Extended periods of cold, wet and windy conditions, known as 'chill events', rapidly increase potential heat loss in livestock. Such events pose substantial risks to cattle, particularly breeds in northern Australia, that are acclimatised and bred for warmer conditions. Yet due to their rarity, very little is known about chill events in tropical climates, particularly during the dry season (May to October). This study provides an observational analysis of two multi-day chill events over Australia's Northern Territory in June 2007 and September 2021; combined, these events caused the deaths of over 700 cattle at two pastoral stations. Here we investigate the meteorological evolution and biometeorological severity of each event using livestock-specific thermal stress indices: the Cattle Comfort Index (CCI) and Livestock Chill Index (LCI). In both events, unusual warmth gave way to extremely cold conditions (CCI < the long-term 1 st percentile), including high to very high LCI values despite the events occurring during the May to October dry season. We find that established LCI absolute thresholds developed for southern sheep do not adequately reflect chill severity in northern cattle systems, underscoring the need for locally derived percentile-based measures. We also present an illustrative example of the early-warning signals for the September 2021 event that could have been produced using the Bureau of Meteorology’s operational numerical weather prediction system ACCESS-G. Several forecast initialisations indicated a shift towards markedly colder conditions up to 8-10 days in advance. Our results highlight the benefit of integrating agri-climatic indices into forecast products and the need to develop region- and species-appropriate chill thresholds for northern Australia livestock. • Examines two major cattle chill events in tropical northern Australia. • Thermal stress indices are used to quantify the severity of the chill events. • Modern weather models can predict these events up to 10 days in advance. • Index-based forecasts can improve livestock risk management and adaptation.