Abstract
Gaseous emissions from livestock facilities pose environmental and health concerns. Monitoring pollutant gases is essential to mitigate impact and enhance the sustainability of livestock systems. Emerging Artificial Intelligence (AI) technologies-particularly Artificial Neural Networks (ANNs)-offer advanced tools to address these challenges by improving livestock monitoring and management. Following PRISMA guidelines, 18 studies published between 2007 and 2024 were selected from Web of Science(®) and Scopus(®). Most research was conducted in Europe (55%), primarily focusing on cattle and swine. Among gases, ammonia (NH(3)) was predicted in 50% of studies and methane (CH(4)) in 35%. The most common ANN architecture was the Multilayer Perceptron (MLP), trained mainly with backpropagation algorithms and validated using the Root Mean Square Error (RMSE). The results show that ANN models consistently outperformed traditional statistical approaches, offering greater prediction accuracy. Future research should focus on identifying optimal ANN structures for precise emission prediction, accounting for environmental variability, reducing dataset bias, and combining ANN with statistical models to develop hybrid approaches that further improve livestock management and sustainability.