Regional trend changes in recent surface warming

Abstract

It has been argued in the literature that global temperature increased at a reduced rate between approximately 1998 and 2013, a phenomenon known as the ‘global warming hiatus’. Statistically motivated studies searching for numerical evidence for this episode typically argue against the detectability of a slowdown. At the same time, process-oriented studies on potential causes of the slowdown are more focused on establishing the consistency between observations and model-based expectations of surface warming. Here, we employ three different gridded temperature data sets, two different statistical tests, and four different sets of periods for defining a baseline period and slowdown period associated to the recent slowdown in surface warming to provide strong evidence against consistent regional patterns of significant changes in the warming trend. Roughly half of the earth’s surface has experienced a reduced warming trend during the slowdown period, which is strongly connected to the Interdecadal Pacific Oscillation as a main source of internal variability of the climate system. This finding is consistent with the understanding of the role of internal decadal variability of the climate obtained from modelling and attribution studies. However, controlling for false discovery rates in the context of multiple testing in the spatial domain, we found that only less than 5% of the earth’s surface have experienced a significant regional slowdown in surface warming. At the same time, for roughly the same small number of grid cells we found the opposite pattern of a significant acceleration in warming. In both cases, the identified areas are not robust against the choice of temperature data set and statistical test, as well as against the delineation of the testing periods for baseline and slowdown periods. Our results demonstrate that similar to previous findings at the global level, statistical testing accounting for the multiple testing nature of the problem argues against a robust detectability of the ‘warming hiatus’ at the regional level.