Machine Learning Detection of Heat Stress in Urban Nesting Birds

Abstract

Introduction: Urbanization and increasing temperatures pose serious problems to birds that nest in urban areas especially during times of high temperatures. Heat stress can influence the behaviors and physiology of the birds which makes the identification of such issues an urgent need for urban biodiversity management. Traditional techniques are expensive and cannot offer a continuous analysis of the problem. This paper focuses on the implementation of machine learning (ML) techniques to identify heat-stressed birds in urban areas.

Materials and Methods: A supervised machine learning model was created based on environmental and behavioral characteristics of the nesting birds. Environmental factors considered in the study include air temperature, humidity, heat index, sunlight, nest temperature, and observation time, while the behavioral factors include nest occupancy, activity level, mouth opening, wing flapping, and any other observable signs of heat-related stress. Data preprocessing was conducted using methods such as missing value treatment, outlier detection, feature scaling, and categorical encoding. Logistic Regression, Decision Tree, Support Vector Machine (SVM), and Random Forest algorithms were compared using accuracy, precision, recall, and F1 score metrics.

Results: Performance of various classification algorithms were different based on evaluation criteria. Performance of Logistic Regression model was 82.4%, whereas Decision Tree and SVM were 85.6% and 87.8%, respectively. Higher performance was achieved using Random Forest model with 91.2% accuracy, 90.6% precision, 91.0% recall, and 90.8% F1-score. Highest performance was achieved with the proposed ML framework with 93.5% accuracy, 92.8% precision, 93.1% recall, and 92.9% F1-score. It can be concluded that combination

Conclusion: The experiment shows that machine learning can be used as an effective tool in identifying heat stress among nesting birds in urban areas. The proposed model presents the computational foundation for automated detection of heat stress. This can be applied in management and conservation of urban wildlife. Further studies on the same should include the use of large multi-species data sets and seasonal observations.