Quarterly Journal of the Royal Meteorological Society Q. J. R. Meteorol. Soc. (2014) DOI:10.1002/qj.2379 Prospects for decadal climate prediction in the Mediterranean region Virginie Guemas, a,b * Javier Garc´ ıa-Serrano, a,c Annarita Mariotti, d Francisco Doblas-Reyes a,e and Louis-Philippe Caron a,f a Climate Forecasting Unit, Institut Catal` a de Ci` encies del Clima, Barcelona, Spain b Groupe de M´ et´ eorologie Grande Echelle et Climat, Groupe d’Etude de l’Atmosph` ere M´ et´ eorologique, Centre National de Recherches M´ et´ eorologiques, Toulouse, France c VARCLIM-TG, LOCEAN-IPSL, Universit´ e Pierre et Marie Curie, Paris, France d National Oceanic and Atmospheric Administration, Climate Program Office, Silver Spring, MD, USA e Instituci´ o Catalana de Recerca i Estudis Avanc ¸ats, Barcelona, Spain f Rossby Center, Swedish Meteorological and Hydrological Institute, Norrk¨ oping, Sweden *Correspondence to: V. Guemas. Institut Catal ` a de Ci` encies del Clima, Carrer Trueta 203, 08005 Barcelona, Spain. E-mail: virginie.guemas@ic3.cat The Mediterranean region stands as one of the most sensitive to climate change, both in terms of warming and drying. On shorter time-scales, internal variability has substantially affected the observed climate and in the next decade might enhance or compensate long-term trends. Here we compare the multi-model climate predictions produced within the framework of the CMIP5 (Coupled Model Intercomparison Project Phase 5) project with historical simulations to assess the level of multi-year climate prediction skill in the Mediterranean region beyond that originating from the model accumulated response to the external radiative forcings. We obtain a high and significant skill in predicting 4-year averaged annual and summer mean temperature over most of the study domain and in predicting precipitation for the same seasons over northern Europe and sub-Saharan Africa. A lower skill is found during the winter season but still positive for temperature. Although most of this high skill originates from the model response to the external radiative forcings, the initialization contributes to the temperature skill over the Mediterranean Sea and surrounding land areas. The high and significant correlations between the observed Mediterranean temperatures and the observed Atlantic multidecadal oscillation (AMO) in the summer and annual means are captured by the CMIP5 ensemble which suggests that the added skill is related to the ability of the CMIP5 ensemble to predict the AMO. Such a link to the AMO seems restricted to western Africa and summer means only for the precipitation case. Key Words: climate prediction; initialization; Atlantic multidecadal oscillation; Mediterranean climate Received 22 June 2013; Revised 24 January 2014; Accepted 24 March 2014; Published online in Wiley Online Library 1. Introduction The Mediterranean region is projected to be among the most heavily affected by the twenty-first century greenhouse gas induced climate change, with significant regional warming and drying by the end of the century (e.g. Mariotti et al., 2008). In the shorter term, a few decades into the future, decadal variability of both internal and external origin will largely determine the conditions to be experienced in the region, as already seen over the course of the twentieth century (Mariotti and Dell’Aquila, 2012). A key question and one of great societal relevance, is to what extent will natural variability enhance or reduce externally forced changes and for how long. In view of the seemingly robust climate change signal in the Mediterranean region, projected external forcing constitutes a source of decadal predictability for this region. What remains to be understood is the degree of predictability of regional internal decadal variability and its impact on our ability to predict future changes a few years ahead. Early potential predictability studies highlighted the Atlantic Ocean as the most promising region for decadal climate prediction (Griffies and Bryan, 1997; Boer, 2000; Pohlmann et al., 2004; Collins et al., 2006). The Atlantic Ocean exhibits strong multidecadal variability which involves variations in the strength of the ocean thermohaline circulation associated with a large-scale signature in sea-surface temperatures (SST) referred to as the Atlantic multidecadal oscillation (AMO: Kerr, 2000; Knight et al., 2005, 2006). Much of the current quest to predict future decadal climate variations hinges on the potential predictability associated with the variations of the ocean gyres and the Atlantic meridional overturning circulation (AMOC). At the edge between seasonal forecasting and climate change projections, the decadal climate prediction exercise consists of initializing a climate c  2014 Royal Meteorological Society