Nearterm prediction of impactrelevant extreme te

Authors: H M Hanlon G C Hegerl S F B Tett D M Smith

Publish Date: 2014/08/12

Volume: 132, Issue: 1, Pages: 61-76

PDF Link

Abstract

A previous study of predictability of European temperature indices revealed significant skill in predictions of 5/10year average indices of summer mean and maximum 5day average temperatures based on daily maximum and minimum temperatures for a large area of Europe particularly in the Mediterranean Here this work is extended to study indices relevant to high heatrelated impacts on energy use human health and maize yields in Europe The skill of predictions of these indices is assessed using decadal predictions of the number of days above critical thresholds of daily maximum mean and minimum Summer temperatures Following comparison of these predictions with observed conditions there is skill found in parts of Europe where the decadal predictions exceed that of using observed climatology and persisting present conditions Areas in the Mediterranean show the most skill in nearterm predictions while skill is small in Northern/Central Europe There is even some evidence of skill on small scales This system is determined to be not appropriate for predicting indices in the UK as the model significantly overestimates the trend in these indices A further test studies the effect of initialising the decadal forecasts with observations Simulations that include external forcing such as greenhouse gas increases show better skill in predicting changes in the frequency of hot events than those that do not and the initialisation of forecasts with the model used here does not improve this skillExtreme temperature events such as the 2003 European Heatwave Schär et al 2004 Fink et al 2004 and 2010 Russian heatwave Barriopedro et al 2011 have had a severe impact on society and nature in particular the impact on human health was profound In terms of impacts it is not extreme seasonal temperatures but increased daily extreme temperatures which are most damaging Díaz et al 2006 Fouillet et al 2006 Grize et al 2005 Pascal et al 2006 Impacts of daily extreme temperatures are now even more important to study It has been shown that daily extreme temperatures show an upward trend in mean summer daily maximum Tmax and daily minimum temperatures Tmin in Europe over the past few decades that has been attributed in part to human influences Christidis et al 2012 and similarly that the frequency Morak et al 2013 and intensity Zwiers et al 2011 of extreme temperatures has increasedIn light of how severe the impact of high temperatures can be and because these events may become more frequent in the future it has become even more important to determine how well we can predict the changing likelihood of such events to enable improved capability for adaptation and planning for the future Hamilton et al 2012 found seasonal forecasts of the number of daily extreme temperatures outside the 10–90 range had significantly better skill than persistence though lower than the skill in predicting the seasonal mean especially in the extratropics The summer season was the most skillful in the northern hemisphere A recent study by Eade et al 2012 demonstrated significantly skillful predictions of moderate 1 in 10 temperature extremes on decadal timescales especially for multiyear periods These assessments of skill were performed using the Spearman rank correlation coefficient and standardised root mean square errorIn this study we build upon the work of Hanlon et al 2013a who found significant skill beyond observed climatology using the mean square skill score Murphy 1988 in predicting the summer average and hottest 5day average daily maximum Tmax and daily minimum Tmin temperatures in Europe with the Met Office Hadley Centre decadal prediction system DePreSys Hanlon et al 2013a determined that this skill is due almost entirely to the forecast recreating the climate change signal rather than from its initialisation Subsequently work shown in Hanlon et al 2013b found similar results using four CMIP5 models CanCM4 HadCM3 MIROC5 and MPIESMLR However for some models there was evidence of improved skill when initialising with observations In particular the Max Planck Institute Earth System Model MPIESMLR Raddatz et al 2007 Marsland et al 2003 shows skill originating both from the external forcing due to climate change and from initialisation with observationsFollowing these previous studies we apply the same methodology as Hanlon et al 2013a comparing decadal hindcasts to observations and assessing the accuracy of them compared to observed climatology and persistence However here we use indices which are based on exceedance of temperature thresholds which are more directly relevant to impacts and this paper investigates the usefulness of such decadal predictions for informing adaptation decisionsThe procedure applied follows the methodology described in Hanlon et al 2013 now termed H13 That paper presented a method to calculate decadal predictions for climate indices from climate model simulations and evaluates them based on a comparison to alternative prediction methods Here this methodology is adapted to assess the predictability of more impactrelevant indices which count the number of days above a critical threshold

Keywords: