Overview

• The team trained a machine learning model on 2006–21 outbreak data and tested it on 2022–23 cases to identify key risk factors.
• The coldest recorded temperature in autumn was the single most influential predictor of outbreak likelihood, with effects varying widely by region.
• Below-average vegetation density between October and December and reduced water levels in lakes and ponds from January to March were linked to lower outbreak risk.
• Regions with established mute swan (Cygnus olor) populations showed a higher probability of highly pathogenic avian influenza occurrence.
• The authors recommend integrating these environmental and wildlife indicators into region-specific surveillance programs to boost early warning capabilities.