We analyze synoptic observational data on extreme daily minimum temperatures in winter months over Ireland from 1950–2022. We model the tail of the marginal distributions of extreme winter minima using a generalized Pareto distribution capturing temporal and spatial sources of non‐stationarity. We disentangle long‐term climate trends from obfuscating shorter‐term (month‐to‐winter long) large fluctuations, caused by, for example, anomalous behavior of the jet stream. We identify extreme spatial events with respect to a carefully chosen risk function and fit an ‐Pareto process to extreme events exceeding a high‐risk threshold. We show that long‐term trends over Ireland of extremely cold winter temperatures are warming at a faster rate than both mean winter temperatures and summer extreme daily maximum temperatures. Critically, we show that if we did not account for large‐scale, short‐term climatic oscillations, we would incorrectly estimate that the extremely cold winter temperatures were getting colder. In terms of spatial extreme events, we find that in periods where the jet stream typically produces the coldest winter temperatures, the estimated rate at which temperatures could fall below the coldest value ever recorded has decreased by a factor of 100 over the study period.