Machine Learning Prediction of Thermal Stress in Native Birds During Heatwaves

Abstract

Introduction: Heatwaves have become an important environmental stressor that affects the physiology and behavior of native birds. Thermal stress identification during heatwaves is difficult since the reactions of birds are affected by many interacting environmental and biological factors. This research aims to present a model based on machine learning to predict thermal stress of native birds in heatwaves.

Materials & Methods: The suggested model incorporates environment parameters, birds’ attributes, behavior signals, and physiological responses. The chosen variables are air temperature, relative humidity, solar radiation, wind speed, heat wave length, species, body weight, age category, seeking shade, wing spread, panting, level of activity, body temperature, and respiratory rate. The data set is split into training, validation, and testing partitions in a 70:15:15 proportion. Three models such as Random Forest (RF), Support Vector Machine (SVM), and XGBoost are designed and compared. The performance of these models is evaluated by the means of Matthews Correlation Coefficient, Cohen’s Kappa, Log Loss, and Balanced Error Rate.

Results: The results from the experiment show that XGBoost has the highest accuracy in terms of predictions with an MCC score of 0.78, Cohen’s Kappa of 0.76, Log loss of 0.41, and Balanced Error rate of 0.15. The Random Forest model is second to XGBoost in terms of predictions while the SVM model shows comparatively poor accuracy in terms of predictions. The findings demonstrate that combining environmental, behavioural, and physiological indicators improves the identification of thermal-stress conditions.

Conclusion: Consequently, it can be observed that the use of machine learning methods in the prediction of thermal stress in native birds under heatwaves is validated through this study. XGBoost proves to be quite relevant for the modelling of the intricate relationship between heatwaves and bird behaviour. The suggested model can help identify thermal stress at an early stage and monitor wildlife based on climate change. Future research can extend the approach using larger multi-species datasets and real-time sensor observations.