Record-breaking statistics detect islands of cooling in a sea of warming

Abstract. Record-breaking statistics are combined here with a geographic mode of exploration to introduce a record-breaking map. We examine time series of sea surface temperature (SST) values and show that high SST records have been broken far more frequently than the expected rate for a trend-free random variable (TFRV) over the vast majority of oceans (83 % of the grid cells). This, together with the asymmetry between high and low records and their deviation from a TFRV, indicates SST warming over most oceans, obtained using a distribution-independent, robust, and simple-to-use method. The spatial patterns of this warming are coherent and reveal islands of cooling, such as the “cold blob” in the North Atlantic and a surprising elliptical area in the Southern Ocean, near the Ross Sea gyre, not previously reported. The method was also applied to evaluate a global climate model (GCM), which reproduced the observed records during the study period. The distribution of records from the GCM pre-industrial (PI) control run samples was similar to the one from a TFRV, suggesting that the contribution of a suitably constrained internal variability to the observed record-breaking trends is negligible. Future forecasts show striking SST trends, with even more frequent high records and less frequent low records.