The Active Red Sea Trough (ARST) is an infrequent weather phenomenon that is associated with extreme precipitation, flash floods, and severe societal impacts in the Middle East (ME). Using reanalysis (ERA-Interim) and observational precipitation (Aphrodite and stations) data, we investigate its underlying dynamics, geographical extent, and seasonality. Twelve ARST events affecting the Levant have the same dynamical characteristics as those associated with a major flood in Jeddah (Saudi Arabia) on 25 November 2009. Hence, the Jeddah flooding was caused by an ARST, which implies that ARSTs can affect a much larger part of the ME than previously assumed. We present an ARST concept involving six dynamical factors: (1) a low-level trough; the Red Sea Trough (RST), (2) an anticyclone over the Arabian Peninsula; the Arabian Anticyclone (AA), (3) a transient midlatitude upper trough, (4) an intensified subtropical jet stream, (5) moisture transport pathways, and (6) strong ascent resulting from tropospheric instability and the synoptic-scale dynamical forcing. We explain the ARST as the interaction of a persistent stationary wave in the tropical easterlies (i.e., the RST) with a superimposed amplifying Rossby wave, resulting in northward propagating moist air masses over the Red Sea. Our findings emphasize the relevance of the AA, causing moisture transport from the Arabian and Red Seas. The particular topography in the Red Sea region and associated low-level circulation makes the ARST unique among tropical-extratropical interactions. The ARST seasonality is explained by the large-scale circulation and in particular the seasonal cycle of the semipermanent quasi-stationary RST and AA.
 Subtropical regions are usually characterized by a semiarid to arid climate with very limited annual precipitation. However, occasionally, extreme precipitation events affect these dry regions, causing flash floods that can have dramatic societal impacts, including major economic damage and loss of lives [e.g., Llasat et al., 2010]. Extreme precipitation events in the subtropics are often caused by tropical-extratropical interactions. They affect the southwestern part of North America [e.g., Knippertz and Martin, 2006; Favors and Abatzoglou, 2013], northwest Africa [Fink and Knippertz, 2003; Knippertz et al., 2003; Knippertz, 2003; Knippertz and Martin, 2005], Australia [e.g., Wright, 1997], South Africa [e.g., Hart et al., 2010], and the Middle East (ME). These tropical-extratropical interactions are associated with phenomena such as tropical plumes [McGuirk et al., 1988; Knippertz, 2007] and atmospheric rivers [e.g., Neiman et al., 2008].
 In the ME (Figure 1), the Levant (eastern Egypt, Sinai Peninsula, Israel, Jordan, Lebanon, and Syria) is frequently under the influence of a synoptic system known as the Red Sea Trough (RST). The RST refers to a low-level pressure trough that extends from the African Monsoon over equatorial Africa, northward over the Red Sea region toward the eastern Mediterranean (EM) [e.g., Alpert et al., 2004a; Tsvieli and Zangvil, 2005]. The RST is attributed to the local topography and thermal forcing factors [Krichak et al., 1997a, 1997b], and is most frequent in autumn while being less prominent in winter and spring [Alpert et al., 2004b; Tsvieli and Zangvil, 2005; Krichak et al., 2012]. The associated weather conditions in the Levant are usually dry and hot. However, when accompanied by a midlatitude upper trough, unstable conditions can develop that trigger convective storms, referred to as the Active Red Sea Trough (ARST) [Kahana et al., 2002; Ziv et al., 2005; Krichak et al., 2012]. Accordingly, the ME can be affected by extreme precipitation resulting from tropical-extratropical interactions, including tropical plumes [Ziv, 2001; Rubin et al., 2007] and ARSTs.
 Climatological studies show that the ARST is a rare weather phenomenon that can occur in late autumn and to a lesser incidence in early winter and spring [Kahana et al., 2002; Krichak et al., 2012; Shentsis et al., 2012]. Dayan et al.  explain the ARST preference for autumn by coinciding favorable latitudinal positions of the African Monsoon and the subtropical jet (STJ) stream. Previous ARST case studies relate several dynamical aspects to the phenomenon [e.g., Krichak and Alpert, 1998; Dayan et al., 2001; Ziv et al., 2005], elaborated in section 2. However, an integral concept fully comprising the involved atmospheric dynamics has not yet been defined [Alpert et al., 2006, chapter 2], which highlights the need for a better and comprehensive understanding of the ARST dynamics at synoptic scales as well as a seasonal perspective.
 The ARST is known to affect Egypt, Israel, Jordan, Lebanon, and Syria [Dayan et al., 2001; Kahana et al., 2002]. For example, a significant part of the severest floods that occurred in the Negev desert (Israel) resulted from ARSTs [Kahana et al., 2002; Shentsis et al., 2012]. Also, the catastrophic flash floods in Egypt in November 1994, leaving 600 casualties and additionally affecting 110,660 people (Emergency Events Database, unpublished data, 2012, available from Centre for Research on the Epidemiology of Disasters; http://www.emdat.be/database, hereafter EM-DAT), were caused by an ARST. Previous ARST studies, however, focused primarily on Israel. Therefore, it is yet unknown to what extent other countries in the ME region are also affected by the ARST.
 On 25 November 2009, Jeddah (Saudi Arabia) was dramatically hit by heavy precipitation and consequent flash floods (hereafter Jeddah 2009 event). The number of reported fatalities ranges from 122 [Haggag and El-Badry, 2013] to 161 (EM-DAT). The EM-DAT database lists the event as the natural disaster with the largest number of casualties in Saudi Arabia in the period 1900–2011 and furthermore mentions that 10,000 people were directly affected, while the estimated financial damage was about US$ 900 million. Haggag and El-Badry  investigated the Jeddah 2009 event based on model simulations and describe synoptic and mesoscale aspects that show substantial similarities with the ARST, which has motivated us to study the Jeddah flooding in this context.
 This study aims to improve the understanding of the ARST dynamics, its geographical extent, and seasonality. We review previous literature addressing ARST associated dynamics (section 2). Section 3 describes the atmospheric reanalysis and observational precipitation data that were used. We analyze the ARST associated precipitation (section 4), and the synoptic dynamics of 12 ARST events (section 5) and the Jeddah 2009 event (section 6). Section 7 addresses the ARST seasonality in view of the large-scale circulation. Section 8 presents a unified concept of ARST dynamics and places the phenomenon in the context of tropical-extratropical interactions. The paper ends with conclusions and suggestions for further research (section 9).
2 ARST Dynamics in Previous Studies
 Several dynamical aspects contribute to the formation of ARSTs. The flow near the surface at the eastern flank of the RST transports warm air masses over the Arabian Peninsula (AP) toward the Levant, while at upper levels, a midlatitude trough and associated cold air approaches the region [e.g., Krichak et al., 2000; Dayan et al., 2001; Kahana et al. 2002; Ziv et al., 2005; Dayan and Morin, 2006; Tsvieli and Zangvil, 2005, 2007; Krichak et al., 2012]. In combination with increased moisture, the troposphere can become highly unstable, triggering strong convection. Upper level positive vorticity advection in combination with low-level warm air advection implies quasi-geostrophic forced ascent [Tsvieli and Zangvil, 2005; Ziv et al., 2005]. In addition, the STJ intensifies over North Africa and the ME [Krichak et al., 1997a; Krichak and Alpert, 1998], associated with strong upper level divergence that favors midtropospheric ascent [Dayan et al., 2001; Ziv et al., 2005; Tsvieli and Zangvil, 2007].
 The origin of the moisture that is involved in ARSTs is debated as previous studies suggest various sources. Case studies indicate the Arabian Sea [Krichak and Alpert, 1998], west equatorial Africa [Dayan et al., 2001], and tropical Africa as the primary and the Red Sea as the secondary moisture source [Ziv et al., 2005]. Tsvieli and Zangvil [2005, 2007] suggest two acting moisture sources: equatorial Africa and/or the Indian Ocean through the southwesterly midtropospheric flow, and the EM by the west-northwesterly flow near the surface. Krichak et al.  indicate transport of large moisture quantities from tropical Africa in the form of an atmospheric river.
 Southerly winds in the midtroposphere over the Levant are considered to be a key feature of the ARST since they are associated with an upper trough that reaches sufficiently far southward to initiate moisture transport of tropical origin [Kahana et al., 2004]. In addition, Ziv et al.  found an intensified high at lower to middle levels over the AP for an ARST event in December 1993, causing persistent southerly winds and tropical moisture transport over the Red Sea. The present study will show that the role of the anticyclone over the AP, hereafter called the Arabian Anticyclone (AA) [after Raziei et al., 2012], is not restricted to the December 1993 event, but is very important for the ARST in general, as well as for the seasonal pace of the low-level circulation.
 In summary, the ARST has been associated with the following dynamical features: the RST, the AA, a transient midlatitude upper level trough, an intensified STJ, enhanced moisture transport, and upward motions resulting from tropospheric instability and the synoptic-scale dynamical forcing.
3 Data and Methods
3.1 Reanalysis and Precipitation Data
 The ARST dynamics are studied with the most recent European Centre for Medium-Range Weather Forecasts reanalysis data (ERA-Interim). The data set covers the period from 1979 and continues in near real time, and is produced with the Integrated Forecast System at a spectral resolution T255 and 60 model levels, reaching up to 0.1 hPa [Dee et al., 2011]. We utilize the monthly mean and 6 hourly analysis and forecasted variables at pressure and surface levels, as well as vertically integrated quantities derived from pressure levels.
 The Aphrodite precipitation data set (version V1101) and station observations are used to provide insight into the evolution and characteristics of the ARST associated rainfall. Aphrodite is a daily gridded data set based on rain gauge observations [Yatagai et al., 2012]. Version V1101 spans the period 1951 – 2007 at a resolution of 0.25° × 0.25° and covers the ME including the larger part of the AP. Station data over Israel are obtained from the Israel Meteorological Service archive.
3.2 Levant ARST Events
 Twelve ARST events (Table 1) that affected the Levant are analyzed, hereafter referred to as Levant ARST events. Their selection is motivated by one or more of the following reasons: (1) ARST events identified by the algorithm presented by Krichak et al. , (2) empirical analysis by D. Edry (personal communication, 2011), and (3) ARST events investigated by previous case studies. Additional ARSTs (May 1982, November 1996 and 2003) are identified via ERA-Interim precipitation, showing events with northward progressing rainfall over the Red Sea region toward the Levant. All Levant ARST events are evaluated on the presence of the RST and upper trough (Figure 2), as well as significant observed precipitation (supporting information Figure S1 and Table 2, see section 4).
12 casualties and 260 people affected in Egypt (flood, 13–18 Nov)b
15 casualties and US$ 40 M damage in Israel (flood from 17 to 19 October), four casualties, and US$ 1 M damage in Egypt (flood, 18–20 Oct) and two casualties and US$ 1 M damage in Jordan (flood, 18–20 Oct)b; at least six casualties in Egypt, nine in Israel, and two in Jordanc
 For the October 1988, November 1996 and 2003 events, negligible rainfall amounts are observed over Israel (Table 2); however, the Aphrodite (and ERA-Interim) data show precipitation over the surrounding region, illustrating that these events should be considered as ARSTs. Moreover, flash floods occurred in Egypt during the October 1988 [Cools et al., 2012] and November 1996 [Moawad, 2012] events, while an extraordinarily long rain spell (13–20 November 1996) affected the west coast of the AP with the most intense rainfall occurring from 16 to 18 November [Almazroui, 2012].
 The October 1979, 1988, and 1997 events deviate slightly from typical ARSTs. Whereas several ARST events are associated with weak cyclogenesis to the east of the Levant, these events show closed isobars at surface levels over the EM Sea (supporting information Figures S11d, S21d, S51d, and S51g), indicating cyclogenesis to the northwest of the Levant, and therefore suggest resemblance of another phenomenon, the Cyprus Low (a midlatitude cyclone over the EM Sea) [Alpert et al., 2004a]. Accordingly, Dayan et al.  describe the October 1997 event as the transformation from a convective storm into a midlatitude baroclinic system. Nevertheless, we include the three events to maintain consistency with previous literature, and because the three events exhibit all ARST dynamical characteristics and had major societal impacts (the October 1979 and 1997 events).
 The selection of Levant ARST events is assumed to be representative for identifying the key ARST dynamics. Note that we did not include all occurred ARSTs affecting the Levant during the period under consideration. Such an effort will be part of a quantitative climatological analysis in a subsequent study.
 Table 2 shows the observed precipitation at stations in Israel for all Levant ARST events. Rainfall affects the Levant for a time period from one to four consecutive days and maxima reach about 10 to 60 mm per day, with the exception of the October 1988, November 1996 and 2003 events (discussed earlier). ERA-Interim and Aphrodite data show that the precipitation during Levant ARST events evolves and progresses northward over the Red Sea, intensifies over the Levant, and eventually decays while moving in a northeasterly direction (Figure 3). In some cases, the precipitation intensifies at a later stage over the Taurus and Zagros mountains due to orographic effects. The intensity and spatial extent of the rainfall differs significantly between the events under consideration. Whereas several events (e.g., October 1979, 1987, 1997, November 1994, 1996, and December 1993) show widespread heavy rainfall over the Levant region for several days, other events (e.g., October 1994 and 2004) affect the region more locally and for shorter periods. However, this does not imply that local amounts are less intense. For example, Shentsis et al.  mentions for the October 2004 event a historical maximum in peak discharge at a station in the Negev desert with a 100 year return period.
 For the Jeddah 2009 event, extremely high precipitation quantities are reported. Whereas annual averages are roughly 50 mm per year, 140 and 180 mm (accumulated over 24 to 26 November) were observed at two stations in Jeddah, while mesoscale model simulations show accumulated maxima exceeding 400 mm for the most severely affected parts in Jeddah [Haggag and El-Badry, 2013]. Rainfall during the Jeddah 2009 event is only shown for the ERA-Interim data (Figure 4) as the Aphrodite data set spans until 2007. The daily precipitation exhibits a maximum over the Red Sea coast near Jeddah on 25 November 2009 (up to ~25 mm day−1) and, although strongly underestimating the amounts, clearly indicates the location and timing of heavy rainfall affecting Jeddah. Similar to the Levant ARST events, rainfall advances northeastward during the following days and decays afterward.
 The daily precipitation during the Levant ARST events shows a roughly similar temporal and spatial evolution in the Aphrodite and ERA-Interim data; see for example Figure 3. However, significant differences in spatial distribution and amounts are apparent between the data sets. For example, ERA-Interim shows for several events (e.g., October 1987, 1988, 1991, November 1994, 2003, and December 1993) more intense and widespread rainfall over the Levant region than Aphrodite. It is not clear if either ERA-Interim or Aphrodite is closer to reality. One should bear in mind that ERA-Interim precipitation is a forecasted quantity, limited by the model resolution and parameterizations, and therefore should be considered with caution. On the other hand, Aphrodite suffers from a relatively low station density over the region, in particular over Egypt, Jordan, and Syria (supporting information Figure S2). Moreover, extreme precipitation quantities are smoothed as a consequence of the interpolation from station to gridded data. A detailed discussion of the strengths and limitations of the precipitation data is beyond the scope of this study.
5 Dynamics of the Levant ARST Events
 The results described in section 5 are based on all 12 Levant ARST events. Minor differences in dynamics among the events are explicitly mentioned in the text. The ARST dynamics are illustrated by one typical event (October 1987) that has, despite its severe societal impacts (Table 1), not been the subject of previous work addressing the synoptic dynamics. Similar figures for all other Levant ARST events are included in the electronic supplement (supporting information Figures S10–S64).
5.1 Tropical and Subtropical Influences
 All Levant ARST events are associated with the presence of the RST at low altitudes and the AA at lower to middle tropospheric levels. The RST is positioned over northeast Africa and the Red Sea region, and is most pronounced at 1000 hPa (Figure 5a) and to a lesser degree at 850 hPa (Figure 5b), illustrating its shallow extent. During ARST occurrence the RST reaches over the Levant up to Turkey and retreats afterward (Figures 5d, 5e, 5g, and 5h). Some events are associated with closed isobars at lower levels to the east of the Levant during or after ARST occurrence, suggesting weak cyclogenesis (e.g., Figure 5g). The RST shows a strong diurnal variation, in agreement with Tsvieli and Zangvil . The 6 hourly data indicate lowest geopotential height values at 12 UTC, increasing at 18 and 00 UTC and reaching highest values at 6 UTC (not shown). The AA is positioned over the AP and demonstrates different structures with altitude. At low levels, a high pressure system stretches from the northeast over the AP (Figures 5b, 5e, and 5h). At higher levels, the AA has a more isolated and pronounced character with a closed circulation (Figures 5c, 5f, and 5i), and reaches up to midlevels as part of a ridge (Figures 6a, 6c, and 6e).
 Both the RST and AA have a strong semipermanent and quasi-stationary character. Hence, the question arises how the RST and AA behave during ARST events as compared to their mean state. Anomalies from the daily climatological mean (not shown) indicate that the RST extends northward prior to ARST occurrence [cf. Dayan et al., 2001; Kahana et al., 2002; Tsvieli and Zangvil, 2005; Ziv et al., 2005], and that the AA intensifies prior to ARST occurrence and weakens at a later stage during, or after the event. The northward extension of the RST and intensification of the AA result in enhanced pressure gradients, strong south-southeasterly winds, and moisture transport over the Red Sea region at lower levels.
5.2 Midlatitude Forcing
 Obviously, all Levant ARST events are accompanied by a midlatitude upper level trough penetrating into the subtropics (Figure 2). The upper level troughs are associated with cold air advection (not shown) and a strong cyclonic, southwesterly flow over the EM and Levant (Figure 6c). The upper troughs show various characteristics for the individual Levant ARSTs. Some events (October 1987, 1988, and 1994) demonstrate a pronounced trough over eastern Europe, extending southward over the EM and northeast Africa (Figure 6). Other events (May 1982, November 1994 and 1996) are characterized by a quasi-stationary cutoff over northeast Africa and the Levant, explaining for example the exceptionally long lasting period of precipitation affecting Jeddah during the November 1996 event [Almazroui, 2012]. Most events (October 1979, 1991, 1997, 2004, November 2003, and December 1993) go together with an eastward moving trough and/or cutoff over the EM and ME.
 The origin of the upper level troughs results for all ARST events from an amplified Rossby wave that imposes a trough over the EM and Levant with ridges at both flanks, positioned over northwest Africa and the AP, respectively (Figures 6a, 6c, and 6e). Most events are associated with an amplification of a stationary wave, while a few events (November 2003 and December 1993) are characterized by an eastward propagating wave, which in both scenarios goes along with wave breaking and cutoff formation. For at least a majority of events (May 1982, October 1979, 1988, 1997, November 1994, 1996, and December 1993) the backward-tilting trough (southwest-northeast orientation) and the anticyclonic shear (increasing wind speed with latitude) indicate anticyclonic wave breaking (supporting information Figures S12, S17, S22, S32, S42, S47, and S52), following the LC1 baroclinic life cycle [Thorncroft et al., 1993] which is characterized by a thinning trough that moves equatorward. Accordingly, wave amplification and breaking causes the intrusion of midlatitude upper level troughs into the subtropics that trigger ARSTs.
 For all Levant ARSTs, an intensified STJ is observed over northeast Africa and/or the AP. During several events (e.g., October 1987, 1997, November 1994, and December 1993), the STJ intensifies and approaches from the west over North Africa prior to ARST occurrence (Figure 6b), reaches over northeast Africa and/or the AP during ARST occurrence (Figure 6d), and ceases and/or moves eastward afterward (Figure 6f). For other events (e.g., May 1982, October 1988, 1991, 1994, and November 1996 and 2003), the STJ intensifies rather in situ over the region. During most events, the STJ streak develops to the southeast of the upper trough and follows its evolution in time and space over the region. Maxima in wind speed at 200 hPa range from 45 up to 65 m s−1. ARST events during late autumn/early winter show relatively higher wind speeds compared to events in autumn, which reflects the seasonality of the STJ that pronounces toward winter. The intensification of the STJ was previously suggested to result from enhanced convection over the tropics [Krichak and Alpert, 1998], indicating an intensified local Hadley Cell overturning. However, visual inspection of upper levels (200 hPa) reveals only for the October 1979, 1987 and November 1996 events substantial poleward outflow from the tropics. In fact, more events (October 1979, 1987, 1994, 1997, November 1994, and December 1993) show a branch of the polar jet, which curves around the western flank of the upper trough and confluences with the STJ (e.g., Figure 6d). Cold upper air advection from the midlatitudes enhances the meridional temperature gradient, implying a thermally driven intensification of the STJ [cf. Dayan et al., 2001].
5.3 Moisture Dynamics
 To investigate moisture dynamics, we analyze the vertically integrated moisture quantities, i.e., total column water (TCW), and vertically integrated water vapor fluxes (VIWVFs). During ARST events, large amounts of TCW (with maxima of 27 up to 39 kg m−2) advance northward over northeast Africa and the Red Sea region toward the Levant, perhaps suggesting central equatorial Africa as dominant moisture source (Figure 7a). However, the VIWVFs reveal pronounced moisture transport from the Arabian and Red Seas curving anticyclonically around the AP (Figure 7b). Figure 8 shows that this moisture transport pathway not only characterizes the October 1987 event, but in fact all Levant ARST events. In addition, several events also show weak westerly moisture transport over the EM (Figures 8a, 8c, 8d, 8f, 8j, and 8l).
 The moisture transport involves a complex picture due to its variability in the vertical direction and in time. Hence, we analyze in more detail the VIWVFs over three layers (Figures 7c–e) as well as its temporal evolution (Figures 7f–h). At low altitudes (~1000 to ~875 hPa), VIWVFs intensify over the Arabian and Red Seas prior to ARST occurrence (Figures 7c, 7f, and 7g), while for several ARSTs (October 1979, 1987, 1988, 1997, 2004 and December 1993) shortly prior to and during ARST occurrence, also weak west-northwesterly VIWVFs evolve over the Mediterranean Sea (Figures 7c and h). At higher levels (~850 to ~700 hPa), southerly VIWVFs over the Red Sea region are present (Figure 7d), while at middle to upper levels (~650 to ~300 hPa), during ARST occurrence, pronounced south-southwesterly VIWVFs are observed over the Levant (Figures 7e and 7h). In addition, several ARST events show minor moisture transport at low levels from central Africa (October 1991), and at middle to upper levels from West Africa (October 1997 and December 1993) and central Africa (October 1979, 1988, 1991 and November 1996).
 The evolution of VIWVFs during the Levant ARST events reveals that the moisture originates predominantly from the Arabian and Red Seas. The persistent anticyclonic flow around the AP and the developing cyclonic flow over the EM, associated with the AA and upper trough, respectively, transport the gradually ascending moisture northward to the Levant. This corroborates that the Red Sea serves as a corridor (and supplier) for moisture transport as suggested by Krichak et al.  and Dayan and Morin . Also, relatively weak moisture quantities from the Mediterranean Sea are advected in the developing cyclonic flow over the EM during several events. Most remarkably, the contribution of moisture from tropical/equatorial Africa that was observed for some events appears negligible in view of the total VIWVF quantities. Hence, our results contrast considerably with several previous ARST studies that address moisture dynamics. This emphasizes the necessity to include not only moisture quantities, but also moisture fluxes in the analysis to reveal the sources and pathways of moist air masses. Also, it appears important to take into account the vertical moisture distribution, since single levels may not be representative of the total moisture transport throughout the troposphere. Finally, relatively coarse data sets may underestimate the moisture contribution from small and poorly resolved water bodies like the Red Sea.
5.4 Upward Motions and Precipitation Generation
 All Levant ARST events are associated with strong midtropospheric ascent (Figure 9a), varying from −0.6 up to −2 Pa s−1 at 500 hPa. The combination of advected moisture, low-level warm air, and upper level cold air causes tropospheric instability, illustrated by high convective available potential energy (CAPE) values over the Levant (Figure 9b) that vary from about 500 to more than 3400 J kg−1, depending on the event. Not surprisingly, events during midautumn show higher CAPE values compared to events in late autumn and early winter. Note that the high CAPE values over the warm Red Sea waters in Figure 9b are common, whereas occasionally high CAPE values over land indicate severe weather.
 Strong upper level divergence is observed, mostly near the left exit region of the STJ streak (Figure 9c), which implies midtropospheric ascent based on continuity principles [Stull, 2000, chapter 13]. During ARST occurrence, the upper level divergence center is positioned over the Levant and reaches maxima of about 0.4 × 10−4 up to 1 × 10−4 s−1. Also, the upper level positive vorticity advection and low-level warm air advection (not shown) imply quasi-geostrophic forced ascent [Holton, 2004, chapter 6]. Finally, orographic effects play an important role in the precipitation generation [Dayan et al., 2001], though this aspect is not discussed for the Levant ARST events in this study.
6 Jeddah 2009 Event as an ARST
6.1 Dynamics of the Jeddah Event in November 2009
 Next we analyze the dynamics during the Jeddah 2009 event and compare them to those of the Levant ARSTs. We find a narrowly shaped and meridionally elongated RST over the Red Sea region near the surface and the AA over the AP at lower to middle levels (Figures 10a, 10b, and 10c). The RST extends northward prior to and during the Jeddah flooding and retreats afterward (Figures 10a, 10d, and 10g), while the AA slightly intensifies prior to the Jeddah flooding and weakens afterward (Figures 10b, 10c, 10e, 10f, 10h, and 10). Furthermore, at upper levels, wave breaking occurs over northwest Africa on 20 November 2009, resulting in a cutoff low that moves eastward over North Africa during subsequent days (Figure 11a). During the Jeddah flooding (the 25th), the cutoff merges with a trough, positioned over the EM and Levant, with ridges at both flanks (Figure 11c). After the Jeddah flooding (the 26th/27th), another (anticyclonic) wave breaking event occurs over the EM and Levant, leading to a next pronounced cutoff low (Figure 11e). The STJ intensifies over northeast Africa prior to and during the Jeddah flooding (Figures 11b and 11d), and stretches over the AP during the subsequent days, while intensifying further (Figure 11f). The wind speed at 200 hPa reaches maxima of ~55 m s−1 on the 25th and ~75 m s−1 on the 27th. As for most of the Levant ARST events, a southward branch of the polar jet merges with the STJ and forces advection of upper level cold air (Figure 11d). High TCW quantities from central Africa reach northward over the Red Sea region (Figure 12a), again suggesting central Africa as a moisture source. However, Figure 12b corroborates that, similar to the Levant ARST events, the moisture originates predominantly from the Arabian and Red Seas, and to a lesser extent from the Mediterranean Sea, and is transported by the persistent anticyclonic flow around the AP and the evolving cyclonic flow over the EM. Closer inspection of the VIWVFs over distinct layers and its temporal evolution reveals that the moisture from the adjacent seas gradually ascents and converges over the Jeddah region during the flooding (Figures 12c–h). In addition, at midlevels, moisture is transported from central Africa (Figure 12e) although the amounts are negligible in view of the total vertically integrated quantities.
 During the Jeddah 2009 event, the upper trough and STJ are positioned somewhat more southerly compared to the Levant ARST events. In addition, the AA shows a closed structure over the southeastern part of the AP, whereas during the Levant ARST events, the AA (at 850 hPa) stretches northward over the entire AP (cf. Figures 10e and 5e). Thus, the configuration of the AA and upper trough explain why the moist airflow converges over the Jeddah region and intrudes into the AP, whereas during the Levant ARST events, the moist air masses propagate farther northward across the Red Sea (cf. Figures 12b, 12h, 7b, and 7h).
 During the Jeddah 2009 event, high CAPE values (>3400 J kg−1) are observed over the Red Sea coast nearby Jeddah (Figure 13b). The upper level divergence over the Jeddah region, ahead of the STJ streak, is relatively weak (Figure 13c). Apparently, the upward motions (Figure 13a) and severe weather on the 25th over the Jeddah region result predominantly from the tropospheric instability and to a lesser degree from the synoptic-scale forcing. Accordingly, Haggag and El-Badry  refer to a quasi-stationary mesoscale convective system. In this respect, the Jeddah 2009 event is similar to the ARST, considering the suggestion by Dayan and Morin  that the convective ascent resulting from tropospheric instability dominates the synoptic-scale forced ascent. During the subsequent days after the Jeddah flooding, the tropospheric instability weakens, while the upper level divergence and related large-scale ascent significantly increase over the central part of the AP (Figures 13d–13f), demonstrating strong cyclogenesis. This evolution of initially strong convective activity (developing into mesoscale convective systems), which at a later stage organizes into large-scale ascent (suggesting cyclogenesis), is mentioned in several ARST case studies [Krichak et al., 2000; Dayan et al., 2001; Ziv et al., 2005]. Accordingly, the December 1993 and November 1994 events show initially high CAPE values, while the upper level divergence at a later stage gradually becomes dominant. Nevertheless, this evolution could not clearly be distinguished during most other Levant ARST events.
6.2 Geographical Extent of the ARST and Jeddah
Haggag and El-Badry  report several important synoptic-scale dynamical aspects that caused the Jeddah 2009 event. First, an eastward migrating Mediterranean cyclone joins an extension of the Sudan low-pressure zone (i.e., the RST), second, a stationary anticyclone over the southeastern AP (i.e., the AA) and associated clockwise flow provide moisture from the Arabian and Red Seas, and third, upper level tropospheric instability and associated deep moist convection result in a quasi-stationary mesoscale convective system and heavy rainfall over Jeddah. Our results are fully in agreement and include the role of the STJ and large-scale ascent. Moreover, we explain the Jeddah 2009 event in terms of the ARST phenomenon and conclude that the associated dynamics and evolution are identical to those observed for the Levant ARST events. This reveals that the Jeddah 2009 event was in fact an ARST. Apparently, the ARST affects not only the Levant, but also the larger part of the AP, introducing a significant geographical extension of the phenomenon.
 Importantly, the Jeddah flooding on 25 November 2009, being a result of an ARST, is not an isolated case. For example, heavy rainfall affected Jeddah on 8 January 1999 [Almazroui, 2011], while more recently flash floods struck Jeddah in December 2010 and caused four casualties [Haggag and El-Badry, 2013]. Based on exploration of the synoptic conditions (not shown), we identify these events as ARSTs. Furthermore, the Levant ARST event in November 1996 also affected Jeddah with large precipitation amounts [Almazroui, 2012]. This illustrates that the Jeddah region has been affected more often by extreme precipitation and resulting flash floods as a consequence of ARSTs. The orography in the region is characterized by steep slopes along the eastern Red Sea coast. Nearby Jeddah, the mountain range is interrupted by a pass that forms a corridor to Makkah with the Hejaz mountains to the north and the Asir mountains to the south. Hence, the local orography promotes channeling of the moist airflow from the Red Sea over the Jeddah region into the AP and potentially makes this region extraordinarily favorable for extreme precipitation and flash floods.
7 Seasonality of the ARST
7.1 Large-Scale Circulation
Dayan et al.  explain the occurrence of ARSTs in autumn by favorable latitudinal positions of the STJ and the African Monsoon, i.e., the local Inter Tropical Convergence Zone (ITCZ). In spring, the STJ and ITCZ reach a similarly favorable setting. In winter, the ITCZ is positioned farther south from the ME, while in summer, the baroclinic zone is positioned farther north from the ME, making intrusions of tropical moist air masses and midlatitude upper level troughs in the ME, respectively, less likely. Moreover, in summer, convective ascent over the ME is inhibited because strong subsidence prevails at middle and upper levels over the ME and EM under the influence of the South Asian monsoon [Rodwell and Hoskins, 1996; Ziv et al., 2004; Tyrlis et al., 2013], the local Hadley Cell circulation associated with the African Monsoon [Ziv et al., 2004], and the heat driven circulation over the Zagros mountains [Zaitchik et al., 2007]. Hence, ARSTs are suppressed in summer. Thus, the seasonal variability of the large-scale circulation suggests ARST favorable conditions during autumn and spring, limited potential in winter, whereas ARSTs are unlikely in summer.
7.2 Origin and Seasonality of the RST and AA
 In summer, a chain of subtropical anticyclones at midlevels dominates over northwest Africa, the AP, and Iran [Galarneau et al., 2008; Zarrin et al., 2010]. When the Hadley Cell circulation migrates southward in autumn, the associated subtropical anticyclonic ridge divides into two subtropical anticyclones over the AP (i.e., the AA) and northwest Africa, respectively. In between, a surface trough extends from the ITCZ, reaching over northeast Africa and the Red Sea region; that is, the RST. The RST and AA are to a greater or lesser extent always present in the daily synoptics during autumn, winter, and spring. In other words, the RST and AA have a strong semipermanent quasi-stationary character and leave signatures in the mean low-level circulation during these seasons. Figures 14a–14f and 14j–14l show the climatology of the RST and AA for January, April, and November (see for the monthly means of all months supporting information Figures S4–S7).
 The RST and AA reach their northernmost position in late autumn (October and November) and spring (April and May), attain their southernmost position in winter (January and February), and are absent in summer (June until September). In addition, the AA is also visible at midlevels (500 hPa) in October and May (supporting information Figures S7e and S7j), consistent with its transformation from an isolated anticyclone near the surface in winter to a subtropical anticyclonic ridge at midlevels in summer. In essence, the vertical and horizontal extent, as well as the latitudinal position of the AA, follows the solar inclination and, more specifically, the intensity of the surface sensible heating over the AP [after Wu et al., 2011, 2012] and the descending branch of the Hadley Cell. The monthly means suggest that the RST during early winter (December and January) is in fact not less frequent [Alpert et al., 2004b; Tsvieli and Zangvil, 2005; Krichak et al., 2012], but is positioned more southerly and therefore affects Israel less frequently. During summer months (June until September), when potentially reaching its northernmost position, the RST is fully absorbed by the Persian Trough, a thermal low-level trough developing from the east under the influence of the South Asian monsoon [e.g., Bitan and Saaroni, 1992; Alpert et al., 2004a, 2004b; Tyrlis et al., 2013], see Figures 14g and 14h. Apparently, the influence of the South Asian monsoon does not only inhibit ARSTs in summer via the imposed subsidence at middle and upper levels over the ME, but also by dominating the atmospheric circulation at lower altitudes.
 The RST and AA show an asymmetric seasonal evolution. The RST and AA (at ~850 hPa) are most pronounced in late autumn and early winter (October until January) and weaker in late winter and spring (February until April). The AA at higher levels (~700 hPa) is most pronounced in autumn (October and November) and spring (April and May), and weakest in winter (January and February). Furthermore, at lower levels (~850 hPa), the AA remains at its southerly position over the southeastern AP and Arabian Sea in spring (February until April), and disappears afterward (May). It appears that the South Asian winter monsoon northeasterly wind over the Arabian Sea reinforces the anticyclonic flow around the AP from autumn until early spring (e.g., Figures 14b and 14k), while in late spring (May), oppositely directed wind, associated with the onset of the South Asian summer monsoon, counteracts the anticyclonic flow. In combination with increasing surface sensible heating over the AP, this may lead to the disappearance of the AA at lower levels during spring. The less pronounced RST in spring compared to autumn (in terms of pressure gradients) is obvious since spring is rather characterized by the Sharav cyclone [Alpert et al., 2004a, 2004b] and suggests at the same time a relation to the less pronounced AA at low levels.
 In summary, the RST and AA (at ~700 hPa) reach their most northerly position in autumn (October and November) and spring (April and May), while the RST and AA (at ~850 hPa) are most pronounced in autumn and early winter (October until January). Thus, the seasonal pace of the RST and AA explains, in addition to the seasonal variability of the large-scale circulation, why ARSTs over the Levant occur predominantly in autumn (October and November) and to a lesser extent in early winter (December and January) and spring (March, April, and May), summarized in Table 3. At more southerly locations (e.g., the Jeddah region), ARST favorable conditions arise later in autumn, end earlier in spring, and can potentially last throughout the winter.
In addition, the ME is dominated by the influence of the South Asian summer monsoon; see text.
ITCZ/African Monsoon and baroclinic zone
RST and AA
Shift southward, most pronounced
Most southerly, weaken
Shift northward (RST and AA at 700 hPa), weak (RST and AA at 850 hPa)
ARST favorable conditions
8.1 Low-Level Circulation Climatology
 Figure 15a schematically depicts the seasonally averaged low-level circulation during autumn, winter, and spring. The RST appears in fact as an inverted trough in the low-level circulation that interrupts the trade winds and the subtropical anticyclones. The RST is characterized by pronounced moisture quantities (supporting information Figure S8) that contribute to the decreased barometric pressure. The high humidity makes the air more buoyant and, under the influence of sensible heating over northeast Africa, causes pressure fluctuations that explain the observed diurnal variations of the RST described in section 5.1. The moisture originates from the Arabian and Red Seas, in particular during the period when the RST and AA (at ~850 hPa) are most pronounced (October until January), as illustrated by the monthly means of the VIWVFs (supporting information Figure S9). The topography plays an important role. While the Arabian and Red Seas moisten the air, the Ethiopian Highlands block the AA intensified easterly low-level flow and enhance its northward propagation into the Red Sea region. In other words, we explain the RST as a persistent and stationary wave in the moist tropical easterlies, induced by the interaction between the low-level circulation, the AA forcing, and the topography in the Red Sea region.
8.2 Unified Concept of the ARST Dynamics
 Based on the 12 analyzed Levant ARSTs and the Jeddah 2009 event, and building on previous studies, we present an ARST concept that includes six important dynamical factors (Figure 15b): (1) the northward extension of the semipermanent quasi-stationary RST, (2) the intensification of the semipermanent quasi-stationary AA, (3) the intrusion of a midlatitude upper level trough, (4) the intensification of the STJ, (5) enhanced moisture transport from the adjacent seas, and (6) upward motions resulting from the synoptic-scale dynamical forcing and tropospheric instability.
 The RST and AA are always present in ARST favorable seasons and show only a relatively weak anomaly from their mean state during ARSTs [cf. Kahana et al., 2002]; hence, they are considered to serve as a precondition. ARSTs are triggered by the upper level trough. Thus, the midlatitude forcing plays a key role and determines the frequency of ARSTs. Knippertz and Martin , Knippertz , and Rubin et al.  found that for heavy precipitation events, associated with tropical plumes, either a quasi-stationary upper level trough or two consecutive upper troughs were essential to initiate sufficient poleward moisture transport from the tropics. Also, Kahana et al. [2002, 2004] highlighted, in relation to the evolution of upper troughs during ARSTs, that a “period of incubation” is required for the southerly transport of tropical moisture. Therefore, the persistence and southward extent of the midlatitude upper level trough plays a crucial part in the amounts of (extreme) rainfall. Accordingly, a fast travelling upper trough that does not reach far southward and is not preceded by another upper trough may result in events with the ARST dynamical configuration without associated precipitation. The moisture originates from the adjacent seas in the region and fuels the precipitation process. Therefore, sea surface temperatures and atmospheric moisture quantities potentially influence the amount of precipitation and the intensity of ARSTs, being of interest in the context of global warming. The combination of all synoptic-scale dynamical factors sets the scene for severe weather and favors strong upward motions. Processes at smaller scales determine the exact location and timing of the convective storms and (localized) heavy rainfall. The dynamical factors and their roles in the ARST are summarized in Table 4.
Table 4. Dynamical Factors of the ARST
Behavior During ARST
Role in ARST
Surface trough from tropics (upward branch Hadley Cell, African Monsoon/ITCZ)
Southeasterly advection warm moist air
Subtropical anticyclone (downward branch Hadley Cell and surface sensible heating)
Southeasterly advection warm moist air
Midlatitude forcing, Rossby wave amplification and breaking
Intrudes into the EM
Cyclonic flow and cold air advection
Positioned over North Africa and the ME
Intensifies due to midlatitude forcing
Upper level divergence
Predominantly Arabian and Red Seas, and to a lesser extent Mediterranean Sea
Transport by AA and upper trough, ascends gradually, converges, and intrudes the ME
Enhances tropospheric instability
Tropospheric instability and synoptic-scale dynamical forcing
Convective and large-scale ascent
Determines location and timing (extreme) precipitation
Krichak and Alpert  proposed an ARST mechanism in five stages, involving enhanced convection over the tropics, an intensification of the STJ, the development of the RST, and the northward advection of moist air masses into the Red Sea area. Our results are partially in agreement; however, the following observations provide additional insights: (1) The northward extension of the surface pressure trough, the enhanced moisture quantities (TCW), and the intensified northward moisture transport (VIWVFs) over the Red Sea evolve rather synchronously, illustrating the involvement of moisture from the Arabian and Red Seas in the northward extension of the RST. (2) The RST extends northward simultaneously with the intensification of the AA and/or the approach of the midlatitude upper trough. Previously, the RST extension was attributed to the positioning and strength of the STJ [Krichak et al., 1997a]. Based on our analysis, we postulate that the northward progression of the RST and associated moisture is forced by the combination of the intensifying anticyclonic flow around the AP and the developing cyclonic flow over the EM and Levant. Thus, the persistent AA and the approaching upper trough undergo phase locking and drive the northward advancement of the RST and tropical moisture. (3) The temporal and spatial evolution of the intensifying STJ progresses simultaneously with the development of the upper level trough, showing that the STJ intensification results predominantly from the midlatitude forcing rather than from intensified convection over the tropics and east-poleward outflow at upper levels, consistent with the results of section 5.2.
 Emerging from the above, we propose the following revised ARST mechanism: (1) Rossby wave amplification and breaking results in the formation of a midlatitude upper level trough that intrudes into the EM. As a result, (2) the upper and lower tropospheric circulations couple and interact. The midlatitude upper level trough and preexisting RST merge, and the ridge over the AP (further) intensifies the persistent anticyclonic flow around the AP at low to midlevels, together forcing pronounced southerly winds over the Red Sea region. Consequently, (3) the RST extends northward, and the associated tropical moist air masses progress over the Red Sea and intrude into the Levant and/or AP regions. (4) The enhanced tropospheric instability and dynamical forcing cause ascending motions, promoting the development of convective storms and heavy precipitation. (5) Subsequently, the midlatitude upper level trough dissipates, the RST retreats, and the AA weakens, and the moisture transport from the Arabian and Red Seas is again directed toward equatorial Africa.
 In summary, ARST occurrence is triggered when an amplified Rossby wave superimposes over the preexisting low-level wave (i.e., the RST). The upper trough deepens and merges with the RST, in line with Tsvieli and Zangvil , while the ridge over the AP (further) intensifies the AA. The anomalies in the upper and lower troposphere phase lock and amplify each other according to the cyclogenetic process as described by Hoskins et al. [1985, Figure 21]. Consequently, the persistent easterly moist airflow toward equatorial Africa is interrupted and progresses northward to the ME, implying that pronounced southerly moisture fluxes over the Red Sea are a key signal of the ARST, which is confirmed by Figure 8.
8.3 AA, Associated Moisture Transport, and Rainfall Over the ME
 Our study underscores that the intensified AA and associated moisture transport not only characterize the December 1993 event [Ziv et al., 2005] and the Jeddah 2009 event [Haggag and El-Badry, 2013], but are relevant for all analyzed Levant ARSTs (Figure 8). Apart from the ARST, the AA and associated moisture transport are also involved in precipitation and floods in Iran [Farajzadeh et al., 2007; Sabziparvar et al., 2010; Raziei et al., 2012]. The latter study found the AA to induce lower tropospheric moisture transport from tropical water bodies (Arabian Sea, Oman Sea, Persian Gulf, Red Sea, and northern Indian Ocean) into cyclonic systems near Iran. In a climatological modeling study, Evans and Smith  identified southerly moisture fluxes, characteristic for a minority of rainfall events, to be responsible for a major contribution to the annual precipitation over the Fertile Crescent (southeast Turkey, northeast Syria, north Iraq, and northwest Iran), suggesting the involvement of the AA. We therefore postulate that the AA and associated moisture transport are not only relevant for the ARST, but also for other precipitation generating phenomena over the ME such as tropical plumes and midlatitude cyclones (e.g., Mediterranean cyclones).
8.4 ARST and Other Tropical-Extratropical Interactions
 The ARST shows strong similarities to other tropical-extratropical interactions that affect semiarid to arid subtropical regions with extreme precipitation. In general, poleward transport of tropical moisture is initiated by the midlatitude forcing. More specifically, important synoptic dynamics include upper troughs and/or cutoffs resulting from (Rossby) wave breaking, intensification of the STJ, wave disturbances in the low-level tropical easterlies (easterly waves), trade surges (strong trade winds), and precipitation generation resulting from orographic lifting, tropospheric instability, quasi-geostrophic forcing, and STJ associated upper level divergence [e.g., McGuirk et al., 1988; Ziv, 2001; Fink and Knippertz, 2003; Knippertz et al., 2003; Knippertz, 2005; Knippertz and Martin, 2005; Knippertz and Martin, 2006; Knippertz, 2007; Rubin et al., 2007; Hart et al., 2010; Favors and Abatzoglou, 2013].
 However, the ARST dynamics differ in particular aspects. Most of the aforementioned studies indicate that the moisture is lifted within the ITCZ through convection and advected in the midtroposphere over the (dry Saharan) planetary boundary layer toward the subtropics, as illustrated by Nicholson [1981, Figure 3] and Knippertz [2003, Figure 1]. For the ARST events, we infer gradually ascending moisture in the anticyclonic flow around the AP. Moreover, the driving mechanism, a subtropical anticyclone at lower levels, is not observed for any other type of tropical-extratropical interaction. Therefore, the particular topography in the Red Sea region and the associated low-level circulation (the RST, AA, and associated moisture transport) makes the ARST a unique phenomenon among tropical-extratropical interactions.
9 Summary and Conclusions
 We investigated the ARST associated atmospheric dynamics, geographical extent, and seasonality. ERA-Interim data were used to study the underlying synoptic dynamics of 12 ARST events that affected the Levant and the flooding that struck Jeddah on 25 November 2009. Further, the seasonal variability of the large-scale and local low-level circulation was analyzed to explain the ARST seasonality. In addition, we considered observational precipitation data to gain insight into the ARST-associated rainfall evolution and characteristics.
 Rainfall during ARSTs evolves over the Red Sea region and progresses in a northeasterly direction. The Levant can be affected for a time period from one to four days with rainfall maxima typically ranging from 10 up to 60 mm per day. Precipitation in ERA-Interim and Aphrodite data show a similar spatial and temporal evolution, however, can locally differ significantly in the timing and amounts of rain. Based on the analyzed events, including the Jeddah 2009 event, and building on previous studies, we present an ARST concept involving six dynamical factors (Figure 15b). This concept comprises (sub)tropical influences (the RST and AA), midlatitude forcing (the upper level trough and intensified STJ), moisture transport pathways, and the resulting upward motion. The AA and associated moisture transport from the Arabian and Red Seas are of central importance for the ARST, and are probably also involved in other precipitation generating phenomena over the ME. The dynamics associated with the Jeddah 2009 event are identical to those associated with the Levant ARST events. Hence, the Jeddah flooding was caused by an ARST. This implies that, in addition to the Levant, also the AP is affected by the ARST, introducing a significant geographical extension of the phenomenon.
 Further, we explain the ARST seasonality by the seasonal pace of the large-scale circulation and in particular that of the low-level circulation. More specifically, the latitudinal positions of the midlatitude baroclinic zone and the African Monsoon (i.e., the local ITCZ) induces ARST favorable conditions during autumn and spring, and limited potential in winter, whereas the South Asian monsoon influence rules out ARSTs during summer. The semipermanent quasi-stationary RST and AA show their most northerly position in autumn and spring, and most pronounced state in autumn and early winter, revealing why ARSTs in the Levant occur predominantly during autumn (October and November) and have a reduced incidence during early winter (December and January) and spring (March to May). ARST favorable conditions in more southerly locations (e.g., the Jeddah region) arise later in autumn and earlier in spring, and possibly last throughout the winter.
 The mean low-level circulation in the Red Sea region is strongly influenced by the local topography. In fact, we explain the RST as a persistent and stationary wave disturbance in the tropical easterlies, induced by the AA forcing, the flow blocking by the Ethiopian Highlands, and the moistening of air masses by the Arabian and Red Seas (Figure 15a). An ARST event is triggered when a midlatitude upper level trough intrudes the EM, going along with a Rossby wave that amplifies and superimposes over the preexisting low-level wave (Figure 15b). Their interaction causes pronounced southerly moisture transport over the Red Sea toward the ME (Figure 8). Hence, the ARST is a tropical-extratropical interaction resulting in a tropical moist air intrusion. The ARST dynamics show substantial similarities with other type of tropical-extratropical interactions affecting semiarid to arid subtropical regions elsewhere, however, differ in particular aspects due to the particular topography in the Red Sea region and the associated low-level circulation, justifying the term “ARST.”
 In ME countries, floods are a main cause of natural disasters and have severe societal impacts. For example, seven of the top ten disasters in Saudi Arabia in the period 1900–2011, ranked by number of casualties as well as numbers of people affected, resulted from floods (EM-DAT). Therefore, the causes (e.g., atmospheric dynamics), potential impacts, and preemptive measures need increased attention. Our study provides a better understanding of the ARST dynamics and should benefit weather forecasting and early warning systems. Further work may quantify the contribution of the identified moisture sources in ARSTs using back trajectory methods. Modeling sensitivity studies could explore the impact of the mountains and land-water surface differences in the Red Sea region on the low-level circulation, and in particular the RST and AA. It will also be important to investigate the role of climate change, for example in view of increasing sea surface temperatures and associated moisture fluxes. The new insights in the ARST dynamics and the expanded geographical extent of the phenomenon, as presented in this work, introduce interesting possibilities to study the ARST in a climatological context. In a subsequent study, we will investigate the ARST climatology, the influence of large-scale circulation patterns on ARST occurrence, and the ARST contribution to total rainfall amounts in the ME.
 The research leading to these results has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement 226144. S.O. Krichak and D. Edry acknowledge financial support by the United States-Israel Binational Science Foundation (BSF) under research grant 2008436. The research was partly supported by The Cyprus Institute and Tel Aviv University through their cooperative agreement. The authors wish to thank Aphrodite, CRED, ECMWF, and IMS for providing their data. The data are visualized with the National Center for Atmospheric Research (NCAR) Command Language (NCL) package (version 6.0.0). We are grateful for helpful discussions with M. Tanarhte from the Max Planck Institute for Chemistry. We greatly appreciate the comments of three anonymous reviewers that improved the manuscript.