LAUSR

Global annual-mean surface temperature set a record high in 2016 in at least three observational datasets— GISTEMP (Hansen et al. 2010), HadCRUT4.5 (Morice et al. 2012), and NOAA (Karl et al. 2015)—exceeding the previous record set in 2015 (Fig. 3.1a). In contrast, the last global mean annual cold record occurred around 1910. Record global warmth implies some record warmth on regional scales as well (Kam et al. 2016), which can cause important impacts such as thermal stress, coral bleaching, and melting of sea and land ice (IPCC 2013). Decreased land ice, combined with ocean heat uptake, contributes to sea level rise, which can exacerbate coastal flooding extremes (e.g., Lin et al. 2016). Figure 3.1 compares observed global-mean temperature anomalies with simulations from the Coupled Model Intercomparison Project 5 (CMIP5; Taylor et al. 2012; Table ES3.1). Record warmth in 2016 largely follows a pronounced century-scale warming trend, and was far outside the range of internal (unforced) climate variability sampled across over 24 000 years of CMIP5 Control simulations (Fig. 3.1c). It was also well outside the range of CMIP5 Natural ForcingOnly simulations incorporating solar and volcanic forcing changes (Fig. 3.1b). In contrast, the observed warming lies within the range of CMIP5 All-Forcing simulations that include both anthropogenic and natural forcing (Fig. 3.1a). These results suggest that observed global-mean temperatures emerged from the natural variability background (natural forcing response plus internal variability) around 1980, and have become increasingly detectable since. The inconsistency of obser ved long-term global warming with simulated natural variability (detection), and its consistency with simulations incorporating anthropogenic forcing (attribution), are in agreement with previous studies and assessments (e.g., IPCC 2001, 2007, 2013; Knutson et al. 2013; Kam et al. 2016). Detection and attribution of human inf luence on global mean temperature is wellestablished in the climate sciences, including through more sophisticated approaches than shown here (e.g., regressions or pattern scaling; Bindoff et al. 2013 and references therein). The adequacy of CMIP5 model simulations of internal variability for detection and attribution has also been assessed previously (e.g., IPCC 2013; Knutson et al. 2013, 2016). Figure 3.1d examines shorter term global-mean temperature variability since 1970, highlighting the timing of four major El Niño events and two major volcanic eruptions. The 2015/16 global temperature event appears as a temporary bump with a magnitude (for January–December 2016) of a little over 0.1°C, superimposed on a long-term warming trend of about 1°C—the latter being largely attributable to anthropogenic forcing according to CMIP5 models (Figs. 3.1a,b). While the El Niño events of 1972/73, 1997/98, and 2015/2016 have apparent warming signatures in global temperature, the 1982/83 event’s imprint was apparently muted by the almost-coincident eruption of El Chichón. Monthly maps of observed surface temperature internal climate variability for 2016 are discussed in the online supplement material. From these and previous studies (e.g., Trenberth et al. 2002) we infer that the short-term calendar-year global mean warmth in 2015 and 2016 is likely to have been at least partly AFFILIATIONS: KnuTson, Zeng, and wiTTenbeRg—NOAA/ Geophysical Fluid Dynamics Laboratory, Princeton, New Jersey; Kam—Department of Civil, Construction, and Environmental Engineering, University of Alabama, Tuscaloosa, Alabama, and Cooperative Institute for Climate Science, Princeton University, Princeton, New Jersey