Clim. Past, 8, 1551–1563, 2012 www.clim-past.net/8/1551/2012/ Climate doi:10.5194/cp-8-1551-2012 of the Past © Author(s) 2012. CC Attribution 3.0 License. Constraining the temperature history of the past millennium using early instrumental observations P. Brohan1, R. Allan1, E. Freeman2, D. Wheeler3, C. Wilkinson4,5, and F. Williamson3,4,5 1Met Office Hadley Centre, Exeter, UK 2NOAA/STG Inc., USA 3Sunderland University, Sunderland, UK 4University of East Anglia, Norwich, UK 5Catholic University of Valparaiso, Chile Correspondence to: P. Brohan (philip.brohan@metoffice.gov.uk) Received: 4 April 2012 – Published in Clim. Past Discuss.: 4 May 2012 Revised: 17 August 2012 – Accepted: 6 September 2012 – Published: 11 October 2012 Abstract. The current assessment that twentieth-century tions – supporting their use for longer-term climate recon- global temperature change is unusual in the context of the last structions. However, some of the climate model simulations thousand years relies on estimates of temperature changes in the CMIP5 ensemble show much larger volcanic effects from natural proxies (tree-rings, ice-cores, etc.) and climate than this – such simulations are unlikely to be accurate in model simulations. Confidence in such estimates is limited this respect. by difficulties in calibrating the proxies and systematic dif- ferences between proxy reconstructions and model simula- tions. As the difference between the estimates extends into 1 Introduction the relatively recent period of the early nineteenth century it is possible to compare them with a reliable instrumental es- The temperature history of the past millennium provides timate of the temperature change over that period, provided vital context for predictions of future change, and attri- that enough early thermometer observations, covering a wide butions of recent change to anthropogenic causes (Jones enough expanse of the world, can be collected. and Mann, 2004). Back to about 1850 large-scale tem- One organisation which systematically made observations perature changes are fairly well-known from thermometer and collected the results was the English East India Com- measurements (Brohan et al., 2006), but longer time scale pany (EEIC), and their archives have been preserved in the reconstructions are based on a variety of natural proxies British Library. Inspection of those archives revealed 900 (tree-rings, ice cores, speleothems, etc.) and have a large log-books of EEIC ships containing daily instrumental mea- uncertainty (Fig. 1). surements of temperature and pressure, and subjective esti- The proxy reconstructions not only disagree amongst mates of wind speed and direction, from voyages across the themselves, but they share a reliance on calibration to re- Atlantic and Indian Oceans between 1789 and 1834. Those cent instrumental (thermometer) records. Calibration is a sta- records have been extracted and digitised, providing 273 000 tistical estimate of the scaling factor relating a change in new weather records offering an unprecedentedly detailed the proxy value (ring width etc.) to a change in temper- view of the weather and climate of the late eighteenth and ature, and it is necessary to assume that this scaling fac- early nineteenth centuries. tor does not change with time or temperature. Recent years The new thermometer observations demonstrate that the have seen a lot of research into improved calibration tech- large-scale temperature response to the Tambora eruption niques, but despite these technical improvements, the uncer- ◦ and the 1809 eruption was modest (perhaps 0.5 C). This pro- tainty in the reconstructions remains large (Frank et al., 2010; vides an out-of-sample validation for the proxy reconstruc- Jones et al., 2009). Published by Copernicus Publications on behalf of the European Geosciences Union. 2 BROHAN ET AL: EEIC EARLY OBSERVATIONS under-represent the effects of large volcanic eruptions (Mann et al., 2012). Because the main difference between the reconstructions is in their calibration, the spread in centennial and longer timescale temperatures is highly correlated with the spread in shorter timescale variability: reconstructions that have a larger variability over (say) the first few decades of the nine- 1552teenth century also P. show Brohan larger et variability al.: EEIC over early the observations whole re- construction. So if the reconstructions could be more tightly constrainedTable 1. Proxy over a seriesshort period, (top) and this modellingwould reduce groups the uncer- (bottom) 0.5 B2000 HCA..2006 MSH..2005 MBH1999 taintyproviding over data the whole used in millennium. Figs. 1 and 9. The early nineteenth cen- BOS..2001 ECS2002 O2005 JBB..1998 DWJ2006 MJ2003 RMO..2005 tury is a strong candidate for such a validation period, as it includesB2000 short-termBriffa temperature(2000); Briffa variability et al. ( of2004 uncertain) size 0 associatedBOS..2001 with largeBriffa volcanic et al. ( eruptions2001) (1809 and Tambora in 1815),DWJ2006 and thereD’Arrigo are some et thermometeral. (2006) observations for the period,HCA..2006 which couldHegerl potentially et al. (2006) be used to quantify the ECS2002 Esper et al. (2002); Cook et al. (2004) −0.5 temperature variability much more precisely. Precise infor- mationMJ2003 on temperatureMann changes and Jones over(2003 this) period would also MSH2005 Moberg et al. (2005) Temperature anomaly (C) Temperature resolveO2005 a major disagreementOerlemans between(2005) the GCM simulations, which vary greatly in the simulated magnitude of the 1809 −1 RMO..2005 Rutherford et al. (2005) andMBH1999 Tambora eruptions.Mann et al. (1999) AccurateJBB.1998 instrumentalJones etweather al. (1998 observations, 2001) have been re- coveredbcc-csm1-1 for limited regions Beijing Climate going back Center, well China before Meteorological 1800 (Ca- muffo and Bertolin,Administration 2011; Alcoforado (Wu et et al. al.,, 2012 2012),) so to do 0.5 bcc−csm1−1 MPI−ESM−P thisGISS-E2-R validation is merely NASA a matterGoddard of Institute recovering forSpace enough Studies such FGOALS−gl GISS−E2−R MIROC−ESM CCSM4 observations, covering(Schmidt a large et al. enough, 2006) area of the Earth, to constrainFGOALS-gl the large-scale LASG, temperature. Institute of Atmospheric Physics, Chi- 0 nese Academy of Sciences (Zhou et al., 2008) MIROC-ESM Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research −0.5 2 New instrumentalInstitute weather (The recordsUniversity for of the Tokyo), early and nine- Na- teenth century tional Institute for Environmental Studies Temperature anomaly (C) Temperature (Watanabe et al., 2010) −1 AmongstMPI-ESM-P the archives Max in Planck the British Institute Library for Meteorology (BL) in London are some 4000 logbooks(Raddatz from et al.ships, 2007 in; theMarsland service et ofal., the2003 En-) CCSM4 National Center for Atmospheric Research 800 1000 1200 1400 1600 1800 2000 glish East India Company(Gent et (EEIC); al., 2011 each) recording the details Year and events of a voyage from England to the Indies (usually India, China or both) and back, typically taking the best part Fig. 1. Northern Hemisphere surface temperature estimates for the of two years. The EEIC received its charter from Elizabeth last millennium. Upper panel: recent proxy reconstructions (after Fig. 1. Northern Hemisphere surface temperature estimates for the I intime 1600, scale and temperatures many of its is earliest highly voyages correlated became with the famous spread lastfigure millennium. 6.10 of Jansen Upper panel: et al., recent2007). proxy Lower reconstructions panel: GCM simula-(after in shorter time scale variability: reconstructions that have a tions from CMIP5 (where possible (MIROC-ESM-P & GISS-E2-R) because of their excellent records of new lands; for example figure 6.10 of Jansen et al. (2007)). Lower panel: GCM simula- thatlarger by Henry variability Middleton over (say) to the the Moluccas first few decades in 1604-6 of (Foster, the nine- tionsthese from have CMIP5 been (wherecorrected possible for drift (MIROC-ESM-P by subtracting a& linear GISS-E2-R) trend fit- 2010).teenth These century early also voyages show larger were variability recorded over in diaries; the whole log- re- theseted to have their been unforced corrected parent for experiment). drift by subtracting In each a case linear the trend black fit- line construction. So if the reconstructions could be more tightly shows instrumental observations (Brohan et al., 2006). Data used is books — formally prepared documents of a standard format ted to their unforced parent experiment). In each case the black line constrained over a short period, this would reduce the uncer- listed in Table 1. — did not begin to appear until the 1650s. Their preparation shows instrumental observations (Brohan et al., 2006). Data used is tainty over the whole millennium. The early nineteenth cen- listed in table 1. was part of an officer’s duties until the gradual expansion of thetury Company is a strong in the candidate 1830s into for such a quasi-military a validation and period, politi- as it includes short-term temperature variability of uncertain size An alternative estimate of temperature changes is given cal body responsible for overseeing British interests in In- diaassociated and beyond. with Those large archivedvolcanic eruptionsin the BL, (1809 therefore, and Tambora extend by general circulation model (GCM) simulations, and the in 1815), and there are some thermometer observations for WCRP Coupled Model Intercomparison Project – Phase