Dating hiatuses: a statistical model of the recent slowdown in global warming and the next one

Abstract

Much has been written about the so-called hiatus or pause in global warming, also known as the stasis period, the start of which is typically dated to 1998. HadCRUT4 global mean temperatures slightly decreased over the 1998–2013 period, although a simple statistical model predicts that they should have grown by 0.016 ∘C/yr, in proportion to the increases in the concentrations of well-mixed greenhouse gases (WMGHGs) and ozone. We employ a statistical approach to assess the contributions of model forcings and natural variability to the hiatus. Our point estimates suggest that none of the model forcings explain more than one-third of the missing heat, accounting for the upper bound of the confidence interval on the effect of tropospheric aerosols, which is the most prominent yet most uncertainly measured of the model forcings that could explain the missing heat. The El Niño–Southern Oscillation (ENSO) explains up to about one-third of the missing heat, and two-thirds and possibly up to 81 % is explained by the unusually high temperature of 1998. Looking forward, the simple model also fails to explain the large increases since then (0.087 ∘C/yr from 2013 to 2016). This period coincides with another El Niño, but the ENSO fails to satisfactorily account for the increase. Instead, we propose a semiparametric cointegrating statistical model that augments an energy balance model with a novel multi-basin measure of the oceans' multidecadal temperature cycles. The model partially explains the recent slowdown and explains all of the subsequent warming. The natural cycle suggests the possibility – depending in part on the rate of increase of WMGHG concentrations – of a much longer hiatus over the period from roughly 2023 to 2061, with potentially important implications for policy evaluation.