Strongly increasing heat extremes in the Middle East and North Africa (MENA) in the 21st century | SpringerLink

The ensemble results of CMIP5 climate models that applied the RCP4.5 and RCP8.5 scenarios have been used to investigate climate change and temperature extremes in the Middle East and North Africa (MENA). Uncertainty evaluation of climate projections indicates good model agreement for temperature but much less for precipitation. Results imply that climate warming in the MENA is strongest in summer while elsewhere it is typically stronger in winter. The summertime warming extends the thermal low at the surface from South Asia across the Middle East over North Africa, as the hot desert climate intensifies and becomes more extreme. Observations and model calculations of the recent past consistently show increasing heat extremes, which are projected to accelerate in future. The number of warm days and nights may increase sharply. On average in the MENA, the maximum temperature during the hottest days in the recent past was about 43 °C, which could increase to about 46 °C by the middle of the century and reach almost 50 °C by the end of the century, the latter according to the RCP8.5 (business-as-usual) scenario. This will have important consequences for human health and society.

The online version of this article (doi:10.​1007/​s10584-016-1665-6) contains supplementary material, which is available to authorized users.

1 Introduction

Even if climate change in the 21st century will be limited to a global mean temperature increase of 2 °C relative to pre-industrial times, warming over land is typically stronger than over the oceans and extreme temperatures in many regions can increase well beyond 2 °C (Seneviratne et al. 2016). The Middle East and North Africa (MENA) are expected to be strongly affected by climate warming, enhancing the already hot and dry environmental conditions (Sanchez et al. 2004; Fang et al. 2008; Giorgi and Lionello 2008; Önol and Semazzi 2009; Lelieveld et al. 2012; Almazroui 2013: Önol et al. 2014; Basha et al. 2015; Ozturk et al. 2015). Assessments of past climate trends in the MENA often suffered from restricted availability of meteorological data sets, and hence are associated with low confidence (IPCC 2013). Several international workshops have been organized to help improve data access, ensure data quality and analyze climate indices (Zhang et al. 2005; Donat et al. 2014).

Based on the available data significant upward temperature trends since the 1970s have been derived for parts of the MENA and Mediterranean Europe (AlSarmi and Washington 2011; Almazroui et al. 2012; Tanarhte et al. 2012; Zarenistanak et al. 2014), accompanied by an increasing number of warm days and high temperature extremes (Kuglitsch et al. 2010; Marofi et al. 2010; Efthymiadis et al. 2011; Donat et al. 2014; Simolo et al. 2014; Tanarhte et al. 2015). While rainfall trends in Mediterranean Europe are significant and predominantly negative, in the MENA they are generally not significant, partly related to the difficulty of establishing representative precipitation measurement networks in this arid region (Xoplaki et al. 2004; Hoerling et al. 2012; Tanarhte et al. 2012; Norrant-Romand and Douguedroit 2014; Ziv et al. 2014).

Analysis of long-term temperature data suggests that since the 1970s the frequency of heat extremes has increased in the MENA (Tanarhte et al. 2015). Based on health statistics and meteorological data Lubczyńska et al. (2015) identified a clear relationship between high temperatures and cardiovascular mortality by cerebrovascular disease, ischemic and other heart diseases, consistent with similar investigations in other regions (Basu and Samet 2002; Gosling et al. 2009). Heat extremes are recognized as the major weather-related cause of premature mortality (McMichael et al. 2006; Kovats and Hajat 2008; Gosling et al. 2009; Peterson et al. 2013), hence their increase in the MENA is of great concern (Lelieveld et al. 2014; Zittis et al. 2015). Heat stress can also cause substantial loss of labor productivity (Dunne et al. 2013; Zander et al. 2015). Further, it is argued that climate change induced weather extremes can impact human security and migration (Barnett and Adger 2007; Piguet et al. 2010; IPCC 2014). Thus heat stress has direct health consequences, while social, economic and political contexts are also important. Both perspectives are relevant for the MENA.

One of the difficulties in the assessment of temperature related climate impacts is the definition of hot weather extremes. Therefore, it is recommended to apply a range of climate indices such as heat wave frequency and warm spell duration (WMO 2009). Zittis et al. (2015) showed that the probability density distributions of daytime maximum temperatures in the warm temperate climate regime north of the Mediterranean are typically wider than in the arid areas to the south. In the latter case the extreme values are comparatively close to the median and mean of the maximum temperature distribution, so that even a moderate rate of warming can lead to the exceeding of heat wave thresholds.

Motivated by the demand for information about regional climate trends, we present projected changes in summertime hot weather conditions in the MENA during the 21st century based on the ensemble output of climate models that participated in the Coupled Model Intercomparison Project Phase 5, CMIP5 (Taylor et al. 2012; Sillmann et al. 2013a, 2013b). We evaluate CMIP5 model uncertainties for the MENA by comparing with observations, and also based on the robustness metric (Knutti and Sedlacek 2012), and show that the models consistently project strong temperature changes whereas precipitation projections are much less consistent. This rationalizes the present focus on temperature extremes, whereas for precipitation higher resolution regional downscaling will be more appropriate. For the latter we refer to the Coordinated Regional Climate Downscaling Experiment framework for the region, MENA-CORDEX (Giorgi et al. 2009; Zittis et al. 2014a; Almazroui 2015).

2 MENA description

The definition of the MENA, applied here, encompasses the larger region between Morocco and Iran, including all Middle Eastern and Maghreb countries, sometimes indicated as Greater Middle East (http://​en.​wikipedia.​org/​wiki/​MENA). The 29 countries included are Algeria, Armenia, Azerbaijan, Bahrain, Cyprus, Djibouti, Egypt, Georgia, Iran, Iraq, Israel, Jordan, Kuwait, Lebanon, Libya, Mauretania, Morocco, Oman, Palestine, Qatar, Saudi Arabia, Somalia, Sudan, Syria, Tunisia, Turkey, United Arab Emirates, Western Sahara and Yemen, with a current population of about 550 million people. While focusing on these countries, we will likewise present climate model output for the MENA-CORDEX area, which includes part of Central Africa, also addressed within Africa-CORDEX (http://​www.​cordex.​org).

Notice that the domain of Africa-CORDEX does not fully cover the Anatolian Peninsula and the Middle East. Because the climate of Central Africa deviates strongly from that in the MENA, being humid – monsoonal rather than hot and arid, respectively, these results are not analyzed in the present context. To evaluate the CMIP5 model data for the MENA we define four subregions that are climatically coherent (Fig. 1, top panel), based on the Köppen-Geiger climate type classification (Kottek et al. 2006). Geographical coordinates of the subregions are the following. A: 9.25 W-12.0E, 30.0–39.0 N; B: 12.0–34.0E, 25.0–33.0 N; C: 25.5–46.5E, 33.0–41.0 N; D: 34.0–60.0E, 18.0–33.0 N. In Section 4 we compare the CMIP5 model output with temperature observations in these subregions. The Supplementary Information includes a Köppen-Geiger analysis of the CMIP5 calculations for recent, mid- and end-century conditions in the MENA, showing that the main climate types are not expected to change much in the 21st century, i.e., remaining hot and arid (Figure S1).Fig. 1

CMIP5 multi-model mean monthly temperature for four MENA subregions compared to observations during the reference period 1986–2005 collected in the CRU and UDEL datasets. Model standard deviations and mean correlation coefficients between model and measurement data are indicated

3 Data and methods

From the CMIP5 output we applied results from an ensemble of 26 models that have been interpolated to a common spatial resolution of 2.5 degrees, about 280 km at the equator (Table S1) (Taylor et al. 2012; Sillmann et al. 2013a, 2013b). Here we use CMIP5 model results for two Representative Concentration Pathways (RCPs), RCP4.5 and RCP8.5, adopted by the IPCC for its fifth Assessment Report (IPCC 2013). Greenhouse gas emissions according to RCP4.5 are assumed to peak around 2040 and then decline, while in RCP8.5 emissions continue to rise throughout the 21st century and hence represent a business-as-usual scenario. Following IPCC (2013) we compare the mid-century period 2046–2065 and the end-century period 2081–2100 with the recent period 1986–2005, the latter used as reference.

We evaluate the CMIP5 model results for the reference period, based on observations compiled in the gridded datasets of the Climate Research Unit (CRU, version 3.22) (Mitchell and Jones 2005) and the University of Delaware (UDEL, version 3.01) (Willmott and Matsuura 1995), also used and inter-compared by Tanarhte et al. (2012). To characterize changes in temperature conditions and hot weather extremes, we follow the procedure of Sillmann et al. (2013a, 2013b) by applying the climate indices defined by the Expert Team on Climate Change Detection and Indices (ETCCDI) of the World Meteorological Organization (WMO) (Klein Tank et al. 2009; Zhang et al. 2011) to the MENA, i.e., over the area of the 29 countries listed in Section 2, as well as the subregions defined in Fig. 1. The indices are calculated for the RCP4.5 and 8.5 scenarios, and are available from the ETCCDI archive at http://​www.​cccma.​ec.​gc.​ca/​data/​climdex/​index.​shtml. The CMIP5 model output is compared with historical observations in the period 1950–2010 from the HadEX2 data set of the UK Met Office Hadley Centre. Monthly and annual indices, based on selected high-quality station observations in 14 MENA countries, have been obtained from http://​www.​climdex.​org/​index.​html (Donat et al. 2013, 2014).

To test the level of consistency of the climate model projections we performed robustness calculations as proposed by Knutti and Sedlacek (2012); see also Proestos et al. (2015). The projection maps of near-surface (2-m) temperature and precipitation for the RCP4.5 and RCP8.5 scenarios are overlaid by stippling and cross-hatching to indicate robustness R, based on the signal-variability ratio and the probability skill score, adopted from weather forecasting applications (Weigel et al. 2007). The signal-variability ratio is calculated from the CMIP5 model spread and the projected change in a particular parameter such as temperature or precipitation. R = 1 – X1/X2, in which X1 is the integral of the squared area between cumulative density functions of individual and multi-model mean projections, and X2 the integral of the squared area between cumulative density functions of the multi-model projection and the historical reference period. A value of R = 1 indicates absolute model agreement.

4 CMIP5 model results

Figure 1 shows a comparison between observations and CMIP5 model output of the ensemble mean monthly temperature, including the standard deviation of all model results. The spatial correlation between CMIP5 and observations is high, especially with the CRU dataset. In all four regions the agreement between model and observations is very good for the summer months (June–August), being the focus of the present work. For the winter, typically October to February, differences can be up to several degrees. In regions B and D the CMIP5 ensemble mean appears to be cold-biased and in region C warm-biased during these months.

Figure 2 presents the projected changes in near-surface temperature, including R, for the mid- and end-century periods and the RCP 4.5 and 8.5 scenarios. The interested reader is referred to Figure S2 for the accompanying results for precipitation. We present projections for winter, i.e., the wet season (December, January, February – DJF) and summer, i.e., the dry season (June, July, August – JJA). While our focus is on the summer, we also present results for the winter to emphasize the seasonal differences in warming trends. This illustrates that the rate of warming is much higher in summer. Please notice the temperature scale differences between the two scenarios. Figure 2 shows that the level of robustness is high, especially for the summer warming, and is highest for the end-of-century projections, in particular for the RCP8.5 scenario, indicating overall agreement among the models. The robustness for precipitation projections, on the other hand, is generally much lower (Figure S2).Fig. 2

Multi-model mean and robustness (dots R ≥ 0.85, and cross-hatching 0.5 ≤ R