Quantifying and predicting extremes in northeastern US temperature

Abstract:

Prediction of heat extremes remains a challenge, despite advances in weather and climate modeling. Here, we analyze 31 years of daily summer temperature data from surface stations in the GHCND archive, and identify regions of the US that experience heat waves synchronously via a cluster analysis. Focusing on the northeastern quarter of the US, a region that contains the major population centers of the eastern seaboard and the majority of the Corn Belt, we find that heat waves are associated with characteristic sea surface temperature (SST) patterns in the extratropical Pacific as early as 40-50 days preceding the heat wave. We perform leave-one-out cross validation to test the skill of predictions based on the Pacific SST patterns and find that they outperform baseline predictions based on persistence and climatology for lead times up to 50 days before a heat wave. Interestingly, other oceanic regions, including the tropics, do not exhibit as clear an association with northeastern heat extremes. The presence of the Pacific anomaly patterns does not have a similarly strong relationship with the median or lower percentiles of the daily temperature distributions, indicating that the increase in extremes associated with the occurrence of the anomaly pattern is not simply a reflection of a uniform upward shift in the temperature distribution.