Machine Learning Evaluation of Air Pollution Effects on Urban Bird Health

Abstract

Urbanization has resulted in increased air pollution levels, creating significant environmental stress for wildlife, particularly urban bird populations. Birds act as sensitive ecological indicators because their health conditions, behavioural patterns, and population distribution are strongly influenced by habitat quality and environmental changes. Continuous exposure to pollutants such as particulate matter, nitrogen dioxide, sulphur dioxide, and carbon monoxide can negatively affect respiratory conditions, habitat suitability, and survival patterns of urban birds. However, understanding the complex relationship between air pollution and bird health remains challenging through conventional ecological assessment approaches. This study presents a machine learning-based ecological assessment framework to evaluate the effects of urban air pollution on bird health. Environmental parameters, including PM2.5, PM10, NO₂, SO₂, CO, temperature, and vegetation availability, were analysed along with bird health indicators such as abundance, species richness, nesting activity, and behavioural responses. Machine learning models, including Support Vector Machine, Random Forest, Gradient Boosting, and Artificial Neural Network, were evaluated to identify relationships between pollution exposure and bird health conditions. The results indicated that increasing air pollution levels were associated with reduced bird health conditions and lower habitat suitability. The model evaluation demonstrated effective prediction capability, with the Artificial Neural Network achieving the highest performance with an accuracy of 94.3%, followed by Gradient Boosting (93.1%) and Random Forest (91.4%). The findings also showed a clear decline in bird health index from 92% under low pollution conditions to 44% under severe pollution exposure. The study demonstrates that machine learning approaches can support ecological assessment by identifying pollution-related risks affecting urban bird populations. Integrating environmental monitoring with predictive models can assist conservation planning, habitat management, and sustainable urban biodiversity protection. Future studies can focus on real-time pollution monitoring and multi-species ecological health prediction.