Environmental Research Letters TOPICAL REVIEW • OPEN ACCESS Recent citations A fluctuation in surface temperature in historical - Influence and seepage: An evidence- resistant minority can affect public opinion context: reassessment and retrospective on the and scientific belief formation evidence Stephan Lewandowsky et al To cite this article: James S Risbey et al 2018 Environ. Res. Lett. 13 123008 View the article online for updates and enhancements. This content was downloaded from IP address 193.174.18.1 on 01/03/2019 at 11:10 Environ. Res. Lett. 13 (2018) 123008 https://doi.org/10.1088/1748-9326/aaf342 TOPICAL REVIEW A fluctuation in surface temperature in historical context: OPEN ACCESS reassessment and retrospective on the evidence RECEIVED 8 February 2017 James S Risbey1 , Stephan Lewandowsky1,2 , Kevin Cowtan3 , Naomi Oreskes4 , Stefan Rahmstorf5,6 , REVISED Ari Jokimäki7 and Grant Foster8 9 November 2018 1 CSIRO Oceans & Atmosphere, Hobart Tas, Australia ACCEPTED FOR PUBLICATION 2 23 November 2018 University of Bristol, Bristol, UK and University of Western Australia, Crawley, WA, Australia 3 Department of Chemistry, University of York, York, United Kingdom PUBLISHED 4 Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA, United States of America 19 December 2018 5 Potsdam Institute for Climate Impacts Research, Potsdam, D14473, Germany 6 University of Potsdam, Potsdam, Germany Original content from this 7 Independent researcher, Finland work may be used under 8 1 Noyes Ct, Augusta, ME 04330, United States of America the terms of the Creative Commons Attribution 3.0 E-mail: [email protected] licence. Any further distribution of Keywords: climate variability, climate trends, temperature fluctuation, pause hiatus this work must maintain attribution to the author(s) and the title of the work, journal citation Abstract and DOI. This work reviews the literature on an alleged global warming ‘pause’ in global mean surface temperature (GMST) to determine how it has been defined, what time intervals are used to characterise it, what data are used to measure it, and what methods used to assess it. We test for ‘pauses’, both in the normally understood meaning of the term to mean no warming trend, as well as for a ‘pause’ defined as a substantially slower trend in GMST. The tests are carried out with the historical versions of GMST that existed for each pause-interval tested, and with current versions of each of the GMST datasets. The tests are conducted following the common (but questionable) practice of breaking the linear fit at the start of the trend interval (‘broken’ trends), and also with trends that are continuous with the data bordering the trend interval. We also compare results when appropriate allowance is made for the selection bias problem. The results show that there is little or no statistical evidence for a lack of trend or slower trend in GMST using either the historical data or the current data. The perception that there was a ‘pause’ in GMST was bolstered by earlier biases in the data in combination with incomplete statistical testing. ‘Nowadays any reference to polywater industrial revolution is imposing climate changes on is always tinged with ridicule, but ten timescales from decadal to centennial, and ultimately years ago many competent and experi- much longer too as the oceans and cryosphere respond enced scientists were quite convinced to the changes in Earth’s energy balance (Hansen et al of its reality. I can see no reason why 1985, Houghton et al 2001). The detection and the scientific and sociological issues attribution of greenhouse climate change (Mitchell raised by this unique episode should be et al 2001) deals with the identification of the ‘signal’ of shrouded in secrecy’—Felix Franks, the forced response to greenhouse gases from the Polywater. ‘noise’ of variability of climate that occurs on the same decadal and multidecadal timescales. The greenhouse climate signal is always accompanied to some degree ‘ ’ ( ) 1. Introduction by noise variation from other forcings of the climate system (such as due to changes in aerosol loading or )( ) The Earth’s climate varies on a vast range of temporal solar variations Marotzke and Forster 2015 and by scales (National Research Council 1982, 1995). The internal variations intrinsic to the coupled climate persistent increase in greenhouse gases since the system (O’Kane et al 2013). © 2018 The Author(s). Published by IOP Publishing Ltd Environ. Res. Lett. 13 (2018) 123008 J S Risbey et al In recent years there have been more than two evidence for the ‘pause’ in the observed GMST record, hundred articles in the climate literature discussing as it is now, and as it was at the time the research was the notion of a ‘pause’ or ‘hiatus’ in greenhouse warm- undertaken. ing that is variously alleged to have taken place some The review provides initial context by describing time in the past couple of decades (Lewandowsky et al temperature fluctuations in the climatological litera- 2016). The form of alleged climate ‘pause’ varies across ture and some issues in constructing observed series of the literature, but essentially involves calculation of a GMST. The consequences of the uncertainties in short-term trend in global mean surface temperature GMST are described for assessment of short-term (GMST) over a decade or two, which is then compared GMST trends. The review then proceeds by providing with either other periods in observed GMST (Stocker a series of retrospective constructions of short-term et al 2013), or with trends estimated from coupled cli- GMST trends on the basis of what was known about mate model projections (Fyfe et al 2013, Risbey et al uncertainty in each of the major GMST series at differ- 2014). This review addresses the former issue (com- ent points in time compared to what is known now parison of observed trends), while a companion about uncertainty in each of these series. This retro- review (Lewandowsky et al 2018) addresses the com- spective analysis provides a framework in which to parison with climate model expectations of trends. assess what was known (or could have been known at When it first emerged, the concept of a global the time) when assessing the evidence for a ‘pause’ in warming ‘pause’ was mostly cast in terms of the obser- global warming. The retrospective (historical) assess- vational record as a period of slower than average ment of trends uses the versions of the GMST data that warming (e.g. Stocker et al 2013). With time, usage existed at the times when researchers carried out their broadened to include a comparison of observed assessments of trend-intervals. warming rates with those inferred from model projec- Because the literature on the ‘pause’ is now so vast, tions. The observations-based view of the ‘pause’ is the review treats the literature primarily as a database perhaps more intuitively accessible, whereas the for statistical assessments. The sets of definitions model-comparison view of the ‘pause’ allows for more implied for the ‘pause’ can be inferred from the pause- complexity in matching variations in the forcing and literature, which provides the range of intervals against the (model simulated) response to that forcing with which to assess potential pauses. We have attempted to observed trends. Neither definition turns out to be summarise some of the key messages and the approach straight-forward in practice. We concern ourselves to statistical methods in this literature, but do not pro- fi ( ) exclusively here with the rst observations-based vide a chronological assessment of individual con- ‘ ’ view of the pause . As such, we do not consider the tributions in the sense more common in reviews. Our role of climate forcing, and we do not conduct any concern is with the definitions, data, and methods fl analysis directed at a causal understanding of uctua- used and their implications for the conclusions drawn. tions in GMST (which would require the use of cli- mate models). We do not discount the worth of the model-comparison view of the ‘pause’, but the issues 2. Climate fluctuations past and present are complex enough that they require separate exam- ination (Lewandowsky et al 2018). While the observa- The field of climatology has long recognized that tions-based view of the ‘pause’ is intuitively appealing climate varies on decadal and longer time scales. The in that one can ostensibly ‘see’ a slowing of warming concept of a ‘climate normal’ was introduced in the rate in (parts of) the GMST record, mere description is early 20th century as a 30 year record or average of not the same as statistical evidence. The complexity climate (Arguez and Vose 2011). The 30 year period here lies in the choices of data, periods, and tests was considered necessary to smooth out at least some employed to quantify whether any part of the record is of the known large decadal-scale variations in climate. indeed unusual. This review attempts to foreground The various GMST datasets have used a 30 year some of those choices and their consequences. ‘climate normal’ period as a baseline against which to The notion that global warming ‘paused’ is now calculate anomalies for similar reasons. The literature entrenched in the journal literature (Stocker et al on climate variability and change has recognized 2013). The ‘pause’ in warming is generally posited in episodes or periods of multidecadal GMST variation this literature as an anomaly about climate that is throughout the 20th century (Handel and Risbey 1992, inconsistent with rising greenhouse gases. Many National Research Council 1995). Thus, the notion of pause-papers commence with the statement that fluctuations in GMST is not new and has been despite rising levels of greenhouse gases, GMST has recognized as a confounding factor in attributing not increased since about 1998 (although the sup- causes of decadal-scale GMST changes in all the IPCC posed start year varies)(Guemas et al 2013, Kosaka reports since their inception in 1990.