with greater variability in abundances performed better. In all models, abiotic factors were ranked highest in contribution to model predictions.
Next, I applied the same model techniques to predicting five insect community metrics (species richness, species evenness, Shannon diversity, Hill number 1 and Hill number 2) with the addition of landscape metrics as predictors. Abiotic and landscape metrics were most important to all community metrics with vegetation metrics providing rare contribution to model variability reduction. Overall, these results indicate that abiotic factors are the strongest drivers of insect populations and diversity with landscape composition additionally contributing to insect community diversity.
Metrics
1 File views/ downloads
23 Record Views
Details
- Title
- Application of Machine Learning Algorithms to Predicting Insect Assemblages in Grass Communities
- Creators
- Megan Blance
- Contributors
- David Crowder (Advisor)Robert Clark (Committee Member)Tobin Northfield (Committee Member)Sanford Eigenbrode (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Entomology, Department of
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University
- Number of pages
- 66
- Identifiers
- 99901019941101842
- Language
- English
- Resource Type
- Thesis