Triatominae in North America - A Comparison of Statistical Machine Learning Approaches for Habitat Suitability Modelling

Aug 5, 2026

Upload by Aria

Modeling historical habitat suitability for Triatominae in North America uses bio-climatic and land cover data combined with observations across three statistical machine learning approaches. The comparison includes MaxEnt for single-species presence-only modeling, a Bayesian Linear Regression model for joint species distributions, and a Bayesian Neural Network adapted for both joint and single-species cases. Single-species models generalize well beyond observed predictor ranges. Performance metrics report MaxEnt AUC greater than 0.9 and test accuracy with BNN fraction correct above 0.7, and future projections under CMIP6 scenarios for 2070–2100 indicate range shifts.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages57

Reading Time2 h 24 min

AccessFree to download

Acknowledgments

Abstract

Chapter 1. Introduction

Chagas Disease

Chapter 2. Implementation

Sources of Data

Model Selection

Maximum Entropy

Bayesian Linear Regression

Bayesian Neural Network

Chapter 3. Results

Historic Spatial Extent

Future Projections

Chapter 4. Analysis

Discussion

Future Work

Appendices

Glossary of Input Variables

Historic Habitat

Future Projections of Habitat Suitability

Bibliography

Vita

The Thesis Committee for Rebecca Dorothy Langdon certifies that this is the approved version of the following thesis: Triatominae in North America: A Comparison of Statistical Machine Learning Approaches for Habitat Suitability Modelling. APPROVED BY SUPERVISING COMMITTEE: Katherine Brown, Co-Supervisor Clint Dawson, Co-Supervisor Triatominae in North America: A Comparison of Statistical Machine Learning Approaches for Habitat Suitability Modelling. by Rebecca Dorothy Langdon THESIS Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of MASTER OF SCIENCE IN ENGINEERING, COMPUTER SCIENCE, AND MATHEMATICS THE UNIVERSITY OF TEXAS AT AUSTIN May 2023 Acknowledgments I would like to thank my supervisor Dr. Katherine Brown for her continued support of my work and her guidance on this project. Professor Clint Dawson and the rest of the FloDisMod team have been invaluable and their knowledge into the subject matter has guided me throughout. Special thanks to my colleague Liting Huang, without whom I would not have been able to complete this work. Finally, I would like to thank Professor Eduardo A. Rebollar Tellez from the Universidad Aut´onoma de Nuevo Le´on and Professor Teresa Feria from the University of Texas Rio Grande Valley for their guidance on species distribution modelling and the data they kindly provided. Triatominae in North America: A Comparison of Statistical Machine Learning Approaches for Habitat Suitability Modelling. Rebecca Dorothy Langdon, M.S CSEM. The University of Texas at Austin, 2023 Supervisors: Katherine Brown Clint Dawson Bio-climatic and land cover data were used alongside observations often species of Triatominae in three statistical machine learning models to determine the historical habitat suitability. The three models selected were Maximum Entropy [MaxEnt], a single species model, a joint species Bayesian Linear Regression model [BLR], and a Bayesian Neural Network [BNN] adapted to model both joint, and single species distributions. It was found that the single species models perform well outside of the observed range of predictor variables, MaxEnt AUC > 0.9 and BNN had a fraction of correct predictionson test data f > 0.7. Future projections for 2070-2100 were produced under four Coupled Model Intercomparison Project (CMIP6) scenarios; SSP1-2.6, SSP2-4 .5, SSP3-7 .0, and SSP5-8 .5. The results project a southward expansion of Triatoma gerstaeckei and a northwards expansion of Triatoma sanguisuga. Table of Contents Acknowledgments 3 Abstract 4 Chapter 1 . Introduction 7 1.1 Chagas Disease .......................... 7 Chapter 2 . Implementation 9 2.1 Sources of Data .......................... 9 2.1.1 Historical Observation Data ................ 9 2.1.2 Bio-Climatic Data ..................... 11 2.1.2.1 Land Cover Data ................. 11 2.2 Model Selection .......................... 12 2.2.1 Maximum Entropy ..................... 12 2.2.2 Bayesian Linear Regression ................ 13 2.2.3 Bayesian Neural Network ................. 15 Chapter 3 . Results 17 3.1 Historic Spatial Extent ...................... 17 3.1.1 Maximum Entropy ..................... 17 3.1.2 Bayesian Linear Regression ................ 20 3.1.3 Bayesian Linear Neural Network ............. 22 3.2 Future Projections ......................... 24 Chapter 4 . Analysis 32 4.1 Discussion ............................. 32 4.2 Future Work ............................ 34 Appendices 35 Appendix A. Glossary of Input Variables 36 Appendix B. Historic Habitat S 39 Appendix C. Future Projections of Habitat Suitability 42 Bibliography 52 Vita 57 Chapter 1 Introduction Climate change is already impacting vector-borne disease [1] . It is established that over the next century the planet will experience more damaging and more frequent extreme weather event

FAQs

Which machine learning models are compared for habitat suitability of Triatominae?

The work compares Maximum Entropy (MaxEnt), a joint-species Bayesian Linear Regression (BLR), and a Bayesian Neural Network (BNN) adapted for both joint and single-species distributions.

How is future habitat suitability projected, and to which time period?

Future projections are produced for 2070–2100 under four CMIP6 scenarios: SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5.

What range changes are projected for specific Triatominae species?

The results project a southward expansion of Triatoma gerstaeckei and a northward expansion of Triatoma sanguisuga.

Aug 5, 2026

Upload by Aria