1. Introduction
The Middle East and southwest Asia are a highly water-stressed region with reduced societal resilience resulting from economic and political challenges. As a result, severe drought in the region can have complex impacts, ranging beyond direct impacts on crops and livestock to an array of indirect impacts associated with sanitation, nutrition, loss of livelihood, displaced populations, and international disputes. As an example, the catastrophic 1999–2001 drought resulted in impacts spanning crop failures; widespread livestock death; significant population migrations; increases in diseases (polio, cholera, diphtheria, typhoid, and tuberculosis); soil and land cover degradation; loss of orchards and fruit trees both as a result of direct drought impacts and through use as fuel; desiccation of internationally important wetlands; increase in household debt, with a disproportionate impact on women and children; and international boundary disputes over both river flows and refugees (Agrawala et al. 2001; Lautze et al. 2002). In terms of standardized precipitation deficits, this regional drought comprised the most severe area of drought in the world during this period (Fig. 1). (Datasets are described in figure captions.) Moreover, future drought impacts may be even worse, as consensus model projections suggest an overall drying trend for much of the region (IPCC 2013a). Understanding the dynamics, predictability, and trends of droughts in this region is, therefore, a critically important problem. This review assesses the current understanding of drought in this sensitive region with regard to both the historical record and future projections and identifies outstanding questions. In addition to drought research, previous results with respect to the region’s climatology, hydrology, vegetation, and precipitation processes that are not focused specifically on drought but are relevant to understanding drought are also surveyed.
The 1999–2001 drought, as measured by the 12-month standardized precipitation index (SPI), calculated based on precipitation from the CPC Merged Analysis of Precipitation (CMAP; Xie and Arkin 1996, 1997), relative to the 1979–2010 period. The red oval denotes the focus area for this review.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
The 1999–2001 drought, as measured by the 12-month standardized precipitation index (SPI), calculated based on precipitation from the CPC Merged Analysis of Precipitation (CMAP; Xie and Arkin 1996, 1997), relative to the 1979–2010 period. The red oval denotes the focus area for this review.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
The 1999–2001 drought, as measured by the 12-month standardized precipitation index (SPI), calculated based on precipitation from the CPC Merged Analysis of Precipitation (CMAP; Xie and Arkin 1996, 1997), relative to the 1979–2010 period. The red oval denotes the focus area for this review.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
The terms “Middle East” and “southwest Asia” both have varying definitions. For the purposes of this study, the region considered extends from the east coast of the Mediterranean Sea through Pakistan, along with the Arabian Peninsula, and includes the following countries (Fig. 2): Israel, Lebanon, Syria, Jordan, Saudi Arabia, Oman, Qatar, the United Arab Emirates, Kuwait, Bahrain, Yemen, Iraq, Iran, Afghanistan, Tajikistan, Uzbekistan, Kyrgyzstan, Turkmenistan, and Pakistan. This area is chosen to be relatively broad while considering an area of generally similar climate (although ranging from Mediterranean climate to semiarid and arid conditions) that is influenced by regionwide drought mechanisms. The region includes an arc of relatively fertile land, the Fertile Crescent, which is indicated schematically on Fig. 2 along with the definitions of Middle East and southwest Asia used in this review.
Countries considered here (in gray) along with approximate definitions of three subdomains often focused on in regional studies (Middle East, in red; Fertile Crescent, in green; and southwest Asia, in blue). Definitions of these regions vary considerably: more expansive definitions of the Middle East include the entire domain; more restrictive definitions focus on the coastal countries. The region as a whole is also sometimes referred to as western Asia.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Countries considered here (in gray) along with approximate definitions of three subdomains often focused on in regional studies (Middle East, in red; Fertile Crescent, in green; and southwest Asia, in blue). Definitions of these regions vary considerably: more expansive definitions of the Middle East include the entire domain; more restrictive definitions focus on the coastal countries. The region as a whole is also sometimes referred to as western Asia.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Countries considered here (in gray) along with approximate definitions of three subdomains often focused on in regional studies (Middle East, in red; Fertile Crescent, in green; and southwest Asia, in blue). Definitions of these regions vary considerably: more expansive definitions of the Middle East include the entire domain; more restrictive definitions focus on the coastal countries. The region as a whole is also sometimes referred to as western Asia.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
As discussed further in the next section, the climate of the region is generally semiarid to very arid, including several deserts, but annual precipitation does exceed 60 cm along the eastern Mediterranean region and on many of the mountain slopes in the region and ranges as high as 180 cm in a small region on the southern coast of the Caspian Sea. Over much of the region, precipitation primarily comes during the cold season (November–April) from synoptic storms, although there are exceptions, including the importance of the summer monsoon in Pakistan and southeastern Afghanistan and the extension of the African summer monsoon/intertropical convergence zone (ITCZ) into the southern coast of the Arabian Peninsula. Because of its common importance for most of the region, drought variability associated with cold season precipitation processes is a particular focus of this review.
Large-scale and subsistence farming as well as pasturing are common throughout the region (e.g., Ryan et al. 2012; Ramankutty et al. 2008), so the region’s precipitation, though modest in many areas, is very important. The combined effects of water scarcity (Oki and Kanae 2006) and frequent drought over the Middle East and central-southwest Asia (Mishra and Singh 2010) affect crop yields and the regional economy (Kaniewski et al. 2012), which increases the potential for the loss of life and property (Agrawala et al. 2001).
Despite the aridity of the region, it is generally well populated outside of the highest mountains and desert regions, with an estimated population density (Fig. 3, top) similar to the southeast United States or South Africa. Within the region, there are areas of high poverty and societal vulnerability; one measure of poverty, infant mortality, is shown in Fig. 3, bottom. This vulnerability is particularly evident during severe droughts, as noted above. Moreover, given the long-standing sociopolitical tensions in several areas of the region in an environment of limited resources, drought variability may raise the risk of regional conflicts (El Kharraz et al. 2012; Selby and Hoffmann 2012; Fröhlich 2013; Gleick 2014; Kelley et al. 2015). In terms of the impacts of drought, therefore, the region is particularly important because of the intersection of population, vulnerability, drought severity, and potential aridification under climate change.
(top) Population count and (bottom) poverty estimates. Estimated infant mortality rate (infant deaths per 10 000 live births) is used here as a measure of poverty. The unit of population is number of persons and the unit of infant mortality rate is infant deaths per 10 000 live births. Both the population data and the infant mortality rate are from the Center for International Earth Science Information Network (CIESIN), Columbia University (Center for International Earth Science Information Network 2005).
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
(top) Population count and (bottom) poverty estimates. Estimated infant mortality rate (infant deaths per 10 000 live births) is used here as a measure of poverty. The unit of population is number of persons and the unit of infant mortality rate is infant deaths per 10 000 live births. Both the population data and the infant mortality rate are from the Center for International Earth Science Information Network (CIESIN), Columbia University (Center for International Earth Science Information Network 2005).
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
(top) Population count and (bottom) poverty estimates. Estimated infant mortality rate (infant deaths per 10 000 live births) is used here as a measure of poverty. The unit of population is number of persons and the unit of infant mortality rate is infant deaths per 10 000 live births. Both the population data and the infant mortality rate are from the Center for International Earth Science Information Network (CIESIN), Columbia University (Center for International Earth Science Information Network 2005).
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Here we review the current state of understanding of drought in the region, considering regional climate in section 2, historical drought episodes in section 3, drought dynamics in section 4, predictability in section 5, and trends and projections in section 6. The review concludes with a summary, including critical research questions for future studies of drought in the region.
2. Regional climate
To provide context for the consideration of drought dynamics, an overview of the regional climate is given. The spatial distribution and seasonality of precipitation, vegetation, and hydrology are reviewed, as well as the primary atmospheric circulation and precipitation mechanisms that constitute the processes through which drought can be expressed.
a. Distribution and seasonality of precipitation, hydrology, and vegetation
The annual mean precipitation is shown in Fig. 4a. The largest amounts occur in five main areas: the coastal eastern Mediterranean Sea, the western slopes of the Zagros Mountains in Iran and Iraq, the south coast of the Caspian Sea, the slopes of the mountain complex on the western edge of the Tibetan Plateau (the Hindu Kush, the Pamir, and the Tian Shan ranges), and the southern tip of the Arabian Peninsula. There is also significant precipitation in the Fertile Crescent, which arcs from the coastal Mediterranean Sea across southeastern Turkey, northern Syria and Iraq, and into northeastern Iran along the western Zagros (Fig. 2). The spatial distribution of precipitation is closely linked to the very significant topography of the region (Fig. 4b), especially the western slopes of the mountains, as much of the regional precipitation occurs during the cold season as a result of orographic capture from eastward-moving storm systems (Martyn 1992) guided by the upper-tropospheric westerlies (Krishnamurti 1961; Schiemann et al. 2009). The distribution of precipitation is a strong control on regional vegetation (Fig. 4c), which has implications for drought feedbacks, as discussed in section 4e. The region also includes areas of little to no precipitation, including several major deserts (nomenclature varies): the Negev in southern Israel; the Syrian Desert, inland from the coastal Mediterranean Sea; the Arabian Desert, over much of the Arabian Peninsula; the Iranian salt deserts in interior and eastern Iran; the Thar Desert in northwestern India and parts of eastern Pakistan; the Registan Desert centered on southwestern Afghanistan; and the Kara-Kum, centered in Turkmenistan.
(a) Annual mean precipitation (cm), (b) elevation (km), (c) growing season vegetation, and (d) coefficient of variation for annual precipitation (ratio of the standard deviation to the mean). The contour intervals for precipitation and elevation are 10 cm and 0.5 km, respectively. The growing season vegetation is estimated by April–August normalized difference vegetation index (NDVI). Both NDVI and the coefficient of variation are unitless ratios, with contour intervals of 0.1 in both cases. The NDVI is shaded from light green (least vegetative vigor) to red (most vigor) to allow visual discrimination of the large range in vegetative vigor over the region. The GPCC version 6 dataset (Schneider et al. 2014; Rudolf et al. 2003) is used for precipitation, with the averaged calculated for the 1951–2010 period. The Global Inventory Modeling and Mapping Studies (GIMMS) NDVIg dataset (Pinzon et al. 2005; Tucker et al. 2005), produced by the Global Land Cover Facility (GLCF) at the University of Maryland, College Park, is used for NDVI for the1981–2006 period. The ETOPO5 dataset (NOAA 1988) is used for topography.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
(a) Annual mean precipitation (cm), (b) elevation (km), (c) growing season vegetation, and (d) coefficient of variation for annual precipitation (ratio of the standard deviation to the mean). The contour intervals for precipitation and elevation are 10 cm and 0.5 km, respectively. The growing season vegetation is estimated by April–August normalized difference vegetation index (NDVI). Both NDVI and the coefficient of variation are unitless ratios, with contour intervals of 0.1 in both cases. The NDVI is shaded from light green (least vegetative vigor) to red (most vigor) to allow visual discrimination of the large range in vegetative vigor over the region. The GPCC version 6 dataset (Schneider et al. 2014; Rudolf et al. 2003) is used for precipitation, with the averaged calculated for the 1951–2010 period. The Global Inventory Modeling and Mapping Studies (GIMMS) NDVIg dataset (Pinzon et al. 2005; Tucker et al. 2005), produced by the Global Land Cover Facility (GLCF) at the University of Maryland, College Park, is used for NDVI for the1981–2006 period. The ETOPO5 dataset (NOAA 1988) is used for topography.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
(a) Annual mean precipitation (cm), (b) elevation (km), (c) growing season vegetation, and (d) coefficient of variation for annual precipitation (ratio of the standard deviation to the mean). The contour intervals for precipitation and elevation are 10 cm and 0.5 km, respectively. The growing season vegetation is estimated by April–August normalized difference vegetation index (NDVI). Both NDVI and the coefficient of variation are unitless ratios, with contour intervals of 0.1 in both cases. The NDVI is shaded from light green (least vegetative vigor) to red (most vigor) to allow visual discrimination of the large range in vegetative vigor over the region. The GPCC version 6 dataset (Schneider et al. 2014; Rudolf et al. 2003) is used for precipitation, with the averaged calculated for the 1951–2010 period. The Global Inventory Modeling and Mapping Studies (GIMMS) NDVIg dataset (Pinzon et al. 2005; Tucker et al. 2005), produced by the Global Land Cover Facility (GLCF) at the University of Maryland, College Park, is used for NDVI for the1981–2006 period. The ETOPO5 dataset (NOAA 1988) is used for topography.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Throughout the Middle East and southwest Asia, interannual variability in rainfall is high. The ratio of the standard deviation to the mean (the coefficient of variation) for annual precipitation is shown in Fig. 4d, with most of the region having values larger than 20% (and often much larger). Most of the highest values are in regions with little to no precipitation or at the margins of those regions (cf. Fig. 4a). As an example of variability, the annual precipitation for Jerusalem is shown in Fig. 5 for 1950–2012; the standard deviation is 31% of the mean, and the annual values span more than a fivefold range in that period, from a minimum of 21 cm to a maximum of 113 cm.
Annual precipitation (cm) in Jerusalem for 1950/51–2012/13. The precipitation data were obtained from the Israel Meteorological Service.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Annual precipitation (cm) in Jerusalem for 1950/51–2012/13. The precipitation data were obtained from the Israel Meteorological Service.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Annual precipitation (cm) in Jerusalem for 1950/51–2012/13. The precipitation data were obtained from the Israel Meteorological Service.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Analysis of precipitation is limited by sparse observations and the presence of steep precipitation gradients associated with the extreme terrain. Figure 6 shows the amount of monthly station observations that underlie the gridded data in the GPCC dataset for the 1951–2010 period, as used in Fig. 4 and Table 1. In addition to the poor spatial coverage in many parts of the domain, there is considerable temporal variability in coverage as well (Hoell et al. 2015). A comprehensive analysis of available precipitation data for the entire region has not yet been done, but the northern part of the domain was considered in Schiemann et al. (2008), Kuwait was considered in Marcella and Eltahir (2008), Saudi Arabia in Almazroui (2011), and Iran in Katiraie-Boroujerdy et al. (2016) for a range of different datasets and sources. In general, data agreement among observed gridded datasets is good enough for qualitative identification of wet and dry years but has a large range of uncertainty for quantitative analysis; model products and satellite estimates can provide value, especially in real-time monitoring, but have even more uncertainty. The high values of precipitation along the southern shore of the Caspian Sea are poorly represented in some observed gridded data as well as in model and satellite estimates. Despite the uncertainties, the large-scale variability appears to be well captured by the available data in the high mountains, where the precipitation is mainly from synoptic-scale systems that are well observed upstream and with precipitation strongly constrained by the orography. That is, while the spatial details of the pattern may not be well resolved, the basin-averaged amounts appear to be adequately represented (Schär et al. 2004; Barlow and Tippett 2008).
Station data availability underlying the GPCC version 6 dataset, as in Fig. 4, for the 1951–2010 period, on the 0.5°-resolution grid. The availability is given in terms of the percent of months in the 60-yr period that have at least one station report per grid box for each month. That is, a value of 100 for an individual grid box indicates that for each month in the period, there was at least one station report within the area of the grid box, whereas a value of 0 indicates that all the monthly values in the period for that grid box were based on interpolation.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Station data availability underlying the GPCC version 6 dataset, as in Fig. 4, for the 1951–2010 period, on the 0.5°-resolution grid. The availability is given in terms of the percent of months in the 60-yr period that have at least one station report per grid box for each month. That is, a value of 100 for an individual grid box indicates that for each month in the period, there was at least one station report within the area of the grid box, whereas a value of 0 indicates that all the monthly values in the period for that grid box were based on interpolation.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Station data availability underlying the GPCC version 6 dataset, as in Fig. 4, for the 1951–2010 period, on the 0.5°-resolution grid. The availability is given in terms of the percent of months in the 60-yr period that have at least one station report per grid box for each month. That is, a value of 100 for an individual grid box indicates that for each month in the period, there was at least one station report within the area of the grid box, whereas a value of 0 indicates that all the monthly values in the period for that grid box were based on interpolation.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Cross correlation of country-averaged annual precipitation for 1951–2010. Correlations ≥0.5 shown in bold and results are rounded to one decimal for clarity. Precipitation is from the GPCC dataset, as in Fig. 4.
The seasonal cycle of precipitation is shown in Fig. 7, in terms of both seasonal amount (left panels) and percentage of annual mean (right panels; brown shading indicates seasonal values less than one-quarter of the annual mean, and green shading indicates values more than one-quarter of the annual mean). There are three main precipitation mechanisms in the region: cold season synoptic precipitation, which is important for much of the region; summer monsoon precipitation, which is important for Pakistan and southeastern Afghanistan; and warm season African monsoon–ITCZ precipitation extending into the southern tip of the Arabian Peninsula from Africa.
Seasonal cycle of precipitation, in terms of (left) total values and (right) percent of annual mean. The total values are shown in intervals of 10 cm, with an additional first interval of 2 cm. The percent of annual mean is shown in intervals of 5% and brown shading for values <25% (the point at which all seasons would be contributing equally) and in intervals of 10% and green shading for values >25%. Calculations are based on CMAP (Xie and Arkin 1996, 1997) for the 1979–2010 period. The CMAP dataset is used here to allow inclusion of ocean precipitation.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Seasonal cycle of precipitation, in terms of (left) total values and (right) percent of annual mean. The total values are shown in intervals of 10 cm, with an additional first interval of 2 cm. The percent of annual mean is shown in intervals of 5% and brown shading for values <25% (the point at which all seasons would be contributing equally) and in intervals of 10% and green shading for values >25%. Calculations are based on CMAP (Xie and Arkin 1996, 1997) for the 1979–2010 period. The CMAP dataset is used here to allow inclusion of ocean precipitation.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Seasonal cycle of precipitation, in terms of (left) total values and (right) percent of annual mean. The total values are shown in intervals of 10 cm, with an additional first interval of 2 cm. The percent of annual mean is shown in intervals of 5% and brown shading for values <25% (the point at which all seasons would be contributing equally) and in intervals of 10% and green shading for values >25%. Calculations are based on CMAP (Xie and Arkin 1996, 1997) for the 1979–2010 period. The CMAP dataset is used here to allow inclusion of ocean precipitation.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Terrestrial hydrological dynamics in the region are extremely diverse and cannot be addressed comprehensively in this review. In general, the prevailing hydrology follows precipitation gradients, with major rivers originating in relatively moist highland zones and often flowing into semiarid or arid regions. In the high mountains of the region, precipitation occurs primarily during the cold season, as snow. As there is little precipitation during the warm season, the spring melt is a primary driver of peak river flows (Schär et al. 2004; Barlow and Tippett 2008). The headwaters of the Tigris and Euphrates Rivers are in the Anatolia Plateau of Turkey and the Zagros Plateau of Iran. These mountain ranges have been found to play a key role in the production of precipitation in the Fertile Crescent region (Evans et al. 2004; Alijani 2008) and, by extension, water resources in the region. Note that the headwaters are partially out of the domain considered here so that a full consideration of hydrological drought for the Fertile Crescent requires consideration of the remote precipitation in Turkey. The waters of the Tigris and Euphrates support extensive irrigation developments in all riparian states and feed the marshlands of southern Iraq, both of which have been found to affect the local climate through their influence on evaporation (Evans and Zaitchik 2008; Marcella and Eltahir 2012b). There has not been much study of soil moisture in the region, but winter and spring precipitation and temperature have a strong influence on soil moisture in the Euphrates plain (Zaitchik et al. 2007) and likely many other parts of the region as well.
Notably, the region includes several major transboundary rivers, including the Indus, the Jordan, the Tigris and Euphrates, and the Amu Dar’ya and Syr Dar’ya, all of which are subject to intense water demands, resulting in considerable political tension. On the northern border of the region, the Aral Sea, fed by the Amu Dar’ya and Syr Dar’ya, was originally the fourth largest lake in the world but was reduced to 10% of its original size by 2007 as a result of the introduction of intensive agricultural practices (Micklin 1988; Micklin and Aladin 2008).
The geographic distribution, seasonality, and interannual variability of precipitation have a clear signal in vegetation cover and variability across the region. The growing season vegetation, as broadly represented by the April–August normalized difference vegetation index (NDVI), is shown in Fig. 4c. The largest values of April–August NDVI are in a narrow band along the southern shore of the Caspian Sea in association with the largest values of precipitation in that region, as well as in the valleys of the Pamir and Tian Shan Mountains. In addition to the local correspondence with precipitation, there are notable anthropogenic fingerprints in the distribution of vegetation. This is true in large rivers with significant vegetation that has an irrigated or nonlocal relationship with precipitation, including the valleys of the Indus, Amu Dar’ya, Syr Dar’ya, Tigris, and Euphrates. Note that, although NDVI is an estimate of vegetative vigor, it does not necessarily reflect the most societally important areas of vegetation, given the widespread subsistence agriculture and livestock grazing through the region that occur in only moderately vegetated areas. Agriculture is sufficiently intensive in the former Soviet Union countries that changes in vegetation associated with the dissolution of the Soviet Union are evident (Kariyeva and van Leeuwen 2012).
b. Circulation and precipitation mechanisms
The upper-level winds, sea level pressure (SLP), and synoptic variability are shown in Fig. 8 for average January conditions (Figs. 8a,c,e) and July conditions (Figs. 8b,d,f). The synoptic variability is shown in terms of 2–8-day-filtered upper-level meridional wind variance, which is a good indicator of upper-level synoptic transient activity in the region, as discussed further below.
(left) January and (right) July circulation fields: (a),(b) 300-hPa wind (vectors and speed colored shading); (c),(d) SLP; and (e),(f) 200-hPa synoptic transients. A 2–8-day-filtered 200-hPa meridional wind variance is used as the estimate of upper-level synoptic transients. The contour interval for wind speed is 10 m s−1, SLP is 4 hPa, and synoptic transients is 10 m2 s−2. The calculations are based on the NCEP–NCAR reanalysis (Kalnay et al. 1996) for the 1951–2010 period.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
(left) January and (right) July circulation fields: (a),(b) 300-hPa wind (vectors and speed colored shading); (c),(d) SLP; and (e),(f) 200-hPa synoptic transients. A 2–8-day-filtered 200-hPa meridional wind variance is used as the estimate of upper-level synoptic transients. The contour interval for wind speed is 10 m s−1, SLP is 4 hPa, and synoptic transients is 10 m2 s−2. The calculations are based on the NCEP–NCAR reanalysis (Kalnay et al. 1996) for the 1951–2010 period.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
(left) January and (right) July circulation fields: (a),(b) 300-hPa wind (vectors and speed colored shading); (c),(d) SLP; and (e),(f) 200-hPa synoptic transients. A 2–8-day-filtered 200-hPa meridional wind variance is used as the estimate of upper-level synoptic transients. The contour interval for wind speed is 10 m s−1, SLP is 4 hPa, and synoptic transients is 10 m2 s−2. The calculations are based on the NCEP–NCAR reanalysis (Kalnay et al. 1996) for the 1951–2010 period.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
At the largest scales, the circulation changes from a cold season regime under the influence of the westerlies and associated synoptic activity to a warm season regime under the influence of the Indian monsoon. The strong westerly jet and associated synoptic activity during the cold season are seen in Figs. 8a and 8e, respectively, and the extension of the northwestern-most branch of the monsoon into northern Pakistan and southeastern Afghanistan is seen in the July–September panels of Fig. 7.
Cold season synoptic precipitation is important over much of the region and has been studied by several authors. Reconciling these previous synoptic analyses of extratropical cyclones and cyclone tracks into an overall picture is challenging, however, because of differences in spatial domain, methodology, and terminology. The transient analysis of Hoskins and Hodges (2002) provides a large-scale, general view. Over the region, they identified a prominent maximum of vertical velocity variance at midlevels and a local branch of transient activity in both meridional wind and potential temperature of the 2-PVU (1 PVU = 10−6 m−2 s−1 K kg−1) surface at upper levels, similar to the branch of transient activity shown in Fig. 8e. Their dynamically oriented analysis found that the principal track at upper levels is from the Atlantic and the principal track at lower levels is from the Mediterranean Sea—these results provide a reasonable regional-scale perspective.
More local analyses have shown considerable detail in the structures of these general paths. The cyclone tracks affecting the Fertile Crescent can be grouped into two main routes: south of Turkey toward the Caspian Sea and near Jordan or Syria toward the Persian Gulf (Trigo et al. 1999). Systems are relatively frequent but precipitation occurs only with strong surface frontal activity associated with the strongest systems (Trigo et al. 1999). For the Middle East, Cyprus lows are more important for precipitation than are Black Sea or Atlas lows. Troughs traveling along the subtropical jet stream can also generate isolated thunderstorms without strong surface fronts via inducing onshore flows from the Persian Gulf (Trigo et al. 2010). A local maximum of cyclogenesis associated with the relatively warm surface of the Caspian Sea and with Caucasus lee waves modulates the pressure field over the southern part of western Asia (Lambert et al. 2002; Paz et al. 2003). Occasionally, Red Sea troughs propagate from the south, and when they coincide with extratropical cyclones entering from the west, there can be devastating floods. This happened, for example, in November 1994 (Krichak et al. 2000; Ziv et al. 2005a). At the margins of the rainy season, Red Sea troughs play a greater role, with most rainy events resulting from a northward extension of this low (Tsvieli and Zangvil 2005). Both the eastern Mediterranean upper trough and the Red Sea trough are important along the southeastern coast of the Arabian Peninsula (Charabi and Al-Hatrushi 2010). Tropical cyclones are also an infrequent source of precipitation in that region (Al-Rawas and Valeo 2009).
The synoptic tracks affecting the southwest Asia part of the domain have not been studied as much as the Middle East part of the domain. Barlow et al. (2006) showed, for heavy precipitation events in Afghanistan, a westerly track into southwest Asia consistent with the branch of transient activity shown in Fig. 8e, as did Cannon et al. (2016) for precipitation in the western Himalayas and Karakoram.
In the summer, the Indian monsoon, in addition to directly affecting northern and southeastern Pakistan, appears to also remotely influence precipitation over the northern part of the domain via an influence on temperature and stability (Schiemann et al. 2007). It has also been suggested that the Indian monsoon may remotely force subsidence over the region (Rodwell and Hoskins 1996, 2001), with orographic interaction playing an important role (Rodwell and Hoskins 2001; Tyrlis et al. 2013; Simpson et al. 2015), and observational data supports this relationship during the monsoon onset (Saini et al. 2011) and peak phase (Ziv et al. 2004). However, this influence occurs after most of the seasonal decline in precipitation has already occurred for much of the region. The degree to which the end of the wet season is due to an increase in inhibiting factors, such as the remote forcing of the monsoon or its tropical precursors, versus a decrease in precipitation-generating factors, such as the northward retreat and weakening of the storm track, is not yet clear.
While orography plays a primary role in generating precipitation via upslope lifting, it can play that role only if sufficient water vapor is transported into the area. The water vapor transport pathways were investigated explicitly in Evans and Smith (2006) for the Fertile Crescent. They showed that most precipitation events in that area are dominated by water vapor coming from the west over the Mediterranean Sea. However, the largest precipitation events were dominated by water vapor coming from the south where the Red Sea and Persian Gulf play a role as source regions. These southerly dominated events are able to transport large amounts of water vapor into the Fertile Crescent through the formation of a mountain barrier jet west of the Zagros Mountains (Evans and Alsamawi 2011). This mountain barrier jet occurs over a relatively small spatial scale such that current global climate models (GCMs) are unlikely to represent the phenomena and hence are probably unable to produce an important class of precipitation event for the Fertile Crescent. Given the extreme terrain throughout most of the region and the scarcity of observations, it is not clear that current gridded global analyses adequately represent regional moisture fluxes.
The seasons in the Persian Gulf area of the region are sometimes divided into the northeast monsoon (December–March), the spring transition (April–May), the southwest monsoon (June–September), and the fall transition (October–November; Walters and Sjoberg 1988), as this area transitions from low-level northeasterly flow in the cold season to low-level southwesterly flow in the warm season. For the full study region, at low levels the northeastern extent is at the margin of the Siberian high during winter and the southern extent is under the influence of a shallow low pressure area over Saudi Arabia and southern Pakistan, typically identified as a thermal low. Over Saudi Arabia, field work shows that within the low pressure area there is subsidence to within 1 km of the surface and then ascent below that (Blake et al. 1983). However, the interaction of orography with remote dynamic forcing may contribute more to the center of the low pressure system over southern Pakistan and northwestern India than does direct sensible heating from the surface (Bollasina and Nigam 2011).
During summer, a low-level jet forms over the Persian Gulf, with orography, mountain slope, and land–sea breeze playing a role in the formation and strength of the jet (Giannakopoulou and Toumi 2012).
3. Historical droughts
This section reviews historical droughts as identified in the previous literature and in a drought disaster database.
In the 1940s–present period, the two most severe droughts for the region as a whole appear to be the 1999–2001 (Agrawala et al. 2001; Barlow et al. 2002; Lotsch et al. 2005; Trigo et al. 2010; Kaniewksi et al. 2012; Hoell et al. 2012) and 2007/08 droughts (Trigo et al. 2010; Kaniewski et al. 2012; Hoell et al. 2012). (The 1999–2001 period given for the first drought indicates the period of broadest impact, but the drought extended from 1998 to 2002 for some areas of the region.) Information about more local droughts as well as droughts prior to the 1940s is sparse and difficult to compare based on different definitions and data periods. As summarized in Bruins (1999), the largest precipitation deficits in Jerusalem for the 1846–1993 period are 1932/33, 1950/51, and 1959/60, all of which were less than 50% of average. For Israel as a whole, based on a hydrologic measure for the 1937–84 period, 1950/51, 1958/59, 1972/73, and 1978/79 were strong drought years. More recently, the period 2004/05–2008/09 was also very dry. The driest years for the Jordan valley identified in Black (2012) for the 1950–99 period are 1960, 1963, 1966, 1970, 1973, 1979, 1986, 1987, and 1999. Section 4d provides some new analysis of regional droughts for the 1951–2010 period in the context of regional coherence.
For an impacts-oriented perspective on historical droughts in the region, drought disasters from the Emergency Events Database (EM-DAT), the Office of U.S. Foreign Disaster Assistance/Centre for Research on the Epidemiology of Disasters (OFDA/CRED) international disaster database (Centre for Research on the Epidemiology of Disasters 2014), are shown in Table 2. A disaster is defined in this database as meeting one of the following criteria: 10 or more people reported killed, 100 or more people reported affected, declaration of a state of emergency, or a call for international assistance. While there may be some reporting issues with this database, the occurrence of a drought disaster generally tracks precipitation anomalies for Asia (Barlow et al. 2006), so it appears to be reasonable. Based on the drought disaster data, the 1999–2001 and 2007/08 droughts are the most represented, sometimes extended to 1999–2002 and 2007–10, along with more local droughts: 1969, 1971–73, and 2006 for Afghanistan; 1969–71 for Iraq; 1964 for Iran; and 1969–71, 1975, and 1977 for Yemen. Given that Yemen has a different primary precipitation mechanism (African monsoon/summer ITCZ) from the rest of the region, it may be expected to have different drought events. Because of the geography and precipitation mechanism, it seems likely that these droughts may be synchronized with drought events in the greater Horn of Africa, but this has not been studied explicitly.
Drought disasters as listed in the EM-DAT disaster database (Centre for Research on the Epidemiology of Disasters 2014).
The historical droughts listed in this section were identified based on different metrics, periods, and areas within the region. A comprehensive ranking of historical drought episodes in the region, accounting for data issues, and especially with respect to the pre-1969 record, is not possible based on the previous literature but would be very useful for assessing the dynamical influences on the region, within-region relationships, and associated predictability.
4. Drought dynamics
This section reviews results related to the dynamics of drought in the region, in terms of large-scale teleconnection patterns, synoptic regimes and climatological features, mechanisms of tropical forcing, regional coherence, vegetation, and dust.
a. Large-scale teleconnection patterns
Several large-scale teleconnection patterns have been shown to have an influence on precipitation in different areas of the region:
North Atlantic Oscillation (NAO) (Cullen and deMenocal 2000; Aizen et al. 2001; Cullen et al. 2002; Mann 2002; Krichak et al. 2002; Syed et al. 2006; Charabi and Al-Hatrushi 2010; Black 2012; Donat et al. 2014; Al Senafi and Anis 2015; Athar 2015),
El Niño–Southern Oscillation (ENSO) (Price et al. 1998; Barlow et al. 2002; Nazemosadat and Ghasemi 2004; Mariotti et al. 2005; Syed et al. 2006; Mariotti 2007; Hoell et al. 2014a; Hoell et al. 2014b; Niranjan and Ouarda 2014; Donat et al. 2014; Yin et al. 2014; Krichak et al. 2014; Al Senafi and Anis 2015; Athar 2015),
Madden–Julian oscillation (MJO) (Barlow et al. 2005; Nazemosadat and Ghaedamini 2010; Barlow 2011; Hoell et al. 2012; Tippett et al. 2015; Pourasghar et al. 2015),
east Atlantic–western Russia (EA/WR) pattern (Krichak et al. 2002; Ziv et al. 2006; Yosef et al. 2009; Black 2012; Yin et al. 2014; Krichak et al. 2014)
Indian Ocean dipole (IOD) (Chakraborty et al. 2006; Pourasghar et al. 2012; Al Senafi and Anis 2015; Athar 2015),
North Sea–Caspian pattern (NCP) (Kutiel and Benaroch 2002; Kutiel et al. 2002; Yosef et al. 2009)
North Africa/western Asia (NAWA) index (Paz et al. 2003; Tourre and Paz 2004)
Mediterranean oscillation index (Yosef et al. 2009; Ziv et al. 2014),
Pacific warm pool index (Lotsch et al. 2005; Ziv et al. 2006),
Atlantic multidecadal oscillation (AMO) (Sheffield and Wood 2008),
Atlantic tripole index (Lotsch et al. 2005),
Circumglobal teleconnection (CGT) (Feldstein and Dayan 2008),
Polar–Eurasia pattern (Yin et al. 2014), and
west Pacific oscillation (WPO) (Aizen et al. 2001).
The strengths of these teleconnections vary considerably within the region. There are also indications of variations in the temporal stability of these relationships (Ropelewski and Halpert 1987; Price et al. 1998; Marcella and Eltahir 2008; Krichak et al. 2014)—although there are perhaps some indications that using a definition based on sea surface temperatures (SSTs) of NAO-like variability gives a stronger, more stable result than a more traditional atmospheric-based definition. ENSO has also been linked to variations in river flow (Barlow and Tippett 2008), soil moisture (Sheffield and Wood 2008), and NDVI (Kariyeva and van Leeuwen 2012) in the region.
For ENSO, it appears that the state of the western Pacific may be an important additional factor (Barlow et al. 2002; Hoerling and Kumar 2003; Hoell and Funk 2013). As noted previously, two of the most severe widespread droughts for the region were during 1999–2001 and 2007/08, both of which were associated with La Niña conditions, a warm western Pacific, and similar hemispheric wind patterns (Agrawala et al. 2001; Barlow et al. 2002; Trigo et al. 2010; Hoell et al. 2012; Hoell et al. 2014a).
For many of these teleconnection studies, the influence on precipitation has been evaluated only for subareas of the full region considered in this review. Moreover, aside from the influence of the tropical Pacific in the 1999–2001 and 2007/08 drought events, the role of these teleconnections in individual drought episodes has not been closely analyzed.
b. Synoptic regimes and climatological features
In an analysis of wet and dry years for the Jordan valley, Black (2012) found that the synoptic regimes noted previously in section 2 were also important in interannual variability and were modulated by the NAO. A wet-minus-dry composite showed that precipitation departures over the Jordan valley were out of phase with precipitation departures over southern Europe and associated with changes in storm-track structure over the Mediterranean Sea.
A major part of eastern Mediterranean precipitation is controlled by a large-scale process associated with two main anticyclonic centers: the Azorian and Siberian highs. During periods with intensive anticyclones over central Europe, the area gets drier (Kutiel and Paz 1998; Krichak et al. 2000). During these dry spells, positive SLP and midlevel height (H-500) anomaly patterns prevail over eastern Europe while negative SLP and H-500 anomalies are found over southwestern and western Europe. A more intensive anticyclonic system over central Europe during dry eastern Mediterranean seasons suggests an intensification of the westward advection of dry Asian air masses into the area (Krichak et al. 2000).
c. Mechanisms of tropical forcing
While there is not yet a complete understanding of the dynamical pathways by which tropical variability can influence the region, there have been several studies of the influence of tropical convection occurring over a region extending from the eastern Indian Ocean to the western Pacific Ocean. Enhanced tropical Indo–west Pacific Ocean convection results in diabatic heating increases, which excites baroclinic (Barlow et al. 2002, 2005, 2007; Barlow 2011; Hoell et al. 2012, 2013) and barotropic (Hoell et al. 2013) stationary Rossby waves over central-southwest Asia and the Middle East. The mean wind appears to be important in increasing the northward extent of the classic Gill–Matsuno response to tropical forcing, which assumes a resting basic state (Barlow 2011; Adames and Wallace 2014). The stationary Rossby waves thermodynamically interact with the mean climate, resulting in modifications to the mid- and upper-tropospheric temperature advection, which is balanced by precipitation-suppressing subsidence (Barlow et al. 2005; Hoell et al. 2012; Hoell et al. 2014b). Anticyclonic circulation associated with Rossby waves also reduces the flux of moisture into the region from tropical Africa and the Arabian Peninsula (Mariotti et al. 2002, 2005; Mariotti 2007; Barlow and Tippett 2008; Hoell et al. 2014b). The Indo–west Pacific convection also appears to excite eastward-traveling barotropic Rossby waves that can travel across the Northern Hemisphere and influence the Middle East and central-southwest Asia from the west (Hoell et al. 2013). Additionally, a hemispherically symmetric influence from ENSO has been proposed in Seager et al. (2003, 2005), resulting from interaction between the tropically forced subtropical jet anomalies and transient activity.
Thus, there are indications that tropical Indo-Pacific Ocean anomalies can influence the region with a modified Gill–Matsuno-like response to the west, a response to the east that propagates around the hemisphere, and a hemispherically symmetric response. Over the region, these wind circulations modify midlevel vertical velocity, moisture flux, and storm tracks, as shown schematically in Fig. 9—these mechanisms appear to be operating in both major regional droughts of the last 50 years, although the relative importance of the three different mechanisms has not yet been evaluated.
Schematic of drought mechanisms for the 1999–2001 and 2007/08 severe regional droughts.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Schematic of drought mechanisms for the 1999–2001 and 2007/08 severe regional droughts.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
Schematic of drought mechanisms for the 1999–2001 and 2007/08 severe regional droughts.
Citation: Journal of Climate 29, 23; 10.1175/JCLI-D-13-00692.1
d. Regional coherence
While there are at least two major factors that link the region as a whole—the widespread nature of the most severe drought episodes and the importance of cold season synoptic precipitation across most of the region—the amount of regional coherence in year-to-year variability is less clear. To provide a very simple first view of this, annual precipitation is averaged over each country and then correlated with every other country for the 1951–2010 period based on the GPCC data; these cross correlations are shown in Table 1. For clarity, the correlations are shown to only one significant digit, and values equal to or greater than 0.5 are shown in bold. The amount and spatial distribution of station data underlying this analysis change considerably over the period (Hoell et al. 2015), and so, while the results are consistent with previous work, they should be interpreted with some caution. In this calculation with annual data, the northern tier of countries (Uzbekistan, Kyrgyzstan, Turkmenistan, and Tajikistan) is distinct from the others. The countries of the Fertile Crescent are generally closely related but do not clearly form a distinct group; instead, they form a chain of relationships, gradually changing from east to west. Afghanistan is correlated at 0.5 or higher only for Iran, Pakistan, and the United Arab Emirates. Pakistan is correlated at 0.5 or higher only for Afghanistan, which is expected, given the large contribution of the summer monsoon to Pakistan’s precipitation (the mutual correlation is 0.79 considering only the November–April months). Yemen is also correlated at 0.5 or higher only for a single other country in the domain, Oman, which is also expected, given the different precipitation mechanism for the southern tip of the Arabian Peninsula (the African monsoon/ITCZ). The countries of the Arabian Peninsula, in general, are not well correlated to the rest of the region, except for Iran.
From this simple analysis, it appears that none of the domains usually discussed in the literature is a close match to the spatial relationships of precipitation variability, at least in the annual mean, although the entire domain can be coherent in severe episodes. This complexity is consistent with the wide range of regional and large-scale influences on the region, as discussed in the previous three subsections.
For the same period of 1951–2010, the yearly variation in the number of countries with precipitation deficits is shown in Fig. 10 for three deficit thresholds: less than average (light brown), less than 90% of average (medium brown), and less than 75% of